 

 

\. .
a

 

5. \-,{§! :5.\..L\\d .1:

 

 

 

 

 

LIBRARIES

3 mljlliiiilliiiuuuuM

(r (‘54,; 8176

 

 

 

 

This is to certify that the

thesis entitled

Species-Site Interactions in a Managed
Subtropical Forest of the Dominican Republic

presented by

Mark A. Hare

has been accepted towards fulfillment
of the requirements for

Masters degmein Forestgy

 

 

 

Date /////‘;/7.5’

0-7639 MS U is an Afﬁrmative Action/Equal Opportunity Institution

 

 

   
 
   

 

LIBRARY
Michigan State
University

 

PLACE IN REFURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.

 

DATE DUE“ . DATE DUE DATE DUE

 

$312529?

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

MSU Is An Affirmative Action/Equal Opportunity Inmitution
WM

pus-9.!

 

SPECIES-SITE INTERACTIONS IN A MANAGED SUBTROPICAL DRY
FOREST OF THE DOMINICAN REPUBLIC

By

Mark A. Hare

A THESIS

Submitted to
Michigan State University
in partial fulﬁllment of the requirements for the degree of

MASTER OF SCIENCE

Department of Forestry

1995

ABSTRACT

SPECIES-SITE INTERACTIONS IN A MANAGED SUBTROPICAL DRY
FOREST OF THE DOMINICAN REPUBLIC

By

Mark A. Hare

In 1986, a thinning study was initiated in a subtropical dry forest of the
Dominican Republic. After six years of inventories, no effects attributable to
thinning were observed. Classiﬁcation and ordination techniques were
subsequently applied to the inventory data collected prior to thinning. Data
consisted of basal area contributions of sixteen dry forest species on 120 100 m2
sites. Using clustering techniques, the sites were partitioned into six groups, each
representing a characteristic species composition. Group One was dominated by
Bursera simaruba, Group Three by Acacia scleroxyla, Group Four by Phyllostylon
brasiliensis, Group Five by Caesalpinia coriaria, Group Six by A. famesiana and
Group Seven by P. brasiliensis' and Pithecellobium circinale. The relative positions
of the groups in correspondence analyses and canonical discriminant analyses
suggested a gradient moving from Group Three to Group Four. Additional
analyses using overstory structures and growth and mortality parameters indicated

this apparent gradient was related to relative productivity.

Copyright by
thAIUKLAﬂNIIRJEVJlilURJS

1995

This study is dedicated to Humberto Checo and to the other individuals past and
present who have given their time and sweat in the work of the ISA-Mao Forestry
Experimental Station. May their work bring sense to the chaos, and better lives

to the communities everywhere entwined in a life or death dance with dry forests.

ACKNOWLEDGEMENTS

I have waited a very long time to write this section. Sometimes, when the end of
this journey was lost in the mists of the farthest horizon, I would imagine what I

would say here. It helped a little.

The ﬁrst people I want to acknowledge are my parents, and not just because I
know I should. My parents are incredible people. After some thirty-two years of
wanderings, I have met only a handful of individuals whose level of integrity,
compassion and dedication can match theirs. They have shown me again and
again the literal signiﬁcance of Christ’s command to serve. I depend on their

presence and support, and on their witness for how life is to be lived.

Next, I need to recognize my brothers and sisters and their respective families.
Their love has been indispensable. It is beyond counting the number of times I
have ended up on one of their doorsteps with little or no warning. Crazy as larks
all of them, but somehow they manage to make it day by day in the "normal"
workaday world, and still keep their senses of humor. I depend on their support,

which they freely give.

I also want to acknowledge Dr. Douglas Lantagne. This study and my degree
would not have happened without him. Doug has always been willing to share his
professional silvicultural experiences and our conversations have been invaluable
in my personal education. Pieces of those conversations are interwoven
throughout this study. Doug has also shown incredible patience with my internal
time table, and he succeeded again and again in ﬁnding ﬁnancial support when I
most needed it. I could not have found a better adviser and friend. I owe him a

great deal. In all likelihood, he will remind me of that.

I also need to express my appreciation to my other committee members, Dr. Peter
Murphy and Dr. Carl Ramm. Dr. Murphy was extremely helpful in directing my
literature search, as well as in bouncing ideas around when I was still unsure of
the direction for my research. Dr. Ramm is almost entirely responsible for my
interest in multivariate methods. His MVA class was well-organized and very
effective in providing a hands on approach to using the statistical procedures. Dr.
Ramm also did an incredibly thorough editing job on my thesis. I don’t know

whether or not to thank him for that, but the ﬁnal product is better for his work.

I was exceptionally lucky to have Humberto Checo at my disposal for a year and a
halt Checo has an intuitive understanding of the dynamics of the dry forest in
general, and the ISA-Mao silvicultural thinning study in particular. Checo is the

person who will ultimately determine if any of my observations are valid and/or

useful for developing long term sustainable management practices of subtropical

dry forest.

The graduate students in the Department of Forestry are an incredible support
group. I cannot count the times I have left the lunch room laughing. In
particular, I would like to thank Michael Powers, Jill Fisher, Peg Payne and Andy
David for giving me rides, making me laugh and just generally for being "bitchin’

dudesﬁ

I owe a lot to any number of other people. Friends who have helped me out, or
bought me a beer, or just told me something good about life- these are the things
which have made life worthwhile these last four years. I especially want to
recognize my current roommate, Chekole Kassa and his "brothers", Yonas,
Nathnael and Berhan. In the last two years, I have frequently depended on them
to keep me sane. Again and again, with beer, music and Ethiopian food they

have set me back on the path and kept me going.

I also want to acknowledge the professors and fellow students from my
undergraduate work at Warren Wilson College in Swannanoa, North Carolina.
After four years in a Big Ten institution, I am keenly aware of both the strengths
and weaknesses of a small liberal arts institution such as WC and I ﬁnd that the
strengths far outweigh the limitations. The institutional opportunities available at

a large university are no match for the intellectual, moral and spiritual resources

vii

available at WWC. Warren Wilson is blessed by a faculty and staff which are
dedicated both to the students and to the college community as a whole. My

experience there was invaluable to my development as a wholly educated person.

Finally, I want to acknowledge friends met and made in the Dominican Republic.
The human and cultural resources available on Hispaniola are truly awesome and
I have had enormous fun in all my visits there. Without hesitation, I commit my

time and strength to the future of that island.

TABLE OF CONTENTS

Page
I. Introduction ............................................................................................. 1
11. Literature Review ................................................................................... 6
Site Classiﬁcation ........................................................................ 6
Site Factors .................................................................................. 9
Site Disturbance .......................................................................... 12
Multivariate Analyses ................................................................. 13
III. Materials and Methods .......................................................................... 21
Site Description .......................................................................... 21
Original Study Design ................................................................ 24
Numerical Methods .................................................................. 28
Variable Selection ........................................................... 28
Species Identiﬁcation ...................................................... 33
Multivariate Analyses ...................................................... 34
Cluster Techniques .............................................. 34
Correspondence Analysis .................................... 39
Canonical Discriminant Analysis ....................... 40
Overstory Analyses .......................................................... 42
Site Characteristics .......................................................... 43
Site Productivity ............................................................... 46
Nonparametric Analysis of Thinning Effects ................ 47
IV. Results and Discussion ............................................................................ 49
Initial Analyses ............................................................................. 52
Cluster Techniques .......................................................... 52
Correspondence Analyses ................................................. 55
Canonical Discriminant Analyses ................................... 71
Summary ............................................................................ 80
Full Data Set Analyses ................................................................ 81
Cluster Techniques ........................................................... 81
Correspondence Analyses ................................................ 91
Canonical Discriminant Analyses ................................... 110
Summary ............................................................................ 131

table of contents (cont’d)

Species, Site and Overstory Characteristics ............................... 133
Species Characteristics ..................................................... 134
Site Characteristics ........................................................... 144
Overstory Structure .......................................................... 148
Summary .......................................................................... 158
Growth and Mortality Within Cluster Groups ........................ 160
Nonparametric Analysis of Thinning Effects ........................... 169
V. Summary .................................................................................................. 175
VI. Conclusions ............................................................................................... 181
VII. Recommendations ................................................................................... 182
VIII. Literature Cited ...................................................................................... 185
Appendix A
Actual basal area and stems removed and assigned cutting levels... 193
Appendix B
Structural characteristics of the 120 silvicultural sites ........................ 196
Appendix C
Site characteristics data from the forty original control sites .............. 199
Appendix D
Ground vegetation data and CRUZALL data from the forty original
control sites .............................................................................................. 201
Appendix E
Growth and morality parameters for 120 sites ..................................... 205
Appendix F
Nonparametric analysis of growth and mortality ................................. 208
Appendix G
Results of hierarchical analyses .............................................................. 215
Appendix H
SAS output from CA for the initial analyses ........................................ 225
Appendix I
SAS output from CDA for the initial analyses ..................................... 230

table of contents (cont’d)

Appendix J
SAS output for CA using the full data set ............................................. 235

Appendix K
SAS output for CDA using the full data set ......................................... 244

Appendix L
Scaled diagrams for the four experimental blocks of the ISA-Mao
silvicultural study ...................................................................................... 252

Appendix M
Representative maps with proﬁle icons indicating cluster group
membership and relative contribution of each of the sixteen species used

in the cluster analyses and ordinal procedures ..................................... 257
Appendix N

A proposal for a study of soil-site interactions in the managed forest of the

ISA-Mao Experimental Forestry Station ................................................ 263

LIST OF TABLES

Table Page

1. A subsample of sites from the thinning study in a subtropical dry forest of the

Dominican Republic demonstrating reassignment of the sites to new
treatment designations based on the percent of initial basal area actually
removed in the cutting. The ﬁrst digit in Site ID indicates the block number
(1 through 4). The second and third digit represent the site number (1
through 30) within each block. The original treatments are listed in the
"Target thinning level" column. The actual percentages of stems and basal
area removed are listed in the last two columns. The assigned cutting levels
are the treatment designations used for analyses in this study ............... 27

Relative measures of density for forty-four species in the silvicultural thinning
study at the Mao-ISA subtropical dry forest experimental station.
Calculations are based on measurements taken at knee height (0.5 m).
Species are listed in descending order of their importance value .......... 30

Local names for species used in the species-site analyses of the ISA-Mao dry
forest with their corresponding Latin names .......................................... 34

Ground vegetation identiﬁed in the forty designated control sites in the ISA-
Mao silvicultural study. "Data designation" refers to the classiﬁcation used
in data analysis .......................................................................................... 45

Prominent canopy species identiﬁed using a factor ﬁve CRUZALL tool on
forty sites representing the original control sites in the ISA-Mao
silvicultural study ....................................................................................... 45

10.

11.

List of tables (cont’d)

Results of two hierarchical cluster techniques and three versions of SAS
Fastclus on species data from a subset of 45 sites from subtropical dry
forest in the Dominican Republic. The ﬁrst digit in the ID number
indicates the block number (1 through 4). The second and the third digit
represent the site number (1 through 30 for each block). Group indicates
the cluster designation assigned based on the ﬁve tests. Numbers indicate
sites consistent across all cluster techniques. Letters indicate the core
group with which the site was most closely associated (i.e. "A" indicates a
site which clustered most consistently in cluster Group One, etc.) ..... 53

Average species composition of the six groups determined for 30 core sites
showing consistent membership across cluster techniques. Order of species
follows the ﬁrst dimension of correspondence analysis. Order of the ﬁve
main groups is based on the ﬁrst canonical axis of canonical discriminant

analysis ........................................................................................................ 56

Singular values, principal inertias and proportional representation for each of
ﬁfteen dimensions of a correspondence analysis of sixteen species and
thirty sites representing core cluster group sites .................................... 59

Singular values, principal inertias and proportional representation for each of
ﬁfteen dimensions of a correspondence analysis of sixteen species and
forty-ﬁve sites in the ISA-Mao silvicultural study ................................ 70

Eigenvalues associated with each axis of CDA applied to 16 species and 30
cere sites representing ﬁve groups from the ISA-Mao silvicultural
study ........................................................................................................... 75

Eigenvalues associated with each axis of CDA applied to 16 species and 45
sites representing ﬁve cluster groups ....................................................... 78

List of tables (cont’d)

12. Results of two hierarchical cluster techniques and two versions of SAS
Fastclus applied to species data from 120 sites in a subtropical dry forest in
the Dominican Republic. The ﬁrst digit in the ID number indicates the
block number (1 through 4). The second and third digit represent the
subplot number (1 through 30 for each block). Final Group indicates the
cluster designation assigned based on the four cluster techniques applied to
all 120 sites. Numbers indicate sites which clustered consistently across all
techniques, referred to as "core" sites. Letters indicate non core sites.
Each letter designation is associated with a core cluster designation. I.e.
"A" is associated with sites which clustered in Group One more than once.
"B" is associated with sites that clustered in Group Two more than once,
etc. A double letter designation indicates that a site clustered in one group
twice and in another group twice. Initial Group indicates the group
designation assigned based on initial analyses using a subset of 45 sites.
NA indicates sites not in the initial analyses ........................................ 82

13. Average species composition of the six groups determined for 67 core
subplots showing consistent membership across cluster techniques. Order
of species follows the ﬁrst dimension of correspondence analysis. Order of
the groups is based on the ﬁrst canonical axis of canonical discriminant
analysis ........................................................................................................ 87

14. Average species composition of six subgroups. Each subgroup is represented
by noncore sites which showed the same clustering pattern. Letters are
associated with the corresponding main cluster group (i.e. Subgroup D
represents sites which clustered most frequently in group four). Order of
species follows the ﬁrst dimension of correspondence analysis. Order of
the groups is based on the ﬁrst canonical axis of canonical discriminant
analysis applied to six core site cluster groups and seven subgroups
represented by a total of 118 sites ........................................................... 89

15. Singular values, principal inertias and proportional representation for each of
ﬁfteen dimensions of a correspondence analysis of sixteen species and sixty-
seven sites representing core cluster group sites .................................... 94

16. Singular values, principal inertias and proportional representation for each of
ﬁfteen dimensions of a correspondence analysis of sixteen species and sixty-
seven sites representing core cluster group sites .................................... 106

17. Eigenvalues associated with each axis of CDA applied to 16 species and 67

core sites representing six groups from the ISA-Mao silvicultural
study ........................................................................................................... 112

xiv

18.

19.

20.

21.

22.

23.

24.

List of tables (cont’d)

Eigenvalues associated with each axis of CDA applied to 16 species and 62
core sites representing ﬁve groups .......................................................... 115

Eigenvalues associated with each axis of CDA applied to 16 species and 118
sites representing six core site cluster groups and seven noncore site
subgroups ................................................................................................... 122

Group means for the core site cluster groups based on site scores along the
ﬁrst canonical axis of canonical discriminant analysis (CDA) using 118 of
the 120 silvicultural experiment sites. Canonical discriminant analysis was
based on the proportional basal area contributions of sixteen dry forest
species, using groupings determined with four different cluster analysis
procedures ................................................................................................ 124

Group means for the noncore site cluster groups based on site scores along
the ﬁrst canonical axis of canonical discriminant analysis (CDA) using 118
of the 120 silvicultural experiment sites. Canonical discriminant analysis
was based on the proportional basal area contributions of sixteen dry forest
species, using groupings determined with four different cluster analysis
procedures ................................................................................................. 124

Speciﬁc gravity and average growth rates for nine species prominent in the
ISA-Mao silvicultural study. Species are listed in descending order based
on average annual growth. "NA" indicates species for which information is
not available from the respective source ................................................ 136

Structural characteristics for nine species prominent in the ISA-Mao
silvicultural study. Species are listed in descending order based on average
height. Data comes from the preharvest inventory done on 120 sites, each
100 m2, in 1986. Means are based on measurements taken at lmee height
(0.5 m) ....................................................................................................... 137

Results of tabulating slope angle by cluster group for the forty original
control sites in the ISA-Mao silvicultural study. The order the rows is
based on values along the ﬁrst axis of CDA applied to 118 sites
representing six core site groups and seven noncore site subgroups. One of
the sites tabulated as group "BG" was originally designated as "BD". One
‘of the sites tabulated with cluster group "G" was originally designated as
"FG". Data was collected in March 1984 ............................................ 145

26.

27.

28.

29.

30.

List of tables (cont’d)

Kruskal-Wallis test statistics for six structural parameters from eight cluster
groups (Wilkinson 1989). The data comes from 88 of 120 sites in the
silvicultural study at ISA-Mao ................................................................. 151

Mean values and rank means for structural characteristics within cluster
groups and across all sites ........................................................................ 151

Rank means and calculated test statistics for growth and mortality
parameters, using the Kruskal-Wallis distribution free test of differences
between rank means. Eight cluster groups are tested for signiﬁcant
differences. The H-Statistic is the calculated test statistic to be compared
with a Chi-square distribution. The alpha level was set at 0.1, with a Chi-
square statistic of 12.02, assuming seven degrees of freedom (k-l, where
k= the number of groups being tested) ................................................ 161

Distribution free multiple comparisons for growth and mortality parameters
found to have signiﬁcant differences between treatment means using the
Kruskal-Wallis distribution free test. Fourteen comparisons were
examined. The test statistic is the critical value calculated using an alpha
level of 0.1 with each test statistic based on the sample sizes of the
treatments being compared. Values in each cell are differences between
the two treatments being compared. Treatments with differences which are
signiﬁcant are underlined ........................................................................ 163

Rank means and calculated test statistics for growth and mortality
parameters, using the Kruskal-Wallis distribution free test of differences
between rank means. Five levels of cutting were tested for signiﬁcant
differences: C= less than one percent of the basal area removed, 1= 1-
15%, 2= 15-36%, 3= 36-55%, 4= 55-72%. H-Statistic is the calculated
test statistic to be compared with a Chi-square distribution. A probability
of 0.1 was used, with a Chi-square statistic of 7.779 assuming four degrees
of freedom ................................................................................................. 170

Distribution free multiple comparisons for growth and mortality parameters
found to have signiﬁcant differences between treatment means using the
Kruskal-Wallis distribution free test. See Table 29 for the treatments
which were compared. The test statistic is the critical value calculated
using an alpha level of 0.1 with each test statistic based on the sample sizes
of the treatments being compared. Values in each cell are differences
between the two treatments being compared. Differences between
treatments which are signiﬁcant are underlined .................................... 170

LIST OF FIGURES

Figure Page

1. Rainfall and temperature patterns for six years during which the data of the
current study was collected. The diagrams follow the criteria established by
Walter (1983). The dotted line represents temperature. The points
represent monthly precipitation. The upper shaded areas represent
moisture in excess'of potential evapotranspiration. The lower shaded areas
represent moisture deﬁciencies. (Source: Checo and Ram,
unpublished) ........................................................................................... 22

2. Plot layout for study of response of native dry forest to thinning. Sample of
1 of 4 blocks .............................................................................................. 24

3. Positions of thirty subplots from the ISA-Mao silvicultural study along the
ﬁrst two principal axes of a correspondence analysis. 45.1% of the total
variance is explained by the ﬁrst two of ﬁfteen axes. Analysis used basal
area contributions of sixteen species. The size of the plotting symbol is
proportional to the sum of the squared cosines in two dimensions. The
label indicates the cluster group assigned to each site using three cluster
techniques ................................................................................................. 58

4. Position of thirty core sites in three dimensions of a correspondence analysis
using the basal are contributions of sixteen species. The size of each
plotting symbol is proportional to the sum of the squared cosines in three
dimensions. The label indicates the cluster designation given based on ﬁve
cluster analyses using the relative basal area contributions of sixteen
species. The length of each spike is proportional to the distance from zero
along the third axis. Sixty-two percent (61.6%) of the total variation is
accounted for by the three dimensions ................................................. 62

List of ﬁgures (cont’d)

5. Position of sixteen species along the ﬁrst two axes of correspondence analysis.
Data comes from basal area contributions in 30 ’core’ plots in the ISA-Mao
silvicultural study representing distinct species compositions. Plotting
symbol is proportional to the sum of the squared cosines in two
dimensions. Symbols are: UV= C. leoganensis, CM= P. juliﬂora, AR= A.
famesiana, BA= P. brasilzcnsrlr, FR= C. cynophallophora, MO= C. ﬂexuosa,
BR= C. emarginata, CI= P. circr'nale, GY= G. oﬁicinale, CF = L.
lanceolatus, AL= B. simaruba, GU= C. con'aria, SA= M. buxzfolia, QU=
E. caribaeum, PA= T. pallida, CA= A. scleroxyla .................................. 64

6. Positions of thirty sites and sixteen species in three dimensions based on a
correspondence analysis. Analysis used basal area contributions of sixteen
species. The size of the plotting symbol is proportional to the sum of the
squared cosines in three dimensions. Numbers represent cluster group
designation of each site. Letters indicate species. UV= C. leoganensis,
CM= P. juliﬂora, AR= A. famesiana, BA= P. brasiliensis, FR= C.
cynophallophora, MO= C. ﬂaxuosa, BR= C. emarginata, CI= P. circinale,
GY= G. oﬂicinale, CF= L. lanceolatus, AL= B. simaruba, GU= C.
con'an'a, SA= M. bwczfolia, QU= E. can'baeum, PA= T. pallida, CA= A.
scleroxyla. The length of each spike is proportional to the distance from
zero along the third axis. Variance explained is 61.6% of the total.. 66

7. Position of forty-ﬁve sites from the ISA-Mao silvicultural study in the ﬁrst
two dimensions out of ﬁfteen of a correspondence analysis. The analysis
used the basal area contributions of sixteen dry forest species. The size of
the pldtting symbol is proportional to the sum of the squared cosines in
two dimensions. The label indicates cluster group designations assigned
using ﬁve different cluster analyses. Thirty-ﬁve percent (35.1%) of the
total variance is represented by these two dimensions .......................... 69

8. Position of sixteen species along the ﬁrst two axes of correspondence analysis.
35.1% of the total variance explained in the ﬁrst two of ﬁfteen axes.
Analysis used basal area contributions on 45 subplots from the ISA-Mao
silvicultural study. The size of the plottingsymbol is proportional to the
sum of the squared cosines in two dimensions. Species code are: UV= C.
leoganensis, CM= P. juliﬂora, AR= A. famesiana, BA= P. brasiliensrls, FR=
C. cynophallophora, MO= C. ﬂaxuosa, BR= C. emarginata, CI= P.
circinale, GY= G. oﬁ‘icinale, CF= L. lanceolatus, AL= B. simaruba, GU=
C. con'an'a, SA= M buxifolia, QU= E. can'baeum, PA= Palo amargo, CA=
A. scleroxyla ................................................................................................ 72

List of ﬁgures (cont’d)

9. Position along ﬁrst two canonical axes of thirty subplots from the ISA-Mao
silvicultural thinning study. The thirty subplots represent ﬁve groups
determined using three cluster analyses on data representing relative basal
area contributions of sixteen dry forest species. Labels indicate cluster

group designation. 89% of the total variance is explained by axes one and
two ............................................................................................................... 74

10. Position along ﬁrst two canonical axes of forty-three sites from the ISA-Mao
silvicultural thinning study. The forty-three sites represent ﬁve groups
determined using three cluster techniques on data representing relative
basal area contributions of sixteen dry forest species. Labels indicate
cluster group designation. Numbers indicate core sites which clustered
consistently across all techniques. Letters indicate sites which did not
cluster consistently. ’8’ represents sites clustering most often in group two,
’0 represents sites clustering in group three, ’D’ in group four and ’E’ in
group ﬁve. 79.2% of the total variance is explained by axes one and
two .............................................................................................................. 77

11. Positions of sites from the ISA-Mao silvicultural study based on their scores
from the ﬁrst two dimensions of a CA procedure using basal area
contributions of sixteen species in the 67 core sites. The size of the
plotting symbols are proportional to the sum of the squared cosines in two
dimensions. Labels represent cluster groups designations. Thirty-eight
percent (37.9%) of the total variance is represented by the ﬁrst two
dimensions ................................................................................................ 93

12. Positions of 67 ’core’ sites in three dimensions of a correspondence analysis
using the basal area contributions of sixteen species. The size of the
plotting symbols are proportional to the sum of the squared cosines in
three dimensions. Labels represent cluster group designations. The length
of each spike is proportional to the distance from zero along the third axis.
Fifty-two percent (51.7%) of the total variance is represented by the three
dimensions ........................................................................................... 96

List of ﬁgures (cont’d)

13. Positions of sixteen dry forest species based on their scores in the ﬁrst two
dimensions of a CA procedure using basal area contributions from 120
sites in the ISA-Mao silvicultural study. The plotting symbols are
prOportional to the sum of the squared cosines in two dimensions.
Symbols are: BA= Phyllostylon brasiliensis, CM= Prosopir juliﬂora, AR=
Acacia famesiana, UV= Coccoloba leoganensis, CF = Lasianthus
lanceolatus, BR= Cassia emarginata, CI= Pithecellobium circinale, MO=
Capparis ﬂexuosa, FR= Capparr's cynophallophora, AL= Bursera simaruba,
GY= Guaiacum oﬁ‘icinale, GU= Caesalpinia coriaria, PA= Palo amargo,
SA= Maytenus buxifolia, QU= Exostema caribaeum, CA= Acacia scleroxyla.
Thirty-eight percent (37.9%) of the total variance is explained by the ﬁrst
two dimensions .......................................................................................... 98

14. Positions of sixty-seven sites and sixteen species in three dimensions based on
a correspondence analysis. The analysis used the basal area contributions
of the sixteen species in each site. The size of the plotting symbol is
proportional to the sum of the squared cosines in three dimensions.
Numbers represent cluster group designation of each site. Letters indicate
species. UV= C. leoganensis, CM= P. juliﬂora, AR= A. famesiana, BA=
P. brasiliensir, FR= C. cynophallophora, MO= C. ﬂaxuosa, BR= C.
emargz'nata, CI= P. circinale, GY= G. ofﬁcinale, CF = L. lanceolatus, AL=
B. simaruba, GU= C. con'aria, SA= M. buxifolia, QU= E. caribaeum,
PA= T. pallida, CA= A. scleroxyla. The length of each spike is
proportional to the distance from zero along the third axis. Variance
explained is 51.7% of the total ................................................................ 101

15 . Positions of 120 sites from the ISA-Mao silvicultural study along the ﬁrst two
principle axes of a correspondence analysis using the basal area
contributions of sixteen subtropical dry forest species. The label indicates
the cluster group designation. Numbers indicate core sites. Letters
indicate noncore sites. The size of the plotting symbol is proportional to
the sum of the squared cosines associated with each point for the two-
dimensions represented here. Thirty-three percent (33.2%) of the total
variance is explained by the ﬁrst two dimensions ................................ 104

16.

17.

18.

19.

20.

List of ﬁgures (cont’d)

Positions of sixteen dry forest species based on their scores in the ﬁrst two
dimensions of a CA procedure using basal area contributions from 120
sites in the ISA-Mao silvicultural study. The plotting symbols are
proportional to the sum of the squared cosines in two dimensions.
Symbols are: BA= Phyllostylon brasiliensis, CM= Prosopis juliﬂora, AR=
Acacia famesiana, UV= Coccoloba leoganensis, CF= Lasianthus
lanceolatus, BR= Cassia emarginata, CI= Pithecellobium circinale, MO:
Capparis ﬂaxuosa, F R: Capparis cynophallophora, AL= Bursera simaruba,
GY= Guaiacum oﬁ‘icinale, GU= Caesalpinia coriaria, PA= Palo amargo,
SA= Maytenus bwcifolia, QU= Exostema caribaeum, CA= Acacia scleroxyla.
Thirty-three percent (33.2%) of the total variance is explained by the ﬁrst
two dimensions ......................................................................................... 107

Positions of 67 core sites along the ﬁrst two axes of a CDA procedure based
on six cluster groups deﬁned by sixteen species. Numbers represent the
group designation given based on four cluster procedures. Seventy-three
percent (72.7%) of the total variance is explained by the ﬁrst two
axes ........................................................................................................... 111

Positions of 62 sites representing ﬁve core site cluster groups deﬁned by
sixteen dry forest species, along the ﬁrst two axes of a CDA procedure.
Numbers represent the group designation given each site based on four
Cluster procedures. Eighty percent (79.8%) of the total variance is
explained by the ﬁrst two axes ................................................................. 113

Positions of 62 sites representing ﬁve core site cluster groups deﬁned by
sixteen dry forest species, along the ﬁrst three axes of a CDA procedure.
Numbers represent the group designation given each site based on four
cluster procedures. Ninety-three percent (93.3%) of the total variance is
explained by the ﬁrst three axes .............................................................. 118

Positions along ﬁrst two canonical axes of six core groups and seven noncore
site subgroups. Groups and subgroups are represented by a total of 118
sites. Groups and subgroups represent the. results of four cluster analyses
applied to data representing relative basal area contributions of sixteen dry
forest species. Labels indicate cluster group designations. Numbers
indicate cores site clusters. Letters indicate subgroups formed from
noncore sites representing the same clustering pattern (i.e. subgroup BG
indicates site clustering twice in group two and twice in group seven).
Sixty-one percent (61.2%) of the total variance is explained by axes one
and two ...................................................................................................... 121

red

List of ﬁgures (cont’d)

21. Positions along ﬁrst and third canonical axes of six core groups and seven
noncore site subgroups. Groups and subgroups are represented by a total
of 118 sites. Groups and subgroups represent the results of four cluster
analyses applied to data representing relative basal area contributions of
sixteen dry forest species. Labels indicate cluster group designations.
Numbers indicate cores site clusters. letters indicate subgroups formed
from noncore sites representing the same clustering pattern (i.e. subgroup
BG indicates site clustering twice in group two and twice in group seven).
Fifty-seven percent (565%) of the total variance is explained by axes one
and three ................................................................................................... .126

22. Positions of six cluster groups along the ﬁrst, second and third axes of a CDA
procedure applied to thirteen cluster groups represented by a total of 118
sites. Only the six core site cluster groups are plotted. Labels indicate
cluster group designations. Seventy-seven percent (76.6%) of the total
variance is explained by the three axes ................................................. 128

23. Box plots of structural characteristics, by cluster group. The horizontal line
. within each box represents the median, splitting the, ordered values in half.

The upper and lower edges of the boxes split the upper and lower halves,
respectively, in half again. The box therefore represents the range of 50%
of the values. The upper and lower edges of the boxes are referred to as
the upper and lower hinges. The lines extending vertically from the upper
and lower hinges extend to the last value(s) lying within one-and-a-half
times the range described by the box. Stars represent points more than
one-and-a-half, but less than or equal to three times the range described by
the box. Circles represent values more than three times the range of the
box away from the upper and lower hinges (Wilkinson 1988). Cluster
group three represents 18 values, group E represents 7, group six, 5 values,
group ﬁve represents 8, group one represents 11, group seven, 14, group D,
14 and group four represents 11 values. The order of the groups is based
on the ﬁrst principle axis of a canonical discriminant analysis using basal
area contributions of sixteen species on 118 sites representing thirteen
cluster groups ............................................................................................ 149

List of ﬁgures (cont’d)

24. Growth and mortality response by cluster group in the ISA-Mao subtropical
dry forest. Symbols represent averages of 100 1112 sites within the same
cluster group as designated based on four cluster techniques using relative
basal area contributions of sixteen tree species. Bars represent plus and
minus one standard deviation. The order of the groups is based on the
ﬁrst axis of a canonical discriminant analysis using 118 sites which
represented thirteen cluster groups. Bars represent plus or minus one
standard deviation. For Group Three, n=18, for Group Six, n=5, for
group E, n=7, for group 5, n=8, for Group One, n=10, Group Seven,
n=14, group D, n=14 and for Group Four, n=11, except for the
parameter, mean DKH increment. For this parameter, Group Four is
represented by only 10 sites. Growth estimates are based on diameter
measurements taken at knee height in 1986 and 1992 .................. 165

25. Growth and mortality responses to thinning in the subtropical dry forest of
the Dominican Republic. Symbols represent averages of 100 m2 sample
plots within the Same thinning level classiﬁcation. Thinning levels are: 0=
less than one percent of the basal area removed (n=32), 1= 1-15%
(n=19), 2= 15-36% (n=22), 3= 36-55% (n=25), 4= 55-72% (n=15), 5=
greater than 72% removal (n=5). Parameters are based on diameter
measurements taken at 0.5 m above ground level taken in 1986 and 1992.
Bars represent plus or minus one standard deviation ......................... 172

Introduction

Deforestation is recognized as one of the most serious environmental and
economic problems for many countries in the tropical and subtropical regions of
the world. In many of these countries, dry forests are the areas most heavily
impacted. Often large portions of the population depend on them for fuel,
lumber, animal forage, food and medicine (Fries 1992, Garcia and Alba 1989,
Cuevas and Hernandez 1987, Murphy and Lugo 1986a). Where communities
depend on the dry forest for their daily existence, the quality of life degrades as
the quality of the forests erodes due to inefﬁcient and excessive exploitation (Fries

1992, Garcia and Alba 1989, Cuevas and Hernandez 1987).

In the Dominican Republic, the dry forest life zone occupies around 21 percent of
the country and accounts for about 29 percent of the total estimated forest cover
(Knudson et a1. 1988, Laureano 1991). Fifty percent or more of the fuelwood
harvested each year comes from the dry forest and nearly two-thirds of the
population still depends on ﬁrewood and/or charcoal for cooking and heating
(Laureano 1991). Many communities are established within or adjacent to the
forest, depending on its resources for building materials, animal forage, honey
production and medicines (Cuevas and Hernandez 1987). Based on current rates
of exploitation, mature dry forest timber is expected to be exhausted by the year

2002 (Laureano 1991).

2

One goal of forest management is to increase forest productivity while assuring
mardmum efﬁciency in the use of the forest. In temperate forests, management
practices include biological and ecological interactions in establishing effective
silvicultural practices (Coile 1952, Ralston 1964, Carmean 1975, Barnes 1984).
Silvicultural treatments are based on species composition and the speciﬁc
environmental characteristics of a given site, to assure the maximum sustainable
yield of the desired products (Cajander 1926, Barnes 1984). Scientiﬁc forest
management has a much more limited history in the tropics and subtropics (Fries
1992, Lamprecht 1989), although plantation forestry has received a great deal of
attention in recent years. In the semiarid regions, there is evidence that the
productivity of non-degraded native forests are as high as plantations of
introduced species (Hardcastle 1992, von Maydell 1992, Montero et al. 1984).
More over, native forests preserve the diversity of species essential to providing
the variety of products upon which rural populations depend (Hardcastle 1992,
von Maydell 1992). Proper management of tropical dry forests has the potential -
as in the temperate zone forests- to increase productivity and assure better
efﬁciencies and sustainable yields of desired products. Understanding the
ecological relationships in these forests is essential to developing appropriate

management techniques (Fries 1992).

While limited in extent, there are projects in semiarid Africa that are successfully
integrating improved management techniques into the structure of daily life in

rural communities (Heermans 1992, Christensen 1992, Lungren 1992). Literature

3

relating to the management of native dry forest in South America, including the
Caribbean, is difﬁcult to ﬁnd. One exception is in the Dominican Republic where
the Instituto Superior de Agricultura (ISA) established the ISA-Mao Experimental
Forestry Station (EEF ISA-Mao) in a subtropical dry forest located near the city
of Mac. The station’s goal is to integrate ecological, silvicultural and economic
information to develop management models for the enhancement of growth and
yield of fuelwood and charcoal from the native dry forest (Knudson et al. 1988).
Since its initial inception, the station’s work has broadened to include research on
the use of the forest for forage and honey production (Checo, personal

communication) .

After initial studies examining variations in species composition and structure
across the landscape (Powell and Mercedes 1986), a silvicultural study was
established in 1986 to determine the effects of thinning in a forest where ﬁfty
percent of the stems were less than ﬁve cm diameter at breast height (DBH).
Treatments were initiated in 100 m2 plots at one of ﬁve target levels of thinning,
including undisturbed control plots. The initial experimental design assumed
variations in species composition and site conditions would be controlled using a
randomized complete block layout with subsampling. In 1988, analyses of
diameter and height growth suggested that a positive response was occurring at
the highest level of thinning (Knudson et al. 1988). In 1992, however, analyses
revealed high rates of mortality and inconsistent growth results within treatments.

Standard univariate statistical analyses were not able to account for the variation

4

in growth and mortality (Checo, personal communication). Therefore,
multivariate analyses were initiated to examine the relationships of species
composition to site productivity. Multivariate statistical analyses (MV A) can
jointly examine many interrelated variables. Using MVA techniques of
classiﬁcation and ordination, species distributions before thinning were examined
in relationship to site quality and disturbance history. This approach was expected

to provide insight into the dynamics of the dry forest ecosystem and explain some

of the effects of controlled thinning.

The goals of this study are to (1) determine whether patterns of species
composition existed among the sampled sites in the unthinned forest, (2) examine
the implications of species distribution with respect to disturbance history and
underlying environmental gradients, and (3) explore the relationship of growth

and mortality with respect to species composition.

The objectives are:
1) Examine a subset of sites for natural groupings of sites with
similar species compositions.
2) Inspect the entire data set to determine if similar groupings can be
detected.
3) Using data ordinations, examine site groupings for implications in terms

of underlying environmental gradients.

5

4) Based on the results of these pattern analyses, use information about
dominant species, site conditions and overstory structure to assess
implications with respect to disturbance histories and underlying

environmental gradients.

5) Based on the results of the pattern analyses, examine groupings of sites

for differences in growth and mortality.

Literature Review

Site Classiﬁcation

Determining the potential productivity of a forest is essential for applying the
appropriate treatment and managing for the optimum species. In his publication,
The Theory of Forest ijes, Cajander (1926) proposed a series of quality classes
that identiﬁed forest units with similar growth potential based on characteristic
understory species associations. Arranging units in the several quality classes from
most to least productive, Cajander (1926) found that variations in growth,
dominant tree height, structural characteristics, and soil conditions all followed

recognizable patterns.

In the ensuing years, many other systems of site classiﬁcation have been
developed, each with its own emphasis and vocabulary (Rowe 1984). Holdridge
(1967) developed a classiﬁcation system to explain global variations of vegetation
and productivity. Using average annual temperature, precipitation, and potential
evapotranspiration (PET), Holdridge divided the globe into a series of Life Zones,
each with characteristic climatic conditions, and consequently, characteristic
vegetation. Holdridge found that, while species varied within a Life Zone from
region to region, the form and structure of the climax vegetation were remarkably
similar even among sites from different continents. Walter (1985) also emphasizes
moisture and temperature relationships in his Climatic Diagrams. Unlike

Holdridge, however, Walter’s system illustrates seasonal variations, rather than

7

annual means. Sites with similar seasonal patterns of rainfall and temperature
would be more alike in their vegetation and potential productivity than all the

sites with the same average annual climatic conditions (Walter 1985).

Macroclimatic conditions set the absolute limits on vegetative development
(Holdridge 1967), but the direct effects of temperature and moisture conditions
on vegetation are modiﬁed by topography, soils, and the vegetation itself (Walter
1985, Thomas and Squires 1991). Within a region of similar climatic conditions,
there are variations in the landscape which must be understood to apply effective
resource management techniques. In Germany, a comprehensive, integrated
approach has been developed called ecosystem classiﬁcation. In this system,
classiﬁcation begins by deﬁning relatively homogeneous units based on
macroclimatic conditions. Within these units, forestland is further divided by
simultaneously using vegetation, soils and topography. Initially, each of these
attributes was studied individually. Then, using interdisciplinary teams of
specialists, the interrelationships between the factors were examined and criteria
developed for determining site classes with homogeneous conditions for growth
(Barnes 1984). A similar system has been developed and tested in Michigan. The
key to this system is understanding the relationships between: (1) vegetation
(overstory, understory, groundcover) and topography, (2) between vegetation and

soils, and (3) between topography and soils (Barnes, et al. 1982).

8

In tropical America, Beard (1944) developed a classiﬁcation of climax vegetation
based on physiognomic characteristics. Just as Holdridge (1967) observed that
under similar climatic conditions, forests will assume a similar structure regardless
of species composition, Beard found that forests with different species
components but with characteristic physiognomy are found repeated throughout
tropical America. Beard suggested that these physiognomic groups can be
organized along gradients, corresponding to decreasing availability of moisture.
Beard (1944, 1953) observed that, while moisture is the primary factor affecting
the vegetation, available moisture is determined by the mutual interactions of
climate, topography and soils. Beard’s system is primarily descriptive and was not
developed for use as a management too]. More over, the system is based on
undisturbed vegetation and is therefore not easily applied to the vast areas of
forest affected by human intervention (Beard 1944, Holdridge 1967).
Nevertheless, Beard’s system does classify forest sites along gradients which can be
interpreted in terms of productivity. The relationships he describes between
separate physiognomic groups may have implications in understanding

successional processes (Beard 1944).

To determine the production potential of a forest in the semiarid tropics, both
Holdridge and Walter’s systems are helpful in establishing limits of productivity
within a relatively broad geographic region. Beard’s classiﬁcation system offers
insights into patterns of forest structure and composition across a landscape, and

suggests implications with respect to moisture availability and disturbance history.

9

Finally, however, within the context of site potential, the total complement of
vegetation-soil-topographic interactions must be described and understood if an

optimum management program is to be developed.

Site Factors

Many studies have examined the interrelationships between species distributions,
site productivity, climate, topography and soil in temperate America (Coile 1952,
Ralston 1964, Carmean 1965, 1975, Kercher and Goldstein 1977, Pregitzer, et al.
1983, Padley 1989, Fisher 1994). While their relative importance differs from site
to site, the total complement of factors found to be important remain constant
among most of the studies. Climate determines the total moisture available.
Aspect, slope length, slope steepness and slope position affect soil development
and soil moisture relationships. They also control angle of light entry and total
irradiation. Soil texture, depth and rockiness are inﬂuenced by topography and in
turn affect the development of vegetation. Plants are ultimately indispensable for
soil accumulation, keeping ﬁne particles in place against the force of gravity,
adding organic matter and cycling nutrients up from the subsurface horizons.
Disturbance, particularly human intervention, may affect species distributions in
random ways (Barnes et al. 1982) and change potential site productivity through
erosion and soil compaction. In the tropics and subtropics, excessive exploitation
of dry forest trees usually leads to a reduction in species diversity and increasing
dominance by more xerophytic species such as cacti and thorny legumes

(Holdridge 1945, 1967, Tamayo 1963, Powell and Mercedes 1986). Nevertheless,

10

post disturbance vegetation often shows characteristic patterns which can be
related to underlying physical conditions (Cajander 1926, Grigal and Goldstein

1971, Kercher and Goldstein 1977, Whitney 1991).

Many factors which inﬂuence soil development, species distributions and site
productivity in the temperate zone appear to be of equal importance in the tropics
(Beard 1944, 1953, Asprey and Robbins 1953, Loveless and Asprey 1956,
Markham and Babbedge 1979, Furley and Newey 1979, Powell and Mercedes
1986, Yair and Shachak 1987, Thomas and Squires 1991). However, the relative
intensities of each factor and the interactions between factors differ in the
semiarid areas (Amon 1992). Parent material is usually more important in
determining soil characteristics due to less leaching (Amon 1992), although run-
off from slopes may greatly increase productivity and soil development at the
slope’s base (Walter 1985, Yair and Shachak 1987, Amon 1992). Also, near the
equator, east and west slopes are the driest, versus the south and southwestern
slopes in the temperate zone. In the arid zones, on ﬂat ground, water will sink to
greater depths on sandy soils than on clay soils, and therefore remain available for
plant growth for a longer period after a single rain event. Rocky soils may permit
even deeper saturation and may therefore present the least drought like

conditions in some situations (Walter 1985, Lamprecht 1989).

Although site factor studies in the tropics which describe changes in species

composition in relation to soil conditions and topographic position tend to be

11

more descriptive than quantitative, they illustrate general trends. In applying
Beard’s system to the vegetation of Jamaica, Loveless and Asprey (1956) noted
that two related formations found on limestone derived soils were associated with
different degrees of slope steepness and soil depth. A third, more complex
formation occurred on the adjacent lowlands where alluvium material overlies
marine clays. In Ghana, Markham and Babbedge (1979) studied the transitions
between forest and savanna along transects laid across slopes representing nine
meters change in elevation. They found that the changes in vegetation were
associated with slope position, soil depth, nutrient status and moisture availability.
In Belize, Furley and Newey (1979) also found distinct species associations
corresponding to slope position. They found soils to be deepest and biomass
greatest on foot slopes. Mid-slope sites had more shallow soils and the vegetation
was shorter and forest structure less complex. Summit forest sites had species
described as typical of more mature forest, although the vegetation was generally
more open and included cactus species. Overall, they found that soil depth and
moisture content tended to decrease from foot slope to summit, while pH,
exchangeable Ca and the percent sand fraction tended to decrease from summit
to foot slope. At the ISA-Mao station in the Dominican Republic, Powell and
Mercedes (1986) found that species and structure changed rapidly when the
topography became more rolling. Ridges were noted to have structure and

complexity similar to highly disturbed areas in more level terrain.

12

Detailed models of ecosystem interactions for semiarid tropical forests are not
available in the literature. However, the studies available show that vegetation,
topography and soil relationships found to be important in the temperate zone are
also important in the tropics, perhaps even more so in the dry regions.
Topography affects soil development, moisture relationships, and total irradiation
which in turn affect the species distributions and the potential productivity of a

given site.

Site Disturbance

The effects of cutting on species distributions in the subtropical dry forest has not
been studied extensively. In their study of forest formations in Jamaica, Loveless
and Asprey (1956) noted that an area representing evergreen bushland (Beard
1944) had been affected by extraction of ﬁrewood and fence posts. The authors
suggested that composition of the forest was essentially unchanged from a climax
formation because harvested trees coppice extensively and can therefore reform
the original canopy rapidly. In Venezuela, Tamayo (1963) noted that the most
heavily disturbed areas of dry forest were nearest to population centers. These
forests consisted of small shrubby legumes such as Prosopis juliﬂora and columnar
cacti, with scattered clumps of ground cacti in the Opuntia genus the only
remaining ground vegetation. In a general review of the dry forests of the
Dominican Republic, Holdridge (1945) noted that local dominance of the
subtropical dry forest by Lemaireocereus hystn'x was probably due to heavy cutting.

In their study of species composition and structure in the Mao forest, Powell and

13

Mercedes (1986) observed that areas along the forest edge and adjacent to major
foot paths appeared to be the most highly disturbed. The authors associated the
cacti, Lemaireocereus hystrix and Consolea moniliformis and the trees, Prosopis
juliﬂora and Phyllostylon brasiliensis' with highly intervened areas. Maxwell (1985)
includes Acacia tortuosa in the list of species dominating disturbed sites. In an
area previously cut and cultivated, Powell and Mercedes (1986) found that the

species Exostema caribaeum dominated the canopy.

Overall disturbance tends to reduce forest diversity, increase the dominance of
cacti and thorny legume species and creates a low open structure of small trees
(Holdridge 1945, 1967, Tamayo 1963, Powell and Mercedes 1986). However,
since cutting for charcoal and construction is selective, large trees in a given area
are not by themselves an indication that the site has not been extensively
disturbed (Maxwell 1985).- On the other hand, an area of forest with large
specimens of species known to be favored for charcoal and/or construction would
suggest that the area has been relatively free of signiﬁcant human disturbance

(Powell and Mercedes 1986).

Multivariate Analyses

Multivariate methods are statistical techniques used to examine the variance
expressed in a data set, particularly the covariance observed among many
interrelated variables. Normally, researchers use multivariate analysis (MVA)

when they are interested in the patterns expressed in a data set rather than in

l4
quantifying a treatment effect. In many ecological studies, unlike most designed
experiments, the levels of the pertinent parameters are uncontrolled, their
distributions are usually not statistically normal, and the relationships between
parameters are often unknown or not well understood. Standard statistical
procedures are therefore not appropriate, nor can they elucidate the relationships

which are of interest (Digby and Kempton 1987).

Two broad categories of MVA techniques in studies of species distributions are
those used for classiﬁcation and those used for ordination. Classiﬁcation assumes
that sites can be numerically partitioned into discrete units while ordination
perceives community variation as continuous along one or more gradients (Pielou
1969, Grigal and Goldstein 1971, Digby and Kempton 1987). Although
communities may be continuous rather than discrete, recognizing discrete points
along the continuum is still useful for understanding the interactions between
species and their environment (Pielou 1969, Grigal and Goldstein 1971, Kercher
and Goldstein 1977, Pregitzer and Barnes 1984, Digby and Kempton 1987).
Digby and Kempton (1987) recommend classiﬁcation techniques along with

ordinations of the data to examine the relationships between groups.

Pregitzer and Barnes (1984) used a combination of classiﬁcation and ordination
techniques to examine differences in soil and topographic characteristics between
site units previously delineated using an Ecological Classiﬁcation System (ECS).

They found that the ﬁeld based ECS had identiﬁed classiﬁcation units which

 

 

15

differed in topographic and soil factors known to strongly inﬂuence tree growth
(Pregitzer and Barnes 1984). Padley (1989) also found that separate ordinations
using environmental and vegetation data sets were highly correlated with each

other and with previous ECS designations.

In their study of an oak hickory watershed in the Smoky Mountains of eastern
Tennessee, Grigal and Goldstein (1971) used four hierarchical clustering
techniques to classify 290 sites. Clustering techniques are numerical analyses
designed to separate units into distinct groups based on some type of distance
matrix. Many different methods exist for cluster classiﬁcation, each of which may
result in a different grouping of the data. Using several different methods and
comparing classiﬁcations across methods removes some of the potential for
subjective interpretation of the data based on a single, well-chosen procedure that
supports the investigator’s preconceptions (Pielou 1969, Digby and Kempton 1987,
James and McCulloch 1990). Grigal and Goldstein observed that within each of
the four cluster techniques, at a level of classiﬁcation which divided the data into
four large distinct groups, each group had a distinctive species composition which
appeared to characterize that cluster group. "Characteristic species" were
determined by comparing the average relative basal area contribution of each
species within cluster groups relative to the average contribution across the entire
watershed. Species which on average contributed more basal area to a particular
group than to the watershed as a whole were deﬁned as characteristic of that

group. Within each cluster technique, the characteristic species composition of a

16
given group corresponded to the characteristic composition of one of the groups

in each of the other three techniques.

Grigal and Goldstein also found that in each of the four major groupings, some of
the sites remained consistent across all techniques, while other sites changed
group membership depending on the technique used. Of the 290 sites, 131
grouped consistently in all four methods. The authors termed these ”core" sites
and interpreted them as representing discrete points along the species distribution
continuum. Sites that were inconsistent in their group membership were noted to
have species compositions intermediate between the characteristic compositions of
the major groups. It was assumed that these intermediate sites changed
membership according to the bias of a particular cluster technique (Grigal and

Goldstein 1971, Kercher and Goldstein 1977).

To examine the relationships between clusters, Grigal and Goldstein (1971) used
an ordination technique referred to as canonical variate analysis. Based on the
values of a particular set of variables, canonical variate analysis (CVA) maximizes
the ratio of between group variance to within group variance to give the best
separation of the groups (Digby and Kempton 1987). Whereas other ordination
techniques such as principal components analysis (PCA) give equal weight to all
the variables, CVA develops a function which gives the greatest weight to the
variables which are the most consistent within each group (Digby and Kempton

1987). Because of this property, groups formed based on cluster analyses will

 

l7

tend to separate in canonical space bases on the species which are common to
sites within a cluster group, but uncommon to sites in other cluster groups.
Species which are erratic within a group, or consistently present across all groups,
will have less impact on the overall ordination relative to analyses which do not
account for any structure within the data set. Grigal and Goldstein (1971) found
that the four major groups of sites identiﬁed using cluster analyses and CVA in

combination could not be identiﬁed using principal components analysis.

Plotting clusters from each of the techniques along the ﬁrst two canonical axes of
four respective canonical variate analyses, Grigal and Goldstein (1971) found that
the four major groupings clearly separated from each other. Minor groups
formed from the results of one classiﬁcation plotted in close association with one
the major groups, suggesting a relationship based on similar species compositions,
as indicated by the average composition of sites within the respective groups.
Grigal and Goldstein interpreted the clear separation of the groups in canonical
space as a strong indication that the cluster techniques had recovered natural

groupings within the data set.

As in most ordination methods, CVA includes a centering of the data, such that
the origin or centroid, represents the grand mean ’of the data set, across all
variables. The position of a site or a cluster with respect to the centroid therefore
represents the site’s degree of variation from the overall mean. Generally, both

distance and direction are signiﬁcant (Digby and Kempton 1987, Greenacre 1993).

18
In the CVA’s applied to the results of each of four cluster techniques, Grigal and

Goldstein (1971) found that the major groups had the same relative positions in
two dimensions for each CVA procedure. More over, when CVA was applied to
the subset of sties which clustered consistently across all techniques, Grigal and
Goldstein found that each of the four main clusters plotted in one of the four
quadrants formed by the juxtaposition of the ﬁrst two canonical axes. The
position of each group with respect to the centroid was in a different direction
from the rest, suggesting that the groups represented different extremes of one or

more underlying environmental gradients.

Using the vegetation groups deﬁned by Grigal and Goldstein (1971), Kercher and
Goldstein (1977) supplemented the data with measurements of environmental
parameters at each site, including slope position, insolation, slope angle and age.
Following a series of procedures developed from their previous use of CVA, the
authors found a high association with the groups as described by the site factors
and those described by the vegetation. In the process, Kercher and Goldstein
(1977) determined that age and slope position were the two variables most
signiﬁcant in the separation of the four groups. They extrapolated this to suggest
that, given the time since disturbance and the position in the watershed landscape,

they could predict the vegetation most likely to dominate a given site.

Fisher (1994) used another ordination method, referred to as correspondence

analysis, to study the species composition of pre-settlement forests in northern

19

lower Michigan. Correspondence analysis (CA) is an ordination technique
operating on a two-way contingency table of counts of objects (James and
McCulloch 1990). It is primarily a graphical technique used to illustrate the
relationship between data points in as few dimensions as possible. CA assumes
chi-square distances for the interpretation of graph plots. If scores calculated for
two or more sites are similar, the sites can be assumed to be geometrically close,
if the data meets chi-square assumptions. As with CVA, the procedure begins by
centering the data set, so that zero represents the grand mean across all variables.
Therefore, direction, as well as distance may be used in interpreting the positions
of the individual sites. Unlike CVA, correspondence analysis does not assume any
a priori grouping of the data. Fisher (1994) found that CA was able to capture
most of the species and site variance in the ﬁrst two axes. Plotted with these two
axes, the species followed an ordination along the primary axis according to
moisture stress tolerance, with species near the centroid representing those with
intermediate tolerance. Ordination of the sites suggested a similar pattern, with
sites associated with particular landforms and soils following a gradient of relative

soil moisture availability.

It can generally be concluded that patterns inherent in the species distributions
across a landscape can be recovered using a combination of classiﬁcation and
ordination techniques. These patterns frequently reﬂect growth related gradients
associated with speciﬁc topographic and soil characteristics. While no studies are

available which describe the use of these methods in the dry forests of tropical or

 

20
subtropical America, the literature does suggest that vegetation-soil-topography
relationships exist and that they are similar to relationships which have been
described for temperate zone forests. Although disturbance has commonly altered
the original species distributions in tropical and subtropical dry forests (Murphy
and Lugo 1986a), distinct patterns may still exist, inﬂuenced by a combination of
environmental and disturbance factors. In as much as they are not random, MVA

techniques should be capable of capturing such patterns.

Materials and Methods

Site Description

The ISA-Mao Experimental Forestry Station is located in the western part of the
Cibao Valley in the Dominican Republic (19°35’ N and 71°4’ W). Occupying
about 1000 ha in a semiarid region of the country, the station experiences two
rainy seasons. One is from March to June, the other is from September to
December (Knudson et al. 1988). Average annual precipitation is 647 mm, but is
irregular from year to year (Figure l) and the average annual temperature is
27.1°C (Checo and Ram, unpublished). The ratio of potential
evapotranspiration (PET) to precipitation is between 2.0 and 4.0 (Knudson et al.
1988). Based on average annual rainfall, temperature and PET, the Mac forest is

classiﬁed as subtropical dry forest, according to Holdridge (1967).

Located between the Mac river ﬂoodplain to the northeast and the Cordillera
Central mountains to the southwest, the station is characterized by rolling hills
with elevations ranging from 78 to 175 meters above sea level. The site ranges
from level terrain with deep soil to steep slopes with shallow soils. Soils are
derived from limestone parent material with pH varying from 7.8 to 8.4 (Knudson
et al. 1988). Soils in the area have been classiﬁed in the subgroup Ustalﬁc
Haplargids, which are arid soils with higher than normal clay content (CRIBS

1977), but Aridic Haplustalfs may be a more appropriate classiﬁcation because

21

 

622 mm 870 mm m

 

 

 

 

iii“

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.SIIOIIII“

 

 

 

 

O
C U I A I J J A I 0 O I

1989

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

ﬁgure 1 Rainfall and temperature patterns for six years during which the data of the current

‘ study was collected. The diagrams follow the criteria establ'nhed by Walter (1983). The dotted
lire represents temperature. The points represent monthly precipitation. The upper shaded areas
represent moisture in excesb of potential evapotranspiration. The lower shaded areas represent
moisture deﬁciencies. (Source: Checo and Ramm. unpublished).

23

soils with an aridic moisture regime are not normally associated with well
developed forest vegetation (Mokrna, personal communication). A high bulk
density soil layer is present at depths between 35 and 50 cm in many parts of the
forest (Checo, personal communication). This may represent a phenomenon
common in semiarid regions where rapid evaporation of moisture limits the depth
of rainfall penetration, resulting in the accumulation of eluviated calcareous.

material at shallow depths, forming a hardpan, or caliche (Arnon 1992).

The species composition and structure of the forest are complex. Like most of
the subtropical dry forest in the Dominican Republic, the ISA-Mao forest has
been subjected to both long term as well as relatively recent cutting, clearing and
burning. Clearing for charcoal production was originally the principal source of
disturbance, with evidence of old charcoal piles still present in many areas.
Clearing for cultivation also occurred as well as considerable animal grazing
(Powell and Mercedes 1986). Since 1978, such disturbances have been controlled,
but incursions still occur by individuals from adjacent communities. These
incursion involve removing individual trees for fence posts and house construction

(Checo, personal communication).

The history of selective cutting in the ISA-Mao forest has resulted in a mosaic of
site histories. Some areas were completely out over, others had only a few trees
extracted and a few areas have remained relatively undisturbed. The exact history

of any given site is not well known. Most dry forest species sprout vigorously

24

when cut, suggesting that trees with numerous stems may have been subjected to
cutting at some point. Observations by Murphy (personal communication) in the
Guanica dry forest of Puerto Rico suggest, however, that some trees may have
multiple stems for reasons other than cutting. Exceptional moisture stress due to
natural conditions may also be responsible for a higher incidence of multiple
stems. Study plots in the Mac forest with the highest proportion of multiple
stems may either represent the greatest level of intervention or the most severe
environmental conditions. Exploitation of the forest also generally results in a

shorter overall height and a greater dominance by smaller boles (T amayo 1963,

Powell and Mercedes 1986). A

 

 

 

 

 

. . . . , , , O O W 1
reduction in relatrve morsture avarlabrlrty OOO// 5],",
can result in similar conditions (Beard 29" M
1944).

40% 80%
Original Study Design
0% 60%

 

 

 

 

Within this diverse landscape, a
silvicultural thinning study was

Figure 2 Plot layout for study of response of
established in 1985-1986 (Knudson et a]. native dry forest ‘0 “ﬁlming- Sample 0“ 0f 4

blocks.
1988). Patterned after a randomized
complete block design, four blocks, each with six 50 X 50 m plots, were located in
sections of the forest representing different structure, topography and species

dominance. Target thinning levels of 20, 40, 60 and 80% were randomly assigned

to four of the 50 X 50 In areas. The two remaining sections were designated as

25
controls. Within each of the 50 X 50 m areas, ﬁve permanent circular subplots

were systematically located, each 100 m2 (Figure 2). Each set of subplots was
originally designed to represent a subsample of their respective treatment plot.
Due to restrictions of time and resources, thinning treatments were applied only
within each of the circular subplots (Checo, personal communication). Each
subplot became its own experimental unit, rather than a representative of the

larger 50 X 50 m area.

Because the subplots are not effective subsamples for 50 X 50 m plots, they have
been considered independent samples of the area of forest corresponding to the
silvicultural study. Statistically, the subplots are not independent since each set of
ﬁve was systematically located within their respective 50 X 50 m area. Even if
each set of subplots had been randomly distributed within their treatment plots,
their independence would be ecologically questionable. In a plant ecosystem, a
random distribution might actually consist of randomly distributed clumps of
individuals rather than a random mix of the individuals themselves (Pielou 1969).
Because of growth and reproductive patterns, two trees next to each other may be
more likely to be of the same species simply because of their proximity, rather
than due to some environmental characteristic. Closely situated sample plots
might show similar species compositions because they all happen to fall within a
random clump of species, rather than because the data points reﬂect some
common underlying environmental characteristic (Pielou 1969). As more plots are

included in the sample and when plots are distributed over a large geographic

26
area, random associations tend to cancel each other out. Pielou (1969) also
observes that the ﬁner the species composition mosaic, the less likely proximate
sites are to show spurious associations. In this study, subplots within a single 50 X
50 m plot may be more likely to have had the same species composition because
of their proximity, not necessarily because of similar environmental attributes. It
is assumed, however, that the study has enough sample points over a large enough
geographic area to represent a diversity of site conditions. This should help

minimize the probability of contriving arbitrary site relationships.

Before thinning (1986), all trees within each subplot with at least one stem greater
or equal to 2.5 cm at breast height were identiﬁed by common species name,
measured for height, diameter at breast height (DBH) and diameter at knee
height (DKH). One height measurement was recorded for each tree. Stems
were then removed based on the target thinning level. Residual trees were
remeasured post-thinning and again each year through 1992. The goal of stem
removal was to reach targeted treatment levels while maintaining a one to one
relationship between the percentage of stem and basal area removed. In practice,
due to the complex structure of the forest, the result was not as precise as desired.
Comparing actual removal of stems and basal area with the target thinning rate,
clearly shows that cutting was not consistent within treatment plots, nor within
treatments across blocks (Table 1). This was discovered while preparing the data

for multivariate analyses.

27

Table l A subsample of sites from the thinning study in a subtropical dry forest of the Dominican
Republic demonstrating reassignment of the sites to new treatment designations based on the
percent of initial basal area actually removed in the cutting. The first digit in Site ID indicates the
block number (1 through 4). The second and third digit represent the site number (1 through 30)
within each block. The original treatments are listed in the ”Target thinning level" column. The
actual percentages of stems and basal area removed are listed in the last two columns. The
assigned cutting levels are the treatment designations used for analyses in this study.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

r ﬂ!

ID Target Assigned Actual basal Actual stems f
thinning cutting level1 area removed (%)

level (%) removed (%)

212 20 C 0.0 0.0

213 20 c 0.0 0.0 II
214 20 C 0.0 0.0
215 20 C 0.0 0.0
111 20 1 1.6 7.4
112 20 1 3.1 8.7
211 20 1 3.9 3.3
114 20 1 10.7 21.1
415 20 1 12.3 13.0
313 20 2 15.9 44.4
115 20 2 16.5 125
113 20 2 18.3 21.4
411 20 2 23.7 27.1
311 20 2 24.2 18.5
314 20 2 26.3 393
312 20 2 27.3 28.9
413 20 2 31.3 41.1
412 20 . 2 35.2 37.8
414 20 3 46.8 42.9
315 20 3 54.8 45.9

 

 

 

 

 

1Assigned cutting level ”C”: < 1.0% basal area removed. "1": 2 1.0% and S 15.0% removal. "2":
>15.0% and .<. 36.0% removal. ”3": >36.0% and S 55.0% removal. "4": >55.0% and S 72.0%
removal. "5": >72.0% basal area removed.

28

Because thinning was not consistent within targeted treatment levels, subplots
were assigned a new treatment designation according to the actual basal area
removed (Table 1). Basal area was used as the sole criteria for reassignment.
Levels for reassignment were

Controls < 1.0% basal area removed

Cutlevel l 2 1.0% and s 15.0% basal area removed

Cutlevel .2 > 15.0% and s 36.0% basal area removed

Cutlevel 3 > 36.0% and s 55.0% basal area removed

Cutlevel 4 > 55.0% and s 72.0% basal area removed

Cutlevel 5 > 72.0% basal area removed.
Because the areas where the data was collected were not actually subsamples of
the 50 X 50 m plots, they will be referred to as "sites" rather than "subplots". See
Appendix A for the complete list of cutting level designations based on basal area

removal.

Numerical Methods

Variable Selection

Forty-four species were identiﬁed in the initial pre-harvest inventory within the
silvicultural study area. For analyses of species-site relationships, sixteen species
were selected which were found to be the most dominant across the entire study
area. To select these species, relative importance values were calculated using

stem density, tree density, basal area density and frequencies as follows:

1v,= RDS, + RDN, + RDMX + RFRX

29

Where:
IV, a relative importance of species x.
12133, = [E stems ofspeciesx/z all stems] x 100
RDN, a [2 trees ofspeciesx/ Z all treeslx 100
RDM. - (E basal areas(DKH) of speciesx / 2 all basal areas(DKI-l)] x 100

RFR, a [frequency of species x / E frequencies of all spp] X 100

The decision to select sixteen species was not arbitrary. Several versions of the
importance values were calculated and it was noted that while each version
ordered the species differently, the top sixteen species were always the same. For

the species selected, each comprised at least 1.4% of the total basal area (Table

2).

To represent species dominance, stem counts, tree counts, and basal area
summations were all possibilities based on the data collected in the silvicultural
study. Each of these measures weights species differently according to the
species’ particular structural form. Stem counts give greater relative weight to
species with multiple stems versus those species which tend to have single stems.
Cinazo (Pithecellobium circinale- #2, Table 2), for example, comprised 13.1% of
all the stems, but only 5.3% of the trees. Quina (Exostema caribaeum- #4, Table
2), on the other hand, made up 6.3% of the stems, but 8.0% of all trees. Using
stem density to compare species contributions gives somewhat greater weight to

Cinazo than to Quina relative to using tree counts as a measure of species

dominance.

30

Table 2 Relative measures of density for forty-four species in the silvicultural thinning study at the
Mao-ISA subtropical dry forest experimental station. Calculations are based on measurements
taken at knee height (0.5 m). Species are listed in descending order of their importance value.

 

 

 

 

  

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

  
  
  
 

 

 

 

 

- Relative stem Relative tie: Relative basal Relative Importance

density (%) density (%) area density (%) frequency (%) value II
24.3 29.3 29.7 8.9 92.1
I 13.1 5.3 5 5 6.7 30.7
I Guatapanal 5.7 4.7 10.1 6.5 27.0
i Quina 6.3 8.0 4.9 6.2 253
I Brucon 5.2 6.4 4.9 7.1 23.6
' Candelén 5.4 6.6 7.8 3.7 235
' Guayacﬁn 4.7 5.2 2.5 7.2 19.6
f Almacigo 2.0 2.7 8.6 4.1 17.3
i Cambron 4.7 4.8 3.5 4.0 17.0
ll 10 Aroma 3.6 3.2 2.7 4.0 135
11 Sangretoro 3.3 2.9 2.4 4.1 12.7
12 Mostazo 2.7 2.2 2.6 43 11.7
13 Cafetén 3.2 2.9 1.5 3.4 10.9
14 Frijol 1.4 1.6 2.8 33 9.2
15 Palo amargo 1.6 1.7 1.4 2.8 75
16 Uvero 2.0 1.5 1.7 22 7 .4
17 Palo de burro 1.1 1.4 0.9 3.0 63
18 Ojo de palorna 1.6 1.5 0.9 2.0 6.1
19 Palo blanco 1.2 0.9 0.6 1.8 4.6
20 Ciavellina 0.6 0.8 0.4 1.6 3.4
21 Sopalpo 0.7 0.4 1.0 1.0 3.1
22 Trejo 05 0.7 0.2 1.6 3.0
23 Tabacuelo 05 0.7 03 13 2.7
24 Canelilla 0.6 0.7 0.2 1.1 2.6
[25 Hueso de chivo 0.5 0.5 0.5 1.0 2.4
26 Frijolillo . 0.6 0.4 0.5 0.9 2.4

27 Amarra carnero 0.4 0.4 03 1.0 2.1 ||

Cereza 0.4 0.4 0.1 1.0 1.9 II
Palo de cairnan 0.2 0.3 0.2 0.7 15

 

 

 

 

 

 

  

.r. h ..w. L .11. h

h r. a

.1... h

a L~

Ire

lug

lug

lies

ﬁlth

31

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 2, continued
density (%) density (%) area density (%) Frequency (%)
02 03 0.1 0.7
02 03 0.1 0.6
02 03 02 0.4
02 02 0.1 0.6
03 03 0.1 0.4
03 02 02 02
[36 Penda 0.1 0.1 on 03 05
37 135666611 0.1 0.1 0.0 02 03
I38 Candeli 0.1 0.0 0.0 0.1 03
I39 Bsyahonda 0.1 0.0 0.0 0.1 02
40 Ciruela 0.1 0.0 0.0 0.1 02
41 Orerno de buey 0.0 0.0 0.0 0.1 02
42 Cabra on 0.0 0.0 0.1 02

 

Una de gato

0.0

0.0

0.1

 

 

Chicharron

Tree density appears to correspond more closely with basal area density for

Cinazo. Quina, however, made up only 4.9% of the total basal area at knee

 

0.0

 

 

0.0

 

0.1

 

 

height, while Guatapanal (Caesalpinia coriaria- #3, Table 2), with only 4.7% of

the trees, made up 10.1% of the basal area. Tree density therefore gives more

weight to Quina than to Guatapanal. Using basal area density on the other hand,

gives signiﬁcant weight to Almacigo (Bursera simaruba- #8, Table 2), with 8.6% of

the basal area, although the species comprised only 2.0% of all the stems and

2.7% of the trees. Because the Mao forest is composed of species which are

structurally diverse, each measure of species dominance gives a different weight to

a different set of species and none of the measures is perfect.

In preli
hated
that die
measur
recur
caliper
less rigi
he 5i;
have all

United

32

In preliminary analyses, all three measures were used. Preliminary calculations
focused on stem and tree counts because an initial review of the data sets showed
that diameter measurements were variable. For example, some stems were
measured as smaller in 1992 than in 1986. At least two factors contributed to this
measurement error. First, pre- and post-thin inventories used the average of two
caliper readings for stem diameter measurements. Subsequent inventories were
less rigorous (Checo, personal communication). Second, many dry forest trees
have signiﬁcant taper and subsequent yearly measurements on the stems may not
have always been at the same point. In the dry woodlands of Arizona in the
United States, diameters are measured at the root collar (DRC) (McPherson
1992) presumably making consistent measurements easier and avoiding the

problem of multiple stems.

In the end it was decided that basal area would be used to represent species-site
interactions in the ISA-Mao silvicultural study. Basal area often shows a strong
relationship with total tree biomass and canopy cover (Barth and Klemmedson
1982, Maxwell 1985, Lamprecht 1989). Grigal and Goldstein (1971) used basal
area contributions in their multivariate analysis of species relationships across an
oak hickory watershed in Tennessee. Finally, in the temperate zone, basal area is
usually assumed to be independent of stand age, versus tree and stem counts,
which change radically over time. To decrease the effect of measurement error,
diameters at knee height (DKH) were chosen over diameters at breast height

(DBH) to calculate basal area. Diameters at knee height may introduce less error

for w.
ltlSll

teas

Rest

and;

1111}:

Data
for 1h
111 P
lain
some
1‘"ng
1‘1,ng

Md

11.1119
Match

ill

11,,

33

for two reasons: (1) More of the trees are single stemmed (82% at knee height
versus 70% at breast height) and are therefore represented by a single
measurement rather than multiple measurements. And (2) stem taper at 0.5 m
above ground level is less acute, making subsequent measurements less variable.
Residual trees were remeasured in 1986 immediately after thinning. Basal area
summations calculated from diameter measurements for trees present in both pre-

and post-thin inventories were averaged and the average used in subsequent

analyses.

Species Identification

Data collected in the silvicultural study used local names for tree species. Sources
for the Latin names of the sixteen species selected for species-site analyses are
van Paasen (1986) and Knudson et al. (1988). Where the two sources conﬂict, the
Latin name used by Knudson et al. (1988) was followed. Further veriﬁcation for
some of the species was possible using Common trees of Puerto Rico and the
Virgin Islands (Little and Wadsworth 1964) and Trees of Puerto Rico and the
Virgin Islands (Little et al. 1974). In general, the description for each species
provided in this source matched the characteristics noted by van Paasen (1986).
However, two species, Cafetan and Palo amargo, which are listed by Knudson et
al. (1988) as Lasianthus lanceolatus and Trichilia pallida, respectively, did not
match with the habitat distributions described by little and Wadsworth (1964).
Nevertheless, the Latin names provided by Knudson et al. (1988) will be used in

this study. Table 3 gives the Latin names for all of the species. Garcia and Alba

34
(1989) and Little and Wadsworth (1964) were used as a source for family

designations and for some of the original sources for the Latin names.

Thble 3 Local names for species used in the species-site analyses of the ISA-Mao dry forest with
their corresponding Latin names.1

Local name Latin name Source Family
Almacigo Bursera simaruba (L) Sarg. Burseraceae
Aroma Acacia famesiana (L.) Willd. Mimosaceae
Baitoa Phyllosglon bmsilienas Cap. Ulmaceae
Bruc6n Cassia emarginata L. Caesalpinaceae
Cafetan Lasianthus lanceolarus (Griseb) G. Maza Rubiaceae
Cambrén Prosopis juliﬂora (Sw.) DC. Mirnosaceae
Candelén Acacia scleraxyla Tuss. Mimosaceae
Cinazo Pithccellobium circinale (L.) Benth. Mimosaceae
Frijol Cappan's cynophallophom L. Capraridaceae
Guatapanal Caesalpinia corian'a (Jacq.) Willd. Caesalpinaceae
Guayacan Guaiacum ofﬁcinale I... Zygophyllaceae
Mostazo Cappan's ﬂwwsa (L) L. Capraridaceae
Palo amargo T richilia pallida Sw. Meliaceae
Quina Exostema caribaeum (Jacq.) R. & S. Rubiaceae
Sangretoro Maytenus brmfolia (A. Rich.) Griseb Celastraceae
Uvero Coccoloba leoganensis Jacq. Polygonaceae

 

‘Source: Little and Wadsworth (1964), van Paasen (1986), Knudson et al. (1988), Garcia and Alba
(1989)

Multivariate Analyses

Cluster Techniques

Cluster analyses operate on distance matrices derived from initial data sets. In
this study, species composition by site was tabulated using DKHBA. Data from
before the initial harvest (1986) was used to create a species-site data matrix.
Before submitting the data to cluster analyses, species densities were converted to

relative basal area contributions by calculating basal area of species J: as a

35

proportion of total basal area for each of the respective sites. No other
standardization of the data was considered necessary. Fisher (1994) used a similar
conversion for cluster analyses with tree counts. Because interpretations of the
MVA procedures are complex, the initial analyses were done with a subset of
forty-ﬁve sites. Forty of these represented the original control sites, plus ﬁve
additional sites designated as controls based on actual basal area removed.
Patterns of species composition observed in these initial analyses were
subsequently used to help interpret the results from the full data set analyses.

The initial cluster analysis used a data matrix representing 16 species and 45 sites
(16 X 45). Subsequent analyses used a data matrix representing 16 species and all
of the 120 silvicultural sites (16 X 120). Data from the pre-harvest inventory were

used in all of the MVA procedures.

Two hierarchical methods were selected to examine the data based on a distance
matrix derived using the Euclidean squared distance metric (the SAS default
option). The two hierarchical methods used were ﬂexible beta and Ward’s
minimum variance, both of which are options within the SAS Proc Cluster
procedure (SAS Institute Inc. 1985). The default beta value of -0.25 was used for
ﬂexible beta. Because Ward’s minimum variance method is sensitive to outliers
(SAS Institute Inc. 1985), the "trim” option was used, with 1% of the values
removed prior to analysis. Results of the hierarchical clustering were examined
visually. A series of preliminary analyses led to the selection of a level of

classiﬁcation which optimized for the maximum number of groups with the

36

greatest stability of group membership, as well as the most consistent species
composition within each group. Using these criteria, the same number of cluster
groups were formed from both hierarchical methods. In the initial analyses, six
groups were formed from the subset of 45 sites. In the full data set analyses,

seven groups were formed from the set of 120 sites.

Based on the groupings apparent from the hierarchical analyses, a range of groups
was selected to use in a series of nonhierarchical cluster techniques, also using a
distance matrix based on the Euclidean squared distance metric. SAS Proc
Fastclus was used to split the sites into k groups, based on a maximum of ten
iterations. In all of these analyses, group membership became stable in fewer
than ten iterations. In the initial analyses, 45 sites were clustered three times
using SAS Fastclus, using k = four, ﬁve and six. Subsequently, all 120 sites were
clustered twice, once with k = six and again with k= seven. A greater range
of k values were used in some additional analyses for both data sets. Higher
values resulted in additional groups of one or two sites. Lower values resulted in
the combination of sites with widely disparate species compositions. The ranges
presented here resulted in groupings of the data which were relatively stable
across clustering techniques based on characteristic species compositions and site
membership. "Characteristic species" for each of the major cluster groups were
selected using two criteria. First, a species was considered characteristic of a
cluster group if the average basal area contribution was at least two times the

average contribution across all sites within the data set. Second, a species was

37
considered characteristic of a cluster group if the average basal area contribution
was greater than the contribution across all sites and the standard deviation was
less than the standard deviation across all sites. In their study of a Tennessee
watershed, Grigal and Goldstein (1971) identiﬁed species as characteristic of a
particular cluster group based on an average basal area contribution which was

large relative to the contribution across the entire watershed.

Groupings within each of the cluster techniques were assigned numbers which
corresponded to one or two of the characteristic species determined to best
represent a particular group of sites. For example, in preliminary analyses, one of
the groupings was observed to represent a large contribution by the species
Bursera simaruba. This grouping was given the designation Group One.
Thereafter, any grouping of sites which showed a strong dominance by B.
simaruba was designated as Group One. A similar procedure occurred for each of
the other group designations. In this way, group membership could be compared

across cluster techniques.

As found by Grigal and Goldstein (1971), within each cluster analysis, the
characteristic species composition of the major groups corresponded to the
characteristic compositions of one of the major groups in each of the other
techniques. Likewise, as observed by Grigal and Goldstein, some of the sites
remained consistent across all analyses, while other sites changed group

membership depending on the technique used and/or the number of k used to

38

separate the data. Sites which clustered consistently were considered "core" sites
representing groups with distinct species compositions. In the ranges used for
SAS Proc Fastclus, certain groups of sites consistently separated out in the same
order moving from a lower k to a higher k. For example, if a given site was
member of Group Six in the hierarchical techniques, k-clustering assigned it to
some other group when k was equal to four in the nonhierarchical analyses and a
member of Group Six only when k was equal to six. Such sites were not
considered to have changed group membership. Thirty of 45 sites were

designated core sites in the initial analyses and 67 of 120 in the full data set.

Sites which did not fall into the same groupings across all techniques are referred
to here as noncore sites. These were assigned letter designations based on the
cluster group with which they were most closely associated. For example, in the
initial analyses, sites which were designated as Group One in at least three of the
ﬁve analyses were given the letter designation "A". Sites which were designated as
Group Two sites in at least three of the analyses were given the designation "B",
and so on. In the full data set analysis, a similar procedure was followed. If a
site was split between two cluster groups, it was given both designations. For
examwe, if a site was assigned to Group Two by both the Ward’s minimum
variance and the ﬂexible beta methods, but was assigned to Group Seven in both
versions of the nonhierarchical techniques, the site was designated as "BG", with
”B" representing the group two assignations and "G" representing the group seven

assignations. This secondary classiﬁcation does not necessarily represent expected

39

associations with other site characteristics, such as overstory structure or growth
and mortality. However, if the core sites represent discrete points along the
species-site continuum, it was expected that characteristics of the noncore sites
might provide insight into the relationships of the sites representing intermediate

areas.

Correspondence Analysis

To examine the relationships between sites from a different perspective,
correspondence analysis was applied to the original species-site data set without
converting basal area summations to relative basal area contributions. CA
assumes no grouping within the data set. Therefore, cluster groups would not be
expected to remain cohesive. CA was used primarily to check for consistency in
the association of sites associated in the cluster analyses. The data set examined
with correspondence analyses can be considered a two-way contingency table of
counts, with basal area (DKHBA) as a weight. Since basal area is continuous
rather than discrete data, the correspondence analyses did not meet the
assumptions of chi-square distributions. Therefore, distances observed in graphing
the principal axes cannot be strictly interpreted. However, apparent distances
were used to approximate the relationships of sites within cluster groups relative
to the relationships between groups to determine if sites grouped using cluster
techniques showed any consistent relationships. The directions of sites with
relative to the origin was also interpreted with respect to possible environmental

and disturbance factors. Since CA allows for an interpretation of species

40

relationships as well as site relationships, CA was used to examine which species
were associated with each cluster group. TWO correspondence analyses were
applied, using ﬁrst only core sites, then using both core and noncore sites. In the
initial analyses, a data matrix representing 16 spp and 30 core sites was analyzed,
followed by a matrix representing 16 spp and 45 sites. Full data set analyses were
ﬁrst applied to a matrix representing 16 spp and 67 core sites, followed by a

matrix representing 16 spp and core sites.

Canonical Discriminant Analysis

Operating'on the same data sets with the same variables as cluster analyses,
canonical variate analysis (CVA) can be used as an ordination procedure to
examine the relationships between the cluster groups (Grigal and Goldstein 1971).
Grigal and Goldstein also used CVA to determine if groupings of sites formed
using cluster techniques actually occupied discreet areas in canonical space. In
SAS, the procedure which separates groups based on the ratio of between group
variation to within group variation is referred to as canonical discriminant analysis
(CDA) (SAS Institute, Inc. 1985). Proc Candisc (SAS Institute, Inc. 1985) was
used to examine the quality of separation between cluster groups and to examine
ordinal relationships which could potentially relate to underlying gradients (Grigal
and Goldstein 1971). As in other ordinations procedures, canonical discriminant
analysis ﬁrst centers the data using means of each variable within each cluster
group. A function is then created which maximizes the ratio of between group

variance to within group variance, optimizing the separation of the groups along a

41

series of axes. The number of axes created is less than or equal to the min (p, g-
1), where p represents the number of variables and g is the number of groups
(Digby and Kempton 1987). Because the data is centered, the origin, or centroid,
represents the grand mean across variables for all groups and the position of the

groups with respect to the origin is indicative of their variation from this mean.

Initially, CDA was used to examine relationships between thirty core sites
representing ﬁve cluster groups. Subsequently, noncore sites were returned to the
data set and forty-three sites representing ﬁve groups (16 spp X 43 sites) were
analyzed. Noncore sites were assigned the group designation with which they
were most closely associated. Two sites designated as a minor cluster group were
eliminated. In the full data set analyses, the ﬁrst CDA was applied to a data
matrix with 16 spp and 67 core sites representing six core site cluster groups. A
second CDA procedure eliminated ﬁve of the initial core sites. The ﬁve sites
represented a single cluster group of limited sample size relative to the other
cluster groups that appeared to distort the results of the CDA procedure. The
third CDA was applied to 16 spp and 118 sites representing six core cluster
groups and seven subgroups (two sites were eliminated which clustered randomly
across all cluster analyses). This analysis was done to examine the relationship of
noncore sites- sites theoretically representing the continuous portion of the species

distribution continuum.

42
Overstory Analyses
Following the results of the classiﬁcation and ordination procedures, structural
characteristics were examined for each cluster group. Diameter class distributions
were analyzed for trees with single stems, based on the following class

distributions:

Diameter Class Diameter distribution (DKH)
3.0 cm

3.0 cm and < 4.0 cm

4.0 cm and < 5.0 cm

5.0 cm and < 6.0 cm

6.0 cm and < 7.0 cm

7.0 cm and < 8.0 cm

8.0 cm and < 10.0 cm
10.0 cm and < 13.0 cm

2 13.0 cm

\OOOQOKUI-ths-t
IVIVIVIVIVIVIVA

A tenth class, labeled MS, was also used, which included all trees with more than
one stem at knee height. Trees in this category consisted of 17.8% of all the trees

in the 1986 inventory before thinning.

Average tree height and average diameter (DKH) for each site were calculated, as
were basal area at knee height (m2 ha"), total trees (ha") and total stems (ha'l).
Total trees and total stems were used to estimate an average measure for the
frequency of multiple-stems at each site. Each of these variables was subjected to
the Kruskal-Wallis distribution-free test of differences between rank means for
cluster groups. Differences between groups were examined graphically using box

plots. All of these structural characteristics were examined for signiﬁcance with

43
respect to disturbance history and potential site quality. The assumptions were
that better sites support more basal area and taller trees, while poorer sites
generally have lower canopy heights and shorter vegetation (Beard 1944, Asprey
and Robbins 1953, Loveless and Asprey 1956). Disturbance generally creates
similar conditions, reducing overall tree height, causing a greater number of
multiple-stemmed trees and increasing total dominance by smaller diameter trees
(T amayo 1963, Holdridge 1967, Powell and Mercedes 1986, Kellman and Roulet
1990, Poynton 1990, Vera and Messerly 1990). Values for each parameter for

each site are listed in Appendix B.

Site Characteristics

Originally, a goal of this study was to collect data on the environmental
characteristics of the study sites to assess the association of species composition
and site characteristics. Due to limitations of time and resources, this segment of
the study was reduced to a brief review of the forty designated control sites.
During a ﬁve day period in March 1994, the principal researcher and an assistant
carried out a rapid inventory of a limited set of variables on each site. The
parameters examined included slope angle, slope position, degree of canopy cover,
surface soil texture, and identiﬁcation of ground vegetation. Basal area estimates
were also made using factor ﬁve of a CRUZALL tool, recording each stem by
local species name. Identiﬁcation of all species was done by the assistant. The
site data collected is found in Appendix C. Ground vegetation and CRUZALL

data is found in Appendix D.

44
Slope angle was estimated visually, with the terms "none", "slight", "moderate",
"steep" and "very steep" denoting increasingly acute slopes. Slope positions were
designated as "plain", "toe", "midslope", "shoulder", "low ridge" and "high ridge".
Canopy cover was estimated as >10% (category 2), but less than 25%, >25% but
less than 50% (category 3), >50% but less than 75% (category 4) and >75%
(category 5). Soil texture was a rough ﬁeld evaluation of the soil horizon
immediately below the organic layer. Designations ranged from silty clay to clay
sand. The stoniness of these samples was noted, using the labels "none", "some"
and "very". Ground vegetation was identiﬁed by common name. Species which
were unfamiliar to the assistant were designated as "herb". In analysis of ground
vegetation information, many of the species occurring only once or twice were
redesignated as "herbs". Table 4 is a list of the ground vegetation species
identiﬁed by the assistant. Data were collected to evaluate the relative dominance

of cacti species, which were not included in the original silvicultural study. Table

5 lists the dominant overstory species identiﬁed.

Data collected from this trip were analyzed using two-way contingency tables
based on frequencies by cluster group. For all of the parameters examined, the
results presented are useful only inasmuch as they serve to help interpret results
from the other analyses. The sample size and expected frequencies were too

small in these analyses to accept the results on their own merit.

45

Table 4 Ground vegetation identified in the forty designated control sites in the ISA-Mao
silvicultural study. "Data designation" refers to the classification used in data analysis.1

Local name Latin name

Cabuya Frucraea hexalpetala
Camphor

Cayuco Lemaireocereus hystrix
Desconocido

Espartillo Andropogon gracilis
Guinea

Guasabara Cylindropuntia caribaea
Hierba

Lamba vaca

Maguey Agave, sp.

Maya Bromelia pinguin

Palo prieto

Pilotera

Tremolina Croton, spp.

Tuna

Verbena Stachytarpheta cayennensis

 

1Source: Burgos et al. (1986).

Family
Amaryllidaceae
Cactaceae

Gramineae
Gramineae
Cactaceae

Gramineae
Amaryllidaceae
Bromeliaceae

Euphorbiaceae
Cactaceae
Verbenaceae

Data designation

Frucraea
Herb
Lhystrix
Herb
A.gracilis
Grass 2
C.caribaea
Herb
Grass 1
Agave
B.pinguin
Herb
Herb
Herb
Tuna
Herb

Table 5 Prominent canopy species identified using a factor five CRUZALL tool on forty sites
representing the original control sites in the ISA-Mao silvicultural study.2

Local Name Latin Name Source

Almacigo Bursera simaruba (L.) Sarg.
Alpargata Consolea monilifonnis

Baitoa Phyllostylon brasiliensis Cap.

Brucén Cassia emarginala L.

Cafetan Lasianthus lanceolatus (Griseb.) G. Maza
Cambrén Prosopis juliﬂora (Sw.) D.C.
Candelén Acacia scleroxyla Tuss.

Cayuco Lemaireocereus hysmbr (Haw) B & R
Cinazo Pithecellobium circinale (L.) Benth

Frijol Capparis cynophallophora L.

Guatapanal Caesalpinia coriaria (Jacq.) Willd.
Guayacén Guaiacwn oﬂ‘icinale L.

Palo de burro Leuceana trichodes

Quina Exostema caribaeum (Jacq.) R & S
Sangretoro Maytenus buxifolia (A. Rich.) Griseb
Uvero Coccoloba leoganensis J acq.

 

Family

Burseraceae
Cactaceae
Ulmaceae
Leguminosae
Rubiaceae
Mimosaceae
Mimosaceae
Cactaceae
Mimosaceae
Capraridaceae
Caesalpinaceae
Zygophyllaceae
Mimosaceae
Rubiaceae
Celastraceae
Polygonaceae

2Source: Britton and Rose (1963), van Paasen (1986), Knudson et al. (1988), Garcia and Alba

(1989).

46

Site Productivity

To evaluate sites on their relative growth potential, some measure of site
productivity had to be selected. In the temperate zone, tree height and age are
used to calculate site index. This is a common measure of site potential that was
developed for single species in even-aged stands (Carmean 1975, Pritchett and
Fisher 1987). Because site index curves have not developed for the Mao forest,
tree heights were used only as a relative indication of site potential. Observations
by Loveless and Asprey (1956) suggest that overall canopy height under dry forest
conditions is inﬂuenced by local environmental conditions. The principal
indicators of relative site productivity used in this study were growth and
mortality. These measures were estimated using diameter increments (DKH),
basal area summations (DKHBA) and stem counts at knee height. One measure
of growth for trees present in both the post-harvest inventory in 1986 and in the
annual inventory of 1992 was calculated as the difference between DKHBA
summations in 1992 and after thinning in 1986. Positive differences were summed
for each site and the variable was named BARGRTH. A second measure, called
BARAVE was calculated as the total DKHBA growth divided by the number of
trees alive per site in 1992. This parameter is an estimate of the average basal
area increment by site, per tree. A third variable, called DKHDIF, was created by
calculating the difference between 1992 and 1986 diameters of single-stemmed
trees. Positive differences were then averaged for each site. One measure of
mortality, STEMMORT, was calculated as the difference between the number of

stems recorded for each tree post-thin 1986 and the number recorded in 1992. A

47
second measure of mortality, BARMORT, was calculated as the difference

between post-thin 1986 and 1992 basal area summations for each tree, assuming
one or more stems (DKH) were missing. For both variables, positive differences
were then summed for each site. Mortality was assumed for missing stems,

although a number may have been cut green for fence posts and poles.

A sixth variable was created based on the difference between basal area growth
and basal area mortality. The variable, referred to as NETGRTT-I, estimates the
net change in biomass on each site over six years. On many sites, there was a net
loss of biomass as estimated by the change in basal area. Although a number of
factors may have affected all of the growth and mortality estimates, they are
considered to be sufﬁciently precise for determining the relative trends in site
productivity. Differences in growth and morality were tested between cluster
groups across all cutting levels. Differences between groups across cutting levels
were tested for a signiﬁcant difference using the Kruskal Wallis distribution free
one-way test of differences, assuming independence of sites. The values for each

parameter in each site are listed in Appendix E.

Nonparametric Analysis of Thinning Effects

The initial premise that thinning effects were not signiﬁcant was retested after the
sites were assigned to cutting levels based on the actual basal area removed.
Nonparametric analyses were deemed most appropriate because assumptions

related to the original experimental design were not considered to be valid. The

tea

(Ho

Cut

(F

The

lot:

For

lest

48

Kruskall-Wallis test of differences between treatments was used because block
means did not accurately represent the range of responses within treatment levels
(Hollander and Wolfe 1973). Sites were considered as independent samples.
Cutting level 5 (> 72% basal area removal) was discarded since the sample size
(n=5) was small relative to the other cutting levels. The treatments tested were:

Controls < 1.0% basal area removed

Cutlevel 1 z 1.0% and s 15.0% basal area removed

Cutlevel 2 > 15.0% and s 36.0% basal area removed

Cutlevel 3 > 36.0% and s 55.0% basal area removed

Cutlevel 4 > 55.0% and s 72.0% basal area removed

Cutlevel 5 > 72.0% basal area removed
The six measures of site productivity used to test differences between cluster

groups were also used to test differences between cutting levels. The hypothesis

to test for overall differences was:

Ho: TC=T1=T2=T3=T4.

For variables found to differ signiﬁcantly between treatments, a distribution-free

test of multiple comparisons was also applied (Hollander and Wolfe 1973). See

Appendix F for details on the test assumptions and test statistic calculations.

Results and Discussion

The ultimate goal of this study is to examine and explain species-site interactions
in a dry forest ecosystem using-multivariate methods of classiﬁcation and
ordination. Understanding species-site relationships is necessary in order to
design effective management techniques. Site evaluation and classiﬁcation have
occurred in tropical and temperate regions, but in the temperate zone
classiﬁcation has been more focused on developing efﬁcient and sustainable
systems for forest management. In the process of establishing these management
systems, a comprehensive understanding of species-site interactions has developed,
relating factors of climate, topography and soils to the total complement of

overstory and understory vegetation.

Multivariate techniques have been used in many temperate zone studies to
explore species-site relationships. Similar methods should be effective with
species composition data from the ISA-Mao subtropical dry forest. However,
there are three important factors which make this study signiﬁcantly different
from the temperate studies. First, studies examining species-site relationships
using multivariate techniques have not previously dealt with ecosystems recently
disturbed by human intervention. In the ISA-Mao forest, relatively recent
disturbance has affected species distributions in unknown ways. Second, in most
such studies, site conditions were known or data were later collected to conﬁrm

and/or modify models and classiﬁcations developed using species distributions.

49

Rou
10h;
Des

hm

dew

SM

50

Rowe (1984) recommends using ﬁeld based criteria to delimit groups of sites with
similar production potential. Numerical analyses can then be used to modify
these classiﬁcations. In this study, information about site characteristics is limited.
More importantly, no deﬁnitive information is available which predicts site
productivity based on speciﬁc site characteristics in the subtropical dry forest.
Finally, most of the multivariate analyses have dealt with temperate species whose
ecologies are relatively well known. The ecologies of dry forest species have
received very little study (Fries 1992). Literature pertinent to the species
prominent in the ISA-Mao forest was found to be limited in scope and in detail,
although the studies do suggest a number of important relationships (Beard 1944,
Record 1944, Holdridge 1945, 1967, Asprey and Robbins 1953, Loveless and
Asprey 1956, Tamayo 1963, Peacock and McMillan 1968, Jacobs 1965, Lugo et al.
1978, Ruskin 1980, Scott and Martin 1984, Murphy and Lugo 1986b, Otis and
Buskirk 1986, Rogers 1987, Stevens 1987, Garcia and Alba 1989, Kellman and
Roulet 1990, Poynton 1990, Vora and Messerly 1990, van Auken and Bush 1991,
Johnson 1992, Lees et a]. 1992, Hunter and Steward 1993, Buskirk and Otis 1994).
Despite these differences, temperate zone studies will be used as a basis for
interpreting the results of classiﬁcation and ordination procedures. Literature
from other studies in the semiarid tropics will be used to modify and clarify ideas
developed using temperate zone concepts. Additional information from the

silvicultural study will also be used to add to the interpretation of species data.

3P?
of f
5::

site.

tit
Sea
(It:

MI

51

Thischapter consists of ﬁve sections, each with a discussion of the analyses
applied to the data. Section one outlines the results of initial analyses on a subset
of forty-ﬁve sites, the forty original control sites and ﬁve undisturbed sites.
Section one is also an introduction to the multivariate methods used to classify
sites and examine relationships between groups. Section two presents and
discusses the MVA analyses for the full data set of 120 sites. Section three
examines the overstory structural characteristics of the sites with respect to the
cluster groups formed from the full data set analyses. Section three also examines
the data collected on site characteristics with respect to the cluster groups.
Section four presents and discusses differences in growth and mortality between
cluster groups. The ﬁnal section examines the effects of thinning on growth and

mortality.

hill:

Gus

and
lend
it a

tech

prev
One
Em:
mi
and
ma
off;
Gro;
5:0:
char.

52
Initial Analyses

Cluster Techniques

Based on the results of the two hierarchical cluster techniques, ﬁve major groups
and one minor group were identiﬁed. See Appendix F for SAS output, including
dendograms. Using a nonhierarchical technique, a range of groups from four to
six were examined. With a k of six, all six groups identiﬁed in the hierarchical
techniques were present, based on the species composition of each cluster. Thirty
of the initial forty-ﬁve sites clustered consistently into the ﬁve major groups
(Table 6). These thirty sites are referred to as "core" sites and are considered as
representing discrete points along a species-site continuum (Grigal and Goldstein
1971, Goldstein and Grigal 1972). The minor group consisted of only two sites
and was not treated as a core site group. Based on the criteria outlined
previously, three species were determined to be characteristic of the four Group
One sites, including Bursera simaruba with 44.0% of the basal area, on average.
Exostema can'baeum (8.0%) and Guaiacum ofﬁcinale (1.9%) were of secondary
importance. In Group Two, also with four sites, Phyllostylon brasiliensis (23.1%)
and Prosopis juliﬂora (20.9%) represented the largest proportions of the basal
area. Pithecellobium circinale (9.8%), Maytenus buxzfolia (6.0%), Guaiacum
ojfﬁcinale (2.4%) and Coccoloba leoganensis (2.0%) were secondary species.
Group Three, with eight sites, was dominated by Acacia scleroxyla (44.5%) and
Exostema caribaeum (20.6%). Tn'chilia pallida (6.9%) was a secondary

characteristic species.

53

Table 6 Results of two hierarchical cluster techniques and three versions of SAS Fastclus on
species data from a subset of 45 sites from subtropical dry forest in the Dominican Republic. The
first digit in the ID number indicates the block number (1 through 4). The second and the third
digit represent the site number (1 through 30 for each block). Group indicates the cluster
designation assigned based on the five tests. Numbers indicate sites consistent across all cluster
techniques. Letters indicate the core group with which the site was most closely associated (i.e.
"A" indicates a site which clustered most consistently in cluster Group One, etc.)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

ID 7T]:— Flexible Ward’s Fastclus Fastclus = Fastclus
Beta k=4 k= =6
101 E II 5 5 3 5 2 I
102 2 ll 2 2 2 2 2
103 E ii 2 5 5 5 2
104 4 ll 4 4 4 4 4
105 4 ll 4 4 4 4 4 ll
110 5 ll 5 5 5 5 5 II
126 1 ll 1 1 2 1 1 II
127 6 ll 6 0 2 2 6
128 5 5 5 5 5 5
129 2 ll 2 2 2 2 2
130 E Jl 2 2 5 5 5
206 5 5 5 5 5 5
207 3 3 3 3 3 3 II
208 3 3 3 3 3 3
209 3 3 3 3 3 3
210 3 7[ 3 3 3 3 3
212 3 3 3 3 3 3
213 1 1 1 2 1 1
214 E 5 5 3 5 2 n
215 c ll 5 5 3 3 3 ll
221 5 5 5 s s 5 II
222 c 3 3 2 2 3
223 3 II 3 3 3 3 3
224 3 3 3 3 3 3
225 3 3 3 3 3 3
301 B 2 2 2 1 1 l
w r

 

 

 

 

 

 

 

 

54

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 6, continued
“1 Group
302 1
303 D
304 4
305 D
316 B
317 B
318 B
319 2
320 s
F 401 4
' 402 4
‘ 403 4
L 404 4
405 D
426 6
427 5
428 2
429 r
430 5

 

 

 

 

 

 

55

Group Four, with seven sites, was dominated by Phyllostylon brasiliemir, with 73%
of the total basal area. Capparis' cynophallophora (3.3%) was a secondary species.
Group Five consisted of seven sites, in which Caeralpinia corian'a dominated with
43.6% of the basal area on average. Cappan's ﬂexuosa (7.4%), Cassia emarginata
(6.9%) and Acacia famesiana (3.8%) were also characteristic. Table 7 summarizes

the species compositions of each group.

Correspondence Analyses

Correspondence analysis (CA) was the next step in the analysis of the initial forty-
ﬁve sites. Correspondence analysis is used here to examine the relationships
between sites as expressed by the cluster analyses and to explore species
relationships more explicitly. Since CA assumes that all sites and species,
respectively, are independent, no a priori reason exists for patterns observed in the
previous analyses should be repeated. Because the data used in the
correspondence analyses do not meet the strict deﬁnition of chi-square
distributions (i.e. continuous versus interval data), the emphasis in interpretation
will be on direction rather than distance. However, the relative difference of
within group spacing versus the spacing between groups will be used as an
estimate of the quality of groups formed using cluster analyses. For groupings of
sites which are consistent with the results of cluster techniques, the relationships
between species and between sites will be examined for implications in terms of

underlying environmental factors.

56

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

. .. -nwvs4 me 3. 14le 3 1858 386V,
1 a... S. M mm a.— 3.1ng
3 I no 5. 543:8 :53
M as 3 3 8 an 6.383 seas:

3 e... H «.8 v.9 6.528 3.3608
, 3 a... o... S. guise,
a 3 he 3 2 u.— 33%3
a H 3 a a 2. 43% .5856
3 as _ 3 3 4336 ﬁghter
, no 3 E a.» 8%.. sad
3 an a... 1. 68.5: 3355
_ an 3 __ 3 3 iguana
,_ 3 an" «3 2 353a 833%
. 3 a... an a.» 330.3 308!
_ «a as a.» 3 5&1 168:,
an 3 z. 3 ﬂee-94363686
3.. 3 as
an as: M dd as:
$.15 :15
~ 965 3 n 9.80 8.8%

 

 

£545. .533 38:8
no «in 3:280 E 05 no .325 a «98% £5: 25 05 mo SEQ .3323 canteen—mote”. .«o coma—85:. Eu 2: «Boaou 360% we 3ch
. 60:35.03 cog—o 80.8. asp—3805 2.03280 wig—w 3m» 28 an no“ comm—520v nan—OB be 05 .0 6033.380 someone omﬂo>< b 033—.

57

In Figure 3, the site scores are plotted for the ﬁrst two dimensions of a CA
applied to the thirty core sites representing ﬁve cluster groups. Site and species
scores are listed in Appendix G. Figure 3 shows that the spread between sites
within groups One, Three and Four is relatively small, while spaces between
groups are relatively "clean." Pielou (1969), Grigal and Goldstein (1971) and
Digby and Kempton (1987) suggest that classiﬁcation systems can recognize
discrete points along a continuous species-site gradient, or gradients.
Correspondence analysis is inherently an ordinal technique, rather than a method
of classiﬁcation. The results of CA using the thirty core sites representing ﬁve
groups supports the concept that three of the ﬁve main cluster groups represent
discrete points along a species-site continuum dominating the ISA-Mao forest.
While Group Five sites suggest a less cohesive relationship within the group, there
is no overlap with the spaces occupied by other sites. This suggests that Group
Five sites also represent a discrete position. Group TWo sites, on the other hand,
cluster relatively closely, but the space they occupy overlaps with group four sites.
This suggests that sites classiﬁed as Group Two represent more complex

relationships than the other groups- relationships which may be continuous, rather

than discrete.

An important advantage of correspondence analysis is its ability to capture a
signiﬁcant proportion of the variance in the data set in a few dimensions. Table 8
lists the singular values for each of the ﬁfteen dimensions necessary to represent

all of the data variance. In the ﬁrst two dimensions of the CA procedure, a total

58

 

 

1.5 l I 5 l l
5., 5o
1.0 — ~
23 O5
05
0.5 — 2. T
8}: 2. 2
cﬁ 43“ 4 §§
‘-* 00 49 °
(\i 444.0 3 043
< o 03 O3
O
—0.5 7 r
10 01
—1.0 ‘7 01. _
10
—1.5 1 f

 

 

 

 

-1.5

—1.0

CA1: 25.9%

—0.5 0.0

0.5

1.0 1.5

Figure 3 Positions of thirty subplots from the ISA-Mao silvicultural study along the ﬁrst two
principal axes of a correspondence analysis. 45.1% of the total variance is explained by the first
two of fifteen axes. Analysis used basal area contributions of sixteen species. The size of the
plotting symbol is proportional to the sum of the squared cosines in two dimensions. The label
indicates the cluster group assigned to each site using three cluster techniques.

 

59

Table 8 Singular values, principal inertias and proportional representation for each of ﬁfteen
dimensions of a correspondence analysis of sixteen species and thirty sites representing core cluster

group sites.

Accumulative

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Dimension Singular Percent
Values inertias percentage
1 0.77763 0.60471 25.90 25.90 .
2 0.6693 0.44881 1922 45.12 N
3 0.62099 038562 16.52 61.64
4 0.49850 024850 10.64 7228
5 03692 0.13662 5.85 78.13
6 032383 0.10487 4.49 82.62 g
7 031414 0.09868 423 86.85 1
8 029566 0.08741 3.74 9059
9 024682 0.06092 2.61 932 i
10 0.21163 0.04479 1.92 95.12 :1
11 0.19412 0.03768 1.61 96.73 I
12 0.16594 0.02754 1.18 97.91
13 0.15726 002473 1.06 98.97
14 0.11476 0.01317 0.56 9953 l
15 0.10361 0.01074 0.46 99.99 5
233479 5

60

of 45.1% of the total variation is represented. This proportion is referred to as
the display quality of the graph. Fifty-ﬁve percent (54.9%) is not represented.
This is the display error. Relative to the challenge of displaying each site using all
sixteen species variables, Figure 3 represents an improvement for a reasonable
analysis of the data. However, although patterns observed are consistent with the
previous analyses, the unexplained proportion of the data could represent serious

glitches for interpretations.

The quality of representation usually differs between sites and species within each
dimension. The total variance explained in one dimension can be deﬁned as the
sum of the individual variances attributable to each of the points in that
dimension. Therefore, the variance explained for one point in that dimension can
be deﬁned as a proportion of the sum of all the variances attributable to that
point across all dimensions. The squared cosine associated with a site or a species
represents the variance explained in a particular dimension for a particular site or
species in that dimension. The squared cosines associated with a site (or a
species) sum to one across all dimensions. The squared cosine in a given
dimension therefore represents the proportion of variance explained for that site

or species (Greenacre 1993).

In Figure 3, the size of each plotting symbol is proportional to the summed values
of the squared cosines for the ﬁrst two dimensions. The median value for Group

Three sites is 60.9%. For Group One sites, the median value is 52.5%. For

61
Group Four sites, the median value is 48.5%. The median value is 44.7% for

Group Five sites and 15.9% for Group Two sites. These values indicate that the
display qualities for Group Three sites are relatively good, even though the overall
quality of the graph is relatively poor (display error equals 54.9%). On the other
hand, the display qualities for Group Two sites are very bad (the display error is
over 80% for two of the sites). Display qualities for the other sites are

intermediate.

If we add the third dimension to the graphical display of the thirty core sites, we
increase the display quality to 61.6%. Figure 4 represents the positions of the
thirty core sites in three-dimensions. The median of the summed cosine values
for Group One sites is now 86.9%. The median for Group Three sites is 71.1%,
for Group Four sites it is 69.6% and for Group Five sites, 61.1%. The median
value for Group Two sites is still a very low 17.3%. The relationships among
Group Two sites appear to be complex- many dimensions would be necessary to
describe the precise position of each of these sites. Nevertheless, in three
dimensions, Group Two sites still cluster together. The relationships within and
between groups One, Three, Four and Five are represented fairly well by three
dimensions. Scatter within Group Four sites is still relatively small. Scatter within
the other groups has increased, but groups One, Three, Four and Five still appear
to occupy discrete areas of space. Based on the results of correspondence analysis
using thirty core sites, the combined use of ﬁve cluster analyses appears to have

captured nonrandom patterns inherent in the data.

62

 

 

 

 

 

 

 

 

’1 3°
1
9 1% 1!
9 §
4 \
1o 1 ”“7
.4» T I
IQ 4
(0 vs
A. '3- I ‘ re
6'0 4 . 1
A 5 ' " Q
U 43
l I
~.'-
9 ’9 ‘
a?»
‘9
r
0
1&4 0 x
6:» ° 025
$6 ﬁg) 0° «4..
a: 6:9
o ,6 x-Q‘
’x Q (2“
""
,\?° ’13.,

Figure 4 Position of thirty core sites in three dimensions of a correspondence analysis using the
basal are contributions of sixteen species. The size of each plotting symbol is proportional to the
sum of the squared cosines in three dimensions. The label indicates the cluster designation given
based on ﬁve cluster analyses using the relative basal area contributions of sixteen species. The
length of each spike 'n proportional to the distance from zero along the third axis. Sixty-two
percent (61.6%) of the total variation is accounted for by the three dimensions.

63

Given that groupings of sites observed in CA correspond with groupings formed
from cluster analyses, species relationships observed in CA should correspond
with species selected as characteristic of each cluster group. Figure 5 represents
the position of each species plotted in two dimensions, in the same space as the
sites. These positions can serve as a reference for interpreting the site scores as
well as providing a description of interactions between species (Digby and
Kempton 1987, Greenacre 1993). These positions correspond with some of the
"characteristic species" listed for each cluster group. Bursera simaruba ("AL") is
located at the extreme bottom, below the origin along the second axis, in the same
position as Group One sites. Phyllostylon brasiliensis ("BA") and Coccoloba
leoganensis ("UV") are located to the extreme right of the origin along the ﬁrst
axis, "near" the same area occupied by Group Two sites. Prosopis juliﬂora ("CM")
and Capparis ﬂexuosa are located to extreme right of the ﬁrst axis and
approximately half way up the second axis, also in the same area as Group Two
sites. Acacia scleroxyla ("CA"), Exostema can'baeum ("QU") and Trichilia pallida
("PA") are grouped to the extreme left along the ﬁrst aids, in the same position as
Group Three sites. P. brasiliensis is located in the same position as Group Four
sites. Towards the top, above the origin along the second axis, Caesalpinia
coriaria ("GU"), A. famesiana ("AR") and Capparisﬂemasa ("MO") are in the
same space occupied by Group Five sites. All of these relationships correspond
with the species nominated as "characteristic" based on average basal area
contributions within core cluster group sites. Several species are missing.

Guaiacum ofﬁcinale ("GY") is missing from Group One, Pithecellobium circinale

64

 

 

 

 

1.5 l i l l 1
AR .
o GU
1.0 — MO *-
. SA
CM .
0.5 — e
R:
. CI
3 0 O FRCF 9 PA
CK}. ' BAO ER 0 CA
<3 UV . '
O o QU
-O.5 - -
—1.0 - e
—1.5 A1“) . r

 

 

—1.5 -1.0 -0.5 0.0 0.5 1.0 1.5
CA1: 25.9%

Figure 5 Position of sixteen species along the ﬁrst two axes of correspondence analysis. Data
comes from basal area contributions in 30 ’core’ plots in the ISA-Mao silvicultural study
representing distinct species compositions. Plotting symbol is proportional to the sum of the
squared cosines in two dimensions. Symbols are: UV= C. leoganensis, CM= P. juliﬂora, AR= A.
famsiana, BA= P. brasiliensis, FR= C. cynophallophora, MO: C. ﬂexuosa, BR= C. emarginata,
CI= P. circinale, GY= G. oﬂ‘icinale, CF= L. lanceolatus, AL= B. simaruba, GU= C. coriaria,
SA= M. buxrfolia, QU= E. can'baeum, PA= T. pallida, CA= A. scleroxyla.

65
(”CI") is missing from Group Two, Cappan's cynophallophora ("FR") from Group
Four and Cassia emarginata ("BR") from Group Five. However, all of the primary

species correspond.

The problem with this interpretation of the species’ orientations is the same as
with the sites. The display error is over 50.0%. Therefore, limited conﬁdence can
be placed in the overall positions of the species. However, also like the site
orientations, the display quality differs among the species. Based on the summed
values of the cosines associated with each species in two dimension, the display
quality for A. scleroxyla is 80%, 66% for C. coriaria, 64% for P. brasiliensis, 60%
for B. simaruba, 51% for E. caribaeum and 53.4% for T. pallida. All other sums
are less than 50%. Again, by adding a third dimension (Figure 6), we increase the
overall display quality to 61.6%. Values for the above species increase to 92%,
83%, 92%, 97%, 53% and 59%, respectively. Summed values for all other species
remain below 50%. Because the values for A. scleroxyla, C. corian'a, P. brasiliensis
and B. simaruba are so high, we can be fairly conﬁdent in the signiﬁcance of their
relative positions in Figure 6. Comparing relative positions is harder in three
dimensions, so species scores have been plotted together with site scores. B.
simaruba is clearly associated with Group One sites, A. scleroxyla with Group
Three sites, P. brasiliensis with Group Four sites and C. coriaria with Group Five
sites. More over, although their display qualities are relatively low, E. caribaeum

and T. pallida are still associated with Group Three sites and A. famesiana and C.

66

 

 

 

 

 

 

 

 

 

 

 

 

W 4.
at T
90%,"
T 9"
F *3
5 I
1 4.. I“ I” N1:
O\° i rFPT
{2 a
co 7 ,,
r‘ % <3
0. ”A I J '5
CO i ‘0" 9 c.)-
A '84 l O 4)
a. NK
0
I .
A
I .,
I.»
5 ~ I
\?->
X9
0?: x?)
1%» 9° ‘6
6} 02>
g
{<3 (233'
o ’9 ~2-
,x Q (3»
ll».
)3; Ix?”

Figure 6 Positions of thirty sites and sixteen species in three dimensions based on a
correspondence analysis. Analysis used basal area contributions of sixteen species. The size of the
plotting symbol is proportional to the sum of the squared cosines in three dimensions. Numbers
represent cluster group designation of each site. Letters indicate species. UV= C. leoganensis,
CM= P. juliflara, AR= A. famesiana, BA=- P. brusiliensis, FR= C. cynophallophora, MO= C.
ﬂexuosa, BR= C. emarginata, CI= P. circinale, GY=- G. oﬂicinale, CF = L. lanceolatus, AL= B.
simamba, GU= C. corian'a, SA= M. bunfolia, QU= E. can'baeum, PA= T. pallida, CA= A.
sclemxyla. The length of each spike is proportional to the distance from zero along the third axis.
Variance explained is 61.6% of the total.

ﬂaw;
man-g
same
111Gb

busi’:

The 1
its
the 11
111111
and t
10111.
Idem.
60w
5960':
relat'r
mgge
Ibran
Dot SC
5011p:

Mp:

Mp.

67

ﬂaxuosa are still associated with Group Five sites. None of the species are
strongly associated with Group Two sites, although P. juliﬂora is located in the
same area of space. This may reﬂect again the complexity of Group Two sites,
which appear to represent a diversity of species loosely associated with P.

brasiliensis.

The results of this correspondence analysis coincide both with the groupings of
sites using ﬁve cluster analyses and with the species nominated as characteristic of
the respective cluster groups. If the species are responding to underlying
environmental or site history conditions, the relative positions of both the groups
and the species could have ecological signiﬁcance. The ﬁrst axis suggests a
contrast between sites dominated by P. brasiliensis versus those dominated by A.
scleroxyla. The second axis suggests a contrast between sites dominated by C.
con'aria versus those dominated by B. simaruba. Given multiple factors affecting
species distributions, the ﬁrst dimension suggests one factor which affects the
relative dominance of P. brasiliensis and A. scleroxyla. The second dimension
suggests a second factor which affects the relative dominance of C. con’aria and B.
simaruba. These two factors appear to be independent, since the ﬁrst axis does
not separate groups One and Five, and the second. dimension does not separate
groups Three and Four. The third axis suggests a complication because both
groups Three and Four are below the origin while both groups One and Five are
above it. This third dimension (by itself) would suggest an association between

groups One and Five (as does the ﬁrst dimension) and between Three and Four

 

(as do
oi the
{1001

wing

Asec

dim
and 1
$1ng1
sites e
Other.
the ﬁn
6101;

1111 1
are inc
Niel
TM a]

0011 si:

68

(as does the second dimension). Since the third dimension combines the effects
of the ﬁrst two dimensions, it may represent an interaction between hypothetical
factors One and Two. Similar results have been found in other ecological studies

using correspondence analysis (Digby and Kempton 1987, Fisher 1994).

A second correspondence analysis examined the relationships between both core
and noncore sites. Figure 7 represents the positions of forty-ﬁve sites in two
dimensions. Scores for sites and species are listed in Appendix G. Groups One
and Three still occupy discrete areas of space. Group Five is intermixed with
subgroup E, but remains spatially separate from the other groups. Group Four
sites are intermixed with a number of other sites, but they remain "close" to each
other. The relative positions of groups Three and Four and One and Five along
the ﬁrst and second axes remain the same. Group Two sites remain mixed with

Group Four sites.

Table 9 lists the proportional representation of each of the sixteen dimensions for
the second CA. The variation represented by two dimensions decreased to 35.1%.
This suggests that species-site relationships are more complex when noncore sites
are included in the analysis. The positions of the sites now suggest a continuous
gradient, moving from Group One sites through sites representing Groups Four,

Two and Five, arriving at Group Three sites at the opposite end of an are. If

core sites represent discrete points along a species continuum and noncore sites

 

1.2

69

 

 

 

 

l L l
5.
5.
E5.
E. 5. 05
5 05
' C
2. go’ 0 (33
6.6. E 30
4. . 0
i3" 4° 2 C E 65333
T2- ' A _
—. D. oTS 03 U
11.9 B.
Do
B.
1O 10
lo
I

 

 

-0.6

I
0.0 0.6

CA1: 20.1%

1.2 1.8

Figure 7 Position of forty-ﬁve sites from the ISA-Mao silvicultural study in the first two
dimensions out of ﬁfteen of a correspondence analysis. The analysis used the basal area

contributions of sixteen dry forest species. The size of the plotting symbol is proportional to the

sum of the squared cosines in two dimensions. The label indicates cluster group designations

assigned using ﬁve different cluster analyses. Thirty-five percent (35.1%) of the total variance is

represented by these two dimensions.

70

Table 9 Singular values, principal inertias and proportional representation for each of ﬁfteen
dimensions of a correspondence analysis of sixteen species and forty-ﬁve sites in the ISA-Mao

silvicultural study.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Dimension Singular Principal Percent Accumulative
Values inertias percentage
1 0.74884 056076 20.09 20.09
2 0.64634 0.41776 14.97 35 .06
3 058767 034535 12.37 47.43
4 054936 0.30179 10.81 58.24
5 052025 0.27066 9.70 67.94
6 0.43015 0.18503 6.63 7457
7 0.38322 0.14685 5 .26 79.83
8 037732 0.14237 5.10 84.93
9 0.35765 0.12797 458 89.51
10 0.30456 0.09276 332 92.83
11 0.26560 0.07055 253 95.36
12 021461 0.04606 1.65 97.01
13 0.20860 0.04351 1 56 9857
14 0.14512 0.02106 0.75 99.32
15 0.13766 0.01895 0.68 100
2.79137

 

 

 

rtpr
betv
mi
rela'

sites

ﬁle

Ida!

Can.
Con
sites
anal}
relax:
that ;
dust:
and I
hem

HUMP

After

We a

71
represent the continuous points, more information (i.e. more axes) should be
between the areas where distributions are discrete. More information is probably
available from the additional axes, but two dimensions sufﬁce to illustrate that the
relationships between the core sites are not radically affected by including noncore
sites in the analysis. Figure 8 illustrates the relationships between species in two
dimensions using both core and noncore distributions. The relative positions of
the minor species are somewhat different, but the positions of P. brasiliensis, A.

scleroxyla, C. corian'a and B. simaruba remain the same.

Canonical Discriminant Analyses

Correspondence analysis conﬁrmed that there is a strong relationship between
sites within groups formed from the cluster techniques. Canonical discriminant
analysis (CDA) can therefore be used with greater conﬁdence to examine the
relationships between these cluster groups. The use of CDA implicitly assumes
that an individual site is a valid samples of a particular population, in this case, a
cluster group. In these analyses, CDA is used as an ordination technique (Digby
and Kempton 1987) rather than as a test of the signiﬁcance of differences
between groups. Probability tests would require multivariate normality within

groups and homogeneity of variances between groups, neither of which was tested.

After removing all noncore sites from the initial analysis, the thirty remaining sites

were analyzed with CDA. Plotting the ﬁrst two canonical axes, 89% of the

 

72

 

 

GUO
Fido

0.6 _ CMR . . SA —
x o PAO CA
:3 0-0 BA. CIGY
Lg UV. BR. . QU
.. CF.
35
U _0.6 ‘7 _

 

 

 

 

—1.8 I A Q ' I I
—L2 —0£3 (10 (16 L2 L8

CA1:ZOJ%

Figure 8 Position of sixteen species along the first two axes of correspondence analysis. 35.1% of
the total variance explained in the first two of fifteen axes. Analysis used basal area contributions
on 45 subplots from the ISA-Mao “silvicultural study. The size of the plotting symbol is
proportional to the sum of the squared cosines in two dimensions. Species code are: UV= C.
hoganensis, CM= P. juliﬂora, AR= A. famesiana, BA= P. brasiliemis', FR= C. cynophallophora,
MO= C. ﬂewosa, BR= C. emarginata, CI= P. circinale, GY= G. ojﬁcinale, CF = L. lanceolatus,
AL= B. simaruba, GU= C. coriaria, SA= M. buufolzh, QU= E. caribaeum, PA= Palo amargo,
CA= A. scleroxyla.

 

73

variance was explained (Figure 9). Scores are listed in Appendix H. Eigenvalues
associated with each dimension are listed in Table 10. Groups One, Three and
Four clearly separate from all other clusters. Groups Five and Two clearly
separate out from groups Three and Four, but remain very close to each other.
Grigal and Goldstein (1971) used canonical variate analysis to determine if site
groupings formed using cluster analyses represented discrete groups in canonical
space. The clear separation of groups One, Three, Four and Five reinforces the
separation of these groups observed in CA. On the other hand, it would be
reasonable to suggest that groups ﬁve and two were arbitrarily separated by the
cluster techniques and in fact, the two groups should be considered as a single
population. However, Grigal and Goldstein (1971) found that the groups which
overlapped had common species associations. In this case, there are no
"characteristic" species common between groups Two and Five. Therefore, there
is no immediate suggestion of why groups Five and Two overlap. On the other
hand, while the CDA procedure ensures that the ﬁrst two axes are the best
representation of the relationships apparent in the data, it does not guarantee that
the axes represent the most important relationships. The remaining eleven
percent of the variance unexplained in the ﬁrst two axes may represent the most
important differences between groups two and ﬁve, differences which are

independent of the relationships expressed by the other four groups.

 

74

30 l I

 

 

N
0:2
m
V—q
(\j 10 T 4.. F
a 4%":
U
a; 2
2..
O '5’- .1
3:33. 3’11 1

 

 

 

_10 ‘l l l
—10 O 10 20 30

CANI: 75.2%

Figure 9 Position along first two canonical axes of thirty subplots from the ISA-Mao silvicultural
thinning study. The thirty subplots represent five groups determined using three cluster analyses
on data representing relative basal area contributions of sixteen dry forest species. Labels indicate
cluster group designation. 89% of the total variance is explained by axes one and two.

 

75

Table 10 Eigenvalues associated with each axis of CDA applied to 16 species and 30 core sites
representing five groups from the ISA-Mao silvicultural study.

Percent Cumulative

75.19

Eigenvalue

Dimension
1 132.7587

24.3619 88.99

13.4948 96.63

5.9431 100

 

Representing 75 % of the variation, the ﬁrst dimension separates Groups One and
Four from Groups Two, Three and Five. The second dimension separates Group
Three from Groups Two and Five. The second dimension also places Group

Four in the northeast quadrant formed by the juxtaposition of the two axes,

Group Three in the southwest quadrant and Group One in the southeast. Grigal

and Goldstein (1971) found that each of their core site clusters were positioned in

a different quadrant in two dimensional canonical space. Site information
indicated that these positions represented ecological extremes created by moisture
and topographic relationships (Grigal and Goldstein 1971, Kercher and Goldstein
1978). In this case, Group One would represent the most extreme environmental
characteristics based on its position at the extreme right of the primary axis. As
in CA, groups Three and Four appear to represent a contrast with each other,
”Sitioned as they are in opposite quadrants. While Group One is at the furthest
enrelme along the ﬁrst axis, its position in a quadrant between groups Three and

Foul- represents a movement away from the origin that is independent of the

Ye\ationships express by groups Three and Four.

 

76
The species composition of each group illustrates these relationships. Group
Three is dominated by A. scleroxyla (44.5 %) with very little P. brasiliensis (2.9%).
Group Four is dominated by P. brasiliensis (73.2%) and has no A. scleroxyla.
Group One, however, while dominated by B. simaruba (44.0%), includes both P.
brasiliensrls (14.5%) and A. scleroxyla (7.3%) (Table 7). Although Group Five does
not clearly dominate the northwest quadrant, Group Five sites show the inverse
pattern to Group One site. Group Five sites are dominated by C. corian'a
(43.6%) with no B. simaruba while both P. brasiliensis (8.2%) and A. scleroxyla
(5.5%) are represented. These contrasts between One and Five and Three and
Four are identical to the relationships observed in the correspondence analyses.
Two independent site factors would appear to control separately the relationships
between species compositions in groups One and Five versus groups Three and

Four.

To examine the effects of the noncore sites on CDA, these sites were temporarily
designated the cluster group number with which they were most closely associated
(i.e- sites designated as "D" were given the designation "4", etc.). The two sites

designated as Group Six sites were removed. The remaining 43 sites were then

submitted to CDA. The scores for each site are listed in Appendix H. Plotting

the sites in two dimensions, 79.2% of the variance is illustrated in Figure 10.

Eigenvalues associated with each dimension are listed in Table 11. Noncore sites
are plotted with their original letter designations. With 20.8% of the variance

unaccounted for, Figure 10 still represents a fairly good visual estimation of the

 

77

 

 

 

 

 

 

15 l l 1 L l
12 7 r
9 _ _
N
CO
1\' 6 ‘- 4. x
N 4474.
C°\i 40. 4.
E; 3 _. 2,39%: 114 B _
C.) 56E, 33'
O 5B. E'E_ 2
5. 5. 1
5. 5 1 1r
. .1 .
_3 C§: C- —
3
33-33.
—6 l l l j l
-6 —3 O 3 6 9 12 15

CANl: 51.6%

Figure 10 Position along ﬁrst two canonical axes of forty-ﬁve sites from the ISA-Mao silvicultural
nullhing study. The forty-ﬁve sites represent ﬁve groups determined using three cluster techniques
data representing relative basal area contributions of sixteen dry forest species. Labels indicate
gusts-J group designation. Numbers indicate core sites which clustered consistently across all
iques. Letters indicate sites which did not cluster consistently. ’8’ represents sites clustering
often in group two, ’0 represents sites clustering in group three, 'D’ in group four and ’E’ in
8'0“]: ﬁve. 79.2% of the total variance is explained by axes one and two.

78

Table 11 Eigenvalues associated with each axis of CDA applied to 16 species and 45 sites
representing five cluster groups.

Dimension Eigenvalue Percent Cumulative

1 21.1195 51.61 51.61
11 .3066 27.63 79.24
6.1641 15.06 94.3

2.3315 5.70 100

 

relationships between the groups, although not as good as the previous CDA plot.
As in CA, including sites which may represent continuous species distributions
increases the complexity implicit in the data, making more information (more

axes) necessary to describe the precise position of each group.

In this analysis, Group One is still positioned to the extreme right of the ﬁrst axis.
Group Three still dominates along the lowest portion of the second axis.
However, the ordination of sites in groups TWO, Three, Four and Five suggests a
continuum, with Group Three sites towards the bottom, Group Four sites towards
the tap, and groups Five and Two in between. Unlike the previous CDA, the
separation between each of these groups is "fuzzy".' The two site factors suggested
by CA would also apply to this CDA procedure. However, one factor would
appeal- to determine the dominance by B. simaruba while a second factor would
result in a continuum of species distributions, moving from A. scleroxyla through
C‘ cOr'z'aria and a mix of other species, to P. brasiliensis at the opposite extreme.
l“ CA, the continuum could be described as moving from A. scleroxyla to C.

corian'a to P. brasiliensis to B. simaruba at the opposite extreme.

 

79

If a great deal were known about the ecology of the species used in these
analyses, inferences could be drawn about the relative positions of each cluster
group. Conversely, if more were known about moisture relationships across the
ISA-Mao landscape, inferences could be made with respect to the ecology of each
species. As mentioned previously, very little information is available in either of
these areas. What is known, is that soil depths and topography vary a great deal
across the landscape. Soil and topography are closely interrelated and together
affect moisture relationships which in turn affect species distributions and relative
site quality. What has also been observed is that the relative intensity of
disturbance also differs from area to area, and even from site to site (Powell and
Mercedes 1986). Disturbance is also known to affect species distributions, and if
sufﬁciently severe, may result in site degradation and reduce relative productivity.
Given that both relative moisture availability (as it relates to topography and soils)
and disturbance history are known factors within the ISA-Mao silvicultural study,

the following suggests a relationships between four of the cluster groups.

 

 

 

ll Undisturbed Disturbed II
II Moist Group 5 Group 4 II
II Dry Group 3 Group 1 II

 

 

 

The positions of the cluster groups in this model are consistent with their
Positions both in CA and CDA. However, a great deal of additional information

would be necessary to test the model’s viability.

80
Summary
Several observations are worth emphasizing at this point: (1) Patterns of species
compositions can be described for a subset of the silvicultural study sites. Based
on a series of cluster analyses, two-thirds of the sites can be assigned to of one of
ﬁve groups, each representing a unique species composition. (2) These groupings
are consistent with an independent analysis using CA. (3) Three of the ﬁve groups
occupy discrete areas of canonical space, while two of the groups overlap. (4) The
positions of four of these groups can be interpreted by comparing their relative
positions using CA and CDA, and by comparing the relationships between species
characteristic of each group. The results of the same procedures applied to 120

sites will be examined next.

81

Full Data Set Analyses

Cluster Techniques
With some basic patterns of species composition established using a subset of the
data set, the full data set was then analyzed, beginning with the same cluster
analyses described earlier. See Appendix F for output from SAS for the two
hierarchical classiﬁcations. The cluster group classiﬁcations established in the
initial analyses served as markers for the groupings of sites exhibited in the new
analyses. Based on the two hierarchical methods and two versions of the
nonhierarchical technique (k=7 and k=8), eight groups were identiﬁed. Of the
eight groups identiﬁed, six were represented in each of the four methods. Sixty-
seven sites clustered consistently into one or another of these six groups. These
sites are those referred to as "core" in subsequent analyses and are designated
with the corresponding cluster group number. Letters indicate "noncore" sites and
represent the core cluster group with which a site is most closely associated. Two
sites clustered with a different group in each analysis and are therefore designated
With an "X" (Table 12). Of the eight cluster groups identiﬁed in the initial
analyses, four corresponded to the major groups identiﬁed in the partial data set
anal)o'ses based on species composition. Groups One (B. simaruba), Three (A.
scleroxyla), Four (P. brasiliensis) and Five (C. coriaria) were clearly represented by
the new results. Group Six (A. famesiana) corresponds with a minor group of two
sites formed from the initial analyses. The species composition for each group

baSed on the core sites is given in Table 13.

82

Table 12 Results of two hierarchical cluster techniques and two versions of SAS Fastclus applied
to species data from 120 sites in a subtropical dry forest in the Dominican Republic. The ﬁrst
digit in the ID number indicates the block number (1 through 4). The second and third digit
represent the subplot number (1 through 30 for each block). Final Group indicates the cluster
designation assigned based on the four cluster techniques applied to all 120 sites. Numbers
indicate sites which clustered consistently across all techniques, referred to as "core" sites. Letters
indicate non core sites. Each letter designation is associated with a core cluster designation. I.e.
"A” is associated with sites which clustered in Group One more than once. "B" is associated with
sites that clustered in Group Two more than once, etc. A double letter designation indicates that
a site clustered in one group twice and in another group twice. Initial Group indicates the group
designation assigned based on initial analyses using a subset of 45 sites. NA indicates sites not in
the initial analyses.

ID Ward’s Flexible Fastclus Fastclus
Beta k=6 k=7

2 7 7

7 7 7

 

 

83

Table 12, continued

ID Ward’s Flexible Fastclus Fastclus
Beta k=6 k=7

124 6 5 6

125 5 5 6

 

 

84

Table 12, continued

Fastclus
k-7

 

Fastclus
k-6

 

Ward’s

 

Flexible
Beta

 

Initial
Ora-P

NA

 

Group

 

ID

 

 

 

 

227

 

 

229

 

 

301

 

 

 

 

 

 

 

 

 

310

 

311

 

312

 

313

 

314

 

315

 

316

 

317

 

318

 

319

 

320

 

321

 

322

85

Table 12, continued

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

WNW-774747484 44434775784174747417
F
ué
Wes-774747484 44434475774744747417
F
9'
W724747464 24234275252722747217
.P.
MM724747464 44434275224624747417
P
WWMMMMMMMM4 44nmmmmmmmmmmmmmmmmm
P
mm7w47474m4 D4D34B75WXDGBD747D17.
mmmmmmmmmm mmmmmmmmmmmmmmmmmmmm

86

Table 12, continued

 

 

 

 

 

 

 

 

 

ID Final Initial Flexible Ward’s Fastclus Fastclus
Group Group Beta k=6 k=7
423 6 M 6 6 7 6
424 A M 6 7 1 1
425 G NA 6 7 7 7
426 6 6 6 6 7 6
427 5 A 5 5 5 5 5
428 G 2 2 7 7 6
429 1 1 1 1 1 1
430 5 5 5 5 5 5

 

 

 

 

 

 

 

 

 

Group Two was not consistently represented in all four clustering procedures.
That is, the sites designated as Group Two sites in the initial analyses did not
group together consistently across methods in the full data set analyses. Group
Two was therefore not considered a viable cluster group. Although Group Two
did not remain coherent in the new analyses, another cluster group appeared
which had sites assigned together consistently across methods. This group was

designated as Group Seven.

Species "characteristic" of each group formed were determined using the same
criteria outlined in the methodology. Based on these criteria, species
characteristic of Group One (n=11) are B. simaruba (46.7%) and E. can'baeum
(7.3%). The dominant species in Group Three (n=18) is A. scleroxyla (44.2%).
E. caribaeum (17.3%), Maytenus buxtfolia (5.4%) and T. pallida (4.5%) are

secondary species. The dominant species in Group Four (n=11) is P. brasilz’ensis,

87

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

a: 3 4.2 «.3 ed e... an 3 2. 3 we 3 ed a... ages 38%.
e.» n.» :2 at an 3 E 2 nm 2 3. an 8 no 533:8 35.3
on 3 3 an 2 e... 5 a... E I an I 3 2 assess 38%:
mm 3 w... 3 ed 3. ed 3. 2 3 3. an as n... use} smears
mu 2. to 3 «.2 n: 9: «.3 an 3 3 e.— 2 3 stars assuage
o... 3 E 2 a... 2 an a." 2 a. 3 a." 3 2 senses sausage
m: 3 2 2 ed a... no 2 0.2 3.. E E. No 3 $5.5» 283m
an 3 3 S 3 3 q: S. 3 e... 2 a... 3 3. Eﬁssaémsﬁ 5&8
S 3 no 3 a... e... we 3 3.. an e... a." 3 a... sense 3&6
2 3 3 on we 3 an 3 3 an a: n2 3 n.— susee §a3~§§m
a.» an 3. 3 S: 3 5 3 3 an 3 3 3 5 5383... s3
2 v.— 3 a... w... e... 3 3 . 2 o... q... a... 3 3 33883 35:33
3 5 3 e... SN 3: no 2 3 so 3 3 3 2 aassuss «3.58
3 an as a... a: «an E 3 v...” a... an .2 e... E asses ass...
3 3 3. e... 0.: 2: 5 3 ed 2 3 n... 3 ea 2&3 anus:
a we" 3 an 3 a... 4.2 3— 2.2 3— :: was no a: sausages 638%:
39 33 38 as 3.. 33 3..
.9m .52 .Qm 5.: .Qm :82 dd 5.: dw 5.: dd 5.: dw 5.:
sans Que Gus Gus Ens Sue Ens
83:. .33. r7 m 9.96 o 9.20 m 9.20 _ macaw n 96.6 v 9.96 860%

 

 

imam—«am Ease—tom?
Romeo—So mo Mia 32:03 «ma 2: no woman am 3:93 2t go 320 $9328 coco—unemmohs we coma—08% «new 2: ”30:8 360% «o 320
don—£533 gonna—e nachos amnesia—o:— Eowmmmaoo wagon» 30:38 28 be ecu beams—~33 anew hm 05 mo 2.38958 momoomw owauo>< mm 033—.

88

with 79.6% of the basal area on average. Cassia emarginata (6.7%) is a secondary
species. Characteristic of Group Five (n=8) is C. coriaria (49.9%). C. emarginata
(6.3%), A. famesiana (5.8%) and Guaiacum oﬂicinale (2.9%) are secondary
species. Dominating Group Six (n=5) is A. famesiana (35.2%). Coccoloba
leoganensis (16.8%), C. coriaria (11.5%), and P. juliflora (10.5%) are all important
secondary species. P. brasiliensis (28.8%) and Pithecellobium circinale (23.3%) are
the two most dominant species in Group Seven (n=14). B. simaruba (7.1%) and

Lasianthus lanceolatus (3.6%) were secondary species.

Because only half of the sites in the silvicultural study could be classiﬁed into a
distinct cluster group, the noncore sites will also be examined to provide insight
into the species-site relationships in the intermediate areas between the main
cluster groups. Table 14 contains the species composition of six noncore cluster
groups referred to here as "subgroups". Each of these six subgroups represents
sites which showed the same clustering pattern (i.e. the Subgroup BG represents
all the sites which clustered in Group Two twice and in Group Seven twice). The
ﬁrst criteria for picking characteristic species was somewhat more rigorous for
noncore subgroups. To be nominated as characteristic, a species needed to
represent an average basal area contribution three times greater within a subgroup

than across all sites. The second criteria remained the same.

In the noncore groups, four species are characteristic of Subgroup B (n=3),

including P. brasiliensis (44.8%), C. con'an'a (17.8%), G. ojﬁcinale (4.3%) and

89

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

he 3 E 98 we 2 do as .3 3. do a... S a... .1858 598V
3 3 v.2 n: S 2 S. a." 2. 3. ed 3 3. 3 5%... 35.3
9m 3 I a... an 2 no no .3 v... S. 3 as a." 3353‘ 3.59:
mm 3 9m an 3 3 o... .3 ed a... I we do .3 ii 3.3ng
0.2 u: an an a.“ 33 no no 3 no 3 2.. No 3 35.8 3.336
3 3 as 2. 3 a... 3 3 2 a." an n. 3 .2 333% 5.8.36
m: 3 do a... do a... a... 3 3 3 ed a... 3 a. 332.: 283m
an 3 3 a... do 3. 3 a... mi 3. no we 3 I 23%: i6
2 E 3 a... 3 an S 3 3 3 3. 3 an 3 Saginaw
no 5 1m 3 S 2 m: .2 .6 3 3 3 an 3 also 5.385.:
3 3 S 3 3. 3 3 2. 3 3 S. 3 E 3 53.93... «.56
2 I do e... do .3 3 S S 3 S S 3 a... 23885 33.5.4
3 5 do 3. 3 3 m2 3 S: 3 an 2 2 E 3.3..de 23388
3 2 9m 3. a.» 2 E as do 3. as an 3 5” 333.5 398V
3 3 ed a... n2 n: 9: a: Z a... 2 3” S. 3 29S} E
SN 3a 2. a... m.» E a: 3: E at. 2 Q3 3. a: gauga nieaek
as as A5 A5 A5 A5 39
dd 5.: d.w 5.: d.m 5.: d.m 5.: dd :82 dd 5.: d.m 5.:
suns Anus ﬁns 3»: an: ﬁns :75
38... Bob 0 9.233 m gains» 0 9.235 cm gasps m mamas d 9.833. 8.83.

 

 

 

 

 

 

 

 

 

 

63? a: no .83 a .3 32—38;?— 230533 :33 and £598 dogma—u 3mm 28 gm» 2 Egg Mum—Sue “nag

38:8 .«0 and 3.5.80 Eu 2: no @033 a 2:55 05 mo EEO imam—ans 8.363%230 no noun—58:. Em 05 25:8 330% “0 ~35

A58 95..» um bacon—v95 «85 333—0 4033 moan acacia G Adena-m .od use» .3330 £85 Savings 2: .23 @8800»? 2a 22qu
.5022” mum—33“. 083 2: $0391 A033 moan 0.525: .3 “3539—92 a daemon» 55 69.9353 5» .«o Eamon—=8 ”£009 owﬁo>< 3 «Sub

90

Coccoloba leoganensis (2.3 %). In Subgroup BG (n =9), Phyllostylon brasiliensis
(34.0%), Cappan’s cynophallophora (16.6%), Coccoloba leoganensts (5.9%) G.
oﬂicinale (2.9%) and L. lanceolatus (2.3%) are characteristic. In Subgroup C
(n=3) C. con‘aria (25.1%) and A. scleroxyla (20.8%) are dominant. E. caribaeum
(17.3%), Cappans ﬂexuosa (8.9%) and A. famesiana (3.8%) are secondary species.
Characteristic of Subgroup D (n= 14) are P. brasiliensis (57.0%), Cassia emargz'nata
(5.6%), P. juliﬂora (4.8%) and A. famesiana (3.7%). In Subgroup E (n=7), C.
con'an'a (38.0%) and Prosopis juliﬂora (21.8%) are dominant. A. famesiana (7.5%)
and T. pallida (1.9%) are secondary species. Three species are characteristic of
Subgroup G (n=8), Prosopis juliﬂora (17.2%), Cassia emarginata (7.9%) and L.

lanceolatus (1.9%).

Among the four core site cluster groups which correspond to the initial analyses,
there are some differences in the "secondary" species characteristic of each group.
Group One lacks Guaiacum oﬁicinale, Group Four lacks Capparis cynophallophora
and Group Five lacks Capparis ﬂexuosa. A change in the data set affected the
apparent associations suggested by the distributions of these species. This
indicates the some of the patterns observed in the initial data set may have been a
random pattern related to sample size (Pielou 1969). On the other hand, E.
caribaeum is still a characteristic secondary species of groups One and Three, T.
pallida is still characteristic of Group Three and Cassia emarginata and A.
famesiana are still characteristic of Group Five. The persistent presence of these

species regardless of sample size indicates that their associations with their

91

respective groups may not be random. Including P. brasiliensis, B. simaruba,
Caesalpinia coriaria and A. scleroxyla, this group of eight species may represent the

most signiﬁcant patterns of species distributions in the ISA-Mao forest.

Correspondence Analyses

Correspondence analysis is used next to examine the relationships between the
sites within the cluster groups and to explore species relationships suggested in the
previous section. The positions of core and noncore sites are examined to see if
groupings of sites were apparent which correlate with those identiﬁed using
cluster analyses. As noted previously, CA does not assume any structure within
the data set. Therefore, there is no a priori exists for sites associated with a
cluster group to plot in close proximity. As noted previously, because the data
used in the CA procedure are continuous rather than interval, they do not meet
the assumption of chi-square distributions normally used to interpret the results of
CA. Therefore, the observed distances between points plotted using CA scores
are not well-deﬁned. Nevertheless, relative distances between points will be used
to estimate the quality of site groupings formed using cluster analyses. The
relative positions of the species will also be examined for associations with the
respective cluster groups. For groupings of sites which are consistent with the
results of the cluster techniques, the relationships between species and between
sites will be examined for implications with respect to underlying environmental

factors.

92
Figure 11 represents the position of 67 sites plotted with the ﬁrst two principle
axes of a CA procedure applied to the core sites. Scores for the sites and species
are listed in Appendix I. The ﬁrst two axes account for 37.9% of the total
variation in the data set. The proportion of variance attributable to each axis is
listed in Table 15. Although the ﬁrst two axes account for a relatively small
proportion of the total variance, the representation is still an improvement over
using 15 axes (16 species less one degree of freedom: Greenacre 1993) to describe
the precise position of each point. Based on 37.9% of the variation, the
relationships between sites suggested by the cluster analyses are repeated. Sites
within each cluster group plot in relatively close proximity to each other and each
of these clusters is located in an area of space which is discrete relative to the
other groups. Group Three sites represent the most discrete cluster. Groups Five
and Six are well separated from the other groups, but suggest a loose association
between each other. Likewise, groups One, Four and Seven appear to be
associated, with little distance between Group Four and Group Seven, or Group
Seven and Group One. Group Four sites appear to form the most tightly
clustered group, while groups Five and Six have the "loosest" associations among

sites within the respective groups.

As noted in the previous section, the display quality differs for each of the sites.
Each cluster group seems to suggest a characteristic level of display quality for
sites within a group. The median display quality for Group One sites is 38.0%,

for Group Three, 67.4%, for Group Four, 43.2%, for Group Five, 39.4%, for

2.4

2.0

1.6

1.2

0.8

0.4

CA2: 17.1%

0.0

-O.8

-1.2

93

 

4o
4o

50

8o

6o

5o

00

1.0

v

0

5o

5o

5o

O3 3
03 03 3

.3 3
O (1%

 

 

7.7.

lc

1o

1o

)

 

lo

U-
03 0303 08 03

,3

to

 

 

l

—O.8

-0.4 0.0 0.4 0.8

l l l l l

1.2 1.6 2.0

2.4

CA1: 20.8%

Figure 11 Positions of sites from the ISA-Mao silvicultural study based on their scores from the
ﬁrst two dimensions of a CA procedure using basal area contributions of sixteen species in the 67
core sites. The size of the plotting symbols are proportional to the sum of the squared cosines in
two dimensions. Labels represent cluster groups designations. Thirty-eight percent (37.9%) of the
total variance is represented by the ﬁrst two dimensions.

94

Table 15 Singular values, principal inertias and proportional representation for each of fifteen
dimensions of a correspondence analysis of sixteen species and sixty-seven sites representing core
cluster group sites.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Dimension Singular _T Principal Percent Accumulative ll
Values inertias percentage
1 0.75483 056976 20.78 20.78
2 0.68434 0.46832 17.08 37.86
i 3 0.61642 037998 13.86 51.72 i
I 4 056887 032361 11.80 63.52 I
5 0.46979 022070 8.05 7157
6 0.43505 0.18927 6.90 78.47
7 034491 0.11896 4.34 82.81
8 0.30906 0.09552 3.48 86.29
9 0.29361 0.08621 3.14 89.43
1' 10 0.27560 0.07596 2.77 922
11 0.25540 0.06523 2.38 94.58
12 0.23938 0.05730 2.09 96.67
13 0.19638 0.03857 1.41 98.08
14 0.17325 0.03002 1.09 99.17
ll 15 0.15150 0.02295 0.84 100.01
L 2.74235

 

 

 

 

95

Group Six, 28.3% and for Group Seven, 14.8%. The relatively high values for
Group Three sites indicate that the visual position of this group is a good
representation of the information inherent in the data. Group Three sites are
relatively homogeneous and they represent a species relationships unique among
all of the sites. However, values for the other groups are poor, and an

interpretation of their positions is difﬁcult to verify based on only two dimensions.

Figure 12 represents the same CA procedure, using three dimensions to plot each
point. The display quality improves to 51.7 in three dimensions, which still leaves
a display error of just under 50%. Median values for the cluster groups are now
68.0% for Group One, 72.0% for Group Three, 69.5% for Group Four, 42.8% for
Group Five, 39.9% for Group Six and 16.8% for Group Seven. The relationships
between and within groups One, Three and Four are moderately well-described by
three dimensions. Scatter within Group Four is very small, within Group Three is
moderate and is greater within Group One. Although distances between sites
within these groups differs, each cluster is positioned in a discrete area of the
plot. Therefore, the sites within these respective cluster classiﬁcations can be
understood to be closely associated. The display errors of sites in groups Five and
Six remain high. Scatter within these groups is also quite high. Species
relationships among sites within these two groups appear to be more complex
than for groups One, Three and Four, but together, groups Five and Six do

occupy a discrete area of the plot. Differences between these two groups are not

96

 

 

 

 

 

 

 

,1
o\o A 6 . §
0? a .
°’°‘ ’ l
U 4.
, “[i i *1” j
M T j I jl 1' i}
) ll «thHLi‘l i ‘ '

 

‘LP‘ '\
(L9 0.
x95
x?“ ;
9 c- Q 0’
3°; ° ,0 m-
t} o!" ’3’ V
'6; Q
’5‘; 0- so o;
/o '5‘ 9 Q . 4'ng
/ Q'
‘L S" ’
/\' N?” /

Figure 12 Positions of 67 ’core’ sites in three dimensions of a correspondence analysis using the
basal area contributions of sixteen species. The size of the plotting symbols are proportional to
the sum of the squared cosines in three dimensions. Labels represent cluster group designations.
The length of each spike is proprotional to the distance from zero along the third axis. Fifty-two
percent (51.7%) of the total variance is represented by the three dimensions.

97
well-deﬁned by CA, but relative to all other sites, sites in groups Five and Six
represent distinct species relationships. Like Group Two sites in the initial
analyses, Group Seven sites represent the most complex species relationships-
many dimensions are required to describe the precise positions of these sites.
Nevertheless, in three dimensions there is relatively little scatter within Group
Seven. These sites also are positioned in a discrete area of space relative to the
other cluster groups. The species relationships for Group Seven sites are poorly
deﬁned in Figure 12, but differences in composition within the group still appear

to be less than differences between Group Seven sites and all other sites.

The graphical description of the relationships between the 67 core sites using CA
represents a high degree of error for some of the sites. However, the similarity
between groupings of sites using cluster analyses and groupings apparent using CA
is a strong indication that the core site cluster groups do not represent random
associations. The signiﬁcance of the visual positions of each cluster group differs
according to the proportion of variance described for sites within the respective
groups. The positions of sites in groups One, Three and Four can be interpreted

with the most conﬁdence.

As in the initial analyses, the similarity between groupings observed in CA and
groupings observed in the cluster analyses suggests that species relationships
implicit in the cluster groupings should correspond to species relationships

observed in CA. Figure 13 represents the position of each species plotted in two

98

 

 

 

 

 

 

2 4 1 L l l L L l
2.0 — *-
AR .
1.6 ~ 7
CM .
1 2 __ UV GU o _.
3_\:
N 0.8 - ~
H
(\i .. _
< 0.4 SA
0 .
00 __ BA BHOGY . 3 CA _
° CF CIFR P‘AQU
—O.4 — ' r
—0.8 — e
AL,
-1.2

,tfl,o?>,oA 0.9 0A 0% t2 L6 29 “LA

CA1: 20.8%

ﬁgure 13 Positions of sixteen dry forest species based on their scores in the ﬁrst two dimensions
of a CA procedure using basal area contributions from 120 sites in the ISA-Mao silvicultural study.
The plotting symbols are proportional to the sum of the squared cosines in two dimensions.
Symbols are: BA= Phyllostylon bmsiliensis, CM=" Pmsopis juliﬂora, AR= Acacia famesiana, UV=
Coccoloba leoganensis, CF= Imamhus lanceolatus, BR= Cassia emarginata, CI= Pithecellobiwn
circinale, MO= Cappan's ﬂexuosa, FR= Capparis cynophallaphora, AL= Bursem simamba, GY=
Guaiacum oﬂicinale, GU= Caesalpima carian'a, PA= Palo amargo, SA= Maytenus bunfolia, QU=
Exostema caribaeum, CA= Acacia sclemxyla. Thirty-eight percent (37.9%) of the total variance '3
explained by the first two dimensions.

99

dimensions. Some of these positions do correspond with the "characteristic
species" listed for each cluster group. B. simaruba ("AL") is located at the
extreme lower portion of the second axis along the origin of the ﬁrst, in the
general area of Group One sites. A. scleroxyla ("CA"), E. caribaeum ("QU"), T.
pallida ("PA") and M buxzfolia ("SA") are all located to the left of the ﬁrst axis,
along the origin of the second, in the same position as Group Three sites.
Phyllostylon brasiliensis ("BA") is located to the extreme left of the ﬁrst axis along
the origin of the second, in the area occupied by Group Four sites. Caesalpinia
coriaria ("GU"), A. famesiana ("AR"), Prosopis juliﬂora ("CM"), and Coccoloba
leoganensis ("UV") are located along the upper portion of the second axis, in the
area occupied by sites representing groups Five and Six. The remaining species
are clustered around the origin of both axes, in the area occupied by Group Seven
sites. Species missing from the cluster groups include E. caribaeum from Group
One and Cassia emarginata from groups Four and Five. All four species
interpreted as characteristic of Group Three sites correspond precisely, as do the
four species characteristic of Group Six sites. Nevertheless, with a display quality
of only 37.9%, the position of each species is not well described and apparent
associations with particular sites may be misleading. The sums of the squared
cosines are low for most of the species. A. scleroxyla has the highest value with
78.9% of the variance explained for this species in two dimensions. Other values
are 58.3% for Phyllostylon brasiliensis, 51.3% for Caesalpinia coriaria, 48.5% for B.
simaruba, 46.0% for E. can'baeum and 44.1% for A. famesiana. All other species

had values lower than 40%. Based on the relatively high values for Group Three

100

sites and the high value for the species A. scleroxyla, the association between this

group and can be considered to be conﬁrmed. All other relationships are suspect.

By adding the third dimension (Figure 14), the values for the sums of the squared
cosines were increased to 88.5% for A. scleroxyla, 92.5% for P. brasiliensis, 58.0%
for C. con'aria, 94.4% for B. simaruba, 46.2% for E. can'baeum and 55.3% for A.
famesiana. All other species still had values below 40%. Because of the high
values of B. simaruba, A. scleroxyla and P. brasiliensis, we can be conﬁdent in the
signiﬁcance of their relative positions. Because the sums of the squared cosines
for groups One, Three and Four tended to be high, we can also be conﬁdent in
the apparent associations between these species and their respective groups. The
associations between the species located in close proximity to groups Five and Six
are less strong. Nevertheless, the positions of C. con'aria and A. famesiana are
clearly in the same direction from the origin as groups Five and Six. The
association of these two species with their respective groups is probably not

random.

In the same way, E. caribaeum is still associated with Group Three sites, although
its precise position is not well described by three dimensions. likewise, T. pallida,
with only 19.7% of its variance describe is still associated with Group Three sites.
Although only 39.0% of its variance is described by three dimensions, P. juliﬂora

is positioned in the same region as groups Five and Six, as is Coccoloba

 

 

 

 

 

 

 

 

 

 

 

 

101
,1
\" A . . ‘1’
8': ~ .
i ” ‘ i
2% . 1 ‘f’
/ i “WI ~ “ P
.. 1,. ~
. ~ I ll .
’ ”W ‘3
.114 ~- “
i l

 

 

 

92>
{.24
o 09> -
‘67.
t; 0" q, N
his: 09 as? N
be y» 6,0
’0' 6° 0 (L0'
,0‘35 Q?‘ C)" '
‘2. ‘P ’
/\ ‘2; /Q

 

"\

Figure 14 Positions of sixty-seven sites and sixteen species in three dimensions based on a
correspondence analysis. The analysis used the basal area contributions of the sixteen species in
each site. The size of the plotting symbol is proportional to the sum of the squared cosines in
three dimensions. Numbers represent cluster group designation of each site. Letters indicate
species. UV= C. leoganensis, CM= P. juhﬂora, AR= A. famesiana, BA= P. brasiliensis, FR= C.
cynophallophora, MO= C. ﬂexuosa, BR= C. emarginata, CI= P. citcinale, GY= G. ofﬁcinale, CF =
L. lanceolatus, AL= B. simaruba, GU= C. corian'a, SA= M. bwafolia, QU= E. caribaeum, PA=
T. pallida, CA= A. scleroxyla. The length of each spike is proportional to the distance from zero

along the third axis. Variance explained is 51.7% of the total.

 

102

leoganensis, with 20.2% of its variance represented. The interpretation of the
relationships suggested by these species is not strong. However, the positions of
the species strongly deviate from the origin in the direction of the respective
groups. Because these same patterns were observed independently using cluster

analyses, the probability that the associations are random is reduced.

On the other hand, clustered close to the origin, Group Seven sites are not clearly
associated with any particular species. Since the origin represents the average
distribution of all the species, the position of Group Seven sites in this area makes
it difﬁcult to interpret. The grouping of these sites together across four cluster
techniques suggests that they represent a discrete position along the species
continuum. Nevertheless, their position in CA suggests they represent
intermediate sites, with characteristics in common with several of the other
groups. The relationships suggested for these sites by cluster analyses cannot be

completely conﬁrmed using CA.

The results of CA for sites in groups One, Three, Four, Five and Six do correlate
well with the results of the cluster analyses. Species associated with the groups
based on CA also generally correspond well with species nominated as
characteristic of the respective cluster groups. Given that species distributions are
responding to underlying environmental or site history conditions, the relative
positions of both the groups and the species may have ecological signiﬁcance.

The relative positions of groups One, Three, Four and Five are the same as those

103

observed using a subset of data. The ﬁrst axis suggests a contrast between sites
dominated by P. brasiliensrs versus sites dominated by A. scleroxyla. The second
axis suggests a contrast between sites dominated by C. corian'a and sites
dominated by B. simaruba. Along the third axis, Groups One, Five and Six have
high values, while groups Three and Four both have low values. As noted
previously, Group Seven sites are positioned close to the origin along all three
axes. As suggested for the results of the initial analyses, one factor appears to
explain the separation of groups Three and Four, while another affects the
distributions of species associated with groups One and Five. At the same time,
either a hypothetical third factor, or some interaction between the ﬁrst two
explains a strong separation between sites dominated by either P. brasiliensls or A.

scleroxyla from sites dominated by either C. coriaria or B. simaruba.

A second correspondence analysis was applied to the full data set to examine the
interrelationships between core and noncore sites. Scores for sites and species are
listed in Appendix I. Figure 15 represents the positions of the 120 sites in the
ﬁrst two dimensions. Total variation explained is reduced somewhat to 33.2%. It
is therefore even less certain that the relative positions of sites and species
represent an adequate picture of the actual species-site relationships.
Nevertheless, sites in Group One still cluster together at the bottom of the second
axis, Group Three sites cluster together to the right along the ﬁrst axis, Group
Four sites form a tight cluster to the left along the ﬁrst axis, groups Five and Six

are spread out along the top portion of the second axis, and Group Seven sites

104

 

 

1 8 J l 1 L l J 1
a0 E0
8
so ° 50
1.2 - ‘° so
BO 30 5°
8 6. 50
G. 5°5 OE
EH. ‘
O 6 3 Cum. 5.
° 0' 1). I10 5° c. E0co 65C 903
as a“ m 0
F n° 03° in up No 03 3 g
. 0.313. x 03 O 3 g 03
ﬂ 0‘0 no N x 3 g 3
u—r O O 3%310579. 0 05’
e e . 41 $ $3). 70 7.
N 7° 7. 11.7. p
< 7 7 A. o3
7 7. c.
o .b 7 L
_Oe6 — mg. 7.
10
no") 10

—1.2 ~ ‘0 .0

 

 

 

—1.8 r '0

,0ﬁ,QA 09 9A

Figure 15 Positions of 120 sites from the ISA-Mao silvicultural study along the ﬁrst two principle
axes of a correspondence analysis using the basal area contributions of sixteen subtropical dry
forest species. The label indicates the cluster group designation. Numbers indicate core sites.
Letters indicate noncore sites. The size of the plotting symbol is proportional to the sum of the
squared cosines associated with each point for the two-dimensions represented here. Thirty-three

CA1: 18.5%

I l l T
of) if; L6 2.9

l

percent (33.2%) of the total variance is explained by the ﬁrst two dimensions.

1

2A 2%

 

105

are spread along the area between groups One and Four, near the origin of both
axes. Sites in Subgroup C (n=3) are associated with Group Three sites. Sites in
Subgroup D (n=14) are all associated with Group Four. Sites in Subgroup E
(n=7) are mostly found in the area occupied by groups Five and Six. Other
subgroups do not suggest any strong patterns. Adding the third dimension would
increase the display quality to 43.9% (Table 16), but the relationships between the
core sites are not radically affected and the interpretation would therefore be
similar to the previous CA. Figure 16 represents the positions of the species
based on scores from the CA using the full data set. The relative positions of B.
simaruba, A. scleraxyla, E. caribaeum, T. pallida, P. brasilrcnsis, Caesalpinia corian‘a,
A. famesiana and P. julzﬂora are almost identical to Figure 13. On the other
hand, M. buxzfolia ("SA") is less closely associated with A. scleroxyla ("CA"), E.
caribaeum ("QU") and T. pallida ("PA"). Coccoloba leoganensis (”UV") is less
closely associated with A. famesiana ("AR") and Prosopis juliﬂora ("CM"). The
visual associations previously suggested for M. bwafolia and C. leoganensis may
have been an artifact of the reduced data set in the ﬁrst CA. Sample size and/or
sample characteristics affect the patterns "recovered" by CA. However, the

primary structures remain the same.

Summary

Correspondence analysis was used in these analyses to examine the relationships

between sites, independent of the results of the four cluster analyses. The results

106

Table 16 Singular values, principal inertias and proportional representation for each of ﬁfteen
dimensions of a correspondence analysis of sixteen species and sixty-seven sites representing core

cluster group sites.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Dimension Singular Principal Percent Accumulative
Values inertias percentage
1 0.72040 051897 18.53 1853
2 0.64242 0.41270 14.74 33.27
3 0.54279 029463 10.52 43.79
4 0.48965 0.23975 856 52 .35
5 0.46161 0.21309 7.61 5 9.96
6 0.43402 0.1 8838 6.73 66.69
7 0.39142 0.15321 5 .47 72.16
8 038930 0.15156 5.41 7757
9 0.36991 0.13683 4.89 82.46
10 035335 0.12486 4.46 86.92
11 0.31169 0.09715 3.47 9039
12 0.29424 0.08658 3.09 93 .48
13 0.29361 0.08621 3.08 9656
14 0.22374 0.05006 1.79 9835
15 0.21527 0.04634 1.65 100
280031

 

 

107

 

 

 

 

18 I I L l I I l
1'2 _ Aft-M —
GUo
0.6 — r
a»: W 0c.
:5 00 ﬂier SA.
c2 . BA, Ci“. of“
U FR.
-0.6 — —
—1.2 — r
ALO
'-113 r l I l

 

l l l
—O.8 —0.4 0.0 0.4 0.8 1.2 1.6 2.0 2
CA1: 18.5%

 

.4 2.8

Figure 16 Positions of sixteen dry forest species based on their scores in the ﬁrst two dimensions

of a CA procedure using basal area contributions from 120 sites in the ISA-Mao silvicultural study.

The plotting symbols are proportional to the sum of the squared cosines in two dimensions.
Symbols are: BA= Phyllostylon brasiliensis, CM= Prosopis juliﬂora, AR= Acacia famesiana, UV=
Coccoloba leoganensis, CF = Lasianthus lanceolatus, BR= Cassia emarginata, CI= Pithecellobium
circinale, MO= Capparis ﬂexuosa, FR= Cappan's cynophallophora, AL= Bursera simaruba, GY=
Guaiacum oﬂ‘icinale, GU= Caesalpinia coriaria, PA= Palo amargo, SA= Maytenus buxifolia, QU=
Etostema caribaeum, CA= Acacia scleroxyla. Thirty-three percent (33.2%) of the total variance is
explained by the ﬁrst two dimensions.

108

of CA indicate that the groupings of core sites suggested by the cluster techniques
are not random. CA also supports many of the species relationships suggested by
the species contributions within each core site cluster group relative to the
average species contributions across all sites. The graphical representation of
groups One, Three and Four were strongest (the highest proportion of variance
expressed). The graphical representation of Group Seven was the most
ambiguous. Groups Five and Six appear to be closely associated and they could

be considered as two components of the same grouping.

Apparent scatter within each group was greatest in groups Five and Six, somewhat
less in Group One and much less in groups Three and Seven. Group Four
represented the least scatter. The positions of groups One, Three, Four and Five
and Six with respect to the origin tend to conﬁrm that each represents a
characteristic species composition which may relate to characteristic site
conditions and/or disturbance history. The scatter within the cluster groups may
relate to the relative homogeneity of ‘site characteristics within a group. Site
characteristics (site conditions and/or disturbance history) would be most alike
among Group Four sites, and least similar among sites in Groups Five and Six.
Because Group Seven sites have very low squared cosine values, their apparent
afﬁnity in three dimensions carries very little weight. Scatter in additional axes

could be quite extensive.

109

The relative positions of the core site cluster groups corresponds with the
positions of four clusters observed in the initial analyses. A. scleroxyla and
Phyllostylon brasiliensis form one contrast (along with the secondary species, E.
caribaeum and T. pallida), explained by hypothetical Factor One. B. simaruba and
Caesalpinia coriaria form a second contrast, explained by hypothetical Factor Two.
A third axis also appears to be very important which either represents a third
factor, or an interaction between the ﬁrst two. From the perspective of this
silvicultural study, the primary question is which, if any, of these factors can
explain differences in site productivity. That is, does the factor which results in
dominance by P. brasiliensis over A. scleroxyla also affect how fast trees grow on
sites classiﬁed as Group Four versus sites classiﬁed as Group Three? Likewise,
does the factor which affects the relative dominance of C. corian'a versus B.
simaruba affect how fast trees grow on sites classiﬁed as Group Five versus sites
classiﬁed as Group One? Moisture relationships related to topography and soil
attributed probably vary a great deal in the rolling landscape within the ISA-Mao
study. Disturbance histories are also known to vary within the silvicultural study
(Powell and Mercedes 1986). These CA procedures have shown that the cluster
groupings are not arbitrary. Subsequent analyses will focus on the relative
positions of the cluster groups with respect to each other, and the implications
these positions have in terms of underlying gradients within the ISA-Mao

landscape.

110
Canonical Discriminant Analyses
Based on the results of the cluster techniques and the CA procedures, core sites
within each cluster group were assumed to represent samples of their respective
clusters. Canonical discriminant analysis was then used to examine the
relationships between the cluster groups. CDA can also be used to test for
overlap between groupings of sites formed using cluster analyses. CDA was ﬁrst
applied to the 67 core sites representing six cluster groups. Figure 17 represents
the orientation of these six core site cluster groups in two-dimensional canonical
space. Probabilities associated with each axis and scores for each site are found in
Appendix J. Table 17 lists the eigenvalues associated with each dimension. With
72.7% of the variation described by two dimensions, the display quality for Figure
17 is moderately good. The positions of groups One, Four, Five and Seven are
very close, with some interspersion of sites in groups Four and Seven. Groups
Three and Six, on the other hand, occupy discrete areas of space. Group Six, in
particular, dominates the ordination, representing the lowest values on both the

ﬁrst and second axes.

Some of the relationships suggested by the relative positions are the same as those
observed in CA. Groups Four and Three occupy space on opposite sides of the
origin along the ﬁrst axis. Groups One and Five occupy positions on opposite
sides of the origin along the second axis. Group Five is also the closest group to
Group Six. However, Group Five and Group Six are widely separated, while the

relationships of the other groups appear compressed. These relationships are very

 

111

 

 

 

 

 

 

10 l L i l l l J l l l
8 r __
.1
6 _ cl ...
4. .1 1.1
4 ‘ 4.°11_r11 *
2 — .35-"4437'? 474’ _
N ' 77. 77.,7 5
N 0 4 '7 5
g .5; 5 .3
° 0 .5'5 “3’33 .3
.. —2 — e3 3 b
g .. 3 432:.
_4 —— ._
5
_6 _ _
._8 __ _
6.
_10 _ 6 t—
6.
_12 _ _
6. 6
—14 l l l l f l 4 l I l T
-14 -12 -10 —8 -6 -4 -2 0 2 4 6 8 10

CANl: 42.5%

Figure 17 Positions of 67 core sites along the ﬁrst two axes of a CDA procedure based on six
cluster groups deﬁned by sixteen species. Numbers represent the group designation given based
on four cluster procedures. Seventy-three percent (72.7%) of the total variance is explained by the
ﬁrst two axes.

112

Table 17 Eigenvalues associated with each axis of CDA applied to 16 species and 67 core sites
representing six groups from the ISA-Mao silvicultural study.

1 Dimension

 

 

 

 

 

 

 

 

 

different from both the results of cluster analyses and CA. Results from
preliminary analyses suggest that Figure 17 may represent a limitation of CDA.
In a number of preliminary CDA procedures, cluster groups with only a few
members often appeared to have a disparate impact on the relative positions of
the other groups. CDA maximizes the ratio of between group to within group
variance. Small groups may have less variance relative to the larger groups. If
variance is affected bysample size, groups with small sample sizes may have
excessive inﬂuence in canonical discriminant analysis. An analogous problem
exists with CA. Species which are present in a limited number of sample sites
tend to have excessive impact on the analyses and are usually removed after the
initial analyses. In this case, Group Six is represented by a limited number of
sites and was removed in the subsequent analysis to examine the relationships

between the remaining ﬁve core site cluster groups.

CDA was applied again to 62 sites representing ﬁve groups. Figure 18 represents

the positions of the groups in two dimensions. Probabilities for each dimension

113

 

 

 

4!
,4 .4...
3 ° .4.4 .4
1o\\°
. 77 7
° 3. 5 555)“ .g .7
R21 inﬁgﬁi? 8' §B°3. . . 7 7
C0\i O gﬁ 'Z-ﬁf
Z "‘73-'77
5
—3 - .1 —
.1
"1.1 1

 

 

 

 

 

CANI: 57.1%

Figure 18 Positions of 62 sites representing ﬁve core site cluster groups deﬁned by sixteen dry
forest species, along the ﬁrst two axes of a CDA procedure. Numbers represent the group
designation given each site based on four cluster procedures. Eighty percent (79.8%) of the total
variance is explained by the ﬁrst two axes.

 

114

and scores for each site are found in Appendix J. Eigenvalues for each dimension
are listed in Table 18. Accounting for 79.8% of the variation, the display quality
of this graph is slightly higher than the previous CDA and also moderately good.
In this display, each of the ﬁve groups occupies a discrete area of canonical space.
This reafﬁrms the results of the cluster and correspondence analyses. The
groupings of sites in groups One, Three, Four, Five and Seven are not arbitrary-
each group does represent a species composition distinct from the other groups.
This conclusion is particularly important for Group Seven, since the relationships

between sites in this group were ambiguous in CA.

Although all ﬁve groups occupy discrete positions, scatter among groups One and
Seven is greater than the other three groups. Scatter among Group Four sites is
least. The relative degree of scatter among sites probably reﬂects the degree of
homogeneity of species composition within the respective group. Groups in which
species compositions are less homogeneous may represent groups where site
characteristics are also less homogeneous. Based on the visual relationships
suggested by two dimensions in Figure 18, Group Four sites would appear to
represent the most homogeneous species-site characteristics while Group One

would represent the least homogeneous characteristics.

Together, the two dimensions in Figure 18 position Group Three in the northwest
quadrant, Group Four in the northeast and Group One in the southeast. Groups

Five and Seven are positioned near the origin along both axes, which suggests

 

115

Table 18 Eigenvalues associated with each axis of CDA applied to 16 species and 62 core sites
representing ﬁve groups.

Dimension Eigenvalue Percent Qrmulative

1 25.4679 57.09

2 10.1031 79.74

3 6.0874 93.39

4 2.9523 . 100.01

 

they have species characteristics in common with the other groups. Groups One,
Three and Four occupy distinct quadrants, with Groups One and Three occupying
quadrants opposite of each other. This suggests they represent the strongest
contrast among the three groups. In a quadrant intermediate between groups
One and Three, Group Four would appear to represent some factor independent
of the contrast between groups One and Three. However, species compositions
do not reﬂect the same relationships. As observed in the initial analyses, Group
One has species characteristics in common with both groups Three and Four. A.
scleroxyla and E. caribaeum are common to groups One and Three and
Phyllostylon brasiliensis is common to groups One and Four (Table 13). Groups
Three and Four, on the other hand, have no species in common which contributes
more than 4.0% to the total basal area, which suggests these two groups represent
the most complete contrast, while Group One would represent an intermediate
position, withsome independent factor accounting for dominance by B. simaruba.

In this analysis, the relationships suggested by the quadrants are not clear.

 

116
On the other hand, looking at the two axes separately, groups Three and Four
occupy positions on opposite sides of the origin along the ﬁrst axis, and groups
One and Five occupy positions on opposite sides of the origin along the second
axis. These positions do appear to reﬂect differences in species compositions
among the four groups, and they are very similar to the relative to positions of the
groups in CA. However, the contrast between groups One and Five is much less
prevalent than in CA. The dominant relationship is a continuum moving from
Group Three to Group Four. Along this continuum, groups Five, One and Seven
represent intermediate points, with groups One and Seven occupying about the
. same position on the ﬁrst axis. The hypothetical factor which controls this
relationship accounts for 57.1% of the variation between the groups. A second
factor accounts for 22.7% of the variation and would explain the separation of
groups One and Seven. Along this second axis, Group Four is positioned along
the upper extreme, opposite of Group One. Groups Three, Five and Seven
occupy the same position along the second axis, intermediate between groups One

and Four.

If the species which characterize each of the groups are responding to underlying
differences in site conditions (soil attributes, topographic characteristics) and/or
disturbance history (time since last disturbance, type of disturbance, severity of
disturbance), the contrasts suggested by CDA could represent different levels of
these factors. For example, if the ﬁrst axis represents differences in moisture

relationships, Group Three would represent one extreme (either greater or lesser

 

117

moisture availability), Group Four would represent the opposite extreme, and
groups One, Five and Seven would represent intermediate levels. If the second
axis represents differences in species composition due to disturbance, Group Four
would represent one extreme related to disturbance (the most severe or the least
severe, the least time since disturbance or the most time since disturbance),
Group One would represent the opposite extreme and groups Three, Five and
Seven would represent intermediate levels (moderately severe, moderate amount

of time since disturbance). The following diagrams represent these potential

 

 

 

 

 

 

 

interrelationships:
II Least available moisture lntennediate Most available moisture II
Group Three Groups One, Five and Group Four
Seven .
Least disturbance [Most Intermediate Most disturbance [Least
time since disturbance time since disturbance
Group One Groups Three, Five and Group Four
Seven

 

 

 

 

 

These models are hypothetical, with the main goal being to put the results of the

ordination in a real world perspective.

Since the relationships suggested by the ﬁrst two dimensions of this CDA
correspond in part to CA, the third dimension was examined to see if it was
similar to the third dimension using CA. Figure 19 represents the positions of the
ﬁve cluster groups in three dimensions and accounts for 93.4% of the total

variation. With a display error of only 6.6%, considerable conﬁdence can be

118

so

 

 

 

 

 

 

 

(DAN‘S: 13.7%

 

 

 

Figure 19 Positions of 62 sites representing ﬁve core site cluster groups deﬁned by sixteen dry
forest species, along the ﬁrst three axes of a CDA procedure. Numbers represent the group

designation given each site based on four cluster procedures. Ninety-three percent (93.3%) of the
total variance '3 explained by the ﬁrst three axes.

119
placed in the visual representation of the relative positions of each group. As
observed with the third dimension of CA applied to 67 sites and 16 species,
Group Five has large values in the third dimension, groups Three and Four have
low values, and Group Seven remains close to the origin in all three dimensions.
One major difference is that Group One remains close to the origin in the third
dimension whereas in CA Group One sites had the same relative relationship as
group Five. Visually, the relative scatter within each group remains the same in
three dimensions: Group Four has the least and Group One appears to have the
most. The most important relationships apparent in Figure 19 is the removal of
Group Five from the origin along the third axis. Based on the two-dimensional
image, Group Five was interpreted as representing an intermediate level in both
of the hypothetical primary environmental factors. The third dimension suggests
that the species characteristic of Group Five sites represent unique site
characteristics, rather than an intermediate level of the factors controlling the
primary contrasts apparent between the other groups. This relationship can be

diagrammed:

 

 

 

Group Five Groups One and Groups Three and
Seven Four

 

 

 

As in the initial analyses, CDA was also applied to a data set including both core
and noncore sites. Applying CDA to the larger data set implies that each site

represents a sample of its respective group. However, noncore sites represent

120

sites which were rejected as samples of the core site cluster groups. This presents
a limitation in this CDA procedure. The advantage of using most of the sites is
that it gives a more complete picture of species-site interactions. Therefore,
noncore sites were submitted as samples of their respective subgroups rather than
as samples of the core site cluster groups. The single exception was a single
noncore site labeled A, which was assigned to cluster Group One. Two sites were
eliminated which grouped with a different core site cluster group in each of the
four cluster techniques. Subgroups represented by fewer than three sites were
combined with other subgroups. For example, there were two noncore sites
labeled BD which were submitted as members of Subgroup B (n=5). The two
subgroups FG (n=2) and FH (n=1) were combined into a subgroup labeled "F"
(n=3). A noncore site labeled EH was combined with the seven sites
representing Subgroup E (n=8). A total of thirteen groups were submitted to
CDA in this analysis. The smallest groups had n=3 and the largest had n=18.
Scores and probabilities are listed in Appendix J. Eigenvalues associated with

each dimension are listed in Table 19.

Figure 20 represents the positions of these 118 sites representing thirteen groups
in two-dimensional canonical space. The display quality of this graph is 61.2%.
With a display error of 38.8%, the visual representations of group positions is not
as good as the previous graphical representations. Nevertheless, with more than

half of the variability visually represented, apparent relationships are probably

 

121

 

 

 

 

 

 

 

8 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
7 — __
6 — so _
5 “a 4o 4b 30 _
‘ED
4 — 30 —
40 ﬂ 0‘
3 r ”683 4030 h 0‘ ‘1 _
2 _ 30 b% on Ob” (9‘ 94 _
O\0 3‘50 1) DO
3 1 — O7 (bﬂlzbonn 070:0 1—
N 0 co CO 9,5996%an CD0140
00 bit) C(JOdﬁg B 7
C2 —1 — Fo F40 “050 c(4 O 0'9 T
_2 .3 1100:: 01 _
< @b (85‘ 5 1
U -3 q 05 0 _
"4 —‘ FC) 0631 —
-5 _ [O «39% 50 _
—6 L ,0 ”0 L
-7 L L.
...8 — __
‘9 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
-9 —8 —7 —6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8

CAle 41.2%

Figure 20 Positions along ﬁrst two canonical axes of six core groups and seven noncore site
Subgroups. Groups and subgroups are represented by a total of 118 sites. Groups and subgroups
represent the results of four cluster analyses applied to data representing relative basal area
contributions of sixteen dry forest species. labels indicate cluster group designations. Numbers
indicate cores site clusters. Letters indicate subgroups formed from noncore sites representing the
same clustering pattern (i.e. subgroup BG indicates site clustering twice in group two and twice in
group seven). Sixty-one percent (61.2%) of the total variance is explained by axes one and two.

122

Table 19 Eigenvalues associated with each axis of CDA applied to 16 species and 118 sites
representing six core site cluster groups and seven noncore site subgroups.

Dimension Eigenvalue Percent Qrmulative

1 14.3750 41.15
7.0065 61.21
5.3744 76.5 9
2.7168 . 84.37
2.1852 90.62
15925 95.18
0.9506 . 97.9
0.4458 99.18

0.1135 99.5

0.1089 99.81

0.0610 . 99.98

0.0051 . 99.99

 

signiﬁcant. Unlike Figure 17, none of the groups with small sample sizes
dominate either of the ﬁrst two axes. This may reﬂect a great deal of variability
within the subgroups with small n since they are composed of noncore sites, which
were those sites which were not consistently associated with any one type of
species distribution. Greater within group variability would decrease the ratio of
between group to within group variability. Separation of the small groups would
therefore also decrease. It is also possible that increasing the number of groups
decreased the between group variability. If we added additional groups from
increasingly diverse habitats, this would not be true- the variability between the
groups would increase. But in this case, sites were added to the data set which
represented points intermediate between the core groups, thereby decreasing the

variability between the groups. Again, if between groups variability decreased, the

123

ratios of between to within group variability also decreased, and no single group

would have an overwhelming impact due to sample size.

As in the initial analyses, applying CDA to both core and noncore sites results in
the position of the groups along a continuous gradient, versus the discrete
positions represented when only core sites are used. This supports the idea that
the subgroups represent intermediate points between the discrete positions of the
core site cluster groups. Of the thirteen groups, Three and Four are the only two
which occupy discrete positions, representing opposite ends of the continuum
along the ﬁrst axis. Subgroups C and D represent two groups with minor
interspersion with other groups. Grigal and Goldstein (1971) found that several
clusters overlapping in canonical space had common species characteristics.
Adjacent to groups Three and Four, respectively, subgroups C and D suggest
relationships based on the common characteristic species A. scleroxyla and P.

brasiliensis. These relationships concur with the relationships observed in CA

(Figure 13).

Adjacent to Subgroup D along the ﬁrst axis is a mixture of sites from groups One
and Seven and subgroups BG, B and G. P. brasiliensis is a characteristic species

of Group Seven and Subgroups BG and B. P. brasiliensis is also a component of
Group One (12.3%) and Subgroup G (18.9%). These relationships are suggested
by the average contributions of each species to their respective groups (Tables 13

and 14), but were not explicitly demonstrated in the CA procedures. Immediately

124

adjacent to Subgroup C along the ﬁrst axis is a mixture of sites from groups One,
Five and Six and subgroups E, F and G. The relationship among Five, Six and E
can be understood as an expression of common occurrence of the characteristic
species Caesalpinia coriaria (Five, Six and E), A. famesiana (Five, Six and E) and
Prosopis juliﬂora (Six and E). These relationships are consistent with explicit
relationships represented in CA (Figures 11-16). The interspersion of Group One
sites among groups Five and Seven is also consistent with the relative position of
Group One sites along the ﬁrst axis of the CA applied to 16 species and 67 core
sites (Figure 11), but cannot be explained by species characteristic of all three
groups. Tables 20 and 21 list the average score for each group and subgroup

along the ﬁrst axis.

Table 20 Group means for the core site cluster groups based on site scores along the ﬁrst
canonical axis of canonical discriminant analysis (CDA) using 118 of the 120 silvicultural
experiment sites. Canonical discriminant analysis was based on the proportional basal area
contributions of sixteen dry forest species, using groupings determined with four different cluster
analysis procedures.

Group Six

7 Four :

 

 

-1.676

Table 21 Group means for the noncore site cluster groups based on site scores along the ﬁrst
canonical axis of canonical discriminant analysis (CDA) using 118 of the 120 silvicultural
experiment sites. Canonical discriminant analysis was based on the proportional basal area
contributions of sixteen dry forest species, using groupings determined with four different cluster
analysis procedures.

I Subgroup Sbgropu Subgrop Sgroubup Subgroup m Subgruop Sugrboup

 

 

 

125

The second axis in Figure 20 positions a mixture of groups Five, Six and E
towards the lower extreme, and represents a common position of groups Three
and Four towards the top. This axis suggests a curvilinear relationship with the
ﬁrst. Digby and Kempton (1991) suggest that an ordination of data along two
principal axes will result in such a curvilinear relationship when the sites represent
ecologically diverse habitats. An interpretation of the curvilinear relationship is
difﬁcult, because the effects of a possible second environmental gradient are
confounded by an interaction with the ﬁrst. Digby and Kempton (1991) suggest
that a clear representation of such an additional gradient may be hidden in a
higher dimension. In the CA analyses, the ﬁrst two axes appeared to represent
two independent gradients, while the third axis suggested some nonlinear
relationship with the ﬁrst axis. Sites with large absolute values along the ﬁrst axis
(groups Three and Four) had relatively high negative values on the third axis,
while sites with values near zero on the ﬁrst axis (groups One and Five) had
relatively high positive values on the third axis (Figure 12). When CDA was
applied to 62 sites representing ﬁve groups no relationship between the ﬁrst
dimension in either the second or the third dimensions was apparent. However,
this CDA examined only half of all the sites and therefore may not represent the

full complexity of species-site continuums.

Figure 21 represents the positions of each group using the ﬁrst and third
dimensions of the CDA applied to 118 sites representing 13 groups. The quality

of this graphical representation is only 56.2%, indicating a display error of close to

126

 

 

 

 

 

 

8 L l l l l l l l l l l l l l l
7 q _
6 — __
5 -— _.
4 " b wt) 1’0 T’
3 -‘ b b b or: L—
’O‘O (5,; 0”
°\0 2 — .9300 to 5008 O‘5 6919' 0° 9 O —
~44 ._ .4» "a. one @8480.
' so ’0 898' 02%;

L0 ’63:?) ’0 ’O O on

H 0 u ‘6 ”at; E c 013 5

6’5 —1 — ‘b ’0 8G 97 70” —

Z ('0 08 0709

<: ’2 ‘ ‘0 07070" . r

O7 07 O

Q -3 .1 01" 07 .—
- _ 0 1 _.
4 @1 O O!
_5 —.1 .—
—6 — ’ob e
-7 _ '90 _

01
_8 — ‘3! 1—
‘9 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
-9 —8 -7 —6 -5 -4 -3 -2 —1 O 1 2 3 4 5 6 7 8

CANl:

41.2%

Figure 21 Positions along ﬁrst and third canonical axes of six core groups and seven noncore site
subgroups. Groups and subgroups are represented by a total of 118 sites. Groups and subgroups
represent the results of four cluster analyses applied to data representing relative basal area
contributions of sixteen dry forest species. labels indicate cluster group designations. Numbers
indicate cores site clusters. Letters indicate subgroups formed from noncore sites representing the
same clustering pattern (i.e. subgroup BG indicates site clustering twice in group two and twice in
group seven). Fifty-seven percent (565%) of the total variance is explained by axes one and three.

127

50%. However, this display is useful because it indicates a discrete position for
Group One, while maintaining most of the separation of groups Three and Four.
Group Seven also occupies a more nearly discrete position, as does the mixture of
groups Five, Six and E, although there is some interspersion with sites
representing subgroups BG and F. Along this third axis, Group One and groups
Five and Six occupy opposite extremes, much as they did in the CA procedures.
The relative positions of groups One, Three, Four, Five and Seven are all very
similar to their relative positions in two dimensions based on CDA applied to ﬁve

of the six core site cluster groups (Figure 17).

Figure 22 represents the positions of groups One, Three, Four, Five, Six and
Seven in all three dimensions. The display quality is 77.2%, which is moderately
good. The groups are plotted with the second canonical axis representing the
third dimension. The second and third dimensions were switched in this display
to maintain the separation of Group One from groups Five and Six. The
separation of Group Five from groups Three and Four is much clearer in three
dimensions. The similarity between this display and Figure 12 (CA) is signiﬁcant.
Groups Three and Four occupy opposite extremes along the ﬁrst axis, groups One
and Five occupy opposite extremes along the second axis (=third dimension) and
groups Three and Four occupy one extreme along. the third axis (=second

dimension), while groups One and Five occupy the opposite extreme. Also as in

CMNhZDﬁé

canteens-moans:

\
13‘».th

,6

I 1

L

\

 

4111

H8

493.2%
1&23

0: dg‘L* i
‘b g

 

 

 

 

 

A
V

A
U

 

 

 

 

<
>)/ 1
D < ‘
.\
b
'5 <
‘74
o x q
0.4 . o ‘o 6
\g 3 A
a ,‘b o \
Q I h ‘7' ‘5’ ‘3 I O
’6 I .‘ $ °
1% ,6 ['1
)9 3

Figure 22 Positions of six cluster groups along the ﬁrst, second and third axes of a CDA
procedure applied to thirteen cluster groups represented by a total of 118 sites. Only the six core

site cluster groups are plotted. Labels indicate cluster group designations. Seventy-seven percent
(76.6%) of the total variance is explained by the three axes.

129

the CA procedures, Group Seven is positioned near the origin in all three

dimensions. These ordinations can be diagrammed as follows:

 

 

 

 

 

 

 

 

 

 

Group Three Groups One, Five, Group Four
Six and Seven
Group One Groups Three, Four Groups Five and Six
and Seven
Groups Three and Group Seven Groups One, Five
Four and Six

 

 

 

 

 

Apparent scatter within the groups is also similar to the results of CA. Scatter
within group One appears to be the greatest, with two sites adjacent to the area
occupied by Group Seven. Scatter within the complex composed of groups Five
and Six is also considerable, although all of these sites remain well separated from
areas occupied by other core sites. Scatter within groups Three and Seven is
somewhat less than Group One and groups Five and Six. Group Four is
represented by the sites with the least scatter. Although similar to CA, the
proportion of variance unexplained in Figure 22 could result in radical changes in
the relative degree of scatter within each group. However, these relationships are
the same as observed in Figure 18, which represented over 95% of the variability
in the data set. Some of these relationships may change in higher dimensions, but
it can be concluded with reasonable assurance that Group Four represents the
most homogeneous species compositions, groups One, Five and Six represent the
least homogeneous, and groups Three and Seven represent an intermediate

degree of within group homogeneity. The degree of homogeneity in species

130

compositions may reﬂect the relative degree of similarity among site

characteristics within each group.

The relationships diagrammed above are similar to the ones suggested by the
CDA applied to ﬁve of the six groups and virtually identical to the relationships
suggested by the three axes in the CA applied to 67 sites representing the six core
site cluster groups. Discrepancies between the two CDA procedures are probably
related to differences in the sample size of the two data sets. These discrepancies
suggest that some of the relationships observed in CDA may be arbitrary. The
patterns which are consistent throughout all of the analyses are least likely to
represent random relationships. Groups Three and Four indicate a primary .
gradient operating in the forest. Group One indicates a second gradient. Groups
Five and Six are closely related, and may represent the opposite extreme of the
gradient affecting Group One. However, Groups One and Seven appear to have
some common characteristics, as do groups Four and Seven, which means that
these three groups represent a species-site continuum which could also indicate an
underlying gradient. In every graphical representation, Group Seven represents
an intermediate point between two other groups at opposite ends of a continuum.
Finally, in the analyses including noncore sites, subgroups C, D and E are
consistently positioned in close association with groups Three, Four and Five/Six,
respectively. Subgroup C is represented by only three sites, but Subgroup E is
represented by seven sites and Subgroup D by fourteen. In subsequent analyses,

subgroups D and B will be examined along with the core site groups as

131

representatives of the points intermediate between the discrete positions along the

species-site continuums.

Summary

The results of MVA techniques applied to the full data set conﬁrm many of the
patterns observed in the subset of 45 sites initially analyzed using the same
procedures. Patterns of species compositions can be described for the entire data
set. Sixty-seven of the 120 sites can be assigned to one of six groups, each
representing a unique species composition. These groupings are consistent with
an independent analysis using CA, although groups Five and Six (dominated by C.
coriaria and A. famesiana, respectively) tend to be closely associated. Groups
One, Three, Four, Five and Seven can be shown to occupy discrete areas of
canonical space, based on CDA applied to these ﬁve groups. Based on the CDA
applied to all six core site groups and seven noncore site subgroups, a continuum
was strongly suggested, rather than discrete positions for each group. In
particular, the second dimension suggested a curvilinear relationship with the ﬁrst
axis, a relationship which has often been observed in temperate zone studies
where samples come from sites representing diverse habitats. Groups Five and

Six and Subgroup B were closely associated along this continuum.

Using the scores in three dimensions from the same CDA procedure, the relative

positions of the core site groups were very similar to the relative positions of each

132

group in three dimensions based on CA applied to the 67 core sites representing
the six core site groups. Relationships common to all of the analyses suggest one
strong environmental gradient represented by the characteristic species A.
scleroxyla, E. can'baeum and T. pallida versus Phyllostylon brasiliensis. Along this
gradient, C. coriaria, A. famesiana, Prosopis julzﬂora and B. simaruba represent an
intermediate species association. In all of the analyses, a second gradient
separated sites dominated by B. simaruba from this primary gradient. In some of
the analyses, the second apparent gradient also separated out sites associated with
C. coriaria, A. famesiana and P. juliﬂora. In the CDA procedure applied to
groups One, Three, Four, Five and Seven, the second axis did not suggest a
contrast between groups One and Five, but the third axis did separate out Group
Five from all other groups, suggesting a third gradient responsible for the
distribution of C. coriaria and A. famesiana. In each ordination, Group Seven
sites were in an intermediate position near the origin, in a loose association with a
number of species, including Pithecellobium circinale. In particular, the
"characteristic species" assigned to Group Seven suggest an intermediate point
along a continuum moving from Group One to Group Four. The primary
question with respect to all of these apparent gradients is which, if any, can
explain differences in site productivity. In the following section, review of the
literature and personal observations, in combination with analyses of the overstory
structures within each group will be used to examine plausible implications for

environmental gradients within the ISA-Mao silvicultural study.

133

Species, Site and Overstory Characteristics

The primary question with respect to the silvicultural study is which, if any, of the
apparent gradients reﬂect differences in site productivity. Differences in available
moisture would affect site productivity most directly. However, site productivity is
not independent of disturbance factors. On sites dominated by young trees,
diameter increment may be relatively rapid. As a site develops, average diameter
increment may slow, but the rate of basal area accumulation may stay the same.
Differences in physiology also affect apparent productivity. Tree species with
highly dense wood may actually accumulate more total mass than trees with a
lower speciﬁc gravity, although diameters and basal area may increase more for
the trees with the less dense wood. These are just some of the factors which

make an interpretation of relative site productivity complicated.

The greatest limitation is a lack of information. Information from the literature
with respect to the ecology and physiology of each species is scarce, but does
suggest some important characteristics for some of the key trees. This literature
will be used to examine possible site characteristics as indicated by different
species dominance. Direct information on site characteristics is limited to the
qualitative "measurements" made on forty of the sites in March 1993, as well as
general observations with respect to topographic characteristics of the
experimental areas. These measurements and observations will be used to suggest

some possible ecological relationships for the dominant species. Some literature

134

is available which suggests possible relationships between structural characteristics
and site quality and structural characteristics and disturbance. Based on
relationships suggested in these studies, overstory structures of the sites within
each cluster group will be examined for indications with respect to site

characteristics.

Species Characteristics

Dominant in Group One and characteristic of Group Seven, Bursera simaruba is
described by Kellman and Roulet (1990) as a major component in a well-
developed selva community on fossil sand dunes with a caliche layer 140 cm below
the surface. Geilfus (1989) notes that B. simaruba is tolerant of rocky soils, salinic
soils and soils with caliche. In the southwest region of the Dominican Republic,
the species can be seen growing out of shallow pockets of soil deposited in
crevices of uplifted coral reefs on the steepest slopes, accompanied by small
shrubby trees and cacti. In Costa Rica the tree is also associated with cacti,
particularly Lemaireocereus aragonii (Weber), on dry limestone outcrops (Otis and
Buskirk 1986). In another study done with B. simaruba in Costa Rican dry forest,
Stevens (1987) found that annual diameter increment varied from 0.0 cm in the
"worst" year to 2.5 cm in the "best" for the same individual B. simaruba. This
suggests the high degree of variability which can exist in terms of growth from
year to year. With live wood composed of over 50% water (Maxwell 1985),
growth for B. simaruba may be particularly dependent on differences in rainfall

patterns from year to year. Johnson (1992) notes that in the Sonoran desert, B.

135

simaruba occurs only in river canyons. Asprey and Robbins (1953) list B.
simaruba and E. caribaeum among the important components of forest described
as "dry limestone scrub" found in Jamaica in hilly or mountainous landscape with
thin soil over hard limestone rock. This literature suggests that B. simaruba may
be indicative of rocky or shallow soils and/or highly alkaline soils developed under

low rainfall ﬁ'om limestone material.

Table 22 lists estimates of average growth and speciﬁc gravity for nine species
which represent important components in one or more of the core site cluster
groups. Table 23 lists some structural characteristics of each species. Bursera
simaruba had the highest average annual growth in the study by Hernandez
(1986). B. simaruba is also the species with the lowest speciﬁc gravity. Powell
and Mercedes (1986) and Maxwell (1985) note that the tree is not favored for
charcoal production, which may explain its apparent dominance in some parts of
the forest. Structurally, B. simaruba trees were the tallest and had the largest
diameters. None of the individuals were multiple-stemmed. These characteristics
indicate that the trees present in the forest in 1986 were not stump sprouts and
may have been relatively old. Removal of other vegetation may have made
resulted in a de facto dominance by B. simaruba on some sites. The tree’s
relatively fast growth may also have allowed it to become even more dominant as
competition was removed. However, the structural characteristics suggest that

disturbance did not result in new regeneration of the tree on sites where it was

136

Table 22 Speciﬁc gravity and average growth rates for nine species prominent in the ISA-Mao
silvicultural study. Species are listed in descending order based on average annual growth. "NA”
indicates species for whidr information is not available from the respective source.

 

 

 

Acacia fanrariana

 

‘ 3 l . . . .
Phﬂloslﬂon bmsilicnsir

 

 

 

 

 

 

 

 

 

 

‘ Pitlwcellobium circinalc

previously not found. Rather, B. simaruba may represent a residual component of

a pre-existing forest type which dominated on highly alkaline and/or shallow soils.

Exostema caribaeum is the species with secondary dominance in both groups One
and Three. Hernandez found that it had a very slow rate of growth. Data over
six years also indicates this species grew quite slowly. Speciﬁc gravity is quite
high. Few of the trees in the silvicultural study were multiple-stemmed and the

average height indicates the tree was part of the upper portion of the wow.

 

lSource: Hernandez 1986. Based on measurements taken at breast height.
230nm: Betances 1983

3Total growth after six years, based on measurements taken at knee height (05 m) in the 120
silvicultural sites examined in the current study.

137

Table 23 Structural characteristics for nine species prominent in the ISA-Mao silvicultural study.
Species are l'mted in descending order based on average height. Data comes from the preharvest

inventory done on 120 sites, each 100 m’, in 1986. Means are based on measurements taken at
knee height (05 m).

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Species 7 if i 2 7#0f 1 Meanheight(m) Mean I 2' number ‘
trees diameter (cm) of trunks per
tree

Burserasimamba 73 5.6 114 1.0 I
i Trichilia pullida 87 50 4.9 13 1
Acacia sclcmrﬂa 182 4.9 6.1 1.1 1
Phylbsrﬁon bmsiliemir 803 4.9 5.9 1.1 I
‘Execumceubceum 219 4.9 5.4 1.1 1
Acacia famesiana 86 4.8 4.1 1.6 H
* Praeopis juliﬂom 133 4.7 5.1 13

i Pillrecdlobium cireinalc 146 45 3.7 33 JI
j Caesalpim'a coriaria 129 43 9.0 1.6 l
;. Totals‘ 2442 4.6 5.8 13 J

Van Paasen (1986) notes that E. caribaeum is used for fence posts, house
construction, ﬁrewood and charcoal. In the ISA-Mao Forestry Experimental
Station, Powell and Mercedes (1986) noted a heavy dominance by E. can'baeum in
an area previously under cultivation. Checo (personal communication) has
observed the tree growing in rocky and shallow soils. As noted above, Asprey and
Robbins (195 3) list both B. simaruba and E. caribaeum as important components
of Jamaican "dry limestone scrub". This species may be favored by some kinds of
disturbance, but its use for a number of subsistent products and its slow rate of

growth would not favor large specimens in areas of forest subjected to human

 

lTotals are based on measurements of all trees across all species for 120 sites.

138

intervention. The tree would appear to be tolerant of shallow and rocky soils

and/or highly alkaline conditions.

Acacia scleroxyla is the species most dominant in Group Three. Almost nothing is
available in the literature concerning this species. Neither Little and Wadsworth
(1964) nor Little et al. (1974) list it as one of the trees of Puerto Rico and the
Virgin Islands. Nor do Asprey and Robbins (1953) mention it as a component of
any forest type in Jamaica. Hernandez (1986) found it to have moderately low
annual diameter increment and Betances (1983) found that it had a moderately
high speciﬁc gravity. Knudson et al. (1988) note that it is one of the species most
favored for charcoal production in the Mao region. With moderately slow growth,
even moderate sized specimens of A. scleroxyla may be relatively old. Because
they are in high demand as a source of charcoal, sites with moderate sized trees

probably represent areas where disturbance has been minimal.

Trichilia pallida (Palo amargo) appeared consistently as a secondary species in
Group Three. The scientiﬁc identiﬁcation of this species is in question, based on
the description provided by Little and Wadsworth (1964). Information available
from the ISA-Mao studies is mostly limited to the growth rate listed in Table 22.
Van Paasen (1986) notes that the tree is commonly used for charcoal, ﬁrewood
and fence posts. As noted for A. scleroxyla, sites with this relatively slow growing

species present may indicate areas where disturbance has been infrequent.

139

A main component of Group Four, Group Seven and Subgroup D, Phyllostylon
brasiliensis is also the dominant species in the ISA-Mao silvicultural study as a
whole (Table .2) and is noted by Knudson et al. (1988) as representing the species
with the most basal area in the whole experimental station. Structurally, P.
brasiliensis did not tend to be multiple-stemmed in the preharvest inventory.
Mean height indicates it was generally part of the upper portion of the canopy,
but on average, the trees had relatively small diameters. Powell and Mercedes
(1986) note that the species appears to be favored by some types of disturbance.
Maxwell (1985) notes that the tree has not traditionally been favored for charcoal
production in the area as a whole and suggests that this is the reason for its
apparent dominance in the forest. However, in a relatively undisturbed dry forest
of Venezuela, Tamayo (1963) found that species dominating the overstory
included Phyllostylon rhamnioides, Caesalpinia coriaria and Cassia emarginata as
well as several cacti. It is likely that P. brasiliensis was also an important

component of the original forest at the ISA-Mao Forestry Experimental Station.

The prevalence of P. brasiliensis in groups Four Seven, and Subgroup D may be
partially related to disturbance. Site may have been cleared of all "valuable"
species, leaving P. brasiliensis as the de facto dominant. The removal of
competition may have also favored the moderately fast growing species. The
small average diameter suggests many of the trees are young, but the low number
of multiple stems indicates that most of the trees are not stump sprouts. Removal

of other species may have also encouraged natural regeneration by P. brasiliensis.

140

However, sites with large specimens of P. brasiliensis~ are probably either areas
where the tree was a ”natural" component of the original forest or where
disturbance occurred a long time ago. Sites with smaller trees may be areas
where disturbance favored regeneration of P. brasiliensis over other species.

These areas may also be sites where poor site conditions have prevented rapid

growth.

Dominant in Group Five and Subgroup E and a characteristic species of Group
Six, Caesalpinia coriaria has a growth rate comparable to P. brasiliensis based on
average diameter increment. However, mean basal area increment was higher
than all of the species examined here. This discrepancy between diameter
increment and basal area increment may be related to the relatively large average
diameters of C. coriaria in the preharvest inventory. This suggests these larger C.
coriaria were vigorous, with relatively high annual diameter increments, and did
not represent older growth which had reached equilibrium. Structurally, the trees
were the shortest of all the species examined and tended to have more than one
stem. Checo (personal communication) indicates that C. coriaria is a preferred
species for charcoal production, which is consistent with its high speciﬁc gravity.
Tamayo (1963) noted that in dry forest adjacent to population centers, in
Venezuela, C. coriaria was managed in open groves for fruit production which was
used as goat fodder and for production of tannins. Van Paasen (1986) also notes
that the fruits of C. coriaria are sold locally in the Mac area as a source of tannin.

As mentioned above, Tamayo (1963) found the species in association with

141

Phyllostylon rhamnioides and Cassia emarginata in undisturbed dry forest. Asprey
and Robbins (1953) list C. coriaria as a component of "thorn scrub" forest in deep
alluvium soils near the coast of Jamaica, along with Prosopis juliﬂora and A.
famesiana among other species. Although C. coriaria is leguminous, Hunter and

Steward (1993) found that specimens grown in Honduras did not ﬁx nitrogen.

A relatively fast growing species, C. coriaria may respond rapidly after disturbance.
The number of trees with multiple stems may indicate the tree sprouts readily
when cut, which would also allow it to respond rapidly to disturbance. Sites with
very large specimens may indicate areas where disturbance related to charcoal
production has been minimal. However, the commercial value of the fruits may
also have provided incentive for the trees to be left alone. In either case, the area
inﬂuenced by the tree would be less affected by the removal of surrounding
vegetation. The literature from Jamaica suggests C. coriaria does best on deep

soils.

The species with the next best annual growth after B. simaruba is Acacia
famesiana, the dominant species in Group Six sites, and one of the characteristic
species for Group Five and Subgroup E. The results of the study by Hernandez
(1986) notwithstanding, A. famesiana was one of the species with the worst
average growth over the six years of data examined in this study. Unlike B.
simaruba, A. famesiana has a moderately high speciﬁc gravity, ranking ﬁfth among

the species with information available. With a density over three times greater

142

than B. simaruba, the same diameter growth would represent proportionally
greater biomass accumulation. Ecologically, A. famesiana is associated with heavy
disturbance (Peacock and McMillan 1968, Powell and Mercedes 1986, Vora and
Messerly 1990). Van Auken and Bush (1991) note that in the south and
southwest United States, A. famesiana is found in old ﬁelds or grasslands which
have been heavily grazed. They also note that the tree grows best in full sunlight
without herbaceous competition. Kellman and Roulet (1990) list it as a primary
component of secondary succession observed in a sand dune formation. As noted,
Asprey and Robbins (1953) list A. famesiana as an important component in forest
found on deep alluvium soils located near the coast in Jamaica, along with
Prosopis juliﬂora and Caesalpinia coriaria. Hunter and Steward (1993) indicate A.
famesiana is a nitrogen ﬁxer. A. famesiana may indicate disturbed forest, but it
would appear to do best on deep soils without impediments to root growth, and
with little competition from other trees. It’s poor performance in the ISA-Mao
silvicultural study could be the effect of the surrounding forest recuperating from
pre-study disturbance. As shade from the other species increased, growth of A.

famesiana may have been reduced.

An important component of Group Six and Subgroup E, the literature available
for Prosopis julzﬂora suggests characteristics which are similar to A. famesiana. It
is a species frequently associated with disturbance (V ora and Messerly 1990,
Poynton 1990) which does best in deep soils where roots can penetrate to

permanent sources of soil moisture (Ruskin 1980). Given adequate conditions, P.

143
juliﬂora is highly tolerant of drought (Ruskin 1980, Lees et al. 1992). Asprey and

Robbins (1953) list it as the primary component of coastal thorn scrub forest on
alluvium soils exposed to ocean spray. In the southwest region of the Dominican
Republic, P. julzﬂora is an important component of the forest on footslopes, but
quickly disappears as slopes become steeper and/or elevation increases. Based on
observations of charcoal production in this area, P. julzﬂora responds rapidly when
cut, sprouting rapidly and vigorously from the remaining stump. Dominance by P.
juliﬂora in this area may be related to historical use of the forest for charcoal
production. In the silvicultural study, the tree tended to be moderately short, with
small stems. The mean number of stems per tree was higher than for trees such
as P. brasiliensis and A. scleroxyla, but less than for C. coriaria and A. famesiana.
Growth for these trees was about average compared to the other species. Hunter
and Steward (1993) indicate P. juliﬂora is a nitrogen ﬁxer. On deep soils or soils
with little impediment to root growth, P. juliﬂora probably responds rapidly to

disturbance.

Of the three associated legumes, C. c0n’an’a,A. famesiana and P. juliﬂora, C.
coriaria probably represents a "natural" component of the original forest. On the
other hand, the literature suggests A. famesiana may be an invasive species. It is
not clear whether or not P. juliﬂora was an original component, but it is an
important species in dry forest throughout the island. Regardless of their

respective origins, all three legumes appear to have attributes which would allow

144
them to respond positively after a disturbance, with soil characteristics being a

possible limitation.

An important component of Group Seven, Pithecellobium circinale is a low
shrubby tree with many small thorny stems. It’s wood is highly dense and it is
slow growing. The average height indicates its place is usually in the lower
portion of the canopy. A legume in the family Mimosaceae (Fable 3), P. circinale
would appear to be a "classic" weed species which might dominate sites with poor
site conditions and/or areas where cutting has resulted in site degradation. A
species with similar physical characteristics, P. unguis-cati is listed by Asprey and
Robbins (1953) in a number of forest types of Jamaica, including "strand-scrub" a
forest growing on sand beaches along the coast, as well as in "thorn scrub" on
alluvium soils farther in from the coast. P. circinale lacked a strong orientation in
the correspondence analyses, which suggests that this species is also a component
of more than one forest type. It’s dominance in Group Seven may relate to

disturbance or to poor site conditions.

Site Characteristics

Based on site characteristics systematically noted for each of forty sites, slope
angle was observed to suggest the clearest relationship with the groupings of sites
based on species composition. Table 24 represents the results of tabulating slope
angle by cluster group. In this table, the seven Group Three sites observed all

occurred on slopes visually identiﬁed as "steep" or "very steep". While clearly a

145

Table 24 Results of tabulating, slope angle by cluster group for the forty original control sites in
the ISA-Mao silvicultural study. The order the rows is based on values along the ﬁrst axis of CDA
applied to 118 sites representing six core site groups and seven noncore site subgroups. One of
the sites tabulated as group "BG" was originally designated as "BD". One of the sites tabulated
with cluster group "G" was originally designated as "FG". Data was collected in March 1984.

Slight Moderate Very steep

Group Four 4 0 0

Subgroup D 1 0

Group Seven 2 0
Subgroup BG

Subgroup G

Group One

Group Five

Subgroup B

Group Six

Group Three

Totals

 

O

qualitative measurement, the designation of slope angle was objective and
consistent for the sites examined. In contrast to Group Three sites, Group Four
sites were all located on sites with "slight" slopes. More over, sites in Group
Seven and Subgroup BG were all on slopes designated either as "slight" or "none".
Three of the ﬁve sites in Subgroup D were also located on sites with one of these
two designations. In all these groups, the species P. brasiliensis is dominant. On
the other hand, three Group One sites, dominated by B. simaruba, were all
located on sites with "moderate" slopes. These relationships suggest a natural
continuum from shallow slopes to steep slopes, with species progressing from P.
brasiliensis through B. simaruba to A. scleroxyla. This relationship is consistent

with the apparent order of the groups based on both CA and CDA. Sites in

146

groups Five and Six and Subgroup E are less consistent. The species association
on these sites may not be related to site angle. The literature suggests C. coriaria,
A. famesiana and Prosopis juliﬂora are highly tolerant of drought conditions, as
long as root growth is unimpeded. Slope angle and soil depth are frequently
related, but other factors are also involved, such as slope position and length.

The information presented in Table 24 represents one possible factor related to
the positions of groups One, Three and Four along the primary axes developed
using the CA and CDA procedures. This relationship may be used in the

development of further studies in the ISA-Mao forest.

A related observation is based on the location of most of Group Three sites in
Block Two. Block Two consists of two sets of three 50 X 50 m plots, each laid
out along the lengths of two parallel and adjacent ridges. Scaled diagrams of each
experimental block are included in AppendixK. The two ridges of Block Two
represent the highest and most exposed areas included in the silvicultural study.

In Appendix L, diagrams are included which show the relative position of the sites
in each block, using proﬁle icons. Each proﬁle icon is labeled with the cluster
group designation and represents the basal area contributions of the sixteen

species within the respective site.

The location of most Group Three sites in Block Two suggests a relationship
between topographic position and the dominance by A. scleroxyla, E. caribaeum

and T. pallida. However, relationships suggested with respect to topographic

147

position are confounded because these sites are also remote from access by
adjacent communities. Since all three of Group Three’s characteristic species are
sources for products essential to daily life in the local communities and the area is
difﬁcult to access, Block Two sites are the most likely to represent areas which
have remained relatively undisturbed. Nevertheless, despite the unique
characteristics suggested by the positions of Group Three sites in Block Two, sites
from cluster groups One, Six and Seven are also represented in Block Two. Only
groups Four and Five are not present. Remoteness and slope position do not by
themselves appear to determine dominance by A. scleroxyla, E. caribaeum and T.
pallida. Differences in slope angle within the block exist and, as shown

previously, do suggest a relationship with species composition.

Conversely, the absence of sites representing groups Four and Five in Block Two
may suggest a relationship with slope position and/or slope angle in the
distribution of P. brasiliensis and Caesalpinia coriaria. C. coriaria is present on the
ridges, but not in the same combination of species which dominates Group Five
sites, which are located lower in the topography. The same is true for P.
brasiliensis. P. brasiliensis is found on the sites in Block Two, but none of the
compositions found on these sites represent the same combinations and relative
dominance found on Group Four sites located lower in the topography. Again,
these observations are based on limited information. Their greatest value is their
use in the development of further studies in the ISA-Mao forest. However, the

available data does suggest the primary ordination observed in CA and CDA is

148

related to a visible environmental gradient. This gradient should have visible

effects in terms of site productivity.

Overstory Structure

If the differences in species composition between sites are related to site histories
and/or site conditions, there are structural characteristics which should be evident
across sites within cluster groups. Based on literature cited previously, the better
sites would be expected to support more basal area. Such sites would also be
expected to have a higher canopy. Poorer sites would generally have lower
canopy heights and scrubbier vegetation, represented by a higher proportion of
trees with multiple stems (Beard 1944, Asprey and Robbins 1953, Loveless and
Asprey 1956). Disturbance often creates situations similar to the poorer sites,
reducing overall tree height, causing greater numbers of multiple-stems and
generally increasing the dominance of trees with smaller boles (T amayo 1963,
Holdridge 1967, Powell and Mercedes 1986, Kellman and Roulet 1990, Poynton

1990, Vora and Messerly 1990).

In Figure 23, six structural measures are represented using box plots to illustrate
the distributions of the values within each of eight cluster groups. The subgroups
D and B have been included because their positions relative to the core cluster
groups were relatively stable throughout CA and CDA procedures. The order of
the groups along the X-axis generally follows the ordination of these groups along

the primary axis in the CDA procedure applied to 13 groups represented by 118

 

 

 

 

 

6" 1. A
a $ 37‘ _ m.
a)
.
é é g @147
lady-1 r 26‘ " ,4:
$-
2 g 8 8.12“
as 55“ " a
234 ~ 7; -10- .
i i" 1 i
o a ._
:2: t. 33- ,. 3
‘3 °‘
F
g 2—1 1- 4‘ .
1-4 " <
1- r 24
0 l l [T T T T T o l T T T T T T T o l l I T T T T T
3 8 E 5 1 7 D 4 3 6 E 5 1 7 D 4 3 6 E 5 1 7 D 4
Cluster Groups Cluster Groups Cluster Groups
woo 1 1 L1 1 L4 1 mo 1 I I l I l L 1 w l I 11 1 1 l 1
s
2.25
70004 ~ 70004 - '
21004 ‘ t
6000-4 — 6000- ~
0 , §L75<
55000-4 4 5000— L I I
5 . 2 arso-
2 o ' E o I
.4000-« e 4000« » 5 12w
x g . '3
U! I!
nonco-J
ﬁaooo- » '"3000« 4 g
(I)
Earls-1
20004 » 2000~ »
0.50-
it
1000 10004 F 054
0 0 0.“) T T T T T

 

 

 

363517D4

 

 

 

 

 

 

 

 

3GE5I7D4

 

 

 

 

 

 

362517D

as

 

 

Cluster Groups Cluster Groups Cluster Groups

Figure 23 Box plots of structural characteristics, by cluster group. The horizontal line within
each box represents the median, splitting the ordered values in half. The upper and lower edges
of the boxes split the upper and lower halves, respectively, in half again. The box therefore
represents the range of 50% of the values. The upper and lower edges of the boxes are referred
to as the upper and lower hinges. The lines extending vertically from the upper and lower hinges
extend to the last value(s) lying within one-and-a-half times the range described by the box. Stars
represent points more than one-and-a-half, but less than or equal to three times the range
described by the box. Circles represent values more than three times the range of the box away
from the upper and lower hinges (Wilkinson 1988). Cluster group three represents 18 values,
group B represents 7, group six, 5 values, group ﬁve represents 8, group one represents 11, group
seven, 14, group D, 14 and group four represents 11 values. The order of the groups is based on
the ﬁrst principle axis of a canonical discriminant analysis using basal area contributions of sixteen
species on 118 sites representing thirteen cluster groups.

 

150

sites. However, the order of groups Six and B have been reversed, because
species characteristics of these three sites suggest that group E sites may represent
site characteristics intermediate between groups Five and Six. The six structural
parameters were tested using the Kruskal-Wallis distribution free test of
differences between rank means of the respective cluster groups, based on the null

hypothesis:

Ho: cluster Group One= cluster Group Three...= cluster group B.

Sites were assumed to be independent. For all parameters except trees per ha,
there were signiﬁcant differences between at least two groups (Table 25). The
tabulated statistic assumes a chi-square distribution, with seven degrees of
freedom. No statistical tests were used to separate signiﬁcant differences between
mean ranks. The means of the original variables, along with rank means are
listed in Table 26. Means and rank means generally suggest the same
relationships among the eight cluster groups examined. Where the two estimates
suggest different relationships, mean ranks is given more emphasis than means for

the original values.

The strongest relationship recognized in the ordinal procedures was the contrast
between Group Three sites, dominated by A. scleroxyla and E. caribaeum, and
Group Four sites, dominated by P. brasiliensis. It was suggested that Group Three

sites represent areas where disturbance has been minimal and where site

 

151

Table 25 Kruskal-Wallis test statistics for six structural parameters from eight cluster groups
(Wilkinson 1989). The data comes from 88 of 120 sites in the silvicultural study at ISA-Mao.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Structural Parameter N D.F. Test Probability
Statistic

Mean height 88 7 19.14 0.008
Mean diameter 88 7 27.14 0.000
Total basal area 88 7 20.83 0.004
Stems per hectare 88 7 20.97 0.004
Trees per hectare 88 7 11.33 0.125
Mean stems per tree 88 7 25.70 0.001

Table 26 Mean values and rank means for structural characteristics within cluster groups and

across all sites.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

‘ Mean Height Mean DKH Mean baSal Mean stems Mean trees Mean stems

area per ha per ha per tree
N Mea- Rank Men Rank Men Rank MOI- Rank Mon Rank Mon Rank
(") mean 1"”) mean (:1) man man man man
Total 88 . 4.5 44.5 5.8 44.5 9.0 44.5 3082 44.5 2292 44.5 1.37 44.5
Group 3 18 4.3 31.0 5.7 41.7 7.7 35.9 2728 43.0 2294 50.9 1.20 28.8
Group 6 5 4.5 44.1 5.0 23.0 6.3 24.4 2700 39.1 1700 30.0 1.55 62.4
Group E 7 4.5 . 45.9 5.3 31.7 6.3 22.6 2514 34.7 1771 30.6 1.42 58.1
Group 5 8 4.2 23.4 5.6 35.2 9.0 38.1 2875 40.5 2062 39.6 1442 55.4
Group 1 11 4.7 55.4 7.0 68.8 12.3 65.5 2845 43.5 2209 42.5 1.36 37.4
Group 7 14 4.5 43.7 5.3 29.0 9.2 50.6 4464 72.1 2829 60.4 1.65 65.6
Group D 14 4.7 55.7 6.2 51.1 9.3 46.8 2671 36.3 2078 40.7 1.31 37.2
Group 4 11 4.3 57.0 6.6 60.7 10.5 54.5 2413 34.9 1955 39.5 1.2.4 35.1

 

 

 

/

152

conditions are relatively drier than on other sites. Group Four sites, on the other
hand, are represented as sites where favorable conditions for growth resulted in
exceptional dominance by P. brasiliensis. In Figure 23, the distribution of height
values for Group Three is consistently lower than the distribution of values for
Group Four. Total basal area is also generally lower for Group Three than for
Group Four. On average, Group Four carries 2.8 1112 ha'1 more basal area than
Group Three. Since total biomass is a function of basal area and height, these
values suggest that Group Three sites carry considerably less biomass than sites
representing Group Four. This relationship would be expected if Group Three
sites represent areas with less favorable conditions for growth than Group Four

sites.

The limited information on the distribution of Phyllostylon brasiliensis indicates the
species is favored by disturbance. Sites dominated by P. brasiliensis would be
expected to have structural characteristics indicative of such disturbance. Less
basal area might be expected, as well as smaller diameters and more stems and
total trees per ha. In fact, Group Four sites tend to have larger trees and fewer
stems and trees per hectare than other sites. Cutting could also have resulted in
more trees with multiple-stems, but Group Four sites had the second lowest
number of stems per tree (Figure 23, Table 26). The structural characteristics of
Group Four are not consistent with the suggestion that dominance by P.
brasiliensis on these sites is indicative of disturbance. Differences between groups

Three and Four are consistent with the suggestion that the two groupings of sites

 

 

153

represent opposite ends of an environmental gradient which affects relative site

productivity.

The relationship between P. brasiliensis and Pithecellobium circinale on Group
Seven sites is suggestive of disturbance. Structural characteristics are consistent
with this interpretation. Group Seven sites have the second smallest diameters
(5.3 cm), the most stems per tree (1.65) and the most stems per ha (4464).
However, tree heights and total basal areas are intermediate between groups
Three and Four. Therefore, total biomass may also be intermediate. The
position of Group Seven sites in the ordinal procedures of the previous section
suggest Group Seven sites represent conditions intermediate between groups
Three and Four, but closest to Group Four. Relative biomass also places group
Seven sites between groups Three and Four, but closest to Group Four. Groups
Seven and Four may represent similar environmental conditions, with different
disturbance histories. Cutting could have been more thorough on Group Seven
sites resulting in more multiple stems, more stems per ha and smaller diameters.
Group Seven may also represent sites with relatively poor site conditions which
prevent P. brasiliensir from attaining the same dimensions as on Group Four sites.
The interaction between Pithecellobium circinale and Phyllostylon brasiliensis‘ on
these sites may also be important. The weedy characteristics of P. circinale may
indicate it as a ﬁerce competitor for site resources. Disturbance may have
allowed this species to dominate under particular site conditions and its removal

may allow other species to make more expedient use of site resources.

154

In the ordinal procedures, the positions of groups Five and Six and Subgroup B
were sometimes ambiguous with respect to the other groups. The structural
characteristic of these groups are also somewhat ambiguous. Comparing groups
Three and Five, Group Five sites have lower heights but more basal area. Site
conditions on Group Five sites may be the same or slightly better than on Group
Three sites. Mean heights and mean basal area values are about the same for
Group Six and Subgroup E. Heights for these two groups are somewhat higher
than for Group Three while basal areas are somewhat lower. Productivity for
both groups would be expected to be less than on Group Five sites and about the

same as Group Three.

Closely related based on species compositions, groups Five and Six and Subgroup
E are not as closely related based on structural characteristics. Mean stems per
tree are very similar among all three. However, Group Five has lower mean
height values. Group Six, on the other hand, has lower mean diameters (DKH).
Group Five has the most basal area and the most stems and trees. Of these three
groups, Group Six is most clearly suggestive of disturbance, with smaller stems,
less basal area and more multiple stems. However, differences among the three
groups with respect to values for height, diameter and stems per tree are small
relative to groups One and Four. Their values are also similar to Group Seven
sites. Disturbance may have played an important role in the species distributions

of all four groups.

155

Based on structural characteristics, the relationships between Group One and the
other groups are not immediately apparent. If groups One and Three both
represent drier site conditions, structural characteristics of the two groups would
be expected to be similar. However, values for height and total basal area
contrast sharply with Group Three sites. Values for Group One sites are not
generally indicative of disturbance, either. Mean heights, diameters and basal
area values are large, while mean stems per tree are relatively low. Overall,
Group One sites are most similar to Group Four sites. These relationships are
also consistent with the relative characteristics of the dominant species, Bursera
simaruba which tended to be tall with large diameters and single stemmed. Both
B. simaruba and Phyllostylon brasiliensis may represent residual components of
former forest. Both species have relatively fast growth rates and may have
increased their relative dominance as competition was removed. However, the
speciﬁc gravity for P. brasiliensir is more than three times greater than for B.
simaruba. The actual biomass indicated by basal area and height values is
therefore proportionally greater for Group Four sites than for Group One sites.

Relative site productivity is expected to be greater for Group Four.

Summary

Information on the ecological relationships of A. scleroxyla and T. pallida is very
limited. The prevalent use of A. scleroxyla, T. pallida and E. can'baeum for
charcoal production suggests that their dominance in Group Three sites is

indicative of relatively undisturbed sites. The position of most Group Three sites

156

in a remote area of the forest is consistent with this interpretation. At the same
time, the extreme slope angles and slope positions represented by these sites
suggest particularly arid conditions. Mean heights tend to be low on these sites,
as do values for total basal areas. Group Three sites are expected to have less
productive site conditions than all other groups, representing the least productive

end of the primary environmental gradient suggested by CA and CDA.

Implications of the known slope angles for some of the Group Four sites as well
as the absence of Group Four sites on the ridges represented in Block Two
suggest that dominance by P. brasiliensis is favored by shallow slopes and a
position low in the topography of the ISA-Mao forest. The literature indicates
that P. brasiliensis is indicative of disturbance. However, structural characteristics
of Group Four sites indicate taller trees with larger diameters and greater total
basal area. P. brasiliensis in Group Four sites may represent a residual component
of former forest. Group Four sites are expected to have better site conditions
than all of the groups, representing the most productive end of the primary
environmental gradient suggested by CA and CDA. Subgroup D sites would be

expected to have similar growth characteristics.

Structural characteristics of Group Seven do indicate disturbance. More over,
both of the dominant characteristic species- Phyllostylon brasiliensis' and
Pithecellobium circinale- may be indicative of disturbed sites. Small in stature,

thorny and composed of many small stems P. circz'nale is particularly suggestive of

157

a species tolerant of harsh conditions. Present in smaller quantities in a number
of other kinds of sites, excessive disturbance on Group Seven sites may have
resulted in particular dominance by P. circinale. The number of small diameter
stems on Group Seven sites may also indicate areas where disturbance favored
regeneration of P. brasiliensir. The presence of B. simaruba as one of the minor
characteristic species may indicate these sites have soil limitations. Total basal
areas are intermediate between Group Three and Group Four and productivity is

also expected to be intermediate.

Like Group Seven, structural characteristics of groups Five and Six and Subgroup
E suggest disturbance. Mean heights are low and the number of multiple stems
is high. Caesalpinia coriaria, A. famesiana and Prosopis julrﬂora are the three
species consistently associated with these groups. Of the three, the literature
indicates A. famesiana and P. juliﬂora are associated with disturbance. All three
tend to have multiple stems and can be relatively fast growing. The literature also
suggests they do best on sites with deep soils with no limitations for root growth.
These sites may be disturbed areas of the forest with deep soils. Group Five sites
have larger mean diameters than Group Six or Subgroup E. On these sites, C.
coriaria may represent a residual component from the original forest. Group Five
sites may be less disturbed than other areas. Based on low values for total basal
area and small mean heights, productivity is expected to be less than groups Four

and Seven, but higher than Group Three.

158

Of the two species characteristic of Group One, observations made in studies of
the ISA-Mao forest suggest that Bursera simaruba is a residual component,
dominating locally because other species were removed for charcoal. The other
species characteristic of Groups One, E. can'baeum, may have been one of the
these formerly dominant species. E. can'baeum has many local uses, including
ﬁrewood and charcoal. Other studies suggest that both B. simaruba and E.
can'baeum are tolerant of rocky, shallow soils which are highly alkaline. This
suggests a contrast with groups Five and Six and Subgroup E which are dominated
by species which may do best on soils without impediments to root growth. Slope
angle relationships suggest that Group One sites are located on moderately steep
slopes, which may represent relatively shallow soils. Structural characteristics of
Group One are similar to Group Four, consisting of sites with tall trees, relatively
large diameters and high total basal areas. These characteristics indicate growth
conditions would also be similar to Group Four. However, B. simaruba’s low
speciﬁc gravity indicates the biomass on Group One sites may not be particularly
high, relative to Group Four sites. Group One sites are expected to have poor

growth, intermediate between groups Three and Four and less than Group Seven.

Summary

Several different types of data have been used to examine differences between
cluster groups. Based on analysis of the structural data and slope characteristics,
the contrast between groups Three and Four is interpreted to represent

differences in productivity. The structural characteristics of groups Five, Six and

159

Seven and Subgroup E are suggestive of disturbance. In these groups, species
composition may be the result of an interaction between disturbance and
environmental characteristics. Structural characteristics would indicate that groups
One and Four should have similar levels of growth. However, the physiological
characteristics of Bursera simaruba may be more indicative of drought adaptations
than of high leVels of productivity. Slope characteristics and species relationships

suggest that Group One sites are intermediate between groups Three and Four.

The results of CA and CDA suggest a contrast between Group One sites on the
one hand and groups Five, Six and E on the other. The underlying factor or
factors which explain this relationship cannot be determined directly from the
available data. However, the available literature indicates the contrast may be
related to soil characteristics. The species dominating Group One may be
tolerant of shallow, rocky soils, while the species dominating groups Five and Six
and Subgroup E may do best on deep soils without impediments to root growth.
Given this relationship, the prediction would be that growth would be higher on

the sites without root impediments.

160
Growth and Mortality Within Cluster Groups

Measures of site productivity were used to examine differences in relative
productivity between cluster groups. These measures of growth and mortality
provided the ﬁnal test for the scenarios presented based on ordinal procedures
and site data. Group Three sites should be expected to show the lowest levels of
growth, while sites dominated by Phyllostylon brasiliensis' should be expected to
show the highest levels. Both groups One and Five should show intermediate
levels, which are closer to level represented by Group Three sites, if they do in
fact represent areas with site characteristics which are more like those represented
by Group Three. Group Seven sites should be expected to have a relatively better
growth response than groups One, Three and Five if in fact Group Seven sites

represent conditions similar to Group Four sites.

Assuming independence among the sites, the relationship between cluster group
designation and growth and mortality was tested using the Kruskal-Wallis
distribution free test of differences between rank means. In addition to the seven
core site cluster groups, subgroups D and B were included to examine the
relationship of sites theoretically representing intermediate areas along the

species-site continuum. The groups were tested assuming the null hypothesis:

Ho: Group Three= Group Six= group E= Group Five, etc.

161

Table 27 lists the rank means for each parameters and the test statistic (H-
statistic). An alpha level of 0.1 was used to test for signiﬁcant differences. The
equation for the H-statistic comes from Hollander and Wolfe (1973). See

Appendix E for details of the test assumptions and the general equations.

Except for stem mortality, all of the parameters tested had at least two cluster
groups with signiﬁcantly different mean ranks. For the ﬁve parameters showing
signiﬁcant differences, a distribution-free test of differences between treatments
was applied. Not all differences were of interest, so only fourteen contrasts were

examined. The equation used assumes that all groups come from the same

Table 27 Rank means and calculated test statistics for growth and mortality parameters, using the
Kruskal-Wallis distribution free test of differences between rank means. Eight cluster groups are
tested for significant differences. The H-Statistic is the calculated test statistic to be compared
with a Chi-square distribution. The alpha level was set at 0.1, with a Chi-square statistic of 12.02,
assuming seven degrees of freedom (k-l, where k= the number of groups being tested).

   
 
   
 

 

31.8
51.2
57.9
38.3
39.6

 

     
        
   
    
   

44.0

 

66.3
48.1
41.2
49.4 41.9 40.1 14
44.7 37.9 43.6 51.1 14 41.1
Four 11 62.2 62.2 63.3 442 35.0 10 624

Ummmmm—‘

51.9 9
36.3 ‘
37.3
54.2

     

 

     

8
10

     

 

     

   

 

   

       
    

  

 

     

    

      

               

 

      

 

 

 

 

 

 

 

 

 

162

population (i.e. the null hypothesis) and is therefore conservative (Hollander and
Wolfe 1973). Although basal area mortality was found to be signiﬁcant, none of
the differences between cluster groups examined exceeded the test for least
signiﬁcant differences. The results for the other four parameters are presented in
Table 28. Figure 24 illustrates the means and standard deviations of the eight

cluster groups across the four growth and two mortality parameters.

Groups Three and Four are signiﬁcantly different across all four growth
parameters, with Group Four sites showing higher levels of growth than Group
Three, as expected based on previous analyses. These results support the
proposition that groups Three and Four represent opposite ends of a species-site
continuum which relates closely to fundamental differences in site factors
associated with site productivity. More over, groups One, Five and Seven fall in
between these two endpoints, which is also consistent with the model developed

based on ordinal procedures and site data analyses (Figure 24).

Although the differences between groups One, Five and Seven are not statistically
signiﬁcant, the relative patterns between these groups show that growth was
consistently higher in Group Seven sites than in sites of either Group One or
Group Five (Table 27). Although not conclusive, this generally supports the
association of Group Seven sites in proximity to Group Four sites in the ordinal

procedures. Conversely, sites representing groups One and Five were generally

 

 

163

Table 28 Distribution free multiple comparisons for growth and mortality parameters found to
have significant differences between treatment means using the Kruskal—Wallis distribution free
test. Fourteen comparisons were examined. The test statistic is the critical value calculated using
an alpha level of 0.1 with each test statistic based on the sample sizes of the treatments being
compared. Values in each cell are differences between the two treatments being compared.
Treatments with differences which are significant are underlined.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Treatment Test BARGRTH NET GRTH BARAVE DKHDIF
comparison Statistic

Three v/s E 32.5 -28.2 M -26.1 -22.0
Three v/s Five 31.0 -7.3 -20.5 -6.5 -6.4
Three v/s One 27.9 -6.6 -13.5 -7.8 -7.5
Three v/s Seven 26.0 -19.9 -21.7 -10.1 -24.4
Three v/s Four 27.9 ;3_l_.§ -_3_1.§ £5 $2.6
Five v/s E 39.0 -20.8 -18.2 -19.6 -15.6
Five v/s Seven 33.8 -12.6 -1.2 -3.6 -18.0
Five v/s Four 35.3 24.3 -14.1 -25.0 -26.2
One v/s E 35.3 -21.5 -25.1 -18.3 -14.6
One v/s Seven 29.4 -13.3 -8.2 -2.3 -16.9
One v/s Four 31.1 -25.0 -21.0 -23.7 -25.1
Seven v/s E 33.8 -8.2 -16.9 -16.0 2.4
Seven v/s Four 29.4 -11.7 -12.8 -21.4 -8.2
Four v/s E 35.3 3.5 -4.1 5.4 105

 

 

 

 

 

 

 

164

Figure 24 Growth and mortality response by cluster group in the ISA-Mao
subtropical dry forest. Symbols represent averages of 100 m2 sites within the same
cluster group as designated based on four cluster techniques using relative basal
area contributions of sixteen tree species. Bars represent plus and minus one
standard deviation. The order of the groups is based on the ﬁrst axis of a
canonical discriminant analysis using 118 sites which represented thirteen cluster
groups. Bars represent plus or minus one standard deviation. For Group Three,
n=18, for Group Six, n=5, for group E, n=7, for group 5, n=8, for Group One,
n=10, Group Seven, n=14, group D, n=14 and for Group Four, n=11, except for
the parameter, mean DKH increment. For this parameter, Group Four is
represented by only 10 sites. Growth estimates are based on diameter
measurements taken at knee height in 1986 and 1992.

165

 

 

 

 

 

 

 

0.04
415-1
.19.}
.15-
.go...
.35-

 

 

 

 

 

 

 

 

L F b P '51 F b
1 .I Q Li. ﬁt.
1 ill... r0
1 OT 17
l a L. 11
l 0 15
1 .1 LC, 1 1E
1 4 $ 16
1% 1 O 11. 13

 

 

 

$4.:me

an: .nx— Nav 5.5.3 sou- ~88

Cluster groups

Cluster groups

 

 

 

 

_ p B r F L L
1 .IT T4
1 Tllrml. 1D
1 Tlmll o?
.1 .r O t .11.
1 Turmll. 16
1 .r O 13
1 1 w 4 Is
1 Tuml. 13

 

 

 

mmwmwmmwm

A53 005 hun— uﬂUEenS ~35 Riel

 

 

 

 

 

 

 

 

 

_ P p p . p k r h
o i .14.
L .1! DAV} 1. ID
4 ill]. I?
l T w 4 Tel
1 TIT I5
1.1 AV 11 IE
m 1r 0 i 16
4 ii 13
a . a J . a «J _

Cluster groups

Cluster groups

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Ann .3.— may 5.33.33 on: .5

._____..t_»»9.
1 iT I4
1 0 L. 1D
1 RCK . 17
1 Q 11
O 6
L Tlmi 1.;
1 KC; 16
1 1 AV 13

4.......4..4.q
.u.....:.. . .1

Ali-359.3883

e . . _irret
1 Tim]. r4
1 1 0 1D
1 T||¢||l 17
1 .‘LLV “11
1 T w 5
1 T¢l1s
1 rll®l1 16
1 r93 - 13

4 a a a 4 44 . 4
wumuwuuumuw

Cluster groups

Cluster groups

Figure 24

166

less productive, which support their relative proximity to Group Three sites in the

ordinal procedures.

On the other hand, subgroups D and E did not perform as expected. Subgroup D
had lower mean values for basal area growth, net growth and mean DKH
increment versus both groups Four and Seven. This was not predicted based on
the close association of Subgroup D and Group Four sites. Based on species
composition alone, sites in Subgroup D would be expected to have good growth
rates if P. brasiliensir by itself indicates sites which are the most favorable for
growth. The species characteristics which eliminated subgroup D sites from the
Group Four core site classiﬁcation may be related to site characteristics which

distinguish Group Four and Subgroup D in terms of productivity.

Conversely, sites representing Subgroup B were considerably more productive
than expected based on their close association with Group Five sites. Subgroup E
sites had higher values for the four growth parameters than all other groups
except groups Four and Seven. Again, based on species composition alone, sites
in Subgroup B would be expected to have poor growth rates, if Caesalpinia
coriaria, A. famesiana and Prosopis juliﬂora indicate sites which are less favorable
for growth. On the other hand, if these species are indicative of disturbance, a
wide range of site qualities are possible, with a lack of root impediments being the

possible unifying site factor among groups Five and Six and Subgroup E.

 

 

167

A contrast between Group One versus the association of groups Five and Six and
Subgroup E might be expected given that B. simaruba and E. can‘baeum may be
indicative of rocky, shallow soils, while C. coriaria, A. famesiana and P. juliﬂora
may be indicative of deep soils. No such contrast is consistently apparent. Group
Five sites in particular consistently shows about the same level of growth as Group
One sites. However, as mentioned above, Subgroup E sites do not show a close
relationship with groups Five and Six. These sites do show greater growth and
lower mortality than Group One. Perhaps groups Five and Six represent areas
with deep soils, but other site characteristics result in growth conditions which are

roughly equivalent to those represented by Group One.

Summary

As expected based on the model of species-site relationships developed from
ordinal procedures and the information available regarding species, site conditions
and overstory structural characteristics, groups Three and Four contrast sharply in
terms of growth as measured over the six years of the silvicultural study.
Differences between these two groups are apparent in spite of effects related to
different levels of thinning. Also expected were the intermediate positions of
groups One, Five and Seven. Differences between these three groups were not
statistically signiﬁcant, but the trends across growth parameters indicate that
Group Seven sites were generally more productive than the other two groups.
This is also consistent with the model of species relationships based on the

primary axes of the ordinal procedures which consistently placed Group Seven

168

sites closest to Group Four sites. It is also consistent with structural
characteristics of these sites which indicated that Group Seven sites carried basal
area similar to Group Four sites. On the other hand, subgroups D and E did not
perform as expected. Subgroup D sites had considerably less growth than Group
Four sites, and subgroup E sites had considerably better growth than Group Five
sites. The growth responses of these two subgroups indicate a complex system

which needs more study.

169

Nonparametric Analysis of Thinning Effects

Assuming independence among the sites, the effect of cutting level on growth and
mortality was tested using the Kruskal-Wallis distribution-free test of differences
between rank means. The same parameters examined in the previous section
were tested for differences between ﬁve cutting levels. At a probability level of
0.1, average basal area increment, average diameter increment, stem mortality and
basal area mortality were found to differ signiﬁcantly between treatments (Table
29). For the ﬁve parameters found to have at least two treatments signiﬁcantly
different, rank mean differences were tested for signiﬁcance (Table 30). See
Appendix D for details of the test assumptions and the general equations. The

means and standard deviations of the original values are plotted in Figure 25.

Sites cut at the second level (IS-36%) had signiﬁcantly higher net growth
compared to control sites. These sites also had higher average diameter
increment compared to sites cut at the ﬁrst level (1-15 %). Sites cut at the second
level also had less stem and basal area mortality compared to control sites (Table
29, Figure 25). Sites cut at the third level (36-55%) showed greater average basal
area increment per tree and greater average diameter increment per tree
compared to control sites. These sites also had greater average diameter
increment per tree than sites out at the ﬁrst level (145%). Sites cut at level three
also had less stem mortality than control sites, although basal area mortality was

not signiﬁcantly different than sites cut at any other level (Table 29, Figure 25).

170

Table 29 Rank means and calculated test statistics for growth and mortality parameters, using the
Kruskal-Wallis distribution free test of differences between rank means. Five levels of cutting
were tested for significant differences: C= less than one percent of the basal area removed, 1= 1-
15%, 2= 15-36%, 3= 36-55%, 4= 55-72%. H-Statistic is the mlculated test statistic to be
compared with a Chi-square distribution. A probability of 0.1 was used, with a Chi-square statistic
of 7.779 assuming four degrees of freedom.

 

Table 30 Distribution free multiple comparisons for growth and mortality parameters found to
have significant differences between treatment means using the Kruskal-Wallis distribution free
test. See Table 29 for the treatments which were compared. The test statistic is the critical value
calculated using an alpha level of 0.1 with each test statistic based on the sample sizes of the
treatments being compared. Values in each cell are differences between the two treatments being
compared. Differences between treatments which are significant are underlined.

Treatment Test
Comparisons Statistic'

C—vs-l 24.2
C-vs-Z 23.2
Ova-3 22.3

C-vs-4

1-vs-2

 

 

1Total N=113. For the variables NETGRTH, BARAVE, STEMMORT and BARMORT

2Total N=112. For the variable DKHDIF.

 

171

Figure 25 Growth and mortality responses to thinning in the subtropical dry
forest of the Dominican Republic. Symbols represent averages of 100 1112 sample
plots within the same thinning level classiﬁcation. Thinning levels are: 0= less
than one percent of the basal area removed (n=32), 1= 1-15% (n=19), 2= 15-
36% (n=22), 3= 36-55% (n=25), 4= 55-72% (n=15), 5= greater than 72%
removal (n=5). Parameters are based on diameter measurements taken at 0.5 m
above ground level taken in 1986 and 1992. Bars represent plus or minus one
standard deviation.

172

 

 

 

o
1}

O

 

O C

 

 

0
z

 

— q M d - J1 u u
0 0 0

s 1 s 1 a s ._. a
As: you «Ev "3.5.3 «02

-l.0 4
-3.0

 

 

 

O

 

O

 

o
i

D
C

 

 

 

2.00

mmmmmwm

As: you NEV nukes» «use :38

1.75 4

Cutting levels

Cutting levels

 

 

 

 

 

 

 

 

 

 

 

 

 

 

» h P h — p p
1 .Iml. - 5
1 if 1 4
1 O L. 1 3
m .lol. - z
11 .lmi 1 .l
1 T|1®l 1 C

q — d d ﬂ « A
0 5 0 5 W 5 0 5 0
0 7. 5 2. 7 5. 2 0.
2 1 L l L 0 0 0 0
AEUV 00.5 hum «£0582: Eva 5.02

p b F p L p p p _
1 O . s
1 O . 4
1 a L. 1 3
1 O . 2
1 . O . 1
1 Tilroill. . c

A d J—I q u q _ _ u
m 8 6 4 0

II. II 1

wwa542

$58 00.3 h2— nukouu 4d ouuuo><

Cutting levels

Cutting levels

 

 

 

 

 

 

 

 

 

 

 

 

 

 

t e . tL t 4 eff. _ e r
1 i®|15
1 TIIII1®|1 1t.
1 T 0 13
1 7|1®i 12
1 krv 11
O 1. 1C
m T. u _ _ _ a d a u — ~ _
mmmmuzuzumzm.
As: no.— uEuuuv 5:795:— 53m
h — _ _ p b h p _
1 .Imwl15
1 1 HQ 114
1 imll 13
1 TIT 12
1 T AW 11
1 v O 1 1C
4 . e _ 4 q _ . _
w a w u u u .1... s m M M
An: aun— NEV 323185 suns imam

Cutting levels

Cutting levels

Figure 25

173
Although variability within treatments remained high after reassigning sites based
on actual basal area removed (Figure 25), trends of the mean values within
treatments suggest that in general, thinning succeeded in concentrating total basal
area growth on fewer stems. For example, although total basal area growth per
hectare was not signiﬁcantly different between treatments based on nonparametric
analyses, Figure 25 suggests that growth increased slightly on sites with 15-36% of
the basal area removed and was not less than the control sites until the second
highest rate of 55-72% removal. This implies that the growth potential of
moderately thinned sites tended to remain constant relative to unthinned sites.
Fewer stems, therefore, produced the same total growth. The same relationship is
apparent for the parameters BARAVE and DKHDIF. Average basal increment
and average diameter increment both tended to increase as thinning increased
through the cutting level of 36-55% (Table 29, Figure 25). Trees within thinned
sites were larger after six years than individuals in unthinned areas. The highest

levels of thinning appear to have caused growth to decline.

Although growth tended to decrease at the highest levels of basal area removal,
stem and basal area mortality tended to continue to decrease as thinning level
increased. This suggests that even at the highest levels of thinning, stem removal
eliminated competition which in the uncut sites resulted in higher mortality.
Assuming that basal area lost from 1986 to 1992 was part of the natural processes
of the dry forest, mortality may represent a harvestable quantity beyond the initial

biomass removal in the original thinning. If biomass after six years is not

174

signiﬁcantly less than uncut sites and average diameter is greater, it could be
considered that the silvicultural thinning was successful. Based on these criteria,
the cutting level of 15-36% basal area removal was the most successful. Sites cut
at this level had the highest net growth (Table 29, Figure 25). These sites were
essentially at equilibrium over the six years, with the same total biomass
accumulating on fewer stems. On the other hand, sites subjected to 36-55%
cutting appear to have had higher average basal area and diameter increments
(Figure 25). However, based on rank means, the differences are very small
between cutting levels two and three (Table 30). A moderate level of cutting

appears to stimulate better growth in the ISA-Mao subtropical dry forest.

Summary

Based on nonparametric analyses, thinning did affect both growth and mortality in
the forest. The growth response was generally positive for the lowest cutting
levels, and increasingly negative for the two highest levels. Mortality, on the other
hand, showed a linear decrease as cutting intensity increased. These results
suggest that competition between trees in the dry forest is signiﬁcant and growth
rates can therefore be manipulated. However, a great deal of variability is evident
in growth and mortality responses among sites within the same cutting level. This
variability may well be associated with the patterns of species compositions and

site characteristics observed in the data collected before thinning.

Summary

The goals of this study were to (1) determine whether patterns of species
composition existed among the sampled sites in the unthinned forest of a
silvicultural thinning study in the dry forest of the ISA-Mao Forestry Experimental
Station, (2) examine the implications of species distribution with respect to
disturbance history and underlying environmental gradients, and (3) explore the
relationship of growth and mortality with respect to species composition and
thinning level. An exploration of the data from the silvicultural study was
considered necessary because initial analyses of the growth data had revealed
extreme variability in the data, including among sites within the same thinning

treatment.

Initial analyses focused on 45 sites representing a subset of the 120 sites in the
silvicultural study. Basal area contributions of sixteen dry forest species
prominent in the study sites were used to classify sites into relatively
homogeneous groups, using a series of cluster analyses. Analyses of the entire
data set resulted in similar groupings. Six major groups were formed based on
analyses of the full data set. Numbered arbitrarily, Group One sites were found
to be associated with the characteristic species Bursera simaruba and Exostema
caribaeum. Group Three sites were found to have three characteristic species,
Acacia scleroxyla, E. caribaeum and Trichilia pallida. Group Four sites were

dominated by one species, Phyllostylon brasiliensis. Group Five was associated

175

176

with Caesalpinia coriaria and A. famesiana. Group Six was associated with A.
famesiana, C. coriaria and Prosopis julzﬂora. Group Seven was associated with P.
brasiliensz's, Pithecellobium circinale and B. simaruba. Of the 120 sites examined in
the second set of analyses, 67 were found to cluster consistently across four

different procedures. These 67 sites were termed "core" sites.

Applied ﬁrst to the core sites, then to all the sites, correspondence analysis (CA)
was used to test the results of the cluster procedures. All of the groups formed
from the cluster analyses were found repeated in CA. The species relationships
implicit in the cluster techniques were also found to compare well with the results
of CA. Commonly used as a tool for the ordination of sites, CA suggested a
strong contrast between sites in groups Three and Four. A secondary contrast
was suggested between Group One sites and sites representing groups Five and
Six. Based on the results of CA used in other studies, these contrasts may be

indicative of underlying environmental gradients.

Based on the results of the cluster analyses and the CA procedures, core sites
were considered to represent samples of their respective cluster groups. These
groups were submitted to canonical discriminant analysis (CDA) to examine their
ordinal relationships. However, Group Six had to be eliminated because of
apparent limitations related to sample size. Subsequently, the remaining ﬁve
groups were submitted to CDA. Based on this CDA, each group plotted in a

discrete portion of the graph and a contrast between groups Three and Four was

177

apparent. A second gradient was associated with the separation of Group One
from the other groups, but this relationship did not appear to be directly related

to the position of group Five.

Another CDA was applied to a data set partitioned into thirteen groups- six core
site cluster groups and seven subgroups represented by the noncore sites. Results
of this analysis suggested a continuum of groups, rather than a series of discrete
positions, as expected given that noncore sites represented intermediate species
compositions. Plotted along the ﬁrst two axes, a curvilinear relationship was
apparent. This effect is common for ordinal procedures applied to sites
representing diverse habitats. The ﬁrst axis suggested the same contrast between
groups Three and Four noted previously. In this CDA procedure, the third axis
was responsible for the separation of Group One ﬁ'om an association of groups
Five and Six and the Subgroup E. Subgroup E consisted of sites closely
associated with Group Five in the cluster analyses. The relative positions of the
core site groups based on three axes were very similar to the relative positions of

the groups based on three axes of CA applied to 67 sites.

To examine the implications of these ordinal relationships with respect to
disturbance history and underlying environmental gradients, several different types
of data examined. With respect to the contrast between groups Three and Four,
species information suggested that A. scleroxyla, E. can'baeum and T. pallida may

be indicative of relatively undisturbed sites, while Phyllostylon brasiliensir may be

178

indicative of disturbed areas. However, observations of slope position and lepe
angle suggest a strong contrast based on site conditions, not disturbance. More
over, the overstory structural data does not suggest that Group Four sites are
highly disturbed. Group Three sites appear to be indicative of ridgetop sites with
steep slopes which are relatively less productive. Group Four sites appear to
represent areas located lower in the topography, with relatively slight lepes which
are relatively more productive. As expected, growth and mortality parameters

showed that Group Three had less growth and higher mortality than Group Four.

The structural characteristics of groups Five, Six and Seven and Subgroup E were
suggestive of disturbance, which was consistent with the information available
about the species dominant in these groups. However, species composition might
also reﬂect an interaction between disturbance and environmental characteristics.
For example, the literature indicated C. coriaria, A. famesiana and P. julrﬂora are
favored on sites without impediments for deep root extension. Structural
characteristics suggested that groups Five and Six and Subgroup E represent less
productive site conditions than Group Seven. All four of these groups suggested
intermediate conditions between groups Three and Four. Growth and mortality
parameters indicated productivity for Group Seven did tend to be higher than for
groups Five and Six, but Subgroup E sites had relatively high growth rates and low
mortality, similar to Group Four.

179

The literature suggests that the species characteristic of Group One, B. simaruba
and E. caribaeum, are associated with rocky and shallow soils. Available
information on slope angles for Group One sites suggested an association with
moderately steep slopes, which may be indicative of shallow soils. Nevertheless,
structural characteristics of Group One suggested that groups One and Four had
similar site conditions. Both groups tended to carry relatively large amounts of
basal area and were dominated by tall trees with large diameters. However, the
physiological characteristics of Bursera simaruba may be more indicative of
drought adaptations than of high levels of productivity. B. simaruba has a very
low speciﬁc gravity and a large proportion of water in the green wood. Growth
and mortality parameters indicated relative productivity on Group One sites was

intermediate between groups Three and Four, and less than on Group Seven sites.

In the process of preparing the data for analysis, the target thinning levels were
found to differ considerably from the actual stems and basal area removed.
Therefore, nonparametric analyses were applied to growth and mortality
parameters to examine the effects of actual basal area removal. In these analyses,
thinning at a moderate level (IS-36% removal) was generally found to have a
positive effect on growth. Analysis of differences among the cluster groups did
net take into account the effect of different levels of thinning. It would be
reasonable to suggest that variability apparent within cluster groups may have
been associated with different levels of thinning and, conversely, variability within

cutting levels may have been associated with differences relating to species

180

composition and site conditions. Implications of the productivity gradient
revealed using multivariate analyses suggest that thinning would have a more
beneﬁcial effect on "good" sites, where competition between individual trees is
greatest. On "poor" sites, thinning would be predicted to have less effect, as
environmental stress may be a more important limitation than competition
between individual trees. The relatively high level of basal area and the extreme
stem density in Group Seven indicate these sites might show a particluarly positive

response to thinning.

Conclusions

Multivariate analyses had not previously been used to address questions of
species-site relationships in subtropical dry forest. In this study, a series of MVA
procedures Were able to illustrate fundamental structures of species composition
which were not previously understood. Limitations in available site data and in
the literature describing the ecologies of species dominant in the forest restricted
the interpretation of the structures apparent in the data. Nevertheless, the
available data clearly suggested a productivity gradient apparently related to

species composition and site characteristics.

Because virtually all of the dry forests of the world have been heavily affected by
human intervention (Murphy and Lugo 1986a), it is important to emphasize that
species relationships existed in the ISA-Mao forest in spite of disturbance,
although disturbance has effected species distributions in a number of ways.
Human intervention clearly makes analyses of forest dynamics more complicated.
However, using MVA procedures, patterns apparent in the data can be simpliﬁed
and, with additional information, these patterns can be understood and utilized to
develop models of forest productivity which incorporate factors related to human

disturbance as well as environmental gradients.

181

Recommendations

This study has resulted in many questions and ignored many others. Why such
high rates of mortality existed in the silvicultural study has not been addressed.
As mentioned previously, some of the mortality is related to cutting. Some
"mortality" may be related to measurement error. Also, some of the trees may
have been misplaced in the inventory. All of these factors would also impact
growth estimates. Much of this information is available in the inventories from
1987 to 1991. Along with new inventories using more precise measurement
techniques, studies of these other inventories would be a great addition to
understanding the real effects of thinning in the dry forest, as well as add to the
information about fundamental dynamics in the natural forest. In particular,
reliable estimates of growth in the subtropical dry forest are necessary for

determining sustainable harvest rates.

An important component of such studies would be the use of the biomass
equations developed by Maxwell (1985). To make use of Maxwell’s equations,
only the measurement at knee height is necessary. In future inventories, it might
be advisable to eliminate the measurements at breast height. Eliminating the
breast height measurement would save time, allowing for greater accuracy in
measuring the remaining parameters. For example, if the time limitation is
reduced, knee height measurements could be taken with a diameter tape rather

than calipers. Also, if possible, the diameters for each stem of trees with multiple

182

183

stems should be recorded together in subsequent inventories, so that the growth
and mortality of individual stems can be analyzed. Because cacti clearly dominate
some areas of the forest, some measure of cactus dominance should also be

included in future inventories.

If quantitative data on slope positions, angles, aspects, soil depths and soil
characteristics were available for each site, suggestions made in this study about
underlying gradients and species distributions could be tested. The relative
effectiveness of different levels of thinning could also be better understood. In
Appendix M, a brief proposal is included for the study of soil-site interactions in
the area of forest comprising the silvicultural study. Data from such a study
would not only clarify the relationships between core cluster groups, they could
potentially permit an integration of all the sites into one comprehensive system of
classiﬁcation which delineates the forest into relatively homogenous management
units. Such a classiﬁcation system would involve an iterative series of dynamic
procedures, where classiﬁcation leads to speciﬁc hypotheses, which result in a
better understanding of the dry forest dynamics, resulting in better classiﬁcations.
Multivariate methods represent an important tool in this process, helping to

simplify relationships between variables which appear exceptionally complex.

Finally, I would like to emphasize that the ISA-Mao forest represents one of the
few subtropical dry forests where systematic silvicultural techniques have been

initiated. Despite the length of this study, it barely scratches the surface of the

 

184

pool of data collected in the forest since the forest was donated to ISA in 1978.
Some of the studies initiated in the forest were concluded and have been analyzed
and summarized sufﬁciently by the professors and students in ISA’s department of
natural resources. Nevertheless, many of the data have never been thoroughly
analyzed. Although there are undoubtedly many limitations in some of the data
sets, the results of this study show that the available numbers have the potential
for illustrating fundamental relationships directly relevant to appropriate
management of the forest. Resources for an institution such as ISA are always
limited. This is a fact of institutional life. However, it is to be hoped that the
fundamental importance of the research at the ISA-Mao Forestry Experimental

Station is never underestimated due to lack of interest in an "unsexy" resource.

LITERATURE CITED

LITERATURE CITED

Arnon, I. 1992. Soils of the dry zones. pp 85-120. _Ir_1: Dry Lands: Principles and
Practice. Elsevier (Amsterdam).

Asprey, GR and R.G. Robbins. 1953. The vegetation of Jamaica. Ecological
Monographs, 23: 359-412.

van Auken, O.W. and J.K. Bush. 1991. Inﬂuence of shade and herbaceous
competition on the seedling growth of two woody species. Madroﬁo 38(3):
pp. 149-157.

Barnes, B.V. 1984. Forest ecosystem classiﬁcation and mapping in Baden-
Wurttemberg, West Germany. pp 49-65. 1&1 Proceedings of a Symposium
on Forest Land Classiﬁcation: Experience, Problems, Perspectives. J.G.
Bockheim (ed.). University of Wisconsin, Madison.

Barnes, B.V., KS. Pregitzer, T.A. Spies and V.H. Spooner. 1982. Ecological
forest site classiﬁcation. J. For. 80: 493-498.

Barth, R.C. and 1.0. Klemmedson. 1982. Amount and distribution of dry matter,
nitrogen, and organic carbon in soil-plant systems of mesquite and palo
verde. Journal of Range Management 35(4): 412-418.

Beard, 1.8. 1944. Climax vegetation in tropical America. Ecology, 25(2): 127-158.

Beard, J .S. 1953. The savanna vegetation of northern tropical America. Ecological
Monographs, 23: 149-215.

Betances, M. A. 1983. Gravedad especiﬁca de las maderas de 19 especies del
bosques seco en la Republica Dominicana. Undergraduate thesis, Instituto
Superior de Agricultura, Santiago, Dominican Republic. 89 pp.

Britton, N.L. and IN. Rose. 1963. The Cactaceae: descriptions and illustrations
of plants of the cactus family. pp 85-93. Dover Publications, Inc., New York.

Burgos M., J.F., F.N. Diloné G. and JR. Mercedes U. 1986. Malezas en el bosque
seco Mao, Valverde, Republica Dominicana. ISA-Nota Técnica No. 22.
Instituto Superior de Agricultura, Santiago, Dominican Republic. 14 pp.

Buskirk, RE. and G.W. Otis. 1994. Annual waxy bands on a Costa Rican Cactus.
Biotropica 26(2): 229-232.

Cajander, AK. 1926. The theory of forest types. Acta Forest. Fenn. 29: 1-108.

185

186

Carmean, W.H. 1965. Black oak site quality in relation to soil and topography in
southeastern Ohio. Proc. Soil Sci. Soc. Amer. 29: 308-312.

Carmean, W.H. 1975. Forest site quality evaluation in the United States. Adv.
Agron. 27: 209-269.

Checo, Humberto, and C. W. Ramm. 1993. Diagramas climaticos de la Estacion
Experimental Forestal ISA- Mao. Unpublished.

Checo, Humberto. Personal Communications. Departmento de Recursos
Naturales, Instituto Superior de Agricultura, Santiago, Republica
Dominicana.

Christensen, RC. 1992. A management model developed for a natural forest in
Burkina Faso. pp 101-108. _Ir_1_: Management of natural forests in the semi-
arid areas of Africa: present knowledge and research needs. J. Fries (ed.).
Swedish University of Agricultural Sciences: International Rural
Development Centre, Uppsala, Sweden.

Coile, TS. 1952. Soil and the growth of forests. Adv. Agron. 4: 329-399.

CRIES (Comprehensive Resource Inventory and Evaluation System). December
1977. Land Resource Base Report No. 77-1. Michigan State University.
AID Contract No. AID/ERS/MSU PASA #A6fFAB-263-14-76.

Cuevas, C., and M. Hernandez. 1987. Identiﬁcacién de productos y subproductos
del bosque seco en Mao Valverde, Republica Dominicana. Undergraduate
thesis, Instituto Superior de Agricultura, Santiago, Dominican Republic. 72

PP.

Digby, P.G.N. and RA. Kempton. 1987. Multivariate Analysis of Ecological
Communities. Chapman and Hall, New York, NY. 206 pp.

Fisher, J .H. 1994. Pre-european settlement forests of northern lower Michigan:
the role of landform in determining composition across the landscape.
Master’s thesis, Michigan State University, East Lansing, Michigan. 101 pp.

Fries, J. 1992. Management of Natural Forests in' the semi-arid areas of Aﬁica:
Present Knowledge and Research Needs. Swedish University of Agricultural
Sciences: International Rural Development Centre, Uppsala, Sweden. 119pp.

Furley, RA. and W.W. Newey. 1979. Variations in plant communities with
topography over tropical limestone soils. Journal of Biogeography 6: 1—15.

187

Garcia, R.G. and N. Alba V. 1989. Estudio ecoﬂoristico comparativo del bosque
seco subtropical de Azua y Monte Cristi, Republica Dominicana.
Moscosoa 5:55-84.

Geilfus, Frans. 1989. El arbol al servicio del agricultor: manual de agroforestéria
para el desarrollo rural, Vol 2: Guia de especies. Enda-Caribe and
CATIE, Santo Domingo. 778 pp.

Goldstein, RA. and DP. Grigal. 1972. Deﬁnition of vegetation structure by
canonical analysis. J. Ecol. 60: 277-284.

Greenacre, M.J. 1993. Correspondence analysis in practice. Academic Press Inc.,
San Diego, CA. 195 pp.

Grigal, DP. and RA. Goldstein. 1971. An integrated ordination-classiﬁcation
analysis of an intensely sampled oak-hickory forest. J. of Ecology, 59: 481-
492.

Hardcastle, PD. 1992. Some thoughts and ideas on savanna woodlands in
Southern Africa. pp 69-74. In: Management of Natural Forests in the semi-
arid areas of Africa: Present Knowledge and Research Needs. J. Fries (ed.).
Swedish University of Agricultural Sciences: International Rural
Development Centre, Uppsala, Sweden.

Heermans, J .G. 1992. The Guesselbodi experiment: bushland management in
Niger. pp 87-99. In; Management of Natural Forests in the semi-arid areas
of Aﬁica: Present Knowledge and Research Needs. J. Fries (ed.). Swedish
University of Agricultural Sciences: International Rural Development
Centre, Uppsala, Sweden.

Hernandez, J. M. 1986. Comparacién del comportamiento del bosque seco nativo
a tres tratamientos culturales en dos afros de investigaciones, Mao,
Valverde, R.D. ISA-Nota Técnica No. 23. Instituto Superior de
Agricultura, Santiago, Dominican Republic. 9 pp.

Holdridge, LR. 1945. A brief sketch of the ﬂora of Hispaniola. pp 76-78. Q:
Plants and plant science in Latin America. Chronica Botanica Company,
Waltham, Massachussettes.

Holdridge, LR. 1967. Life Zone Ecology. Tropical Science Center, San Jose,
Costa Rica. 206 pp.

Hollander, M. and DA. Wolfe. 1973. Nonparametric statistical methods. Wiley
and Sons, Inc. New York, NY. 503 pp.

188

Hunter, LR. and J .L. Steward. 1993. Foliar nutrient and nutritive content of
Central American multipurpose tree species growing at Comayagua,
Honduras. Commonwealth Forestry Review 72(3): 193-197.

Jacobs, M. 1965. The genus Cappanlr (Capparaceae) from the Indus to the Paciﬁc.
Blumea 12(3): 385-541.

James, EC. and CE. McCulloch. 1990. Multivariate analysis in ecology and
systematics: Panacea or Pandora’s Box? Annu. Rev. Ecol. Syst. 21: 129-
166.

Johnson, MB. 1992. The genus Bursera (Burseraceae) in Sonora, Mexico and
Arizona, USA. Desert Plants 10(3): 126-144.

Kellman, M. and N. Roulet. 1990. Nutrient ﬂux and retention in a tropical sand-
dune succession. J. Ecology 78(3): 664-676.

Kercher, J .R. and RA. Goldstein. 1977. Analysis of an East Tennessee oak
hickory forest by Canonical Correlation of species and environmental
parameters. Vegetatio 35(3): 153-163.

Knudson, D.M., W.R. Chaney and FA. Reynoso. 1988. Fuelwood and Charcoal
Research in the Dominican Republic. Purdue University. Department of
Forestry and Natural Resources. 181 pp.

Lamprecht, H. 1989. Silviculture in the Tropics. Deutsche Gesellschaft fiir
Technische Zusammenarbeit (GTZ) GmbH. Technical Cooperation-
Federal Republic of Germany. Eschborn 1989.

Laureano, E]. 1991. A resource policy analysis for the forestry sector of the
Dominican Republic. Ph.D. Dissertation, Michigan State University, East
Lansing, Michigan. 198 pp.

Lees, 8.6., E]. Russell, R.L. Bingham and P. Felker. 1992. Discovery of
thornless, non-browsed, erect tropical Prosopir in 3-year-old Haitian progeny
trials. Forest Ecology and Management 48(1-2): 1-13.

Little, EL. and EH. Wadsworth. 1964. Common trees of Puerto Rico and the
Virgin Islands. USDA, Forest Service, Washington, DC. 556 pp.

Little, E.L., R.O. Woodbury and EH. Wadsworth. 1974. Trees of Puerto Rico and

the Virgin Islands (Volume 2). USDA, Forest Service, Washington, DC.
1024 pp.

189

Loveless, AR. and G.G. Asprey. 1956. The dry evergreen formations of Jamaica.
J. of Ecology, 44: 799-822.

Lugo, A.E., J .A. Gonzalez-Liboy, B. Cintron and K. Dugger. 1978. Structure,
productivity and transpiration of a subtropical dry forest in Puerto Rico.
Biotropica 10: 278-291.

Lungren, C. 1992. The Nazinga game ranch pilot research in sustainable
development. pp 109-119. E: Management of Natural Forests in the semi-
arid areas of Africa: Present Knowledge and Research Needs. J. Fries
(ed.). Swedish University of Agricultural Sciences: International Rural
Development Centre, Uppsala, Sweden.

Markham, RH. and A.J. Babbedge. 1979. Soil and vegetation catenas on the
forest-savanna boundary in Ghana. Biotropica 11(3): 224-234.

Maxwell, D.G. 1985. Biomass, moisture contents and energy equivalents of tree
species in the subtropical dry forest of the Dominican Republic. Master’s
thesis, Ohio State University, Columbus, Ohio. 88 pp.

von Maydell, H]. 1992. The importance of agroforestry and of secondary forest
products in the management of natural forests in semiarid Aﬁica. pp 77-86.
In: Management of Natural Forests in the semi-arid areas of Africa:
Present Knowledge and Research Needs. J. Fries (ed.). Swedish University
of Agricultural Sciences: International Rural Development Centre, Uppsala,
Sweden.

McPherson, GR. 1992. Ecology of oak woodlands in Arizona. pp. 24-33. In:
Ecology and Management of Oak and Associated Woodlands: Perspectives
in the Southwestern United States and Northern Mexico. USDA Forest
Service, General Technical Report RM-218. Fort Collins, Colorado.

Mokma, D. Personal Communication. Department of Soil Science, Michigan
State University, East Lansing, Michigan.

Montero, V., D.M. Knudson and F. Reynoso. 1984. Metodologia de investigacion
del programa de desarrollo de madera como cumbustible. ISA-Nota Tecnica
No. 1. Instituto Superior de Agricultura (ISA), Santiago, Dominican
Republic. 21 pp.

Murphy, P.G. Personal Communication. Department of Botany and Plant
Pathology, Michigan State University, East Lansing, Michigan.

Murphy, P.G. and A.E. Lugo. 1986a. Ecology of tropical dry forest.
Annual Review of Ecology and Systematics 17: 67-88.

190

Murphy, P.G., and A.E. Lugo. 1986b. Structure and biomass of a subtropical dry
forest in Puerto Rico. Biotropica, 18: 89-96.

Otis, G.W. and RE. Buskirk. 1986. Annual growth bands in a tropical dry forest
cactus. Cactus and Succulent Journal 58(1): 25-29.

Ott, L. 1988. An Introduction to Statistical Methods and Data Analysis. PWS-
Kent, Boston. 835 pp.

van Paasen, M. 1986. Guia para especies arbéreas y arbustivas del bosque seco en
la Republica Dominicana. Instituto superior de Agricultura, Santiago,
Dominican Republic. 234 pp.

Padley, EA. 1989. Associations among glacial landforms, soils, and vegetation in
northeastern lower Michigan. Ph.D. dissertation, Michigan State
University, East Lansing, Michigan. 279 pp.

Peacock, J .T. and C. McMillan. 1968. The photoperiodic response of American
Prosopis and Acacia from a broad latitudinal distribution. American J. of
Botany 55(2): 153-159.

Pielou, EC. 1969. An introduction to mathematical ecology. Wiley-Interscience,
New York, NY. 286 pp.

Powell, M. and J. Mercedes. 1986. Composicién, estructura y estratiﬁcacién de un
bosque seco subtropical en Mao, Valverde, Republica Dominicana. ISA-
Nota Técnica No. 25. Instituto Superior de Agricultura, Santiago,
Dominican Republic. 22 pp.

Poynton, R.J. 1990. The genus Prosoprlr in southern Africa. South Africa Forestry
Journal No. 102: 62-66.

Pregitzer, K.S., B.V. Barnes, and GD. Lemne. 1983. Relationship of topography
to soils and vegetation in an upper Michigan ecosystem. Soil Sci. Soc. Am.
J., 47: 117-123.

Pregitzer, KS, and B.V. Barnes. 1984. Classiﬁcation and comparison of upland
hardwood and conifer ecosystems of the Cyrus H. McCormick Experimental
Forest, upper Michigan. Can. J. For. Res., 14: 362-375.

Pritchett, W.L. and RF. Fisher. 1987. Properties and Management of Forest Soils.
John Wiley and Sons, Inc. New York, NY. 493 pp.

191

Ralston, CW. 1964. Evaluation of forest site productivity. pp 171-201. Q: Int.
Rev. For. Res. Vol 1. Romeberger and Mikola (eds.). Academic Press,
New York: 171-201.

Record, SJ. 1944. Miscellaneous notes on tropical American woods. Tropical
Woods 79: 105.

Rogers, GK. 1987. The genera of Cinchonoideae (Rubiaceae) in the southeastern
USA. Journal of the Arnold Arboretum 68(2): 137-183.

Rowe, J .S. 1984. Forestland classiﬁcation: limitations of the use of vegetation. pp.
132-147. In: Proceedings of a Symposium on Forest Land Classiﬁcation:

Experience, Problems, Perspectives. J.G. Bockheim (ed.). University of
Wisconsin, Madison.

Ruskin, FR. 1980. Firewood crops. Shrub and tree species for energy
production (V. 1). BOST'ID Report No. 40. National Academy of Science,
Washington, DC. 92 pp.

SAS Institute Inc. 1985. SAS User’s Guide: Statistics, Version 5 Edition. Cary,
NC. SAS Institute Inc. 956 pp.

Scott, RE. and RR Martin. 1984. Avian consumers of Barrera, Ficus, and
Ehretia fruit in Yucatan. Biotropica 16(4): 319-323.

Stevens, GO 1987. Lianas as structural parasites: the Bursera simaruba example.
Ecology-USA 68(1): 77-81.

Tamayo, F. 1963. Los bosques xeroﬁlos de Maracaibo. Boletin, Sociedad
Venezolano de Ciencias Naturales, Caracas 23(103): 294-299.

Thomas, DA, and V.R. Squires. 1991. Available soil moisture as a basis for land
capability assessment in semiarid regions. Vegetatio, 91: 183-189.

Vora, RS. and J.F. Messerly. 1990. Changes in native vegetation following
different disturbances in the lower Rio Grande Valley, Texas. Texas
Journal of Science 42(2): 151-158.

Walter, H. 1985. Vegetation of the Earth. Springer-Verlag, Berlin. 318 pp.

Whitney, G.G. 1991. Relations of plant species to substrate, landscape position,
and aspect in north central Massachusetts. Can. J. For. Res., 21: 1245-1252.

Wilkinson, Leland. 1988. SYGRAPH. SYSTAT, Inc. Evanston, Illinois. 980 pp.

192

Wilkinson, Leland. 1989. SYSTAT: The system for statistics. SYSTAT, Inc.
Evanston, Illinois. 822 pp.

Yair, A. and M. Shachak. 1987. Studies in watershed ecology of an arid area. In
Progress in Desert Research. L. Berkofsky and MG. Wurtele (eds.).
Rowman and Littleﬁeld, Totowa, New Jersey: 145-193.

APPENDICES

 

APPENDIX A

Actual basal area and stems removed and assigned cutting levels

193

0
or
U)

\DQQGUIhNNI-I‘

ID

101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315

SUBPLOT

01
02
03
04
05
06
07
08
09
10
11
12
13
14
15
16
17
18
19
20

194

APPENDIX A

CLUSTOT
BG

BG

C}Q

m

~q-Jwtjﬁimtnurm~qcratjhiqs4u~qr-wtaurw'qurqriwtnb1054Htaurwtaurwtdurwtuurwtuc)m<n+-wrap)Utouunr‘Ulvﬁrbtjbtjuro

TRTMT CUTLEVEL

wmwuwwuwmunnwwnonmwbnnwnnwbmwhnno0HonHnnww~¢wnnwnphwwwwoobhbwHMHHnwwwwnnan

PERBARCT

0.... .00. O.
QWHGQOHQU‘IOOOUO

UlbU'Ih
“WPONQOU‘OOOHO

0

HHH
mo
m

69.2

sum

WNOOOOUOONO

.COOIIOUI I.
OOOWOODQGNQOOOOWOOHOO

U'IQO
How

9

PERTRKCT

000000

44

21.

21
12

54.

61
51
54
S6
54
65

65.

00000

\DNDWU‘ImmeGOOQQWQWNGDOOWOOQH\IOU'IOOOOUOOOOOH¢QO#OOOONOO\W

 

 

CBS

76
77
78
79
60
81
02
83
64
65
06
87
88
09
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120

ID

316
317
318
319
320
321
322
323
324
325
326
327
326
329
330
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430

sass sqoqouq Hm m
U a O 8 0000

11411110141111.1161
0

UIPQUIGQ’MQHUQbQUUQU

CUTIEVEL

HOOOOUUWU“MUMﬁHI-‘WNNNDUIWWU'OOHOHNNNNNN.NUDOPHOH

APPENDIX B

Structural characteristics of 120 silvicultural sites

 

 

196

197
APPENDIX B

OBS ID BLOCK SUBPLOT CLUSTOT ALTINI DKHDIF BARPRB TRUNKINI TREEINI AVESTEM

1 101 01 01 86 4.3 0.9 7.0 2700.0 2600.0 1.0
2 102 01 02 7 3.7 0.4 6.1 3300.0 2200.0 1.5
3 103 01 03 3G 4.3 0.8 14.9 3900.0 3700.0 1.1
4 104 01 04 D 4.7 0.3 5.9 2200.0 2100.0 1.0
5 105 01 05 3D 4.5 0.5 12.3 5800.0 5700.0 1.0
6 106 01 06 BD 4.8 0.7 11.6 5300.0 4000.0 1.3
7 107 01 07 C 4.3 1.3 6.4 2700.0 2200.0 1.2
8 108 01 08 36 5.2 0.9 16.2 4200.0 3900.0 1.1
9 109 01 09 4 5.4 0.9 18.0 2700.0 2200.0 1.2
10 110 01 10 5 4.7 0.9 11.6 5100.0 3500.0 1.5
11 111 01 11 D 5.1 0.7 18.6 2700.0 2300.0 1.2
12 112 01 12 4 4.9 0.4 10.8 2300.0 1700.0 1.4
13 113 01 13 D 5.0 0.7 12.4 2800.0 2700.0 1.0
14 114 01 14 D 5.2 0.5 11.3 1900.0 1600.0 1.2
15 115 01 15 FG 3.9 1.2 5.6 2400.0 1900.0 1.3
16 116 01 16 BC 4.5 0.8 10.2 3700.0 3100.0 1.2
17 117 01 17 D 4.7 1.0 12.8 4400.0 3900.0 1.1
18 118 01 18 1 4.7 0.6 17.9 3700.0 3500.0 1.1
19 119 01 19 5 4.0 0.7 6.6 3500.0 2600.0 1.3
20 120 01 20 3 4.3 0.6 9.2 4600.0 3700.0 1.2
21 121 01 21 E 4.9 1.4 5.6 2400.0 1700.0 1.4
22 122 01 22 D 5.1 0.8 10.3 2000.0 1400.0 1.4
23 123 01 23 EB 5.4 0.7 8.3 2900.0 2600.0 1.1
24 124 01 24 6 4.9 1.0 4.9 1100.0 1000.0 1.1
25 125 01 25 E 5.1 1.2 6.2 2400.0 1500.0 1.6
26 126 01 26 1 5.3 0.3 14.8 4600.0 4400.0 1.0
27 127 01 27 6 5.3 0.4 5.8 2500.0 1400.0 1.8
28 128 01 28 E 4.3 0.6 4.9 2500.0 1900.0 1.3
29 129 01 29 G 4.3 0.8 4.7 3300.0 2900.0 1.1
30 130 01 30 E 4.7 0.7 9.6 4300.0 2900.0 1.5
31 201 02 01 3 4.2 0.4 7.6 2800.0 2400.0 1.2
32 202 02 02 3 4.8 0.6 8.8 2500.0 2300.0 1.1
33 203 02 03 3 5.5 0.6 11.5 3000.0 2200.0 1.4
34 204 02 04 BC 4.5 0.7 8.7 3900.0 3300.0 1.2
35 205 02 05 3 4.6 0.9 9.5 3600.0 3000.0 1.2
36 206 02 06 E 3.9 0.5 4.9 1500.0 1100.0 1.4
37 207 02 07 3 3.5 0.8 8.5 1700.0 1600.0 1.1
38 208 02 08 3 4.4 0.4 8.0 2500.0 2200.0 1.1
39 209 02 O9 3 4.3 0.6 5.7 1900.0 1800.0 1.1
40 210 02 10 3 4.4 0.3 7.7 2700.0 2300.0 1.2
41 211 02 11 3 4.4 0.5 6.9 3000.0 2700.0 1.1
42 212 02 12 3 4.3 0.4 5.7 2000.0 1800.0 1.1
43 213 02 13 1 5.0 0.7 12.8 2000.0 1800.0 1.1
44 214 02 14 X 4.2 0.6 7.7 3000.0 2000.0 1.5
45 215 02 15 C 4.8 0.3 6.8 1900.0 1700.0 1.1
46 216 02 16 1 5.4 0.6 13.5 2300.0 2100.0 1.1
47 217 02 17 3 3.7 0.4 5.6 2300.0 1500.0 1.5
48 218 02 18 1 4.8 1.0 13.2 3300.0 1500.0 2.2
49 219 02 19 C 4.4 0.7 8.0 3500.0 2600.0 1.3
50 220 02 20 7 4.6 0.6 8.5 4100.0 3300.0 1.2
51 221 02 21 E 3.9 1.1 6.5 1800.0 1200.0 1.5
52 222 02 22 PG 4.7 0.5 4.3 2000.0 1300.0 1.5
53 223 02 23 3 4.3 0.5 6.7 2900.0 1800.0 1.6
54 224 02 24 3 4.3 0.5 7.8 3000.0 2800.0 1.1
55 225 02 25 3 4.5 0.6 10.3 2600.0 2600.0 1.0
56 226 02 26 3 4.1 0.3 4.6 2600.0 2200.0 1.2
57 227 02 27 1 4.1 0.6 8.5 3300.0 1600.0 2.1
58 228 02 28 7 4.5 1.3 3.9 2700.0 1200.0 2.2
59 229 02 29 3 ' 4.2 0.5 8.9 3800.0 3000.0 1.3
60 230 02 30 1 4.5 . 5.3 1300.0 1000.0 1.3
61 301 03 01 7 4.7 0.5 10.1 4400.0 3100.0 1.4
62 302 03 02 1 4.5 1.7 8.9 2300.0 2000.0 1.1
63 303 03 03 D 4.7 0.7 5.8 2100.0 1900.0 1.1
64 304 03 04 4 5.0 0.6 12.4 3200.0 2500.0 1.3
65 305 03 05 D 4.8 0.3 6.4 2100.0 2000.0 1.1
66 306 03 06 7 4.4 0.9 10.3 4500.0 1900.0 2.4
67 307 03 07 6 4.3 0.6 8.7 4500.0 2500.0 1.8
68 308 03 08 5 4.5 . 5.4 2100.0 1300.0 1.6
69 309 03 09 5 4.3 0.6 7.6 1700.0 1100.0 1.5
70 310 03 10 G 4.2 1.0 7.4 4300.0 1700.0 2.5
71 311 03 11 E 5.0 0.6 6.4 2700.0 2100.0 1.3
72 312 03 12 D 5.0 0.5 10.9 3800.0 2900.0 1.3
73 313 03 13 B 4.7 0.9 11.5 3600.0 2800.0 1.3

 

198

OBS ID BLOCK SUBPLOT CLUSTOT ALTINI DKHDIF BARPRE TRUNKINI TREEINI AVESTEM

 

 

74 314 03 14 7 4.6 1.1 8.3 2800.0 2000.0 1.4
75 315 O3 15 7 4.8 1.1 10.9 3700.0 2100.0 1.8
76 316 03 16 G 4.8 0.5 5.3 3100.0 1900.0 1.6
77 317 03 17 G 4.5 0.6 14.0 3100.0 2200.0 1.4
78 318 03 18 BG 4.5 0.7 10.2 2800.0 2000.0 1.4
79 319 O3 19 G 4.5 0.6 8.8 4300.0 2100.0 2.0
80 320 03 20 5 4.2 0.5 7.3 1500.0 1300.0 1.2
81 321 03 21 1 4.7 0.6 16.9 2300.0 2000.0 1.1
82 322 03 22 BG 4.6 0.9 16.2 2800.0 2500.0 1.1
83 323 03 23 7 4.3 0.8 7.3 3100.0 2200.0 1.4
84 324 03 24 BG 4.8 0.8 15.5 3300.0 2400.0 1.4
85 325 03 25 4 4.6 1.2 16.5 3500.0 2700.0 1.3
86 326 03 26 7 4.7 0.7 10.2 4900.0 3100.0 1.6
87 327 03 27 4 4.9 0.7 8.0 2200.0 2000.0 1.1
88 328 03 28 7 5.2 0.8 9.7 3900.0 2200.0 1.8
89 329 03 29 4 5.1 1.0 11.3 2600.0 2300.0 1.1
90 330 03 30 PH 5.1 . 7.9 2700.0 2000.0 1.4
91 401 04 01 4 4.5 0.6 4.6 2000.0 1700.0 1.2
92 402 04 02 4 4.2 1.0 6.0 1100.0 1000.0 1.1
93 403 04 03 D 4.7 0.8 5.1 2000.0 1400.0 1.4
94 404 04 O4 4 4.6 0.7 9.2 2500.0 2200.0 1.1
95 405 04 05 D 4.3 1.2 3.8 1100.0 1000.0 1.1
96 406 04 06 3 4.2 1.0 5.6 1600.0 1400.0 1.1
97 407 04 07 4 4.4 1.0 7.3 2800.0 2000.0 1.4
98 408 04 08 B 4.3 0.9 9.8 3800.0 2300.0 1.7
99 409 04 09 7 4.3 1.2 9.8 5200.0 3700.0 1.4
100 410 04 10 5 3.9 0.7 9.7 4300.0 3300.0 1.3
101 411 04 11 BG 4.4 0.7 7.6 4800.0 2900.0 1.7
102 412 04 12 X 3.8 0.6 6.4 3700.0 3000.0 1.2
103 413 O4 13 D 4.5 0.7 11.8 5600.0 3100.0 1.8
104 414 04 14 G 3.8 0.4 8.9 4200.0 2500.0 1.7
105 415 04 15 B 4.6 0.7 5.8 2300.0 1800.0 1.3
106 416 04 16 D 4.3 0.2 9.5 2200.0 1100.0 2.0
107 417 04 17 7 4.1 0.8 14.6 7900.0 5500.0 1.4
108 418 04 18 4 4.7 1.7 11.0 1700.0 1200.0 1.4
109 419 04 19 7 4.7 0.8 9.5 5200.0 4200.0 1.2
110 420 04 20 D 4.2 0.7 5.6 2500.0 1700.0 1.5
111 421 04 21 1 4.9 0.4 13.8 3800.0 2400.0 1.6
112 422 04 22 7 4.6 0.5 9.7 6800.0 2900.0 2.3
113 423 04 23 6 4.5 2.1 9.1 3800.0 2600.0 1.5
114 424 O4 24 A 4.0 1.2 7.5 2300.0 2100.0 1.1
115 425 04 25 G 4.3 0.4 6.4 3800.0 2600.0 1.5
116 426 04 26 6 3.7 0.4 3.0 1600.0 1000.0 1.6
117 427 04 27 5 4.2 0.8 8.3 2300.0 1300.0 1.8
118 428 04 28 G 4.3 0.6 5.6 2700.0 1600.0 1.7
119 429 04 29 1 4.1 0.4 9.1 2400.0 2000.0 1.2
120 430 04 30 5 4.1 0.4 7.1 2500.0 2100.0 1.2

 

 

APPENDIX C

Site characteristics data from the forty original control sites

199

 

200

O
u:
U)

waqmmbwup

APPENDIX C

ID GROUP POSITION SLOPE CANOPY HERB SOIL STONY
101 BG lowridge slight 5 2.0 siltclay none
102 7 midslope slight 3 3.0 siltclay very
103 ES midslope slight S 3.0 siltclay none
104 D midslope slight 3 0.5 siltclay none
105 BD lowridge slight 5 4 0 siltclay very
126 1 midslope moderate 5 4.0 sandycla very
127 6 midslope slight S 4.0 sandycla very
128 E toe slight 4 2.0 sandysil none
129 G midslope slight 4 2.0 sandycla very
130 E midslope slight 4 4.0 siltclay very
206 E highridg steep 3 2.0 rockycla very
207 3 shoulder steep 3 2.0 sandycla none
208 3 midslope verystee 4 3.0 sandycla very
209 3 midslope verystee 3 4.0 rockycla very
210 3 shoulder steep 3 3.0 sandycla none
221 E shoulder moderate S 3.0 sandycla some
222 FG highridg slight 2 4.0 claysand none
223 3 midslope steep 2 2.0 sandycla some
224 3 midslope steep 4 5.0 sandycla some
225 3 midslope steep S 4.0 siltysan very
301 7 midslope slight S 4.0 claysilt none
302 1 shoulder moderate 3 3.0 siltclay none
303 D midslope moderate 5 3.0 claysilt some
304 4 toe slight S 5.0 sandycla very
305 D midslope moderate 4 3.0 siltclay none
316 G shoulder slight 4 4.0 sandycla none
317 G lowridge none 4 4.0 sandycla none
318 BG shoulder slight 5 5.0 claysand none
319 G toe slight 3 3.0 sandycla none
320 5 toe slight S 5.0 sandycla none
401 4 lowridge slight 3 1.0 sandycla some
402 4 midslope slight 4 4.0 siltclay none
403 D plain none 5 5.0 siltclay none
404 4 toe slight 4 3.0 siltclay none
405 D plain none 3 1.0 siltclay none
426 6 midslope moderate 5 1.0 claysilt none
427 5 shoulder moderate 5 5.0 some
428 G shoulder moderate 4 1.0 siltclay none
429 1 shoulder moderate 3 2.0 siltclay some
430 5 midslope moderate 5 1.0 claysilt none

 

 

 

 

APPENDIX D

Ground vegetation data and CRUZALL data from the forty original control sites

201

OBS

WOQQU‘DWUP

ID

101
101
101
101
101
101
102
102
102
102
103
103
103
103
103
104
104
104
105
105
105
105
126
126
126
126
126
127
127
128
128
128
128
129
129
129
129
129
129
130
130
130
130
206
206
206
206
207
207
207
208
208
208
208
208
209
209
209
210
210
210
210
210
221
221
221
221
221
222
222
222

GROUP

202

APPENDIX D

HERB

B.pingui
Herb
Agave

Agave
B.pingui

C.cariba
Agave
B.pingui
B.pingui
Herb

Agave
B.pingui

Agave

Agave

B.pingui
Agave

B.pingui

Agave

Agave
C.cariba
Herb
L.hystri
A.gracil

Agave
Gramal

Agave
Herb

A.gracil
Herb

L.hystri
Agave

A.gracil
A.gracil

C.cariba
Gramal
C.cariba

INDEX

NOO

OONWO

00

NH

NNHO Hh NM NOD-‘Oé

ONO
wow

00 OU'IU'I

UIUIOOU'I

00 CO CO OUIOUIO

OOOU‘I

TREES

GUATAP
BAITOA
QUINA
CAFETAN
ALMACIG
CAYUCO
CAYUCO
PABURR
CINAZO
BAITOA
PABURR
ALMACIG
GUATAP
BAITOA
FRIJOL
BAITOA
GUATAP
FRIJOL
ALMACIG
GUATAP
BAITOA
FRIJOL
CANDEL
GUAYAC
ALMACIG
CAYUCO
BAITOA
OJPALO
CAYUCO
CAMERON
GUATAP
CAYUCO
MOSTAZO
BRUCON
CINAZO
FRIJOL
OJPALO
BAITOA
CAYUCO
CAMERON
CAYUCO
BAITOA
GUATAP
ALPARGAT

GUATAP
CAYUCO
FRIJOL
CAYUCO
QUINA
PAAMAR
CAYUCO
GUATAP
ALMACIG
CANDEL
CANDEL
CAYUCO
FRIJOL
QUINA
GUAYAC
CAYUCO
CANDEL
FRIJOL
CANDEL
CAYUCO
SANGRE
TABACU
GUATAP
GUATAP
ALMACIG
AROMA

CRUZALL

HPUQWHUHFHPHWNNO‘H‘NHHHWHNPHPNMUNHHNUHNUUHHNN

HPJHFJQIJ-

HDJO-NFJHIJF‘UOJUJHQJF‘thHDJhJN

 

 

 

ID

222
223
223
223
224
224
224
225
225
225
225
301
301
301
301
302
302
302
302
302
302
303
303
303
303
303
304
304
304
305

305
305
316
316
316
317
317
317
317
317
317
318
318
318
318
319
319
319
319
319
320
320
320
320
401
401
401
401
401
401
402
402
402
402
403
403
403
403
403
404
404
404
404
405
405

GROUP

"'1
Q

06)QtDG>Q<DCimtJUt3tﬂﬁ¢>ht3t1Ut1Ur4thrJhaH~Jq~a\chsuiwcsuiwtpuaw

UU‘hﬁuhnbUUUUUb##bééhﬁﬁﬁmmmeQQQQ

203

HERB

Gramal
A.graci1
C.cariba
Herb
Agave
A.graci1

Gramax

Agave
A.gracil
Herb

Gramal
Frucraea

Gramal
Herb

Herb
C.cariba
Gramal

Frucraea
C.cariba

Gramal

C.cariba
Gramal
Herb
C.cariba
Herb

Cactil
C.cariba
Herb
Herb

C.cariba
C.cariba
Herb

L.hystri
Herb
C.cariba

L.hystri
Herb

C.cariba
C.cariba
Herb

Herb
Cactil

L.hystri
Cactil
Herb
L.hystri
C.cariba
Cactil
Herb

INDEX

U UUIOOND

NNN

HOW tow

PM

U‘
0

NNUO WWNUO

DUN

OU'I
U10

HOPONOO

OOUIU'OO

OOO

OU'lO

OOOU‘I OOOOU‘

NNU‘I
OOO

OUIU'I 000

#00

OUIOU‘OU‘U'

O

OO

TREES

ALPARGAT
CAYUCO
CANDEL
CINAZO
SANGRE
UNYA
QUINA
CAYUCO
CANDEL
PABURR
GUATAP
CAYUCO
CAFETAN
BAITOA
ALMACIG
ALAMCIG
QUINA
FRIJOL
BAITOA
CAYUCO
CAFETAN
CAYUCO
ALMACIG
QUINA
BRUCON
FRIJOL
CAYUCO
FRIJOL
BAITOA
QUINA
BAITOA
CAYUCO
GUAYAC
FRIJOL
CAYUCO
CAMBRON
ALPARGAT
CAYUCO
CAMBRON
FRIJOL
BAITOA
CAFETAN
FRIJOL
UVERO
CAYUCO
BAITOA
CINAZO
CAYUCO
BAITOA
CAMERON
UVERO
BAITOA
GUATAP
FRIJOL
CAMBRON
CAYUCO
UVERO
MOSTAZO
BAITOA
ALPARGAT
BRUCON
CAYUCO
BAITOA
GUAYAC
BRUCON
CAYUCO
BAITOA
ALPARGAT
GUATAP
FRIJOL
GUATAP
BAITOA

CAYUCO
CAMBRON
CAYUCO

CRUZALL

mtahawbduiwtdhabtarahihrdhluJH+JL:w+J»IM\o+‘qron:wr4u:wua+¢wupoxprahawtauaHrahIHoasiwlanaHtah‘”basintahawtohaptakaneostp

UHN-

 

 

 

 

 

 

OBS

148

150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167

ID
405

405
426
426
427
427
427
427
428
428
428
429
429
429
429
430
430
430
430

GROUP

WOOUUU

UlUIUIU'll-‘HPHQQQU'IUIUI

204

HERB

C.cariba
Herb
A.gracil

Herb

C.cariba
Gramax
Herb

A.gracil

L.hystri
C.cariba
Frucraea
Herb

INDEX

N

ONOO

H+ao
o<su1

OU'IO

raova

U'IOU'IU'I

O

TREES

AROMA
ALPARGAT
BAITOA
GUATAP
UVERO
SANGRE
MOSTAZO
GUATAP
CINAZO
CAYUCO
SANGRE
BAITOA
CAYUCO
QUINA
ALMACIG
PABURR
CAFETAN
GUATAP
CINAZO
GUACON

CRUZALL

HF‘NB’HBOPF‘ﬁFJwFJbFJ

F‘Nl‘hJHFJ

APPENDIX E

Growth and mortality parameters for 120 sites

205

206
APPENDIX E

OBS ID BLOCK SUBPLOT CUTLEVEL CLUSTOT BARGRTH BARAVE DKHDIF BARMORT TRUNKMRT NETGRTH

1 101 01 01 C BG 1.4 7.1 0.9 1.1 600.0 0.3
2 102 01 02 1 7 0.7 3.1 0.4 0.1 100.0 0.6
3 103 01 03 C BG 3.7 12.8 0.8 3.7 1000.0 -0.0
4 104 01 04 C D 0.5 4.2 0.3 2.8 900.0 -2.3
5 105 01 05 C BD 1.2 3.0 0.5 3.2 1800.0 -2.0
6 106 01 06 3 BD 1.5 9.8 0.7 2.8 1000.0 -1.3
7 107 01 O7 3 C 1.5 11.3 1.3 0.1 100.0 1.4
8 108 01 08 3 BG 1.4 7.2 0.9 2.8 700.0 -1.4
9 109 01 O9 3 4 1.0 9.7 0.9 0.6 300.0 0.3
10 110 01 10 C 5 1.7 8.9 0.9 6.2 3100.0 -4.5
11 111 01 11 1 D 0.9 6.1 0.7 2.4 1100.0 -1.6
12 112 01 12 1 4 1.0 9.2 0.4 0.6 500.0 0.4
13 113 01 13 2 D 1.2 6.3 0.7 0.5 300.0 0.7
14 114 01 14 1 D 0.5 4.0 0.5 3.0 300.0 -2.5
15 115 01 15 2 FG 1.5 9.5 1.2 0.5 200.0 1.0
16 116 01 16 4 BG 0.8 7.2 0.8 0.5 300.0 0.3
17 117 01 17 4 D 1.4 9.2 1.0 0.3 100.0 1.1
18 118 01 18 4 1 0.9 6.0 0.6 1.4 300.0 -0.5
19 119 01 19 4 5 0.9 6.9 0.7 0.2 200.0 0.7
20 120 01 20 4 3 0.6 3.9 0.6 0.7 500.0 -0.1
21 121 01 21 3 E 1.2 23.3 1.4 1.0 500.0 0.2
22 122 01 22 2 D 0.5 10.5 0.8 0.1 100.0 0.5
23 123 01 23 3 EH 0.8 13.7 0.7 2.4 300.0 —1.6
24 124 01 24 3 6 0.2 9.4 1.0 1.3 200.0 -1.1
25 125 01 25 4 E 0.4 18.5 1.2 0.9 500.0 -0.6
26 126 01 26 1 1 1.1 2.6 0.3 0.6 200.0 0.4
27 127 01 27 C 6 0.2 1.8 0.4 2.7 500.0 -2.5
28 128 01 28 1 E 2.0 12.6 0.6 0.7 300.0 1.3
29 129 01 29 C G 1.4 5.9 0.8 0.9 700.0 0.5
30 130 01 30 C E 1.7 6.6 0.7 0.6 700.0 1.1
31 201 02 01 3 3 0.2 3.5 0.4 1.8 800.0 -1.6
32 202 02 02 4 3 0.1 4.5 0.6 3.2 600.0 -3.0
33 203 02 03 2 3 0.7 9.6 0.6 2.8 800.0 -2.2
34 204 02 04 3 BG 1.2 9.8 0.7 0.3 200.0 0.9
35 205 02 05 3 3 1.0 11.6 0.9 1.0 400.0 0.0
36 206 02 06 C E 0.3 3.4 0.5 1.3 400.0 -1.0
37 207 02 07 C 3 1.0 10.0 0.8 4.6 700.0 -3.6
38 208 02 08 1 3 0.3 2.6 0.4 2.7 1100.0 -2.4
39 209 02 09 C 3 0.5 4.2 0.6 1.5 600.0 -1.0
40 210 02 10 C 3 0.4 2.1 0.3 1.8 600.0 -1.4
41 211 02 11 1 3 1.0 5.3 0.5 2.3 1000.0 -1.3
42 212 02 12 C 3 0.5 3.5 0.4 1.1 500.0 -0.6
43 213 02 13 C 1 0.8 5.5 0.7 3.6 400.0 -2.7
44 214 02 14 C X 0.6 4.8 0.6 3.8 1200.0 -3.2
45 215 02 15 C C 0.4 6.3 0.3 4.1 1100.0 -3.7
46 216 02 16 4 1 0.5 6.3 0.6 2.6 400.0 -2.1
47 217 02 17 3 3 0.2 4.1 0.4 1.4 400.0 —1.1
48 218 02 18 5 1 0.4 7.4 1.0 0.7 400.0 -0.2
49 219 02 19 4 C 0.4 5.4 0.7 1.0 500.0 -0.5
50 220 02 20 3 7 0.6 5.3 0.6 1.9 600.0 -1.3
51 221 02 21 C E 1.6 14.7 1.1 0.2 100.0 1.4
52 222 02 22 C PG 0.2 4.1 0.5 2.4 1100.0 -2.1
53 223 02 23 1 3 0.7 8.7 0.5 3.1 1400.0 -2.4
54 224 02 24 C 3 1.2 7.0 0.5 3.7 1200.0 -2.5
55 225 02 25 C 3 1.3 6.1 0.6 1.5 500.0 -O.2
56 226 02 26 4 3 0.4 10.1 0.3 0.4 300.0 -0.0
57 227 02 27 3 l 0.3 5.2 0.6 0.8 400.0 -0.5
58 228 02 28 5 7 0.2 6.8 1.3 0.1 100.0 0.1
59 229 02 29 4 3 0.3 6.7 0.5 1.4 400.0 -1.2
60 230 02 3O 4 1 . . . 1.8 900.0 .
61 301 03 01 C 7 0.9 3.8 0.5 2.4 1500.0 -1.5
62 302 03 02 1 1 1.9 24.0 1.7 2.4 1100.0 -0.5
63 303 03 03 1 D 0.7 5.5 0.7 3.3 900.0 -2.6
64 304 03 04 C 4 1.1 4.8 0.6 0.8 600.0 0.2
65 305 03 05 C D 0.3 2.5 0.3 2.1 800.0 -1.7
66 306 03 06 2 7 1.4 10.6 0.9 0.3 300.0 1.1
67 307 03 07 2 6 1.0 7.5 0.6 1.9 800.0 -0.9
68 308 O3 08 2 5 0.5 8.2 . 0.2 200.0 0.3
69 309 03 09 2 5 0.3 3.3 0.6 0.3 300.0 —0.1
70 310 O3 10 3 G 1.8 16.4 1.0 0.7 300.0 1.1
71 311 03 11 2 E 1.1 6.7 0.6 0.5 500.0 0.7
72 312 03 12 2 D 0.6 4.0 0.5 1.2 700.0 -O.5
73 313 O3 13 2 B 1.7 12.7 0.9 1.1 500.0 0.6
74 314 O3 14 2 7 1.3 12.7 1.1 0.5 400.0 0.8

207

OBS ID BLOCK SUBPLOT CUTLEVEL CLUSTOT BARGRTH BARAVE DKHDIF BARMORT TRUNKMRT NETGRTH

75 315 03 15 3 7 0.7 8.6 1.1 2.8 1100.0 -2.1
76 316 03 16 1 G 1.1 6.7 0.5 0.7 600.0 0.4
77 317 03 17 C G 0.5 5.1 0.6 9.8 2600.0 -9.4
78 318 03 18 1 BG 1.5 9.4 0.7 0.6 500.0 1.0
79 319 03 19 1 G 0.8 5.4 0.6 3.3 1700.0 -2.5
80 320 03 20 C 5 0.9 8.8 0.5 0.8 300.0 0.0
81 321 03 21 4 1 0.5 16.9 0.6 0.2 100.0 0.3
82 322 03 22 3 BG 1.0 16.7 0.9 0.0 0.0 1.0
83 323 03 23 2 7 0.9 7.2 0.8 1.3 400.0 -0.4
84 324 03 24 4 BG 1.2 29.3 0.8 0.5 300.0 0.6
85 325 03 25 2 4 2.8 23.0 1.2 1.0 700.0 1.7
86 326 03 26 2 7 1.3 7.6 0.7 1.1 500.0 0.2
87 327 03 27 2 4 0.5 6.3 0.7 1.4 500.0 -0.9
88 328 03 28 2 7 0.6 5.4 0.8 1.6 1000.0 -0.9
89 329 03 29 2 4 0.9 15.3 1.0 2.5 900.0 -1.6
90 330 03 3O 2 PH . . . 5.2 1400.0 .
91 401 O4 01 1 4 0.9 6.9 0.6 1.0 500.0 -0.1
92 402 O4 02 C 4 1.5 16.8 1.0 0.3 100.0 1.3
93 403 04 03 1 D 1.0 11.3 0.8 1.8 600.0 -0.8
94 404 04 04 C 4 1.8 11.5 0.7 1.4 900.0 0.5
95 405 04 05 C D 0.7 14.1 1.2 0.9 500.0 -0.1
96 406 04 06 5 3 0.1 3.1 1.0 0.7 100.0 -0.7
97 407 04 07 3 4 1.6 39.6 1.0 0.7 200.0 0.9
98 408 04 08 3 B 0.3 5.6 0.9 1.4 400.0 -1.0
99 409 O4 09 5 7 1.1 12.2 1.2 0.0 0.0 1.1
100 410 04 10 4 5 0.3 5.6 0.7 2.3 300.0 -2.0
101 411 04 11 2 BG 1.5 7.3 0.7 1.1 600.0 0.4
102 412 04 12 2 X 1.3 7.5 0.6 0.4 200.0 1.0
103 413 04 13 2 D 1.3 5.9 0.7 0.9 500.0 0.5
104 414 04 14 3 G 0.3 2.9 0.4 1.9 800.0 -1.6
105 415 04 15 l B 0.9 8.5 0.7 1.1 700.0 -0.1
106 416 04 16 1 D 0.8 27.9 0.2 7.7 1200.0 -6.8
107 417 04 17 4 7 1.4 7.1 0.8 2.2 1000.0 -0 9
108 418 04 18 5 4 0.8 15.7 1.7 1.0 600.0 -0 2
109 419 04 19 3 7 1.5 9.1 0.8 1.7 800.0 -0.2
110 420 04 20 2 D 0.8 11.6 0.7 1.8 700.0 -1 0
111 421 04 21 3 1 0.6 8.8 0.4 0.4 300.0 0.2
112 422 O4 22 3 7 0.3 3.2 0.5 1.8 1400.0 -1.5
113 423 04 23 3 6 1.5 16.9 2.1 1.3 400.0 0.2
114 424 04 24 3 A 0.7 11.6 1.2 1.9 600.0 -1.3
115 425 04 25 3 G 0.8 5.8 0.4 0.3 300.0 0.5
116 426 04 26 C 6 0.5 10.7 0.4 1.6 900.0 -1.0
117 427 04 27 C 5 0.7 8.9 0.8 0.7 500.0 -0.0
118 428 04 28 C G 0.5 4.6 0.6 1.9 1400.0 -1.3
119 429 O4 29 C 1 0.3 2.6 0.4 2.4 1400.0 -2 1
120 430 04 30 l 5 0.6 3.1 0.4 0.6 100.0 -0 1

APPENDIX F
Nonparametric analysis of growth and mortality:

Assumptions and Calculations

208

209
APPENDIX F

(I) Assumptions and equations for Kruskal-Wallis distribution-free test.1

A1: The basic model is:

X§=u+rj+e§, i=1,....,nj, j=1, ..... ,k,
I:
Where u is the overall mean, 7 is the effect of treatment j, and E 1']- = 0.
i=1

A2. The e’s (error variables) are mutually independent.
A3. Each e comes from the same continuous population.

To test
H0: 1'1 = 72 = = 7k,
(1) Rank all values from lowest to highest.
(2) Determine the H statistic, where the H statistic is calculated as:

12 " _giz
H=( N(N+1) 2.; nj ) - 3(N+1)

 

Where:
N = the total sample size,
P7- = the sum of the ranks for treatment j,
nj = the sample size for treatment j,
k = total number of treatments.
(3) Reject H, if H z x 2 (k-1.cr)and

Accept H, ifH < x 2(k-1, a),

where or represents the approximate probability assuming a Chi-square distribution
with (k-l) degrees of freedom.

 

lTaken from Hollander and Wolfe (1973), pp 114-120.

210

(II) Calculations of H-statistic for test of differences in growth and mortatlity
between cluster groups (N =87).

 

 

 

 

 

 

 

 

 

 

 

Cluster n BARGRTH NETGRTH BARAVE STEWORT BARMORT N DKHDIF

group (ﬂnk W) (ﬂInk W) ("t11k W) (ﬁnk W) (m3 m) (M3 m)
Three 18 550.0 498.0 572.0 933.5 1012.0 18 537.0
Six 5 175 .0 164.0 256.0 226.5 280.0 5 220.0
E 7 411.0 464.0 405.0 253.5 179.0 7 363.0
Five 8 303.0 385.0 306.0 226.5 259.0 8 290.0
One 10 372.0 412.0 396.0 351.5 435.0 10 373.0
Seven 14 707.0 691 .0 5 86.0 662.0 5 62 .0 14 759.0
D 14 626.0 530.0 611.0 688.0 716.0 14 575.0
Four 1 1 684.0 684.0 696.0 486.5 385 .0 10 624.0

 

 

 

 

 

 

 

 

 

 

 

 

 

BARGRTH NETGRTH BARAVE STEMMORT BARMORT DKHDIF

 

12
( N(N+1)) 0.0015674 0.0015674 0.0015674 0.0015674 0.0015674 0.0016038

 

k 3 .2
( Z 1151) 178603.65848 182118.73149 17733422136 173251.12165 17711537103 17265235714

s=1

 

 

3 (N +1) 264.00000 264.00000 264.00000 264.00000 264.00000 261.00000 II

 

 

 

 

 

 

 

"fl-statistic 15.43361 21.452897 13.953656 7.5538055 13.610631 12.899839 ||

 

 

 

 

 

 

211

(III) Calculations of H-statistic for test of differences in growth and mortatlity
between cutting levels (N =113).

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Outing n BARGRTH NETGRTH BARAVE S'IEMMOR'I‘ BARMORT N

level ("1* m) ("31‘ m) (“3“ m) (Mk “1”) ("‘3k m)

Control 32 1770.0 1429.0 1430.0 2372.0 2288.0 32

One 19 1179.0 983.0 983.0 1239.0 1215.0 19

TWO 22 1480.0 1638.0 1445.0 1104.5 942.0 21
IF'hree 25 1403.0 1502.0 1680.0 11765 1310.0 25 1738.0
ll Four 15 609.0 889.0 903.0 548.5 686.0 15 904.0

 

 

 

 

 

 

 

 

 

 

 

 

BARGRTH NETGRTH BARAVE STEMMORT BARMORT DKHDIF
12

( N(N+ 1)) 0.0009315 0.0009315 0.0009315 0.0009315 0.0009315 0.0009482

" R .2
( 2 nj)) 37408857399 379555 .86916 376927.26806 387494.16922 381639.84657 369250.015 66

est ﬂ
3 (N + 1) 342.0 342.00000 342.00000 342.00000 342.00000 339.00000

ll H~statistic 6.4756162 115685786 9.1199516 18.9633621 135098710 11 5685786

”

 

 

 

#

212

(IV) General equation used for distribution-free multiple comparisons based on
Kruskal-Wallis rank sums: an approximation valid for unequal sample sizes
(Hollander and Wolfe 1973).

To decide if Tu ¢ 1", , determine if:

N+1

_1_ ....1_
1/2 In
:R... ' R..: 228/1km») l 12 l l( “u + “v )l '

- R. ' is the difference between the rank mean of "treatment" u and the
rank mean of "treatment" v.

2(a , [Mk-1H) is the z-value associated with the upper-tail area of a normal curve,
based on a pre-determined probability level, and the number of
groups being examined.2

is the number of samples for "treatment" u.

is the number of samples for "treatment" v.

 

2Taken from Table 2, Appendix A in Ott (1988).

213

(V) Calculation of test-statistics for multiple comparisons of cluster groups.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Comparison N( N +1 I .i. _l_

z(a/[k(k-1)])3 [ 12 ]1/2 [( n. + 5)]172 Test statistic
Three v/s six 2.92 25.2587 0.5055 37283356722
Three v/s E 2.92 252587 0.4454 32.8506569416
Three v/s 5 2.92 252587 0.4249 313386711596
Three v/s one 2.92 252587 03944 29.0891313376
Three v/s seven 2.92 25 2587 03563 262790504452
Three v/s D 2.92 252587 0.3563 262790504452 ll
Three v/s four 2.92 252587 03827 282261931108 ll
Six v/s B 2.92 252587 05855 43183789042 ll
Six v/s ﬁve 2.92 252587 0.5701 42.0479558204
Six v/s one 2.92 25.25 87 05477 403958347708
Six v/s seven 2.92 252587 05210 38.426565484
Sex v/s D 2.92 252587 05210 38.426565484
Six v/s four 2.92 252587 05394 39.7836649176
E v/s five 2.92 252587 05175 38.16842157
E v/s one 2.92 25 2587 0.4928 363466630912
E v/s seven 2.92 25 2587 0.4629 34.1413765116
E v/s D 2.92 252587 0.4629 34.1413765116
E v/s four 2.92 25 2587 0.4835 35 .660737834
Five v/s one 2.92 25 2587 0.4743 34.9821881172
Five v/s seven 2.92 25 2587 0.4432 32.6883950528
Five v/s D 2.92 25 2587 0.4432 32.6883950528
Five v/s four 2.92 252587 0.4647 342741362388
One v/s seven 2.92 252587 0.4140 30534737256
One v/s D ' 2.92 252587 0.4140 30534737256
One v/s four 2.92 252587 0.4369 322237360076
Seven v/s D 2.92 252587 03780 27.879542712
Seven v/s four 2.92 252587 0.4029 29.7160522716
D v/s four 2.92 25.2587 0.4029 29.7160522716

 

 

 

 

 

 

3 0.00179, where a was set at 0.1 and k(k-1)=56.

 

214

(VI) Calculations of test statistics for multiple-comparisons of cutting levels.

 

 

 

 

 

 

 

 

 

 

 

 

Comparison of N +1 1_ L

cutting levels 201/M4”; [ 12 11/2 [( n, + 11, )11/2 Test statistic
Control v/s one 2575 32.4756 02896 24217704432
Control v/s two 2575 32.4756 02769 23.155671123
Contorl v/s three 2575 32.4756 0.2669 22319424423
Control v/s four 2575 32.4756 03129 26.166159243
One v/s two 2575 32.4756 03131 26.182884177
One V/s three 2575 32.4756 03043 25.446987081
One v/s four 2575 32.4756 03454 28.883961018
TWO v/s three 2575 32.4756 0.2960 24.75290232
Two v/s four 2575 32.4756 03380 2826513846
Three v/s four 2575 32.4756 03266 27311817222

 

 

 

 

 

 

‘ 0.005 where a was set at 0.1 and k(k-1)=20.

 

APPENDIX C

Results of hierarchical cluster analyses

215

 

216

APPENDIX G

(I) Results of SAS analyses using 45 sites and 16 species

Ward’s Minimum Variance Cluster Analysis

2 observation(s) trimmed with estimated density 6.5762413-28 or less.

Eigenvalues of the Covariance Matrix

Eigenvalue Difference Proportion Cumulative

1 871.347 511.868 0.456237 0.45624

2 359.479 103.516 0.188223 0.64446

3 255.962 126.921 0.134022 0.77848

4 129.042 63.536 0.067566 0.84605

5 65.506 8.488 0.034299 0.88035

6 57.018 15.078 0.029855 0.91020

7 41.940 1.682 0.021960 0.93216

8 40.257 13.815 0.021079 0.95324

9 26.442 5.334 0.013845 0.96709

21.108 4.954 0.011052 0.97814

16.155 6.164 0.008459 0.98660

9.990 4.411 0.005231 0.99183

5.579 1.604 0.002921 0.99475

3.975 0.353 0.002081 0.99683

15 3.622 1.190 0.001896 0.99873
16 2.432 . 0.001273 1.00000

Root-Mean-Square Total-Sample Standard Deviation = 10.92547
Root-Mean—Square Distance Between Observations = 61.80378

Number
of
Clusters

HNWIBUIGQG

217

Ward’s Minimum Variance Cluster Analysis

Frequency
of New Semipartial
Clusters Joined Cluster R-Squared R-Squared
401 402 2 0.000963 0.999037
304 404 2 0.001024 0.998013
207 210 2 0.001202 0.996811
209 223 2 0.001477 0.995334
129 428 2 0.001772 0.993561
208 212 2 0.001821 0.991741
126 213 2 0.001826 0.989915
CL41 CL42 4 0.002325 0.987590
302 429 2 0.002408 0.985182
303 305 2 0.002746 0.982436
206 215 2 0.002878 0.979558
130 316 2 0.003595 0.975964
110 430 2 0.003610 0.972353
128 320 2 0.003838 0.968515
104 105 2 0.003876 0.964639
317 319 2 0.004008 0.960631
CL33 405 3 0.004301 0.956330
CL39 225 3 0.004404 0.951927
CL28 403 3 0.004919 0.947008
CL38 CL27 4 0.005408 0.941600
CL32 221 3 0.005775 0.935825
102 301 2 0.006696 0.929130
222 224 2 0.006775 0.922355
101 CL30 3 0.007419 0.914936
CL37 CL25 5 0.007445 0.907491
CL21 318 3 0.009631 0.897860
CL29 427 3 0.009714 0.888146
CL19 214 4 0.011422 0.876724
CL24 CL35 7 0.011738 0.864986
CL23 CL31 6 0.013481 0.851505
CL40 CL18 7 0.015566 0.835939
CL36 CL34 4 0.017648 0.818291
CL15 CL22 7 0.018550 0.799742
CL17 CL26 6 0.020807 0.778934
CLlO 103 8 0.024323 0.754611
CL8 CL16 11 0.032049 0.722562
CL9 CL13 12 0.034288 0.688274
CL12 CL20 9 0.034369 0.653905
CL6 CL11 16 0.086081 0.567824
CL7 CL4 27 0.130356 0.437467
CL3 CLS 36 0.173705 0.263763
CL2 CL14 43 0.263763 0.000000

Tie

HDWHNIV'HIOU

QOHOCDM'X!

.3

.25

.15

.05

GM“
ONO

9

 

0-———

 

218

8A8

15:54 Thursday. October

word's Hintnun Variance Cluster Analysis

1
0
1

OH”
DU.
OHM
090M
up!)
warm
UOH
ON“
ONU
4N5

ID
1 3 3 3 3 4 1 4 3 3 1 3 1 2 3 4
0 0 1 0 0 0 2 2 l l 3 1 2 1 0 2
2 1 8 3 5 5 9 8 7 9 0 6 6 3 2 9

~10”
OH”
DON
”HM
WON
UN”
UNA)
”NU
ON”

12. 1995 3
1 1 4 3 4 4 4
0 0 0 0 0 0 0
4 5 3 4 4 1 2

 

XXXXXXXXXXXXXXXXXXKAX
XXXXXXAAIXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXX
XXXIXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXX XXXXX

XXXXXXX XXXXX . XXXXX

AXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

\DQQO‘UIthb-INH

Flexible—Beta Cluster Analysis

Eigenvalues of the Covariance Matrix

Eigenvalue

837.729
345.195
264.611
149.565
88.712
60.890
58.253
39.957
35.322
26.022
19.458
18.093
14.415
5.204
3.782
2.458

Difference

Beta

492.534
80.584
115.046
60.853
27.822
2.637
18.296
4.635
9.300
6.564
1.365
3.678
9.211
1.422
1.324

= -0.25

Proportion

0.425315
0.175256
0.134343
0.075934
0.045039
0.030914
0.029575
0.020286
0.017933
0.013211
0.009879
0.009186
0.007319
0.002642
0.001920
0.001248

Cumulative

0.42532
0.60057
0.73491
0.81085
0.85589
0.88680
0.91638
0.93666
0.95459
0.96781
0.97768
0.98687
0.99419
0.99683
0.99875
1.00000

Root-Mean-Square Total-Sample Standard Deviation = 11.09523

Mean Distance Between Observations

Number
of
Clusters

PMWIbU‘O’tQQ‘D

Flexible-Beta Cluster Analysis

Clusters Joined

401
304
207
209
129
208
126
CL43
302
303
206
130
110
128
104
317
CL35
CL41
CL30
CL40
CL34
CL39
102
222
101
CL31
CL26
CL22
CL20

402
404
210
223
428
212
213
CL4
429
305
215
316
430
320
105
319
405
225
403
CL2
221
CL2
301
224
CL3
427
CL3
318
214
CL3

4

9

7

2

7

3

Frequency
of New
Cluster

Normalized
Flexible
Distance

\DQ0019QQNO‘hWQNUNMLﬂWhWthNNNNNMNNIbNNNNNNN

0.207675
0.214093
0.231930
0.257167
0.281676
0.285489
0.285871
0.302795
0.328336
0.350597
0.358919
0.401146
0.402006
0.414509
0.416573
0.423571
0.434201
0.442452
0.466544
0.477776
0.501193
0.523151
0.547476
0.550704
0.571936
0.639743
0.651031
0.651550
0.695780
0.723903
0.752431
0.786435
0.843440
0.865587
0.918513
1.084151
1.120361
1.134807
1.155165
1.714345
1.882564
1.962854
2.172582
3.877186

= 59.86433

Tie

mnemna~UMXOH1

GODWHOP'U

220

SAS 15:54 Thursday, October 12. 1995 6
Flexible-Beta Cluster Analysis

ID

 

WW
WWW XXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXX XXXXXXXXXXXXX
XXXXXXLKXXXXXXXXXXK XXXXXXXXXXXXXXXXX WWW XXX XXXXXXX XXXXXXXXXXXH
1 .5 +XXXXXXXXXXXXXXXXXXX XXXKXXXJOLXXWLXX XLXXXXXXXXXXXJOOCXXJO‘XXXXX XXX 1000000! 10000113000000!
WW XXXXYXXXXXAXXXXXX XXXXXIXXXXXXXXXXXXXXXXXXX 100‘ 1000000! JDOOOOOOOQXXX
XXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXKX XXX

 

WWWKWW XXXXXXXXXXXXXX XXXXXXXXXXXXX
WWWXXXWXXXXX XXXXXXXXXAXXXXXXXXXXW
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx W XXXXXXXXXXXXXXXXXXX
05+XXX.XXXXXXXX.XXXXXXXXXXXX .XXX moothxxxmxxxxxxmxxxxxxxmxx
.XXXXXXXXX. ..X.XX.. .HXXXXXX. .XXXXXXX

. .XXXXXX

221

(II) A representative dendogram from SYSTAT (Willkinson 1989) of Ward
Minimum Variance hierarchical cluster analysis using 16 species and 120 sites
(Distance metric is Euclidean distance)

(1) Numbers in bold along the right margin represent points above which
cuts were made to form clusters used in these anayses.

(2) Numbers in italics represent range between which cuts can be made
without affecting the clusters formed.

 

 

 

 

 

 

 

 

 

0.000 DISTANCES 500.000
406 -
+- 2.389
217 'l
+- 7.862
223 °I
+ 8.849
209 -
+- 18.482
208 -
+ 3.692
20? -I.
+- 8.604
225 -l
+- 4.498
203 -
+ ............... 25.960
224 '
6.825
120
8.995
205 -
I 4.680
229 +
++ 6.929
201 -I
. +-- 14.831
212 -
I 3.370
211 *
++ 6.615
210 *
I 3.471
207 -
8.323
226
+ ................... 141.111
302 -
+- 4.914
429 °
8.446
421 -
++ 5.421
230 +
l 3.285
216 - '
+-- 9.963
321 -' I
+ ........... 23.668
213 '
+- 5.130
126 +
I 2.961
118 °
+-- 13.670
218 'l
+- 7.036
227 -
+---- 112.211
115 ---
+- 15.440

 

222
414
424
425
423
426
127
307
124
309
427
308
320
128
125
311
I30
121
2l9
206
215
221
107
214
110
410
430
119
123
330
329
402
327
407
401
404
325
304
109
418

 

 

 

 

.-——- *n'

o—+++"’-'

 

l“*+'

0 8 8
D—+-++_--'

 

 

I++—'

 

 

 

 

..+ ++—-+-+-+-+‘-'

222

6.530
5.881
9.304
29.815
9.695
21.672
11.607
3.954
53.449
8.970
7.155
23.334
6.204
13.445
4.814
7.417
5.265
*833.206
6.511
5.372
13.130
8.946
10.795
24.058
3.780
9.938
4.355
7.239
20.493
1 ................ 283.685
2.289
2.570
12.323
4.928
3.880
2.935
3.227
3.140
7.041
2.439

 

112
104
103
324
322
108
316
317
319
129
428
114
413
117
122
415
105
106
408
116
318
411
204
313
101
412
405
303
305
416
420
403
312
111
113
306
328
315
314
323

--+++-—.
I

 

+-—l

I-_;.I_?"""'

v—-++I
I

 

 

 

 

 

 

6.256
108.700
7.627
9.722
3.687
26.169
7.135
6.339
8.267
4.216
32.361
7.625
11.461
4.316
8.044
4.779
5.938
3.915
15.847
5.662
11.712
6.276
4.350
7.376
8.000
24.851
6.582
5.247
12.660
3.424
7.084
4.785
11.336
3.503

**52.916

5.427
4.577
4.060
11.071
6.004

301
419
220
102
326
417
409
310
422
228

224

 

 

 

9.791
2.942
7.170
7.993
6.280
3.861
24.719
9.692
7.431

APPENDIX H

CA scores from the initial analyses

22.5

226
APPENDIX H

(I) Results of CA procedure using 16 spp and 45 control sites

Inertia and Chi-Square Decomposition

Singular Principal Chi-
Valuoa Inortiaa Bquaroa Parcanta 8 8 12 16 20
----+----+----+----+----+---
0.78888 0.56076 18057.8 20.098 *'*'*'**‘*****""*"‘**'
0.68638 0.81776 13852.7 18.978 *****"*"******"*
0.58767 0.38535 11121.1 12.378 *'***'*'*"*"'
0.58936 0.30179 9718.85 10.818 "*'*"""*"
0.52025 0.27066 8715.88 9.708 *'*"""*"
0.83015 0.18503 5958.39 6.638 '***""
0.38322 0.18685 8729.05 5.268 *“'***
0.37732 0.18237 8588.58 5.108 '*""
0.35765 0.12792 8119.17 8.588 '*'***
0.30856 0.09276 2986.98 3.328 "*'
0.26560 0.07055 2271.78 2.538 88*
0.21861 0.08606 1883.12 1.658 8*
0.20860 0.08351 1801.23 1.568 **
0.18512 0.02106 678.138 0.758 9
0.13766 0.01895 610.268 0.688 '
2.79137 89888.5 (Degrees of Freedom . 660)
Row Coordinates
Initial Diml Dim2 Dim3 Din8 Dims
Cluster
8 0.25685 0.17066 -0.32319 0.18267 -0.12138
2 -0.37605 0.06933 0.08087 -0.02055 0.00858
E -0.87621 0.88573 -0.63961 -0.98061 1.53217
8 -0.62833 0.09186 -0.60617 0.09867 -0.19591
8 -0.89200 0.15981 -0.58789 -0.26008 0.37219
5 -0.15239 0.71683 0.02711 -0.80882 -0.21836
1 -0.08872 -1.10889 0.03361 -0.38382 0.09358
6 -0.71688 0.25991 2.68092 0.99886 0.81781
5 -0.80090 0.85033 0.88126 -0.66813 -0.80238
2 -0.80051 0.33305 -0.08088 0.28687 -0.58589
8 -0.88935 0.69570 0.18812 —0.21788 -0.87595
5 0.59322 0.88021 0.13670 -0.16221 -0.10379
3 1.21886 -0.09139 -0.31252 0.58298 0.09081
3 1.31127 0.29722 —0.09957 0.13755 -0.06839
3 1.69132 -0.02796 -0.27072 0.62189 0.07562
3 1.00855 -0.07136 -0.16507 0.28272 0.22056
3 1.52630 0.03987 -0.15777 0.35280 0.03283
1 0.29725 -1.09608 -0.08093 -0.36278 -0.01267
E 0.55592 0.02586 -0.18769 0.06299 -0.26888
E 1.05656 0.81676 0.08698 -0.08217 -0.07256
5 0.89282 0.73995 0.17893 -0.38337 -0.27322
C 0.11663 0.10969 1.13398 0.21805 -0.36208
3 1.50015 0.03888 -0.13161 0.61567 0.03387
3 1.19521 0.18329 0.07528 0.27236 0.03265
3 1.37057 0.33639 ~0.18828 0.21036 0.07575
A -0.13788 -0.82608 0.06665 -0.08357 -0.18781
1 -0.01033 -1.78102 0.68310 -0.95068 -0.01802
D -0.38995 -0.50331 0.07026 0.15382 -0.18228
8 -0.78280 -0.19215 -0.58186 0.68956 0.08888
D -0.83808 -0.11229 -0.82699 0.15573 0.00182
8 -0.59207 0.35161 -0.06725 0.07911 -0.68186
8 -0.81920 0.02786 -0.38988 0.37970 -0.66589
3 -0.55068 -0.19053 -0.07661 0.89956 1.07098
2 -0.70890 -0.01670 0.31698 0.85917 -0.80670
5 -0.18028 0.90966 0.33029 -1.01913 -0.29588
8 -0.82572 -0.15885 -0.62511 0.39793 -0.29853
8 -0.77082 -0.16581 -0.87887 0.37589 -0.31906
8 -0.82579 0.10809 -0.60336 0.31225 0.00858
8 -0.82313 -0.10623 -0.76878 0.89221 -0.28971
D -0.73927 -0.03215 0.10233 0.05951 -0.32583
6 -0.85733 0.25071 2.53169 1.86927 2.06818
5 0.12959 1.06068 0.81188 -0.82579 -0.27518
2 -0.57378 0.38791 0.03106 0.28823 -1.10558
1 0.12588 -1.39280 0.51088 -0.98800 0.11863
5 -0.26752 0.87898 0.65022 -0.17768 -0.00126

Initial

Cluster

mlahaUI«(nonpuna-mroquIUCIétjhaytuuouaninurmraunutuu:wtnpjntnoswtnanalnham

SPP

BAITOA
GUATAP
QUINA
BRUCON
CANDEL
CINAZO
ALMACIG
GUAYAC
CAMERON
AROMA
MOSTAZO
SANGRE
FRIJOL
CAFBTAN
PAAMAR
UVERO

Diml

0.121895
0.037388
0.083622
0.292232
0.260176
0.018809
0.001319
0.085815
0.050181
0.193378
0.108569
0.282180
0.533589
0.832327
0.773860
0.505397
0.856098
0.087080
0.066777
0.675076
0.368999
0.003810
0.679379
0.323562
0.681059
0.008379
0.000022
0.186671
0.353887
0.176016
0.262000
0.380681
0.078237
0.238062
0.011038
0.368888
0.376283
0.838367
0.329360
0.288096
0.082207
0.008616
0.073088
0.008827
0.050802

Diml

-0.68751
0.18387
0.91565

-0.35065
1.50285

-0.33323
0.08775

-0.18923

-0.78852

-0.68761

-0.80126
0.87063

-0.50008

-0.03519
1.28098

-0.80588

227

Squared Cosine. for the Row Points

Diu2

0.053985
0.001270
0.137583
0.006271
0.027313
0.318859
0.808858
0.006023
0.225573
0.033873
0.211356
0.158693
0.003019
0.082763
0.000211
0.002550
0.000588
0.639530
0.000188
0.105039
0.250929
0.003016
0.000886
0.007610
0.038616
0.300757
0.657163
0.288382
0.021295
0.011565
0.092399
0.000828
0.008888
0.000130
0.281086
0.013821
0.017827
0.006965
0.005886
0.000585
0.003609
0.309188
0.033391
0.581618
0.160119

Dim2

-0.09658
0.89923
-0.22913
-0.30758
0.05203
-0.10686
-1.79288
-0.16998
0.89899
0.88778
0.63288
0.58359
0.78157
-0.80502
0.03821
-0.28158

Dim3

0.193607
0.001710
0.078691
0.275888
0.322688
0.000856
0.000785
0.680807
0.220789
0.000508
0.018808
0.012860
0.035311
0.008800
0.019817
0.013686
0.009187
0.003887
0.008713
0.008570
0.018678
0.322338
0.005229
0.001283
0.011589
0.001958
0.096673
0.008761
0.195278
0.167220
0.003380
0.086228
0.001837
0.086798
0.037052
0.208873
0.182702
0.238018
0.288291
0.005520
0.368087
0.181183
0.000218
0.072912
0.300105

Column Coordinates

Dim3

-0.52988
0.20313
-0.02898
0.69758
-0.22326
0.38357
0.36787
-0.08538
0.27155
2.92878
0.18250
0.51308
-0.78959
-0.07858
-0.06563
0.61297

Dim8

0.061851
0.000112
0.188967
0.007300
0.072681
0.101682
0.077778
0.088888
0.139263
0.018332
0.020723
0.018108
0.106578
0.009159
0.108837
0.080031
0.085781
0.070068
0.000857
0.008083
0.067357
0.011919
0.118831
0.016802
0.015101
0.003078
0.187283
0.022708
0.283363
0.022288
0.008677
0.081783
0.198128
0.098201
0.352758
0.088682
0.089375
0.062678
0.117772
0.001867
0.200685
0.187818
0.018835
0.288989
0.022801

Dim8

0.27285
-0.67279
0.26708
-0.17921
0.38988
0.50867
-0.92150
0.32251
0.16058
1.05265
-1.18882
-0.18887
-0.78795
0.11898
0.20118
1.53512

Dims

0.027288
0.000019
0.851556
0.028778
0.188898
0.029585
0.005770
0.015568
0.050500
0.103887
0.335065
0.007818
0.002981
0.002268
0.001586
0.028368
0.000396
0.000086
0.015575
0.003188
0.038211
0.032865
0.000338
0.000281
0.001958
0.009577
0.000081
0.019526
0.001376
0.000002
0.387889
0.251529
0.280833
0.077039
0.029653
0.086370
0.068530
0.000018
0.080800
0.055826
0.288668
0.020812
0.271219
0.003671
0.000001

Dims

-0.10028

-0.20331
0.12002
-0.30229
0.08921
0.08685
0.07227
-0.01798
-1.10880
0.81068
0.87697
-0.36813
1.96852
-0.25299
-0.18269
1.55018

SPP

BAITOA
GUATAP
QUINA
BRUCON
CANDEL
CINAZO
ALHACIG
GUAYAC
CAMERON
AROMA
MOSTAZO
SANGRE
FRIJOL
CAFETAN
PAAMAR
UVERO

Dial

0.873890
0.019172
0.833885
0.071828
0.838639
0.085191
0.000526
0.018781
0.188938
0-032987
0.080095
0.038375
0.035676
0.000288
0.856610
0.061238

228

Squared Coainea for the Column Points

Dim2

0.010533
0.860578
0.027181
0.058955
0.001001
0.008612
0.780978
0.019117
0.059682
0.018688
0.099729
0.085859
0.078863
0.032897
0.000833
0.005501

Dim3

0.316878
0.023501
0.000323
0.282665
0.018830
0.059877
0.031129
0.008825
0.017663
0.673861
0.005057
0.080856
0.080170
0.001118
0.001277
0.035827

Dilu8

0.088186
0.257825
0.036877
0.018656
0.056202
0.103652
0.195751
0.068852
0.006178
0.087050
0.326180
0.005281
0.088586
0.002837
0.011996
0.222196

Dims

0.011357
0.023588
0.007886
0.053081
0.000895
0.003070
0.001208
0.000218
0.298275
0.051629
0.056658
0.020578
0.550658
0.012836
0.006037
0.226567

229
(11) CA scores from the analysis using 16 spp and 30 core sites

Inertia and Chi-Square Decomposition

Singular Principal Chi-
Valuea Inertia. Squares Percenta 5 10 15 20 25
----+----+----+—---+----+---

0.77763 0.60‘71 13319 25.90‘ 0*.Itltitiﬁtiﬁﬁititttiﬁﬁti

0.66993 0.88881 9885.31 19.228 ***"*""*"*'***‘

0.62099 0.38562 8893.59 16.528 ****'*'*********'

0.89850 0.28850 5873.38 10.688 "**'***"'

0.36962 0.13662 3009.19 5.858 8"**'

0.32383 0.10887 2309.78 8.898 *“'

0.31818 0.09868 2173.53 8.238 "**

0.29566 0.08781 1925.31 3.788 '**'

0.28682 0.06092 1381.78 2.618 '8'

0.21163 0.08879 986.885 1.928 '*

0.19812 0.03768 829.966 1.618 8'

0.16598 0.02758 606.885 1.188 8

0.15726 0.02873 588.73 1.068 *

0.11876 0.01317 290.097 0.568 9

0.10361 0.01078 236.861 0.868

2.33879 51825.1 (Degree. of Freedom - 835)

CLUSTERS DIMl DIMZ DIM3 SUMC012 SUMC0123 800C081 800C082 890C083
2 -0.888 0.255 0.088 0.070 0.070 0.053 0.017 0.001
8 -0.797 0.138 -0.853 0.885 0.636 0.871 0.018 0.152
8 -0.598 0.078 -0.308 0.237 0.298 0.233 0.008 0.061
5 -0.226 0.868 0.223 0.527 0.560 0.038 0.898 0.032
1 -0.221 -0.905 0.397 0.783 0.878 0.082 0.701 0.135
5 -0.331 1.136 1.005 0.355 0.611 0.028 0.327 0.256
2 -0.876 0.310 -0.327 0.192 0.216 0.171 0.021 0.028
5 0.537 0.558 0.128 0.887 0.859 0.217 0.230 0.012
3 1.009 -0.277 -0.597 0.896 0.658 0.861 0.035 0.162
3 1.138 0.222 -0.232 0.760 0.790 0.732 0.028 0.031
3 1.828 -0.260 -0 666 0.780 0.896 0.716 0.028 0.156
3 0.856 -0.286 —0.386 0.871 0.582 0.835 0.036 0.071
3 1.326 -0.118 -0.819 0.807 0.887 0.801 0.006 0.080
1 0.061 -0.911 0.323 0.573 0.685 0.003 0.571 0.072
5 0.762 0.792 0.232 0.658 0.683 0.318 0.380 0.029
3 1.255 -0.119 -0.606 0.611 0.753 0.606 0.005 0.181
3 1.083 0.063 -0 351 0.317 0.352 0.315 0.001 0.036
3 1.130 0.165 -0.365 0.608 0.670 0.595 0.013 0.062
1 -0.112 -1.367 1.338 0.878 0.932 0.003 0.878 0.855
8 —0.968 -0.168 -0.738 0.877 0.786 0.863 0.018 0.269
2 -0.776 0.121 0.028 0.203 0.203 0.198 0.005 0.000
5 -0.167 1.152 0.931 0.885 0.730 0.009 0.836 0.285
8 -1.010 -0.182 -0 633 0.538 0.785 0.527 0.010 0.207
8 -0.935 -0.138 -0.505 0.581 0.696 0.530 0.011 0.158
8 -0.953 0.016 -0.713 0.880 0.687 0.880 0.000 0.286
8 -1.033 -0.128 -0.808 0.513 0.822 0.506 0.007 0.309
5 0.158 1.357 0.950 0.537 0.796 0.007 0.530 0.259
2 —0.718 0.818 -0.299 0.126 0.182 0.098 0.032 0.017
1 0.050 -1.051 1.172 0.388 0.861 0.001 0.383 0.876
5 -0.238 0.796 0.588 0.293 0.838 0.023 0.270 0.185
BA —0.885 -0.118 -0.560 0.680 0.916 0.629 0.011 0.276
GU 0.190 1.101 0.578 0.657 0.830 0.019 0.638 0.178
QU 0.980 -0.336 -0.207 0.512 0.538 0.858 0.058 0.022
BR -0.298 -0.123 0.781 0.058 0.398 0.089 0.008 0.380
CA 1.262 -0.150 -0.502 0.797 0.921 0.786 0.011 0.128
CI -0.288 0.181 -0.136 0.038 0.080 0.027 0.006 0.006
AL -0.108 -1.867 1.150 0.600 0.968 0.003 0.597 0.367
CY -0.278 -0.198 -0.202 0.095 0.128 0.062 0.032 0.038
08 -0.898 0.550 -0.020 0.192 0.192 0.139 0.053 0.000
AR -0.209 1.213 1.099 0.287 0.888 0.007 0.280 0.197
HO -0.290 0.860 1.153 0.176 0.860 0.018 0.158 0.288
3A 0.363 0.671 0.171 0.131 0.138 0.030 0.102 0.007
FR -0.331 0.003 -0.526 0.022 0.078 0.022 0.000 0.056
CF -0.180 0.050 0.323 0.011 0.082 0.010 0.001 0.032
PA 1.232 0.000 -0.381 0.538 0.585 0.538 0.000 0.051
UV ~0.885 -0.287 —0.263 0.158 0.166 0.139 0.015 0.012

APPENDIX I

CDA scores from the initial analyses

 

230

231

APPENDIX I

(1) CDA scores from analysis of 16 spp and 43 sites

Canonical Discriminant Analysis

82 DF Total
38 DF Within Classes
8 DP Between Classes

83 Observations
16 Variables
5 Classes

Class Level Information

Statistic

Wilks’

Lambda

CLUSTEMP Frequency Weight Proportion
l 8 8.0000 0.093023
2 8 8.0000 0.186087
3 10 10.0000 0.232558
8 10 10.0000 0.232558
5 11 11.0000 0.255818

Pillai’s Trace

Hotelling-Lawley Trace
Roy’s Greatest Root

Multivariate Statistics and F Approximations

S=8

Value

0.00015391
3.83378835

80.92173632
21.11986598

M-5.5

N-10.5
F Num DP Den DP
12.1813 68 92.31698
9.8588 68 108
13.7871 68 86
38.3191 16 26

NOTE: F Statistic for Roy’s Greatest Root is an upper bound.

Test of H0: The canonical correlations in the current row

Adjusted Approx Squared
Canonical Canonical Standard Canonical
Correlation Correlation Error Correlation
1 0.977138 0.965886 0.006976 0.958791
2 0.958511 0.982672 0.012538 0.918783
3 0.927586 0.907639 0.021538 0.860816
8 0.836568 0.796772 0.086316 0.699839
Eigenvalues of INV(E)*H
s Canqu/(l-Canqu)
Eigenvalue Difference Proportion Cumulative
1 21.1195 9.8129 0.5161 0.5161
2 11.3066 5.1825 0.2763 0.7928
3 6.1681 3.8326 0.1506 0.9830
8 2.3315 . 0.0570 1.0000

Likelihood
Ratio Approx F Num DF Den DP
1 0.00015391 12.1813 68 92.31698
2 0.00380850 9.2861 85 72.07888
3 0.08189779 6.9383 28 50
8 0.30016099 8.6631 13 26

Pr > P

0.0001
0.0001
0.0001
0.0008

Pr > P

0.0001
0.0001
0.0001
0.0001

and all that follow are zero

OBS

\DOQGUI‘DWNH

830

CLUSINI

midnrmtjsse.sasmtouawtntintjhawisu:w(juantq+swcsu:wcsuerourptno-btunrm

CLUSTEMP

mwwmnoobemwmwwoonwwwwwwmwmwwwwwwmmNmI-Jmobmwm

232

CANl

-2.1356
-0.5228
-2.8857

1.1229

0.9988
-2.3890
11.5606
-2.8271
-l.5852
-2.0679
-8.0018
-2.0905
-2.8861
-0.9913
-2.2267
-1.1001
11.0868
-8.2270
-2.7827
-2.6393
—0.9253
-1.3665
-2.2199
-1.8233

1.7931
15.3105

0.7271

1.2026
-0.6187
-1.0193
-1.3826
-0.2918
-1.7553
-2.9856

1.2733

0.9295
-0.2705

1.7185
-0.0086
-8.8973
-0.6958
13.0957
-8.5688

CANZ

0.02905
2.87778
0.60001
8.06835
2.28029
0.99886
-1.01991
-l.08899
1.75917
1.15122
-1.62957
-8.95521
-8.66582
-6.08820
-8.16199
-5.55769
-2.30885
0.17260
-3.52391
-2.28376
-3.90821
-5.21338
-3.85882
-8.75386
2.22571
-2.27851
2.86527
8.76197
3.01869
2.53706
2.11528
1.61389
2.98602
1.22737
5.61366
5.21788
8.28377
8.60263
8.08803
-0.98757
0.18688
-2.70298
0.81827

CAN3

-0.32766
—3.56876
-0.05177
2.75282
1.82393
-1.89097
0.02857
-0.79220
-8.23983
-3.09086
-0.67168
2.60887
2.08372
2.01173
2.18819
2.69351
-0.36167
-2.20957
0.92716
-0.55383
1.25282
2.01110
1.21887
-0.81208
-2.97753
-1.87888
2.21357
3.30812
2.51210
-2.10787
-3.22208
-8.03920
-8.13910
-2.78296
5.08867
3.71237
1.30881
3.85992
8.11888
-1.00121
-3.05238
-2.18808
-1.27093

CAN8

0.52805
-2.88013
2.58331
1.17133
1.38887
2.11183
0.81812
2.61608
-2.67998
0.95086
1.27822
-1.31058
0.59820
-1.75806
-1.09098
-0.98119
1.25990
1.01798
0.60085
1.60180
-0.85568
-2.26789
-2.30827
-0.23283
-0.08312
0.90617
-0.02202
-0.55185
-0.18806
-0.07185
-2.17809
-2.52806
-2.91397
3.36988
-0.83851
-0.62137
-0.96087
-0.l8172
-0.60862
8.38827
-2.55963
0.50130
2.18683

233

(II) CDA scores from analysis of 16 spp and 30 core sites

Canonical Discriminant Analysis

29 DP Total
25 DP Within Classes
8 DP Between Classes

30 Observations
16 Variables
5 Classes

Class Level Information

CLUSINI Frequency Weight Proportion
1 8 8.0000 0.133333
2 8 8.0000 0.133333
3 8 8.0000 0.266667
8 7 7.0000 0.233333
5 7 7.0000 0.233333

Multivariate Statistics and P Approximations

S-8 M-5.5 N-8
Statistic Value P Num DP Den DP Pr > P
Wilks’ Lambda 0.00000293 16.1211 68 81.82823 0.0001
Pillai’s Trace 3.78007736 11.6912. 68 52 0.0001
Hotelling-Lawley Trace 176.55852938 23.8892 68 38 0.0001
Roy’s Greatest Root 132.75869213 107.8668 16 13 0.0001

NOTE: P Statistic for Roy’s Greatest Root is an upper bound.

Adjusted Approx Squared
Canonical Canonical Standard Canonical
Correlation Correlation Error Correlation

1 0.996255 0.993955 0.001388 0.992528

2 0.980087 0.967797 0.007322 0.960571

3 0.968889 0.950178 0.012811 0.931010

8 0.925188 0.901885 0.026785 0.855973

Eigenvalues of INV(E)*H
= Canqu/(l-Canqu)
Eigenvalue Difference Proportion Cumulative

1 132.7587 108.3968 0.7519 0.7519

2 28.3619 10.8670 0.1380 0.8899

3 13.8988 7.5517 0.0768 0.9663

8 5.9831 0.0337 1.0000

Test of H0: The canonical correlations in the current row and all that follow are zero
Likelihood

Ratio Approx P Num DP Den DP Pr > P
1 0.00000293 16.1211 68 81.82823 0.0001
2 0.00039179 9.6828 85 33.85879 0.0001
3 0.00993688 7.7817 28 28 0.0001
8 0.18802715 5.9831 13 13 0.0015

 

 

OBS

\OQQMU'IubWNH

CLUSINI

mwwmoneemwewwuwmr—awwwwwmwun-Imuss-N

234

CANl

-5.7318
3.5366
8.7931

-7.7223

23.2065

-8.6893

-5.0583

-9.5820

-8.9508

-5.0305

-6.2952

-8.8376

-5.8222

23.6176

-7.9861

-8.0350

-5.9668

-6.8301

26.8906
5.7136

-6.6785

-7.7875
8.2505
2.9232
3.5088
5.8936

~8.9052

-5.9857

22.8852

-8.6882

CAN2

0.96066
6.85705
6.07886
2.29787
-1.91667
0.85877
1.80282
-0.58357
-5.38835
-8.18571
-5.63288
-8.06398
-8.22188
~8.82719
-1.29038
-5.20395
-8.25557
-8.76921
-8.75889
6.52958
0.07878
0.56963
8.71183
7.36280
7.57275
7.25337
0.16356
-0.78807
-5.80218
0.03805

CAN3

-8.73807
1.80729
3.77552

-3.86619

-O.67711

-2.16883

-5.79960

-2.01576
8.87368
3.15657
8.55897
5.20120
8.02973

-0.70861

-0.87186
3.78803
2.62518
2.19553

-3.25678
2.00115

-5.95695

-3.33828
3.59516
2.79155
1.35829
2.81519

-2.68828

-5.81517

-3.18578

-2.73015

CAN8

-3.29251
1.23832
0.50796
2.25516
0.59856
3.38070

-8.36082
2.98855

-0.78638
1.25197

-1.78836

-0.76039
0.03826
0.88129
2.20807

-1.50507

-2.68825

-0.58162
1.06375

-0.52086

-8.81819
2.96502

-0.18628

-0.95586

-1.88071

-0.93892
5.79967

-8.37020
0.68238
3.01758

APPENDIX J

SAS output for correspondence analyses using the full data set

 

235

ID

101
102

108
105
106
107
108
109
110
111
112
113
118
115
116
117
118
119
120
121
122
123
128
125
126
127
128
129
130
201
202
203
208
205
206
207
208
209
210
211
212
213
218

Singular

Va

OOOOOOOOOOOOOOO

CLUSTOT

OIDUIUII~JU
U13 0 0

m
0

OIUO\HIMGIM!301wIﬂF‘UIﬂ DISOWSUIQ
:1: «>8

0

xtdhiwldhlutuﬂiwlﬂh’wtdu)

236

APPENDIX J

(I) Results for CA procedure using 120 sites

lues

.72080
.68282
.58279
.88965
.86161
.83802
.39182
.38930
.36991
.35335
.31169
.29828
.29361
.22378
.21527

The Correspondence Analysis Procedure

Inertia and Chi-Square Decomposition

Principal Chi-

Inertias

.51897
.81270
.29863
.23975
.21309
.18838
.15321
.15156
.13683
.12886
.09715
.08658
.08621
.05006
.08638

2.80031

OOOOOOOOOOOOOOO

Squares Percents

51676
81098.2
29336.9
23873.1
21217.7
18757.2
15255.7
15091.1
13628.9
12832.7
9673.88

8620.7
8588.13
8988.62
8618.38

18.538
18.788
10.528

8.568
7.618
6.738
5.878
5.818
8.898
8.868
3.878
3.098
3.088
1.798
1.658

8 8 12 16 20

.......................

..................
.............
...........
..........
........
.......
.......
......

......

....

....

....

..

..

278837 (Degrees of Freedom - 1785)

Row Coordinates

Diml

0.83163
-0.19395
-0.82118
-0.58903
-0.81682
-0.05228

0.68376
-0.85925
-0.71888
-0.08710
-0.57776
-0.68818
-0.80751
-0.55920
-0.83852
-0.89153
-0.31853

0.18570
-0.01727

1.22879
-0.50971
-0.87568
-0.16765
~0.57228
-0.38336

0.10710
-0.55876
-0.29627
-0.67610
-0.81169
.61088
.63111
.22323
.11038
.08886
.75170
.55832
.55681
.07189
.31210
.83860
.83707
.87998
.59097

OOHHHNHHOHOHHH

Dim2

0.27358
-0.18582
0.05908
0.08763
-0.07573
0.22888
0.52167
-0.69233
-0.07893
0.72988
-0.88808
0.10133
-0.72888
0.88857
0.58878
0.16973
0.02370
-0.85590
0.68771
0.17633
1.21021
0.28818
0.65561
1.51896
1.58192
-0.89983
0.92820
1.05819
0.58738
1.02953
0.15229
0.88616
0.50039
0.18636
0.33910
0.56188
0.09283
0.81990
0.16996
0.07076
0.17787
0.18822
-0.88505
0.11799

Dim3 Dim8 Dims
-0.33312 -0.15258 0.02828
-0.11556 -0.00969 ~0.39785
-0.81291 -0.66895 1.88205
-0.87192 -0.32589 -0.00895
-0.36859 -0.37805 0.58889
-0.25152 -0.32067 0.11708

0.08652 -0.35816 0.87389
~0.35928 -0.27186 1.07220
-0.70822 -0.16889 -0.00282

0.27822 -0.89958 0.36060
-0.31009 -0.13218 -0.08780
-0.85857 0.05286 -0.19871

0.00528 -0.16856 -0.08856

0.20069 0.61287 -0.28630

0.80676 -0.02178 0.28089
-0.18173 1.19299 0.60053
-0.58282 -0.28672 -0.08779

0.38958 -0.18397 0.18283

0.87190 -0.28395 -0.05875
-O.39866 0.12286 -0.13166

0.93885 0.18172 0.02009
-0.15828 -0.29513 0.08201

0.80600 -0.38988 -0.29659

1.89767 0.38121 -0.29523

1.38373 -0.27966 0.05809

0.32896 0.09851 0.08878

1.13698 0.78009 -0.89233

0.91875 -0.50580 0.08858

0.17699 -0.22368 -0.15107

0.60898 -0.55866 -0.02128
-0.87112 0.22189 -0.36987
-0.18726 -0.08127 0.18738
-0.31958 -0.17878 0.07652
-0.38828 -0.21087 0.02627
-0.21205 -0.08587 0.03810

0.23798 -0.23182 0.18305
-0.S8760 0.26758 -0.03071
-0.22730 -0.05731 0.06326
-0.66667 0.25011 -0.07123
-0.80919 0.16896 0.10613
-0.38268 0.25225 -0.03830
-0.37898 0.15186 -0.02S88

0.22386 -0.10981 -0.00067
-0.28502 —0.08533 -0.28182

ID

215
216
217
218
219
220
221
222
223
228
225
226

CLUSTOT

Q

C)

C)

9‘10th84WHmeQQQQWUMQmeQUhUI-‘QHWQHwwww’leQOI-‘UHO

waGDUI¢¢stm~4hIU~Js-qtjn!Ot3><gtn~Juiecscjetjuss.g

Diml

1.26719
0.76232
2.11251
-0.00360
0.82919
-0.00965
1.07085
0.30802
1.87159
1.89372
1.70878
1.20087
0.19856
0.07679
1.88867
0.58585
-0.06807
0.18231
-0.27381
-0.68086
-0.31698
-0.38698
-0.88838
-0.13582
0.12287
-0.31508
-0.33171
-0.58851
-0.39803
-0.36579
-0.58010
-0.89315
-0.73807
-0.39581
-0.57800
-0.10807
-0.09812
-0.36169
-0.82638
-0.57990
-0.68675
-0.15959
-0.62523
-0.38508
-0.63391
-0.25022
-0.73807
—0.66786
-0.72838
-0.73029
—0.62053
2.09883
-0.76075
-0.82562
0.21689
0.13582
-0.83636
0.00899
-0.18182
0.55076
-0.81552
-0.78360
-0.12386
-0.75859
0.03805
-0.73859
0.18865
-O.18628
-0.36301
0.50871
0.05399
-0.68186
0.21005
—0.83225
"0.30895
-0.17820

237

DimZ

0.52981
-1.89308
0.25087
-0.79589
0.86855
-0.80698
0.77635
0.38513
0.18729
0.03395
0.88702
-0.30280
-0.51875
-0.13285
0.23108
—1.27585
-0.66300
-1.37171
-0.26011
-0.12836
-0.10822
-0.18308
1.38336
1.18129
1.31386
0.59789
1.80228
0.03159
0.17896
0.03878
-0.28015
0.62878
0.27108
-0.09836
0.82958
0.89869
-1.63220
-0.61951
-0.18599
0.02577
-0.09639
-0.37726
-0.09516
-0.08688
-0.06960
0.30271
-0.06830
-0.01397
0.03090
-0.03832
0.18662
0.30628
-0.18938
0.89701
-0.82837
0.52166
-0.00378
0.16977
-0.08891
-0.81861
0.25800
0.81099
-0.57771
0.21078
-0.33605
0.15851
-1.73183
-0.09258
0.57356
-0.32861
-0.19361
0.93859
1.05358
0.70788
-1.19757
0.59822

Dim3

0.08877
0.93885
-0.68855
0.58152
0.32808
-0.17092
0.19739
0.71779
-0.59818
-0.31292
-0.82355
-0.23382
0.72108
0.13093
-0.33520
1.21556
0.06018
1.07755
-0.16379
-0.70007
-0.58218
-0.80738
1.38902
0.97803
0.96780
0.52151
1.05887
-0.37978
-0.25682
-0.26892
-0.21201
0.17308
-0.29157
-0.30588
0.15788
0.65383
0.87781
-0.18951
-0.37761
-0.81896
-0.73667
0.03167
-0.69176
0.01303
—0.61295
0.20360
-0.72081
-0.60036
-0.53309
-0.73216
-0.18998
-0.73588
-0.86838
0.05307
-0.17885
0.11768
-0.38913
-0.11890
-0.25286
0.87171
-0.10108
-0.09585
0.05591
-0.32701
-0.20773
-0.37227
1.29511
-0.00183
0.89086
0.66600
0.23888
1.01053
0.85956
0.30732
0.98979
0.81763

Dim8

-0.15276
-0.08768
0.20107
-0.13282
0.07697
0.10262
-0.83173
0.32079
0.25310
0.29879
-0.10805
0.26281
-0.07582
0.51832
0.16918
0.08320
0.05985
-0.06531
0.06783
0.33508
-0.06281
-0.00368
0.88223
-0.26750
-1.00118
-0.10968
-0.63882
0.08818
0.21557
-0.03735
0.39628
-0.36326
-0.16875
1.58783
0.88858
-0.75805
-0.21097
—0.36188
0.12610
0.18553
-0.09279
-0.02801
-0.10527
0.28881
-0.09892
-0.12881
-0.13567
-0.12319
-0.18878
-0.15537
-0.12578
0.15706
-0.16193
-0.25258
0.17637
—0.80977
-0.28888
-0.15828
-0.12806
0.08621
-0.28555
0.05990
0.89061
0.09288
0.06973
0.17338
-0.06258
0.15888
3.86165
0.23213
0.05175
2.96183
-0.58283
-0.17075
-0.05666
-0.19876

Dims

0.13068
0.03158
-0.09220
-0.18583
0.08690
-0.52312
0.21888
-0.88l92
-0.25509
-0.12878
0.28718
0.01258
-0.28720
-0.95202
-0.07581
0.09578
-0.51839
-0.17210
-0.86189
-0.18008
-0.15078
-0.83181
-0.20595
0.22006
0.70606
-0.88001
0.08259
-0.30928
0.00228
-0.20330
-0.30097
-0.17151
-0.82889
0.72518
-0.38961
0.32538
0.17118
1.20319
-0.85878
0.98039
-0.38678
-0.15888
-0.88631
-0.56572
-0.85908
-0.83608
-0.36359
-0.37928
0.06585
-0.21260
-0.50088
-0.03051
-0.21065
-0.05388
-0.88581
0.15528
0.51255
-0.08639
-0.11688
-0.32870
-0.05668
-0.19662
-0.21258
-0.33300
-0.56558
-0.06l77
0.09817
-0.80857
1.20385
-0.20639
—0.51626
0.92565
0.30850
—O.80008
-0.05199
-0.10558

 

 

 

CLUSTOT

nuaurwra>twtsuawtshawtnuogtsuawtnclulmtaa:ag§t1Mtsurwrjggatat3s~uins-gtugggpazg~38

Dim1

0.258683
0.015155
0.031381
0.268899
0.198305
0.008095
0.116566
0.083017
0.821578
0.003819
0.388563
0.378262
0.157522
0.115018
0.021293
0.079372
0.077261
0.028526
0.000168
0.359378
0.075219
0.168611
0.009596
0.039038
0.022959
0.010110
0.038678
0.028908
0.118886
0.058868
0.618589
0.789996
0.519839
0.023391
0.753925
0.270516
0.583096
0.728789
0.788827
0.568536
0.752896
0.788895
0.180132
0.088680
0.608272
0.180975
0.720303
0.000010
0.260862

238

Dim2

0.102299
0.013852
0.000617
0.006850
0.006558
0.078866
0.076588
0.188665
0.008637
0.267862
0.233670
0.008211
0.897820
0.087795
0.039091
0.009865
0.000828
0.605988
0.266766
0.007889
0.828035
0.088809
0.186752
0.278988
0.863003
0.713680
0.108275
0.317711
0.075031
0.383138
0.005896
0.070181
0.086922
0.081158
0.079811
0.151125
0.002069
0.052726
0.005281
0.001682
0.007077
0.008235
0.876866
0.001939
0.105871
0.580796
0.010126
0.885019
0.083168

Dim3

0.151672
0.005380
0.030162
0.198673
0.152012
0.098791
0.000609
0.050795
0.818228
0.038968
0.111928
0.165231
0.000026
0.018818
0.018659
0.010850
0.228082
0.125526
0.125608
0.037318
0.255197
0.018215
0.056280
0.267331
0.351625
0.095378
0.162862
0.239899
0.007886
0.118885
0.052597
0.010812
0.035856
0.289687
0.031052
0.027105
0.072003
0.015850
0.081286
0.058905
0.032752
0.033387
0.030373
0.011318
0.000758
0.212007
0.075636
0.228761
0.039777

Squared Cosines for the Row Points

Din8

0.031808
0.000038
0.078220
0.098510
0.160008
0.158078
0.036080
0.029091
0.023886
0.125683
0.020337
0.002201
0.026951
0.138151
0.000053
0.867562
0.086350
0.027999
0.085879
0.003616
0.005815
0.063367
0.051899
0.017320
0.015230
0.008552
0.076878
0.072587
0.012531
0.099595
0.011625
0.000506
0.010600
0.085108
0.001853
0.025680
0.017188
0.000982
0.011835
0.009361
0.018238
0.005361
0.007338
0.001018
0.008781
0.000551
0.006526
0.013881
0.002288

Dims

0.001093
0.063682
0.626621
0.000071
0.390698
0.020528
0.063058
0.852898
0.000007
0.065862
0.002660
0.031575
0.006782
0.030188
0.006522
0.118878
0.001739
0.027530
0.001987
0.008153
0.000117
0.001288
0.030033
0.010388
0.000570
0.002098
0.100067
0.000563
0.005716
0.000186
0.032818
0.010821
0.002033
0.001326
0.001002
0.009796
0.000226
0.001197
0.000928
0.003693
0.000263
0.000150
0.000000
0.011030
0.006827
0.000282
0.001372
0.016300
0.002861

 

 

CLUSTOT

‘nh‘Otﬂ0\Oiim~Jb'U~dlrdlﬂﬂiﬂld>ﬂgtn~lebwtjlbddhabadbslb~dlbg‘dggthCIBGDQ~J\lwt3Ula‘n0!m~atlﬁ!3hhdhiU~JHIthwtdzgm~4

Dim1

0.000088
0.383780
0.028182
0.753179
0.878707
0.681651
0.385785
0.017993
0.000817
0.753058
0.077588
0.001538
0.005103
0.069880
0.330266
0.098620
0.078867
0.033862
0.008692
0.002623
0.082827
0.020893
0.882538
0.219358
0.228287
0.318752
0.155179
0.270302
0.081859
0.183008
0.003500
0.002292
0.087809
0.082331
0.131608
0.287521
0.017388
0.239585
0.110236
0.288811
0.008268
0.385580
0.380897
0.392389
0.323893
0.207058
0.603589
0.303605
0.222891
0.035895
0.017205
0.117290
0.000028
0.007837
0.085312
0.182866
0.331009
0.012882
0.358029
0.000781
0.375652
0.006890
0.009798
0.008378
0.107811
0.000883
0.033117
0.010909
0.033657
0.030298
0.023872

239

D102

0.088910
0.201867
0.038781
0.008665
0.000285
0.055378
0.028860
0.127907
0.002885
0.018237
0.828022
0.168729
0.878087
0.063026
0.011032
0.011896
0.017655
0.309503
0.356986
0.301663
0.158255
0.373357
0.001275
0.085286
0.002028
0.063022
0.289069
0.036852
0.002560
0.080081
0.282018
0.689258
0.139088
0.008963
0.000260
0.005663
0.096985
0.005589
0.001638
0.002998
0.012101
0.002651
0.000189
0.000706
0.000718
0.018727
0.012858
0.011700
0.303932
0.137813
0.255298
0.000009
0.027610
0.002681
0.025678
0.070388
0.091060
0.273300
0.027935
0.057802
0.017302
0.605791
0.002818
0.020916
0.088830
0.010888
0.062290
0.278835
0.090165
0.867271
0.260983

Dim3

0.018979
0.013050
0.138578
0.076927
0.020833
0.081882
0.018578
0.287188
0.002375
0.038385
0.385136
0.001357
0.292558
0.028992
0.389583
0.288886
0.087822
0.312116
0.282709
0.163688
0.117360
0.211291
0.188253
0.092893
0.121220
0.089115
0.019106
0.082183
0.028687
0.010818
0.127986
0.199360
0.013016
0.033201
0.068698
0.330835
0.000683
0.293285
0.000126
0.232256
0.005878
0.333169
0.275798
0.210188
0.325185
0.019808
0.078222
0.395559
0.003865
0.028808
0.012991
0.093278
0.012688
0.023776
0.033237
0.010798
0.008911
0.002560
0.067238
0.022086
0.095832
0.338982
0.000001
0.015319
0.187028
0.015959
0.072828
0.182678
0.017013
0.293916
0.128915

D1m8

0.005800
0.062826
0.026879
0.013778
0.018998
0.002726
0.018819
0.002708
0.037221
0.009778
0.000887
0.001328
0.001075
0.008286
0.080088
0.003873
0.000007
0.033581
0.018306
0.175270
0.005187
0.076825
0.002889
0.065655
0.002339
0.171569
0.088196
0.018130
0.667172
0.087321
0.172198
0.011516
0.087889
0.003702
0.008288
0.005288
0.000393
0.006791
0.060315
0.005570
0.002191
0.011803
0.011613
0.025281
0.018682
0.008501
0.003381
0.013755
0.078867
0.023736
0.157526
0.036819
0.023989
0.006098
0.001110
0.063720
0.001935
0.197103
0.005819
0.002889
0.020690
0.000790
0.007092
0.761883
0.022720
0.000775
0.625598
0.083872
0.005252
0.001086
0.029198

Squared Cosine. for tho Raw Points, continued.

Dims

0.180312
0.016080
0.060662
0.013992
0.003313
0.018259
0.000082
0.029086
0.125570
0.001963
0.002391
0.100707
0.007863
0.198397
0.013997
0.022308
0.098080
0.007275
0.012389
0.087170
0.099823
0.001295
0.122189
0.000007
0.069283
0.098982
0.018770
0.089580
0.139188
0.065888
0.031726
0.007581
0.528635
0.171708
0.376158
0.073311
0.016287
0.122065
0.237961
0.130261
0.092315
0.088768
0.110075
0.003207
0.027816
0.138689
0.000128
0.023279
0.003520
0.151653
0.022610
0.161827
0.007189
0.005082
0.016139
0.003395
0.020880
0.036992
0.069723
0.163703
0.002627
0.001792
0.188610
0.092188
0.017962
0.077100
0.061103
0.022926
0.028829
0.000881
0.008239

 

 

BAITOA
GUATAP
QUINA
BRUCON
CANDEL
CINAZO
ALMACIG
GUAYAC
CAMERON
AROMA
MOSTAZO
SANGRE
FRIJOL
CAFETAN
PAAMAR
UVERO

BAITOA
GUATAP
QUINA
BRUCON
CANDEL
CINAZO
ALMACIG
GUAYAC
CAMERON
AROMA
MOSTAZO
SANGRE
FRIJOL
CAFETAN
PAAMAR
UVERO

Dim1

—0.56388
0.18860
1.09298

-0.22888
1.80517

-0.13180
0.18081

-0.00637

-0.66983

-0.51687

-0.28821
0.63865

-0.87978

—0.16019
1.13969

-0.60378

240

Column Coordinates

Dim2

-0.10253
0.90909
-0.11399
0.12772
0.22866
-0.17965
-1.55519
0.07627
1.07832
1.10830
0.08676
0.08007
-0.32853
-0.21853
-0.18838
0.30815

Squared Coaines

Dim1

0.865889
0.016897
0.858279
0.018269
0.825918
0.006910
0.005678
0.000020
0.093362
0.080020
0.018802
0.072608
0.036181
0.008812
0.223710
0.030209

Dim2

0.015806
0.800115
0.008985
.005916
.012792
.012837
.696898
.002858
.280168
.184003
.000653
.001156
.016557
.007913
0.003590
0.007870

OOOOOOOOOOO

Dim3

-0.89623
0.55728
-0.03999
0.11708
-0.50362
0.07817
0.95526
-0.09928
0.70235
1.18580
0.15887
0.10695
-0.57018
-0.33177
-0.25888
0.17169

Dim3

0.360868
0.150335
0.000613
0.008968
0.068283
0.002188
0.262778
0.008839
0.102687
0.210896
0.007501
0.002062
0.051102
0.018923
0.011185
0.002883

Dim4

-0.07037
-0.53777
0.30760
0.00608
0.08321
0.23815
-0.12015
-0.03131
-0.22569
0.79231
-0.36396
0.12011
-0.37222
0.15288
0.08697
3.01672

for the Column Points

Dim8

0.007256
0.180013
0.036296
0.000013
0.001755
0.022559
0.008157
0.000882
0.010599
0.098037
0.039569
0.002600
0.021781
0.003997
0.001303
0.758239

Dims

-0.11187
0.38805
-0.02810
-0.57602
0.01039
-0.62816
0.10163
0.15656
-0.28760
-0.31697
0.28936
-0.28828
1.92636
-0.88851
-0.39823
1.13811

DimS

0.018211
0.072903
0.000303
0.120335
0.000027
0.158951
0.002975
0.012085
0.012757
0.015051
0.018573
0.010756
0.583378
0.033970
0.027313
0.106599

 

241

(II) Results of CA procedure for 67 core sites.

The Correspondence Analysis Procedure

Inertia and Chi-Square Decomposition

Singular Principal Chi-
Values Inertias Squares Percents 12
----+----+----+----+—---+---
0.75483 0.56976 32661.9 20.78‘ tttttttttttttttttttttﬁtit.
0.68438 0.46832 26846.7 17.08% '*******'**'*********
0.61642 0.37998 21782.3 13.86% *'************8**
0.56887 0.32361 18551.4 11.80% '**************
0.46979 0.22070 12651.8 8.05% **********
0.43505 0.18927 10849.8 6.90% *********
0.34491 0.11896 6819.62 4.34% '****
0.30906 0.09552 5475.75 3.48% ****
0.29361 0.08621 4941.86 3.14% ****
0.27560 0.07596 4354.31 2.77% ***
0.25540 0.06523 3739.31 2.38% '**
0.23938 0.05730 3284.78 2.09% ***
0.19638 0.03857 2210.82 1.81% **
0.17325 0.03002 1720.73 1.09% *
0.15150 0 02295 1315.82 0.84% *
2.74235 157207 (Degrees of Freedom - 990)
Row Coordinates
CLUSTOT ID Dim1 Dimz Dim3 Dim4 Dims
1 118 -0.00955 -0.58550 0.42091 —0.20944 -0.09184
1 126 -0.12839 -0.60824 0.38635 0.13021 -0.07725
1 213 0.16309 -0.66852 0.24699 -0.05803 0.08211
1 216 0.43354 -1.09950 0.96011 -0.09051 0.23366
1 218 -0.17026 -0.48114 0.48194 -0.28288 -0.11205
1 227 0.08573 -0.23325 0.63850 -0.38428 -0.18915
1 230 0.28906 -0.76371 1.16490 —0.10770 0.43648
1 302 -0.07570 -0.94572 1.01496 -0.14754 0.17292
1 321 ~0.31354 -1.13066 0.90124 -0.21509 0.05764
1 421 -0.01300 -1.16448 1.35905 -0.23497 -0.02342
1 429 0.10857 —0.84191 0.94990 -0.20575 0.09043
3 120 1.03161 0.07481 -0.47345 0.18625 -0.06162
3 201 1.17770 -0.10902 -0.47617 0.35842 0.33260
3 202 1.35532 0.37947 -0.21997 -0.06203 -0.10646
3 203 0.94292 0.41156 -0.35029 -0.17048 -0.15501
3 205 0.85023 0.27457 -0.27423 -0.05190 -0.14319
3 207 1.16812 -0.10912 -0.51948 0.36146 0.23536
3 208 1.23710 0.29386 -0.24300 -0.05943 -0.05377
3 209 1.57742 -0.12871 —0.60263 0.43299 0.38292
3 210 1.00571 -0.08215 -0.37822 0.28254 0.18369
3 211 1.45571 -0.02343 -0.37423 0.30527 0.21120
3 212 1.42479 -0.00767 -0.35687 0.22809 0.21161
3 217 1.66752 -0.02366 -0.65884 0.39918 0.28311
3 223 1.41579 -0.11180 -0.56404 0.37133 0.34158
3 224 1.26807 -0.09248 -0.36416 0.34404 0.11358
3 225 1.37173 0.31760 -0.40062 0.00619 0.00663
3 226 0.87140 —0.38175 -0.15371 0.30147 0.22479
3 229 1.19439 0.07564 -0.36093 0.18082 0.07695
3 406 1.65100 0.01883 -0.70025 0.38213 0.27078
4 109 -0.86641 -0.04843 -0.74485 —0.00997 0.01969
4 112 —0.85398 0.21004 -0.52549 0.06410 0.09605
4 304 -0.88700 -0.02956 -0.60732 0.83816 -0.29593
4 325 -0.90598 -0.09296 -0.78730 -0.00441 -0.06621
4 327 -0.84058 -0.12109 -0.77015 0.02216 -0.11703
4 329 -0.82294 -0.09509 -0.70409 0.00985 -0.13017
4 401 -0.98030 -0.06516 -0.80946 -0.00703 -0.06948
4 402 -0.85360 -0.02231 -0.70965 -0.01823 -0.03100
4 404 -0.92478 -0.02450 -0.81978 -0.02658 -0.02290
4 407 -0.97168 -0.14608 -0.89072 -0.02536 -0.20312
4 418 -0.95332 0.34189 -0.49857 0.07460 0.52855
5 110 -0.02966 0.95623 0.11855 -0.72534 —0.80171
5 119 0.02708 0.87489 0.22350 -0.54861 -0.37644
5 308 -0.06735 1.58897 0.65491 -0.75355 0.02279
5 309 0.33165 1.63924 0.70319 -1.44791 -1.08508
5 320 -0.00730 1.10162 0.50271 -1.11882 -0.74662

 

CLUSTOT

QQQQQQQQQQQQQQGGGGOWU‘U‘

CLUSTOT

05MtnUIm(”UimtnllﬂublbbibibbIblbbtdhawldhlwtdhlwtdh’wtdhlwtn08wOJF‘HOJh‘HidF‘PFJP'

ID

810
827
830
128
127
307
823
826
102
220
228
301
306
318
315
323
326
328
809
817
819
822

ID

118
126
213
216
218
227
230
302
321
821
829
120
201
202
203
205
207
208
209
210
211
212
217
223
228
225
226
229
806
109
112
308
325
327
329
801
802
808
807
818
110
119
308
309
320
810
827
830
128

Dim1

0.08696

0.38755
-0.20888
-0.62390
-0.70393
-0.87856
-0.37275
-0.70279
-0.26907
-0.25856
-0.11880
-0.32862
-0.56863
-0.88833
-0.71910
-0.68916
-0.28572
-0.59860
-0.06306
-0.33798
-0.21018
-0.33821

Dim1

0.000101
0.022970
0.026630
0.069827
0.035110
0.008850
0.032698
0.001897
0.036968
0.000088
0.008982
0.273966
0.851270
0.778668
0.857879
0.688119
0.875818
0.688225
0.696876
0.832985
0.688208
0.706170
0.668505
0.675159
0.376737
0.657612
0.288590
0.691858
0.558763
0.387075
0.871763
0.872865
0.827085
0.367178
0.389605
0.868888
0.883259
0.813785
0.819882
0.887711
0.000369
0.000350
0.001068
0.015913
0.000011
0.005832
0.031967
0.026518
0.036225

242

D1u2

0.62998
1.35712
0.79182
2.05629
1.32018
1.85668
1.35931
1.78636
-0.12583
-0.37681
-0.09703
-0.52667
-0.18822
0.07088
-0.08136
-0.17176
-0.25635
0.10285
-0.36059
-0.33758
-0.32698
-0.10682

Dim2

0.380278
0.508759
0.887871
0.889105
0.280361
0.035900
0.228250
0.296003
0.880718
0.383858
0.297191
0.001881
0.003867
0.060727
0.087158
0.071386
0.008152
0.036576
0.008680
0.002889
0.000177
0.000020
0.000135
0.008210
0.002008
0.035253
0.086981
0.002773
0.000073
0.001209
0.028539
0.000525
0.008897
0.007620
0.008668
0.002250
0.000303
0.000290
0.009880
0.062728
0.383380
0.365218
0.598680
0.388750
0.251985
0.306080
0.391992
0.395781
0.393898

Dim3

-0.03013
0.57891
0.26653
0.91200
0.68619
0.91088
1.88858
1.52875

-0.18763

-0.20986
0.02961
0.05961

-0.36866

-0.31821

-0.15162

-0.81865

-0.00277

-0.11982

-0.16606
0.17233

-0.26559

-0.05926

Dim3

0.196535
0.207998
0.061078
0.382853
0.281298
0.269003
0.531085
0.380933
0.305828
0.522310
0.378321
0.057706
0.073770
0.020806
0.063138
0.071168
0.098108
0.025011
0.101708
0.061236
0.085218
0.088302
0.108359
0.107157
0.031070
0.056092
0.007610
0.063183
0.100518
0.286079
0.178629
0.221685
0.322523
0.308232
0.255915
0.387178
0.306365
0.325159
0.352859
0.133393
0.005893
0.023838
0.101021
0.071536
0.052875
0.000700
0.070385
0.088889
0.077803

D1n8

-0.55387
-0.99238
-0.53675
-0.18856
0.26116
-0.12108
3.82398
2.70999
-0.19806
0.08039
0.11007
-0.03358
-0.11381
-0.09358
0.38603
0.01979
-0.12018
0.09870
0.10829
0.83278
0.08089
-0.11680

Squared Cosine- for the Row Points

Dim4

0.088661
0.023626
0.003372
0.003088
0.096916
0.097838
0.008539
0.007208
0.017397
0.015612
0.017789
0.008930
0.081798
0.001623
0.018955
0.002589
0.085560
0.001896
0.052508
0.038172
0.030088
0.018097
0.038310
0.086883
0.027730
0.000013
0.029278
0.015888
0.029933
0.000051
0.002658
0.115388
0.000010
0.000255
0.000050
0.000026
0.000202
0.000382
0.000286
0.002987
0.220595
0.183608
0.133783
0.303297
0.259916
0.236593
0.209586
0.182088
0.003309

Dims

-0.87789
-0.80138
-0.19822
2.11980
1.99820
1.88963
-1.88838
0.50831
-0.28895
0.09563
0.36982
0.01836
-0.11928
o0.06587
-0.16356
-0.05873
-0.08366
0.51968
0.05580
-0.19295
0.08888
0.00289

Dims

0.009357
0.008316
0.006750
0.020283
0.015206
0.023608
0.078556
0.009896
0.001289
0.000155
0.003829
0.000978
0.035993
0.008779
0.012368
0.019803
0.019317
0.001225
0.081065
0.018885
0.018802
0.015577
0.019270
0.039300
0.003023
0.000015
0.016276
0.002870
0.015030
0.000200
0.005967
0.052638
0.002281
0.007118
0.008787
0.002558
0.000585
0.000258
0.018329
0.189921
0.269890
0.067618
0.000122
0.170336
0.115788
0.176136
0.136682
0.028829
0.818018

CLUSTOT

QQQQQQQQQQQQQQGGQG

127

EAITOA
GUATAP
QUINA
ERUCON
CANDEL
CINAZO
ALMACIG
GUAYAC
CAMERON
AROMA
MOSTAZO
SANGRE
FRIJOL
CAFETAN
PAAMAR
UVERO

EAITOA
GUATAP
QUINA

ERUCON

CINAZO
ALMACIG
GUAYAC
CAMERON
AROMA
MOSTAZO
SANGRE
FRIJOL

PAAMAR
UVERO

OOOOOOOOOOOOOOOOOO

Dim1

.062728
.032279
.007626
.032313
.025368
.085566
.002837

088367
165070

.295592

528880

.097558
.057981
.308918
.008859
.122137
.030055

036858

Dim1

-0.75871
0.38228
0.93595

-0.32752
1.88838

-0.27052

-0.05023
0.00858

-0.78861

~0.66398

-0.21228
0.88921

-0.09950

-0.88235
0.73573

-0.60872

243

Dim2

0.220619
0.898092
0.101812
0.208768
0.005587
0.099880
0.001782
0.113963
0.018388
0.006330
0.006770
0.006829
0.086683
0.009171
0.185793
0.121818
0.072720
0.003677

Dim3

0.059603
0.118911
0.120965
0.152897
0.007636
0.030851
0.000162
0.001860
0.070373
0.127599
0.023513
0.080572
0.000005
0.012861
0.030922
0.031752
0.087988
0.001132

Squared Cosines for the Row Points

Din4

0.008638
0.002100
0.683850
0.880866
0.013193
0.001187
0.002281
0.000863
0.006707
0.011035
0.152809
0.000091
0.010251
0.008512
0.013189
0.200260
0.001138
0.008396

Dims

0.503808
0.318087
0.121579
0.016908
0.028887
0.006831
0.025297
0.000139
0.007362
0.005867
0.027361
0.000798
0.008968
0.235975
0.003892
0.039807
0.001623
0.000002

Column Coordinates

Dim2

-0.10618
1.22211
-0.21690
0.06082
-0.00962
-0.17187
-1.12998
0.00739
1.31878
1.76858
0.05118
0.18518
-0.23029
-0.25635
-0.36386
1.18832

Squared Cosine.

Dim1

0.571887
0.085708
0.836963
.058082
.788782
.036123
.000956
.000013
.091538
.058786
.010016
0.188330
0.002627
0.032760
0.182381
0.018696

OOOOOOOO

Dim2

0.011188
0.867208
0.023867
0.001865
0.000035
0.018582
0.883968
0.000035
0.287115
0.386658
0.000581
0.007050
0.018073
0.011002
0.038825
0.072196

Dim3

-0.58708
0.86829
-0.05887
-0.03086
-0.50599
0.02939
1.10007
-0.15668
0.26207
0.98701
0.86887
-0.12109
-0.21268
-0.28822
-0.27809
1.87103

Dim3

0.342128
0.067433
0.001400
0.000400
0.096266
0.000426
0.858716
0.015607
0.011339
0.111366
0.048795
0.003016
0.012003
0.009985
0.019761
0.110634

Diu8

0.00838
-0.89780
0.28870
0.00212
0.26826
-0.02088
-0.19873
-0.01151
-0.37988
0.39801
-0.70201
0.05726
0.17086
0.05378
0.19095
3.63996

for the Column Points

Dim8

0.000069
0.252188
0.081576
0.000002
0.026257
0.000207
0.018970
0.000088
0.023770
0.019278
0.109571
0.000678
0.007710
0.000888
0.009591
0.677392

Dims

-0.08103
-0.67825
0.13068
0.00280
0.20582
0.09137
0.07895
-0.11976
1.21710
1.79386
-0.52171
-0.27880
0.08873
-0.08577
0.27388
~1.50560

Dims

0.006518
0.182210
0.008513
0.000008
0.015867
0.008121
0.002363
0.009119
0.288561
0.399817
0.060515
0.015983
0.000630
0.001232
0.019668
0.115896

APPENDIX K

Results of CDA procedures using the full data set

244

 

245
APPENDIX K

(I) The results of the CDA procdure using 16 spp and 118 sites representing 13
groups.
Canonical Discriminant Analysis
117 DP Total

105 DP Within Classes
12 DP Between Classes

118 Observations
16 Variables
13 Classes

Class Level Information

CLUSTOT Frequency Height Proportion
1 12 12.0000 0.101695
3 18 18.0000 0.152542
8 11 11.0000 0.093220
5 8 8.0000 0.067797
6 5 5.0000 0.082373
7 18 18.0000 0.118688
E 5 5.0000 0.082373
36 9 9.0000 0.076271
C 3 3.0000 0.025828
D 18 18.0000 0.118688
E 8 8.0000 0.067797
F 3 3.0000 0.025828
G 8 8.0000 0.067797

Multivariate Statistics and P Approximations

8-12 M-l.5 N-88
Statistic Value F Hum DP Den DP Pr > F
Hilks' Lambda 0.00001118 10.5143 192 895.1545 0.0001
Pillai's Trace 5.74286224 5.7937 192 1212 0.0001
Hotelling-Lawley Trace 34.93521466 16.0423 192 1058 0.0001
Roy’s Greatest Root 14.37501015 90.7823 16 101 0.0001

NOTE: F Statistic for Roy's Greatest Root is an upper bound.

Adjusted Approx Squared
Canonical Canonical Standard Canonical
Correlation Correlation Error Correlation
1 0.966933 0.958859 0.006013 0.938959
2 0.935868 0.917721 0.011587 0.875101
3 0.918217 0.908816 0.018503 0.883122
8 0.858955 0.816981 0.028878 0.730988
5 0.828280 0.805388 0.029025 0.686088
6 0.783755 0.763512 0.035661 0.618272
7 0.698087 0.668289 0.087397 0.887325
8 0.555300 0.508569 0.063982 0.308358
9 0.319298 . 0.083025 0.101988
10 0.313328 . 0.083378 0.098175
11 0.239786 0.230256 0.087138 0.057897
12 0.071883 -.180896 0.091978 0.005110
Eigenvalues of INV(E)'B
- Canqu/(l-Canqu)
Eigenvalue Difference Proportion Cumulative

1 18.3750 7.3685 0.8115 0.8115

2 7.0065 1.6321 0.2006 0.6120

3 5.3788 2.6576 0.1538 0.7659

8 2.7168 0.5316 0.0778 0.8836

5 2.1852 0.5927 0.0625 0.9062

6 1.5925 0.6819 0.0856 0.9518

7 0.9506 0.5087 0.0272 0.9790

8 0.8858 0.3323 0.0128 0.9917

9 0.1135 0.0087 0.0032 0.9950
10 0.1089 0.0879 0.0031 0.9981
11 0.0610 0.0559 0.0017 0.9999
12 0.0051 . 0.0001 1.0000

Test of H0: The canonical correlations in the current row and

CBS

WGQGU‘IOWNH

wmqmmawwla

HFJH
NHO

ID
101

103
108
105
106

108
109
110
111

113
118
115
116
117
118
119
120
121
122
123
128
125
126
127
128
129
130
201
202
203
208
205
206
207
208
209
210
211
212
213
215
216
217
218
219
220
221
222
223
228
225
226
227
228

Likelihood

0
0
0
0
0
0

0
0
0
0
0

0

Ratio

.00001118
.00017191
.00137682
.00877378
.03260998
.10386917
.26928065
.52528680
.75981907
.88562968
.93768675
.99889018

CLUSTOT

EG

w.0
Q

Q

ClMthP‘M<DFIUlﬂthr4C78’0C1U4>C1mlbulntnﬂ!U

QFJUJwtdhlﬂtnantJhJH()Fiwtdhlwtdu’MCJUIWtdhlm

246

Approx F

10.5183
8.2889
6.7981

.3568

.8388

.8397

.8098

.5035

.8789

OHNM‘IU'I

CANl

-0.81675

2.23279
-0.29055

5.16966

2.91590

1.82592
-3.85375

2.01636

5.85285
.82812
.55815
.98578
.58877
.39850
.16297
.01809
.98185
.87929
-0.90908
-3.93686
-1.91788

8.06589
-1.88058
-1.95166
-3.50689

0.98858
-0.86768
-2.38067
-0.82362
-l.28698
-5.38767
-6.99895
-8.78323

1.03515
-8.32989
-3.31652
-5.91083
-6.17088
-7.20182
-6.01713
-7.08679
-6.52718
-0.35609
-8.99525
-1.82663
-7.78891

1.77709
-8.17700

2.16602
-8.33296
-2.58881
-6.75512
-8.38696
-6.85735
-5.08009
-0.28919

0.53657

OUNNU‘SUIIbO

Num DF

192
165
180
117
96
77
60
85
32
21
12
5

CANZ

1.59891
-0.39889
0.02582
1.31896
0.65693
0.78350
-0.88050
0.60889
2.86356
-2.82172
0.99600
2.38513
0.58781
0.12827
-2.03379
-1.10302
2.50682
-0.39825
-3.18827
8.26763
-3.51989
0.88365
-2.83532
-8.89756
-5.23798
-0.57398
-3.92508
-3.68502
-0.85013
-2.82888
2.18582
2.88813
2.87991
0.81859
2.25627
-0.70828
5.69878
2.39568
8.65115
8.55509
8.61857
3.58079
0.18966
0.27099
-2.88227
5.91671
-1.92032
0.32978
-0.89815
-1.30522
-1.35525
5.37239
2.82138
2.71600
2.88868
-2.68802
-0.51218

all that follow are zero

Den DP

895.1585
838.2959
772.1261
708.5608
683.8827
576.723
508.0298
837.0078
363.0013
288.8287
200

101

CAN3

1.12975
-1.85107
2.93572
1.86726
1.19883
1.80708
1.16978
-0.60862
1.00867
2.22638
-0.89856
1.59068
-2.87918
2.31882
0.03519
1.87682
0.97690
-3.73588
.99988
.17003
.38992
.55282
.86890
.82061
.86136
.95537
.71015
.86252
.88758
.51785
.31892
0.81868
0.30285
0.75018
0.69108
1.66778
0.32887
0.98858
-0.80603
0.88181
0.32127
-0.08623
-8.32917
1.28073
-7.08760
-0.87609
-8.68912
1.35087
-2.07811
1.65637
-0.15868
—1.07213
-0.08072
0.29868
-0.73038
-8.38735
-3.28530

OHHUNWHHNOP

I
OH

Pr > P

0.0001
0.0001
0.0001
0.0001
0.0001
0.0001
0.0001
0.0226
0.6598
0.7001
0.8833
0.9912

CAN8

0.82675
1.07361
-1.15631
0.98382
0.23706
1.21870
0.69985
-1.33182
-0.02502
2.36623
-0.09856
-1.00132
-0.32858
-3.59205
1.97381
-1.20781
0.28687
-0.28577
1.70196
0.80928
-0.31898
0.17890
2.39518
-8.96191
0.80853
-0.71889
-6.22655
1.72668
1.63681
3.28067
-0.32802
0.28302
-0.05088
0.86257
0.50967
1.71722
-0.80707
0.29823
-1.03191
-0.83389
-0.35861
-0.81680
-0.68635
1.00529
-0.56316
-1.19810
0.36302
0.01570
0.36783
2.08266
0.31368
-1.38698
0.56552
-0.80797
-0.81751
0.82888
-0.86208

 

 

 

a"

 

ID

229
230
301
302
303
308
305
306
307
308

310
311
312
313
318
315
316
317
318
319
320
321
322
323
328
325
326
327
328
329
330
801
802
803
808
805
806
807
808
809
810
811
813
818
815
816
817
818
819
820
821
822
823
828
825
826
827
828
829
830

CLUSTOT

Q

C)

th~Jﬂlétdtjbtjﬁ-b'qdhxlﬁ~J¢Ig~JUIHtﬂCIw‘DC)Q~dU!UI06)mtnO\QtjibUl4\JHtJ

Q

Gib

UIHQU‘GQHG‘Ql-‘UQDQUw

247

CANl

-5.66791
-2.01269
1.73967
-0.93205
2.28692
6.38062
3.10318
8.01738
-1.77056
-0.78237
-3.03096
-0.38887
-3.80189
8.20088
2.78271
2.69928
3.08526
0.73391
1.39878
1.17573
-0.19601
-1.80103
1.86879
0.71917
1.91236
0.28800
6.08971
1.57829
5.93506
3.18232
5.05879
-1.18931
6.32335
5.18303
8.35338
6.07233
3.05691
-7.88778
7.62698
2.82268
0.82093
-0.76396
1.96953
3.33627
-1.28153
2.72326
3.58828
2.15953
5.67011
2.68967
8.29086
0.03292
1.18361
-2.85065
-1.82096
0.23272
-1.78205
-2.21302
-0.87175
-1.35722
0.17793

CAN2

2.88652
-2.88251
-0.71180
-3.88181

1.01328

2.79701

2.89076

0.85285
-5.12195
-5.26765
-6.11398
-2.36535
-8.22355

1.53321
-0.32858

0.88583
-0.27360
-0.62672
-1.03959

0.10200
-1.31069
-3.71092
-2.78153
-0.17703
-0.89838

0.28982

3.07829
-0.53092

2.71102
-0.38025

2.20185
-2.21626

3.03768

2.87026

2.87161

2.98589

1.09378

5.38582

3.73892
-0.76630

0.95329
-1.19988

0.10888

1.29703
-1.25596
-0.38012

1.85365
-0.26827

1.92978

0.28859

1.70153
-3.98101
-0.51386
-8.69865
-2.11231
-0.53837
-5.78268
-8.82208
~0.89511
-2.13397
—2.80537

CAN3

-0.36818
-6.62687
-2.63071
-6.38225
-0.17083
1.08207
1.12885
-1.08531
2.88037
3.88720
3.65388
-0.76513
1.68125
0.81108
1.35838
0.57502
-1.16515
0.97923
-0.22188
1.02133
0.05297
2.39677
-7.61311
—0.52010
-0.89387
2.11178
1.18285
-2.28872
1.23285
-0.78601
1.09839
1.11839
1.20211
1.11722
0.59806
1.12725
1.02715
-0.57879
1.27626
1.63082
-2.83010
1.52591
0.51585
0.87256
-1.61550
0.88637
1.08788
-3.28987
1.85096
-1.83281
0.95966
v8.22272
-2.18900
2.96257
-1.93615
-1.63990
3.76639
3.36879
-0.28397
-5.98668
1.55579

CAN8

0.16707
-1.76110
0.23772
-0.36508
-0.26037
-0.81970
0.29991
0.02189
-8.06861
-0.65672
8.69777
1.30908
3.87213
-0.98988
0.82259
-0.08097
-0.21618
2.81918
1.50222
-3.07157
0.77289
3.17052
-0.75017
-1.15523
0.72650
-1.33980
0.35372
0.83757
0.15759
-1.67013
0.33375
1.81315
0.28026
0.89780
0.01252
0.58197
0.16858
-1.18736
0.01917
0.89020
-0.81082
2.08950
0.07838
1.51856
1.76812
0.89698
-0.30033
-0.93318
-1.93328
0.82058
0.00922
-0.81226
0.85753
-5.68069
-0.58850
0.67953
-9.37585
2.96617
2.50165
-0.16683
0.86681

 

248

(II) The results of the CDA procedure using 16 spp and 67 sites representing

seven groups

Statistic

Wilks'

Lambda

Pillai's Trace

Hotelling-Lawley Trace
Roy’s Greatest Root

Canonical Discriminant Analysis

67 Observations
16 Variables

6 Classes

Class Level Information

66 D? Total

61 DP Within Classes

5 DP Between Classes

CLUSTOT Frequency Height
1 11 11.0000
3 18 18.0000
4 11 11.0000
5 8 8.0000
6 5 5.0000
7 18 18.0000

S-S M-S N-22
Value F
0.00001082 27.7618
4.33793292 20.4753
59.05591596 32.7760
25.11599918 78.4875

Proportion

0.168179
0.268657
0.168179
0.119803
0.078627
0.208955

Multivariate Statistics and F Approximations

Den DP
80 225.7661
80 250
80 222
16 50

NOTE: F Statistic for Roy’s Greatest Root is an upper bound.

MOWMH

0|.qu

Canonical

Correlation

0.980668
0.973077
0.987237
0.918208
0.838852

Eigenvalue

25.1160
17.8287
8.7331
5.0893
2.2928

Adjusted Approx
Canonical Standard
Correlation Error
. 0.004713
. 0.006539
0.935223 0.012687
0.901003 0.020218
0.813896 0.037382

Eigenvalues of INV(E)*H

. Canqu/(l-Canqu)

Difference

7.2913
9.0916
3.6838
2.7968

Proportion

0.8253
0.3018
0.1879
0.0862
0.0388

Squared
Canonical

Correlation

0.961709
0.986878
0.897258
0.835777
0.696311

Cumulative

0.8253
0.7271
0.8750
0.9612
1.0000

Pr > P

0.0001
0.0001
0.0001
0.0001

Test of no: The canonical correlations in the current row and all that follow are zero

U'IOIUNH

Likelihood
Ratio

0.00001082
0.00027220
0.00512808
0.08987286
0.30368928

Approx F

27.7618
22.2538
16.7590
13.1088

9.5535

Rum.DF

80
60
82
26
12

Den DP

225.7661
185.6923
183.1568
98
50

Pr > P

0.0001
0.0001
0.0001
0.0001
0.0001

 

 

 

OES

UOQOUI.UNH

ID

102
109
110
112
118
119
120
128
126
127
201
202
203
205
207
208
209
210
211
212
213
216
217
218
220
223
228
225
226
227
228
229
230
301
302
308
306
307
308
309
318
315
320
321
323
325
326
327
328
329
801
802
808
806
807
809
810
817
818
819
821
822
823
826
827
829
830

a
2:

MHU‘GGQPdﬁQUIQDUbﬁhﬁﬁhdeHdemmmdﬂﬁ-‘QHUQPUNWNQHUHHUUWUUUUUUUOHUtUUIHﬁUI‘Q

249

CAN1

-1.17203
-3.38387
0.52398
-5.00517
0.51770
-0.01069
5.98850
-9.57178
-1.18508
-7.88818
5.27532
7.95953
5.31099
5.39283
7.55885
6.78073
7.72221
7.68190
8.38681
7.35939
0.67655
1.29118
8.73886
-1.78589
-1.76886
7.19318
6.69116
7.88753
6.21736
-0.35882
-1.75029
6.07258
0.75888
-1.71806
-0.90581
-8.96815
-2.71180
-7.68531
-1.25688
1.17582
~2.01928
-8.08730
-0.96882
-1.85276
-1.91127
-8.57553
-1.68903
-8.10606
-3.81708
-3.93897
-8.66167
-8.78668
-8.98022
8.33950
-8.38059
-0.89261
0.36795
-3.02580
-6.08367
-2.29181
-0.99628
-2.31891
-8.82306
-8.56680
1.99188
0.16312
-1.09385

CAN2

2.2250
2.1317
0.6768
1.7278
8.3808
-1.8377
-1.8770
-13.0035
3.7319
-8.8277
-2.3822
-2.0962
-0.8886
-1.8908
-1.5378
-2.2689
-2.6888
-2.1816
-2.8767
-2.1080
3.8536
8.6855
-3.0667
5.0618
2.3088
-2.3800
-2.1809
-2.3990
-1.6372
3.6105
1.0918
-1.9808
3.7810
2.3783
8.1189
2.8229
2.9088
-10.8235
-3.0106
-1.5938
1.2193
1.1768
-1.3280
7.2203
0.5070
2.6828
2.5565
3.1529
0.8887
2.5738
3.1209
1.9985
2.6788
-3.0335
8.8882
2.3617
0.0999
2.7118
0.2871
2.1031
6.2633
1.3618
-11.2888
-13.1578
-1.7698
3.9228
-0.7558

CAN3

0.50098
8.18218
1.09153
3.85399
-2.75875
0.03869
1.27568
-2.38212
-2.69878
-2.28878
0.76852
0.65712
0.80208
0.91116
2.03919
0.61891
1.08653
1.23118
1.16338
0.69888
-2.66618
-6.11676
0.98073
-3.26238
0.38288
0.70683
0.87361
0.83285
0.02160
-3.82662
-0.70965
0.63668
-7.01008
-0.86193
-6.16879
8.30878
1.55071
-2.87598
-1.09979
°1.28518
1.92852
1.02527
-0.86882
-6.35056
1.59791
8.75211
-0.61383
8.53639
0.73056
8.05193
8.85118
8.28891
8.88307
0.88207
5.81031
-0.15995
0.95136
-1.18559
3.71778
0.88876
-7.62735
-0.10187
-2.87179
-3.39375
-1.09852
-5.69300
-0.09008

CAN8

1.26661
-2.03331
5.54160
-0.54534
-0.29125
4.39540
-1.12500
-2.24274
-1.22719
-1.53257
-0.30110
0.77773
0.61069_
0.49436
-1.34223
0.76269
-1.62233
-2.07169
-1.23329
-0.90301
-1.05223
-1.30355
-2.53115
-0.09179
-0.33355
-3.01231
-1.25612
0.11163
-1.35466
0.45127
-1.6l638
0.09903
-1.13030
-0.42676
-0.53322
-1.00229
-0.53395
-0.61160
6.29043
3.72995
0.51303
-1.20903
5.88886
-1.43369
0.55331
-0.60353
0.36731
-0.32021
-0.35205
0.22346
-0.30443
-0.42791
-0.41460
-1.79477
-0.11160
-1.50765
3.93759
-1.37373
-1.03539
-0.53091
-0.73639
-0.49032
-1.31563
—1.40374
7.37973
-1.11739
4.03572

 

I}

250

(III) Results of CDA procedure using 16 spp and 62 sites representing ﬁve
groups.
Canonical Discriminant Analysis
61 DP Total

57 DP Within Classes
8 DP Between Classes

62 Observations
16 Variables
5 Classes

Class Level Information

CLUSTOT Frequency Weight Proportion
1 11 11.0000 0.177419
3 18 .18.0000 0.290323
8 11 11.0000 0.177819
5 8 8.0000 0.129032
7 18 18.0000 0.225806

Multivariate Statistics and P Approximations

3-4 H-5.5 Nh20
Statistic Value P Rum DP Den DP Pr > P
Wilks' Lambda 0.00012148 23.4523 64 166.6986 0.0001
Pillai’s Trace 3.87803910 18.7808 68 180 0.0001
Hotelling-Lawley Trace 88.61065180 28.2302 68 162 0.0001
Roy's Greatest Root 25.46787882 71.6284 16 45 0.0001

NOTE: P Statistic for Roy's Greatest Root is an upper bound.

Adjusted Approx Squared
Canonical Canonical Standard Canonical
Correlation Correlation Error Correlation
1 0.980927 0.978852 0.008837 0.962218
2 0.953905 0.980259 0.011532 0.909935
3 0.926771 0.912590 0.018065 0.858905
8 0.868280 0.888408 0.032396 0.786981
Eigenvalues of INV(E)*H
- Canqu/(l-Canqu)
Eigenvalue Difference Proportion cumulative
1 25.8679 15.3688 0.5709 0.5709
2 10.1031 8.0156 0.2265 0.7978
3 6.0878 3.1352 0.1365 0.9338
8 2.9523 . 0.0662 1.0000

Test of HO: The canonical correlations in the current row and all that follow are zero

Likelihood

Ratio Approx P Rum DP Den DP Pr > P

1 0.00012148 23.4523 64 166.6986 0.0001

2 0.00321530 16.8625 45 128.5226 0.0001

3 0.03569969 13.4910 28 88 0.0001

4 0.25301937 10.2194 13 45 0.0001

038 ID CLUSTOT CANl CANZ CAN3 CAN4
1 102 7 1.74456 -0.01409 1.52315 -2.60925
2 109 4 4.94853 3.66201 -2.89968 0.58737
3 110 5 -0.41120 1.50288 4.83269 1.00703
4 112 4 5.89994 3.85338 -1.15891 1.66025
5 118 1 1.66292 -4.17186 -0.71727 1.65455
6 119 5 -1.00367 1.39656 5.18511 0.18513
7 120 3 -5.67635 1.15112 —1.71587 0.71687
8 126 1 2.43948 -3.50271 -0.77623 0.85465
9 201 3 -5.59763 0.92409 -0.76935 -0.93101
10 202 3 -7.61256 0.51675 -0.28272 1.73566
11 203 3 -4.72748 0.44354 -0.23201 0.67548
12 205 3 -5.22746 0.82958 -0.09903 0.60810
13 207 3 -6.65106 1.26637 -3.23801 2.08280

 

 

 

ID

208
209
210
211
212
213
216
217
218
220
223

7-228

225
226
227
228
229
230
301
302
308
306
308
309
318
315
320
321
323
325
326

CLUSTOT

UIHUIQH\IhxlmxlubthﬁbdeIbQHlelxllﬂUldﬁHQHWQHwWWWQPWHl-‘wwwww

251

CANl

-7.02580
-7.98760
-7.38625
-8.09590
-7.26610
1.52897
1.15885
-8.68818
8.18360
2.55337
-7.15353
-6.67719
-7.76818
-6.57520
2.08801
2.08789
-6.31891
1.36188
2.63897
2.85358
5.31069
3.50527
-0.63580
-1.81682
2.07883
3.65233
0.19866
5.15902
1.95613
6.27832
2.00967
6.08681
3.90251
5.36088
6.91807
6.19371
6.76800
-8.52637
7.05381
1.76828
-0.27868
3.02398
6.68567
3.05863
3.83336
2.68871
-2.88632
1.83833
0.10327

CAN2

0.77838
0.90278
0.68856
0.88922
0.51282
-3.88099
-7.25967
0.78221
-8.83588
-0.16005
0.36739
0.87381
0.80272
-0.09808
-8.55858
—0.80810
0.72809
-7.57033
-1.28033
-6.80010
3.62912
0.89923
1.88988
1.18820
1.60713
1.52128
1.25560
-8.12831
1.68798
8.19578
-1.08003
3.29270
1.05693
3.18059
3.92188
8.05008
8.58351
0.76287
3.87318
-0.99807
1.61766
-1.83378
8.61229
0.36860
-8.88978
-0.27065
0.89062
-6.83236
0.98129

CAN3

0.88239
-1.78006
-2.72778
-2.10258
-1.33291
-1.80373
-0.96375
-2.96923
-0.26308

0.08098
-3.01175
-1.80892
-0.07789
-0.88087

0.78951
-0.28513

0.01359
-0.63981

0.09197
-0.08812
-1.11388
-0.20582

7.19903

8.39589

0.75689

0.08891

5.83838
-1.38336

1.01797
-1.88979

1.27256
-2.08520

0.86538
-1.11335
-2.65206
-1.56059
-1.75588
-2.07187
-2.81085
-1.36268

3.51085
-0.33033
-1.33583
-0.31221

0.15869

0.37101

7.11939
-0.68967

8.78323

CAN8

0.08393
-0.52280
0.91829
1.11897
0.19321
1.37388
0.98225
0.38150
0.85620
—3.03673
-1.18210
-0.70835
-0.27257
-1.50698
-1.08598
-5.68158
-0.56971
2.58089
-2.32099
1.76618
-0.22763
-3.90387
1.88851
2.76310
-1.62857
-2.37828
1.32812
2.07860
-3.61083
1.00057
-2.86758
1.72685
-1.90382
1.26828
2.83285
1.77266
2.08628
0.09909
2.28059
-2.03717
1.11366
-2.61601
2.16011
-2.17060
0.89852
-8.07083
1.93218
1.82061
-1.17725

 

 

APPENDIX L
Scaled diagrams for the four experimental blocks of the

ISA-Mao silvicultural study

 

252

 

+—

253
APPENDIX L

BLOCK ONE

 

0“ I...
.’ ‘3 I ‘s‘
: o : 4
. y a... “Q “ "
' O O. .-‘
.2— I “
1 3
‘ s‘ 5:"; ‘
~ 0-.
a, ‘\ 4 ' “ '
. .' ‘. t ‘.
‘ \ I 4 s
\\£’I \53.o'

 

\Umjl'2.5wx ;'.": ." a:
‘24 .' 8';
4_./ ['7‘ 9.0

I.'

8

‘Mo '
3 0.

 

{’ K: / ‘~.
1 3, . l 28 ‘.
‘\—/ \0"
. - 90 y, 6 0 70
a" .‘s /~-‘\ ' ' “
‘8 g. \ ’ : ' . "
.e--o’ \~_.I’ 1"”, ‘ .

LU
I'-‘\
s:
- \c'
’.
E
\
8 &

 

 

13.

 

1 .1
l
6

I

o

4

0

(new

 

 

 

 

 

 

 

.. t .. new. . \ll.
. / o \ . 7’
a s. of / s
,0 I304\\ ’8\
a a ..
/..v
\\‘-"’ ‘\88‘/
1 L 1 £
9 , .
I 0’08\9

 

 

 

254
BLOCK TWO

135.3»..43.

 

 

 

 

.. da—OUM
\8"3" “8'.’
i . . .
. . 4.
/. II-8\ l’l‘\e
do §\\I‘Ie
o
b .9 I.
.o \s
\‘ 6’ \\"/
o o .
o a’. . IV
I- \\ I '8‘

 

 

 

 

W08

 

'

 

' -8

_

255
BLOCK THREE

 

50

H.

1.1

 

 

 

 

 

 

 

 

2 0 70 IE‘ ’6‘
. (‘7‘) x'o‘ .‘Q‘. 'Ir q“ I ' o e\
. 3 :
\500’ \f'" kn.) ‘\~:"’.
0' ‘. ' “0
. .
\“.'£"l. 8%. ”’ 1 L .
’00.‘ "-~. . '.O\ . 'O..
\. , s ‘l ' \
l I," 9‘" /’ \30' A” I
\o I .O- -e’ ‘~.O.'
. 4 0 ‘/. 9 0 '/4
'/ \‘. f o-\ (’ 0% \‘ {"e‘
o .
“‘2. /: .‘70 -i') ‘I' ". \eéots ;.
’ o - \
0‘ 80 . -
..,, ~.u.
I. .‘s t'.‘ 'P.‘ - ’~ ‘
.I . ' 0' 3. 0' " I” “
\ ' '
ks...) “s..." K950; ‘é..2’
fey’a’ 6070
.4 “"s-‘ f’ﬁ‘ ’4 “e I "e‘
f 3 , . : .1 0 ‘.
‘4 a", “‘\L f ‘ 9. . a". ‘\.l. f”
.' ‘3 1' ‘: “‘
\..a/ .. 1' I
.Oo“ ' ".-‘ '0‘. -’ o‘.e
. ‘ ' . ‘. ‘0 ‘. .l \‘
.Q‘..?"7 I“- . '1. x. .1:’ I. :‘é.’ '0}. h
. ' “mu“
so u: 1
€Cal'4- '3 [M t \O “‘5

 

256
BLOCK FOUR

 

40000. 600/.. 00000 .....7€5//Go
4.4. : -. 3:. \212' “5

to.

 

 

14...}

q
:5
LL

 

H3

 

 

 

in

 

4'! 68’ F'

APPENDIX M
Representative maps with proﬁle icons indicating cluster group membership and
relative contribution of each of the sixteen species used in the

cluster analyses and ordinal procedures

257

258
APPENDIX M

Key to the diagrams, the cluster groups and the proﬁle icons

The number below each icon indicates the site designation (1-30 in each block).
Squares represent original 50 X 50 m treatment plots (See Figure 2). Contiguous
plots are reprented by contiguous squares. Lines connecting plots represent plots
which are contiguous. See scaled diagram in Appendix N.

The peaks in each proﬁle represent the relative basal area contributions of the
sixteen species used in these analyses. In order from left to right, the peaks
represent P. brasiliensis, Prosopis juliﬂora, A. famesiana, C. leoganensis, L.
Ianceolatus, Cassia emarginata, Pithecellobium circinale, Cappan's ﬂexuosa, C.
cynophallophora, B. simaruba, G. ofﬁcinale, Caesalpinia coriaria, T. pallida, M.
buxifolia, E. can'baeum and A. scleroxyla. This order is based on the scores of
these species in the ﬁrst dimension of a CA procedure using 16 species and 120
sites.

"CI=" indicates cluster designation based on four cluster procedures. Numbers
indicate core site cluster groups. Letters indicate the core site cluster group with
which a noncore site was most closely associated. "A" indicates Group One, "0'
indicates Group Three, etc. Group One (Cl=l) represents sites with the
characteristic species B. simaruba and E. caribaeum. Group Three (Cl=3) sites
are characterized by A. scleroxyla, E. caribaeum, M. buxifolia and T. pallida.
Group Four (Cl=4) sites are characterized by P. brasilzbnsis and Cassia emarginata.
Group Five (Cl=5) sites are characterized by Caesalpinia coriaria, Cassia
emarginata, A. famesiana and G. ofﬁcinale. Group Six (Cl=6) sites are
characterized by A. famesiana, Coccoloba leoganensis, Caesalpim’a coriaria, and
Prosopis juliﬂora. Group Seven (Cl =7) sites are characterized by Phyllostylon
brasiliensis, Pithecellobium circinale, B. simaruba amd L. lanceolatus.

Characteristic species for subgroups include Phyllostylon brasiliensis and
Caesalpinia coriaria for Subgroup B (Cl=B) sites, P. brasiliensis and Cappan's
cynophallophora for Subgroup BG (Cl=BG) sites, Caesalpinia coriaria, A.
scleroxyla, and E. can'baeum for Subgroup C (Cl =C) sites, P. brasiliensir for
Subgroup D (CI=D) sites, C. coriaria and Prosopis juliﬂora for Subgroup E
(Cl=E) sites and P. juliﬂora for Subgroup G (Cl=G) sites. "Characteristic
species" were not determined for subgroups represented by only one or two sites.

259

 

Cl-D r Cl-D h
14 11

Cl-FG

15

CH) II ° Cl-4 II
13 12

 

 

01-4 II CI-=BDII
09 06
cr-s I
10
CI-BG II I? cu—c II II II
08 07

 

 

 

Cl-D II 01-96 l‘
04 O1

CF80

O5
CI=BG I I I I CI=? II I I I
03 O2

 

 

 

CI-G I I Cl-1
29 26
CI-E
30
C'-E I\ III II Cl-6 II Ii
28 27
Cl-6 II Cl-E I I II
24 21
Cl-E
25
Cl-EH II II Cl-D II
23 22
1
01-5 II Cl-BG I\
l
19 16
Cl=3
20
Cl=1 Cl=D II
18 17

 

 

Block One

 

 

260

 

C

MI

 

 

 

 

 

 

I M

 

 

1-3 I
02 03
01-3
05
01-3 CI-BG
[I NI II AMA
01 04
01-3 II 01-3 I II
07 08
01-3
10
Ct—E ‘I\ 01-3
1 9 III a II
06 — 09
01-3 II 01-1
I
_m
12 13
or=c
15
cu=x
14

 

 

 

 

 

CI-S I Cl-1 I
17 18
Cl-7
20
CI-1 II CI-C I II
16 19
Cl-FG I I I I Cl-3 II
22 23
Cl-S
25
Cl-E II I Cl-3
2" _ 24
Cl-1 I II II I Cl-7
27 28
CI=1
30
Cl=3 CI=3
[Ll MI A
26 29

 

 

Block Two

 

261

 

 

 

 

01-? I Ct-E I I OH! I I O CI-G
14 11 19 16
Cl-7 0.5
15 20
Ci-B I OH) I CI-BG I I OI-G I
13 12 18 17
01-5 II 01- 7 I I Co-BG I 0 Cl- 1 II
09 06 24 21
CI-G 01-4
10 25
01-5 I II 01-6 I I II 01-7 II I I 01-86 II I I
08 07 23 22
OH I end I I A 01-4 I 0.7 I II
04 O1 29 26
01-0 Ct-FH
06 30
0‘0 II W 01-1 II 01-7 II I 01-4 II
03 02 28 27

 

 

Block Three

 

 

 

 

262

 

CHAIM (HAL

CF1 I CFO I
29 26

 

 

 

24 21
CFO I I I C¥6 “
25 80
Che Ch? ChG 0P5
23 22 28 27
ChG II ChBG I Ch? I Cho I
14 11 19 16
01-8 I Cl-D I
15 20
01-0 I L‘F-X I I CH I 01-7 I I I
13 12 18 17
Ch4 I Ch4 I Ch? I I Che I
04 01 09 06
Cho II I Obs I H
05 10
U
08

L: CHIMtAM

 

 

 

Block Four

 

APPENDIX N
A proposal for a study of soil-site interactions in the managed forest of the ISA-

Mao Experimental Forestry Station

263

 

 

 

264

APPENDIX N

A soil-site study at the ISA-Mao station should ﬁrst examine the effects of
topography on soil development and available moisture. Within each of the four
study blocks, two soil pits should be randomly situated in each of three
topographic positions: toe slope, mid slope and ridge top. For each soil pit,
standard taxonomic information should be collected (Soil Survey Staff 1992),
including organic material in surface horizon, horizon depths and characteristics
(including moisture content), textures, stoniness, and depth to root limiting layer if
present. In addition, root biomass estimates should be taken, by depth, as
outlined by Lugo and Murphy (1986b). Of particular interest would be the ability
of roots to penetrate the layer of soil with a high bulk density which is present in
much of the forest (Checo, personal communication). In total, six sets of data will
be collected in four areas of the forest. Based on this data, it should be possible
to determine the site characteristics which show the greatest correlation with
available soil moisture and root distributions. These parameters are the ones

most likely to be related to differences in site quality.

Based on this initial data, a set of speciﬁc site parameters will be selected which
most clearly represent differences in site quality. Ideally, parameters should be
selected which can be measured easily and relatively rapidly. This would make
site selection unnecessary, as resources would be sufﬁcient to collect data on each

of the 120 original sites where species composition and growth and mortality data

 

265

is available. Using site data, multivariate techniques can then be applied to
determine if site relationships correspond to species composition groupings.

Initial analyses would focus on the subset of 67 sites which could be classiﬁed into
distinct species association groups. Clusters of sites which correspond both in
terms of species composition and site characteristics would be classiﬁed as species-
site groups. Based on relationships observed among these groups, sites with
intermediate species composition could be evaluated based on site characteristics
and correlated with the species-site group they most clearly seem to represent.
This aspect is essential, as only about half of the study sites were classiﬁed into a

deﬁnitive species based group.

The ﬁnal goal is to develop a comprehensive model of species-site interactions
which, (1) describes the effects of site characteristics on species composition and
(2) helps predict growth response to systematic silvicultural treatments. With such
a model, hypotheses could be developed and tested in other areas of subtropical
dry forest, both within the ISA-Mao station and in areas of forest managed
communally by local residents. Understanding the limits of the forest as a whole
and describing potential productivity as it changes across the landscape are
important steps in transforming exploitation of a deteriorating resource into wise

and sustainable use. Science meets campesino.

 

 

 

111! IIIIII- ll 1'14-

.: r
(5.... t. :13.

.v .111-

. u.
to... 44.4.; .5... 4.
(It I:

. alpha.

IRA!

 

Eva!!! : I. it
1.3. xtlvc. .
1.1.2.... .