I A. v x . . Li"! :‘ ‘ I“; *- ‘,§ eiifiigégé; ' 5,12. $5.35} 9‘35} ’ fih .- .. «.ng .. 3‘ .— 3 1a.! ft". 31*, 6 I l ~ T74 @2455} This is to certify that the dissertation entitled THE ROLE OF DISTURBED CARIBBEAN DRY FOREST FRAGMENTS IN THE SURVIVAL OF NATIVE PLANT DIVERSITY presented by IAN ALFRED RAMJOHN has been accepted towards fulfillment of the requirements for the Ph.D degree in Plant Biology/Ecology, Evolutionary Biology and Behavior ajor Pg‘é’ssomfi \1 Mom‘ ”>30 “I; Date TVI MSU is an Affirmative Action/Equal Opportunity Institution LIBRARY Michigan State University PLACE IN RETURN Box to remove this checkout from your record. TO AVOID FINES return on or before date due. MAY BE RECALLED with earlier due date if requested. DATE DUE DATE DUE DATE DUE MAY 132W Tgfi’l 2 II‘ ‘- 6/01 cJClFlCJDateDue.p65.p.15 {E lell THE ROLE OF DISTURBED CARIBBEAN DRY FOREST FRAGMENTS IN THE SURVIVIAL OF NATIVE PLANT DIVERSITY By Ian Alfred Ramjohn A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Plant Biology Program in Ecology, Evolution and Behavioral Biology 2004 THE RULI TrOl‘i me along '4' and are restrl hments. I su's'il'al OI “- Wilforeslk ilS 0%) “'55 offragmems W. I m. Onl} In ong "from“ IO.3 ABSTRACT THE ROLE OF TROPICAL DRY FOREST FRAGMENTS IN THE SURVIVAL OF NATIVE PLANT DIVERSITY By Ian Alfred Ramjohn Tropical dry forests are a globally endangered ecosystem. Like most dry forests, those along the south coast of Puerto Rico have experienced a long history of disturbance and are restricted to a single large (4000-ha) protected area and an array of smaller fragments. Evidence suggests that small fragments can play an important role in the survival of native plant diversity, especially in the absence of large protected areas. In 1993, forest cover stood at 16900 ha (23.2% of the overall dry forest life zone); 13100 ha (18.0%) was Closed Forest and 3800 ha (5.2%) was Open Forest. Nine distinct clusters of fragments were identified across the dry forest zone based on a separation distance of 500 m. Only one fragment was isolated by a distance of over 1 km. In one of the few studies of its kind, an array of forty fragments (ranging in size from 6 x 10'3 ha to 11372 ha) formed the basis of a detailed study. Guanica Forest, the 4000-ha reserve, was selected as the reference community. Nineteen fragments were classified as Relict (>7S% ‘old growth’), three were classified as Mixed (25-75% ‘old growth’) and seventeen were classified as Regrowth (<25% ‘old growth’). One fragment was unclassified. Even small Relict fragments were able to support species assemblages that were representative of those found in Guanica Forest. On average, more of the reference species (sampled from Guanica Forest) were present in Relict (S4i3.6%; mean i- 1 standard error) than in Regrowth fragments (24i2.7%). Nineteen fragments wtaoncd'*7 ll: smallcx' miss that .. hgtzmts ix. lflmmmfii- $$fooiR Species rich: bird on Ian laments tar Reiict sites (i Clngrtmrh fifihi Ids dttlsltles 0f 5 ml Were 31 I:OF‘I‘IOIJI‘ p supported >50% of the reference species and five fragments supported >75% of them. The smallest fragment which supported >5 0% of the reference species was 0.04 ha. The species that dominated the reference sites in Guanica Forest were present in most Relict fragments but were absent from most Regrowth fragments. Four of these species (Gymnanthes lucida, Eugenia foetida, Croton humilis and C. discolor) were present in 63-73% of Relict fragments but were only present in 6-13% of Regrowth fragments. Species richness was a function of fragment area and disturbance history. Clustering based on Jaccard similarity in species composition produced five distinct groups of fragments (and two unassigned sites). Three of these clusters consisted predominantly of Relict sites (including one group of coastal fragments) and two consisted predominantly of Regrowth sites (one dominated by Leucaena Ieucocephala and the other by Pisom'a albida). Like other dry forests, both the fragments and Guanica Forest consisted of high densities of small, multi-stemmed trees; between 0.2 and 5.2 m2 ha'l (up to 55% of basal area) were accounted for by stems between 1 and 2.5 cm diameter at breast height (dbh). Forty-four percent of all trees were multi-stemmed. Trees averaged 2.43 stems per tree; multi-stemmed trees averaged 4.22 stems. Of the 53 rare or endangered species present in southwestern Puerto Rico, 12 turned up in at least one of the sampled fragments. Twenty-three fragments supported at least one rare or endangered species. Based on the presence or absence of plant species among fragments, six species were designated potential indicators of sites with high conservation value (Antirhea acutata, Coccoloba diversifolia, Cordia rickseckeri, Guettarda krugii, Plumeria alba and Savia sessilzflora). I \l t” RRhnuh final outconi “fluid also I. Herr) Cam; lhis “time of i Department ‘1 Relation} E This Iiilii‘il and b Forestry. I \ Colon. Carll Discussions deaeloping Melina Coll We Flam ACKNOWLEDGEMENTS I would like to thank my major professor Peter Murphy for support, patience and < belief in what I could do, in addition to his advice on the project and role in shaping its final outcome and for giving me areas sense of the adventure of the field of ecology. I would also like to thank my Guidance Committee — Thomas Burton, Frank Ewers and Henry Campa III, and former member Jose Panero. This research was fimded in part by the USDA Forest Service International Institute of Tropical Forestry, with additional support from the Taylor Fund of the Department of Plant Biology, Michigan State University, and the Ecology, Evolution and Behavioral Biology Program at Michigan State University. This work was vastly assisted by Miguel Canals Mora, Manager of Guanica Forest and by Ariel Lugo of the USDA Forest Service International Institute of Tropical Forestry. I would also like to thank Peter Weaver, Frank Wadsworth, Julio Figueroa Colon, Carlos Rodriguez Pedraza, Millie Alayén and Alberto Rodriguez of the IITF. Discussions with Miguel Canals, Ariel Lugo and Peter Weaver were crucial in developing an understanding of the land, the landscape and the ecosystem. Sandra Molina Colon of the Pontifical Catholic University of Puerto Rico introduced me to many of the plant species of Puerto Rican dry forest and provided undergraduate students who assisted as volunteers. I would like to thank the property owners who allowed me to work on their land, especially Eddie Sepulveda, Miguel Canals, Roberto Carlo, Dr. Acosta and the F as family. iv Genet. . llcttl: . , ‘M illIlE-J ..I .9.' MINA. t; Pal Dir. Stnbji lured ileslc' Clubbi ln’mdi Iiicsl Vv .‘. . . ' [Th I“ I l n I Dr: IC-I Our WOL M iii ng r .aml inal I would also like to thank Kurt Stanley, Skip Van Bloem, John Genet, Kristen Genet, Joseph Cook, Robert Hollister, Lissa Leege, Emilio Font, Eugenio Toro, Karol Morel and Curt Peterson for assistance with data collection, Curtis Clevinger and Winston Johnson for their assistance with plant identification, Brian Dunphy for his assistance in preparing me for the experience of data collection in Puerto Rico and Jason and Pam Kilgore, Bill Chu, Salmaan Quader, Jennifer Clevinger, Christine Podos, Kristi and Eric Thobaben, Timothy Tibbetts, Alicia and Mike Bosela, Virginia Baker, Taylor Reid, David Johnson, Alex Hernandez, Tad Eichler and Ethan Nedeau. I would like to thank Drs. Patrick Webber, Kay Gross, Ray Hammerschmidt, Rich Kobe, Don Hall, Oliver Schabenberger, Susan Hill, Steve Hamilton, Alan Tessier and Sarabjit Tokhie. Also Tony D’Angelo, Tom Harpsted, Sonya Lawrence, Dr. Craig Tweedie, John Mugg, Rajesh Lal, Ournatie Marajh, Rev. Rick Erickson, the MSU Wesley Foundation and Espresso Royale Caffe. I would like to than Sandra Dieffenthaler for leading me to biology, Dr. Colin Clubbe for introducing me to the wonder of the field of ecology and Tyron Kalpee for introducing me to experimentalism. Also Dr. Sanjay Ramdath, Nigel Foster, and Drs. Videsh Rarnbissoon and Barry and Parvi Ollivierra. Finally I would like to thank my family — Ian, Helga, Carol and Karl Ramjohn, Floyd Lucas, and most especially, my wife Lindsay. llSI OF I.‘ IISI 0F lil- CHAPTER Habitat Puerto 05}ch Outline CHIPTER ; SIN.) Rfil: Site Sclec DmColl CHAPTER DRY FORE IUUoduct Ohm \i Clhod: Land; TABLE OF CONTENTS LIST OF TABLES ............................................................................................................. xi LIST OF FIGURES ......................................................................................................... xiii '- CHAPTER 1: INTRODUCTION ....................................................................................... 1 Habitat Fragmentation and the Preservation of Biodiversity ...................................... 5 Puerto Rican Dry Forests .......................................................................................... 14 Objectives ................................................................................................................. 21 Outline of the Study .................................................................................................. 23 CHAPTER 2: METHODS ................................................................................................ 27 Study Region ................................................................................................................. 27 Site Selection ........ 28 Data Collection ............................................................................................................. 34 CHAPTER 3: LANDSCAPE CHANGE AND THE LANDSCAPE ECOLOGY OF THE DRY FOREST ZONE OF SOUTHWESTERN PUERTO RICO .................................... 41 Introduction ................................................................................................................... 41 Objectives ................................................................................................................. 44 Methods ......................................................................................................................... 45 Landscape Characterization ...................................................................................... 45 Landscape Change .................................................................................................... 46 Dynamics of Focal Fragments .................................................................................. 46 Fragment Networks and Connectivity ...................................................................... 47 Results ........................................................................................................................... 48 Landscape Characterization ...................................................................................... 48 Landscape Change .................................................................................................... 52 Dynamics of Focal Fragments .................................................................................. 52 vi Fmgfllt‘ Discussrtr' Forest 1 Laridst .. Dunn: I Pragma- Implica Summa CHAPTER 4 Introducti Fragment Networks and Connectivity ...................................................................... 53 Discussion ..................................................................................................................... 60 Forest Cover .............................................................................................................. 60 Landscape Change .................................................................................................... 64 Dynamics of Focal Fragments .................................................................................. 67 Fragment Networks and Connectivity ...................................................................... 68 Implications ............................................................................................................... 69 Summary ................................................................................................................... 70 CHAPTER 4: COMMUNITY STRUCTURE OF PUERTO RICAN DRY F ORESTS .. 72 Introduction ................................................................................................................... 72 Nested Subsets in Fragmented Communities ........................................................... 73 Objectives ................................................................................................................. 75 Methods ......................................................................................................................... 75 Data Collection ......................................................................................................... 75 Community Characterization .................................................................................... 76 Nestedness in Fragment Species Assemblages ......................................................... 76 Results ........................................................................................................................... 78 Community Characterization .................................................................................... 78 Nestedness in Fragment Species Assemblages ......................................................... 88 Discussion ..................................................................................................................... 91 Community Characterization .................................................................................... 91 Nestedness in Fragment Assemblages ...................................................................... 97 Summary ....................................................................................................................... 98 CHAPTER 5: SPECIES-AREA RELATIONSHIPS OF PUERTO RICAN DRY FOREST PLANTS ON A FRAGMENTED LANDSCAPE. ......................................... 100 Introduction ................................................................................................................. 1 00 vii \lc’i ids. RCS ls- Dis 5er Sun tar. (IN ER l\' PL' ill lntr» uct: Mil Ids Res :5 E lllt‘ 3" isii C lCI;. Sum an (“AP 1R‘ ON A ‘ Iller JCI: Objectives ............................................................................................................... l 04 Methods ....................................................................................................................... 105 Data Collection ....................................................................................................... 105 . Data Analysis .......................................................................................................... 105 Results ......................................................................................................................... 107 Discussion ................................................................................................................... 1 13 Summary ..................................................................................................................... 119 CHAPTER 6: PLANT SPECIES RESPONSES TO LONG-TERM FRAGMENTATION IN PUERTO RICAN DRY FOREST LANDSCAPE .................................................... 121 Introduction ................................................................................................................. 1 2 1 Objectives ............................................................................................................... 1 23 Methods ....................................................................................................................... 124 Results ......................................................................................................................... 125 Range-Abundance-Incidence Patterns .................................................................... 125 Species Profiles ....................................................................................................... 128 Relative Abundance Profiles ................................................................................... 135 Site History ............................................................................................................. 135 Seed Mass ............................................................................................................... 137 Exotic Species Abundance ...................................................................................... 141 Discussion ................................................................................................................... 141 Conclusions... ........................................................................................................... 145 Summary ..................................................................................................................... 146 CHAPTER 7: THE CONSERVATION POTENTIAL OF DRY FOREST F RAGMENTS ON A TROPICAL LANDSCAPE .................................................................................. 147 Introduction ................................................................................................................. 147 Conservation in Fragmented Landscapes ............................................................... 147 viii . If“ . M"; lit” f‘ What constitutes a valuable fragment? ................................................................... 150 Objectives ............................................................................................................... l 53 Methods ....................................................................................................................... 153 Data Collection ....................................................................................................... 153 Species Richness/Species Density .......................................................................... 154 Representativity ...................................................................................................... 1 5 5 Rare and Endangered Species ................................................................................. 161 Indicator species ...................................................................................................... 161 Cluster Analysis ...................................................................................................... 165 Results ......................................................................................................................... 165 Species Richness/Species Density .......................................................................... 165 Representativity ...................................................................................................... 1 66 Rare and Endangered Species ................................................................................. 169 Indicator Species ..................................................................................................... 172 Cluster Analysis ...................................................................................................... 172 Discussion ................................................................................................................... 1 79 Species Richness/Species Density .......................................................................... 179 Representativity ...................................................................................................... 1 80 Rare and Endangered Species ................................................................................. 182 Indicator Species ..................................................................................................... 183 Cluster Analysis ...................................................................................................... 184 Summary ..................................................................................................................... 186 CHAPTER 8: CONCLUSIONS AND RECOMMENDATIONS .................................. 188 Conclusions ................................................................................................................. 1 88 Further Questions ........................................................................................................ 192 ix APPENDIX APPENDIX {CREST I: E llTEl’clll APPENDIX 1: F RAGMENT-SPECIES OCCURRENCE MATRIX ............................ 196 APPENDIX 2: DOMINANCE-DIVERSITY CURVES FOR PUERTO RICAN DRY FOREST FRAGMENTS ................................................................................................ 223 LITERATURE CITED ................................................................................................... 263 ' ‘L‘I'zfi . [$4.th -- . ~ driiil. ~ Ml -. 50"“1‘13‘. AV I, ‘ 1“ . IaDlC ' RRRT Inklif Gumi lTRiLE LIST OF TABLES Table 2.1: Sampling design — the measurements collected for each of the three sampling designs used to study dry forest fragments and reference plots in Guanica Forest, southwestern Puerto Rico. ........................................................................................ 35 Table 2.2: A summary of the sampling design employed for dry forest fragments and reference plots in Guanica Forest, southwestern Puerto Rico. ................................. 37 Table 2.3: A summary of the data collected for dry forest fragments and reference plots in Guanica Forest, southwestern Puerto Rico (see text for details). ............................. 39 Table 3.1: Distribution of forest cover (based on 1993 aerial photographs), by size class, in the dry forest zone, southwestern Puerto Rico. Column totals may not precisely match the values in the columns due to rounding. .................................................... 49 Table 3.2: Transition probabilities between Agriculture, Open Forest and Closed Forest in dry forest fragments of southwestern Puerto Rico in the period 1936-1989. Transitions are based on the probability of a point in a given land-cover class ending in the same or another land-use class between the named time periods. .................. 51 Table 3.3: History of studied dry forest fragments, southwestern Puerto Rico, in the period 1936-1993 with additional notes on changes in the period 1995-1998. ........ 54 Table 3.4: The characteristics of dry forest fragment clusters separated from each other by more than 500 m, southwestern Puerto Rico ....................................................... 58 Table 4.1: Summary of the structural characteristics of the plant community in Guanica Forest, Puerto Rico based on this study and published data. .................................... 84 Table 4.2: Summary of the structural characteristics of studied dry forest fragments: basal area and stem density. ............................................................................................... 86 Table 4.3: The distribution of stems among single- and multi-stemmed trees in Puerto Rican dry forest. ........................................................................................................ 90 Table 5.1: Results for a General Linear Model analysis of the relationship between fragment species richness, fragment area and fragment history (see text for definitions of the terms). ......................................................................................... 107 Table 5.2: The values of parameters c and z (and standard errors of the estimates) and the proportion of variance explained by the regression (R2) obtained from species-area regressions for all fragments and various subsets of dry forest fragments, southwestern Puerto Rico. ...................................................................................... 109 xi labia 5.3: ‘l'lfiii r Guiu‘i: Title 5.4: l‘ fr arr: V lib-it Di I. I lab-166.1: I Specie Table 6.3: ‘ and tr Tame C 5 Table 5.3: Parameter estimates obtained using the sigmoid Hillsm,e function (Lomolino 2000) for an array of Puerto Rican dry forest fragments alone and together with Guanica Forest, a 4000-ha reserve. ......................................................................... l 10 Table 5.4: Estimates of the parameters c and z of the species-area curve for intra- fragment sample curves of dry forest fragments, southwest Puerto Rico ............... 111 Table 5.5: Least square mean estimates of species density (sensu Whittaker, 1975) based on fragment history in Puerto Rican dry forest fragments ...................................... 110 Table 6.1: Criteria used to assign a Range score to Puerto Rican dry forest plant species ................................................................................................................................. 124 Table 6.2: Pearson correlations between Abundance (among fragments) and Fragment Species Richness for each of 10 dry forest species in southwestern Puerto Rico. . 135 Table 6.3: The relationship between the distribution of selected dry forest plant species and fragment history in southwestern Puerto Rico ................................................. 136 Table 7.1: Reference List of Representative Species sampled from Guanica Forest ..... 156 Table 7.2: Rare and endangered plant species which are either present in the dry forest zone, southwestern Puerto Rico or have been recorded there in the past. The list was compiled on the basis of Federally listed endangered species, herbarium collections (Figueroa Colon and Woodbury 1996), or on their distribution in Guanica Forest (Quevedo et al. 1990). ............................................................................................ 162 Table 7.3: Standardized residuals of the species-area curve (standardized residuals by x2 transformation) and “species density” (calculated from the species-area curve) of dry forest fragments in southwestern Puerto Rico. Symbols refer to fragments with significantly more species than expected (+), significantly fewer species than expected (-) or that did not differ significantly from expectations (0). Rank refers to the rank order of fragments in terms of species density. ........................................ 167 Table 7.4: Proportion of the species composition of the three main associations in Guanica Forest that are represented in sampled dry forest fragments, southwestern Puerto Rico. DeF = Deciduous Forest, SEv = Semi Evergreen Forest, ScF = Scrub Forest ....................................................................................................................... 170 Table 7.5: Distribution of rare and endangered species among studied dry forest fragments in southwestern Puerto Rico. ................................................................. 171 Table 7.6: The number of plant species among the six proposed Indicators of fragments of high conservation potential which were present in studied dry forest fragments in southwestern Puerto Rico. ...................................................................................... 177 Table A1: Species-site occurrence matrix for plant species in Puerto Rican dry forest fragments ................................................................................................................. 197 xii Fieure 1.1. 1973 l Diagr. F igure 2.3. Puertt traini- Getter: detail: Figure 3.\ '. south' the g: detaii south delin Figure 3; Puer frag: Figure 4. Sent I‘I‘ifit‘ and Figure 4. con- Clu Clus FIEUre 4 Pue ta.- t Figure \ - w F‘é’flre S CU:- LIST OF FIGURES Figure 2.1. Map of Puerto Rico showing the dry forest zone (after Ewel and Whitmore 1973) and the approximate locations of weather stations used to construct Climate Diagrams. Letters refer to the order of the Climate Diagrams in Figure 2.2. .......... 29 Figure 2.2: Climate Diagrams (Walter and Lieth 1967) for the dry forest life zone in Puerto Rico based on NOAA 1971-2000 climate normals. Heavy shading areas (rainfall >100 mm mo") represent water excess, while light shading represent water deficits. ...................................................................................................................... 29 Figure 2.3: The location of studied forest fragments in the dry forest life zone of southwestern Puerto Rico. For details of the construction of the base map see Chapter 2. Approximate scale 1: 200 000. Inset map of Puerto Rico showing the dry forest zone (after Ewel and Whitmore, 1973) and the area from which the detailed map was selected. ........................................................................................ 33 Figure 3.1: The distribution of forest cover from west to east in the dry forest zone, southwestern Puerto Rico. The boxed area on the map shows the area covered by the graph. Each block was approximately 7.5 km wide. See text for additional details. The scale refers to the map of Puerto Rico. The highlighted area in the south of the island lies within the dry forest life zone (sensu Holdridge 1967) as delineated by Ewel and Whitmore 1973) .................................................................. 50 Figure 3.2: Map showing the extent of forest cover in the study region of southwestern Puerto Rico. Arrows indicate gaps of more than 500 In between clusters of fragments. Approximate scale 1:170 000. ............................................................... 59 Figure 4.1: Dominance-diversity curves for dry forest fragments and Guanica Forest, southwestern Puerto Rico. Fragments are presented grouped according to hierarchical clustering of Jaccard similarities of species composition (see Figure 4.2 and text for details). .................................................................................................. 79 Figure 4.2: Hierarchical cluster dendrograms of Jaccard coefficients based on the species composition of Puerto Rican dry forest fragments, southwestern Puerto Rico. Clusters were delineated on the basis of a cut-off distance of 0.8. Numbers refer to clusters referred to in the text .................................................................................... 89 Figure 4.3: Dominance-diversity curves for four dry forest fragments in southwestern Puerto Rico. Site 38 (circles) and Site 9 (triangles) are the two lowest-diversity fragments, while Sites 4 (squares) and 5 (crosses) are among the highest diversity. 94 Figure 5.1: Species-area curve for Puerto Rican dry forest fragments. Note log scale. 108 Figure 5.2: Estimates of the fitted parameters c and z of intra-fragment species-area curves of dry forest fragments in southwestern Puerto Rico. Species-area curves )(iii uere r for all Figure (3.1 '. tlhcrt‘ ‘5 m‘ ~ lorest = Figure 6.1. I per 25 bio-get. the cat abundt Figure 6.3: I specie plot in Specie Figure 6.4; lragrr tet‘ere lf-ii. Fig‘xe 6.5 l 60-.- h‘iStt: CYTOI Figure 6. in u ...... were modeled using the power fiinction S = cAz. Correlation coefficient r = 0.564 for all data, and 0.776 if the outlier, Site 9, is excluded. ........................................ 112 Figure 6.1: The relationship between Frequency (number of plots in Guanica Forest where a species occurs) and Mean Abundance (the mean number of individuals per . 25 m2 plot in Guanica Forest where the species is present) for Puerto Rican dry forest species. Data are presented as mean abundance d: 1 standard error ............. 126 Figure 6.2: The relationship between Mean Abundance (the mean number of individuals per 25 m2 plot in Guanica Forest where the species is present) and Range (the biogeographic breadth of the species distribution; see Table 6.1 for the meaning of the categories) for Puerto Rican dry forest species. Data are presented as mean abundance i 1 standard error. ................................................................................. 127 Figure 6.3: The relationship between Incidence (the number of fragments in which a species is present) and Mean Abundance (the mean number of individuals per 25 m2 plot in Guanica Forest where the species is present) for Puerto Rican dry forest species. .................................................................................................................... 129 Figure 6.4: Abundance Profiles (probability of the species being present as a function of fragment species richness) for dry forest species with the highest abundance in reference plots in Guanica Forest (a-e) or the highest incidence among the fragments (f-j). ......................................................................................................................... 130 Figure 6.5: The number of species with small (01-20 mg), medium (20-60 mg), or large (60-473 mg) seeds in Puerto Rican dry forest fragments with different disturbance histories. Data are present as mean number of species per fragment :t 1 standard error. ........................................................................................................................ 138 Figure 6.6: The relationship between seed mass and Incidence (the number of fragments in which a species is present) in dry forest fragments in southwestern Puerto Rico. ................................................................................................................................. 139 Figure 6.7: The mean number of exotic species per 25-m2 plot in Puerto Rican dry forest fragments with different land-use histories. Data are presented as mean number of individuals per 25 m2 plot :I: 1 standard error. Relict fragments are those in which 2 75% of the fragment had supported forest cover continuously since 1936 (‘old growth’), Regrowth fragments were 5 25% ‘old growth’, while Mixed fragments are 25-75% ‘old growth’ forest ..................................................................................... 140 Figure 7.1: Incidence functions (sensu Diamond 1975) for the six species present in dry forest fragments in southwestern Puerto Rico which met the criteria selected to identify indicators of sites with high conservation value. Species richness classes each consist of 5-6 fragments grouped on the basis of total plant species richness. Indicator species were defined as those which were present in less than 20% of the two most species-poor classes and were present in more than 80% of the fragments in the most species-rich class. ................................................................................. 173 xiv Fieure I33. Fi.‘ure 83 Figure B: . \ Figure BC. Figure 82. Figure B'.‘ i ,— Figure 83. Figure 82.? Faeuzj Fig-“re BIL HaeBlt Hamult HemBz; Pierre 82.: EFWBlZ Figure B21 Figure Bl} Figure82.: EfimBl: h. F ‘EJie Bl: HfleBrx Figure 7.2: Hierarchical clustering dendrograms of Puerto Rican dry forest fragments based on scores of their Conservation Potential. .................................................... 178 Figure 32.1: Dominance-diversity curve of Site 1, southwestern Puerto Rico .............. 224 Figure 32.2: Dominance-diversity curve of Site 2, southwestern Puerto Rico .............. 225 Figure 32.3: Dominance-diversity curve of Site 3, southwestern Puerto Rico .............. 226 Figure 32.4: Dominance-diversity curve of Site 4, southwestern Puerto Rico .............. 227 Figure 32.5: Dominance-diversity curve of Site 5, southwestern Puerto Rico .............. 228 Figure 32.6: Dominance-diversity curve of Site 6, southwestern Puerto Rico .............. 229 Figure 32.7: Dominance-diversity curve of Site 7, southwestern Puerto Rico .............. 230 Figure 32.8: Dominance-diversity curve of Site 8, southwestern Puerto Rico .............. 231 Figure 32.9: Dominance-diversity curve of Site 9, southwestern Puerto Rico .............. 232 Figure 32.10: Dominance-diversity curve of Site 10, southwestern Puerto Rico .......... 233 Figure 32.11: Dominance-diversity curve of Site 1 1, southwestern Puerto Rico .......... 234 Figure 32.12: Dominance-diversity curve of Site 12, southwestern Puerto Rico .......... 235 Figure 32.13: Dominance-diversity curve of Site 13, southwestern Puerto Rico .......... 236 Figure 32.14: Dominance-diversity curve of Site 14, southwestern Puerto Rico .......... 237 Figure 32.15: Dominance-diversity curve of Site 15, southwestern Puerto Rico .......... 238 Figure 32.16: Dominance-diversity curve of Site 16, southwestern Puerto Rico .......... 239 Figure 32.17: Dominance-diversity curve of Site 17, southwestern Puerto Rico .......... 240 Figure 32.18: Dominance-diversity curve of Site 18, southwestern Puerto Rico .......... 241 Figure 32.19: Dominance—diversity curve of Site 19, southwestern Puerto Rico .......... 242 Figure 32.20: Dominance-diversity curve of Site 20, southwestern Puerto Rico .......... 243 Figure 32.21: Dominance-diversity curve of Site 21, southwestern Puerto Rico .......... 244 Figure 32.22: Dominance-diversity curve of Site 22, southwestern Puerto Rico .......... 245 Figure 32.23: Dominance-diversity curve of Site 23, southwestern Puerto Rico .......... 246 Figure 32.24: Dominance-diversity curve of Site 24, southwestern Puerto Rico .......... 247 XV Figure 83.. Figure Bil Ewall FlEure BIL Figure 33. FREE: Figure 32.25: Figure 32.26: Figure 32.27: Figure 32.28: Figure 32.29: Figure 32.30: Figure 32.31: Figure 32.32: Figure 32.33: Figure 32.34: Figure 32.35: Figure 32.36: Figure 32.37: Figure 32.38: Dominance-diversity curve of Site 25, southwestern Puerto Rico .......... 248 Dominance-diversity curve of Site 26, southwestern Puerto Rico .......... 249 Dominance-diversity curve of Site 27, southwestern Puerto Rico .......... 250 Dominance-diversity curve of Site 28, southwestern Puerto Rico .......... 251 Dominance-diversity curve of Site 29, southwestern Puerto Rico .......... 252 Dominance-diversity curve of Site 30, southwestern Puerto Rico .......... 253 Dominance-diversity curve of Site 34, southwestern Puerto Rico .......... 254 Dominance-diversity curve of Site 35, southwestern Puerto Rico .......... 255 Dominance-diversity curve of Site 36, southwestern Puerto Rico .......... 256 Dominance-diversity curve of Site 37, southwestern Puerto Rico .......... 257 Dominance-diversity curve of Site 38, southwestern Puerto Rico .......... 258 Dominance-diversity curve of Site 39, southwestern Puerto Rico .......... 259 Dominance-diversity curve of Site 40, southwestern Puerto Rico .......... 260 Dominance-diversity curve of Site 41, southwestern Puerto Rico .......... 261 xvi CHAPTI \i'ii l excellent 3‘ ludseape : Like much wsystemt What surm inethe it “3:" OF lent be eonside ”that ha: I already 0t Tl Guanicai L“NESCt Itiirest‘ ll MOT)". I fix Charc er 01‘ 19. 20in ”"~l ar 30?)? ' .-r “impara CHAPTER 1: INTRODUCTION With along history of human disturbance, the south coast of Puerto Rico is an excellent place to study the ecology of tropical dry forest fragments. Scattered across the landscape is an archipelago of fragments that vary in size, age and disturbance history. Like much of the world’s tropical dry forest, only a small portion of the original ecosystem survives (Janzen 1988a, b, Allen 2000, Trejo and Dirzo 2000), and most of what survives is in small fragments. Unlike many other areas that have been studied to date, the fragmentation of this landscape is not a new phenomenon. While there is no way of knowing the extent to which the distribution of species among the fragments may be considered to represent a final “equilibrial” distribution, many of the early changes that have been documented elsewhere (e. g. , Laurance et al. 2002b) are likely to have already occurred in this system. This array of fragments is complemented by Guanica Forest (Bosque Estatal de Guanica), a relatively large, well-studied protected area that forms the core of a UNESCO Man and the Biosphere Reserve. Although Guanica Forest is not a pristine dry forest, it forms the best reference against which to compare these fragments. During its history, Guanica Forest has experienced a wide range of disturbances including cutting for charcoal and fence-post production, plantation forestry based on exotic species (Lugo er al. 1978, Canals Mora 1990, Wadsworth 1990, Molina Colon 1998, Erickson et al. 2002) and agricultural crop production (based on corn, squash and peas; Erickson et al. 2002). The fact that Guanica Forest has experienced a range of impacts that are comparable to those experienced by the fragments makes it a better reference against which to 5 condition ' fragment: 1 \\ I: prior to. th. mnpuun these liner. forest or - and tried Fro. llhitmore forest e‘ieu' inniallx SI Occur in II Effects of hitter (I M file C It) land area “(1113:“ which to compare the effects of fragmentation. Had Guanica Forest been in a ‘pristine’ condition (if such a thing existed), it would be impossible to differentiate the effects of fragmentation from those of disturbance. While the presence of dry forest fragments outside of Guanica Forest was known prior to the initiation of this project, little information was available about their species composition and habitat quality. This study addressed the question of what exists in these fragments - whether they supported forest cover that resembled that of ‘intact’ dry forest, or whether they supported species-poor communities dominated by exotic species — and tried to identify some of the factors that correlated with habitat quality. Tropical deforestation constitutes a major threat to biodiversity (Wilson 1985, Whitmore and Sayer 1992). While some species are eliminated directly by the process of forest clearing, others survive in remnant patches. Among the species that were present initially, some will subsequently be lost as a consequence of post-isolation changes that occur in the fragments (Turner 1996, Laurance et al. 2002b). In addition to the direct effects of forest removal, deforestation also tends to result in the degradation of remnant habitat (Laurance et al. 2002b). Skole and Tucker (1994) estimated that while only 6% of the closed-cover Amazonian forests had been cleared by 1988, an additional 9% of the land area had been affected by fragmentation and edge-related phenomena. Between 1978 and 1988 twice as much land was subject to human-induced modification as was actually cleared (Skole and Tucker 1994). It is likely that deforestation in Puerto Rico followed a similar pattern to what has been observed elsewhere in Latin America since the 19605 (Rudel er al. 2002). Deforestation along a forested frontier usually follows a pattern of selective logging ,. l r ~t l l'ugh ET“L tBierregli‘i clear land tlkrserup l dee‘tines 5‘; clear than ; cieared lari f :tility has tuliiration ln‘tasion b Once this. Feature tR OI {DI the (‘high grading’) followed by the migration of settlers along the logging access roads (Bierregaard and Dale 1996, Turner et al. 2001 , Laurance et al. 2002a). These settlers clear land for agriculture, usually a mixture of subsistence agriculture and cash crops (Boserup 1964, Browder 1996, Turner et al. 2001). Land is cultivated until its fertility declines sharply and then new land is cleared. Since secondary forest is usually easier to clear than primary forest (e. g., Freeman, 1955 cited in Lawrence et al., 1998), previously cleared land is frequently cultivated again once it has recovered to the stage where fertility has been restored (Kleinman et al. 1996, Turner et al. 2001). Repeated cultivation with too short a fallow can lead to permanent loss of fertility and subsequent invasion by fire prone grasses (e. g., Albers and Goldbach 2000, Turner et al. 2001). Once this stage is reached, land is often turned over to large landowners who convert it to pasture (Rudel et al. 2002). In other areas, forest is directly converted, either to pasture or for the establishment of large-scale agricultural schemes often based on government subsidies (Geoghegan et al. 2001, Steininger et al. 2001a). Rudel er al. (2002) considered this “hollow frontier” model to reflect one of the major drivers of permanent deforestation since in these cases secondary forests are not allowed to reclaim the landscape once the agricultural frontier advances. On the other hand, Boserup (1964) considered agricultural intensification as a consequence of “increasing population size to be the main driver of deforestation. Changes of both sorts have been documented in Amazonia (Browder 1988, Ozorio de Almeida 1992, Laurance 1999), but in the absence of a detailed historical study there is on way to determine whether either of these models describes the actual pattern of landscape transformation that occurred when Puerto Rico was initially deforested. P.; Laurance Forests; sf. Brazil. bu let'ill‘il} 0T loragrieul tropics int and fire (l Elut’. 199s. 2093)- Or. oleattle a1 (a 5 Hz IFE c Blackmok As Clt’len md the Carihi. 190.39,.“ r... Patterns of deforestation are similar in the wet and dry tropics, but deforestation usually begins sooner and is more complete in dry forest zones (Steininger et al. 2001b, Laurance et al. 2002b). Burn efficiencies are higher in tropical dry forests than in wetter - forests; slash fires consume 60-90% of pre-bum biomass in dry forests in Mexico and Brazil, but only 40-60% of the biomass in wet forests (Kauffman er a1. 2003). The higher fertility of dry forest soils and the fact that they are easily cleared makes them attractive for agriculture. An alternative pathway to deforestation observed in the seasonally dry tropics involves a process of gradual degradation through the combined action of grazing and fire (Uhl and Buschbacher 1985,Murphy and Lugo 1986a, Janzen 1988a, b, Nepstad er al. 1999, Blackmore and Vitousek 2000, Nepstad et al. 2001, Cochrane and Laurance 2002). Once grazing enters the equation the resulting forest is more fire-prone; removal of cattle also makes these forests even more fire-prone, since tall ungrazed grass is able to carry fire deeper into forest fragments (e. g. , J anzen 1988a, Johnson and Wedin 1997, Blackmore and Vitousek 2000). As a group, dry forests are the most threatened biome in the Neotropics. Of the eleven major habitat types recognized by Dinerstein et al. (1995) in Latin America and the Caribbean, dry broadleaf forests were the most threatened with 28 of 31 ecoregions (90.3%) falling in the Critical or Endangered categories, with the other three (9.7%) falling in the Vulnerable or Relatively Stable categories (Dinerstein et al. 1995). In many cases, dry forest landscapes consist of little more than a scattering of disturbed fragments in a sea of deforestation (Janzen 1986, 1988b, Murphy and Lugo 1995, Murphy et al. 1995, Smith 1997, Allen 2000, Trejo and Dirzo 2000). l - x ' en UCElI‘dLllti J . Q ‘ ' GEEIm\.llIli Habitat Fragmentation and the Preservation of Biodiversity Definition of Fragmentation Some authors have used the term “fragmentation” synonymously with habitat destruction (e. g. , Andre'n 1994) while others have used it to describe both habitat destruction and the processes by which the remnant habitat is broken into smaller pieces (e.g., Diffendorfer et al. 1995, With and Crist 1995, With et al. 1997, Dooley and Bowers 1998). Others have used the term fragmentation in the narrow sense, clearly distinguishing between habitat destruction and fragmentation (e. g., Bascompte and Solé 1995, McGarigal and McComb 1995, Lindenmayer and Possingharn 1996, Neve et al. 1996). In this study the term “fragmentation” has been used in the second sense. Bunnell (1999) considered the term ‘habitat fragmentation’ to include six discrete concepts. Modification of intact habitat may include (1) reduction of the total area; (2) increase in the amount of edge; (3) decrease in the amount of interior habitat; (4) isolation of a habitat fragment; (5) increase in the number of habitat fragments; and (6) decrease in the average size of a habitat fragment. Different taxa may respond in different, and even contradictory, ways to these processes, making “habitat fragmentation” a complex phenomenon (Haila 2002, Laurance et al. 2002b) . In addition, the term ‘habitat fragment’ is largely undefined in the literature. For the purpose of this dissertation, a habitat fragment is considered to be a unit of an ecosystem whose boundaries and associated conditions have been determined by human-influenced or other disturbances. 'agrrienn hedgerou » major ile‘o uhether it Stnberlol' EXlt’lll l0 \‘ filiahalent 0\ IBDFFP) I in the hurr BUFFP m in man}. 0 fragment-a lh'ubbcn - Sjatial Owization of Remnant Habitat Habitat destruction tends to proceed in a non-uniform manner which leaves forest fragments concentrated in ravines, riparian strips, steep hillsides, fence lines and hedgerows (Forman 1995, Kahn and McDonald 1997, Lamb et al. 1997). One of the major debates in community ecology in the 19703 and 803 was the SLOSS debate — whether it is preferable to protect a Single Large gr Several Small reserves (e. g. , Simberloff and Abele 1979, Wilcox and Murphy 1985). At the heart of the issue was the extent to which a certain amount of habitat distributed over several parcels was equivalent to the same area of habitat in a single parcel. Over the last 22 years, the Biological Dynamics of Forest Fragments Project (BDF F P) near Manaus, Brazil, has convincingly demonstrated the value of large reserves in the humid tropics (Laurance et al. 2002b), although Thomas (2004) suggests that the BDFFP model represents an extreme case of sensitivity to fragmentation, and that forests in many other areas (such as Malaysia, where he worked) are much less sensitive to fragmentation. The fact that many species are rare and have patchy distributions (Hubbell 1979, Hubbell and Foster 1986, Pittman et al. 1999) makes the location of a fragment an important determinant of its ability to conserve biodiversity. The initial inclusion of a species in a fragment is essentially a sampling effect, although mobile species displaced by deforestation may move into remnant habitat (Bierregaard and Lovejoy 1989). Scattered fragments may also be able to sample more habitat types than can a large block of habitat, but they are not guaranteed to do so, since certain types of habitat are more prone to be cleared (Lamb et al. 1997). Kahn and McDonald (1997) suggested that three factors dictate where forest tends to persist — the physical nature of the tartan to non-lie treasuh... IleDona'. Fr \lu Anhl The size . fragment “egalh’e flfldlnm C. x the terrain, the legal status of the land, and whether it is economically viable to convert it to non-forest uses. Deforestation is likely to concentrate on highly productive lowland areas which lack legal or cultural protection (Lebbie and F reudenberger 1996, Kahn and McDonald 1997, Lamb er al. 1997). Remnant fragments often occupy the least desirable land and may not provide suitable habitat for many species. Fragment Area Fragment area is a primary determinant of species richness (Arrhenius 1921, MacArthur and Wilson 1967, Rosenzweig 1995, Lomolino 2000, Whittaker et al. 2001). The size of a fragment will influence the number of species that are present when the fragment is created and the probability of species’ persistence. Given the relatively short time frame of most fragmentation studies, area effects have mostly been demonstrated for animals, but some studies have also demonstrated this for plants (Leigh er al. 1993, Turner 1996). MacArthur and Wilson (1967) predicted that extinction rates should be negatively correlated with fragment area, a prediction that has been supported by the findings of the BDFF P study (Laurance et al. 2002b and references therein). On the other hand, some studies have produced results that conflict with these predictions. As Thomas (2004) pointed out, patchily distributed species may be less sensitive to fragment area than are unifome distributed species. Once a patchily distributed species is present in a fragment it is more likely to exist in a viable population than is an evenly dispersed species (which is likely to be uniformly rare throughout its range). Evenly dispersed species exist in populations that are a function of fragment area. In their review of fragmentation experiments, Debinski and Holt (2000) found many studies \\ tere ear: concur: lzllllll It! propomo‘ external t influence tonseoui laadsear Slii‘S 3W finitely studies which did not show an area effect, although they pointed out that many of them were carried out on a time-scale that may be too short to elicit area-driven changes in the community. In addition, Turner et al. (1996), Pither and Kellman (2002) and Thomas (2004) found that small tropical forest fragments were able to support a substantial proportion of the native community. Edge Effects Natural areas surrounded by human-altered landscapes are subject to “the eternal external threat” (Janzen 1986). The smaller the fragment, the more strongly itwill be influenced by these external factors collectively termed “edge effects”. One of the major consequences of habitat fragmentation is an increase in the proportion of edge in a landscape (Chen et al. 1992, Skole and Tucker 1994, Kapos et al. 1997). While larger sites are likely to consist of both “edge” and “interior” habitat, smaller sites may be entirely edge (Laurance 1991, Irnre 2001). Air temperature, air moisture, vapor pressure deficit (VPD), soil moisture, light intensity and levels of photosynthetically active radiation (PAR) change at edges. Kapos (1989) found differences in air temperature and VPD up to 60m into fragments at the BDF F P site in Amazonia but this distance decreased as the edge matured and became more closed (Williams-Linera 1990, MacDougall and Kellman 1992, Camargo and Kapos 1995, Kapos et al. 1997). Murcia (1995) considered orientation and physiognomy to be important in moderating the intensity of abiotic edge effects. While edge orientation is likely to be more significant at mid- to high latitude sites (e. g. , Palik and Murphy 1990, Matlack 1994), Turton and Freiburger (1997) found that abiotic factors N were inta. lautudet T | \\ microeliri' noting a. dounstre. “ top) Ill. older and Uuranee . F‘lfifiu‘ 7 “I lift t l- . L1m‘iresu were influenced more by edge aspect than by distance from the edge in a relatively low latitude (17° S) Australian forest fragment. Wind is a major source of damage to trees along edges and unlike changes in microclimate, wind damage does not decline as edges age (Laurance et al. 2002b). When moving air encounters a forest edge, eddies are created upstream and turbulence downstream (Ghuman and La] 1987) which can be powerful enough to break or uproot canopy trees (Williams-Linera 1990, Esseen 1994, Laurance 1997). As edges become older and less permeable, downwind turbulence increases (Savill 1983, Laurance 1997, Laurance et al. 2002b). Since fragments generally have a larger proportion of “edge” and a smaller proportion of “interior” than does continuous forest (Laurance 1991, Forrnan 1995, 11an 2001), one would expect to find more wind damage in fragments than in continuous forest. However, Van Bloem et al. (in press) found no significant difference in the amount of damage experienced by fragmented or continuous dry forest in Puerto Rico following Hurricane Georges in 1997. High levels of tree mortality have been recorded near edges (Lovejoy et al. 1986, Laurance 1991, Leigh et al. 1993, Ferreira and Laurance 1997, Laurance et al. 1998, Mesquita et al. 1999, Laurance et al. 2002b). Higher levels of tree mortality and wind- throw result in an increase in disturbance-associated species (e. g. , the pioneer tree Cecropia sciadophylla increased 33-fold along the edges of the BDFF P fragments over a 20 year period; Laurance eta1., 2001). These changes, which have been termed hyperdynamism, result in an intrinsically less stable community in forest fragments (Laurance 2002, Laurance et al. 2002b). lr. th‘mls‘ l. W. Tu: ul. 1997. 3 olehango same m. . 551376 dens Thomas s; llhere La; Worked is [10,? [I'Pl‘ior' Breeze“ “ Ed. (“Eda”) allUW‘ing ll likely to at 836‘“ 0f c make the ll Patterns 0f are What Wall BlOer his a 103g 1 FL‘ . “spend [0 In contrast with the BDFF P model (which he calls the “things fall apart” model), Thomas (2004) pointed out that other studies (Kellman 1996, Kellman et al. 1996, Turner 1996, Turner and Corlett 1996, Corlett and Turner 1997, Harrington et al. 1997, Turner et- al. 1997, Kellman er al. 1998, Pither and Kellman 2002) have found relatively slow rates of change in tropical forest fragments in some areas (what he calls the “more of the same” model) —- in essence, that community change in forest fragments is driven by the same demographic processes that drive the dynamics of continuous blocks of forest. Thomas suggests that one of the important differences between the Amazonian fragments where Laurance formulated his theory and the Southeast Asian fragments where Thomas worked is one of wind velocity — Peninsular Malaysia experiences neither trade winds nor typhoons, while Central Amazonia experiences a continent interior “Amazon River Breeze” which commonly exceeds 15 km h'l. Edge effects may be experienced differently by dry forest fragments. Dry forests (especially in the insular Caribbean) have more open canopies than do wetter forests, allowing more light and wind penetration. These differences in canopy architecture are likely to affect interactions with air currents. The drier conditions may increase the effects of desiccation, but the fact that these trees are adapted to drier conditions may make the impact of this desiccation less severe. Research is needed to determine if the patterns of edge effects observed in wetter forests are the same for dry forests, or if there are qualitative differences between wet and dry forests. As Van Bloem and colleagues (Van Bloem et al. 2003, Van Bloem et al. in press) pointed out, Puerto Rican dry forest has a long history of wind disturbance via hurricanes and tropical storms, and is able to respond to wind disturbance through the production of sprouts even in undamaged trees 10 ([0 \K‘lllt‘ii history 07 related [‘3' importan. directl} a. Fraenent Ill invasion I W. La; | and Gilfe. IOgether u. TaSFIlElnia: and “ind- AUS'JZIllat‘. Th in‘v‘élsion h Resting 51 5&ch In It l9? (.1 I - Hills". (to which phenomenon they attribute the multi-stemmed nature of these forests). This history of wind disturbance in evolutionary time suggests that dry forests in the insular Caribbean are likely to be pre-adapted to the wind stresses that are likely to be associated with fragmentation. However, in the absence of any studies directly addressing edge- related phenomena in these forests this remains purely speculative. Despite the importance of edge-related phenomena in fragmentation studies, this study did not directly address edge-related questions. Eggment Invasibility Habitat fragments surrounded by a matrix of altered vegetation are susceptible to invasion by the species that dominate the matrix (Janzen 1983, 1986b, HOpkins et a1 . 1990, Laurance 1991, Fensham 1995). J. B. Kirkpatrick and L. Gilfedder (Kirkpatrick and Gilfedder 1995, Gilfedder and Kirkpatrick 1998) found that surrounding vegetation, together with grazing, were the major determinants of fragment integrity in subhumid Tasmanian habitat fragments. Willson and Crome (1989) found both animal-dispersed and wind-dispersed seeds were transported up to 80 m inward from the edge into Australian rainforest fragments. The position of fragments on the landscape may also increase the probability of invasion by non-forest species. In open landscapes, forest fragments can be important roosting sites for birds and bats. The behavioral patterns of birds and bats concentrate seeds in roosting or feeding areas (Snow 1962, Livingston 1972, Howe and Primack 1975, Howe 1977, Fleming and Heithaus 1981, Glyphis et al. 1981, Uhl er al. 1981, Debussche et al. 1982, Uhl et al. 1982, McDonnell and Stiles 1983, Guevara et al. 1986, ll I— l ‘37? .v_-_-,.._ -t""._._ I 3-.....— llcfiana 1093.\L Graffiti t'. ' 1110\6 St‘e‘ Gazaar. heahun; Hannei enakdtr "* "AHLI Pl“.\quir‘.1 .,‘____ Ri PUpulalit‘ir lCEUghle} gFUCUC div ”isms ll the}: are It moderate} l-ial‘le p0I IF“ 63th: I996). Hi pgpuhnhn McClanahan and Wolfe 1987, Janzen 1988a, Guevara et al. 1992, Guevara and Laborde 1993, McClanahan and Wolfe 1993, Guevara et al. 1998, Toh et al. 1999, Galindo- Gonzalez et al. 2000). While much has been said about the ability of frugivorous birds to . move seeds from forest fragments into abandoned pasture (da Silva et al. 1996, Martinez- Garza and Gonzalez-Montagut 1999, Ortiz-Pulido et al. 2000) , little has been said about the ability of frugivores to move the seeds from the matrix into fragments (Aldrich and Hamrick 1998), despite the fact that this may be a significant factor in the degradation of isolated fragments. Population Dynamics and Extinction Rare species are usually assumed to have high extinction probabilities since small populations are at risk as a consequence of environmental or demographic stochasticity (Caughley 1994). The viability of small populations can also be affected by the loss of genetic diversity. Most tropical trees are outbreeders with complex incompatibility systems (Bawa 1974, Zapata and Arroyo 1978, Bawa er al. 1985, Bawa 1990). Since they are less prone to inbreeding in natural conditions, outbreeders are likely to carry a moderately high genetic load of slightly deleterious alleles (Lande 1995). Minimum viable population sizes for tropical forest trees needed to ensure long-term survival have been estimated at effective populations sizes (N e)l of about 5000 (Alvarez-Buylla er a1. 1996). However, the studies that led to these conclusions were not done on insular populations. Species native to the Greater and Lesser Antilles should be adapted to much Ne rs defined as the srze of an Idealrsed population that would have the same amount of Inbreeding or random gene drift as a given real population (Kimura and Crow 1963, Alvarez-Buylla et al. 1996). 12 imlllf l:l.el}' l Altar: partiallj risi; bee the rum depend l ettirtctie seneseen of comp] F. fifigmentt isolated p 'reseue el‘ lSllltelj' it: ”We are ; [halt l C 10.: 'll'k‘iralia, R‘mfi er i; ”errant l I. ‘1 smaller populations than are mainland populations; historically small populations are likely to be more inbred and, as a consequence of this, to carry lower genetic loads (Alvarez-Buylla er al. 1996). This makes them less susceptible to inbreeding depression (reduced viability, seed production and growth rates caused by the segregation of partially recessive lethal alleles). Species that are tightly tied to mutualists are at added risk because they are likely to go extinct if fragments do not support viable populations of the mutualist (Bond 1994, Nason er al. 1997). Species that are “seed limited” — those that depend on seed production to replace senescent stems — are likely to be at higher risk of extinction than are “resprouters”, which depend primarily on sprouting to replace senescent stems (Kruger et al. 1997, Bond and Midgley 2001). These factors add levels of complexity beyond the simple assumptions relating rarity with extinction risk. The spatial arrangement of fragments may also influence the ability of a fragmented system to maintain plant populations. The probability of extinction of an isolated population can be reduced as a consequence of immigration — the so-called ‘rescue effect’ (Brown and Kodric-Brown 1977). While the distance between fragments is likely to be a key factor in determining the amount of seed flow among fragments, there are few data as to the actual inter-fragment distances that birds and bats are likely to move seeds. Da Silva et al. (1996) found that most forest birds are unlikely to cross more than 100-200 m of open pasture. Graham’s (2001) calculations suggest that toucans are unlikely to move more than 300 m across pasture. Lamb et al. (1997) stated that Australian birds are known to readily cross 500 m distances between fragments. While Ranta et al. (1998) used a distance of 350 m as a distance across which rainforest animals are unlikely to move, this value appears to be an assumption that was not tied to any 13 sreeitic . motes h lllOVClllt" ' tlirahan‘; iota. \“t’ll relatively luster e1 liars; M. LIE: have unusual r ,. *msments Were awai [‘7‘th 0f th puerto r: Ir l‘l.‘;l ‘ lelidgc am] SUI-ll“ L? Ernie . hid] In specific data. The probability of a bird (or bat) actually transporting viable seed as it moves between fragments is also likely to decrease as the distance increases, since longer movements are likely to include st0ps at trees at various points between the fragments (Graham 2001), at which point seeds may be defecated or regurgitated, thus reducing the total volume of seed that the bird may be carrying. Spatial and Temporal Scale Most studies of habitat fragmentation have looked at systems that have a relatively short history of fragmentation (on a scale of a few years to a few decades; Turner et al. 1996, Debinski and Holt 2000) or very long time scales (a few thousand years; Morrison 2002, Pither and Kellman 2002). Few studies have looked at systems that have been fragmented on an intermediate time scale. In that regard, this study is unusual. In addition, few studies of forest fragments in the tropics have looked at small fragments; Pither and Kellman (2002) stated that only two published studies that they were aware of (theirs and Thomas, 2004) have looked at fragments 1 ha or smaller. In both of these regards then, this study is at a scale that is unlike that of other studies. Puerto Rican Dry Forests Tropical dry forests are one of the major tropical biomes. As defined by Holdridge (1967), dry forests may have once covered 42% of the land area in the tropics and subtropics (Brown and Lugo 1982). They are also among the most heavily impacted tropical forests (Lerdau et al. 1991). Dry forests have been settled longer than wetter forests, and deforestation has “preceded and exceeded that of evergreen forests” l4 (Steinin: *6 “err. Caribbt; Hurrah} and Lug. late or” al. 1095 1990 (Fr an are 0: Santa Cr' affected E m Boll“ D Retirer s: m‘ on an foreSis: 1.. l ”.40 IT, 19863.1. l liar of “r. forms l 7; {Wests ha (Steininger et al. 2001a). Human population densities are higher in drier regions than in the wetter parts of the tropics (Tosi and Voertman 1964). The dry forests of the wider Caribbean and Central America have been largely eliminated (e. g. , J anzen 1986b, Murphy and Lugo 1986a, Janzen 1988b, Kimber 1988, Dinerstein et al. 1995, Murphy and Lugo 1995, Gonzalez and Zak 1996, Gillespie 1999, Gillespie et al. 2000), as have those of the Caribbean coast of Venezuela and Colombia (Ceballos 1995, Dinerstein er al. 1995). Only 27% of the original dry forest remained in an intact state in Mexico in 1990 (Trejo and Dirzo 2000) and deforestation rates remained high. In South America, an arc of deciduous and semi-deciduous forests stretches from eastern Para, Brazil to Santa Cruz, Bolivia (Steininger et al. 2001a). Most of this dry forest has been heavily affected by development, and the last remaining large block of dry forest, the Chiquitania in Bolivia, is at present experiencing the highest rate of deforestation in the world (Steininger et al. 2001a, Steininger et al. 2001b). Dry forests tend to be shorter in stature, have more open canopies and have greater stem densities (Murphy and Lugo 1986a) than do wetter forests. Canopy heights are, on average, 50% that of wet forests (10-40 m for dry forests vs. 20-84 m for wet forests; Murphy and Lugo 1986a) and basal areas are about 30-75% that of wet forests (17-40 m2 ha.l for dry forests vs. 20-75 m2 ha.l for wet forests; Murphy and Lugo 1986a). Productivity varies with soil moisture; net primary productivity is about 50-75% that of wet forests (Murphy and Lugo 1986a). Aboveground biomass is lower in dry forests (78000-32000 kg ha'l) than in wetter forests (26900-118600 kg ha'l), but dry forests have a greater proportion of their total biomass belowground (Murphy and Lugo 1986b). Dry forest soils tend to be richer in nutrients than wetter forest soils (Lugo 15 and NIH for agrit degree e classifie- serub (. Bioelim; sithin a definitio more or». 10 the et: hnhholc Slems pa FalTlSVi‘tfj fOrests h. 1Elic‘l'tt ll‘ SOUIllem I'dlSClLSSt; and Murphy 1986a, Steininger et al. 2001a, Laurance et al. 2002a) and are thus preferred for agriculture. These forests are often classified physiognomically on the basis of structure and degree of deciduousness (Schimper 1903, Beard 1944, 1955). Beard (1944, 1955) classified dry forests as semi-deciduous forest, deciduous forest, thorn forest and thorn scrub. Other definitions of dry forest have been made on climatic or bioclimatic bases. Bioclimatic definitions (e.g., Holdridge 1967) include patches of more mesic forest within a more xeric overall community, such as gallery forests, within the overall definition of dry forests. Caribbean dry forests tend to be shorter in stature and have more open canopies than do Central American dry forest -— a fact that has been attributed to the effects of hurricanes by some (e. g., Van Bloem et al., in press). Canyon and sinkhole forests in Guanica Forest, Puerto Rico are taller, less deciduous and have fewer stems per tree than do forests in more exposed areas (Lugo et al. 1978, Castilleja 1991, Farnsworth 1993). While the obvious explanation for this lies with the fact that canyon forests have more access to moisture, it is impossible to rule out a reduced hurricane impact in these more sheltered areas. Sarmiento (1972) discussed the convergence between dry forests of northern and southern South America. Despite the fact that taxa are shared between the two regions (discussed by Pennington er a1. 2000), dry forests of northern and southern South America are dominated by different genera. While Caribbean dry forests are characterized by genera typical of the northern Neotropics, Jamaica and Puerto Rican dry forests are unusual in that the dominant families (in terms of numbers of species) are the Myrtaceae and the Rubiaceae, while continental dry forests are dominated by the 16 legumit' 300:). \ relatit'e : Most drj. al. 1995 Dominic. largest re and 199: Antilles l 1994. M L forests in mm Sm d Leguminosae and the Bignoniaceae (Gentry 1995, Gillespie et al. 2000, Trejo and Dirzo 2002). While this in part reflects a radiation of the genus Eugenia in the Caribbean, the relative unimportance of legumes in intact Puerto Rican dry forest is striking. Dry forests have been described as an “endangered ecosystem” (J anzen 1988b). Most dry forests have been highly impacted by agriculture and urbanization (Murphy et a1. 1995). Caribbean dry forests remain under threat. Losses continue to be high in the Dominican Republic (Roth 1999, 2001). Deforestation in the Hellshire Hills, Jamaica’s largest remaining tract of dry forest, was almost twice the national average between 1987 and 1992 (Tole 2002). Little intact dry forest remains in Puerto Rico or the Lesser Antilles (Kimber 1988, Ray 1993, Francis et a1. 1994, (Gonzalez 1994, Ray and Brown 1994, Murphy et a1. 1995, Gonzalez and Zak 1996, Government of Grenada 2000). Dry forests in Grenada are threatened by housing development, fuel wood harvest, and tourism development (Government of Grenada 2000), in Tobago by tourism development (Boodram 2001) and in Trinidad by petroleum production, plantation forestry and agricultural encroachment (Ramjohn et al. 2002a, Ramjohn et al. 2002b, Ramjohn et al. 2003) Puerto Rican dry forest remains among the best studied in the Caribbean. Guénica Forest, which was described as “an excellent example of subtropical dry forest” (Ewel and Whitmore 1973), is one of the major sites of dry forest research in the northern Neotropics. Five forest associations have been documented within Guanica Forest: mangrove forest, dwarf forest, dry scrub forest, deciduous forest and semi-evergreen forest (Lugo er a1. 1978). The largest of these, the deciduous forest, is by far the best studied (Lugo et a1. 1978, Dunevitz 1985, Murphy and Lugo 1986b, Castilleja 1991, 17 ()uigle} 20009.b the dry s Fertile-2m. (”I'Ltllli lugo 19' C'ifliiitlfi. Ut‘lt'lt’ulti dominate Redlh.ir IUIESLg a: Illere are . lCorlen ] hResin (I mmhae; ab‘“d0noi h. ”F COm Quigley 1994, Murphy and Lugo 1995, Dunphy 1997, Molina Colon 1998, Dunphy et al. 2000), but the community composition and metabolism have also been documented for the dry scrub forest and the semi-evergreen forest (Lugo et al. 1978, Castilleja 1991). The deciduous forest association is dominated by Gymnanthes lucida Sw., Exostema caribaeum (J acq.) R. & S., Pisom'a albida (Heimerl) Britton ex Standl., Coccoloba microstachya Willd. and Amyris elemifera L. (Lugo et al. 1978, Murphy and Lugo 1986b, Castilleja 1991). The semi-evergreen forest association is dominated by Gymnanthes lucida, Bucida buceras L., Bursera simaruba (L.) Sarg. and Pictetia acuIeata (V ahl) Urban (Lugo et a1. 1978, Castilleja 1991). The scrub forest association is dominated by Bucida buceras, Bursera simaruba, Pictetia aculeata, Thouim'a striata Radlk. in Engler & Prantl and Pilosocereus royenii (L.) 3yles & Rowley (Lugo et al. 1978, Castilleja 1991). Succession in (Laribbean Dry Forest Although the term succession is applied to the recovery process on both cut-over forests and those that have been converted to non-forest land uses and then abandoned, there are substantial differences in the pattern of succession between these two ‘types’ (Corlett 1994, Molina Colon 1998, Boucher et al. 2000, Mesquita et al. 2001 Burgos and Mass in review). Corlett (1994) distinguished two groups of successional tropical forests — those that have regrown on land that was converted to a non-forest land-use prior to being abandoned, and those that have regrown on land that was disturbed but which was never fully converted to an alternative land-use. He suggests that the term ‘secondary forest’ 18 should t resemh I, support :0” VCLL' offuu '. ,. Tandem I ilgh‘lx 11. forest ire Species ‘ should only be applied to the former example since the latter type of forest usually still resembles primary forest in species composition, while the former type of forest tends to support a distinct community. Forests that have simply been cut recover relatively quickly — on the order of 50- 200 years — while intensively used agricultural land was estimated to require on the order of 500 years or more to recover (Guariguata and Ostertag 2001). One of the major differences between cut and converted forests involves what Vanderrneer et al. (1996) called the “direct regeneration pathway”. In cut-over forest or lightly used agricultural land, rootstocks remain intact and are capable of sprouting. Dry forest trees are especially prone to be resprouters (Kruger et al. 1997, Bond and Midgley 2001), and are able to rapidly regenerate a discontinuous canopy dominated by the species that were present prior to being cut (Ewel 1971 , Ewel 1980, Dunevitz 1985, Murphy and Lugo 1986b, Murphy et al. 1995). Forest succession on lands that have been converted to non-forest tends to proceed differently from natural forest openings or lands that have been cut but not converted. Land that has been converted to non-forest uses tends to be depleted in organic matter and have suffered alterations of soil organic properties and have altered rates of organic matter decomposition and biomass accumulation (Aide er al. 1996). The species that dominate this type of secondary succession are often different from those that dominate natural gaps and cutover sites (Greig-Smith 1952 a, b, Uhl et a1. 1988, Parrotta et al. 1997, Mesquita 2000, Mesquita et al. 2001). Two key elements can account for these differences — the fact that forest conversion (but not cutting alone) eliminates rootstocks and seedling banks (Ewel 1980, Murphy and Lugo 1986b, Corlett 1994, 19 4 LT- ‘ L) H :1" ad over for that had terns of commur. Islands. J abandon ‘While a 5 51‘1" Fill t’ll a"ell-’lcultti and Man, @ficuh u SFCCICST (Burgos . abandOFit a“Oldir C MD“ Ill ,. Molina Colon 1998, Boucher et al. 2000), and the fact that some pioneer species appear to inhibit the regeneration of primary forest species (Sim et al. 1992, Mesquita er al. 2001). In addition, the duration and intensity of use influences the pattern of succession (Hughes et al. 1999). In Guanica Forest, Molina Colon (1998) found a distinct difference between areas that had been used for agriculture, housing and a baseball diamond, and forest land cut- over for charcoal production; in the 53 years since the villagers were relocated, the sites that had been used for charcoal production were indistinguishable from uncut forest in terms of species richness and basal area, while the other sites supported a species-poor community dominated by Leucaena leucocephala (Lam) de Wit. In St. John, US. Virgin Islands, Ray and colleagues (Ray 1993, Ray and Brown 1995) found that a 33-year-old abandoned pasture supported a species-poor community dominated by L. leucocephala, while a 50-year-old site supported a far richer community dominated by Bourreria succulenta Jacq. Similar successional patterns have been observed on abandoned agricultural land in the Dominican Republic (Roth 1999) and in J alisco, Mexico (Burgos and Maass in review). Roth (1999) found that as much as 29 years after the cessation of agricultural activity, dry forests in Jacqui Picado in the Dominican Republic supported species-poor forests dominated by one native and two exotic leguminous trees, while (Burgos and Maass in review) found a similar situation in Mexico where 25-year-old abandoned agricultural land was dominated by one of two exotic legumes (which differed according to topographic location). Castilleja (1991) found that, while there was enough light for seed germination below the forest canopy in Guanica Forest, and that seedling germination correlated 20 "fl . -A_ _ . ' ' 711m £3.11 WEI " Fa" that they assume z Species r season is leueocgfn Objectii inversely with canopy cover, seedling survival through the dry season was dependent on dry season canopy cover. Similarly, Ray (1993) found that out-planted seedlings survived better below shade cloth in St. John, US. Virgin Islands. In addition, McLaren and McDonald (in press) found that dry forest seedlings grew best under light shade but that they survived best under heavy shade. As a consequence, it seems reasonable to assume that one of the factors related to the failure of L. leucocephala forests to accrete species richness is a consequence of the fact that seedling survivorship through the dry season is low under its fairly open deciduous canopy, although the fact that L. leucocephala is dry-fruited (and thus, unattractive to frugivorous birds and bats) may substantially reduce seed inputs. Objectives The objectives of the study were: 1) To quantify the current extent of forest cover in the dry forest zone in southwestern Puerto Rico; 2) To examine the drivers of land-use change in the period 1936-1993 as they pertain to the maintenance of forest cover in the dry forest zone; 3) To quantify the Spatial patterns of forest cover from the perspective of connectivity across the landscape and identify gaps in the overall network of dry forest fragments; 4) To describe the historical dynamics of a subset of dry forest fragments that were the focus of a detailed study of their plant community structure and their role in the conservation of the dry forest biota (see Chapters 4-7). 21 bl 8) 3 rt- l3)]‘ 5) 6) 7) 8) 9) To determine whether the fragments support a species-rich native forest community or whether they are depauperate stands dominated by weedy exotic species; To determine whether the basal areas and stem densities of the fragments are comparable with those found in Guanica Forest; To determine whether distinct assemblages can be delineated on the basis of their plant species composition; To evaluate the degree of nestedness present in the assemblage of dry forest fragments. To explore the relationship between plant species richness and area in studied dry forest fragments; 10) To compare the effectiveness of a power function (Arrhenius equation) and a sigmoid function (the Hills]0pc equation) as suggested by Lomolino (2000) in explaining the relationship between species richness and area; 11) To investigate the relationship between the per plot species richness (“species density” sensu Whittaker et al., 2001) and total species richness in studied dry forest fragments. 12) To determine whether there is a correlation between abundance of species in the reference community (Guanica Forest) and their geographic range; 13) To determine whether there is a relationship between local abundance in Guanica Forest and frequency in sample plots within Guanica Forest; 22 l6) l7)? l8ll 19) l 20.)] be 14) To determine whether there is a relationship between local abundance of plant species within Guanica Forest and the number of fragments within which a species occurs; 15) To determine whether there are differences in the distribution of species that are locally abundant in Guanica Forest and species present in most of the fragments in terms of the factors which determine of their presence in fragments of differing species richness and history; 16) To determine whether there are differences in seed size among fragments with different disturbance histories; 17) To determine whether there is a difference in the abundance of exotic species among fragments with different disturbance histories. 18) To develop methods to determine the conservation value of Puerto Rican dry forest fragments; 19) To evaluate the conservation potential of the studied dry forest fragments; 20) To designate and evaluate indicators of high quality dry forest fragments that can be used to prioritize conservation decisions. Outline of the Study The overall goal of this project was to expand knowledge of Puerto Rican dry forest from Guanica Forest, the largest intact patch of dry forest, to the array of fragments across the dry forest zone in southwestern Puerto Rico. To properly understand this dry forest system, it was useful to consider it at a variety of scales from species to the landscape level. The problem of managing this landscape to conserve the dry forest 23 system: predict. data Us. rules lil- data net Batista: “TIC Sll. ilperiet: Unf‘ragr: referent compare filming heme: lhe Slru. system in Puerto Rico requires a multi-scale approach. Studies that attempt to make predictions about the biology of a system only on the basis of remotely sensed land-cover data usually attribute an unrealistic level of generality and predictive power to simple rules like species-area curves. While useful, such studies do not provide the multi-scale data needed for realistic conservation and management of fragmented tropical landscapes. In order to provide the multi-scale data needed a group of 39 forest fragments were studied. The fragments ranged in size from 6 x 10'3 ha to 1372 ha and had experienced a wide range of disturbance types. In the absence of undisturbed, unfragmented examples of dry forest in Puerto Rico, Guanica Forest was chosen as the reference community against which the plant communities in the fragments were compared. Land-cover dynamics in the dry forest zone of southwestern Puerto Rico were examined over the period 1936-1993. Forest and community structure of dry forest fragments were documented and the patterns of species distribution across the landscape were used to designate indicators of high-quality fragments. The structure of the dissertation is as follows: Chagter 1: Introduction ( this chgpter) This chapter supplies background information on the impacts of habitat fragmentation on dry forests, and introduces the overall project. 24 Chflglt’f (“Triplet St‘Uillll . —— p) Chapter 2: Methods This chapter introduces the study region and outlines the methods of data collection used in the study. Chager 3: Landscape Change and the Landscape Ecology of the Dry Forest Zone of Southwestern Puerto Rico This chapter addresses the changes in land-use since 1936 as they pertain to this study, and supplies detailed histories of the 39 forest fragments that were the focus of the remainder of the study. Chapter 4: Community Structure of Puerto Rican Dry Forests This chapter describes the forest structure in a series of dry forest fragments, and groups them on the basis of plant species composition. C_h§pter 5: Species-Area Relationships of Puerto Rican Plant Species in a Fragmented Landscape This chapter examines the species-area relationships within and among the suite of studied forest fragments. gamer 6: Plant Species Responses to Long-term Fragmentation in Puerto Rican Dry Forest Landscgp_e_ This chapter looks at the distribution of plant species among the forest fragments and examines the factors that may be the drivers of these patterns. 25 I ~ 1 i- p (RAIL —_._ firmer 7: The Conservation Potential of Forest Fragments on a Dry Tropical Landscape This chapter evaluates the conservation potential of the forest fragments and designates species that can serve as indicators of fragments with high conservation potential. Chapter 8: Conclusions and Recommendations This chapter connects the whole work and seeks to make recommendations as to how this landscape may be managed for the conservation of native plant diversity. 26 C HA Dudl Hdhi coastal Alma memu mmen lwmul Nul‘eml mdghj 2000 ”K Were COI CHAPTER 2: METHODS Study Region This study was carried out in the western half of the dry forest life zone (sensu Holdridge 1967) of southwest Puerto Rico as delineated by Ewel and Whitmore (1973), a coastal strip between approximately 18°N 66° 35’W and 18°N 67° 12’W (Figure 2.1). All studied fragments were located west of the city of Ponce within a few kilometers of the southern coast of the island. The study area was located on the lee side of the island, in the rain shadow of the Cordillera Central and was classified as subtropical dry forest (sensu Holdridge 1967) by Ewel and Whitmore (1973). The climate is seasonal with most rainfall occurring between August and November (Figure 2.2). Precipitation varies between 600 and 1000 mm annually (Ewel and Whitmore 1973). Climate Diagrams (Walter and Lieth 1967) based on the 1971 - 2000 monthly climate normals (National Oceanic and Atmospheric Administration 2001) were constructed for nine sites in and around the study area (Figures 2.1, 2.2). On average, Puerto Rico experiences one hurricane every eight years (Quifiones 1992) but return rates on the south coast are about one every twenty-five years (Van Bloem et al. in press). While 37 hurricanes have hit Puerto Rico between 1700 and 1999, the eyes of only 12 of these came near the dry forest zone (Van Bloem et al. in press). The dry forest life zone of southwestern Puerto Rico consists of alluvial valleys scattered among low hills. South out of the Cordillera Central there is a sudden onset of dryness. The green hills are replaced by faded yellow grasslands. The forests are short in stature and the trees are multi-stemmed. Many of the trees are dry-season deciduous (Murphy and Lugo 1986); they drop their leaves in the dry 27 RBOU pill. 11 deeidu Ema their sh impene- | mm.l PTUSUP Site Sc OhmlnC( Dakota ZQF160;" mags mild-Iii igdy‘ season as a means of water conservation. Dry-season deciduous trees are, for the most part, facultatively deciduous — the drier the year, the more pronounced the degree of deciduousness (see Medina et al. 1990, Eamus 1999, for discussion of the relationship between evergreenness and deciduousness in tropical dry forests). As a consequence of their short stature and multi-stemmed grth form, forests tend to form dense, almost impenetrable thickets consisting of a mass of thorny trunks, spiny lianas and patches of cacti. Large areas of successional vegetation dominated by Leucaena leucocephala and Prosopis pallida (H. & 3. ex Willd.) HBK are also characteristic of the landscape. Site Selection Aerial photographs (1 :33 000 color-infrared photographs December 1993), obtained from the United States Geological Survey, Eros Data Center, Sioux Falls, South Dakota, were used to locate, classify and map forest fragments across the dry forest life zone of southern Puerto Rico. All potentially suitable fragments were identified using these aerial photographs and a subset of those was randomly selected for study. Access to these sites and actual conditions were determined by ground surveys. Sites that appeared to be forested on aerial photographs but which were actually wooded pasture were discarded. The original study design (Murphy and Burton 1993) envisioned dividing the dry forest zone into four 28 i Figure: M 1973) at Diagran fa) Prcclphn “on (m m) 1 FIBER I F‘s’Ure 2 . P‘Ji‘rto R >100 mn e. Coamo a. Ensenada I . P N b. Isla Magueyes c once I 01020304050km I_I.._l_.l_.l._I Figure 2.1. Map of Puerto Rico showing the dry forest zone (after Ewel and Whitmore 1973) and the approximate locations of weather stations used to construct Climate Diagrams. Letters refer to the order of the Climate Diagrams in Figure 2.2. g 200 - +Precipitation 100 G E 150 - + Temperature -_ 80 2: :8 -- 60 a 5 100 - ‘3 'E. -e 40 ’2‘ g 50 - . -L 20 E a fl 0 i J. i i i i i i i i 1‘ 0 J F M A M J J A S O N D Month a. Ensenada Figure 2.2: Climate Diagrams (Walter and Lieth 1967) for the dry forest life zone in Puerto Rico based on NOAA 1971-2000 climate normals. Heavy shading areas (rainfall >100 mm mo'l) represent water excess, while light shading represent water deficits. 29 22:: :33-32:99...— b. lsla} a... .l l A... 2: 2:36:93???— \ «2. .u.» ==.: .Cma—TdQth ‘éUlr. M Figure 2 AOL 0.52209th 0 0 0 0 oo 6 4 2 O . _ . _ a - u - ”_....- tot p. . . O 0 5 0 150 Jr 100 J- :55 5333395 JFMAMJJASOND Month b. Isla Magueyes AOL 9.520358 r0e0000 186420 _ w n . 0 0 O O 0 0 5 O 5 211 3:5 nowaumamooum JFMAMJJASOND Month c.Ponce AU.» unauauoafioh JFMAMJJASOND Month :55 5533695 Figure 2.2: (continued). (I. Aguirre 30 e. Coa‘ Figure App-pfiv :33-31:00...- E 200 100 Erso- -80 ~60 G _ g 100 .40 Ié 50' -20 §0ttititt4ttto “ JFMAMJJASOND Month e.Coamo Figure 2.2: (continued). 31 Temperature (°C) U "9.". get . lmflllt bill the} Studied a: _ , “'3ng geographic quadrants (northeast, southeast, southwest and northwest) with study sites evenly distributed among the four quadrants. The distribution of forest fragments (as determined from the aerial photographs) made the design impractical. Most forest fragments were concentrated in the southwestern quadrant. The flat land in the Lajas Valley (northwestern quadrant) and the areas east of the town of Ponce had little forest cover. In addition, the area between Ponce and Guayanilla provided few suitable sites as the area forms a large continuous patch of secondary dry forest without definable fragments. A total of 39 fragments, ranging in size from 0006-1392 ha were fully inventoried (Figure 2.3). Data were also collected from two other sites (Sites 3 and 41), but they were not fully inventoried. In one case the site was accessible when initially studied in 1996 but not in 1997 (Site 3, Figure 2.3). In the other case, the sampled “fragment” (Site 41) was later shown to be part of a much larger block of forest. Existing forest fiagrnents formed a continuum of sizes, slopes and aspects; many sites included more than one well-defined ‘aspect’. In addition, the sites had variable disturbance histories — several were mosaics of patches with difi‘erent histories —- and the feasibility of many standardized comparisons among classes was limited. When selecting fragments, no attempt was made to control for the magnitude or recentness of past disturbance except that active pasture sites supporting monodominant stands of Prosopis pallida (an exotic leguminous tree) were excluded. Reference plots within Guanica Forest were selected through discussion with the Park Manager, Miguel Canals Mora. Semi-evergreen and Deciduous plots were located 32 688—8 83 ASE 3:83 05 533 80% «one 05 28 A93 .8953? 23 3.5 H33 0:3 388 be 2: wag/cam SE 8.5an mo man: wows .25 com J ofiom Begging .N SEED com 92: 33 Ba mo nowonbmaoo 06 mo 2.5% 8m .ooE otozm 80603538 mo 28a om: «Show be ofi E 358mg ~88.“ 3:53 we gouge. 2:. "m." 0.5“:— _ EN £1 _l...m .G 85l— Nm .cm 33m .. \c. r. .v fem _lI_ Samamgm 8% . _ _ mm .2 .: saw an .3 K 82m fig 33 r“ Positi Fropi studie Data to char in: in 3 Within record; atahei Identil} which 1 number he abla.’ We- lafiabtt in areas that supported closed forest cover in 1936 (based on the maps of Velez Rodriguez 1995a). The location of each studied forest fragment was recorded using a Geographic Positioning System by staff of the USDA Forest Service International Institute of Tropical Forestry at Rio Piedras, Puerto Rico to facilitate the use of these sites in future studies. Data Collection Three separate plot-based methods (Table 2.1) and one plotless method were used to characterize the plant community of the sites. A variable number of circular plots (25 m2 in area) were used to collect data on the identity and abundance of plant species within each fragment. Initially the height, species, and height of first branching were recorded for each individual over 0.5 m tall; the diameter at breast height (dbh; recorded at a height of 1.5 m) was recorded for each stem with a dbh of 1 cm or greater. The identity of any vines climbing these individuals and any epiphytes and the heights at which they were present were also recorded. Species less than 0.5 m tall were identified and abundances were recorded, by species. To streamline data collection, these methods were later modified. A variable number of 25 m2 circular plots were used in each site to record the species present, and the abundance of each species in each plot. Canopy height and the height of the tallest individual in each plot were also recorded. Belt transects (generally 2 m wide and of variable length) were used to collect data on the forest structure. The location and 34 .CU.—V~ C~.—.u:- -LU~ZU>¢C~33Z .uZUL. ML 35.2.3va 2.. v.~::~ oUUZQLUKUL 3:: twp—Ur:13.: .135:— >t._—v >.—Z..1. Cu 197.: 7.21.7.0: 3:2»:532 0.9:: U.:..—: £030 L...» qvbubaiteb V..Cb.2.vh:r.:..v:~ .1: . Initiu: 1C.:L::~.l. uh .N .v-~=.~- 8000.8 0.060% 3000 Sm @0880.— 003 SE 05 8 8000a £82368 mo 00988 38 05... B08800 m0? B08209 z 5033 9 360% 05 88 .c0mm802 0:38.. 0 80>mm 003 136368 comma 36368 2000 :0 80.3880 .00 £80: use «.3802 0060mm .3 40203208 £000 838:0 88> no 30803 36368 .3 856 80 _ N 80% £000 £80805 0008 “won mo 88 05 85m? :8 8 m6 a0>o E802 B80 4.50 80 fl N 88> .«o $5802 @088 3.6368 “00:5 05 .«o £30m £82308 :0 $830805 08m mo E803 #030368 860% .3 3 .30 so a N sea 68 02088 202 2050 .30 e no 85 $3385 _a E s33. new 80 w m saw a as. 8365 58 E .582 seam 3 .masesee =#0 .8 .35 :9 a 3 5% 3.2305 58 .8 $552 0880; :5 SEE: - SE 33000 NE mm 3580 - SE 3386 NE mm .002 2005 800003558 .0008» 83:5 8 308 008000.000 8a $80883 608.“ be .880 3 B08 8800c 9:388 0005 06 mo £000 com “0080—30 3808080008 05 I 8200c mam—98$ "flu 030,—. 35 speci Slfm data \ the 5; menu were r' comp. Specie search; made 1 four 5n implir .Sile “-2 encoun- Ypflifig QT? 33m 5115qu species of each tree with at least one stem 2 1 cm dbh were recorded. The dbh of each stem 2 1 cm was recorded for each of these trees. See Table 2.2 for a summary of the data collected at each studied fragment and in the reference plots in Guanica Forest. The plot-based methods did not record every species in a fragment. To expand the species lists, ‘site walks’ were undertaken. Each fragment was searched as thoroughly as possible (given the limitations of time) and all new species encountered were recorded. A system of diminishing returns was used as a means of estimating the completeness of the search. If one hour of search time failed to record any additional species, a fragment or portion of a fragment was considered to have been adequately searched. Small fragments were completely searched. In larger fragments attempts were made to search each major feature of the site (e. g. , each major slope and each valley). Different methods were used in several fragments (Table 2.2). In Sites 37-40 (the four smallest fragments) the entire fragment was treated as a single plot and the height, species, and height of first branching were recorded for each individual over 0.5 m tall and the dbh was recorded for each stem with a dbh of 1 cm or greater. Species less than 0.5 m tall were identified and abundances were recorded, by species. No plot—based sampling was done in Sites 31 and 32; species lists for the sites were compiled using a ‘site walk’. No overall species list was compiled for Site 3 because access problems were encountered in 1997, or for Site 41 because it turned out not to be a fragment. Overall species lists were not compiled for the reference sites within Guanica Forest. Table 2.3 presents an overall summary of the data collected for each fragment. Unknown plants were given field codes. Voucher specimens were collected for subsequent identification. Exceptions were made in the case of species thought to be rare 36 Ari: 32.021 FCU~IU>2FSSCI Qty—C... 30:32.... v C_ m.»..& UUCOLMTJL 7:: metastuwrmt 3.4.2.; .32 L3,. ~0r¥fi€7§2b 2.4.2.4? 12.:Letzv. .22-:0 >..:::::..... T. ..N.N .,u-~=.h 2 8m 8 E 2 -- § mm 2: OR -- o v _ vm 2: 8m € 2: w -- _ mm m: o2 -- m h 2 mm o: o? 5 o: S -- S a 08 RN E com a -- 2 om 8N m? 6 8m 2 -- 2 2 m: m: -- -- n 2 M: 3 SN -- m o S 2 8m 8m 6 8m 2 -- mm 8 RN 8m 5 2: h m 3. 2 SN 08 -- -- m o z m: m: -- -- B N 2 *2 E 5 «mm 2 -- m 2 com ma -- n w 3 : o2 mam -- m o 3 2 mu m: -- m m 3 a 8m E 6 8m : -- n w 8N m2 5 on n x mm 5 8m 8m 6 8m om -- we 0 8m o2 5 o: w o z: m m: 8m 5 E om -- we v EN EN -- -- 2 Q m on 3m 6 on S -- o: N 2; 8v -- -- 2 Q: ~ «8 Aguasbmv NE Amomooqmv A3035: mo .05 Arno $625 906on + 232—av NOE WHOM. ~02 30% NE .muoomddb SOD mo woe 339 N8 mN MO .oz 303 we mN mo .OZ «.02 3mm down 8.53% 5833558 JmEom «3530 E 803 8:0me 98 ficofimfim “was.“ >6 8% woke—n80 :wmmou mam—95m 05 .«o bagm < "fin 935. 37 .AuuUZCCCSUV N-N .U\~\-rh. .So 3E3 $3 32953 03939856 Ho: Sn a8% 08% a 3:380 225 mm: 83on H .05 8:5 05 8m 938:8 825 8% 860% 98 35885 93 $03 oflwamm mm 3358 0.53 3% 05:0 82F .r 6me 33 was no 8288 o: ucofiwmh non—Movie? a 8: 33 3F .. “meow o; E g o; 2 -- 88 855 of of -- -- o ... s 8 8 -- -- Q. wood S 2: 2: -- -- JP :3 a 2: 2: -- -- + :3 mm com com -- -- + 86 R 8 o: S R w -- 3o 3. 2: m2 € 9: m -- So 2 ca o2 6 own 2 -- :3 E 2: 2: -- -- v 23 mm -- -- -- H -- 3 mm -- -- -- H -- :6 R VS com S EN w -- do 2 2 fl -- -- m 3 am o8 own Q on 2 -- no mm of ma -- m o to 8 ”8 com € ”8 w -- _ 8 NE Augusbmv NE Amomoommv @8853 mo .05 3:5 momoommv 928% + 0.58ng mag 30H 8?. 38. “8 £885 :3 mo no.3 32a NE mm .«o .02 $03 Nw... mm mo .02 83 8mm .Goscwcoov Nd 03:5 38 [\II'M" I Tab] in G: Table 2.3: A summary of the data collected for dry forest fragments and reference plots in Guénica Forest, southwestern Puerto Rico (see text for details). Site Structure Abundance Overall Species List + + + + 5? +++++++++++++++++++++++++++++ +++++++++++++++++++++++++++++ DJ ~ I l +++++++++++++++++++++++++++++++++++++. 1.» co ++++++++++- ++++++++++. Reference Plots 39 or endangered; in such cases field notes were used for later identification. Newly encountered species represented by single individuals were not collected except in the case of larger woody plants where it appeared that collections would not adversely affect . the survival of the plant. Voucher specimens were prepared using a plant press and were dried using a field dryer constructed using a 100-watt light bulb. Voucher specimens were deposited at the Beal-Darlington Herbarium at Michigan State University. Species nomenclature follows (Liogier 1985, 1988, 1994, 1995, 1997) except where otherwise stated. 40 C H.- [(1 PH [Hm chart. 1998 I Roth 1 recolo Consec farm “hEiht' 6’! a1. 2 18.79 1. F350 UH Cchr: 1‘0}be a 31am CHAPTER 3: LANDSCAPE CHANGE AND THE LANDSCAPE ECOLOGY OF THE DRY FOREST ZONE OF SOUTHWESTERN PUERTO RICO Introduction The survival of native species on a fragmented landscape requires the preservation of adequate amounts of habitat. The dynamics of this habitat profoundly affects species conservation. Dry forests are resilient after cutting (e. g. , for firewood or charcoal production; Ewel 1980, Dunevitz 1985, Murphy et al. 1995, Molina Colon 1998) but are slow to recover from prolonged agriculture (Ray 1993, Molina Colon 1998, Roth 1999, Burgos and Mass in press). Most of the native tree species are slow to recolonize areas from which rootstocks have been eliminated (Molina Colon 1998). As a consequence of this, second grth forest can be almost indistinguishable from uncut forest after fifty years, or can have a radically different species composition depending on whether it was used for charcoal production or row crops (Molina Colon 1998, Erickson et al. 2002; see Chapter 4). The dry forests of Puerto Rico have a long history of human impacts. As early as 1879 local Spanish officials expressed concerns about the quality of surviving wood resources in the southern part of the island (Wadsworth 1950). The early twentieth century saw the expansion of sugar cane cultivation and the demise of the coffee industry following the American annexation of the island (Dietz 1986, Santiago 1992). This led to a major shifi in the rural population away from the mountainous interior into the coastal lowlands. Between 1899 and 1910, while the population of Puerto Rico grew by 17.3%, the population of Guanica Municipality grew by 121.4% (Dietz 1986). This increase in 41 per: 10! t‘ .' population density would have increased the local demand for forest products, especially for charcoal, the main fuel used for cooking (Murphy 1916). Sugar production also contributed to the degradation of the forest resource. In addition to competing directly with forest as a land use, sugar refining put heavy demands on remaining forest land; Murphy (1916) reported that the sugar refineries were major consumers of fuel wood. Seven sugar refineries were located in the southwestern portion of the island in the later 1920s including Central San Francisco, the largest sugar refinery in the island (Chardon 1927) Puerto Rico is likely to have been almost entirely forested at the time of European contact in 1493 (Wadsworth 1950). By the late 19403 forest cover had been reduced to about 7% of the island (Birdsey and Weaver 1982, Birdsey and Weaver 1987). A shift in government policy from the 19405 onward changed the economy from one based on agriculture to one based on manufacturing industries. This was coupled with increased emigration to the US. mainland in the postwar period. These changes reduced dependence on the land and absorbed labor surpluses (Dietz 1986, Santiago 1992). This resulted in reforestation, as abandoned land returned to forest. Between 1948 and 1990 reforestation averaged 0.63% of the island per annum (Rudel et al. 2000) and forest cover now stands at about 35% (Birdsey and Weaver 1987, FAO 1998). Puerto Rico has experienced the highest rates of reforestation in the world in the postwar period (Rudel et al. 2000). The countries with the next highest rates of reforestation in this period were Germany and Austria, whose rates of 0.25% per annum (Rudel et al. 2000) were considerably lower than that of Puerto Rico. 42 an ofc wit fire hfls dunn SPCL‘h tome emhq “hkh [Rudd Elfin Most of the forest land that survived the peak era of deforestation was located in the Luquillo Mountains in the northeast of the island or in one of several small tracts in the Cordillera Central (Wadsworth 1950, Brash 1978, Figueroa Colon 1996). In southwestern Puerto Rico, forest cover was concentrated around Guanica Forest, a tract of dry forest that was set aside in 1919 as a 2079-ha reserve and subsequently expanded to its present 4000 ha. Gleason and Cook (1926) also referred to a “relict tract” of dry forest near Tallaboa. Figueroa Colon (1996) considered the Sierra Bermej a, a range of hills to the east of Guanica Forest, to be another important repository of plant diversity during this period. The Sierra Bermeja supports six of the 13 federally listed endangered species potentially present in the areas (see Chapter 7). Two of these species are endemic to the summit of a single hill, the Cerro Mariquita. This makes the area richer in endangered species than Guanica Forest (which supports five endangered species, two of which are endemic to the reserve). Most of the reforestation has occurred in the mountainous interior of the island (Rudel et al. 2000) but Lugo et al. (1996) recorded extensive reforestation in the lowland dry forest zone in the vicinity of Guanica Forest. Rudel et al. (2000) discussed two models to explain recovery. One model, the “forest transition hypothesis” posits the change to be driven by reduced demands on the land as a consequence of industrialization. In more industrialized societies workers are drawn to jobs in the manufacturing and service sectors and away from agriculture. With more cash income and less time available to spend on the land (Preston 1989, Rudel et al. 2002), goods and services previously obtained from forest fragments are now purchased from other sources. Thus, the pressure on the land decreases. Their other model, the “special 43 {.2 reIaz incr; rcdu. mad. 13nd. imp» [1131C Obje 4) relationship hypothesis”, suggests that Puerto Rico’s special relationship with the USA. increased the cost of labor through the provision of Federal Assistance programs and reduced the supply through emigration to the mainland. This increase in the cost of labor ~ made agriculture economically unviable and thus led to the abandonment of agricultural land, which was then allowed to revert to forest. Historical land use and current patterns of forest cover are likely to play an important role in determining the distribution of plant species across the landscape. Understanding these land-use patterns is important in interpreting species patterns. Objectives 1) To quantify the current extent of forest cover in the dry forest zone in southwestern Puerto Rico; 2) To examine the drivers of land-use change in the period 193 6-1993 as they pertain to the maintenance of forest cover in the dry forest zone; 3) To quantify the spatial patterns of forest cover from the perspective of connectivity across the landscape and identify gaps in the overall network of dry forest fragments; 4) To describe the historical dynamics of a subset of dry forest fragments that were the focus of a detailed study of plant community structure and their role in the conservation of the dry forest biota (see Chapters 4 — 7). 44 f7“ ."t’l Lan< G‘s-13'. Dal» zone dry 17. Rico. dll'ide‘ “id? 0 onme Methods Landscape Characterization Aerial photographs (1:33 000 color-infrared photographs December 1993), obtained from the United States Geological Survey, Eros Data Center, Sioux Falls, South Dakota, were used to locate, classify and map forest fragments across the dry forest life zone of southern Puerto Rico. Aerial photograph coverage amounted to 73 000 ha of the dry forest life zone (about 60% of the total dry forest life zone of the island of Puerto Rico; Ewel and Whitmore 1973). Landscape elements dominated by woody vegetation were classified as either ‘Closed Forest’ or ‘Open Forest’, based on the amount of ground cover that was visible between tree crowns, using FAO criteria for the classification of vegetation (FAO 1993). Fragments with more than 50% of the ground visible were classified as Open Forest, while those with less ground visible were classified as Closed Forest. Forest fragments were mapped using acetate overlays. The entire area was divided into 15 blocks to facilitate data collection. Each block was approximately 7 km wide (east to west) and a variable depth inland (north to south) from the coast, depending on the width of the dry forest zone at that point. Fragment areas were estimated using squared paper and were grouped into five size classes (< 5 ha, 5-10 ha, 10-50 ha, 50-100 ha and > 100 ha). The forest cover in the entire study region and in each of the 15 blocks was calculated. 45 La C01'er each F We t: indm Landscape Change Land cover data were collected from a series of six published land cover maps spanning a 53-year time period (Vélez Rodriguez 1995a, b, c, d, e, 0 covering two USGS ‘ Quadrangles surrounding Guanica Forest Biosphere Reserve, Puerto Rico. The published maps were prepared using traditional photo-interpretive methods (Lugo et al. 1996). Time series land cover data were collected by overlaying a 2 cm x 2 cm grid on each map and recording the land cover class present at each point in the grid. Efforts were made to register the grid in the same manner on each map to ensure continuous monitoring of the same set of points through time. A total of 529 data points were collected from each map. Land cover classification followed those used in the maps (Anderson et al. 1976) except that the three wetland classes and the three barren ground cover classes were aggregated into a single wetland and a single barren ground class respectively. Transition probabilities (the probability that a point would ‘transition’ from one cover class to another or that it would remain in the same cover class) were calculated for each point between each pair of maps. Thus, the probability of change from one cover type to another was not calculated for the whole area, but instead was calculated for individual points. Dynamics of Focal Fragments A chronosequence of aerial photographs was used to assess changes in forest cover and land-use in and around a subset of fragments that were the subject of a more detailed study of plant community structure and conservation potential (see Chapters 2, 4 — 7). Study-sites were selected as outlined in Chapter 2. Aerial photographs from 1936 46 ' -3. (‘1 ; 1. Autc and ‘ Carri phm I D313 1 const area 1 basis forest “go“: Fragm 1””931. < h “0305? 9L hit arm: (1:18 000 black and white photographs, obtained from the Oficina de F otogrametria, Autoridad de Carreteras y Transportacion, San Juan, Puerto Rico), 1963 (1 :20 000 black and white photographs, obtained from the Oficina de F otogrametria, Autoridad de Carreteras y Transportacion, San Juan, Puerto Rico), and 1993 (1:33 000 color-infrared photographs December 1993, obtained from the United States Geological Survey, Eros Data Center, Sioux Falls, South Dakota) were used to assess change. Site histories were constructed for each of the fragments with the exception of one that was outside of the area for which aerial photo-coverage was obtained. Fragments were classified on the basis on the proportion of the site that was ‘old growth’ — i. e., areas that had supported forest cover continuously since 1936. Each fragment was classified as Relict (> 75% ‘old growth’), Mixed (25-75% old growth) or Regrowth (< 25% old growth). Attempts were made to ascertain the causes of decreases in fragment size. When patch size decreased, land-use in the area lost from the fragment was used to classify the cause of the decline in patch size. These changes were classified as i) Agricultural or ii) Urban, Industrial or Infrastructural. On-the-ground observations were used to describe changes that had occurred during and after the fragment inventories (see Chapter 2) were carried out. Fragment Networks and Connectivity Unrectified aerial photographs (1: 33 000 color-infrared photographs December 1993), obtained from the United States Geological Survey, Eros Data Center, Sioux Falls, South Dakota were scanned and a photomosaic of the study area was constructed using Photoshop 5.5 (Adobe Systems, 1999). A forest cover map was constructed by tracing the areas of forest cover on this image. To assess potential connectivity between 47 .‘l'll "ti 1. we: 1.191 a ti» p311; ClISL: l R831 Lam OCCU: forest Vane: Elgar ~ "0 l Rico‘~ . fragment clusters, gaps between clusters greater than 500 m and greater than 1000 m were identified and recorded. These distances were selected based on da Silva et al. (1996), Lamb et al. (1997) and Graham (2001) — the smaller distance (500 m) represented - a distance at which seed movement was thought to be unlikely (given the movement patterns of potential seed dispersers), while the larger distance (1000 m) represented a distance at which almost no seed movement was to be expected. Results Landscape Characterization Forest cover stood at 23.2% in the dry forest zone in 1993 (Table 3.1); 18.0% of the forest cover was Closed Forest and 5.2% was Open Forest. Closed Forest was the dominant wooded land-cover category in 1993, covering more than three times the area occupied by Open Forest (Table 3.1). Forest cover peaked toward the middle of the dry forest zone, and decreased toward the eastern and western end. Closed-Forest cover varied across the landscape more than did Open Forest cover. Closed Forest cover ranged from 0.6 to 67.4% of the area of each block, while Open Forest ranged from 0.6 to 9.2% of each block (Figure 3.1). The city of Ponce, the second-largest city in Puerto Rico, was located in Block 10. The town of Guanica was located in Blocks 4 and 5. F ifty-two percent of all fragments (322 fragments) were below 5 ha, and accounted for 4% of the total forest cover (Table 3.1). Three percent of all fragments (20 fragments) were over 100 ha, and these accounted for 63% of all forest cover (Table 3.1). The mean fragment size was 42.6 ha for Closed Forest fragments and 12.3 ha for Open 48 22:22.39... 3:: 22:25CU 3.: C.— .I...u::..> mi: £.:....: >7..J..LLLC :2. >_:: n..:..:: :::~\C. v .Cb‘vx Aer-9:»— :LDZLUBSZJCI .UCCN awful»..— AL—u UP: Cm Jazz—U UNmm >.£ .AISQG..HQCHCSQ ~.~...-U-.~ NCO~ :3 UUVCFFV L.J>CU .20.: {KC :C.::~:.:.J..~A\ u~ .N. .v§-=.~. Na 2: 5: N: 3: 9: m2 2 8m ask 2 SN 85 e 02 Ni was 3 E 2: A 2 new as 2 S 2 so 3 2 83m 3 ”.2 :2 8 mm «:2 £2 2. 2 of: S w: E E so 2 am a. 2 an 2 Q S... e: no 3 as we 2 w v oaoN ocoN 3.24 598m ficoawfim £5 82 “mecca 3.38me awe—U onm e: co as 309 L0 .x. as 32 Lo .02 L0 .x. was .6 .x. as 82 Lo 52 Evans; “moan ammo 520m @820 $3988 9 26 3828 05 E 839 05 :32: 3369a 8: 38 £88 EEO .83 883 8333858 6:8 888 be 05 5 .350 BR 3 .Amnamowouofi 3.5m 32 so comma :58 ~88.“ no 5955me "fin «fish 49 .353 “dogma? Ea _oam .3 388:3 we ASE owwtgom amemnv econ om: “mono.“ .06 05 £55, 3: cam—mm 05 mo 5:8 05 5 no.8 uofimmeE 2C. .SE 8.53 .«o use 2: 8 Eamon 038 BE. .mzflov 128228 How ~on oom .023 :2 ms boumgxoaqa .83 x83 zoom .nmfiw o6 E Booze nos 05 9.56% 92: 05 no «one waxes 2:. .83 otosm 8333558 .38 528 be 05 E «woo 8 “no? 80¢ c260 “mono.“ mo companion? 2F "fin aha—ME 9. 3. 9. NF 2. 0.. m w w 0 m v m N r ....... ...... ,... . _ .,-._ . .. .‘ '. ' .- s "a?" . y‘.‘ , . . . . . . .. . . r. . ' .t’.’ .;.".‘.;. . . . . -- .‘.' : 7 "1.2.3.: I‘ 'a 31.: ' .‘. ‘. r . . t Pi . 8 M 3,.” m m“ ....... _1 CM 1“! ..I L w .d ,Hmonom come—OE n. ........ .. ow m . 880m 590. cm 1 m 5. om 3 on on S o m .. _ 2 Z - 8 mHvHMHNHHHonmhmvaNH 50 III +CCCT£CSF~. 3533...?de Ont: we‘d-YES: U-~ Humvnv>)~.dr~ Tut-CU .JZaul~v~2w~ Lbntczfv h: 32:31. nu»: _..: man—.750 IRS—U L9>CUI—qu~_ -U>mw 5 C.- uC‘OQ 3;: \A~..:£.~£:LP~ Hz: :3 7.47.35 UL... RCAICZCTRfiF .313 ~ ICwN. ~ uvgvmhbr~ 3:» p... :U-VN CHAD: N puquIU>lu._—:»vl Lav Ian-Unpnuwuuhtu ulnvhfv'u \ALFJ cm aluhfiuuh nvmdmmau—Tv .vcz unmrvhpu..~ :rVLA v .UL-a.—.~U.—Lnfl< n.3,)».irunvrn ?.q.v.~u.-.~r~...r~nvhnk :;v.-s.-7.n~=k.§‘ nN-~u .Uhninwha 880m @820 8 880m @820 I m m 880m :25 8 880m 9820 I m m 8830im< 8 880m @820 I _ m 888 R820 8 880m :30 I m m 880m :30 3 880m :85 I N N odeowax 8 880m ammo I g m 88cm “820 8 o§~=otw< I m _ 880m 5&0 8 o§_=oflw< I N ~ 2838m< 8 EBEotw< I ~ _ Jr 3o 82 c8 :3 So 86 So one 8.0 3.23" m; 85 3o mod 3o 86 one So Ed 225 mod Ed Rd 3o 85 3.0 33 mod 86 $82 3.0 as Ed 8o 2.0 who 23 3c 86 8-582 85 one 86 3o :5 Po :3 8.9 2; 58-82 _N mm mm _m mm mm __ - m_ +5385 .mvotom 0:5 v25: 05 5953 820 89-25 .8528 8 088 08 E macaw 830 B>oo-w§_ 53w a E “Eon a mo 32330:” 05 co v83 08 mcoummcwfi. 63733 woven 2: E SE 955 8883558 mo mucofiwma 880m 5 5 880m @820 28 88cm :85 .o§_:om&< 5833 82—3308 :28ch “Na ozah 51 Forest fragn the median Landscap In I“ predominam over I18 [0 ( Open Fores1 Open Fores In II most preval Closfid For Dynamic: Tht 6 x 10.3 ha maintainm p836m of formed 0 fence DOSI basis M b Elem” ) . Forest fragments. The median size of Closed Forest fragments was less than 5 ha while the median Open Forest fragment was approximately 5 ha. Landscape Change In two of the three classes considered (Agriculture and Closed Forest) the predominant transition was retention in the same class (Table 3.2), although it fell from over 0.8 to 0.52 for Agriculture in the 1983-1989 interval. The primary transition for Open Forest was to Closed Forest; retention was more prevalent than the transition from Open Forest to Agriculture. In two of the land-cover classes (Agriculture and Closed Forest) retention was the most prevalent transition; this declined over time for Agriculture, but increased for Closed Forest. Dynamics of Focal Fragments The dry forest fragments that were the focus of detailed study ranged in size from 6 x 10'3 ha to 1372 ha in 1993. Twenty-three of the studied fragments (59%) have maintained some amount of forest cover throughout the 1936-1989 period. Eighty-nine percent of the total area occupied by the fragments was ‘old growth’ (i. e. , had been forested continuously since 1936, although stems are likely to have been harvested for fence posts and charcoal production). Nineteen fragments were classified as Relict on the basis of being 75-100% ‘old growth’, three were classified as Mixed (25-75% ‘old growth’) and 17 fragments were classified as Regrowth (O-25% ‘old growth’). 52 Tu L size and 1w H fmgmelt agricultural or infrastmt consequenc and 1998, t dt‘VClOpme during the l Frélgment A to between frag identified _ lOrest fram Twenty-three of the studied forest fragments increased in size, 14 decreased in size and two remained about the same size between 1936 and 1993 (Table 3.3). Of the 14 fragments that decreased in size, six fragments appear to have shrunk because of agricultural development, six appear to have shrunk as a consequence of urban, industrial or infrastructural development and two fragments appear to have declined in size as a consequence of both causes. Of the four fragments that were eliminated between 1995 and 1998, three were eliminated as a consequence of urban, industrial or infrastructural development and one was destroyed by fire. Five other fragments also decreased in size during the 1995-1997 interval (Table 3.3). Fragment Networks and Connectivity A total of nine clusters and eight major gaps were identified using a 500-m cut-off between fragment clusters (Table 3.4). Using a 1000-m cut-off only one gap was identified - the gap separating the forests on the Cabo Rojo peninsula from all other forest fragments (Figure 3.2). 53 ~02 filav G... SUL< 4&0? ~ IVKXV ~ 3.1qu 3.: a: ......va=~2~0 :5. mundane: _:Cnu_~:¢.J-w 2:)? moo-lefdv— «Javmhnvhh 9;. C.- .anm Vn .vuLU-.A~ CLUu2032u3Avnv. .Zuuhhvnuhuflnwh.‘ ~7..J.=w\ \fi.;v fiunumfizuumkflv >..=:.I..-~ HF:-N.- .UNhh-wyht $88 9 Bug“: mm: 85 2:3 _SBBotwm wocovfinm mm EoEwaa 2: .22 H88? .330: mo couosbmqoo 2: 98 aouosbmcoo 38 3 333% m9» “coast £5 . _ m EH 5 mozm mm .8qu mo x83 06% 05 mo tam 3353.5 3me + mm _.m S .383 05 5:85 2.3:. Mo 56593 . Eu 3 vowaofiwfim :89 mm: x .333? 32898 was 550 388 £23 “020% + 3 oh a .38 a we coaosbmaoo 05 3 8:2ng 803 >2: :5 Km 85 fit» mnoswucoo bfifiwto mm? 26 mat. 8:3“ - o3 N. w .couosbmcoo 38 98 “cacao—26v 53.59% E vogmcoo 203 mean 850 233 Juana 9 US$280 203 £83 £5 go 353 Am 98 3 83m vows—2: 35 628 .«o x85 owufl a mo tam mm? 8% mg 820% - 3 mm m .Eofiwfim $88 :5on 05 BE @3338 3:88 628 8:3va Begum 820m + on vo o .32 .3 528 3 3:32 :5 dmfi E 3528 Ho: 803 mxozg HESS 088 98 8mm 05 .«o :3 8883 EC. .5388 $28050on a mo “cacao—gov 05 E 98 532538 38 E 3838 mm? x 523 88m €52 3588 «0 x83 «Foo? comb: a mo comtom a fl 8% ma gazed - mm 2: m .88 .63on 20. 05 53> 3:388 was 3.88 3 wow—32 BR 05 mo 5:8 28 5.8: 05 2 A86 05 «o .8525 m 509$ 95 Buzzer—mm .«o macs owes 820m + mm mm_ v .wougmxo £3:va Ba .33 E mono—m wcmommfiots 23 5.5: 3 @8352 we? Bo: 8>oo “munch gamed + 3 3 m 63.53% ummcmcgm 28 53:398sz c5323 3 onmEEov bomb: “mono.“ fiBBmom E 86 £5 mo 55% a Sons :5 JcoEwEm £5 .«o 28 2: 258 x83 3383 own: < @832 + E o: N .3058 come—o 0.88 a 2:3 8:8 980%? “mono.“ 05 :5 .32 3:3 88 S. communes. bows; 6:3” -\+ 3 NS: _ meaoaaoo 3% 036 5305 82 3a 52 20 .x. 82 $2.32 85m 2: a 83% so 38: 128323 53> m3 732 cocoa 05 E .ooE 055 8283528 finennwmfi 588 be @3me mo ESE—m and «Bah. 54 ”acorn—CCU EZEV. buccab ....... Drowcuc VOA: 085. ~02 ~40 o\o fibkdvx .AUUZ 2.: CCUV Riva .UNQ ark. .32 8% 88238 Eco b>8 380m 55on32 - o 2 mm .32 EB 8888.» .3328 team .3 8880 8? 32m gewom + o N; mm .38 £5 .«o 8885—0 05 3 850$ mm? 86 £5 .88 Hawk: a 8 H88.“ .«o 3.8% :95 a .3 88288 33535 820% - 02 N2 3 .82 8t“ 88—9628 :88 “8.8.2 8083 2.: ”$2 98 85 8953 8820 83 : £88888 3.8m 022 05 5 8803 33 86 $5 @852 5388M + o m2 cm .82 new $2 80253 288238 :83 $28 882a 05 .mfiauwoaona 3.8m £2 05 E 88 2: E 2888 803 805 8888 26% a swag—«3. 5388M + o m._ 2 .Eofiwmfl $5 2888 .88.. mo €9.30: m .3 wuwauamfi DRESS 83 sea 5:. .3 Ba m 3% 8305 35 :38 meow 2-82 35 8o Ea E3 - 8 2 M: 8:32 -\+ 2: m .0 2 .88m «.35 mo £32 05 mo 868on 2.: 3 8582 mm? 85 3.28 «o 83 832 m .«o :8 bfifiwto 820m - 02 m.m @— .mm Ea : new 8305 35 “seems“ see .532 £0 Ea $35qu 820m - 2: 2 2 £88 35 mo couficofimma 8 SE :88 v.28 numb: a mo 8m 8:32 .. 02 e 3 .82 E 856% Eu 203 :38 05 .8 88858 06 .8888“ 809 was 858 row 8232 + 02 m 2 of ”388.3. .858 flow a 35$, :38 8803 a bfifiwto mm? 8% ma mzafiwouona Eton mo 8m 05 no 88080 05 0238 mm 85 -- 88 o: m 2 Bnmuwouona Eton mo 883888.30 05 Bob H8398 8: mm mu: :5 .2398 3 86 WE“ “.588 8 088 083 39:83. .vofiofiwfim mm? 98 83m 2:3va 8:32 - o2 9m 2 35 03m 832 m E mm 98 2 8am 5:» 8888 bfifiwto 83 van mat. 35888 3% .855 5320 82 25 «oz 20 .x. 32 33.58 2 2.5 55 mHCOE—COLV @2237. 03.23:.V Errantmv M00 ~l- ~32 uih v WM» ~w.d.~ < .uzw, .Afived:-.-n~nvnv.v R|-Fa .Uhkfinwrht .8800 .30 82.32 05 a 883 :38 2F .82 .3 358 23288 a; an .32 5 880. no 8.88 58.8: 0 .5 888 .«o 5008 8022 0 8 88088 00.5 02m 25. «020m - on: cod cm .82 .32 88 82 .32 80585 880—0 305:0 85 0% 05.2. .82 E 8:88 85 0:0 08 c055 8805 00.5 :55 :05 8025 8.5 598on83 2.2.200 82 05 5 50008 508m 05. .82 .5 “08:08.5 :08 85 8820.58 85 :5 .598888 B80 32 05 E 82005 00.5 02m 25. 550832 + o no.0 mm .8388 3:8 82 20 8&3 50.0.05 85 80.58 888 38 o: :5 .82 5 8088 805 805 @8088 .508 < 558w0m + o no.0 Vm .82 8:5 Ba 32 23 5253 Emma 8803 85 : 55888058 20200 82 05 5 80.58 888 8x8— 85 .2 88 2 82m 8 $88 mo 5028 08m 08 no 88 5:898 85 02m 25. 5.588% + 0 mod mm 682 5 8828? 55800 504 5588M + o fio mm 502025 82-82 05 :8: 5 02m 8 88088 00.5 0:0 3:. 8:08 - c2 :6 2m .02 a 88002, 5588 33 52:08 + c 8 8 502005 8782 05 8 .58 888-8: 85 02m 252. 5.58w0m + o Nd 8m .82 28 $2 80.305 8852 8.5 02m 25. .2 v8 m 0:m 88:05 85 0.005 05-2% _ .2020— mmo 8m 52052.5 820% - o2 wd mm .82 v8 32 80.585 88 0.8 88888 05 .5 w 0:m 88% 88:38 8:0M - 2: Yo um .82 80% 8820.58 Eco 80.58 880m 5.58m0m + o _ om 022:5 850885 05 85 :05 88 .0558 8000.28 05 85 88 89898 85 “088 go 508m :88 < 852 + 2 2m 8 .8805 80.» :58 0: 8:83 805 85 0:0 05 85 200—00.58 E0288 0 52.5 80880280 :5 .Emmuwowonq 3.200 06 .3 @8058 8 Z 5.588% + o 2 Va 35868 mag 820 5320 82 ea #02 BO 8 08¢. 8:52:82 mm as“... 56 VACUCZCOrV @3227. bicarfiv CreaCLmv MOO N 3:“ .92 30 o\.. FE< -AhUuU~un~.~u-AVuvv #1.F4 “hag“. .mnm8w80fi _0800 $2 05 8 a: 8030 mm 020 0.5%. 5.5030% + o wood 8. .mna8w8onq .0800 $2 05 8 as 8030 2 8% 05¢. 5.583% + o Ed on 05283852 _0800 $2 05 8 as 8030 2 0:0 05:. 558wo% + o 56 mm 052888052 3.000 82 05 8 a: 8030 2 0:0 05:. 5.583% + o mod 0m $808800 080% 028:0 5.580 32 02m U.02 20 ..\° 08¢. .9350on 3. 2.3 57 al.. !”b_ WJJUIFIW'J 33..“ r q l A ~ I‘lo I IV 1 u ‘ I vN-Nh‘rv V K . i“! hohMVb; N WV ‘5' MI! . . U .. I‘ niA:V.‘ sh ‘V i 0 Ch? ~Wt h.. I 8:? ‘ ~ I. .-,sl~‘\P luv MT FF-C ~UFH I F‘H’V F KARL le-H I In!“ . F‘fl-AI' "I uh , HP 93 HIIIIVQAPA- Fr .OOE Oahu—9m MO OEOUXD 0 80805530 05 :0 888 .80 080 580—08 :88 0 ”0.50% 0900 05 885002 a - .080 809800 E 05 .80 5:8 8:80: 50:05:090 50 00:0 :< w 809800 8 .8 .8 .2 .mm .8 .m _m 08 02880 00 833 05 8258 £2 32 8 .880 500888 n a 00.000 850.5 08002 0.80 800 20 0.580 .880 50888 0 80:33 09 .80 :58 05 .80 5:0: 888 59 5:0 .805 .80 5:8 598050808 5m .3“ .8 Km .3. . _m .m .5 8:80: 50:05:090 .05080m 088m 05 @5205 8580.5 .«0 :8 089:8 < m mm .3 5.58w8 509.585 5588: .59 0:88:00 .80 5:8 888 .80 8.8 5:095 :< v 908:00 0:055:00 05 5:0 880 m 08:0:0 :002509 0: 505 8:0009 - E08095 09 508 505 E9 888 509585 ”030% .80 :.58 05 .80 80.5980 Z m :18 .8 .8 .8 .8 .2 .8 .0388: 08.05000 .8005 0000000 .2 .3 .2 .2 .2 .5: .5 .5 .N .9 80.58 0909 3:058 5:0 880% 005083 00:0 58088: 80909 05 505305 N 8.00:0 58088: E08880 0 2 .00:0% mm .mm .M: .m («0 580 05 m0 :8 m0 85:80.0 :05 888 .80 0900—9 0809 0 .80 E808 80805 095. 9 00820 505208 008m 88888008090 EOEw0£ 08% 800:: 80805538 .8 com :05 808 59 0050 9000 :88 588008 80820 E0308 888 5:5 0:0 0888808090 09%. can 030,—. 58 .\ I < .ooo oi; 0908 80:88:84 880808 .8 8:08.20 :00389 8 00m :05 08:: .8 80m 80855 83080V .005: 080:: 50803580 .80 8&0: 55:8 05 :: 880 888 (:0 888 05 85508 :02 “NM 088: 59 Dis M- to:- {CC w Discussion Habitat destruction is one of the major causes of biodiversity loss (Wilson 2002). While some forest species are able to survive in a matrix of altered habitats, in general, forest species require forest habitat. While it is unlikely that the magnitude of forest recovery observed in Puerto Rico will be repeated in many other parts of the tropics, an understanding of the conditions that facilitated this recovery may help in the identification of other areas where reforestation is feasible. Understanding the drivers of this change can help to identify areas where reforestation is likely to succeed and areas where the prospects of success are poor. Rudel et a1. (2000) identified other limited areas where a similar set of conditions exist, for example in parts of the Greater Antilles (Zweifler et al. 1994) and in the Andean region of South American (Rudel et al. 2002). Forest Cover At 23.2% forest cover (18.0% Closed Forest and 5.2% Open Forest), the dry forest zone is well below the overall estimate of 35% forest cover for Puerto Rico (FAO 1998). While there was relatively little forest cover east of the city of Ponce and west of the town of Guanica, forest cover between these towns was high, and included substantial areas that supported Open or Closed Forest even in the 1936 aerial photographs, including most of Guanica Forest. These areas of high forest cover are the core of the dry forest in Puerto Rico and are probably the key to the long-term survival of the dry forest biota. Most dry forest fragments were small. Fragments smaller than 5 ha accounted for 54% of all Closed Forest fragments and 50% of all Open Forest fragments, but only 2.9% 60 7.4. of the Closed Forest and 9.5% of the Open Forest cover. Fourteen Closed Forest fragments greater than 100 ha accounted for 4.5% of all Closed Forest fragments, but 75.2% of all Closed Forest cover. Ranta et al. (1998) found 1839 forest fragments in a 2674 km2 portion of the Brazilian Atlantic rainforest that ranged in size from 0.06 to 1539 ha and had an average size of 34 ha; this amounted to 0.69 fragments per hectare, a figure that was close to the density of 0.51 fragments per hectare observed in southwestern Puerto Rico. Forest cover in the Brazilian landscape was about the same as in southwestern Puerto Rico — 23% in both cases. Forty-eight percent of Brazilian fragments were smaller than 10 ha (versus 70% in Puerto Rico) while only 7 % were larger than 100 ha (as compared with 3% in southwestern Puerto Rico). While Ranta et al. (1998) considered the distribution of forest cover in Brazil to be a stable end state, the distribution of forest cover in Puerto Rico is probably still changing (as shown by changes observed between 1995 and 1998) — agriculture remains unprofitable (thus, likely to lead to increased forest abandonment) but forest cover is threatened by housing and resort development. Large forest fragments are likely to support more species than small fragments (see Chapter 5). Large fragments are less likely to lose species after isolation than are small fragments (Hanski and Ovaskainen 2002). Laurance et al. (2002b) and Alvarez- Buylla et al. (1996) suggested that reserves in the mainland Neotropics should be on the order of hundreds of square kilometers; optimal reserve sizes for many continental species are likely to be larger than many Caribbean islands. Dry forests are less species rich (Murphy and Lugo 1986a), and thus less likely to have as many rare species (Hubbell 2001). As a consequence of this, the optimal size of conservation units is likely 61 __ mflVr . Iol Ca: Pol ha‘ {Ct 65 to be smaller for dry forests than for wetter ones. In addition, dry forests in the insular Caribbean are likely to carry lower genetic loads and to have smaller minimum viable population sizes (Lande 1995, Alvarez-Buylla et al. 1996) by virtue of the fact that they have evolved as relatively small populations. Fourteen Closed Forest fragments (comprising a total of almost 10000 ha) each exceeded 100 ha in size. Another 14 fragments were in the 50-100 ha range. If these can be protected from deforestation they may form a core area that can be managed for biodiversity conservation. The conservation value of many of the small fragments is likely to be fairly low, especially for those that are primarily Regrowth (see Chapter 7). Nonetheless, fragments may support viable populations of certain species. For any such species these fragments may be valuable elements of the entire genetic diversity of the species (e. g., Aldrich and Hamrick 1998, Aldrich et al. 1998). Many dry forest trees would fit the definition of resprouters (Kruger et al. 1997, Bond and Midgley 2001, Kruger and Midgley 2001), in which case the life span of an individual genet is likely to be very long. Such species are likely to have a long persistence time even in small fragments. The presence of small fragments may also reduce the gap between large areas of forest. This may reduce the movement cost (Graham 2001) for bats, birds and insects between fragments and thus increase the probability of gene flow among fragments. Taken broadly, it seems reasonable that it should be possible to explain forest cover in economic terms (Kahn and McDonald 1997). Areas that remain under forest cover (and thus, by extension, areas that are allowed to revert to forest) are areas in which the marginal cost of converting the land to non-forest cover exceeds the post-conversion benefits derived from that land. Similarly, land is likely to be allowed to revert to forest 62 CO C.‘ naval"- . .1 .- t h. _...- lillllllll. cover if the cost of maintaining it as non-forest exceeds the marginal benefit derived from the current land use. In this context, the findings of Ramos Gonzalez (2001) are to be expected — with the collapse of the agricultural sector in Puerto Rico, former agricultural lands are likely to either become urbanized or to be allowed to revert to forest. This change has resulted in the observed pattern in which urban development has encroached up to the boundary of protected areas (as has also happened in the case of the community of La Luna and the city of Guanica which have grown right up to the border of Guanica Forest). Distance decay theory (Clark 1951, in Wickharn et al. 2000) and retail gravity (Carrothers 1956, in Wickham et al. 2000) predict that distance from economic centers and the size of these centers are the principal determinants of land demand; for example, deforestation was greatest near major population centers in the Mexican state of Morelos (Trejo and Dirzo 2000). This would then predict that forest cover should be lower near urban centers and higher away from them. However, this explanation is likely to be most applicable either in actively developing frontiers or in old agricultural landscapes which have reverted to forest as a consequence of economic transitions and are, in effect, secondary frontiers. The pattern observed in southwestern Puerto Rico was not in keeping with the latter hypothesis. Forest cover was high adjacent to the large towns (Ponce, Yauco, Guanica and Guayanilla) and low further away from these urban centers. It would appear that topography is a key factor in predicting forest cover in this system, as has been found elsewhere (Lamb et al. 1997, Smith 1997, Trejo and Dirzo 2000). Forest cover was highest in areas with rough topography near to urban centers, while it was low on flat ground further from the urban centers. Flat lands with low forest cover are, for the most 63 \l .II-L part, active pasture. Unlike the situation that Wickham et al. (1999, 2000) analyzed, changes in forest cover do not appear to be driven by land demand — or, rather, that land demand cannot be predicted simply as a function of distance from urban centers. Forest cover reflects land that has been abandoned, and much of what is non-forested is land that was agriculturally productive enough to retain under agriculture. In addition, agriculture is likely to persist on lower-value lands. Higher property values (and as a consequence, property taxes) near urban centers can make agriculture a less viable prospect near cities (although this is likely to be countered by greater ease in getting produce to market). Landscape Change Studies of landscape change in the dry tropics are few. Endress and Chinea (2001) looked at land use change in the Republic of Palau in a period of agricultural decline, while (Kramer 1997) looked at changes in what later became the Area de Conservacion Guanacaste (ACG), Costa Rica at a time when forest cover was increasing as a direct consequence of programs associated with the formation of the ACG. In contrast, Turner et al. (2001) examined land-use change in the southern Yucatan over a time period in which development pressures were high and population was increasing rapidly. The level of retention of Closed Forest in Puerto Rico was not high compared to most studied landscapes because of the low overall forest cover. Endress and Chinea (2001) and Turner et a1. (2001) looked at landscapes that were 70-90% forest throughout the study period. Kramer's (1997) work is more comparable in terms of the amount of forest cover, but she subdivided forest types differently making comparisons difficult. 64 Generally, over the period 1936-1989, agricultural land had a high probability of remaining under agriculture. Transition probabilities to agriculture from other cover classes were low, and remained low throughout the study period. Transitions 1 2 and 1 3 (Agriculture to Open and Closed Forest respectively) represented the abandonment of agricultural land. In total, the probability of agricultural abandonment remained fairly constant between 1936 and 1983 (ranging from 0.11 to 0.15) but jumped sharply in the final interval (1983-1989) to 0.36. Similarly, transition 1 1 (the retention of agricultural land in agriculture) remained fairly constant in the 1936- 1983 period, but declined sharply to 0.62 in the 1983-1989 transition. Throughout this period the probability of retention was much higher in Puerto Rico than it was in the southern Yucatan (where the retention probability was 0.35; Turner et al. 2001) and in Guanacaste Conservation Area in Costa Rica (where the retention probability of pasture was 0.41; Kramer 1997). Although the area under agriculture declined sharply (Lugo et al. 1996), the retention rate for agricultural land remained high. The probability of conversion to agriculture from other cover classes (2 1 and 3 1) was low, even when taking into account the relatively small amount of forested land available for conversion. In the Area de Conservacion Guanacaste, which saw forest cover increase from less than 15000 ha to over 17000 ha in the period 1979-1985, transition probabilities to pasture remained between 0.09 and 0.25, depending on the forest type (Kramer 1997), while those in the study area in Puerto Rico were in the range 0.01 to 0.13. On predominantly forested landscapes such as those in Palau (Endress and Chinea 2001) and the southern Yucatan (Turner et al. 2001) transition probabilities to agriculture were much lower, since the ratio of intact to cleared land is much higher. 65 Tu agi an (a I") 0 int Turner et al. (2001) found transition probabilities of only 0.02 from Upland Forest to agriculture, although the transition probability for secondary forest was higher (0.18). The transition from Closed Forest to Open Forest (3 2) declined over the study period from a high of 0.17 in the first interval (1963-1950/51) to a low of 0.04 in the penultimate interval (1971-1983). This transition is likely to represent biomass harvest — charcoal production, for example — or the degradation of forest by factors such as fire or the incursion of cattle. E.M. Sepulveda (personal communication) reported that his uncle attempted (unsuccessfully) to convert Site 11 to pasture at some time in the past by a combination of cutting and burning of the forested ravine. Actions of this sort could also be responsible for the 3 2 transition. While the use of charcoal as a fuel has declined (Murphy et al. 1995), charcoal pits were observed both in fragments (Sites 1 and 2) and on a private farm where Prosopis pallida trees in pasture were cut and used to make charcoal. Thus, biomass harvest continues to be a factor in Puerto Rican dry forest fragments albeit at a much lower level than in the past. With some exceptions (most notably, Coastal Scrub Forest) Open Forest is not a stable natural community in southwestern Puerto Rico. Left to natural processes, most Open Forest is likely to be converted to Closed Forest by succession. In this regard, it is not surprising that Open Forest had consistently lower retention probabilities (between 0.14 and 0.36) than did the other two target land-cover classes (0.62-0.84 for Agriculture and 077-090 for Closed Forest). In the absence of burning, pasture is likely to become open woodland (Janzen 1988a) through the colonization of spiny trees such as Prosopis pallida and unpalatable shrubs such as Calotropis procera (Ait.) Ait. f. and Lantana involucrata L. These areas are likely to show up as Open Forest in aerial photographs 66 (although pasture dominated by P. pallida appeared as Closed Forest in some cases). When these pastures are ‘improved’ and the trees are cut and burned, the transition from Open Forest to Agriculture is recorded, when in fact the land use has remained pasture throughout the transition. Dynamics of Focal Fragments The studied fragments provided a pattern of change that differed from what was observed at the landscape level. Twenty-five of these fragments supported forest cover in 1936. Twenty-three maintained forest cover throughout the period, while two were cleared and subsequently reforested. This suggests that forest fragments have a high ‘inertia’ — once a patch becomes forested, it tends to remain forested, a fact which agrees with the landscape-level findings. However, 14 of these fragments decreased in size, suggesting that drivers of deforestation remained active even when other land was being allowed to revert to forest. On the other hand, the fact that about half of the deforestation was caused by non-agricultural land transformation suggests that the causes of this deforestation may have been different from the ones that allowed agricultural land to revert to forest. Ramos Gonzalez (2001) found that built-up areas, forest and shrub land expanded at the expense of agricultural land. Similarly Lopez et al. (2001) stated that most urban and suburban development came at the expense of agricultural land. 67 Fragme were the t lowest ex isolation. that there the thresh Differenc< realized g; Th bigger the ”heap. I the Specie wPfidict Setitrated Sptties-jx fraEments “More 1 thine. W between ( lOIESI frag ofcanele Fragment Networks and Connectivity Figueroa Colon (1996) suggested that Guanica Forest and the Sierra Bermeja were the two main refuges for biodiversity during the period when forest cover was at its ' lowest extent. When a gap of 1 km between forest patches is considered significant isolation, there is connectivity between these two putative refugia, and it may be assumed that there is gene flow across the landscape. On the other hand, if a 500-m gap is used as the threshold at which connectivity ends, two gaps exist between these areas (Figure 3.2). Differences in species composition between Relict and Regrowth forest means that realized gaps are likely to be consistently larger for many species. The occurrence of gene flow across the landscape is a stochastic process. The bigger the gap between two forest patches, the lower the probability of gene flow across the gap. In addition, the severity of a gap or even the existence thereof, is a function of the species in question. In the absence of species-specific empirical studies it is difficult to predict thresholds of connectivity across the landscape. The only fragment that is separated from other forest patches by a gap of more than 1 km is Site 9, one of the most species-poor fragments studied (Chapters 4, 7 ). However, it also differed from other fragments in several important ways (see Chapters 4 and 7). Differences in community structure may be a function of environmental conditions and may not reflect isolation alone. Whatever the thresholds may be, if maintenance of some degree of connection between Guanica Forest and the Sierra Bermeja is seen as a priority, protection of the forest fragments between them is a priority. Two key areas are the strip of hills that run north of the resort town of La Parguera and the strip of forested hills immediately south of Carretera 16 in the Lajas Valley (Figure 3.2, Clusters 4, 5 in Table 3.4). These areas 68 lack fortt Ottrred ht Foundatit Parguera Implica Fo pattern of island's fc land-use c forest cot. by 1.8%p pasture. u- PIOleitt' det‘elopm ln Luna. whi has glmtr The ”Earl “Cut-tied t hmmgu hm ten Si lack formal protection for the most part (with the exception of Isla Magueyes which is owned by the F undacién Puertoriquefia de Conservacion (Puerto Rico Conservation Foundation) and are likely to be subject to development pressures as the town of La Parguera grows. Implications Forest recovery in Puerto Rico has been one of the major exceptions in the global pattern of tropical deforestation, but not the only exception. Important threats to the island’s forests remain including urbanization (Lopez et al. 2001). In a recent study of land-use change in northeastern Puerto Rico, Ramos Gonzalez (2001) found that while forest cover increased by 1.2% annually between 1978 and 1995, built-up areas increased by 1.8% per annum. Most of these changes came at the expense of agricultural land and pasture, with the net effect of bringing urban and suburban development into immediate proximity with forest. As a consequence, future expansion in urban and suburban development is likely to come at the expense of forested land. In the study area, several examples of this are evident. The community of La Luna, which was established by families that were resettled from within Guanica Forest, has grown to the point where it immediately abuts Guanica Forest (Lugo et a1. 1996). The creation and expansion of Barrio Belgica created Sites 7, 10 and 31; the area occupied by the community was formerly a forested hilltop. Site 35 was eliminated for housing construction, while the land formerly occupied by Site 24 was offered for sale as housing lots after the forest was cleared. The construction of homes also appears to threaten Sites 4 and 14, while Sites 21 and 36 appeared to be earmarked for subdivision 69 for boost and infra shrtnlt'ag agrteuht the tutor forest ret patch of high. the ettinetio factors tl enough t across th 311d exttt dlSlribm factors. Summ; 1’ FOIC. for housing lots when the fragments were visited in 1997. Between 1936 and 1993 urban and infrastructural development were the factors responsible for most fragmentation and shrinkage of the forest fragments that were the target of this study. Given the demise of agriculture in Puerto Rico, urban development can be expected to be the main threat to the future of forests. The limited area of forest that remains means that the threat to forest remains high. On predominantly forested landscapes, the probability of any given patch of forest being eliminated is fairly low; even if the absolute rate of deforestation is high, the relative rate is likely to be lower than in a largely deforested landscape. While many studies have attempted to use remotely sensed data alone to infer extinction risks and predict extinction trajectories (e. g., Ranta et al. 1998, Tole 2002), the factors that influence the relationship between species richness and area are still not well enough understood to confidently make such predictions. The distribution of species across the landscape is driven by factors such as resource availability, dispersal limitation and extinction debt. Conservation decisions need to reflect the actual patterns of species distributions within communities. The remaining chapters attempt to incorporate these factors. Summary 1) Forest cover was 23.2% in the dry forest zone in 1993; 18.0% Closed Forest and 5.2% Open Forest. 2) A total of 308 Closed Forest fragments totaling 13100 ha and 312 Open Forest fragments totaling 3800 ha were mapped. 70 3) Flllj for: fort 4) Fort east. 5) The malt “Cit Were 7.) On I] clust 1000- 87) Cont relati Rica Tellet lets] 3) 4) 5) 6) 7) 3) Fifty-two percent of all forest fragments were below 5 ha in area, but only accounted for 4% of the forest area; 3% of all forest fragments were over 100 ha, but accounted for 63% of the total forest cover. Forest cover peaked toward the middle of the dry forest zone and declined toward the eastern and western extremes. Twenty-nine of the forty fragments that formed the basis of the more detailed study maintained some forest cover throughout the period 1936-1993. Nineteen fragments were classified as Relict (> 75% ‘old growth’), three were classified as Mixed (25- 75% ‘old growth’) and seventeen were classified as Regrowth (< 25% ‘old growth’). Twenty-three of the focal fragments increased in size in the period 1936-1993, fourteen decreased in size, and two remained about the same size. Four fragments were eliminated in the period 1995-1998. On the basis of using a 500-m separation as the basis for ‘isolation’ between fragment clusters, nine distinct clusters were identified on the landscape. On the basis of a 1000-m separation, only two clusters were present; Contrary to expectations, the probability of conversion of forest to non—forest was relatively high in the ‘recovering’ landscapes of Puerto Rico and Guanacaste, Costa Rica, but was very low in the ‘deforesting’ landscape of the southern Yucatan; this reflects the limited area of forest on the ‘recovering’ landscapes rather than a high level of deforestation. 71 CHAP DRY F ltttrod lr Central). as suhtrot stnp alon kilometer system he et al. 197 Castilleja Dunphy p Ottstde o: (1998) ha nonhtteg TI asSociatit clliltacter Stems be I olthe lite natural gr Sllltlled ] CHAPTER 4: COMMUNITY STRUCTURE OF PUERTO RICAN DRY FORESTS Introduction In the rain shadow created by the mountainous spine of the island (the Cordillera Central), Puerto Rico’s south coast is dry. Ewel and Whitmore (1973) classified this area as subtropical dry forest (sensu Holdridge 1967). These dry forests originally covered a strip along the south coast about 120 km long (from west to east) and three to 20 kilometers inland (Ewel and Whitmore 1973). Most studies of the plant ecology of this system have focused on Guanica Forest, a 4000—ha protected area (e. g., Ewel 1971, Lugo et a1. 1978, Dunevitz 1985, Lugo and Murphy 1986, Murphy and Lugo 1986b, 1990, Castilleja 1991, Quigley 1994, Murphy et al. 1995, Dunphy 1997, Molina Colon 1998, Dunphy et al. 2000). Little has been done to document the extent or ecology of dry forest outside of this protected area (Murphy et al. 1995), although Vazquez and Koltennan (1998) have described the vegetation of the Punta Guaniquilla Nature Reserve at the northwestern comer of the dry forest zone. The most common forest association in Guanica Forest, the Deciduous Forest association, is the best studied. Murphy and Lugo (1986b) found this community to be characterized by a large number of relatively small-stemmed trees, with 56.9% of all stems belonging to multi-stemmed individuals. This was attributed to historical cutting of the trees, but Dunphy et al. (2000) found evidence to suggest that this may reflect the natural growth form of many of these species. The other major associations are less well studied. The Semi-Evergreen association has taller trees with fewer stems per individual 72 (lugo (I more it n i forests th and Cool much oft (1994)— 3). If the H998) an W has t agncultun likely to b Regtttttth hand, if st Sites in lllt a‘éfitultur; and REgrc lmllS Ofs NeSted 5 ll) Scales Will] (Lugo et al. 1978, Castilleja 1991) while the Scrub Forest association is more open, with more widely spaced trees (Lugo et a1. 1978, Castilleja 1991). Forest fragments outside of Guanica Forest are likely to have originally supported . forests that were broadly similar to the associations present in Guanica Forest (Gleason and Cook 1926). However, these areas have experienced a wide range of disturbances; much of the dry forest present outside of Guanica Forest is secondary forest sensu Corlett (1994) — that is, forests that have developed after agricultural abandonment (see Chapter 3). If the successional trajectories of these forests are similar to what Molina Colén (1998) and Erickson et al. (2002) have documented within Guanica Forest (and Roth, 1999 has described in the Dominican Republic), where 50-year-old forests on abandoned agricultural land are species-poor and dominated by exotic leguminous trees, then there is likely to be a large difference between the community compositions of Relict and Regrowth forest fragments (see Chapter 3 for a definition of these terms). On the other hand, if successional trajectories are more similar to what Ray (1993) described in older sites in the US. Virgin Islands, where 50-150-year-old successional forests on abandoned agricultural land were dominated by a species-rich mixture of native species, then Relict and Regrowth fragments are likely to be less different, and instead form a continuum in terms of species composition. Nested Subsets in Fragmented Communities While species-area curves describe the overall pattern by which species richness scales with increasing areas, they do not describe the overall patterns in species occurrence on a landscape. The theory of island biogeography (MacArthur and Wilson 73 l%7) pr predict l pointed overlap distribut island tt more spt islands i relations 3000, an extinctic lhhere s abandon Winning is a con Commur pmlectir 1967) predicts how species richness will scale with area in an archipelago, but it does not predict how species will distribute themselves. As Simberloff and Abele (1979, 1982) pointed out, the number of species present in an archipelago is a function of the degree of , overlap between the species compositions of individual islands. Nested subset theory (Patterson and Atmar 1986) predicts that species distributions within archipelagos are likely to be nested. The species present in any given island will be present in all islands that are more species-rich, and all species present in more species-poor islands will be present in that island. The species composition of islands is thus considered to be deterministic. Extinction, colonization, disturbance, habitat distribution, hierarchical niche relationships and passive sampling can all produce nested patterns (Patterson and Atmar 2000, and references cited therein), but these patterns are especially apparent in extinction-driven systems. The studied fragments are a mixture of relict patches of forest (where species composition is likely to be driven by extinction) and regrowth on abandoned agricultural land (where species-composition is likely to be driven by colonization; see Chapter 3 for a detailed history of the studied fragments). Nestedness as a community characteristic does not seem to have been described for dry forest plant communities, but can it yield information on the incremental conservation value of protecting additional fragments smaller than those that are already protected. 74 Object Met/10c Data Cc fmm a tc elil’e’lfiet abundam c0mplete COllectjm Objectives 1) To determine whether the fragments support a species-rich native forest community or whether they are simply depauperate stands dominated by weedy exotic species; 2) To determine whether the basal areas and stem densities of the fragments are comparable with those found in Guanica Forest; 3) To determine whether distinct assemblages can be delineated on the basis of their plant species composition; 4) To evaluate the degree of nestedness present in the assemblage of dry forest fragments. Methods Data Collection Vegetation structure, species composition and abundance data were collected from a total of 40 dry forest fragments and in reference plots located in the semi- evergreen, deciduous and scrub forest associations in Guanica Forest. Structure and abundance data were not collected from two of these sites (Sites 31 and 32) and a complete species list was not compiled for Site 3. Methods of site selection and data collection are described in Chapter 2. 75 Commt fragmer Forest. plots co for plan interpret and hem height 0) hectare, based on stemmed l ftagmen melett Wilde Nested Using the PallilSOn Community Characterization Dominance-diversity curves (Whittaker 1965) were constructed for the 38 fragments for which abundance data were collected and for the reference plots in Guanica' Forest. Relative abundance of each species was calculated within a pooled sample of all plots collected at a single site. In keeping with Hubbell’s (2001) definition of a “guild” for plant species (all species competing for a fixed pool of resources, in this case interpreted broadly) comparisons were limited to plants rooted in the ground. Epiphytes and hemi-parasites were thus excluded. Basal areas were calculated from diameter at breast height (dbh; measured at a height of 1.5 m) measurements of stems 2 1 cm, and were expressed as square meters per hectare. Data were pooled across plots and the basal area per fragment was calculated based on the total pooled sample area, rather than on the mean of the plots. The multi- stemmed nature of the community was also investigated. Jaccard’s coefficient of community (J accard 1900) was calculated for the 39 fragments for which complete species lists were compiled. Hierarchical clustering with complete linkage was used to group these fragments on the basis of their Jaccard coefficients (see Legendre and Legendre 1998). Nestedness in Fragment Species Assemblages The degree of nestedness in this overall assemblage of species was measured using the Nestedness Temperature Calculator (Atmar and Patterson 1995; see Atmar and Patterson, 1993 for the underlying theory). The extent to which the assemblage deviates 76 from p “Temp the one frame by re-c ttas m: bottorr tt‘ideSp line“ is occum Where diagor Where “l litre from perfect nestedness (the amount of disorder in the system) is termed the “Temperature” of the presence-absence matrix. The software (Atmar and Patterson 1995) calculated “Temperature”, a measure of ~ the unexpectedness in the presence-absence matrix was calculated across species and fragments. The presence-absence matrix was arranged into a state of maximal nestedness by re-ordering entire rows and columns until unexpectedness in the occurrence of species was minimized. The top row thus represented the most “hospitable” fragment and the bottom row the least hospitable one. The left-most column was occupied by the most widespread species, and the right-most column the least widespread one. The “boundary line” is the hypothetical line that separates the portion of the matrix that is expected to be occupied from the portion of the matrix that is expected to be unoccupied. Local unexpectedness of cell if was calculated according to the formula: _ 2 where d,_-,- measures the distance of the cell from the boundary line along the skew diagonal and Dij is the length of the matrix parallel to the skew-diagonal. Total unexpectedness was calculated according to the formula: U =1/(mn)2 Z uij where m is the number of rows and n the number of columns. System temperature, T was calculated as T = k U where k = 100/Umax. 77 ha (.4 r. \t‘hic calcu stud/'60" I ///2 Comma The probability that this degree of nestedness in this species-site matrix could have been produced at random was tested using a Monte Carlo resampling method (Atmar and Patterson 1993). Nestedness Temperature was calculated for plant species in the 39 fragments for which complete species lists were compiled. For comparison purposes, it was also calculated for lizard (Genet 1999b) and termite (Genet 1999a) communities which were studied in a subset of these fragments by other investigators. Results Community Characterization Studied dry forest fragments ranged in size from 6 x 10'3 ha to 1372 ha (see Chapter 3 for details of the history of each of these fragments). Total species richness ranged from 17 to 173 species per fragment. A total of 64511 individuals in 10435 m2 (380 25-m2 plots and 4 larger plots ranging from 60-200 m2 in the fragments and I9 25- m2 plots in Guanica Forest) were inventoried in species plots. (A further 5142 m2 were inventoried in transects in which only trees _>_ 1 cm DBH were inventoried). Dominance-diversity curves showed a large range in terms of their degree of dominance (relative abundance of the most abundant and least abundant species) and their diversity (slope of the curve). Fragments are presented grouped according to hierarchical clustering of Jaccard similarities of species composition (Figure 4.1). The individual curves are presented in Appendix 2. 78 .5 4 .9" r e A: 10000 3 + Sits—{3.77 1000 i 8 . a a t 13 100 3 5 s .9 «< 10 f 1 4 r - — e~e— 80 100 120 a. Jaccard Cluster 1 Figure 4.1: Dominance-diversity curves for dry forest fragments and Guanica Forest, southwestern Puerto Rico. Fragments are presented grouped according to hierarchical clustering of Jaccard similarities of species composition (see Figure 4.2 and text for details). 79 10000 : :*+”§1;§7 1 ‘ -I- Site 28 ‘ —A— ' 1000 . 1 Site 391 0 ‘ , ,, §E¢740 l 0 E a "U fl 5 .D < 40 60 80 100 120 Rank b. J accard Cluster 2 10000 '3 #7-, 7 7, 7 7 1 + Site 28 -l— Site 4 ‘ l -A— Site 7 + Site 10 1000 1 ‘ +Site 12 +Site 15 j —— Site 6 —-— Site 8 3 1 —<>—Site 21 —U—Site 25 fl , 1 .g 100 1‘ . I A—A— Site 27 —X—Site 36 g 1 .D < 80 100 120 c. J accard Cluster 3 Figure 4.1: (continued). 80 JOE-n-vfinun-(x d. Jaccar fig... OOH-nutshI-Audyx tlaccard Figure 4. 10000 i 4.: Site '1 1 + Site 13 + Site 23 —x— Site 26 + Site 30 —1— Site 29 —-- Site 34 -- Site 19 ‘°‘ £6 20 , :‘3’ 3,0635 Abundance 80 100 120 d. J accard Cluster 4 10000 ‘ %_._’"’sg§ ‘ {-I-Sitel81 l—A—Sitezzi 100° 1 z—x—Sitel 3 ' i—x—SiteZ i ~+Site l4) t—i:Site17/i Abundance 100 120 Rank e. J accard Cluster 5 Figure 4.1: (continued). 81 Abundance Uaccart Abundance g' GUénic; HEW” 10000 ~ .,_, _fi _- 1 1+‘Site 161‘ 1 1+. £099.”; 1000 3 1 5 “g 100 1 :i 4: <1: 10 1 1 r ;—-— , »-.—T——————-—~ O 20 40 60 80 100 120 Rank f. J accard Cluster 6 10000 g + Guanica Forest I 8 :1 a 1: :1 :1 .n 4 Rank g. Guanica Forest Figure 4.1: (continued). 82 Guinic; 11375 t support: 25.8 m: reflect 1} of the be the two I than did 1112 1121'l 1 l Deciduo fragmentl cm dbh 1 by lluipf {figment FCrest, b (1936b). A $1th (7: ”it {Tab ”111th Baseline structural data gathered from the three main forest associations in Guanica Forest are presented in Tables 4.1a and 4.1b. Stems densities ranged from 11275 stems ha'l to 24857 stems ha'l (Table 4.1a) with the Deciduous Forest association ' supporting the highest stem density. Total basal area (BA) ranged from 16.9 m2 ha.l to 25.8 m2 ha.l and was highest in the Scrub Forest association. This anomaly appears to reflect the cactus Pilosocereus royenii; eight individuals of P. royenii accounted for 50% of the basal area of one of the two Scrub Forest transects (40% of the total basal area of the two transects combined). The Semi-Evergreen Forest association had a higher BA than did the Deciduous Forest association. Basal area of the fragments ranged from 5.1 2 -1 2 -1 m ha to 47.6 m ha (Table 4.2). Twelve of the 41 fragments had basal areas that were lower than that of the Deciduous Forest measured in Guanica Forest (stems 2 1 cm dbh). Nine of these fragments had basal areas lower than the Guanica Forest values when only stems 2 2.5 cm dbh were considered, although 18 of them had basal areas lower than those recorded by Murphy and Lugo (1986b). When only stems 2 5 cm dbh were considered, six of the fragments had basal areas lower than the value recorded for Deciduous Forest in Guanica Forest, but 12 of them had basal areas lower than that reported by Murphy and Lugo (1986b) A total of 11412 stems Z 1 cm dbh on 4699 trees were measured. Of these, 8797 stems (77.1%) were part of multi-stemmed clumps with between two and 34 stems per tree (Table 4.3). The mean number of stems per tree was 2.43 for all trees and 4.22 for multi-stemmed trees (Table 4.3). Among stems 2 2.5 cm dbh, 68.3% belonged to 83 #:an fivnvctflm~n-uu «0:3 \Andqawm Hie: CC 30,865 Cal: O~L03n~ .aZUhflvuh 302.3“:de C.— xflum:3~r:nCCU unumw~h~ “viutho muUmquhhunU-hLECU :HLDuUZLuW tU£~KC bannuahkzvn «ht? “hanwrh £92383; 3.5m oma 05 5 EB— RcB—sotwe 803 35 32m 528 begooom .o .mnmfiwouonm Econ 98— ofi E 620.“ @820 3:0ng 86 $03 $28 2332 .3 833683 388m 32668 23 58330483 5233 52889 d 9% $3 $2 mmnw mm"? emcee .Eem a: _ : £88 .3 -- -- Sofie -- -- -- e eafiem :8: -- been -- -- e: m -- $2an $93: own..— Smm ES -- -- -- -- a. 29:2 2: $3: comm .. -- ace 5 t : .3 B owsq 82 Ba seem as: we m E: Seam were fie a... m N fie so 2 N fie so _ N 520m 288 in Eu m N new 80 Wm M nan Eu _ N “meow nooawcgoimfiom fie .5 m N fie so 3 N fie :5 _ N amp—Ohm SOSUMUOQ ATE 2:83 base 8on .a .see eeemfia 9.8 38m $5 no women SE 885m .329“ 8330 5 Q5888 Ema 05 mo 8358.2.an 35.8.5 05 mo $885 "3. 035. 84 A a: z — ~ I ~ . N r V neiv (‘9 Nah-.IJE‘MW 0R. flung... . .i mnmecwoponm Eton 0mm“ 05 E was assuage. 825 35 82a 328 bewaooom d .mamfiwouona 38a 32 05 E “mecca 8on votommsm $5 8me $23 2822 .o 3% :5 ~ 263m 805 “mi 88 macaw =e mo no.8 Ewen 30H .3 823888 388m 3052026 use 52305283 5953 aoEmcfih d e 2: e ”.2 o a: £83 .3 -- -- -- t- .. omdfi .. -- -- be 8325 :3: -- mm -- -- o9 -- -- 3N -- e358 Aeewofiv owsq -- -- t- ed QC fiAm. av .. -- .1 a. .3952 we 68: Nd -- 1 fim .. .. e53 .. -- .3 B ownq NS 3N mam 3 ed 3: 5: N2 mom aeam fie fie :5 m Nfie :5 2 N fie :3 _ N fie :5 m N fie Eu 2 N fie a... _ N fie :5 m N fie :5 2 N fie :5 _ N “88..“ naom “mecca msosEooO “modem 52w8>9€6m pee NE 82 33m .fi .Aeeeeeeee a... 2%... 85 {AsmmCUU nF—Uum flea-w 30L“ 2321K. nfiuCUHCWw-whbq HZULn: \fiLnV Unvmuuzum 1N0 MUmuImLUuUBCJN-LU ~n~h3~UEHm ozuoho \fiawEEert uNt‘t .U~n-~y~t mi 2: wmm \tmmH Sec Scam mm as v.3 c.c~ mm: mmcm Soc _m mm“ v.3 3: 83 cmhm ccmm cm c.cm m.mm 5mm cmmm cc: ccec 2 mm v.3 5: cog ch Seem 3 cc cm _.m Sm Svm 532 2 ad ed as sea comm mmmc_ 2 5.2 $2 c.cm 3Q 83 acme 2 Emfi _.E 2: cc? ccG ccwmc 3 Wm omcfl mmcm Km: 2 cc 5.: <2 omm_ mwcm mmh: m_ v.2 m.cm m.mm cmm_ cmnc ccci 2 3: NE cm. mmE cccv mflw cc mc me m6 of cvcm ccnmm c Tmm m.mm wém cccm mmcv $3 w 2.. 5mm m.cm comm cccw 3me 5 QB 02 Q: ccm~ ccwm some c c.cm m.mm 9mm 3mm mmmm cog" m as m.: m.m~ 83 came coww v m. cm _.nm c.cm cccm came cgwc m Om 5.3 Q: cwcm coon cow“: m mm «.3 0.2 cm? whom 8me c gnu Bo m N :3 :8 mm N 5c :8 c N saw :5 m N 5c Eu mm N new :8 _ N 3538: 6c NEV 82 38mm 358: com 583 bacon 88m .bmmaoc 88m 28 8.8 Emma ”ScoEwec “8.8.“ it @295 Mo mowmtowoeano 3502.8 05 co Egm “mé «Sam. 86 S 3; n: 82 83 88m 3 08 e8 EN 28 88 88 8 3: m8 SN 88 8mm 82: 8 28 8.8 5.8 88 282 85 8 :8 8.8 2.8 82 82 82 R 3: $2 88 :R 8 EN EN EN 88 88 ~82 mm at 0.: 2: E 82 38 em 3. on em 8: 88 88 mm 5.2 8 RR 8: 8 8.2 e8 mm 88 86 $8 8 e8 5; ms 88 8% 8w: 8 in w. _ m an. 88 $2: 8: R q: n: 8.2 88 we: 88 em 28 e8 68 88 :8 :2: 8 we no 8: o8 88 88 8 m. : _.ee e5. 88 8; 85 mm fie :5 m N fie es 3 N fie so _ N fie :5 m N fie :5 3 N fie :3 _ N @880: com mac 8.3. 885 A288: cum 2:89 .96qu 8on .Aeeeeeeee "S. 289 87 multi-stemmed trees while 45.9% of stems Z 5 cm dbh were part of multi-stemmed trees (Table 4.3). The clustering algorithm produced a total of 7 clusters based on a 0.8 cut-off (Figure 4.2). Clusters 1 and 7 each consisted of a single Regrowth fragment (see Chapter 3 for details of the successional history) with unusual species compositions. Cluster 2 consisted of four Regrowth fragments. Cluster 3 consisted of 10 Relict, two Mixed, and one unclassified fragment. Cluster 4 consisted of eight Regrowth and three Relict fragments. Cluster 5 consisted of five Relict, one Mixed and one Regrowth fragment. Cluster 6 consisted of two Relict fragments. Nestedness in Fragment Species Assemblages All three species-occurrence matrices were significantly nested. The plant species x fragment matrix consisted of 393 species present in a system of 39 forest fragments with a matrix fill of 19.4%. The lizard x fragment matrix consisted of 10 species in 10 fragments (Genet 1999b) with a fill of 59.5% and the termite x fragment matrix consisted of 9 species in 10 fragments (Genet 1999a) with a fill of 53.8%. The Nestedness Temperature of the species x island matrix was 14.6°, while those of the lizard and termite matrices were 4.l6° and 11.31° respectively. Based on Monte Carlo estimation, the probability of a matrix of the temperature of the plant matrix or less being drawn at random was < 0.00001. The probability of a matrix of the calculated temperature or less being drawn at random was less than 2.75 x 10'5 for the lizard matrix and less than 2.15 x 105 for the termite matrix. 88 LREGROWTH Site 24 2.REGROWTH Site 40 2.REGROWTH Site 38 l'_| 2.11130110wr11 Site 37 l I 2.11130 ROWTH Site 39 3.1113er Site 31 3.RELICT Site 21 3.RELICT Site 36 3.RELICT Site 28 3.UNKNOWN Site 12 3.REL|CT Site 15 _‘ 3.RELICT Site 7 : :l 3.RELICT Site 4 3.M1XED Site 10 3.11EL1CT Site 6 3.MIXED Site 25 —' 3.RELICT Site s l""] 3.111:le Site 27 4.REGROWTH Site 20 4.REGROWTH Site 34 I I_— 4.REGROWTH Site 19 4.11120 ROW’I‘H Site 35 4.RELICT Site 13 , 4.1113er Site 11 4.REGROWTH Site 26 4.REGROWTH Site 30 4.REGROWTH Site 32 4.REGROWTH Site 29 4.1113er Site 23 S.RELICT Site 1 5.REL1CT Site 5 1'3— 5.MIXED Site 2 5.RELICT Site 14 5.RELICT Site 18 5.REGROWTH Site 22 5.RELICT Site 17 I ‘ 6.RELICT Site 16 6.RELICT Site 9 7.11130110er Site 33 l l I l I l l 0.0 0.2 0.4 0.6 0.8 1.0 1.2 Distances Figure 4.2: Hierarchical cluster dendrograms of Jaccard coefficients based on the species composition of Puerto Rican dry forest fragments, southwestern Puerto Rico. Clusters were delineated on the basis of a cut-off distance of 0.8. Numbers refer to clusters referred to in the text. 89 owe mm 8.». 34 9mm v.3 new Eu m N 85% m.ww on Qum cod hem «:3 new :8 Wm M £8on fl: em mm.v med 034 5.3 now 80 fl N 288m 8:. @088on 888on 388on 388on 8820 388on 6m 8:on -232 Bwfim -252 038m -258 E 88% .«o 83:52 mo owficoeom 988“er 02,—. con macaw mo 838: Z mood. we owficoobm .828 be SSE 8.55 5 much BEEBméBE use -336 macaw 258 .«o :ouspwammu 2E. and 933—. 90 Due to the differences between the plant matrix and the animal matrix (in terms of the number and range of fragments sampled) the two results should not be taken to be based on identical samples. However, Since all but one of the lizard and termite samples are a subset of the plant plots, some comparisons are certainly warranted. Discussion Community Characterization As is typical of Caribbean dry forests (Murphy and Lugo 1995) both the fragments and the reference plots in Guanica Forest consisted of large numbers of small- stemmed trees (Tables 4.2 and 4.3). AS Lugo et al. (1978) pointed out, if a cut-off of 10 cm stem diameter (standard in forestry work) were used to characterize this community, there would be very few trees. While a 2.5 cm diameter at breast height (dbh) cut-off has been fairly standard in community studies (e. g. , Gentry 1982, Murphy and Lugo 1986b, Gillespie et al. 2000), a considerable proportion of the biomass is accounted for by stems between 1 cm and 2.5 cm. Between 0.2 m2 ha'1 and 5.2 m2 ha.l (or up to 55% of the basal area in stems over 1 cm dbh) is accounted for by stems in this range. This is Similar to what Murphy and Lugo (1986b) found in their study in the Deciduous Forest association in Guanica Forest, where stems under 2.5 cm dbh (all measurable stems; they did not use a 1 cm cut-off) accounted for 16% of the total basal area. The findings of this study for the most part do not deviate from previous studies of Guanica Forest (Lugo et al. 1978, Murphy and Lugo 1986b, Castilleja 1991) except in the case of Scrub Forest. Stern densities and basal area in the Scrub Forest differed 91 sharply in magnitude and pattern from the findings of previous studies. Lugo et al. (1978) found the density of stems Z 5 cm dbh to be 540 stems per hectare, and the basal area 4.2 m2 ha'1 in the Scrub Forest association. Similarly, Castilleja (1991) found the density of stems 2 2.5 cm dbh to be 1556 stems per hectare, and the basal area to be 3.3 m2 ha". Both studies found that the Scrub Forest had the lowest stem density and basal area among the three associations. This is in sharp contrast with the findings of this study in which the Scrub forest had the highest density of large stems and the highest basal area among the three associations. There are several factors that may be responsible for this lack of agreement. One is the definition of Scrub Forest. There is no study of which I am aware that seeks to define the various associations that are present in Guanica Forest. The associations appear to be a modification of Beard’s system of vegetation classification (Beard 1944, 1955) and are fairly intuitive, but they represent real communities that blend into one- another, usually without Sharp borders. As a consequence of this, it is possible that I defined Scrub Forest more broadly than did other authors, with the result that my sample included more large trees than did the previous studies. While this explanation may account for some of the differences in basal area between this study and previous studies, it fails to account for the qualitative differences. 1 found that Scrub Forest had the largest basal area of all the associations. Thus, the discrepancy cannot simply be explained in terms of the use of a broader definition of Scrub Forest than other studies. Another possible explanation for this discrepancy lies with the fact that sample areas are relatively small. While both Lugo et al. (1978) and Castilleja (1991) used a sample area of 1000 m2, this study used a 300 m2 sample; however, the fact that a smaller minimum stem 92 diameter was used in this study means that similar numbers of stems were measured in all three studies. Whether the sample measured was large enough to obtain a stable estimate of stem density or not, the major factor influencing stem density and basal area estimates for the Scrub Forest association was the cactus Pilosocereus royem'i. If the eight individuals of this species had not been present in the sample the estimated basal area would have been 15.5 1112 ha]. This would have resulted in a pattern that qualitatively agreed with previous studies, with Scrub Forest having a lower estimated basal area than the Deciduous Forest or Semi-Evergreen Forest associations. On the other hand, there is no reason why these individuals should be excluded — this species is, in fact, one of the characteristic Species of the Scrub Forest association (Lugo et al. 1978, Castilleja 1991). Five fragments had extremely low basal areas (< 10 m2 ha"l based on stems Z 1 cm dbh; Table 4.3). Three of these (Sites 9, 16 and 17) were coastal fragments with fairly open canopies resembling Scrub Forest. The other two were disturbed fragments. Fragments differed substantially in terms of species-abundance relationships. As illustrated in Figure 4.3, the difference between the most diverse (Sites 4 and 5) and the least diverse (Sites 9 and 38) fragments are substantial, both in terms of the relative abundance of the dominant species and in terms of the overall diversity of the sampled species. The fact that sample sizes were unequal between species-rich and species-poor fragments confounds the relationship (a larger sample increases the probability of encountering rare species) but it is still apparent that there are marked differences in species diversity among fragments (see Appendix 2). 93 8:836 8052 06 macaw 08 @0883 m use $08:me v £35 223 .mEoEwem 365368032 95 05 2e Amofiwcaud m BE 28 Ame—86v mm 8mm .83 855 5883:5an E mpcofiweh “meow be So.“ 8m 833 bmmuozfirooswfiaom 21:9 25w:— xnam 35:52:. _288 voooooooooA III-IIlllI-IIIIIIIII See H a m m. 3 w 1 Ed n u D. B u t aillllllil 3 _ msax e eema m asa< _ -~8 ramsaoim «_ 94 The clustering algorithm produced three large clusters and four small clusters when the cut-off distance of 0.8 was used (Figure 4.2). In reality, two fragments (Site 24 and Site 33) were not assigned membership in any cluster and were only joined to the other clusters at a distance of 1.0 (which amounts to a similarity of 0.0). Thus Clusters 1 and 7 are simply fragments that were not assigned to any cluster. They were also not Similar to one-another. The deepest cleavage among the other clusters is between Clusters 2, 3, and 4 on one hand, and Clusters 5 and 6 on the other. Cluster 2 was a group of small Regrowth fragments each centered on a single Pisonia albida tree. Cluster 4 included the remaining Regrth fragments and two disturbed Relict fragments. These fragments were dominated (to varying extents) by Leucaena leucocephala. General observations of secondary dry forest in southwestern Puerto Rico suggest that there are two or three successional pathways possible on abandoned agricultural land. The most common type involves Leucaena leucocephala and Prosopis pallida. P. pallida is a common pasture tree; cattle will eat it and disperse its seeds (Janzen 1986a). Active pasture can develop a continuous canopy of P. pallida (personal observation). Abandoned pastures are invaded by L. leucocephala, resulting in a simplified community that appears to be stable, at least in 50-year—old abandoned agricultural land (Molina Colon 1998). In a similar situation in St John, US. Virgin Islands, Ray (1993) found that a 33-year-old abandoned pasture was dominated by L. leucocephala while a 50-year-old pasture had a more Species-rich community dominated by Bourreria succulenta. Ray (1993) hypothesized that the difference was a consequence of grazing by feral donkeys in the L. leucocephala Site which was arresting succession. However, Molina Colén (1998) 95 found a species-poor community dominated by L. leucocephala on 50-year-old agricultural land in the absence of grazing. This suggests that this community is fairly stable and not just a transient successional assemblage. An alternative successional pathway, which appears to involve Pisonia albida, is typified by Cluster 2. Patches of P. albida-dominated regrowth are often nested within larger areas of L. leucocephala-dominated regrowth and can also develop in active pasture (personal observation). These patches Show a marked difference in Species composition relative to the surrounding L. leucocephala-dominated regrowth; they are much more Species rich and often consist of fleshy-fruited trees. While both Leucaena leucocephala and Pisonia albida are dry fruited, P. albida often shows high levels of infestation with mistletoes (Phoradendron spp.) These fleshy-fruited hemi-parasites will attract frugivores (Watson 2001) and thus enrich the seed rain under P. albida trees (Guevara et al. 1986, Guevara et al. 1992, Guevara and Laborde 1993). See Janzen (1988a) for a description of the process by which ‘nuclear trees’ can develop into forest islands. A third successional pathway may involve colonization by Bourreria succulenta. Ray (1993) found a 50-year-old pasture to be dominated by B. succulenta. Similarly, in Puerto Rico, disturbed areas can be dominated by this Species. No fragments were distinctly dominated by this Species, but roadside areas and bulldozed areas adjacent to one fragment (Site 14) were dominated by this species. As a fleshy-fruited species, B. succulenta should be attractive to frugivores, which are likely to deposit a seed-rain rich in fleshy-fruited species. 96 The proportion of trees that were multi-stemmed (44.3%) was consistent with previous studies; Dunphy et al. (2000) found 43.3% of all trees were multi-stemmed. Similarly, Murphy and Lugo (1986b) found that 57% of all stems 2 2.5 cm dbh were members of multi-stemmed individuals. This is smaller than the values of 68.3% for stems 2 2.5 cm dbh and 77.1% for stems Z 1 cm dbh that I found in this study. The average number of stems per individual ranged from 1.95-2.43 depending on whether a 1 cm, 2.5 cm or 5 cm minimum diameter was employed. Similarly, the number of stems per multi-stemmed tree ranged from 3.72-4.22. The method used here is an underestimate of the number of trees that are multi- stemmed Since it is based on a cut-off diameter at breast height of 1 cm. Several trees that are reported as being Single stemmed had additional stems that were below the 1 cm cut-off (personal observation). In addition, the proportion of single-stemmed trees is higher than might have been found if the sample were restricted to Relict fragments since Leucaena leucocephala, which dominated most Regrowth fragments, is generally single stemmed. Nestedness in Fragment Assemblages As expected, the system shows significant nestedness, not just for plant species, but also for animal Species. The very low Nestedness Temperature (high degree of nestedness) indicates that fragments with fewer species have a subset of the species that are present in more species-rich fragments. This means that smaller fragments are unlikely to add species that are absent from larger fragments. While secondary growth on abandoned agricultural land has fewer species than older forests, these species are also 97 present in relict vegetation. Many of these Species are components of the native forest (or at least they appear to be in modern disturbed forests). Others, like Leucaena leucocephala, are exotic species, but have either successfully invaded disturbed relict forest, or have colonized the edges of expanding fragments. When the comparisons are made between inventory lists, these species appear as part of the vegetation of all fragments. These issues are expanded and further discussed in Chapters 5 — 7. Summary 1) 2) 3) Relict fragments supported a species-rich assemblage that was dominated by native plant species; Regrowth fragments supported fewer species and many were dominated by exotic species, especially Leucaena leucocephala. Like other dry forests, both the fragments and Guanica Forest consisted of high densities of small, multi-stemmed trees; much of the basal area was accounted for by small trees. Although overall stem densities and basal areas of Guanica Forest and the fragments fell within the range observed in previous studies, the estimated basal area for Scrub Forest in Guanica Forest was not consistent with either expectations or previous studies. Five distinct clusters (and two unassigned sites) were found among the fragments on the basis of a matrix of Jaccard Similarities. These included three predominantly Relict community types (including one coastal community) and two predominantly Regrth community types (one dominated by Leucaena leucocephala and one dominated by Pisonia albida). Overall there appeared to be a real separation between 98 Relict and Regrowth fragments on the basis of species composition, although the main Regrowth cluster was nested between two clusters of Relict fragments; Mixed fragments were not distinct from Relict fragments on the basis of Species composition. 4) Nested subset analysis found that fragment communities were significantly nested (as 5) were lizard and termite communities in a subset of the fragment assemblage). A large proportion (44.3%) of all trees was multi-stemmed. Trees averaged 2.43 stems per tree; multi-stemmed trees averaged 4.22 stems per tree. The tree with the largest number of stems, 34, was a Coccoloba microstachya individual present in Site 21. 99 CHAPTER 5: SPECIES-AREA RELATIONSHIPS OF PUERTO RICAN DRY FOREST PLANTS ON A FRAGMENTED LANDSCAPE. Introduction The increase in species-richness with increasing area is considered one of the closest things to a general rule in ecology (Schoener 1974, Lomolino 2000). This relationship is most commonly formalized using the power function or Arrhenius equation (Arrhenius 1921) S = cAZ (where S = number of species in an area of size A, and c and z are fitted constants). Values of the parameter 2 usually range from 0.1 to 0.5 (Rosenzweig 1995). Values at the lower end of the range are common in samples taken from continuous habitat while those taken from islands within an archipelago are usually around 0.26 (the ‘canonical’ value found by Preston, 1962). Values are higher when environmental thresholds are crossed, and are highest when they include areas that draw on different source pools (i. e. , when evolution plays a different role in structuring the biotas of different areas). While there is an established body of theory regarding the value of the parameter 2, there is none regarding the parameter c (Lomolino 2000). Mathematically, S = c when A = 1. Thus, it appears reasonable to consider c to be a measure of Species per unit area (or, as such, ‘species density’); Hubbell (2001) equated c with p, the number of species per unit area. Since species per unit area is a function of the units of area used, it is reasonable that values of c should vary over orders of magnitude, as Lomolino (2000) pointed out, depending, for example, on whether the unit of measurement is square meters or square kilometers. While the power function is the most common model of the 100 species-area relationship, others are also used. The semi-log or Gleasonian model (Gleason 1922) S = k0 + k1 log A (where k0 and k] are fitted constants) is ofien used in plant ecology (Rosenzweig 1995, Lomolino 2000). Less theory exists regarding these constants than about those of the Arrhenius equation. Both of these models have been criticized on two grounds: that they are unbounded (the relationship does not asymptote), and that they do not account for the ‘Small island effect’. Several criticisms of the power function as a model of the relationship between Species richness and area derive from the idea that species-area curves Should asymptote. The assumption is that island biotas are drawn from finite Species pools, and aS such Should not increase indefinitely. An equation that adequately models the species-area relationship should, thus, asymptote. One weakness in this criticism that has been pointed out by He and Legendre (1996) is that it is statistically unwarranted to use a regression curve to extrapolate beyond the range of the data. It should not matter what a species-area curve predicts beyond the range of the data. While some of the criticisms that the power function vastly overestimates the species richness at large scales are based on fitted data (Plotkin et 01.2000), in other cases they are based on extrapolated predictions from much smaller samples. Thus, the criticism may be based on a failure of the model to perform when used improperly — more a criticism of the methodology than of the model. In addition, as Hubbell (2001) pointed out, there is room for an infinite number of species among infinite individuals in infinite area. Thus, despite what has been said elsewhere, the idea that species—area curves should tend towards infinity rather than asymptote should not be problematic in and of itself. However, most people think in terms of the fact that a finite area will have a finite number of Species, and expect that all 101 of those species will be found before the final plot is sampled. From this idea comes the expectation that any real species-area curve Should asymptote. The expectation that species-area curves should asymptote may also reflect a bias that comes from working in species-poor systems (usually in the Temperate Zone). Species-area curves do not asymptote in even the largest data sets for moist tropical forest trees (Condit et al. 1996, He et al. 1996, Plotkin et al. 2000). While it has been argued that the power function overestimates species-richness at large scales (May 1975, He and Legendre 1996, Plotkin et al. 2000), it has also been observed that species-area curves tend upward at larger spatial scales, when areas with different evolutionary histories are included (Preston 1962, Shmida and Wilson 1985, Rosenzweig 1995, Hubbell 2001). The existence of a small island effect (SIE) is a valid concern with regards to the use of the power function to model the species-area curve. Unfortunately, there seems to be disagreement as to what form an SIE should take. According to Lomolino and colleagues (Lomolino 2000, Lomolino and Weiser 2001), the SIE is the lack of a relationship between species richness and area in small islands; many small islands have fewer Species than are predicted by the power function. However others (May 1975, He and Legendre 1996, Plotkin et al. 2000) have found that there was a lack of fit of the power function in small islands because the observed slope of the species-area curve was steeper than the predicted curve. MacArthur and Wilson (1967) noted that species richness is independent of area for relatively small islands. Williams (1996) pointed out that many small islands have no species, and their omission from species-area curves biases our view of species-area relationships. As a consequence of this, Lomolino (2000) makes a case for the use of 102 sigmoidal models of the species-area curve, and in particular recommends the sigmoidal Hills]0pc function as a model with biologically meaningful parameters. He and Legendre (1996) had a different idea of the small island effect; they expected small fragments to have steeper species-area curves because they are prone to lose species more rapidly than are larger fragments. These two views may be reconciled of one considers the former to be the case in a system which is stable or accreting species, while the latter is the case in a system which is losing species afier fragmentation. The applicability of the idea of the small island effect to plant species in forest fragments is questionable. When studies of fragmentation look at species on islands or animals in forest fragments, it is possible to have zero Species richness in non-zero area. An island might support no Species (or no species in the target taxon) and a forest fragment might lack the group of animals upon which a study focused. However, if a study focuses on the plant species composition of a forest fragment (as this one did) it is impossible to have zero species richness in non-zero area. Once there is something which can be called a forest fragment, at least one tree species must be present. In fact, plant species are likely to be present even in the absence of trees — thus, fragments could be considered to have non—zero species richness when the fragment size is equal to zero. One aspect of the construction of Species-area curves that is rarely addressed is the fact that there are really two types of species-area curves (but see Hubbell 2001). Studies that look at an archipelago of islands (or fragments) tend to construct Species-area curves with each island as a separate point in the regression. Each island is an independent sample, and only the species richness (not the species composition) is considered. On the other hand, when species-area curves are constructed within a single 103 patch (e.g., collectors curves) points are often nested or averaged. He et al. (1996) and Plotkin et al. (2000) constructed species-area curves based on subsets of a larger 50-ha plot. In some cases average species richness per plot is calculated for each Size-class, while in other cases plots are random-ordered and species-area curves are calculated through Monte Carlo re-sampling. In yet other cases, simple collectors’ curves are employed: the species-area curve is constructed on the basis of the order in which plots are collected. These two types of species-area curves may be termed cumulative or nested Species-area curves (in the latter case) and non-nested curves (in the former case). There are fundamental differences (both biological and statistical) between these two types of Species-area curves, yet I have never seen this difference acknowledged, let alone seen it actually investigated. The degree to which a system is nested (see Chapter 4) will determine the degree of divergence between these two types of species-area Clll'VCS. Objectives 1) To explore the relationship between plant species richness and area in studied dry forest fragments; 2) To compare the effectiveness of a power function (Arrhenius equation) and a sigmoid function (the Hills]ope function) as suggested by Lomolino (2000) as models of the Species-area curve; 3) To investigate the relationship between the per-plot Species richness (“species density” sensu Whittaker, 1975) and total species richness in studied dry forest fragments. 104 Methods Data Collection Data on vegetation structure, species composition and species abundance were collected from a total of 40 dry forest fragments. Species compositions were recorded for a total of 380 25-m2 plots in 35 of the fragments and 19 25-m2 plots in three reference communities in Guénica Forest. The selection of study sites and the methods of data collection are described in Chapter 2. Data Analysis Inter-fragment species-area curves were modeled by means of a power function of the form S = CA2 (Preston 1962, MacArthur and Wilson 1967) where S represents the Species richness of the fragment, A represents the area of the fragment in hectares and c and z are fitted parameters. The parameters were estimated by maximum likelihood nonlinear regression of the untransformed data. To examine the effects of unequal levels of error (based on the fact that species lists would be more complete in smaller sites, as a consequence of the sampling method) and to look for discontinuities in the underlying species-area relationship, regressions were calculated for subsets of the fragment array: < 100 ha, < 10 ha, < 5 ha, < 1 ha, < 0.5 ha and for the fragments and Guanica Forest combined. Separate analyses were also carried out on the basis of fragment history (see Chapter 3 for a description of the land- use history of each of the fragments over the period 1936-1993). 105 Additional species-area curves were fitted using the Hills.ope function (Lomolino zoooy (Log (A50 / Area)) S = Smax/ [l + (HiHSIOpe )] where Smax is the maximum species richness or asymptote, Hills.ope is a direct measure of the slope of the curve through the inflexion point, A50 is the area yielding a species richness that is half the maximum value and Area is the fragment area. The curves were fit with Smax unfixed and with Smax = 393 (the actual number of species recorded in the array of fragments) and Smax = 650 (the total number of Species recorded for Guénica Forest according to Figueroa Colon (1996). Curves were fit for the fragments alone and for the fragments and Guénica Forest combined. The relationship between the number of species per 25-m2 plot (‘species density’ sensu Whittaker, 1975) for the 35 fragments for which inventories based on 25-m2 plots were carried out (see Chapter 2 for details of the data collected in each fragment) and fragment species richness, fragment area and fragment history (see Chapter 3 for details of the history of these forest fragments) was investigated using analysis of variance (ANOVA) Intra-fragment species-area curves (species accumulation curves) were constructed for each of the 35 fragments for which inventories based on 25-m2 plots were carried out (see Chapter 3 for details of the data collected in each fragment) and for the three reference communities in Guanica Forest. Species-area curves were created using a re-sampling method: in each fragment, the curves were constructed by random-ordering the samples. Five sets of random-ordered plots were combined to make a Single species- 106 area curve. Curves of the form S = CA2 were fitted using non-linear regression. All analyses were carried out using Systat 9 (SPSS Inc., 1998). Results Initial examination of the data suggested that there might be confounding in the data set between fragment size and history, Since most of the Regrowth fragments are small and most of the large fragments are Relict. However, examination of a General Linear Model analysis of the relationship between Species Richness, Fragment Area and Fragment History revealed that there was no significant interaction between Area and History in the linearized (log-transformed) data (Table 5.1a). When the complete model was analyzed, History was not Significant (p<0.085), but when the (nonsignificant) interaction term was omitted, Fragment History was a Significant predictor of Log Fragment Species Richness (Table 5.1b). Table 5.1: Results for a General Linear Model analysis of the relationship between fragment species richness, fragment area and fragment history (see text for definitions of the terms). a. Analysis of the complete model Source Sums-of-Squares df Mean-Square F -ratio P Fragment History 1.352 1 1.352 17.389 0.085 Log Fragment Area 0.414 2 0.207 2.662 < 0.001 StatuS*Log 0.281 2 0.141 1.809 0.180 Fragment Area Error 2.847 32 0.078 107 Table 5.1: (continued) b. Analysis of the model without the non—Significant interaction term Source Sums-of-Squares df Mean-Square F-ratio P Fragment History 0.722 2 0.361 4.437 0.019 Log Fragment Area 2.580 1 2.580 31.680 < 0.001 Error 2.768 34 0.081 E 200311] 111111111 111111111 711111111 11111 a) O Eb 160 ‘ (U a 120 '- L. 8. U, 80 r G) '8 0) O. V) ‘5’ 40 s E 45 .0 112=0.715 g 1 g o 2 W 0.01 0.1 1 10 100 1000 Fragment Area (ha) Figure 5.1: Species-area curve for Puerto Rican dry forest fragments. Note log scale. Species richness increased Significantly with fragment area. The power function was able to account for 71.5% of the variability in the relationship between species 108 richness and area (Figure 5.1). The estimates of the parameters c and 2 were 69.3 and 0.127 for all fragments combines (Table 5.2). Parameter estimates based on a subset of the data are given in Table 5.2 below. Table 5.2: The values of parameters c and z (and standard errors of the estimates) and the proportion of variance explained by the regression (R2) obtained from species-area regressions for all fragments and various subsets of dry forest fragments, southwestern Puerto Rico. c 2 R7 All Fragments 69.3 (3.6) 0.127 (0.013) 0.715 Relict Fragments 76.5 (5.6) 0.116 (0.019) 0.654 Fragments < 100 ha 68.6 (3.5) 0.143 (0.024) 0.557 Fragments < 10 ha 68.3 (3.4) 0.145 (0.031) 0.499 Fragments < 5 ha 70.3 (3.6) 0.166 (0.035) 0.547 Fragments < 1 ha 82.2 (11.3) 0.215 (0.059) 0.570 Fragments < 0.5 ha 94.3 (17.8) 0.263 (0.077) 0.572 Fragments + Guanica 51.4 (9.2) 0.259 (0.027) 0.689 Overall a sigmoid function yielded a poorer fit than did a power function. When Smax was not specified, the curve-fitting algorithm failed to converge — without “forcing” a fit, the sigmoid curve did not fit the data. If Smax was specified at 393, the explanatory power of the equation was high (0.727) for the fragments alone (Table 5.3). When Smax was set at 650, the curve-fitting algorithm failed to converge for the fragments alone, but was able to explain over 60% of the variance observed for the fragments and Guanica Forest combined (Table 5.3). The area in which half the species would be present (A50) varied greatly among estimates, ranging from 158 ha when Smx was set at 393 (for the fragments and Guanica Forest) to 7331 ha when Smax was set at 393 and the fragments alone were considered. 109 Table 5.3: Parameter estimates obtained using the sigmoid Hills]ope function (Lomolino 2000) for an array of Puerto Rican dry forest fragments alone and together with Guanica Forest, a 4000-ha reserve. Fragments Fragments + Guanica Forest Smax=393 Hillsmpe 1.188 1.408 A50 7330.9 158.3 112 0.727 0.540 Smax=650 11111Slope -- 1.359 A50 -- 2255.2 112 -- 0.603 Intra-fragment species-area curves also displayed an increase in Species richness with area sampled in a pattern that was consistent with the power function. Estimates of the parameter c ranged from 2.64 to 10.92 (the lowest estimate, 2.03, was not statistically Significant). Estimates of the parameter 2 ranged from 0.302 to 0.595 (Table 5.4). There was a positive correlation between the parameters c and 2 (Figure 5.2). Species density (based on constant plot size, sensu Whittaker, 1975) was not significantly related to either fragment area or species richness, but was significantly related to fragment history (Table 5.5). There was also a significant interaction between fragment Species richness and history. Table 5.5: Least square mean estimates of species density (sensu Whittaker, 1975) based on fragment history in Puerto Rican dry forest fragments. History Adjusted Least Square Mean of Standard Error Number of Plots Species Density MixedT 30.36 1.96 41 ‘ Regroth 24.70 1.41 76 Relict'i‘ 20.69 0.44 230 'l’ Fragments that have been forested continuously since 1936 are called Relict, those that have been non-forest at some point since 1936 are called Regrowth, and those that are a mosaic of both types are called Mixed. See text for a more complete explanation. 110 Table 5.4: Estimates of the parameters c and z of the species-area curve for intra- fragment sample curves of dry forest fragments, southwest Puerto Rico. Site c z 1 8.98 0.418 2 5.42 0.435 3 6.76 0.414 4 6.23 0.443 5 6.93 0.392 6 10.68 0.328 7 9.56 0.389 8 7.97 0.393 9 2.64 0.302 10 9.08 0.384 11 8.04 0.401 12 8.38 0.405 13 5.37 0.465 14 6.00 0.479 15 8.15 0.366 16 3.65 0.461 17 3.79 0.449 18 10.65 0.359 19 8.46 0.337 20 3.43 0.486 21 7.69 0.384 22 4.07 0.440 23 5.78 0.432 24 2.86 0.455 25 10.92 0.353 26 7.15 . 0.406 27 3.57 0.538 28 3.66 0.538 29 203* 0.595 30 9.37 0.381 33 2.20 0.552 34 5.34 0.422 35 9.33 0.316 36 5.82 0.466 41 4.23 0.514 SEvT 2.84 0.560 DeF’r 9.34 0.304 ScFT 1.94 0.452 * estimate not significant (p < 0.0816). 1' Reference plots within Guanica Forest: SEv = Semi-Evergreen Forest, DeF = Deciduous Forest, ScF = Scrub Forest. 111 ‘1 A relict + Regrowth .1 16 . , XA X mlxed 8 4 +11 9 Guanica + V unknown A + C 1‘: pm 4 + ' I fi+ ‘ ‘ + e o + + 0.0 0.2 0.4 0.6 0.8 1.0 Z Figure 5.2: Estimates of the fitted parameters c and z of intra-fragment Species-area curves of dry forest fragments in southwestern Puerto Rico. Species-area curves were modeled using the power function S = cAz. Correlation coefficient r = 0.564 for all data, and 0.776 if the outlier, Site 9, is excluded. 112 Discussion The evidence suggests that, for fragments larger than 1 ha, the relationship between species richness and fragment area was fairly consistent across the range of subsets of the fragment array. Although a small island effect (SIE) was observed for fragments smaller than 1 ha, in contrast to the expectations of the SIE (which predicts a fairly flat species-area relationship in very small sites; see Lomolino 2000), the slope of the species-area curve was steeper among the small fragments than it was for the full array of fragments. On the surface this pattern among the small fragments is more in keeping with the predictions of He and Legendre (1996) that smaller fragments will have a steeper slope because they are prone to lose species more quickly. This will result in a steeper Species- area curve. Closer examination of the data fails to support this interpretation. While the slope parameter (2) is steeper for the small fragments (those < 1 ha in Size) than it is for the system as a whole, c is also larger for these small fragments. Thus, it would appear that the small fragments have more species per unit area than does the system as a whole. It is reasonable, because of historical factors, to expect the smallest fragments to have a greater species density than larger ones. In the case of Regrowth fragments, since successional dry forests on abandoned agricultural land are species-poor (Molina Colén 1998) and there was no significant relationship between species richness and area, it would appear that second growth adds species slowly. The basic set of species present in Regrowth fragments is already present in small fragments. Small Relict fragments also appear to have the basic components of the mature dry forest communities (see Chapter 7). Large areas of species-poor Regrth may cause the slope of the species-area curve 113 to be flatter than it would otherwise be. Comparable phenomena are absent from small fragments regardless of their history. Relict fragments had a flatter species-area curve than did the other subsets of fragments. The z-value, 0.116 is close to the value of 0.1 that is expected for samples drawn from continuous habitat (Rosenzweig 1995) and the c-value of 76.5 is second only to that of the group of fragments smaller than 0.5 ha. This may be the consequence of either less “relaxation” in Relict fragments, or undersampling in large fragments. Relict fragments may have fewer species and thus may resemble samples taken from continuous habitat. Alternatively, the flatness of the curve could be explained by under-sampling in the largest fragments, which could reduce the overall rate at which species are added, and so depress the entire regression. Including Guanica Forest in the regression led to a large increase in the estimate of the Slope parameter 2 from 0.127 to 0.259, and a decrease in the estimate of the parameter c, from 69.3 to 51.4. The estimate of z is almost identical with the ‘canonical’ value of 0.26 (Preston 1962). This suggests that the overall Species-area curve is a sigmoid (which seems to be the overall consensus on the true form of species-area curves; Leitner and Rosenzweig 1997, Rosenzweig and Ziv 1999, Lomolino 2000, Lomolino and Weiser 2001). Small fragments (those below 1 ha in area) and large fragments have steeper species-area curves than do intermediate-sized areas. Despite this fact, use of the sigmoidal Hillsmpe function as suggested by Lomolino (2000) did not improve the overall regression. When used in the basic three-parameter form, the curve-fitting algorithm failed to converge and a regression could not be fitted to the data. Using the published species richness of 650 for Guanica Forest (Figueroa Colon 114 1996) or the observed Species total of 393 are both unsatisfactory since neither figure reflects the total potential species pool. There are species found within the dry forest zone (and which are thus part of the total species pool) which are not present in Guanica Forest (see Chapter 7) and there are species recorded for Guanica Forest that are not part of the species-pool available to the fragment; these include mangrove species (fragments were, by definition, non-mangrove) and the more mesic species which have been recorded in Sink holes (F amsworth 1993), a habitat that was not sampled in any of the fragments. The intra-fragment species-area curves were steeper than those usually observed for samples drawn from continuous habitat. Several explanations may be proposed to address this issue. Species-area curves were constructed from scattered, rather than contiguous, plots. It has been shown (e. g., Rosenzweig 1995) that scattered plots tend to accumulate species more quickly than do contiguous plots. This is a reasonable expectation given that dry forest trees are clumped (Hubbell 1979); plots are expected to Show spatial autocorrelation. Thus, scattered plots are more likely to pick up new species than are contiguous plots. In addition, these are very small samples. He and Legendre (1996) Showed that small samples behaved differently, and yielded different estimates of the parameters c and 2 than did larger plots (when both were drawn from the same data). Leitner and Rosenzweig (1997) found a positive relationship between c and z in simulated data sets. However, this relationship was found among samples drawn from species pools of different sizes. This was probably a consequence of the fact that both c and z scaled positively with the size of the source pool. 115 The observed pattern does not fit into most of the existent theory regarding the construction of communities with regards to species richness and species accumulation. As Whittaker et al. (2001) pointed out, if a factor does not vary consistently with species richness using equal-sized plots, then that factor cannot be a driver of species diversity in that context. By extension, factors that drive species richness should be observable at a local scale. The observed patterns of c- and z-values could be best described as being ‘complementary’. Sites that have higher c-values have lower Slope parameters (2), while sites that have lower c-values have higher z—values. This pattern means that fragments either had high species richness or that species were rapidly added between plots — in other words, fragments either had high (it-diversity or high B-diversity (sensu Whittaker 1975), not both. This is a reasonable conclusion if these fragments draw upon a pool of common species that account for the majority of all individuals. For the most part, diversity is driven by rare species, while sampling is most likely to encounter common species. Collectors curves compiled during the course of data collection appeared to level well before all species in the fragment were sampled. Murphy and Lugo (1986) observed a similar situation in Guanica Forest where their species-area curve approached a plateau after only 34 tree species were recorded. Scheiner et al. (2000) discussed the importance of knowing the form of species- area curves through space -— whether the curves are parallel or they intersect at some point has a major influence on the relationship between species richness and productivity, and is likely to affect other such relationships. Many of these curves intersect even within the 116 area sampled. Thus, correlates with Species richness are unlikely to be scale invariant, and functional relationships may well differ among fragments. No relationship was observed between per-plot species richness (species density, sensu Whittaker, 1975) and either fragment area or species richness. The absence of such a relationship suggests that fragment species richness is not a function of species density. Species richness at the plot level is a function of fragment history. Examination of the least square means shows that Relict fragments were the most species poor (on a per plot basis) and Mixed fragments were the most Species-rich. The difference in effect size between these groups is large. Mixed fragments had almost 50% more species per plot than did Relict fragments, despite the fact that Relict fragments on average had the highest species richness. This differs substantially from what Ross et al. (2002) found in Australian fi'agments. Their study was one of the few to look at Species density (sensu Whittaker, 1975) in forest fragments of varying age. Contrary to the findings of this study, Ross et al. (2002) found that species density declined as fragments aged, especially when they were subject to disturbance (mostly fire) and invasion by exotic species. This relationship suggests that disturbance is the real driver of species density (sensu Whittaker, 1975). This agrees with the predictions of the intermediate disturbance hypothesis (Connell 1978) and with Dunevitz's (1985) findings in Guanica Forest. As long as rootstocks remain in place, dry forest recovers rapidly from cutting, and the species that were present before cutting remain dominant (Ewel 1980, Molina Colon 1998). Once they have been eliminated, succession tends to be much slower and involves species that can establish from seed. These two processes involve different source pools — one, the trees that persist (Bond and Midgley 2001) and the other, species that establish 117 from seed. In Mixed fragments, both source pools are present and plots are thus likely to be more species-rich. Although it may be suggested that Relict fragments are more species-poor because of the inclusion of several ‘coastal’ fragments that have very few species per plot, these account for only a small proportion of all Relict fragments and omitting them does not materially alter the results. Taken together, these findings suggested that local (plot) species richness was not the driver of fragment species richness. Samples taken from continuous habitat yield species-area curves with flatter slopes than do similarly sized islands. Rosenzweig (1995) suggested than this was due to the presence of “sink species” in the samples drawn from continuous habitat. “Sink species” are species that are present in the sample because they are present in the large area of habitat that the continuous habitat provides; they would not be able to persist in an island the Size of the area sampled from the continuous habitat. These species are only present in larger islands. Species richness increases more quickly with island area than it does in similarly sized mainland samples because certain species will be present in samples that are smaller than the minimum island size in which they can persist. If this is the case, then the species density should be higher in larger Sites, so as to compensate for the flatter Slope of the Species-area curve, a process comparable to the ‘mass effect’ of Shmida and Wilson (1985). In this system, fragment species richness appears to be a function of turnover between plots (Ii-diversity, sensu Whittaker, 1975) and not of ‘point’ species richness (a- diversity, sensu Whittaker, 1975). Whether this is caused by resource heterogeneity, negative density dependence in recruitment or disturbance history is not something that 118 this study can discern, but it is clear that heterogeneity and not local species richness is the main driver of species richness. On the other hand, is there any a priori reason to assume that species density Should be a predictor of fragment species richness? Hubbell and colleagues (Hubbell et al. 1999, Hubbell 2001) suggested that local factors do not drive species richness. Instead, it was suggested that dispersal limitation is one of the key factors in the maintenance of diversity in tropical forest communities. Overall species richness (what Hubbell, 2001 calls ‘metacommunity’ species richness) is driven by local-scale differences in Species composition. If this is the case, then Whittaker et al.'s (2001) assumption that plot species richness Should drive site Species richness may be questionable. Summary 1) Fragment species richness was a function of fragment area and fragment history. 2) Inter-fragment species-area curves were better fit by a power function (the Arrhenius equation) than by a sigmoid function (the Hills]0pc function). 3) A small-island effect (SIE) was observable among fragments smaller than 1 ha. 4) Inclusion of Guanica Forest into the species-area curve altered the parameters of the relationship but did not worsen the fit of the relationship; Guanica Forest has a larger range of habitat-types than did the fragments, and so would be expected to have a higher fl-diversity than did the fragments. 5) The Arrhenius function explained 71.5% of the variance in the overall inter-fragment species-area curve and 65.4% of the variance in the Relict fragment species-area 119 6) 7) curve; Regrth fragments did not Show a significant relationship between species- richness and area using either a power function or a linear regression. Mixed fragments had the highest species density (sensu Whittaker, 1975) and Relict fragments had the lowest; there was no relationship between species density and fragment species richness. There was a significant negative correlation between the parameters of the Arrhenius equation (c and z) in intra-fragment species-area curves. 120 CHAPTER 6: PLANT SPECIES RESPONSES TO LON G-TERM FRAGMENTATION IN PUERTO RICAN DRY FOREST LANDSCAPE Introduction In studies of habitat fragmentation, species richness and diversity indices are among the primary descriptors of community patterns. However well these summary patterns describe the patterns of species distribution, they remain summary patterns, and can hide as much as they reveal. Community patterns are made up of individuals and species. The way these species (and the individuals that make up these species) distribute themselves on the landscape is what structures a community and drives community dynamics. The existence of a trade-off between competitive ability and dispersal ability among plants is one of the basic features of models which attempt to explain coexistence in competitive communities. Species that are better competitors are likely to be locally dominant. Hubbell (2001) has shown that, in general, space is a limiting resource that is fully used. For a new individual to establish itself, a space must become available. Poorer competitors are able to survive in the community by being better dispersers. When a Space becomes available, a superior disperser is more likely to find that space and become established. Jennings et al. (2001) and Vandermeer et al. (2001) have suggested that most competition among trees occurs at the seedling stage — it is difficult for a seedling to displace an adult tree regardless of its competitive advantage. Models have shown (Chesson and Warner 1981, Chesson 1986) that a trade-off between competitive ability and dispersal ability is adequate to allow coexistence over long 121 periods of time. Hubbell (2001) has shown that even without that trade-off, the simple fact of dispersal limitation can allow for extremely long extinction times in local communities comprising a few thousand individuals. Tilman and colleagues (Tilman et a1. 1994, Tilman et al. 1997) have shown that, given the existence of a competition-colonization trade-off, dominant species are at risk in fragmented landscapes because they are less able to recolonize fragments from which they have gone extinct. Surprisingly, these models predict that weedy species will gain an advantage in a fragmented landscape even in the absence of firrther disturbance within the fragments. The relationship between abundance and range is one of the fundamental relationships in macroecology (Gaston et al. 1997). Locally abundant species tend to have wider geographic ranges than do less abundant species. On local scales, where dispersal limitation is not likely to be a major factor, local dominance hierarchies are likely to be repeated across the landscape (Hubbell 2001). Thus, locally abundant species are likely to be superior competitors and are likely to be more widely dispersed geographically. However, because dominant species are likely to be relatively poor dispersers, they are less likely to recolonize disturbed areas. Seed dispersal characteristics are likely to be important with regards to what species are able to colonize regrowth. If other things (such as dispersal syndrome) are controlled for, one would expect that small seeds would have a higher probability of being dispersed into a new site than would large seeds. On the other hand, large seeds are likely to have more reserves, which may be useful in establishing in new habitats. This may be especially relevant in seasonally dry areas, since seedlings need to have 122 access to soil moisture in order to survive the dry season. Leishman and Westoby (1994) found that large seeds had an advantage in establishing under conditions of low soil moisture — it seems probable that this would also be true for surviving low soil moisture in the first dry season to which a seedling is exposed. Objectives 1) 2) 3) 4) 5) 6) To determine whether there is a correlation between abundance of species in the reference community (Guanica Forest) and their geographic range; To determine whether there is a relationship between local abundance in Guanica Forest and frequency in sample plots within Guanica Forest; To determine whether there is a relationship between local abundance within Guanica Forest and the number of fragments within which a species occurs; To determine whether there are differences in the distribution of species that are locally abundant in Guanica Forest and species that are present in most of the fragments in terms of the drivers of their presence in fragments of differing species richness and history; To determine whether there are differences in seed size among fragments with different disturbance histories; To determine whether there is a difference in the abundance of exotic species among fragments with different disturbance histories. 123 Methods Inventories were carried out in a total of 39 dry forest fragments as outlined in Chapter 2. Species recorded in a total of 19 25-m2 plots in Guanica Forest were used to examine differences in the distribution of forest species in continuous and fragmented Puerto Rican dry forest. Each species present was assigned a Range score based on its biogeographic distribution (Liogier 1985, 1988, 1994, 1995, 1997; see Table 6.1). Table 6.1: Criteria used to assign a Range score to Puerto Rican dry forest plant species Score Range 1 Endemic to Puerto Rico and adjacent islands. 2 Puerto Rico plus Hispaniola and the Virgin Islands. 3 Insular Caribbean (including the Bahamas and offshore islands administered by Venezuela). 4 Caribbean and either Florida, Central American or South America. 5 Tropical America (present in both Central American and South America). 6 Pantropical or extra-tropical. Mean abundance was calculated for each species sampled in Guanica Forest on the basis of plots where the species was present, not from total area sampled. Frequency was calculated as the number of plots in Guanica Forest where the Species was present. Incidence was calculated as the number of fragments where the Species was present. Distribution profiles were constructed for each of the five most abundant species and the five Species with the highest Incidence with presence or absence plotted against species richness. LOWESS methods (Cleveland 1979, Cleveland and Devlin 1988) were used to fit the curves. Seed mass (in mg) was obtained from Castilleja (1991) for each of 45 common tree species. These were grouped into three categories: small (0.1-20 mg), medium (20- 124 60 mg) and large (60-473 mg) seeds. The proportion of small, medium and large seeded Species was compared among fragments by site history. For each of these 45 species, the number of fragments in which it was present (Incidence) was graphed against seed mass. The mean number of individuals of exotic Species per 25-m2 plot was calculated on a per-site basis. All grasses except Lasiascis divaricata (L.) Hitch. were included, since most grasses were lumped into three morphospecies and were not identified to species. Not all grasses are exotic, but the majority of pasture grasses are exotics, and these include almost all grasses likely to be found in the fragments except L. divaricata. Results Range-Abundance-lncidence Patterns Frequency-Abundance There is a positive relationship between Frequency (the number of reference plots in Guanica Forest in which a species occurred) and Mean Abundance (mean number of individuals per reference plot in Guanica Forest in which a species occurred; Pearson correlation = 0.678; a significant positive exponential relationship R2 = 0.738), but this pattern is largely a consequence of three species (Gymnanthes lucida, Croton humilis L., and C. discolor Willd.) which had both high frequencies and very high mean abundances (Figure 6.1). 125 coho vangm ~ 8 vengeance :88 we pox—omega 08 Sam .86on 828 5 SEE 8.62m com 958% mm 36on 05 98:3 898m 8330 E 83 8 mm com £3233: mo confide 508 23 consequences :82 98 @580 36on e. 80:? 820m 83:0 E $03 no 39835 >28:on 502509 9:82:28 BE. #6 0.53m 5:259:— om w. 3 E .2 E w o v m o S. e m- 0 o 1 o m a o a o m c w 0 O i m m u 0 T S V a. o , a m. T om m a 1 mm - om mm 126 .820 23288 ~ 8 859598 :38 me @2583 2a Sam .86on 828 be 58E 0225 How 83.8328 05 .«o wagon. 05 com _.o 2an com 82:552.”? 86on 05 .«o 532: oEmeuwoowoB 28 owned 28 Geneva mm 86on 05 20:3 820m «05:0 5 SE «8 mm can “213232: me 238:: 52: 23 oo§e§n< 522 503:5 2:88:22 2F an.» 9.53 own—«m o w v m N ~ o 5 OT 1 WI 1 o m . . m I r m m I 2 w n T m~ W. I m r om m e mN t Om mm 127 Range-Abunda_ng§ There was no statistically significant difference among the six classes (ANOVA, 45 d.f., p < 0.543). Species in Range class 4 (species found in the Caribbean and either Central America, South America or Florida) had the range of abundances in the reference plots in Guanica Forest (Figure 6.2). Incidence-Abundance There was little overall trend of abundance with increasing incidence. The species with the highest mean abundances were present in an intermediate number of fragments (Figure 6.3). Species Profiles The five species with the highest mean abundance in Guanica Forest (based on those reference plots in which they were present) were: Gymnanthes lucida, Croton humilis, Eugenia foetida Pers., C. discolor and Erithalisfiuticosa L. LOWESS regressions yielded similarly shaped profiles for four of these five species; the curves had an intermediate peak at fragments of a species richness of about 60 species and then declined before increasing again (Figures 6.4a-d). The fifth species, Erithalisfruticosa had a monotonically increasing distribution profile, but was only predicted to occur in fragments with over 100 species (Figure 6.4e). The five species with the highest incidences were Bourreria succulenta, Distictis lactiflora (V ahl) DC., Lantana involucrata, Pilosocereus royenii, and Stigmaphyllon emarginatum (Cav.) A. J uss. Their distribution profiles were approximately flat. The LOWESS regression predicted that these species would be present in all fragments (Figure 6.4f-j). 128 .36on $28 be 88E 0:25 How 958a mm 86on 2a 82? $0.8m 85:0 5 83 «8 mm “on mfiszwE .«o “2:58 988 23 85222 5.62 new Gunman mm 86on a £033 E 35895 mo 89.85: 2.3 35205 5253 @230322 BC. and Bawa ow 8522: om 2 S m o w IIIIILiIIli LIP _ C I -- II — III In I I I - m I I I I - 2 w 9 m 7 WM V I a. - cm m u I 1 mm a I - cm mm 129 1.2 l l I l 1.0 .O 00 T Presence 9 N l 0.0 l 0 50 100 150 200 Species Richness a: Abundance profile for Gymnanthes lucida. 1.2 LOP 0.8 * Presence l 0.4 0.2 *- 0.o - 1 o 50 100 150 200 Species Richness b: Abundance profile for Croton humilis. Figure 6.4: Abundance Profiles (probability of the species being present as a function of fragment species richness) for dry forest species with the highest abundance in reference plots in Guanica Forest (a-e) or the highest incidence among the fragments (f-j). 130 1.2 1.0 r '- 8 0.8 — 4 § <1) __ a 5: 0.6 0.4 r - 0.2 r r 0.0 - l 0 50 100 150 200 Species Richness c: Abundance profile for Eugenia/oetida. 1.2 I I I no >— 0 C .0. O O O O - 8 _.I c: O a - IL 4 l ' 0 50 100 150 200 Species Richness d: Abundance profile for Croton discolor. Figure 6.4 (continued). 131 1.2 I I I 1.0— a... o e e — Presence ' O 50 100 150 200 Species Richness e: Abundance profile for Erithalisfi-uticosa. 1.2 I I I .—n O I 1 Presence .O .0 .O A ex oo 1 l T l l l g: N I l 4:1: 1 1 50 100 1 50 200 Species Richness .9 (Dc: 1': Abundance profile for Bourreria succulenta. Figure 6.4 (continued). 132 1.2 I T I 1.0— W — l l 0.8 Presence 0.2 — ‘i 00 $1.- 1 l 'o ?o' 100 150 200 Species Richness g: Abundance profile for Distictis lactiflora. 1.2 I I I 1.0 0.8 *— I I 1 Presence 0.6 l L 0.4 l l 0.2 00 "I = l 1 ° 0 50 100 150 200 Species Richness h: Abundance profile for Lantana involucrata. Figure 6.4 (continued) 133 1.2 I I I l 0 *— W -< 8 0.8 *- — c: 8 CD _. a: 0.6 r 0.4 ~ — 0.2 r 4 0.0 “ l = ‘ l L O 50 100 150 200 Species Richness i: Abundance profile for Pilosocereus royenii. 1.2 I I I 1.0 >- W -‘ 8 0.8 r - c: 3 Q) __ —I a: 0.6 0.4 r '- 0.2 *— - 0 0 —O* 1 c 1 1 ' O 50 100 150 200 Species Richness j: Abundance profile for Stigmaphyllon emarginatum. Figure 6.4 (continued). 134 Relative Abundance Profiles Four of the five species with the highest abundances in Guanica Forest had small (0.21- 0.59) positive Pearson product-moment correlations between their relative abundance in fragments and the species richness of the fragments. Erithalisfiuticosa had a correlation of —0.06. The five species with highest incidences among fragments had small negative correlations (-0.08 to —0.14) between their relative abundance and the fragment species richness (Table 6.2). Table 6.2: Pearson correlations between Abundance (among fragments) and Fragment Species Richness for each of 10 dry forest species in southwestern Puerto Rico. Species Correlation with Species Richness High Abundance Gymnanthes lucida 0.227 Croton humilis - 0.585 Eugenia foetida 0.428 Croton discolor 0.21 1 Erithalisfruticosa -0.064 High Incidence Bourreria succulenta -0. 144 Distictis lactiflora -0.084 Stigmaphyllon emarginatum -0.127 Lantana involucrata -0.142 Pilosocereus royem'i -0.098 Site History Four of the five species with the highest relative abundance in Guénica Forest showed a significant relationship between Incidence and fragment history (Table 6.3). Erithalisfiuticosa was the sole exception. All of these species had a higher probability of occurrence in Relict than Regrowth fragments (Table 6.3). Four of the five species with the highest Incidences showed a significant relationship between Incidence and fragment history (Table 6.3). The sole exception was 135 m. 5 is 53 God assesses remainsmem 36 2: 23 :3 use? seeasi 33 is :35 ~86 assesses 33:3 new 3% :3 22V asses sense m. a $5 33 God 823383 esteem 8822: a5 3 N8 83 2: :2 uses $55353 n2 «.8 god a? anemoxszemé 92 new and 83 emoeestfiasmtm 3 m8 mood 5.2 £53: 2220 me :5 Sod v 8&2 seas €36 855.? new oocogooo mo b23303 082.586 mo mum—530a ”meaofiwflm 5383M “flcofiwfim 820m tom—Benoa nx :omHaom 86on SE 8.55 5833538 5 bog: «aofiwmb cam momooam ESQ 328 be @8028 we nousnwummu 05 5053 3:80:52 at. "we oEmH. 136 Pilosocereus royenii. All species showed a higher probability of being present in Relict than Regrowth fragments (Table 6.3). The Incidence of Distictis lactiflora showed almost no difference at all across fragment histories (p < 0.811), while Lantana involucrata showed the largest non-significant difference (p < 0.081). Seed Mass Most of the species in Relict and Mixed fragments were small seeded (0.1-20 mg). The number of species in each size class differed among fragments with different histories. Per fragment species totals differed with history for small seeded species (ANOVA, 36 d.f., p < 0.0005, R2 = 0.543), medium seeded species (ANOVA, 36 d.f., p < 0.0005, R2 = 0.431) and large seeded species (ANOVA, 36 d.f., p < 0.001, R2 = 0.321). There was a high degree of collinearity between the distribution of the three seed sizes (small vs. medium, Person correlation r = 0.809, small vs. large, Pearson correlation r = 0.814, medium vs. large, Pearson correlation r = 0.814). As a result of this, use of an overall MANOVA was unwarranted. Overall this suggested that the real difference between sites was a function of overall species richness, which was supported by the fact that Relict and Mixed sites had very similar patterns, while the pattern in Regrowth fragments was different (Figure 6.5). Since there were approximately twice as many small-seeded species as medium- or large-seeded species (22 small, 12 medium and 11 large seeded species), the pattern in the Relict and Mixed fragments did not differ from the expectation that the distribution of seed sizes among fragments was a random sample from the pool of species being tested (x2 test, 3 d.f., a = 0.05). The pattern in the Regrowth fragments was significantly l37 coho @838», H H Eofiwam 5m 360% we 598:: S38 mm Emmoa 2m Sam .onBmE 8:3:sz #:20wa SE $583M“ 520% be 58E otosm E £58 38 mnvéov emu: Ho Awe owiomv 83on .38 omédv :95 5? 860% mo Saga 2:. "m6 95E..— gokwom @852 620M o - m m B u N n m m 9 u I: S d m. II n E 35:93 628 be 883 otosm E 83 ~8-mm com 860% 385 we “BASE 508 05. ”he nun-ME Essex Biz 626m o i. H S W .— u 3 X . - om m. 4 is 3 S d .8 m. m v a. . 9. m D. B u a gr I om ow 140 different from expectations. There were fewer small seeded species and more large- seeded species than would be expected by chance (x2 test, 3 d.f., a = 0.05). There was no apparent relationship between seed mass and Incidence (Figure 6.6). Exotic Species Abundance The abundance of exotic species (expressed in terms of the number of individuals per 25 m2 plot) was higher in Regrowth fragments (36.1 :t 14.5 individuals per 25 m2) and was lower in Mixed (16.0 d: 6.0 individuals per 25 m2) and Relict (12.2 i 2.2 individuals per 25 m2) fragments (Figure 6.7). The overall relationship was not significant (ANOVA, 38 d.f., p < 0.214). Discussion One of the more general ecological patterns observable at large scales is that locally abundant species are more geographically widespread than locally uncommon species (Gaston et al. 1997). This pattern was not observed for these data. This may have been a consequence of the manner in which the geographical ranges were calculated. A species in Range Class 1 (Puerto Rican endemics) could conceivably occupy a larger range than a species in Range Class 2 (Puerto Rico, the Virgin Island and Hispaniola). For example, a Puerto Rican endemic that is distributed across the whole of the island (2. g. , Thouinia striata) may have a larger range geographically than a species which is present in Puerto Rico and Hispaniola but possesses a restricted distribution in both islands (e. g., Stahlia monosperma). In addition, “whole community” measures may be inappropriate. Trends 141 may have been stronger if analyses were restricted to more narrowly defined “guilds” or to more closely related taxa (e. g., single families or genera). Unlike the broader Range-Abundance pattern, the Range-Frequency pattern met these expectations, although this relationship was primarily driven by three species (Gymnanthes lucida, Croton humilis, C. discolor) with especially high mean abundances and frequencies. This pattern fits what Hubbell (2001) found in his analysis of tree species abundance in Peru; seven species showed a visible competitive advantage while the remaining species appeared competitively equivalent. On the other hand, some species, such as Bursera simaruba had a high frequency (it was present in nine out of 19 plots) but a low abundance (its abundance averaged 1.9 individuals per plot). At the other end of the spectrum Erithalisfiuticosa had very high local abundances, but was present in only three plots. Species like this probably use a strategy much like what Bolker and Pacala (1999) called ‘phalanx competition’ — they are restricted to specific resource patches, but within those patches they can be very common and are able to out- compete other species (often by clonal spread). The relationship between Abundance and Incidence also met theoretical expectations. Dominant species had intermediate Incidences. Four of the five dominant species were more likely to be present in Relict fragments than in Regrowth ones. Four of the five most widespread species showed no significant difference between their distributions in Relict, Mixed and Regrowth fragments. The species that are dominant in Guanica Forest (and to some extent in many of the Relict fragments) are absent from most of the Regrth fragments. While it is impossible to disprove the hypothesis that these dominant species were not part of the pre-fragmentation community in the 142 Regrowth fragments, there is no a priori reason to assume that this should be the case. It seems more reasonable to assume that the absence of these species from Regrowth fragments is due to their failure to recolonize these areas as they reverted to forest; this pattern is expected given a trade-off between competitive ability and colonizing ability. It is not surprising that Gymnanthes lucida is a poor colonizer given its dry, mechanically dispersed fruit (Castilleja 1991). It is more surprising that Eugenia foetida is a poor colonizer since it has small fleshy fruits. The fruit and seeds of E. foetida are smaller than those of Bursera simaruba and Guaiocum ofiicinale L.; the fruits of both of these species are readily removed by frugivorous birds (Ricart Morales 1999). It is possible that it diverts relatively little energy to seed production, as might be expected if this species fits the profile of a resprouter rather than that of a reseeder (Kruger et al. 1997), or that seed production is limited by low pollinator visitation, pollen dilution (Aizen and Feinsinger 1994a, b) or limited disperser movement across the landscape. It is also possibly a consequence of low seed viability as a consequence of inbreeding depression. The widespread species are likely to possess a suite of characters that helps them to colonize Regrowth fragments. Three of them are fleshy fruited, while the other two have winged fi'uit that are presumably wind dispersed. These species also do not appear to be forest dependent. Bourreria succulenta is an early successional species (Ray 1993, Ray and Brown 1995). Pilosocereus royenii is often found in open grassy areas (personal observation), while the vines Stigmaphyllon emarginatum and Distictis lactiflora can be found on fences and isolated trees (personal observations). The ability of these species to 143 utilize the non-forested matrix probably plays a key role in their ability to colonize Regrowth. While Regrowth fragments had a different distribution of seed sizes than did Relict and Mixed fragments, the prediction that there were more small-seeded species in Regrowth fragments was not supported. Instead, fewer small-seeded species were recorded in Regrowth fragments, possibly reflecting an advantage conferred by larger seeds in establishment in drier conditions. Seedlings establishing in abandoned pasture are likely to experience more severe dry season conditions than would seedlings establishing under tree cover. There was no clear pattern between seed mass and the number of fragments in which a species was present. The higher proportion of exotic species in Regrth relative to Relict fragments fits expectations. That the trend was not statistically significant is not surprising, since the standard errors were so large, especially for Regrowth fragments. While there is adequate light in the understory to allow germination (Castilleja 1991), it is likely that belowground competition (Coomes and Grubb 2000) will limit establishment in intact forest. It seems probable that trees will only recruit successfully when an established adult dies. The identity of the species that manages to capture an opening depends on the species that are present and able to get seeds into the opening — a gap poses no opportunity for a tree that produces no seed in the year that the gap appears. When a large area is opened for colonization and it is far from established forest patches, species that are nearby and that are able to produce large amounts of seed have an advantage in establishing (e. g., Clark et al. 1999). Weedy species in general have an advantage, and exotic species are often weedy. Thus, it is reasonable to expect high exotic species 144 abundances in Regrowth fragments. On the other hand, some exotic species are able to invade intact forest fragments; 0eceoc1ades maculata (a terrestrial Afn'can orchid) was primarily observed in ‘high quality’ Relict fragments. Conclusions Species are the drivers of community patterns. History is an important driver of the distribution of individual species. While many species appear to be able to readily colonize Regrth forest, the typical community dominants are not among them. Abandoned agricultural land and young regrowth forest are widespread on the landscape. Any species that is able to exploit this habitat is likely to be abundant on the landscape, and thus, is likely to be an important component of the seed rain into newly available habitat. This study did not address the question of seed rain and species dynamics, but it seems reasonable to conclude that widespread, readily dispersed species are not only likely to be common in disturbed areas; they are also likely to form a disproportionate amount of the seed rain into undisturbed areas (see Janzen 1983). If the dominants are unable to colonize Regrowth, it may be necessary to reintroduce them. It remains unclear as to whether Leucaena leucocephala hinders the establishment of native species or whether the native species are simply unable to disperse into regrth dominated by L. leucocephala. In the moist forest zone of Puerto Rico, L. leucocephala was the best ‘nurse crop’ for native tree species among various exotic plantation species used for reforestation (Lamb et al. 1997). If it actually inhibits the establishment of native species in the dry zone, then it is probably a function of belowground competition. On the other hand, it may simply be that seed dispersers are 145 not attracted to the dry-fruited L. leucocephala or that, as a deciduous species it fails to produce enough shade to allow seedlings to survive the dry season. Since Relict and Regrowth fragments differ in species composition, they are likely to differ in their role in the conservation of the native biota. Given these differences, it is not safe to assume that the conservation value of a fragment is a simple function of its species richness. It is important to also incorporate information about the species composition of a fragment and how well it reflects the reference community. Chapter 7 addresses these concerns. Summary 1) 2) 3) 4) 5) Species abundance correlated weakly with Range and Frequency, while the relationship with Incidence was complex. The distribution of dominant species from Gua'nica Forest (the reference community) among fragments is a function of fragment history and species richness; dominant species are predominantly present in Relict fragments. The distribution among fragments of the most widely distributed species correlated negatively with fragment species richness and was independent of fragment history (except in the case of Pilosocereus royem'i). The distribution of seed mass differed between Relict and Regrowth fragments; contrary to expectations, there were fewer small-seeded species in Regrowth fragments than in Relict fragments. Exotic species had higher abundances in Regrowth fragments than in Relict or Mixed fragments but the difference was not statistically significant. 146 CHAPTER 7: THE CONSERVATION POTENTIAL OF DRY FOREST FRAGMENTS ON A TROPICAL LANDSCAPE Introduction Conservation in Fragmented Landscapes Guanica Forest (Bosque Estatal de Guanica) occupies about 4% of the dry forest zone in southwestern Puerto Rico (Murphy et al. 1995); it is by far the largest area of protected dry forest on the island. This makes it the key resource for the conservation of dry forest biodiversity. Despite this fact, it cannot sustain the long-term survival of all dry forest species. While over 650 plant species have been recorded from Guanica Forest (Figueroa Colon 1996), several important elements of the dry forest biota are missing from this site. Of the 49 plant species formally recognized as threatened or endangered by the US. Fish and Wildlife Service, 13 have been recorded from dry forest habitats in southwestern Puerto Rico. Existing populations of only five of these species have been documented within Guanica Forest. Consequently, eight of these species depend entirely on habitat outside of Guanica Forest. Other protected areas including the Cabo Rojo National Wildlife Refuge and Laguna Cartagena National Wildlife Refuge provide critical habitat for some of these species (e. g. , Aristida chaseae and A. portoricensis; US. Fish and Wildlife Service 1994a, b), but others are entirely dependent on privately owned lands (e.g., Catesbaea melanocarpa Krug & Urban in Urban; Silander 1999), which are often under considerable development pressure. Similar patterns are likely to exist for other rare species. In addition to harboring species that may not be present in the main reserve, additional populations provide insurance against catastrophic events. While natural fires 147 are rare events in this system, fires set by human agency are not infrequent in the dry season (personal observation) and occur regularly along roadsides in parts of Guanica Forest (M. Canals Mora, personal communication). Events like this can cause the extinction of a population (or a species if it is restricted to a single site). Similarly, outbreaks of pests or pathogens tend to spread faster across contiguous populations. Populations broken into several isolates may have a better chance of surviving a disease outbreak (see Hess 1994, 1996). Breckon et al. (1998) and Breckon (2000) documented the apparent extirpation of Opuntia repens Bello (a species endemic to Puerto Rico and the Virgin Islands) from the offshore islands of Monito and Desecheo, presumably as a result of infestation by the cactus moth Cactoblastis cactorum. The same isolation that makes 0. repens unlikely to recolonize these islands may also have stopped the outbreak from spreading to other populations. Subdivided plant populations experience restricted gene flow (in the form of pollen and seed transfers among populations). This can affect the extinction probability of a small population. Inbreeding depression can reduce viability, seed production or growth rates as a consequence of the segregation of partially recessive lethal alleles. The loss of potentially adaptive variation in quantitative characters due to genetic drift can reduce the ability of the population to adapt to changing environmental conditions. The effects of new mildly deleterious mutations can accumulate and become fixed by genetic drift in small populations, thus lowering the overall viability of the population. The existence of forest fragments outside of Guanica Forest can play a role in reducing the degree of isolation experienced by populations in the main reserve. Fragments can encourage the movement of pollinators and seed dispersers across the landscape. 148 Adequate levels of seed and pollen movement among populations in Guanica Forest, the fragments, and perhaps the forests of the Cordillera Central, can ameliorate many of the aforementioned negative genetic consequences. It is important to be able to assign a value to forest fragments for two main reasons: to identify valuable habitat that may be acquired for conservation purposes or for which conservation easements may be obtained, and to be able to determine whether or not a site proposed for development is important in a conservation context. Forested lands outside of Guanica Forest are under the control of several different bodies and differ in the degree of protection afforded to them. Protected lands include public lands administered by the Puerto Rico Departamento de Recursos Naturales y Ambientales (DRNA) and the US. Fish and Wildlife Service (US FWS) and private lands under the control of the F undacion Puertoriquefia de Conservacion (Puerto Rico Conservation Foundation). While other lands, including those under the control of government agencies and those in private hands, lack formal protection, many development activities are subject to laws and regulations that require a permitting process. This allows the government some measure of control over the fate of forest fragments on lands lacking formal protection. Ideally, a conservation management plan for the whole landscape should be devised that protects enough natural habitat to ensure the long-term survival of all native species. In heavily deforested landscapes this option does not present itself; it is unlikely that enough habitat exists to ensure the survival of all species in the absence of management interventions. Instead of attempting to determine whether any given patch of habitat can ensure the long-term survival of any given species, the objective is one of 149 identifying the remaining habitat patches which will make the largest possible contributions to the survival of as much as possible of the native biota. Most modern attempts to select optimal sets of reserves are based on the complementarity principle (Vane-Wright et a1. 1991 ): the overall idea is that new reserves should be selected to bring in the maximum number of species not already present in existing reserves. The objective of determining a minimum set of reserves is that each (target) species should be present in at least a certain (predetermined) number of reserves. No method of site selection can be better than the information used in the selection process. However, since the resources employed in data collection are likely to originate from the same pool of funds that can be used for conservation it is imperative that the methods of assessing the value of potential conservation areas should maximize cost effectiveness. What constitutes a valuable fragment? While any fragment that preserves a viable population of a native species may be potentially valuable to the conservation of that species, valuable fragments should harbor viable populations of as many native species as possible, and should also preserve interactions between these species and the cycles of energy and nutrient flow through the system. More specifically there are three criteria that can be used to assess value. 1. Conservation: whether the species present are considered to be in need of protection. 2. Representativity: whether the species present are representative elements of the community from which they are (presumably) drawn. 3. Connectivity: where the site lies in proximity to other fragments, and its overall setting in the landscape (see Chapter 3). 150 Species richness is one of the most commonly used criteria for assessing conservation value (Dufréne and Legendre 1997). The use of species richness poses some problems since species richness increases with area. Thus, this can amount to simply assigning conservation on the basis of fragment area. While this is not necessarily a bad criterion, it introduces potential confounding that should be acknowledged. Species density (a measure of the number of species per unit area) would appear to compensate for the relationship between species and area. Sites that have more species per unit area than expected (“hotspots” as such) might be seen as important since they can protect more species in a given area than can less “species dense” sites. However, one must account for the fact that the relationship between species and area is non-linear (see Chapter 5 for a discussion of species-area curves). As a consequence, one must control for this in computing species density. Whittaker et al. (2001) recommends the use of fixed-area plots to measure species density because statistically normalized estimates of species richness tend to average the range of variability in the site, but it is precisely this ‘summary’ property of statistical prediction that makes this measure of species density appealing in this situation. The conservation value of a fragment is the product of these interactions. In addition, as mentioned in Chapter 5, there does not seem to be any compelling reason why overall fragment species richness should be a fimction of local species richness (at-diversity) - in fact, it is difficult to explain the species richness of Guanica Forest or any of the fragments without stressing the importance of habitat heterogeneity and B-diversity. Measures of species richness are neutral with regards to the species involved. Metrics that regard species simply as numbers (e. g. , species area curve and diversity 151 indices) mask the identity of the species involved. The presence of exotic species in a fragment can inflate the species richness and yet is likely to decrease, not increase, its conservation value (see Chapter 6). An alternative to using a measure of species richness is to compare the species composition of the fragment to some standard list of species (e. g. , Webb 1989). Dufréne and Legendre (1997) considered this a more satisfactory means of assessing conservation value. The actual species composition of the fragment is what is considered, and species not characteristic of the system (e. g., exotics) can be discounted. One weakness of this method lies in the construction of the reference list; as Dufréne and Legendre (1997) pointed out, “representativity” is a subjective concept and it requires that a “typical” or “pristine” example of the community be identified. If a reference sample can be identified, then (relatively) objective reference lists can be compiled through random sampling. This is less of a problem philosophically for adherents of the Zurich- Montpellier school, where releve selection requires the identification of a “typical” portion of the community. This method, which values sites on the basis of how typical they are, can be complemented by searching fragments for less common elements of the community. The number of rare or endangered species present in a fragment can serve as a measure of its conservation value even if the site is not a typical example of the community. Kirkpatrick and Gilfedder (1995) found that sites that contained endangered species in Tasmania were not necessarily those with the highest biological integrity. An alternate method of identifying sites of high conservation value would be to identify species that have high fidelity for valuable sites. Such indicator species could be 152 used to identify valuable fragments. Indicators are suitable tools whenever the data are too complex to handle without aggregation (Miiller et al. 2000). Indicator species can serve as ‘flags’ if vegetation descriptions are required as baseline elements in the preparation of permits for development activities. Unfortunately, the selection of indicators requires that some measure of conservation ‘value’ be made beforehand, thus making indicators sensitive to the biases inherent in the selection of the measure of ‘value’ selected. Objectives 1) To develop methods to determine the conservation value of Puerto Rican dry forest fragments; 2) To evaluate the conservation potential of the studied dry forest fragments; 3) To designate and evaluate indicators of high quality dry forest fragments that can be used to prioritize conservation decisions. Methods Data Collection Inventories were carried out in a total of 39 dry forest fragments in southwestern Puerto Rico. The selection of study sites and the methods of data collection are outlined in Chapter 2. 153 Species Richness/Species Density Inter-site species-area curves were modeled by means of a power function of the form S = CA2 (Preston 1962, MacArthur and Wilson 1967) where S represents the species richness of the fragment, A represents the area of the fragment in hectares and c and z are fitted parameters. The parameters were estimated by maximum likelihood nonlinear regression of the untransformed data (See Chapter 5, Figure 5.1 and Table 5.1 for more details of the species-area relationship). Residuals of the regression were scaled relative to the value predicted by the regression via a x2 transformation (Observed - Expected)2/(Expected) and were assigned a sign (positive or negative) based on whether the observed species richness was greater than or less than the predicted value. Values of the standardized residuals that were greater than 3.84 were considered to be significantly different from the predicted value (at the 0.05 level, based on a 1 degree of freedom x2 test). All analyses were carried out using Systat 9 (SPSS Inc., 1998). As shown above, the relationship between species richness, S, and area, A, is nonlinear. If species density were calculated as S/A, it would decrease with area, and larger sites would necessarily have a lower species density than smaller sites. If the functional form of the relationship between S and A is accounted for, then the measure of species density becomes S/Az which is c (since S = cAz). Species densities were calculated for each fragment and fragments were ranked on the basis of species density. 154 Representativity Reference lists were compiled based on the species composition of a total of 19 25 m;2 plots in three forest associations in Guanica Forest (see Chapter 2 for details). The species composition of each fragment was compared with reference lists complied from randomly sampled plots located in each of the three main forest associations present in Guanica Forest: Semi-Evergreen forest, Deciduous forest and Scrub Forest. See Table 2.1, Chapter 2 for sampling design. The reference lists for each association are presented in Table 7.1. Species lists for all studied fragments were compared with these reference lists and the proportion of species in each reference list present in the fragment was calculated. 155 c.3333 3333.5 E33832 SEED 38me 3.35569 333.36 65.3390 6369333 338er 623233 32830 833.433 32.5 n33 33330 3.3: 6.3639 ewezxmvxutemmeb 4333:33S8 MEQQQGU BQRLGENM kahk3m hfikmu3fl GENURQ mB~=m~zbbzh kamxkfibm _.mmm 55:52ng £353 38.6 333.3 33gb 383E ESQBQQEEU 3333 Sufiocfiob 33329835 338.30 mm:.& 33839 33.432633 33830 uSESEi 33:39 39552. 33.3% .3363 333m M 833.85. 33.38% _.mam QEEQSQLV .333 33.0 L383» 33.5 353.33 3.3.6 38.35 533833.89 3333 3.833809 .38me 62.35280 3.23:» czteéeb 33333.8 633339 tweak 38830 Esteban “.335 3.938% mismmeb eux39§~33=® £33330 33.839. 33.3% 3233 3.83% maumflsooa Dickson 352393 3335mm £833sz Stubs _dam 3:59?wa N 3833 33.3an 833.3 3333‘ 5323» utAEV 33:83 MEAEV 53:3» 22th “mucou— 52wb>m mEom 620m 305289 £30m 3300 580m 83:0 Boa 33:8m momooqm o>_§:omoaom mo 65 vacuumed "as 033—. 156 53333333333333NM3¢N 3333333533 33335333333 53353333333333.3399 .3353 3.3333333 533333\333333333M3¢~ 33.33 3.333533% 3333333533 3333533333 .3323 3.3333333 3.3333333 333593 5.3333333.» 3335333 3. 3.3333333 333533 333.3333 333~3mkm 333353333. 33333.5 33.33333. 353333333: 33.3353Q 3533R33333m 33L3E3Q 3§333.3.~333~.~ 33.333 33353336 33.83 333533536 33.83 333533536 33.33 33.33333mv 3M3 33.3339 333.33 33.33333U 33.3333 33.33333w 33.33333 33.333330 5333333 5333330 5333333 5333.333mv mdw 880 mdm $330 0% $330 3% 380 333.3533 3333333333 333M333“. 333.333.33R 35333.23... 353.83% m353333.233 353.83% 353333.233 353283 dm 3.53%3m 3333.33.33 353M3m~ 33.333333 353M3m 33533.3 3.533% 33533.3 3.333333% 3353.3 3.333w3m 33.333333 353M3m 33.333333 333M333 33.333\3.53M3M 33.333\3.~33M3m~ 33.333.\3.53m3m~ 3,233.33 353M3m~ 53333333333 53~§3333bm 533333333 53~Ax3333bm 3333.33.53.33333m 333333§333333m 53533339 3333333333~m 3333333 3.33.3me 333K333 33333.5 338,.“ 3333w33>m 383m 68o”— msosEoom 33.3% 33.800 €8,583 "E. 2...? 157 Ezsxwmm3mm=3 £33333mmm 5333333353 £333~3M 3333333 333383335 5333533. 333.55% n33333~3§323§33§m 3x33333333 33AE3M. NEH 339%.sz33 5333333335 3333 3.5an 3398.3 333338.85 3333333 3.383% 333.333. 3.383333% .Qm 3.353% 333-3335 3:33A3\33V< 333333333 3333333 magmas“. 383NM33§$M 3333333 333x3§3~m 353333.835 3.3me3% 3333833 3Eo$®- 353% 3N33=Am- N g c H 33333.33 3.333% 3333 3.53mi .3333 ”33333333 3333383 3.323% 336 3533333333 333333333 333.6535 333E3N 3333.833 3E3S©- 3533M SSSQM 3333383 3.3333. 33 3.23533 332:3 3.383.: .3333 mzmxmuaoam 3333383 3.32.3.5 3333333393 333.5 3.8333533 3.33233 3.3.83.3 $333333: 3~3-Q33833~ 3333333 ~33LU3NQ>§N QSQEKQN 620m 5233m— mEom meHOnw mficfifimooa seem 338 €3,583 "E «zap 158 3333335 33333339 0332: 532 n .3835 3.235539 ova—05 532 3 33333333.... .m .«o 23 £382 “mo“: :2: “mowwsm £886on 830:? ”Gog Noswtuom-ouo>oo< oomv 3333383 3.33.8.0 03205 538 33333333.» 3.:mtzcm mo 388% m 38.333 .3 as 22... $3 Ewe: N 86on 25 mo 23 3808 ~88 ~35 wcumowwsm 633333333 .V bwfififififigo 83 80862; 55:32 .225 05 E wow—macaw? 5:88 on 8: 2:8 36on 805 £333. .V 35 333333333 .V n dam 3.333333wa _ 33338.33 «.3353NNN 3.3333: 3,3333% Edam 32.533 «3% 05> 2 .% 05> 3.3%3 3333.3 5 3.3%3 3x35: 5 Aogogzv 085 32535335 333353533 5 323536.35 3.53.53533 5 333E333 3.43352 5 3333333 333333 5 3333333 3.3.3333 5 333.23. 33.33.55 3333.”. 3.333335 333.5,... 3.5.5335 333%283 3.35333 5 323%222‘ 3.35333 5 3-A§m3§3§ 3.3533 5 2333333353 =3-§53£M3m 233353353 33:5:535M3m. 233353353 =3-§53EM3m .885 accumugm gum 6985 303289 85m 3800 gauges u: 2.3. 159 .3533 2333 .33 3333333 .3 3:3 .3 5383033 853333 .3 33823 was I 3% 333332 .335 35 5 3323330 33838 :3 8 333% 333i? 335.53 8 nwsoco 3333538 _335 “o: 03 H 35 .323 5333 35 5 33333333333333. 33> 333.333 .5 355 8 8398 3353 H 3353533, 93 35 no 30555333 ofimfiwoow 3338333 35 50 £333 35 :0 3353333» 95 35 3332533 833338385 3333 30 333.333 53> 333.333 .5 50 305933333 35 333838 5350 53333333235 33> 333.333 .5 50 «855.5333 35 5 23333335 35 .«o 3308 25>» 3me 32on “3333333333335 .H3> 3333.33 5 33333 33353, 9 333323.— 3333 33: 3333333333335 3.5.3835 o 602 Noswt33mé33>3omn do 83 3.55m 5883:58 9 sea «case a 855% as 8 8 A82 Essa as :28 Wobwfieco Ba: bfifiowom mo £93 05 so @3388 $3 a: £9 .63 2.: 8 22: 3282 econ 52o.“ be 05 E “comma 556 0.8 €33 860% :33 @203qu 23 max ”S. 033. 162 163 8888§K8t888-§ 088820805 8888.888 8.88% 82:38:88 8.838% 888%888 8.838%. ommofifiswoq 8.98m? 88%. 8888282 888288 8.888988% owoomfiwfiom 888% 8.:88mmz omoomgé 8.883888% 8.88% omoomfizm 5:838». 888888.8883m 083883 888.8%.8388 888N888n~ 88.853888 888N888a~ cavemen—stem 85288888 8~8MA~8K 88888888 8~8MA~8K omoomfimbom 33:88 8~8mb8m omooafiwbom 8888.: 882K 88833.5 828N828: 8.58.8mmnw 83835 8.38888 88SE8R omoofiocmmmam 8888.8 8.888888% 88888533 2358885 85888886 88088882 83888880 8888682 8888.88.88 8.8830 0888880 888888.888 58.88.: 0838.32 38888888 «858888: 388~8A~8Q 3&8888: 888883888 8858888: oncommnsm 88.38885 88388882 8858388 38888.8: 88588888 883888SV< omooafism 88888 888M888: oncomflmfifiz 8.883% 8888.8. .8> 88888.8 8.393 .8: 8888.8 8.8893 oncomotm 8888838888 8888888883 88888828888 88888838N 8888880 888mg.” 88883 oncomaofio> 88.8888 88.8888 umoomfi§o< 88828888 8.8588885 08383839005. . 5i :_€wmn_a—b_§ueuew2iow £35 - 88: .8 8 2:325 $3: 8.5883 8 :25 835E «8:53 2: an: .3. £5..— .Awosaflcoov NH 038,—. é? saga $83 z £23 8 35% 2t oEEuBou 8 Eamofismfi E 3.69: 05 $30on 08% 05 8 Home MESNB firmwam Ea b.2823 ”SKA: 50m v .8038 3:08 ES £325 M0 £85 m \>ow.mé.couow§u:o\\dtn 3m Burr—om £355 93 5E .m.D 2: B 862% wouowqmuco nu ma warm: 8m 369% 83693 a mm 5833 3.569 N \>om.m£.wouow§vao\\umtn an 8585 05 ~96 ofimzd>< _ Eztcmmufiofi EfibnofizuN Eazfincficfi ESbSSSVN omoofism b.8893 322:»; b.6893 9:85»; 0862380 $25»qu Smfitk 3385b thflfi 38232 33.4232: Smnzazfi 08320895 11 ANNE“: Etoeoeoatom - 33: .3 3 c2250 63: b.5253 a. :25 «23»:— oEEB 23 firm .3. 2:5; €2,532; 3 2...? 164 Cluster Analysis Hierarchical clustering with complete linkage was used to group studied sites on the basis of similarity among the various metrics used to evaluate fragment conservation potential (see Legendre and Legendre, 1998 Chapter 8, for justification for the use of this method of clustering). Values used in the cluster analysis were transformed so as to scale to values between zero and one, in order to avoid imbalance in the weighting of the variables. Site history (as determined in Chapter 3) was compared with the clusters produced by this method. Analyses were carried out using Systat 9 (SPSS Inc., 1998). Results Species Richness/Species Density The species richness of more than half of all fragments did not differ significantly from the value predicted by the species-area curve. Twenty-two fragments had residuals that were not significantly different from predicted values, seven had significantly more species than predicted and 10 had significantly fewer species than predicted (Table 7.3). Most Relict fragments (fragments that were more than 75% ‘old growth’; see Chapter 3) had either significantly more species than predicted (six fragments) or did not differ from predicted species richness (11 fragments). Only three Relict fragments had fewer species than predicted (Table 7.3). Most Regrowth fragments (fragments that were more than 75% post-1936 regrowth; see Chapter 2) had either significantly fewer species than predicted (six fragments) or did not differ from predicted species richness (eight fragments). Only one Regrowth fragment had more species than predicted (Table 7.3). 165 Two Mixed fragments (fragments that were more than 25% but less than 75% post-1936 regrowth) did not differ significantly from predicted species richness, while one had significantly fewer species than predicted (Table 7.3). The pattern of distribution among the categories was not significantly different from what would be expected at random (Pearson’s 12, 4 x 3 contingency table, n = 39, p < 0.397 for all categories; if the ‘unknown’ category was omitted, 3 x 3 contingency table, n = 38, p < 0.257). Species density matched the pattern displayed by the Standardized Residuals (Pearson correlation r = 0.932). Species densities ranged from 31 species/ha to 104 species ha.l (Table 7.3). If the same calculation is performed using the area and species richness of Guanica Forest, a value of 227 species ha'1 is obtained (however, Guanica Forest does not fit the species-area curve calculated for the fragments; see Chapter 5). If the ‘canonical’ value of z is used (0.26; Preston, 1962) a species density of 89 is obtained for Guénica Forest, a value which is in line with the upper limit of values obtained for the fragments. Representativity Individual fragments supported between 7.1% and 85.7% of species in the reference lists compiled for each of the three main associations in Guanica Forest (Table 7.4). Individual fragments scored similarly against each list, reflecting the fact that the three associations overlap substantially in species composition (Table 7.1). Five fragments (Sites 1, 2, 4, S, and 7) supported more than 75% of the species present in at 166 mm is o to 2- E3 8 3 mm em _.8 o so 3- escape Q :3 mm mm 4.8 o 3 3. saewom we 2 mm a :3 o 3 3- EE R N 2 om 6% o 3 so 20m 4: NE a 2 S: o 3 so as; we no mm M: 2K 0 9o 3 .gofié 5 m S t 0.9 o No 3 323 S 3 : 2 3K 0 so 3 588m 2 _ em 2 a9 0 Z 2 gamma a 8.0 «m 5 3» o E 3: £3 8 :.o a 2 tom 0 mm 3; Eve 2: e E S «.8 o 2 3; E3 m: e w 2 _.S o an MS £3 E so R a as o 2 n: 352 8 em a 2 0.3 o 2 we 520m 2: me e : 2” o 3 ”.2 Ba: 2: 3 2 w E» + 3 9mm 2% w: z: m 8 2m + on c.cm 20m 3 2 M: m mg + 5 m2 532me R No 2 4 mg + _.: mom seam N: 5. 2 _ ”.2: + E _ as 322 8 sod 3 m 32 + v.2 _.mm 20m 2: 2 a m 93 + 0.2 0.: same 0: mm 5 gm bacom momooqm gamma/0Q 3660M wonmvbwvcfim 1.328“ , boummm $053M 360mm 8.3. 8%. .333 86on mo £83 5 aaoawmh mo 620 see 05 3 Eamus E .on 2038830 88m b23336 8&6 8a Be 35 8 3 38096 52: 860% 526m €52.“ng At @8898 55 86on 808 DES—mama 53, $533 8 Home $383 .83 885 8883598 5 "nugget 828 be MC 633 83-86on 05 80¢ 35:2on 233% 363m: 28 3285—85 Nx B mum—Ema BEEN—653v 0E3 moaéflooam 05 mo mas—ouch vofivfiwaflm “ms «Bah 167 168 mm o: - R. R 2a. 5388a : _ R R QR - 92 ”.3. 820m on em a R new - Rd R. R- goaom : So mm mm 2... - 3: OR- goaom 9. 2 R R 3... - 3: 6R- 20m mm 3 t R 5% - 3 an- Biz R4 E R 8 we. - we ma- 538% R 83 8 mm wt. - E _.2- sieve mm 8.0 R R c.cm - 3 RR- 20m 3 8 e R E... - 3. Rd- gamma R Rod R 2 ”Rm o S 92- 638me m... Re R 2 mom 0 3. 2:- geese R So R R is o R ”.2- eaewom :. S R R 3% o 3 ea- 30m we 3 S R 0% o S 3 7 aaoaom S 2 2 63mm bmmcom 83on cozmm>om _msEmom BNESESm 328% ©onme $2303 86on meg 02m 88583 as as; least one of the reference lists, and three sites (Sites 1, 5 and 7) had more than 75% of the species in the deciduous forest, semi-evergreen forest and scrub forest lists (Table 7.4). The smallest site that scored above 75% for any of the associations was Site 7 (33 ha). A total of 19 sites supported > 50% of the reference list for at least one of the three associations. The smallest site to support > 50% of the reference list was Site 31 (0.11 ha) for the semi-evergreen forest association and the scrub forest association, and Site 36 (0.04) ha for the deciduous forest association (Table 7.4). Relict fragments supported 17.1-85.7% of the reference species while Mixed fragments supported 39.2- 78.6% and Regrowth fragments supported 7.1 -47.6% of the reference species. All fragments larger than 100 ha had > 75% representation. All fragments over 33 ha had > 50% representation. Rare and Endangered Species Twelve of the 53 rare or endangered species were present in at least one of the fragments (Table 7.5). The most widespread species, Leptocereus quadricostatus (Bello) Britton & Rose, was present in 18 fragments and the second most widespread, Guaiacum sanctum, was present in 11 fragments. Psychilis krugii (Bello) Sauleda was recorded from eight fragments, Jacquinia umbellata DC. from five fragments, Eugenia woodburyana Alain and Passiflora bilobata Juss. from four, Polygala cowellii (Britton) S.F. Blanke, Reynosia guama Urban and Trichilia tn'acantha Urban were present in two and Bourreria virgata (Sw.) G. Don, Cordia rupicola Urban and Randia portoricensis (Urban) Britton & Rose were present in only one fragment (Table 7.5). A total of 25 fragments supported at least one rare or endangered species. One fragment (Site 1) had nine rare or endangered species, two fragments (Sites 2 and 4) had five species, one 169 Table 7.4: Proportion of the species composition of the three main associations in Guanica Forest that are represented in sampled dry forest fragments, southwestern Puerto Rico. DeF = Deciduous Forest, SEv = Semi Evergreen Forest, ScF = Scrub Forest. Site Area History Percent DeF Percent SEv Percent ScF 1 1372 Relict 85.7 84.1 81.1 2 770 Mixed 78.6 73.0 78.4 4 125 Relict 67.1 79.4 71.6 5 101 Relict 84.3 84.1 78.4 6 64 Relict 52.9 57.1 52.7 7 33 Relict 75.7 79.4 78.4 8 7 Relict 50.0 55.6 48.6 9 5.9 Relict 31.4 30.2 36.5 10 5.1 Mixed 48.6 60.3 51.4 1 1 2.6 Relict 60.0 65.1 55.4 12 3 unknown 44.3 49.2 51.4 13 2 Relict 38.6 41.3 35.1 14 6 Relict 52.9 54.0 51.4 15 3.7 Relict 40.0 41.3 41.9 16 3.3 Relict 48.6 42.9 41.9 17 6.3 Relict 55.7 42.9 52.7 18 1.5 Relict 54.3 60.3 52.7 19 1.5 Regrowth 35.7 41.3 36.5 20 1.5 Regrowth 20.0 27.0 20.3 21 1.2 Relict 57.1 65.1 60.8 22 1.2 Relict 62.9 50.8 59.5 23 1 Relict 17.1 23.8 23.0 24 1 Regrowth 7.1 7.9 9.5 25 2.4 Mixed 42.9 54.0 39.2 26 1 Regrowth 37.1 41.3 36.5 27 0.4 Relict 38.6 39.7 35.1 28 0.8 Relict 50.0 46.0 51.4 29 0.2 Regrowth 25.7 31.7 21.6 30 0.2 Regrth 42.9 47.6 37.8 31 0.1 1 Relict 52.9 54.0 51.4 32 0.1 Regrowth 25.7 33.3 31.1 33 0.09 Regrowth 12.9 19.0 14.8 34 0.07 Regrowth 30.0 38.1 32.4 35 0.07 Regrowth 27.1 33.3 24.3 36 0.04 Relict 50.0 46.0 48.7 37 0.02 Regrowth 22.9 22.2 23.0 38 0.01 Regrowth 15.7 15.9 14.9 39 0.01 Regrowth 18.6 17.5 18.9 40 0.006 Regrowth 14.3 17.5 16.2 170 mm x. R R R RER R 2 2 t 2 2 3 2 2 : E w R e m 898: Z EoEwSm >< XXXX ><><><><><>< 656666.66 6236.25 6566M 6.266666% 22266266266. 6.666% muss £3666 33:66 6~6Mb66~ 6666626 6.263666% 666626636666 wzmxmoefisq 66626656 6.2566665 £68666 5666.666b 669066669: 6.56M6M 636.56g 6.26269 6268.: 6.226226% 86on 171 fragment (Sites 7) had four species, four fragments had three species, seven had two species, and 10 had one rare or endangered species (Table 7.5). Indicator Species A total of six species fit the criteria selected to identify potential indicator species (Figure 7.1a-f). These species were: Antirhea lucida (Sw.) Hook. f. (Rubiaceae), Coccoloba diversifolia Jacq. (Polygonaceae), Cordia rickseckeri Millsp. (Boraginaceae), Guettarda krugii Urb. (Rubiaceae), Plumeria alba L. (Apocynaceae) and Savia sessiliflora (Sw.) Willd. (Euphorbiaceae). Three fragments supported all six Indicators (Sites 1, 5 and 7), two fragments supported five of them (Sites 4 and 15) and two fragments supported four of them (Table 7.6). Sixteen fragments supported none of the six Indicator species. Cluster Analysis The clustering algorithm yielded two groups of fragments based on measures of conservation potential, using a cut-off distance of 0.8 (Figure 7.2). One group consisted of 16 Relict fragments, three Mixed fragments and two Regrowth fragment and one fragment of unknown history. The second cluster consisted of six Relict fragments and twelve Regrowth fragments. The pairs of fragments that were most similar were Sites 8 and 14 and Sites 29 and 32. Fragments in cluster 2 were mainly located in the western and southern parts of the study area (Figure 7.3). Based on the constituent fragments, the first cluster appears to be fragments of higher conservation value, while the second cluster appears to be those of lower conservation value; fragments in cluster 1 averaged 2.7 Indicator species per site, while those in cluster 2 averaged 0.7 Indicators. No 172 Antirhea lucida Proportion of Sites Occupied 1 2 3 4 5 6 7 Species Richness Class a. Antirhea lucida Figure 7.1: Incidence functions (sensu Diamond 1975) for the six species present in dry forest fragments in southwestern Puerto Rico which met the criteria selected to identify indicators of sites with high conservation value. Species richness classes each consist of 5-6 fragments grouped on the basis of total plant species richness. Indicator species were defined as those which were present in less than 20% of the two most species-poor classes and were present in more than 80% of the fragments in the most species-rich class. 173 Coccoloba diversifolia 1.0 0.8 a 0.6 ~ 0.4 - Proportion of Sites Occupied 0.2 a 0.0 i l 1 1 1 2 3 4 5 Species Richness Class b. Coccoloba diversifolia Cordia rickseckeri O) 1.0 0.8 - 0.6 4 0.4 ‘ Proportion of Sites Occupied 0.2 - 0.0 C i I v 1 2 3 4 5 Species Richness Class c. Cordia rickseckeri Figure 7.1 (continued). 174 Guettarda krugii 1.0 Proportion of Sites Occupied 1 2 3 4 5 6 7 Species Richness Class (I. Guettarda krugii Plumeria alba 1.0 t, 0.8 J .9 Q 3 8 c 0 . w 0.6 a) .t a) s... O .5 0.4 - 1: O Q 9 0- 0.2 - 0.0 e c i . . 1 2 3 4 5 6 7 e. Plumeria alba Figure 7.1 (continued). Species Richness Class 175 Savia sessiliflora 1.0 Proportion of Sites Occupied Species Richness Class f. Savia sessiliflora Figure 7.1 (continued). 176 Table 7.6: The number of plant species among the six proposed Indicators of fragments of high conservation potential which were present in studied dry forest fragments in southwestern Puerto Rico. Site Area History Species Number of Richness Indicators 1 1372 Relict 174 6 2 770 Mixed 125 4 3 45 Regrowth -- O 4 125 Relict 147 5 5 101 Relict 148 6 6 64 Relict 95 3 7 33 Relict 149 6 8 7 Relict 103 4 9 5.9 Relict 50 1 10 5.1 Mixed 101 0 11 2.6 Relict 82 O 12 3 unknown 81 2 1 3 2 Relict 75 O 14 6 Relict 101 3 15 3.7 Relict 112 5 16 3.3 Relict 68 2 17 6.3 Relict 58 1 18 1.5 Relict 93 2 19 1.5 Regrowth 62 l 20 1.5 Regrth 46 O 21 1.2 Relict 104 2 22 1.2 Relict 68 2 23 1 Relict 64 0 24 l Regrowth 31 0 25 2.4 Mixed 92 3 26 1 Regrowth 76 O 27 0.4 Relict 74 3 28 0.8 Relict 68 3 29 0.2 Regrowth 43 2 30 0.2 Regrowth 77 O 31 0.11 Relict 63 O 32 0.1 Regrowth 41 0 33 0.09 Regrowth 35 O 34 0.07 Regrowth 57 1 35 0.07 Regrowth 45 O 36 0.04 Relict 69 2 37 0.02 Regrowth 29 0 38 0.01 Regrowth 17 O 39 0.01 Regrowth 28 O 40 0.006 Regrowth 21 O 177 MIXED RELICT RELICT RELICT RELICT RELICT RELICT MIXED RELICT RELICT RELICT RELICT REGROWTH RELICT MIXED RELICT REGROWTH RELICT REGROWTH RELICT REGROWTH * RELICT RELICT RELICT REGROWTH REGROWTH REGROWTH REGROWTH RELICT REGROWTH REGROWTH REGROWTH REGROWTH RELICT REGROWTH REGROWTH RELICT REGROWTH Site 2 Site I Site 7 Site 5 Site 4 Site 6 Site 28 Site 25 Site 8 Site 11 Site 21 Site 36 Site 30 Site 15 Site 10 Site 31 Site 26 Site 13 Site 19 Site 16 Site 22 Site 12 Site 27 Site 14 Site 18 Site 34 Site 29 Site 32 Site 35 Site 23 Site 39 Site 37 Site 33 Site 40 Site 17 ”if” Wig W J l __ Site 20 Site 38 Site 9 Site 24 it i: 0.0 0.2 0.4 0.6 0.8 Distances l 1.0 l 1.2 Figure 7.2: Hierarchical clustering dendrograms of Puerto Rican dry forest fragments based on scores of their Conservation Potential. 178 fragments in cluster 2 had more than three Indicator species, and only one (Site 6) had more than two of them. Discussion Species Richness/Species Density There were no surprises among the fragments whose species richness differed significantly from predicted values. The fragments with the largest negative standardized residuals (Sites 9 and 24) were also outliers in the fragment community classification (Chapter 4). Site 9 was very species-poor. The community was dominated by two woody species, Erithalisfiuticosa and Coccoloba microstachya; 66.3% of all individuals sampled in this fragment belonged to these two species. This site was located on a windswept peninsula and probably had the most extreme environment of all the studied fragments. It was also the most isolated — it was part of the only cluster of fragments that was more than 1000 m from other clusters of forest fragments (Chapter 3). Site 24 supported a community that did not gain membership in any other cluster on the basis of species composition (Chapter 4); it was also more mesic than any of the other studied fragments @ersonal observation). Most of the fragments with large positive residuals had shrunk significantly in the 1936-1993 period. It seems reasonable to assume that these fragments are ‘oversaturated’ (sensu Diamond 1972): they currently have more species than they would support at equilibrium. It is likely that enough time has not elapsed for them to have reached equilibrium. However, in the case of Site 21 (the fragment with the second- largest standardized residual) it seems more reasonable to invoke the intermediate 179 disturbance hypothesis (Connell 1978). This fragment had an open canopy and was grazed by cattle (personal observation) — it seemed a prime candidate for weed invasion, but much of the original species complement appears to have still been present (it supports 51-61% of the species on the reference lists; Table 7.4) Measures of species per unit area are qualitatively similar to the standardized residuals of the species-area curve. Whittaker et al. (2001) argues against using measures of species density that are estimated statistically since they tend to incorporate ‘other information’ about the site, and in doing so may mask the real relationship between per- plot species richness and fragment species richness (see Chapter 5). It is precisely this property of the estimated value of c that makes it a useful measure of fragment conservation value. If species richness (or some metric derived from species richness) is to be used as a measure of fragment conservation potential, it should reflect the factors that make the site richer in species than might be predicted on the basis of area. Habitat heterogeneity is a factor of this type. Representativity This is perhaps the most intuitively appealing of the measures of fragment conservation value. It would seem reasonable that a fragment with a species composition resembling that of the reference community would be functionally similar to that community. However, macroecology theory suggests otherwise. Common species are likely to be widespread, while rare species are likely to have more restricted distributions (Gaston et al. 1997) although the pattern of incidence of Guanica Forest species in the fragments is somewhat more complicated than that (see Chapter 6). Nonetheless, there is 180 a general pattern where some species (that are ‘common’ inasmuch as that they were included in the relatively small sample that was used to compile this reference list) are present in most of the fragments regardless of the overall ‘quality’ of the fragments. These ‘ubiquitous’ species are, in fact, likely to be present in almost all fragments, which may be a function (at least in part) of their ability to utilize matrix (non-forest) habitat. As was shown in Chapter 6, the species that are most abundant in Guanica Forest are not those that are the most widespread among fragments — in fact, these dominant species are absent from many Regrowth fragments. See Chapter 6 for a more detailed analysis of these types of species patterns. Large fragments show high representation. Five of the six largest fragments that were inventoried had > 75% representation across all three of the reference lists. The only large fragment with < 75% representation (Site 6) was partly grazed in the 19303 (R. Carlo, personal communication). Several very small fragments had high representation, consistent with the predictions of the “more of the same’ model (Thomas 2004). Two fragments smaller than 0.5 ha supported > 50% of the reference lists. It was immediately apparent in the field that these fragments were high-quality remnants of dry forest. The fact that fragments this small could support such a large number of the reference species illustrates the resilience of dry forests, especially since the smaller of them (Site 36, 0.04 ha) was already mostly isolated in 1936. Unfortunately, this fragment was destroyed by fire in the 1998 dry season (S. Van Bloem, personal communication). Thus illustrates a fimdamental weakness of small fragments — their elevated susceptibility to disturbance (Janzen 1983, 1986b, Viana and Tabanez 1996, Viana et al. 1997, Cochrane and Laurance 2002). 181 The idea that small fragments can support a substantial proportion of the native community agrees with findings by (Pither and Kellman 2002) who found that 25 small fragments (ranging in size from 325 m2 to 3625 m2 were able to support 106 of the 160 tree species present in large fragments of gallery forest in Belize, and by (Thomas 2004) who found that 85% of the species in five focal genera present in a 50—ha plot at Pasoh Forest Reserve, Malaysia, were present in 12 l-ha forest fragments which had been isolated for about 25 years. Rare and Endangered Species The presence of a species of concern in any fragment makes it valuable. On the other hand, the presence of endangered species does not always indicate the presence of high quality habitat (Gilfedder and Kirkpatrick 1998). In this instance, the presence of rare and endangered species was not the best tool for identifying fragments of high conservation value. The two most widespread species on the list, Leptocereus quadricostatus and Guaiacum sanctum L. were present in fragments that were not otherwise seen as having high conservation value. Leptocereus quadricostatus is a narrowly endemic species restricted to the dry forest zone of southwestern Puerto Rico. Any species endemic to so heavily degraded a habitat should be seen as being at risk. Concern for this species was also expressed regarding disease (Quevedo et al. 1990). Similarly, G. sanctum is a rare species that has been over-harvested historically. However, it does not appear to be as rare as Quevedo et al. (1990) thought it to be. Both of these species are fleshy-fruited and appear to be vertebrate-dispersed (see Ricart Morales, 1999, for details of the dispersal of G. oflicinale, an ecologically similar species 182 with similar fruit; Liogier 1988). In addition, L. quadricostatus appears to be able to colonize regrowth and survive in heavily disturbed areas; it was present, for example, in Site 26, a Regrowth fragment in the middle of a cattle pasture. As a spiny cactus, it is likely to be resistant to browsing by cattle. Most of the species on the list of rare and endangered species were not present in the sampled forest fragments, although the possibility exists that they were misidentified or overlooked. The fragments that supported the largest number of rare and endangered species were large and species-rich. As a measure of fragment conservation potential, rare and endangered species added several fragments that would not have been otherwise considered valuable, but the information is difficult to interpret. Indicator Species Seven of the eight largest fragments were among the fragments that had four or more of the six proposed indicators. The sole exception was Site 6, with only three indicator species which was also the only large fragment with > 75% representation. Unlike the other assessment systems, the system of indicator species did not confer a high value to small fragments; Site 36 had only two of the six indicators and Site 31 had none. The system was developed using species with high fidelities for the most species-rich sites. Given the relationship between species richness and area, these tend to be large fragments. The rationale for using several species as indicators was to minimize the effects of the presence or absence of a single species: metapopulation theory predicts that species will be absent from some portion of their suitable habitat. Similarly, a single species may be a relic of a different community that had been incorporated into a 183 Regrowth fragment as it expands. The use of several species amounts to a measurement that is somewhat closer to a ‘community’ measure. All six of the species selected are distinctive and readily identifiable in the field. This should make them easy to use in field surveys. Another advantage of highly detectable species as indicators is that the probability of them turning up in a survey if they are present should be high. Even if species lists are incomplete, easily detectable species are likely to turn up, while more cryptic species might not be recorded. This feature of the proposed indicators was not selected a priori, but nonetheless is likely to increase the utility of these species as indicators. Cluster Analysis The clustering algorithm produced two well-separated clusters. The ‘high conservation value’ cluster consisted mostly of Relict or Mixed fragments. Most of the Relict fragments in the ‘lower value’ cluster were either highly disturbed (e. g., Site 11) or were purely ‘coastal’ fragments. Despite the fact that these fragments are likely to have a low species density (as a consequence of their being rocky with very little soil), they are valuable in that they represent a distinct community which, by virtue of its location on the seas shore, is likely to represent especially favored sites for the deve10pment of resorts and holiday homes. These fragments also supported littoral species which were not present in other sites. The location of the two groups of fragments (Figure 7.3) indicates that there is some degree of spatial separation between high and low conservation value clusters. In addition to the coastal fragments, fragments at the drier end of the spectrum may also be 184 more likely to end up in the lower-value cluster. It is possible that the two clusters represent wetter and drier groups of fragments, rather than hi gher- and lower-value fragments, but this interpretation is not supported by differences in history between the two fragments clusters. It is also possible that the difference may relate to recovery after disturbance. The recovery of dry forest after disturbance may be slower in drier sites than in wetter sites. It is also possible that deve10pment pressures were greater in drier sites, resulting in fewer Relict sites at the dry end of the spectrtun. The results of the cluster analysis agree broadly with fragment history and the system of proposed Indicator species in terms of the definition of more and less valuable fragments. Unfortunately, the classes are somewhat broad — while it might be desirable to protect all fragments in the ‘higher value’ cluster, it is probably impracticable to do so. There are two philosophical positions from which to approach the search for sites of high conservation value. One can either approach the question mechanistically, and in doing so try to use the correlates of ‘valuable’ sites to attempt to predict other sites which would be likely to be of high conservation value, or one could approach it from a purely practical point of view and try to identify sites that need protection on a case-by-case basis. The weakness of the former case lies in the fact that ecology remains a complex and often poorly understood science — identification of correlates of high value sites does not guarantee that the underlying causal factors will be discovered. On the other hand, simply carrying out inventories to search for valuable sites or to find species of concern is likely to be inefficient and require a large number of specially trained staff. In addition, if little is known about the conditions that allow a site to support a ‘valuable’ species assemblage, then one has no idea how to ensure that the site remains ‘valuable’. While 185 the idea of designating Indicator species is a first attempt to streamline the process of surveying fragments, it still needs to be tested and refined. Summary 1) Twenty-two fragments did not differ significantly from the species richness predicted by the species-area curve, seven fragments had significantly more species and 10 had significantly fewer species than predicted. 2) Of the seven fragments with significantly more species than predicted, six were Relict and one was Regrowth. 3) Of 10 fragments with significantly fewer species than predicted, three were Relict, six were Regrowth and one was Mixed. The pattern of distribution of histories among the groups did not differ significantly from random. 4) Fragments supported between 7.1% and 85.7% of the reference species. 5) Five fragments supported more than 75% of the species on at least one of the three reference lists; three of them supported more than 75% of the species on all three reference lists. 6) Nineteen fragments supported more than 50% of the species on at least one of the reference lists; the smallest fragment with more than 50% of the species on at least one of the reference lists was 0.04 ha. 7) Twelve of 53 rare or endangered species were present in at least one of the fragments; the most widespread of these species were Leptocereus quadricostatus and Guaiacum sanctum. 186 8) Twenty-five fragments supported at least one of the rare or endangered species; only two Federally listed endangered species (Eugenia woodburyana and Trichilia triacantha) was present in any of the fragments (in four and two fragments respectively). With nine of these species, Site 1 was overwhelmingly the most important fragment in terms of rare and endangered species 9) Six species were designated potential indicators of sites of high conservation value on the basis of their meeting the criteria designated; these species were Antirhea lucida, C occoloba diversifolia, Cordia rickseckeri, Guettarda krugii, Plumeria aIba and Savia sessiliflora. 10) Two well separated clusters were identified on the basis of the criteria outlined to determine conservation value of these fragments. 187 CHAPTER 8: CONCLUSIONS AND RECOMMENDATIONS Conclusions Puerto Rican dry forest fragments (including the large fragment, Guanica Forest) represent the last remnants on the island of Puerto Rico of a vanishing biome. The abandonment of agriculture in Puerto Rico and the concomitant return of forest cover provide an opportunity to manage the landscape for the preservation of the native biota and the survival of this commrmity type. It is not only important to attempt to preserve the species that are present; it is also important to attempt to preserve their genetic diversity. Relict forest fragments are able to harbor a representative assemblage of plant species (up to as much as 86% of the reference species; Chapter 7), but many Regrowth fragments do not — Gymnanthes lucida, Croton humilis and C. discolor were only present in 6.3% of all Regrowth fragments (Chapter 6). Instead, most Regrowth fragments are dominated by Leucaena leucocephala. While forest cover has been able to recover without intervention (Chapter 3) this forest often lacks the community dominants, even as much as 50 years afier abandonment (Molina Colon 1998). In light of this, it may be necessary to re-introduce these species into Regrowth forest. It is impossible to attempt to restore Puerto Rican dry forest to its pristine condition, given the uncertainty as to what would constitute ‘pristine’ vegetation. One can attempt to re-create a hypothetical ‘climax community’ much like Gleason and Cook (1 926) did, but there is no guarantee that restoration would yield a forest that resembled their ‘Bucida series’. While Murphy and Lugo (1986b) suggested that the short, multi- 188 stemmed nature of the forest was a consequence of historic cutting, Dunphy et al. (2000) found that the condition was natural. Nonetheless, this does not constitute proof that the original vegetation was short and multi-stemmed historically. It was suggested that the original vegetation might be unrecoverable because erosion may have stripped much of the topsoil of the original community after the trees were cut, resulting in the present short-stature, multi-stemmed forest (F. Wadsworth, personal communication). It is also possible that the draining of Laguna Guanica altered the rainfall regime, resulting in a more xeric community than existed initially. On the other hand, Van Bloem et al. (in press) have shown that the multi-stemmed condition may be caused by wind storms, a fact which suggests that the forest may have always been short-statured and multi- stemmed. Nepstad and colleagues (Nepstad et al. 1994, Jipp et a1. 1998, Moreira et al. 2000) have shown that deep soil moisture plays an important role in the water relations of dry forest trees in Para in the Brazilian Amazon. Trees that are able to access deep soil moisture continue to transpire much later into the dry season than do pasture grasses. While some trees appear to have access to soil moisture in Puerto Rican dry forest because they remain green and appear to be transpiring throughout the dry season (e. g. , Guaiacum officinale; Gleason and Cook, 1926, Castilleja, 1991), no one has measured the role of this resource. In that light, it is impossible to attempt to estimate the change that may have occurred with the loss of the original forests, and whether trees in second growth are able to access deep soil moisture at all. What we are left with then is a need to manage an already highly impacted community whose ecology is still not well enough known. However, the reality of the 189 situation is that some sort of pro-active management is essential. Development will continue along Puerto Rico’s south coast. As development proceeds, forest will be lost, even as more pasture is abandoned and reverts to forest. Molina Colon (1998) has shown that even fairly old Regrowth does not begin to resemble relict forest (in terms of species composition), although, as has been shown in Chapter 4, some Relict forest can be degraded to something that resembles Regrowth without ever actually being cleared. The preservation of Relict forest fragments must be seen as a priority by agencies responsible for granting permits for development. Chapter 7 presents some tools that can be used to identify fragments that are likely to be of high conservation value. These tools must be used, tested and refined, and their predictive power relative to animal species must be tested. The most effective tools for conservation are also the most expensive. The purchase of land for the purpose of conservation and the purchase of conservation easements are valuable tools for conservation, but they are expensive. Without public support, it is also expensive to police the borders of reserves (see Allen 2001). Education is another valuable tool — based on my interactions with members of the public, they were all aware of the value of bosque seco (dry forest), but they were aware of it in the singular sense — the Bosque Seco (i. e., Guanica Forest). Many people did not seem to connect the idea of dry forest with the forest fragments amongst which they live. Half the task has been achieved, but more needs to be done to create an appreciation of the dry forest fragments outside of Guanica Forest. Priority areas for conservation outside of protected areas must be identified. The expansion of the Laguna Cartagena National Wildlife Reserve to include land in the 190 Sierra Bermeja is a valuable start, but the whole Sierra Bermeja must be formally identified as a priority area. Similarly, a good spatial perspective is essential. As demonstrated in Chapter 3, the strips of forest across the Laj as Valley should also be high priority areas. The southern strip, just north of the town of La Parguera, is likely to be at especially high risk. For the most part these are not Relict forest, although there are many Relict patches embedded in a larger matrix of Regrowth; it is likely that these forests would not be identified as being of high conservation value using the tools in Chapter 7. Their value lies in their spatial context. If these forests are preserved and expanded, it may be possible to create and maintain gene flow between Guanica Forest and the Sierra Bermeja. Several patches of forest to the north of Guanica Forest may constitute another hi gh-priority corridor. The Commonwealth Forests of Susua and Maricao lie in the southwestern portion of the Cordillera Central, and are the nearest large protected areas to Guanica Forest. The southern SIOpes of these forests support vegetation that shares many species in common with Guanica Forest — although they are in the moist forest life zone, the steep south-facing slopes are fairly dry. It is important to maintain gene flow between the plants on these slopes and their conspecifics in the dry forest zone. Thus, the corridor north of Guanica Forest is likely to be another priority area. Forest sites adjacent to Guanica Forest are also likely to be priority areas. There is privately owned forestland between Guanica Forest and the protected area that was studied as Site 4. Maintaining this band of forest between the two protected areas is important to the maintenance of the biological integrity of Site 4. It is also important to maintain a buffer around Guanica Forest where still possible — already the community of 191 La Luna has grown to the very edge of the forest, and the forest also abuts the town of Guanica. Maintaining a forested buffer along the northern and eastern edges of the park is critical. The largest of the studied fragments was Site 1. This fragment, to the northeast of Guanica Forest, supports Trichilia triacantha, a federally listed endangered species. Most of the fragment is privately owned, and is adjacent to housing developments in several areas. Protection of this biologically rich fragment is important, as are the fragments to the east of it in the Tallaboa area. East of Tallaboa and west of Ponce is a large patch of dry forest that was not considered under this study due to the absence of defined fragments. However, cursory examination has shown that large parts of this area support mixed-species forest and not Leucaena leucocephala-dominated regrowth. This status is predictable since the area was forested in the 1936 aerial photographs, even though it appeared to be heavily disturbed. Several endangered species and a species that is a candidate for listing have also been recorded in a portion of this block of forest. This area consists of at least 1000 ha of dry forest; if much of this forest is actually mixed-species forest, it may have a critical role to play in the conservation of Puerto Rican dry forest. It is critically in need of further study. F urther Questions As questions are answered, further ones arise. Some of the areas which seem to need to be addressed are: 192 1) Examine the role of species in accelerating and encouraging succession. Young secondary forests tend to be dominated by Leucaena leucocephala and are species-poor, but not all Regrowth forest is of this sort. Several distinct hypotheses are available regarding successional pathways, and these need to be tested a. Shade: evergreen species cast deeper shade, which is likely to encourage succession. b. Focal trees: fleshy-fruited species are attractive to frugivores, which are likely to be seed dispersers. Higher levels of seed input are likely to accelerate succession. Additionally, species which are not themselves fleshy-fi'uited but which are parasitized by mistletoes (mostly members of the Loranthaceae and Viscaceae) provide food to frugivores. Pisonia albida is one of the most consistently and heavily infested trees in the study area (personal observation). c. Water/Nutrients: strips of forest have been observed to develop along seasonal streams in the study area. Although these areas are Regrowth, they are support species-rich forest which clustered with Relict fragments on the basis of species composition. These hypotheses need to be tested experimentally. 2) Deeper understanding of the role of Leucaena leucocephala in the process of succession; does it inhibit succession through competition, does it provide an inhospitable environment for the establishment of seedlings because it is dry- 193 3) 4) 5) 6) season deciduous, or does it simply fail to attract seed dispersers because it is not fleshy fruited? Based on the identification of ‘old’ patches (areas within existing fragments which have been forested continuously since the 19303), actively search out rare species. Use spatially explicit models to predict the movement of species - seeds, seed dispersers and pollinators — across the landscape, and use these to improve connectivity through restoration and enrichment planting. Use population genetic tools to examine the levels of genetic diversity within fragments and among fragments to determine how well the existing genetic diversity is captured in protected areas. Focus on a larger suite of species to address the question of how different species and groups of species have responded to long-term habitat fragmentation. 194 APPENDICES 195 APPENDIX 1 FRAGMENT-SPECIES OCCURRENCE MATRIX 196 ca X a— m— : 3 m— X 6— m— n— _— :— ><><><><>< X X ><><><><><><>< X X X 66666.56 6.66.669 6663.3 6.666656% 66663666.... 6.265666% 6636MSE66 6.266666% 6.626666% 663D 666:. 6E666.~ 6666366666: 6.656665 we _ m 666-63M66 666669.66: 6x6~\.666~ 6.66.6qu 666.666 6.6266.»st 66636663 6.666666 Eon— amtéfi Stung dm 666663.6x 6.663: 6S6>66- 66.6.5 £69562 6~6=63MM6 6.366666% 6.33 6.26563 66666666 6.6566560 6.-6\.6§M66 65.66.: 66.6.5 6.3.630 6.8%6 665666366 dm ugmw 56.46.6666 5666626663 6.66.8.6“. 626~M6=663m 666.868 6.566% 6666.3». 6.3630 66.3%935 6.5665 86on oncomfiwfiom oaooeoapaom oaooameoewmm 3388?. 663862 oaooa§666< 66866885. 8365535 omooa>aw< omooa5§o< base .fiaofiwfim 626% be 68E 9.5an 5 $663 8m x568 85:38 oatmofiogm "36 635. 197 X X X X X XXXX :6 an um um on mm 6m mm mm 3” an an an bu on X mu 6N mu an an 66666.56 6.66.660 6663.; 6.266666% 666636666 6.666663% 6.3%3566 66.66.6666m 6.6666666R 66.660 6666. 656665 6666366266 6.656663 mo _ m 666-66M66 66666966: 6666665 6.66.6qu 6666.6 666669.660 66366663 6.666666 _ mom 6.5.6.6666 366.66: .mm 666666A6- 66.2: 669566 66.6.66 6S6>66- 6666.65MM6 6.666686% 6.56.6 6.666666% 66666666 666666560 6.66\.6§M=6 66.68.: 66.6.66 6.6.6660 6.66366 666666636V dm 666MV 366666666 E666=6666m 6.66.266 686~M6666Afi 6.8.6566 6.566% 6666.63. 6.66630 66.366.66.62 6.66666 «38% omooafiwfiom omooaomofiom 663686:me 83863.6 33362 6666656636 68626668566 omooafigg omooa>mw< oaooafieaoe. 2:56 6666668 "36 636,—. 198 an a— X m— b— e— m— X : n— n— X ~— X :— XXXXX XXXXX XXXXXX X XX X XXXX X XX XXXXXXX XXXXX X XXXX X X XXXX X X XXXX XXXX 6666.66 6.6666660 6666x6~xmt6ua60 66666.6 .0 S 6L66Q6666Q66A6 6.26QQ60 6.666666 6.26666 0 66666663 666660 66666 36.6666666K Ndm 6.6530 6% 6.6530 666%; 6.666Q0 6.66666 666666666: 6666666616666 6666666360 6666M66 66666665 6666666.... 6.66.260 666666666 6.6666665 66626666 66666-6& moon 3am 8am 6.536% 6.668660 6.3666 6.6.66\66.S6.~ 6666366665 6.6.66.\66.S6.~ 66663666666 6.6L6\666- 8666666366 56.66.66.662 £66.36; 6.66.660 266666.366? 6.66.660 666666666 6.66.660 666666 6.66.660 666666666666 6.66.66 0 8625 6686268660 66668660 omoofioflzm 6636:6865 oaooafiwflom Esq... 66:66:66 ":6 636,—. 199 X :6 an an em X on mm vn an an —m X an an an XXXXXX X X X X 5N on ma 6N mm XXXX X «N 3 6666666 6.26336 0 6866‘ 6.26336 0 66.665 .0 66 6666366663650 6.26336 0 6.666666 6.263360 666.666.: 66666 0 66636.6 66666666663 Ndm 6.6530 6% 6.6630 66636.6 6.6530 6.6.6665 666666666: 6666666626666 6666666360 66666366 666666633 666.665.: 6.66.260 666666666 6666663 66636666 6.266663 moan 686 mean 66636.3 6.66295 6.5666 66.66.66.363 666336.668 666666.363 666636666666 6.6.8.3662 £66S666m=6 56.36.66.661 6666366 6.66.660 666666.366? 6.66.660 666666666 6.66.660 666666 6.66.660 6666666266 6.66.660 8.36m 666662666660 66668660 omoofioflsm 6686:6865 ououafiwflom $5.... 66:66:66 “36 636,—. 200 an a @— XXXX 5 6— m— : 2 ~— ~— 3 XXXX 666666666663 6636666 6666.66 6360866600 6666666666666 «6669666600 663.6666 6560866366 6666666666 6566866600 6.6666666 6&6666: need 356 $06 mmoo 680 66663 6.666N mdm 6666336600 6% 26633650 6666663 66666336690 3 _ m .63 6666666666666 66366-63 66666656 6.266666 666.66% 6666.66 666666660 6666666 336666660 66.6666 66.66% 66666 .0 .8 666.66 6.560 amok 66666636666§66660 .66 666636: 6666 665666 6366663.». 66.66660 669363 66.260 636on oauomfiuonasm 66666368505 6668936665 66386me 6666666630 omoogammwhnv 66666323660 66685—68860 6666826660 66666320 663668—60 66668660 £5... 6626668 :4 635. 201 666666666663 66366666 6.666666 666660866360 66666666666 66660866360 663.5666 66660866360 66666666666 6666086636m 66666666 6.666.666: M696 N6?— _ 6cm mmoU 680 66663 6.666N Ndm 66666636600 6% 6666336600 666666663 6666666636600 6o; dam 666666666666 666366-663 66666666 6.6666666 6666663 666666 6666666660 566666 6636666660 6666666 66.66am 00.606 .U ho 60:.266 0~§~U amok 66666636666§6666m in 66666966 6.66.6666 666.6630 66.666669. 66.66660 669363 66.26 0 6636366666”.— emcee—xxofisbm 666666636665 666666696 oncomemoxo omoom_5mwm._0 oaoom—g—O>GOU GNDOMGSOEEOU vacuum—069800 oncommmgnv 686668.60 666666660 :6 an an R on mm 6m mm um Em an an an bu 6N mu wanna" 5 8666.5 6:56 66:66:66 "3.. 633. 202 X X X X on 3 m— X X XXXXX X X X 5 6— m— X XXXX X 3 n— X XXXX «— XXXX 3 XXXXXX X X X X XXXXXX X X X XXXXXX XXXXX X X XXXXX X XXXXX XX 66666666663 366v. 66666 26.686 66.66 6.6666666X 6.666666 6666666364 6666666666 66566.6 6.6666666 6.66663 666666.866 6.36% 6366666666 366666666 666666366 dmm 6666666666666 366666663 686 6666666666 66366666 666.63.666M 6636.666 66.6666 6666666630 66666666 66MM6~3 666666.663 66666360 666666 6666666360 66663666669 6666663660 66666666 6.366666666Q 666.36 666660 6.6.6636 666660 666666.66 666660 666.6666 666660 dam 6666666666 0 666666.666 6669666666 0 6666666666 666666660 666666666\666E6666&6V 6.66666 .V .8 666666666 6666666669666 6666666666 6.666V 8.36m osmosemmoq 66666663 omooatog 6686666863 203 668636ofism 6:5... 66:66:66 "3.. 63:. X X X X 3. an um um wm mm vn mm an mm X X X X X an a" an bu on .3228me 3.5 V ”NEQEKEEQU 32.3 chafing 23%sz 3808?? x ukhzuumhoh uhAEum. x 3332 Ema: EoRifid» 3.53% EESEAQ ascsim SB§~§M§ d8 umEoEE‘SE mafincsfimk ooom chosfihnusmt 23235 x unobmbadm 930th x x 3395 «mfizunfiab $83.3 cmmmék $.83qu Sfexmam x 883 Sausxmam Eminuun EmfiufiuBQ x 33m? negro x MEEE .856 L883» 2896 x x 32:53 5:95 dam quumuEuab 2532.8 muAanEuSU 3522336 53.82qu x QB-§8\DEE§§QLV .555. x x .V ho 38:33 GEESFQKV £353.: 33% X X X ms an E 83on 8858334 838:3 088528.0— omuommgoommnm 204 oaooflfionasm Esau defiance "3‘ «Ban. cN @— XXXX @— : e— X n— N— _— 3 X X X X X X NXXXX uEzum micnogm auto‘ufimom .23-».5wza Efinozuumfii flathmmcfigg 533i 333.3 SEEK m_~m o¢om whom «Von oven NNoA o_om m_om econ Noam _ o?— ngmuouafi utmuuamq $3.58 autmufiat Ezzfitummfiuu SSARSQEmQI dam 2333b .% EamheswmQ “.3533 9.35:3:qu Emmi bnoEQ EBXSS 3.25820 sweats 33.2300 dm uEwmohth uahtoa 553330 5.53 303‘ 3:285Q393V 83on 38:?an 2:5... 626355 "3‘ 033. 205 av QM an hm en mm vn X mm an 3 cm a" an us on X vs 6335 unmask chtov‘uunom £35393 Eamacauumfii mmwzmtmmcfiuau 533$ 833qu 83qu 2 E coon— whom «2E oven NNom Son mam econ Noon _ com UNEQmuouam‘ $3.033 $3.33 3535A: EchuumEuu BEARoSEmam dam 3283b .9. 23.39:qu ”Sums? 5:38.:qu £me RESmQ SNQSS KISSED 3:923 utmfaob dm 3539580 0%.th ESQEQQU can? 589V uzumwmngfifiouuw 8625 3858534 2:5... 603533 £< 03.2. 206 x x x 835$.» SSE 83802 x x 2223qu 323,925 x x x x x x x x x x 8.53 355.25 x 28%:qu Snaumim vacuum—DE x omonm x x x x x x x Scam x x x x x 395 x x x x x x vvonm Son. x x 38 x x x x x x x x x econ x one x mmom x 2cm x x x x x x x x x x x ”Sm x x Eazuutmfiu $3.833: §Nu=mo~xm 3&th flammabu Efiflt .3 £335.86 x Eden; 3.8.6»me 288332 EEGSMSE» =02§§EM=M 3.333% abmaoxuumt uno~§§~m 5.43.055 «consign—«E Sfiuwabh utmngcm x x 98> 3:“ 03825 awnmutotom aufiomuN x x x x x 8.55 “snatcfiuk x flaw 325% x x x x x uzénbom 23%. 3858534 cu a w— 2 e— m— z 2 Nu : 3 a w b e n v n N u .83on .mzfiam X X X X X X X X X X XX X XX XX X X X X X X X XXXX XX X X XXXX X X X X X X XXXX XXXX X XXXX X 602550 "3‘ «Bars 207 3. mm mm hm on mm vn X X mm NM 3 X X X X X X XXXXX an an wN bu on mu X X X XXXX x VN an an : safe 3: 35:33.5 aziutk SL3 saint.“ .EBMSBE 5:88.30 oncnm mvcnm wvcnm Scam _ 8m nnom oven— emcm mmom 20¢ a _ om Succutmsc E§§£§< 3.8523 nausmt 33kb,» gnfimt dm EEKAGQO 988.5 SESwum EEQfiMSEm noaémuEth. 3.353% “883063: EGEEENM Snotuzam 53365;» umxmngum. Em: no; flutmutotoa uuSomuN 8%5 ”untrufick 3? 2.2%. caémbok 358m omoogoz 088202 omoow>1w2 03833232 808:3 unmocmfiawox— 83on 2:53 dog—ESQ £4. «Bah 208 XXXXX X X X en a an S c.— 3 men chm _ mom 833.88 8338»qu 35.8on 32%983 uncutmfiu 3.285% mazmfiszvxfiazgwufi Sumfima. Sutuuxok smntsfi $33530 839835. 3:85 53.53 3.32.5 83333 323$ 23533 382 Ecumukuhmuab _moh amok amok u~§£§~mfiu .553qu €n§to§§t§ Sagas: Ezmmfiw unabaengi 3:89AM 53:8“? Sawmfiw 333R Sammam 58.585 559$ 62332.: Stumflw uEBa\u.EuM=m 96¢? E§M=m .aEEQ firmwaw 5.88:: 53qu 8325 888—0 goons—$832 388.52 88982 Esau 625980 :< 93.; 209 3. an an hm on mm X en mm an 3 on an an 6N on ma cu m" an a mmom Nmom fl mom 833qu 3333886 §S§A~ou .3538: 0803520 3232.826 5sz? m§m=~25§ £355.38 Summtmg Emanugok EmExESx 35:32030 08820 859633. 3:8: 333: Snoflm 833.93 5:3: u-ob§¢ sunk EEMuKENQcaw oaooafiwfioxz :8. 38‘ amok 3333235. EEEum SuzafionELEuLS 323mg: firmwam ancbaeuoci Sammfiw REES? 3.3»me 58on Szumwfi 53.565 firmwam organ: firmwam chamoxsEmMam Eovg Samm=m $633 Sammfiw 80an2 3.68:: 538: 83802 832$ E5!— 210 .vosauaoo "3‘ Saab an a w— : e— X m— X X : ><><><>< X n— «— X X = A: X ><><><>< >< XXXXX >< I‘KXXXX XXXXX X X XXKXXX XXXXX XXXXX X X X XXXK xxxx 36% ShoSmux Eamtfizofiumhommfk 328%on 5:389 2:821 ESSNQQQEU 323:» 35.3300 388.6?3 ”33300 mmom ENQQEEE awfinuamgw 9595 338.89 tutu?» 38880 ubxuSwohE 388.39 zws 328.30 czobuxuéh 338.80 .5338 Suméom Suuméfin Mahatma wdm ”8&6 «:3 .830 Ndm 3.30 _.am 2.3% $83.89“. omuefiaa 3.5433 SEPSQNR azutfiamau Efioawmuk 53.82800 uEthofitk 3333.4 EoSaudm 6233.5 EoKEdk uuflzmnnu EoSadk dam 55:6; ES £395 8.3mm 08355.2 «386258 oncomcombom omoosmwbom 888m oncomeBEEm 38.2qu mangoes—cinn— 833320 Esau 3:538 i 23 211 X x x mm mm vm mm mm —m an an an XXXX v" mu NN 353% 5.880% Eamtfizofinmhomms 2.8%on 3:989 3585} EaNSquEED 8.2%.5» 35.3360 Qwummxenxc uttefiob 08855.2 mmom SNAtQESE 8:33:30 umoufluombom Egg: acSouooU tutu?» BSoauoU 333E335 GQSQUUQU 2M3 333800 magnuméh 339800 «wooacowbom .5338 Sumbom canon—ambom 238.89% £98364 adv. “.ng min .230 Ndm 2.ng flaw. ”8&0 333m haunts». owners: oncomfiwmaga .8533 Sonmmmm czummamuw £898me omoofiumi E53288 cfiwtwofitk omoumoofioinm 3333.4 EShuuuk 838.5 EoSuwum 93.533 ESbfidk dam SE8; ES ”.535 888.220 9.39on E88— .wosaunoo £< «Bah. 212 Eazémozos SSAHQSEQN .3383:ch ESESSSVN Sneak £5585ch EthuNm MIAEV x Embsgfia ”thaw 8385M x uEEEE umxmfiahm. x unaxgu mucouztmmm .352: uEmnzox Mazuutotom 336% 833.9: 538% .% Ehofiéfi each x 32 3o; x 3: x £3822on Stooge: 83th Status NEE £23235 SEAS» Stutmab £33.23 uEmadxm 3982: umhosm aoucésusrm EQEPEQ U .8 2:3 388030 PERSLQR 38$.me unfié 3233‘ omooafism SEES USEEV 233.923 §£A§N 33523 3.859% 28855.2 8325 225% X X XXXX X X X X X X X X X X X X X X X X XXXXX X X X X X X X X X X X X X X XXXX X XX X XX XX X X X X X X X X X caa—w—Se—m—v—m—N—u—c—awbe X X X X IDXXXX NXXXX F‘XXXXX .uoscuaoo "3‘ «Bah 213 Var- X Ezzxxnozofi £3~AR3£33N 333333333335 E3~§3£33N 3353K £3~Ax3£=3N 3.3352333 fixAEV 3.33% 53.333 thEV omooawnm 335.235 SEEEG 33:33 3.333353% 3.33me 3533332 3.333288% 3.333% 33383.33 3.833% .% 3.333399% coon awom 2.0m amon— awunmototom 3.333333: 23.633 33.8333D mm; 33.3325 3333.33 33.83335 23333233 3533.8xm 3.33.333: 33335m 38333§33§tm 333$§3Q U .5 333 3333333U 3.333333% 3339.830 33.85 3233=V omoofinsm 333333 3mf33V 3333333 §A§N 3333.833 Euozkuk 03333523 3. an an FM on mm 3n mm an —m an an mm .3 on mm 3N MN nu mm 830% £83,.— .Bsaucoo £< «Bah 214 X X XXXXX XXXX X XXXXX X X X XXXXXXX X X X X X cm a w— .a% 3333333333~E 333M333 3333333338. 333.6585 3=3~=3< 3.33533 3333334 38333.3.5383356 33:33.8 3.530 33%.? 3§E§ 33.2.3 3.833.330 33323333 3.33.3335 3.33:3 3.353333% 3333:3533 3.3533335 38333833 33333: 3.3333333E3x 33333:»: 3.39383 323N336 333333.45 3.23%‘ Ezacbowfimm 53333. 3533333353 53.83% omom 333.233 53.33% 33333 3.23.539 53:35.33“. =3~A33§3Nm £83333 3268.33.54 omen 33:35.53 ES~§QS§U 333.23. 3.33.539: 3§E§~3Q 3.53me 3.33933 38333:»: 333.3333 23~3QAE xuxEE... ESbSSgN 86on 88385 omoowcon3u> 08885 0388035... 833:. omoofimaEn—oafi vacuum—Sohoum Seaman—om caveman—_Em omooataifiobm 888033 0883538 8883. 2:53 30:33:00 ":3 053,—. 215 333333533 3.333333: 3.333333353x33333333m 3% 333333333333 33388; 333M333 S33E33£3SM 3333333353 33333333 333533 3333333 3333.335 5333333339 3333333333> x 33533.53 3.333 3333385 333%.33 3333335 333333353... 33333.33 333333330 omoofizh 33333353 33.333333 33.33533 3.3333333 3533.35.33 3.3333335 ouoofimfinnoofi. 216 x x 333353 35333335 x x 33333333333 3.3.33.3 3833:3338m x x x 5333333333333 533.33% x x 5333333353 53333.0 333 omega—om 3333.533 333.35% omooaoflmfim 333333 35333.30 oncomtflasaobm 5333333333 33~AR33333M 53333333 333933333. 383 x 35333933 5335333530 33335335 x x x x x x x x x x x x 333.533 3.33.3333R x x x x x x x x x x x 3-§QA~3Q333353M x x x 333M333 333333333: x 333.3333 33333531 333333533 3333.33 53~AN33333N 33338.8 3 an an 3m on mm 3n mm an —m an an an 5 on m" 3N mu nu 3 33325 3:53“— .33333333 "33 333,—. x xx xxx x 3; x 83 x 3:3— x E: x x CE x SE x x x x x x x x SE x x x x x 8; x x x x x x x x SE x 8: x 32E x wmom x x x nmcm x x x x 383 x nmom x 383 x 33m x x x x x NS.“ x :9— x can Nmom x N32.— 301 x 38E x x bmo: 38$ 3335.33 53.33333 5333.836 3.3353333 5333.630 omuoaanqowaN 333.3333 3333.6 33338; X X X X X X X X X X X X X X X X X X X X X X ”XX [\XXX X IOXXX VXX NXXX v-«XXX an 3 w— 5 3 m— 3 n— 2 2 c— a 3333.; 3:533 33333333 "3.. 335. 217 X X X X X X X X X X X :3 an an FM em mm 3n mm mm _.m an an mu X X X X 3N en mu 3N MN an XXXX _N_m om~m m—_m m__m b__m v__m x o__m x ao_m _c_m cc_m 3333 ”mom wmom omen mmom vmom v—om x NMcm __om o_om Nmowm mvamm ovowfi 383 x 383 one: 3333333 5333333 53333339 x 3.335333 5333.636 8333:3333me x 333.3333 3333.5 333385 _N 33325 33:53.3 33:33:33 "3.. 333,—. 218 a" a— w— b— X e— m— v— n— «— x «mom x mac; «mom x x x one; x upon cham «5cm aha; who; _pom cue; coca woo; x x X x boon— woo; woo; x «can moon Neon coon one; x cmom x «no; «mom _mcm x x x x x x x avom x hqom 9.8 Ease: : 3 a a b e n v n n x 3.8% Esau X 602588 ”3.. «Bah 219 av on an FM on mm vm mm an —M an an an XXXX X X 6N cu mu vn mm «N ~N vwom mwom «mom owom when when ens; when when _nom one; moon «com been even noon «com mwom Noom coon omen Vmom mmom mmom _mom avom hvom 30m 565:5 86on ESE 603588 ”3. 033—. 220 an a w— b— 3 m— X 2 n— N— _— c— a w h e X _ 2:. x me. x m 5H m fl 8. mock 88. week cook x 88. mmcnw — monm x wvonm hvonm mvonm mmonm _Nonm omonm m_onm wmom x 30m 28m 38 «do; x ”mom Sam 38 much. 535:5 v m N _ 863m 55a..— .35538 “3. «Sun. 221 x M 3H m mob m 5H m ~ 8. mach 89—. x woe. x woo... ~ o8. mmcnm x x x x x x fl monm x x x 55.5 Sues—w m vonm mmonm _ Noam omonm m Sam waom Son 3cm 30m mmom ”mom gem cwom Son 850515 3. an an pm on mm vm mm um 3 an an an R cm mu VN an an un “Econ—m .553.— .voscucoo :4 03:. 222 APPENDIX 2 DOMINANCE-DIVERSITY CURVES FOR PUERTO RICAN DRY FOREST FRAGMENTS 223 83 8.5an 5333598 J 3mm mo ago Egofiubofifiaoa "fimm 9.5mm nuance—SE 02 o3 om ow ow cm 0 good - Sod T 8d aaunpunqv ”pupa I fio 224 8E 0:25 8883528 .m 8mm mo ego bmmbzubofiwfifiom "finm 953m .520 :53 ow on ow om ov cm cm 3 o _ , . Seed 1 Sod T Ed aaucpunqv ”pupa I fio 225 ONH 8E 9.53 8333898 .m 8% .«o 258 bmfichéoSwEEoQ "Gnm «Sufi 5.5 ism 2: cw 8 8 om o 585 - Sod . So aaucplmqv )[uuu r “.o 226 SE 8.25 5833558 on cam go 955 ngZVoogEEom “Yum PSME 55x52 2: 2: cm 8 3 om o - . . 58¢ - 53 m m M. 3 r :3 w n u 0.. B u a - _.o 227 8E cream 50395558 .m 86 mo 256 bapozuéocmfifiom "Qum— PEEK uoEOw—uam om“ o2 ow ow ow om o . . _ Seed 1 Sod 1 Ed aounpnnqv GAQBIOfi - fio 228 SE 3.25 5383598 .0 02m %0 958 bmflozuéocmfifiofl "Wan 95»?— uoEO :53 02 2: ow co ow cm 0 _r Seed 1 Sod 1 _od oaunplmqv ”pupa I fio 229 cm— 83 8.55m 8883558 K ozm go 955 EmmkoznéofiamEoQ "hum 9:.me nuance—cam o2 ow ow ow om o good I 50.0 m B m. 3 - :3 w n u D. 3 u m I _.o 230 8.3 8.8an 8883558 .w ozm 90 996 EmmSZVrowEEOQ "QNm PSME hear—015"“ cg cog ow ow ow om c _, _ . . Seed 1 Sod I Hod aaucpunqv ”pupa 1 fio 231 omfi 8E 8.55m 58835:me d ozm mo 0E3 bmflgmvbofiwfifioa "Qum— 9.sz SEOu—uam o3 ow om ow cm 0 Seed 1 Sod 1 Ed oouapunqv OAQBIQH - fio 232 8.2 8855 5.883558 .3 8% go 0350 3883580585550Q "afium 0.5—ma .5556 555% ad o2 ow cc o5 om o . 5 . . Sood 1 Sod I 5.0 oannplmqv M9393 233 SE 8855 5883558 .2 85 .«o 0&8 388358085an "2.?" 25w:— 8556555— omfi 2: ow ow ov om o . _ _ _ Sood I Sod T 86 aauepunqv “pupa 234 08m 885 8883558 “N5 85 mo 0350 55883580855509 "N“.Nm 9.53.5 8550:53— oS co“ ow oo 05 on o _ 5 _ Seed 1 Sod m m A. a I 56 W n n P B u a a .. _.o 235 SE 8.55 8883558 .2 85 .«o 988 Emmuozvéoawfifiom "Mada «guru .8qu Jana o2 o3 ow ow ow om o _ h P Sood - Sod m w A. a I 86 W n u n. B u m I _.o g 236 83 9.8an 5883538 .3 8% .«o 033 bmmuozuéocmfifiom "afinm charm Sauce—:5— oS o2 om ow ov om o L F Seed 1 Sod - _od aaunpunqv “pupa 237 ooE 8.53m 5383538 .2 mam no 033 bubzwéoqmfifioa fifiam «Sufi SEGA—3m omfi o3 ow ow cw on o h _ Seed 1 Sod I 56 aouupnnqv aAycpu 238 o9 8E 8.85m 5383538 .m: 8% go 256 bmmHoZnéoSwEEoQ "Sdm «Sufi pounce—cam 2: 8 8 8 om o . . . . 53o , 52 m m A. 3 I :3 av. n u D. B u a w _.o 239 8mm 9.55m 8883938 .5 85 no 255 bmmuofinéofigom unfimm 25E SEC x5e o2 o3 ow co ov om o . , . _ Sood I Hood I Ed aaunpunqv M98133 240 83 3533 3.8803538 .3 03m mo 0330 38.833-85.5809 "m“.nm charm uoEO :35— omfi o3 ow ow ov om o I _ good I 80.0 I 5.0 aaunpunqv aAynlau 241 comm 8.83m 3.8803538 .3 3% mo 0330 bmmuofiuéoaufifion "a“.nm 033E...— ..%..o€§ cs 9: 8 8 9. cm o . _ . 883 I :56 H m. m A. 3 I :3 w n u D. B u 3 9 I _.o 242 oNH cog 8.53m 3.6803538 .om 03m mo @330 383035-353:qu 5N.Nm 2&3 .85 :52 2: ow 8 o... om o good Sod Ed fio aauepunqv ”pupa 243 83 3.3033 80803538 .fim 03m mo 0330 €20>€0030580Q "HNdm 0.59m hocucxnam OS 2: ow 8 8 om o _ _ I. 88¢ I 83 m B m. 3 I :3 W. n u D. B u a I _.o 244 003m 2.83m 50803538 .NN 03m mo 0330 38003500305809 "NN.Nm 033mm,.“ noeucxaam om_ o3: ow ow ow cm 0 . . 88.0 I Sod I 86 aaunpnnqv “papa 245 8E 9.55m 8383838 .8 3mm mo ago bmfluzwéogufiaom "mnfim 0.53m 5.5.1:: as 2: ow 8 8 om o _ 88.0 1 :56 H 9 m M. 9 I :3 w n u D. 8 u a r _.o 246 o2 8E 9.53m 5883638 .vm “Em no 033 bmm~o>€éo§§om 3N.Nm 25E 3201:5— 2: cm ow ov on o H Sood T Sod m m. m. a a Ed W. m P a u a 1 <0 247 8E 3.55 8883558 .3 86 mo «>55 €Eo>€bo§£80fl "mNdm 95$“.— hour—0:5;— om_ o3 ow cw ow om o F 5006 I Sod 1 56 aanupunqv analog 248 8mm 98an 8283538 dm 8% LS 955 bmmuozwboawfifiom $N.Nm 0.5»; 8qu2:5— oe o2 ow 8 8 om o . 58¢ 1 god 8 P B m" 3 , :3 w m P 8 u a , 3 249 08M 883m 8883538 Km 85 .«o 93.330 388358088809 KNdm 23w; 8580:35— dmfi d3 ow do ow dm 0 Seed r Sod 1 3d aaunpunqv M98133 250 08¢ 8.83— 8883538 .mm 85 .«o 0830 388358288an £N.Nm charm 8.5352 8 2: 8 8 8 om o 88.0 , god I So aounpunqv OAQBIQH 251 02m 883m 8883538 .3 8% do 0830 388358888809 "aufim 85E 8.80 :35— 8 2: cm 8 8 cm o 58¢ I 88. H n. m u. r :3 w n u D. B u a I 3 252 08M 883m 8883538 dm 8% m0 0830 388358088800 dm.um 9.33m .320 is— 8 8 ex 8 S cm a I _I 88.0 I Sod H 3 w A. 3 I 8.0 w n u D. B u 3 9 I _.o 253 08m 883m 8883538 4% 8% m0 8830 8835-008880Q :QQ— 0.58m hofihOu—fié 8 8 8 8 8 om o 385 I 53. H 3 m. U. A 0 I :3 w n u p B u 3 3 I _.o 254 00E 0803& 8883538 .mm 8% .80 0830 bmmho>580§880fl "afium 83w:— 32003.8 8 8 ow 8 3 cm o _ good I 356 H m. B m. I 5.0 w n u D. B u a I 3 255 00E 883m 8883538 6m 8mm m0 8830 38358088800 "mmfim mun—ME 5:085. 8 8 8 8 8 cm o I . 388 I 88 m m A. 9 I :3 w m D. B u a I 3 256 d2 00E 883; 8883538 Km 8m .50 8:30 88835808880Q "33m charm 89—Aug—83 8 8 8 8 om o I . I :86 3 86 I 3:. aouuplmqv GAQBIOH I ~d 257 08m 083m 8883538 .mm 8% I80 0830 88388088809 "mm.mm 9.33m nocquaam omfi o2 ow ow ow cm 0 8 Seed I Sod I 5d aoucpnnqv “pupa 258 d2 08% 8.83m 8883538 .3 8m .50 8830 bmfiofinéogfifioa 6m.um «Sufi 8.080818% 8 8 8 8 8 o 38¢ I 86 3 So aauupunqv “pupa j ~d 259 es 008 088m 8883538 dv 8m m0 @830 888>6805880Q "bmfim charm Vigil It}? hocuO :89— dd_ ow do dv om o 8 _ good I Sod I 3d aaucpunqv alumna 260 08m 883m 8883538 .3 8% .80 8830 3835-88880Q unflnm 9.3mm,”— SEO 0.58 oi 2: 8 8 8 om o 58.0 W I Sod d a u a d w m. H . a I 80 W M n ,0 u » D. m B m u . 3 _ a w I S m 261 LITERATURE CITED 262 LITERATURE CITED Acevedo-Rodriguez, P. 1996. Flora of St. John: US. Virgin Islands. New York Botanical Garden, New York. Aide, T. M., J. K. Zimmerman, M. Rosario, and H. Marcano. 1996. Forest recovery in abandoned cattle pastures along an elevational gradient in northeastern Puerto Rico. Biotropica 28:537-548. Aizen, M. A., and P. F einsinger. 1994a. Forest fragmentation, pollination, and plant reproduction in a Chaco dry forest, Argentina. Ecology 75:330-351. Aizen, M. A., and P. Feinsinger. 1994b. Habitat fragmentation, native insect pollinators, and feral honey bees in Argentine "Chaco Serrano". Ecological Applications 4:378-392. Albers, H. J ., and H. J. Goldbach. 2000. Irreversible ecosystem change, species competition, and shifting cultivation. Resource and Energy Economics 22:261- 280. Aldrich, P. R., and J. L. Hamrick. 1998. Reproductive dominance of pasture trees in a fragmented tropical forest mosaic. Science 281. Aldrich, P. R., J. L. Hamrick, P. Chavarriaga, and G. Kochert. 1998. Microsatellite analysis of demographic genetic structure in fragmented populations of the tropical tree Symphonia globulifera. Molecular Ecology 7:933-944. Allen, W. F. 2000. Restoring Hawaii's dry forest. BioScience 50:1037-1041. Allen, W. F. 2001. Green phoenix: restoring the tropical forests of Guanacaste, Costa Rica. Oxford University Press, Oxford. Alvarez-Buylla, E. R., R. Garcia-Barrios, C. Lara-Moreno, and M. Martinez-Ramos. 1996. Demographic and genetic models in conservation biology: Applications and perspectives for tropical rain forest tree species. Annual Review of Ecology and Systematics 27 :3 87-421 . Anderson, J. R., H. E. E., J. T. Roach, and R. E. Witmer. 1976. A land use and land cover classification for use with remote sensor data. Geological Survey Professional Paper 964, US. Department of Interior Geological Service, Washington, DC. Andrén, A. 1994. Effects of habitat fragmentation on birds and mammals in landscapes with different proportions of suitable habitat: a review. Oikos 70:355-366. Arrhenius, O. 1921. Species and area. Journal of Ecology 9:95-99. 263 Atmar, W., and B. D. Patterson. 1993. The measure of order and disorder in the distribution of species in fragmented habitats. Oecologia 96:373-3 82. Atmar, W., and B. D. Patterson. 1995. The nestedness temperature calculator: a visual basic program, including 294 presence-absence matrices. AICS Research Inc., University Park, NM. Bascompte, J ., and R. Solé. 1995. Rethinking complexity: modelling spatiotemporal dynamics in ecology. Trends in Ecology and Evolution 10:361-366. Bawa, K. S. 1974. Breeding systems of three species of a lowland community. Evolution 28:85-92. Bawa, K. S. 1990. Plant-pollinator interactions in tropical rain forests. Annual Review of Ecology and Systematics 21:399-422. Bawa, K. S., D. R. Perry, and J. S. Beach. 1985. Reproductive biology of tropical rain forest trees. 1. Sexual systems and incompatibility mechanisms. American Journal of Botany 72:331-345. Beard, J. S. 1944. Climax vegetation of tropical America. Ecology 25: 127-158. Beard, J. S. 1955. The classification of Tropical American vegetation-types. Ecology 36:89-100. Bierregaard, R. 0., and V. H. Dale. 1996. Islands in an ever-changing sea: the ecological and socioeconomic dynamics of Amazonian rainforest fragments. Pages 187-204 in J. Schelhas and R. Greenberg, editors. Forest Patches in Tr0pical Landscapes. Island Press, Washington, DC. Bierregaard, R. O., and T. E. Lovejoy. 1989. Effects of fragmentation on Amazonian understory bird communities. Acta Amazénica 19:215-241. Birdsey, R. A., and P. L. Weaver. 1982. The forest resources of Puerto Rico. USDA Forest Service Southern Experiment Station, New Orleans, LA. Birdsey, R. A., and P. L. Weaver. 1987. Forest Area Trends in Puerto Rico. USDA Forest Service Southern Experiment Station, New Orleans, LA. Blackmore, M., and P. M. Vitousek. 2000. Cattle grazing, forest loss, and fuel loading in a dry forest ecosystem at Pu'u Wa'aWa'a Ranch, Hawai'i. Biotropica 32:625-632. Bolker, B. M., and S. W. Pacala. 1999. Spatial moment equations for plant competition: understanding spatial strategies and the advantages of short dispersal. American Naturalist 1532575-602. 264 Bond, W. J. 1994. Do mutualisms matter? Assessing the impact of pollinator and disperser disruption on plant extinction. Philosophical Transactions of the Royal Society, London B 344:83-90. Bond, W. J ., and J. J. Midgley. 2001. Ecology of sprouting in woody plants: the persistence niche. Trends in Ecology and Evolution 16:45-51. Boodram, N. 2001. The ecology of plant species on Little Tobago, Tobago, W.I. M.Phil. The University of the West Indies, St. Augustine. Boserup, E. 1964. The Conditions of Agricultural Growth: The Economics of Agrarian Change under Population Pressure. Aldine, New York. Boucher, D. H., J. H. Vandermeer, I. Granzow de la Cerda, M. A. Mallona, and I. Perfecto. 2000. Post-agricultural versus post-hurricane succession in southeastern Nicaraguan rain forest. Plant Ecology: 1-7. Brash, A. R. 1978. The history of avian extinction and forest conservation on Puerto Rico. Biological Conservation 39:97-1 l l. Breckon, G. 2000. Revision of the flora of Desecheo Island, Puerto Rico. Caribbean Journal of Science 36: 177-209. Breckon, G., D. A. Kolterman, V. Santiago-Ve'lez, and F. Lopez-Arroyo. 1998. Flora of Monito Island, Puerto Rico: Observations and new records. Caribbean Journal of Science 34:132-136. Browder, J. O. 1988. Public policy and the deforestation of the Brazilian Amazon. Pages 247-297 in R. Repetto and M. Gillis, editors. Public Policies and the Misuse of Forest Resources. Aldine, New York. Browder, J. O. 1996. Reading colonist landscapes: social interpretation of tropical forest patches in an Amazonian agricultural frontier. Pages 285-299 in J. Schelhas and R. Greenberg, editors. Forest Patches on Tropical Landscapes. Island Press, Washington, DC. Brown, J. H., and A. Kodric-Brown. 1977. Turnover rates in insular biogeography: effects of immigration on extinction. Ecology 58:445-449. Brown, S., and A. E. Lugo. 1982. The storage and production of organic matter in tropical forests and their role in the global carbon cycle. Biotropica 14:161-187. Bunnell, F. L. 1999. Let's kill a panchreston - giving fragmentation meaning. Pages vi- xiii in J. A. Rochelle, L. A. Lehman, and J. Wisniewski, editors. Forest fragmentation - wildlife and management implications. Brill Academic Publishers, Leiden, Netherlands. 265 Burgos, A., and J. M. Mass. in press. Vegetation change associated with land-use in tropical dry forest areas of Western Mexico. Agriculture, Ecosystems and Environment. Camargo, J. L. C., and V. Kapos. 1995. Complex edge effects on soil moisture and microclimate in central Amazonia forests. Journal of Tropical Ecology 11:205- 221. Canals Mora, M. E. 1990. El futuro del Bosque de Guanica como una unidad efectiva de conservacion. Acta Cientifica 4: 109-1 12. Carrothers, G. A. P. 1956. A historical review of the gravity and potential concepts of human interaction. Journal of the American Institute of Planners 22294-102. Castilleja, G. 1991. Seed germination and early establishment in a sub-tropical dry forest. Ph.D. Yale University, Ithaca Caughley, G. 1994. Directions in conservation biology. Journal of Animal Ecology 63:21 5-244. Ceballos, G. 1995. Vertebrate diversity, ecology and conservation in neotropical dry forests. Pages 195-220 in S. H. Bullock, H. A. Mooney, and E. Medina, editors. Seasonally dry tropical forests. Cambridge University Press, Cambridge. Chardon, C. E. 1927. The varietal revolution in Porto Rico. Journal of the Department of Agriculture of Porto Rico 13:9-41. Chen, J ., J. F. Franklin, and T. A. Speis. 1992. Vegetation responses to edge environments in old-growth Douglas-fir forests. Ecological Applications 2:387- 396. Chesson, P. L. 1986. Environmental variation and the coexistence of species. Pages 240- 256 in J. Diamond and T. J. Case, editors. Community ecology. Harper and Row, New York. Chesson, P. L., and R. R. Warner. 1981. Environmental variability promotes coexistence in lottery competitive systems. American Naturalist 117:923-943. Clark, C. 1951. Urban population densities. Journal of the Royal Statistical Society Al 14:490-496. Clark, J. S., M. Silman, R. Kern, E. Macklin, and J. HilleRisLambers. 1999. Seed dispersal near and far: patterns across temperate and tropical forests. Ecology 80:1475-1494. 266 Cleveland, W. S. 1979. Robust locally weighted regression and smoothing scatterplots. Journal of the American Statistical Association 74:829-836. Cleveland, W. S., and S. J. Devlin. 1988. Locally weighted regression: an approach to regression analysis by local fitting. Journal of the American Statistical Association 83:596-610. Cochrane, M. A., and W. F. Laurance. 2002. Fire as a large-scale edge effect in Amazonian forests. Journal of Tropical Ecology 18:311-325. Condit, R., S. P. Hubbell, J. V. LaFrankie, R. Sukumar, N. Manokaran, R. B. Foster, and P. 8. Ashton. 1996. Species-area and species-individual relationships for tropical trees: a comparison of three SO-ha plots. Journal of Ecology 84:549-562. Connell, J. H. 1978. Diversity in tropical rain forests and coral reefs. Science 199:1302- 1310. Coomes, D. A., and P. J. Grubb. 2000. Impacts of root competition in forests and woodlands: a theoretical framework and review of experiments. Ecological Monographs 70:171-207. Corlett, R. T. 1994. What is secondary forest? Journal of Tropical Ecology 10:445-447. Corlett, R. T., and I. M. Turner. 1997. Long-term survival in tropical forest remnants in Singapore and Hong Kong. Pages 333-346 in W. F. Laurence and R. O. Bierregaard, editors. Tropical Forest Remnants: Ecology, Management, and Conservation of Fragmented Communities. University of Chicago Press, Chicago. da Silva, J. M. C., C. Uhl, and G. Murray. 1996. Plant succession, landscape management, and the ecology of frugivorous birds in abandoned Amazonian pastures. Conservation Biology 10:491-503. Debinski, D. M., and R. D. Holt. 2000. A sm'vey and overview of habitat fragmentation experiments. Conservation Biology 14:342-355. Debussche, M., J. Escarré, and J. Lepart. 1982. Omithochory and plant succession in Mediterranean abandoned orchards. Vegetatio 48:255-266. Diamond, J. M. 1972. Biogeographic kinetics: estimation of relaxatiOn times for avifaunas of the southwest Pacific. Proceedings of the National Academy of Sciences USA 69:3199-3203. Diamond, J. M. 1975. Assembly of species communities. Pages 342-444 in M. L. Cody and J. M. Diamond, editors. Ecology and evolution of natural communities. Harvard University Press, Cambridge, MA. 267 Dietz, J. L. 1986. Economic history of Puerto Rico: institutional change and capitalist development. Princeton University Press, Princeton. Diffendorfer, J. E., M. S. Gaines, and R. D. Holt. 1995. Habitat fragmentation and the movement of three small mammals (Sigmodon, Microtus and Peromyscus). Ecology 76:827-839. Dinerstein, E., D. M. Olson, D. J. Graham, A. L. Webster, S. A. Primm, M. P. Bookbinder, and G. Ledec. 1995. A conservation assessment of the terrestrial ecoregions of Latin America and the Caribbean. The World Bank, Washington, DC. Dooley, J. L., Jr., and M. A. Bowers. 1998. Demographic responses to habitat fragmentation: experimental tests at the landscape and patch scale. Ecology 79:969-980. Dufréne, M., and P. Legendre. 1997. Species assemblages and indicator species definition: the need for an asymmetrical and flexible approach. Ecological Monographs 67:345-366. Dunevitz, V. L. 1985. Regrowth of clearcut subtropical dry forest: mechanisms of recovery and quantification of resilience. MS. Thesis. Michigan State University, East Lansing. Dunphy, B. K. 1997. The multi-stemmed growth of trees in a subtropical dry forest. M.S. Michigan State University, East Lansing. Dunphy, B. K., P. G. Murphy, and A. E. Lugo. 2000. The tendency for trees to be multiple-stemmed in tropical and subtropical dry forests: studies of Guénica forest, Puerto Rico. Tropical Ecology 41:161-167. Eamus, D. 1999. Ecophysiological traits of deciduous and evergreen woody species in the seasonally dry tropics. Trends in Ecology and Evolution 14:11-16. Endress, B. A., and J. D. Chinea. 2001. Landscape pattems of tropical forest recovery in the Republic of Palau. Biotropica 33:555-565. Erickson, H. E., E. A. Davidson, and M. Keller. 2002. Former land-use and tree species affect nitrogen oxide emissions from a tropical dry forest. Oecologia 130:297- 308. Esseen, P.-A. 1994. Tree mortality patterns after experimental fragmentation of an old- growth conifer forest. Biological Conservation 68. Ewel, J. 1980. Tropical succession: manifold routes to maturity. Biotropica 12:2-7. 268 Ewel, J. J. 1971. Experiments in arresting succession with cutting and herbicide in five tropical environments. Ph.D. University of North Carolina, Chapel Hill. Ewel, J. J ., and J. L. Whitmore. 1973. The ecological life zones of Puerto Rico and the US. Virgin Islands. USDA. Forest Service, Institute of Tropical Forestry, Rio Piedras, Puerto Rico. FAO. 1993. Forest resource assessment 1990: tropical countries. Food and Agriculture Organization of the United Nations, Rome. FAO. 1998. State of the world's forests: 1997. Food and Agriculture Organization of the United Nations, Rome. Farnsworth, E. J. 1993. Ecology of semi-evergreen plant assemblages in the Guanica dry forest, Puerto Rico. Caribbean Journal of Science 29:106-123. F ensham, R. J. 1995. Floristics and environmental relations of inland dry forest in north Queensland, Australia. Journal of Biogeography 22: 1 047-1063. F erreira, L. V., and W. F. Laurance. 1997. Effects of forest fragmentation on mortality and damage of selected trees in central Amazonia. Conservation Biology 11:797- 801. Figueroa Colon, J. C. 1996. Phytogeographical trends, centers of high species richness and endemism, and the question of extinctions in the native flora of Puerto Rico. J. C. Figueroa Colon, editor. The Scientific Survey of Puerto Rico and the Virgin Islands: an Eighty-Year Reassessment of the Islands' Natural History. New York Academy of Sciences, New York. Figueroa Colon, J. C., and R. O. Woodbury. 1996. Rare and endangered plant species of Puerto Rico and the Virgin Islands. Pages 65-71 in J. C. Figueroa Colon, editor. The Scientific Survey of Puerto Rico and the Virgin Islands: an Eighty-Year Reassessment of the Islands' Natural History. New York Academy of Sciences, New York. Fleming, T. H., and E. R. Heithaus. 1981. Frugivorous bats, seed shadows, and the structure if tropical forest. Biotropica l3 (suppl.):45-53. Forman, R. T. T. 1995. Land mosaics: the ecology of landscapes and regions. Cambridge University Press, Cambridge. Francis, J ., C. Rivera, and J. Figueroa. 1994. Toward a woody plant list for Antigua and Barbuda: past and present. General Technical Report SO - 102, United States Department of Agriculture Forest Service, Southern Forest Experiment Station, New Orleans, Louisiana. 269 Freeman, D. J. 1955. Report on the [ban of Sarawak. Government of Sarawak, Kuching. Galindo-Gonzalez, J ., S. Guevara, and V. J. Sosa. 2000. Bat- and bird-generated seed rains at isolated trees in pasture in a tropical rain forest. Conservation Biology 14: 1693-1703. Gaston, K. J ., T. M. Blackburn, and J. H. Lawton. 1997. Interspecific abundance-range size relationships: an appraisal of mechanisms. Journal of Animal Ecology 66:579-601. Genet, J. A., III. 1999a. The diversity and abundance of termites (Isoptera) in a fragmented subtropical dry forest landscape. MS. Thesis. Michigan State University, East Lansing. Genet, K. S. 1999b. The resilience of lizard communities to habitat fragmentation in dry forests of southwestern Puerto Rico. MS. Thesis. Michigan State University, East Lansing. Gentry, A. H. 1982. Patterns of Neotropical plant species diversity. Evolutionary Biology 1521-54. Gentry, A. H. 1995. Diversity and floristic composition of neotropical dry forests. Pages 146-194 in S. H. Bullock, H. A. Mooney, and E. Medina, editors. Seasonally dry tropical forests. Cambridge, Cambridge University Press. Geoghegan, J ., S. Cortina Villar, P. Klepeis, P. Macario Mendoza, Y. Ogneva- Hirnmelberger, R. Roy Chowdhury, B. L. Turner, II, and C. Vance. 2001. Modeling tropical deforestation in the southern Yt'lcatan peninsular region: comparing survey and satellite data. Agriculture, Ecosystems and Environment 85:25-46. Ghuman, B. S., and R. Lal. 1987. Effects of partial clearing on micro-climate in a humid tropical forest. Agriculture and Forest Meteorology 40: 17-29. Gilfedder, L., and J. B. Kirkpatrick. 1998. Factors influencing the integrity of remnant bushland in subhumid Tasmania. Biological Conservation 84:89-96. Gillespie, T. W. 1999. Life history characteristics and rarity of woody plants in tropical dry forest fragments of Central America. Journal of Tropical Ecology 15:63 7-649. Gillespie, T. W., A. Grijalva, and C. N. Farris. 2000. Diversity, composition, and structure of tropical dry forests in Central America. Plant Ecology 147:37-47. Gleason, H. A. 1922. On the relation between species and area. Ecology 3: 1 58-162. 270 Gleason, H. A., and M. Cook. 1926. Plant ecology of Porto Rico. Scientific Survey of Porto Rico and the Virgin Islands 7:1-173. Glyphis, J. P., S. J. Milton, and W. R. Siegfried. 1981. Dispersal of Acacia cyclops by birds. Oecologia 48: 1 38-141. Gonzalez, O. J. 1994. Dynamics of tropical dry forests in St. Lucia, West Indies. Ph.D. Dissertation. University of Michigan, Ann Arbor. Gonzalez, O. J ., and D. R. Zak. 1996. Tropical dry forests of St. Lucia, West Indies: vegetation and soil properties. Biotropica 28:618-626. Government of Grenada. 2000. Grenada: Dry Forest Biodiversity Conservation - Submitted to the World Bank Global Environmental Fund. Forestry and Parks Department, Ministry of Agriculture, Lands, Forestry and Fisheries, Government of Grenada, St. Georges, Grenada. Graham, C. H. 2001. Factors influencing movement patterns of Keel-billed Toucans in a fragmented tropical landscape in southern Mexico. Conservation Biology 15:1789-1798. Greig-Smith, P. 1952a. Ecological observations on degraded and secondary forest in Trinidad, British West Indies. 11. Structure of the communities. Journal of Ecology 40:316-330. Greig-Smith, P. 1952b. Ecological observations on degraded and secondary forest in Trinidad, British West Indies: I. General features of the vegetation. Journal of Ecology 40:283-315. Guariguata, M. R., and R. Ostertag. 2001. Neotropical secondary forest succession: changes in structural and functional characteristics. Forest Ecology and Management 148:185 - 2006. Guevara, S., and J. Laborde. 1993. Monitoring seed dispersal at isolated standing trees in tropical pastures: consequences for local species availability. Vegetatio 107/108:319-338. Guevara, S., J. Laborde, and G. Sénchez. 1998. Are isolated remnant trees in pasture a fragmented canopy? Selbyana 19:34-43. Guevara, S., J. Meave, P. Morena-Casasola, and J. Laborde. 1992. Floristic composition and structure of vegetation under isolated trees in neotropical pastures. Journal of Vegetation Science 3:65 5-664. Guevara, S., S. E. Purata, and E. Van der Maarel. 1986. The role of remnant trees in tropical secondary succession. Vegetatio 66:77-84. 271 Haila, Y. 2002. A conceptual genealogy of fragmentation research: from island biogeography to landscape ecology. Ecological Applications 12:321-334. Hanski, I., and O. Ovaskainen. 2002. Extinction debt at extinction threshold. Conservation Biology 16:666-673. Harrington, G. N., A. K. Irvine, F. H. J. Crome, and L. A. Moore. 1997. Regeneration of large-seeded trees in Australian rainforest fragments: a study of higher-order interactions. Pages 292-303 in W. F. Laurance and R. O. Bierregaard, editors. Tropical forest remnants: ecology, management and conservation of fragmented communities. University of Chicago Press, Chicago. He, F., and P. Legendre. 1996. On species-area relations. American Naturalist 148:718- 739. He, F., P. Legendre, and J. V. LaFrankie. 1996. Spatial pattern of diversity in‘a tropical rain forest in Malaysia. Journal of Biogeography 23:57-74. Hess, G. 1994. Conservation corridors and contagious disease: a cautionary note. Conservation Biology 8:256-262. Hess, G. 1996. Disease in metapopulation models: implications for conservation. Ecology 77:1617-1632. Holdridge, L. R. 1967. Life zone ecology. Tropical Science Center, San Jose', Costa Rica. Hopkins, M. S., J. G. Tracey, and A. W. Graham. 1990. The size and composition of soil seed-banks in remnant patches of three structural rainforest types in North Queensland. Australian Journal of Ecology 15:43-50. Howe, H. F. 1977. Bird activity and seed dispersal of a tropical wet forest tree. Ecology 58:539-550. Howe, H. F., and R. B. Primack. 1975. Differential dispersal by birds of the tree Casearia nitida (Flacourtiaceae). Biotropica 7:278-283. Hubbell, S. P. 1979. Tree dispersion, abundance, and diversity in a tropical dry forest. Science 203:1299-1309. Hubbell, S. P. 2001. The unified neutral theory of biodiversity and biogeography. Princeton University Press, Princeton. Hubbell, S. P., and R. B. Foster. 1986. Commonness and rarity in a Neotropical forest: implications for tropical tree conservation. Pages 205-232 in M. E. Soulé, editor. Conservation biology: the science of scarcity and diversity. Sinauer, Sunderland, Massachusetts. 272 Hubbell, S. P., R. B. Foster, S. T. O'Brien, K. E. Harms, R. Condit, B. Wechsler, S. J. Wright, and S. Loo de Lao. 1999. Light gap disturbances, recruitment limitation, and tree diversity in a neotropical forest. Science 283:554-557. Hughes, R. F ., J. B. Kauffrnan, and V. J. Jaramillo. 1999. Biomass, carbon, and nutrient dynamics of secondary forests in a humid tropical region of Me'xico. Ecology 80: 1892-1907. Imre, A. R. 2001. About the ranking of isolated habitats with different shapes: an interior- to-edge ration study. Acta Biotheoretica 49:115-120. J accard, P. 1900. Contribution au probleme de l'immigration post-glaciaire de la flore alpine. Bull. Soc. Vaudoise Sci. Nat. 36287-130. J anzen, D. H. 1983. No park is an island: increase in interference as park size decreases. Oikos 41:402-410. J anzen, D. H. 1986a. Chihuahuan Desert nopaleras: defaunated big mammal vegetation. Annual Review of Ecology and Systematics 17:595-636. Janzen, D. H. 1986b. The eternal external threat. Pages 286-303 in M. E. Soulé, editor. Conservation Biology - the Science of Scarcity and Diversity. Sinauer Associates, Sunderland, Massachusetts. J anzen, D. H. 1988a. Management of habitat fragments in a tropical dry forest: growth. Annals of the Missouri Botanical Garden 75:105-116. J anzen, D. H. 1988b. Tropical dry forests, the most endangered major tropical ecosystem. Pages 130-137 in E. O. Wilson, editor. Biodiversity. National Academy Press, Washington, DC. Jennings, S. B., N. D. Brown, D. H. Boshier, T. C. Whitmore, and J. d. C. A. Lopes. 2001. Ecology provides a pragmatic solution to the maintenance of genetic diversity in sustainably managed tropical forests. Forest Ecology and Management 154:1-10. J ipp, P. H., D. C. Nepstad, D. K. Cassel, and C. R. Carvalho. 1998. Deep soil moisture storage and transpiration in forests and pastures of seasonally-dry Amazonia. Climate Change 39:395-412. Kahn, J. R., and J. A. McDonald. 1997. The role of economic factors in tropical deforestation. Pages 13-28 in W. F. Laurance and R. O. Bierregaard, Jr., editors. Tropical Forest Remnants. University of Chicago Press, Chicago. Kapos, V. 1989. Effects of isolation on the water status of forest patches in the Brazilian Amazon. Journal of Tropical Ecology 52173-185. 273 Kapos, V., E. Wandelli, J. L. Camargo, and G. Ganade. 1997. Edge-related changes in environment and plant responses due to forest fragmentation in Central Amazonia. Pages 33-44 in W. F. Laurance and R. O. Bierregaard, editors. Tropical Forest Remnants: Ecology, Conservation and Management of Fragmented Communities. University of Chicago Press, Chicago. Kauffman, J. B., M. D. Steele, D. L. Cummings, and V. J. Jaramillo. 2003. Biomass dynamics associated with deforestation, fire, and, conversion to cattle pasture in a Mexican tropical dry forest. Forest Ecology and Management 17621-12. Kellman, M. 1996. Redefining roles: plant community reorganization and species preservation in fragmented systems. Global Ecology and Biogeography Letters 52111-116. Kelhnan, M., R. Tackaberry, and J. Meave. 1996. The consequences of prolonged fragmentation: lessons from gallery forests. Pages 37-58 in J. Schelhas and R. Greenberg, editors. Forest patches in tropical landscapes. Island Press, Washington, DC. Kellman, M., R. Tackaberry, and L. Rigg. 1998. Structure and function in two gallery forest communities: implications for forest conservation in fragmented systems. Journal of Applied Ecology 35:195-206. Kimber, C. T. 1988. Martinique revisited: the changing plant geography of a West Indian island. Texas A&M University Press, College Station, Texas. Kimura, M., and J. F. Crow. 1963. The measurement of effective population number. Evolution 17:279-288. Kirkpatrick, J. B., and L. Gilfedder. 1995. Maintaining integrity compared with maintaining rare and threatened taxa in remnant bushland in subhumid Tasmania. Biological Conservation 74: 1-8. Kleinman, P. J. A., R. B. Bryant, and D. Pimental. 1996. Assessing ecological sustainability of slash-and-bum agriculture through soil fertility indicators. Agronomy Journal 88:122-127. Kramer, E. A. 1997. Measuring landscape change in a remnant tropical dry forest. Pages 386-399 in W. F. Laurence and R. O. Bierregaard, editors. Tropical forest remnants: ecology, management and conservation of fragmented communities. University of Chicago Press, Chicago. Kruger, L. M., and J. J. Midgley. 2001. The influence of resprouting forest canopy species on richness in Southern Cape forest, South Afiica. Global Ecology and Biogeography 10:567-572. 274 Kruger, L. M., J. J. Midgley, and R. M. Cowling. 1997. Resprouters vs reseeders in South African forest trees: a model based on forest canopy height. Functional Ecology 1 1 : l 01 -105. Lamb, D., J. Parrotta, R. Keenan, and N. Tucker. 1997. Rejoining habitat remnants: restoring degraded rainforest lands. Pages 366-3 85 in W. F. Laurance and R. O. Bierregaard, Jr., editors. Tropical forest remnants: ecology, management and conservation of fragmented communities. University of Chicago Press, Chicago. Lande, R. 1995. Mutation and conservation. Conservation Biology 92782-791. Laurance, W. F. 1991. Edge effects in tropical forest fragments: Application of a model for the design of nature reserves. Biological Conservation 57:205-219. Laurance, W. F. 1997. Hyper-disturbed parks: edge effects and the ecology of isolated rainforest reserves in tropical Australia. Pages 71-84 in W. F. Laurence and R. O. Bierregaard, editors. Tropical forest remnants: ecology, management and conservation of fragmented communities. University of Chicago Press, Chicago. Laurance, W. F. 1999. Reflections on the tropical deforestation crisis. Biological Conservation 91:109-117. Laurance, W. F. 2002. Hyperdynamism in fragmented habitats. Journal of Vegetation Science 13:595-602. Laurance, W. F ., A. K. M. Albemaz, G. Schroth, P. M. Fearnside, B. Scott, E. M. Venticinque, and C. Da Costa. 2002a. Predictors of deforestation in the Brazilian Amazon. Journal of Biogeography 29:737-748. Laurance, W. F ., L. V. F erriera, J. M. Rankin-dc Merona, and S. G. Laurance. 1998. Rain forest fragmentation and the dynamics of Amazonian tree communities. Ecology 79:2032-3040. Laurance, W. F ., T. E. Lovejoy, H. Vasconcelos, E. M. Bruna, R. K. Didham, P. C. Stouffer, C. Gascon, R. O. Bierregaard, S. G. Laurance, and E. Sarnpaio. 2002b. Ecosystem decay in Amazonian forest fragments: A 22-year investigation. Conservation Biology 16:605-618. Laurance, W. F., D. Pérez-Salicrup, P. Delaménica, P. M. Fearnside, S. D'Angelo, A. J erozolinski, L. Pohl, and T. E. Lovejoy. 2001. Rain forest fiagmentation and the structure of Amazonian liana communities. Ecology 82:105-116. Lawrence, D., D. R. Peart, and M. Leighton. 1998. The impact of shifting cultivation on a rainforest landscape in West Kalimantan: spatial and temporal dynamics. Landscape Ecology 13:135-148. 275 Lebbie, A. R., and M. S. F reudenberger. 1996. Sacred groves in Afiica: forest patches in transition. Pages 300-324 in J. Schelhas and R. Greenberg, editors. Forest patches on tropical landscapes. Island Press, Washington, DC. Legendre, P., and L. Legendre. 1998. Numerical Ecology, Second edition. Elsevier, Amsterdam. Leigh, E. G., Jr., S. J. Wright, E. A. Herre, and F. E. Putz. 1993. The decline of tree diversity on newly isolated tropical islands: a test of a null hypothesis and some implications. Evolutionary Ecology 7:76-102. Leishman, M. R., and M. Westoby. 1994. The role of seed size in seedling establishment in dry soil conditions - experimental evidence from semi-arid species. Journal of Ecology 82:249-258. Leitner, W. A., and M. L. Rosenzweig. 1997. Nested species-area curves and stochastic sampling: a new theory. Oikos 79:503-512. Lerdau, M., J. Whitbeck, and N. M. Holbrook. 1991. Tropical deciduous forest: death of a biome. Trends in Ecology and Evolution 6:201-202. Lindenmayer, D. B., and H. P. Possingham. 1996. Modelling the inter-relationships between habitat patchiness, dispersal capability and metapopulation persistence of the endangered species, Leadbeater's possum, in south-eastern Australia. Landscape Ecology 11:79-105. Liogier, H. A. 1985. Descriptive flora of Puerto Rico and adjacent islands: Spermatophyta. Volume I: Casuarinaceae to Connaraceae. Editorial de la Universidad de Puerto Rico, Rio Piedras, Puerto Rico. Liogier, H. A. 1988. Descriptive flora of Puerto Rico and adjacent islands: Spermatophyta. Volume II: Leguminosae to Anacardiaceae. Editorial de la Universidad de Puerto Rico, Rio Piedras, Puerto Rico. Liogier, H. A. 1994. Descriptive flora of Puerto Rico and adjacent islands: Spermatophyta. Volume III: Cyrillaceae to Myrtaceae. Editorial de la Universidad de Puerto Rico, Rio Piedras, Puerto Rico. Liogier, H. A. 1995. Descriptive flora of Puerto Rico and adjacent islands: Spermatophyta. Volume IV: Melastomataceae to Lentibulariaceae. Editorial de la Universidad de Puerto Rico, Rio Piedras, Puerto Rico. Liogier, H. A. 1997. Descriptive flora of Puerto Rico and adjacent islands: Spermatophyta. Volume V: Acanthaceae to Compositae. Editorial de la Universidad de Puerto Rico, Rio Piedras, Puerto Rico. 276 Livingston, R. B. 1972. Influence of birds, stones, and soil on the establishment of pasture juniper Juniperus communis, and red cedar, J. virginiana in New England pasture. Ecology 53:1141-1147. Lomolino, M. V. 2000. Ecology's most general, yet protean pattern: the species-area relationship. Journal of Biogeography 27:17-26. Lomolino, M. V., and M. D. Weiser. 2001. Towards a more general species-area relationship: diversity on all islands, great and small. Journal of Biogeography 28:431-445. Lopez, T. d. M., T. M. Aide, and J. R. Thomlinson. 2001. Urban expansion and the loss of prime agricultural lands in Puerto Rico. Ambio 30:49-54. Lovejoy, T. E., R. O. Bierregaard, Jr., A. B. Rylands, J. R. Malcolm, C. E. Quintela, L. H. Harper, K. S. Brown, Jr., A. H. Powell, G. V. N. Powell, H. O. Schubart, and M. B. Hays. 1986. Edge and other effects of isolation on Amazon forest fragments. Pages 257-285 in M. E. Soulé, editor. Conservation biology: the science of scarcity and diversity. Sinauer, Sunderland, Massachusetts. 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. Lugo, A. E., and P. G. Murphy. 1986. Nutrient dynamics of a subtropical dry forest. Journal of Tropical Ecology 2:55-72. Lugo, A. E., 0. Ramos, S. Molina, F. N. Scatena, and L. L. Vélez Rodriguez. 1996. A fifty-three year record of land use change in the Guanica Forest Biosphere Reserve and its vicinity. USDA Forest Service Intemational Institute of Tropical Forestry, Rio Piedras, Puerto Rico. MacArthur, R. H., and E. O. Wilson. 1967. The theory of island biogeography. Princeton University Press, Princeton. MacDougall, A., and M. Kellman. 1992. The understorey light regime and patterns of tree seedlings in tropical riparian forest patches. Journal of Biogeography 19:667- 675. . Martinez-Garza, C., and R. Gonzzilez—Montagut. 1999. Seed rain from forest fragments into tropical pastures in Los Tuxtlas, Mexico. Plant Ecology 145:255-265. Matlack, G. R. 1994. Vegetation dynamics of the forest edge: Trends in space and successional time. Journal of Ecology 82:113-123. 277 May, R. M. 1975. Patterns of species abundance and diversity. Pages 81-120 in M. L. Cody and J. M. Diamond, editors. Ecology and evolution of communities. Belknap Press of Harvard University Press, Cambridge, Massachusetts. McClanahan, T. R., and R. W. Wolfe. 1987. Dispersal of omithochorous seeds from forest edges in central Florida. Vegetatio 71:107-112. McClanahan, T. R., and R. W. Wolfe. 1993. Accelerating forest succession in a fragmented landscape: the role of birds and perches. Conservation Biology 7:279- 288. McDonnell, M. J ., and E. W. Stiles. 1983. The structural complexity of old field vegetation and the recruitment of bird-dispersed plant species. Oecologia 56:109- 116. McGarigal, K., and W. C. McComb. 1995. Relationships between landscape structure and breeding birds in the Oregon Coast Range. Ecological Monographs 65:23 5- 260. McLaren, K. P., and M. A. McDonald. 2003. The effects of moisture and shade on seed germination and seedling survival in a tropical dry forest in Jamaica. Forest Ecology and Management 183:61-75. Medina, E., V. Garcia, and E. Cuevas. 1990. Sclerophylly and oligotrophic environments: relationships between leaf structure, mineral nutrient content, and drought resistance in tropical rain forests of the Upper Rio Negro Region. Biotropica 22:51-64. Mesquita, R. C. G., P. Delaménica, and W. F. Laurance. 1999. Effects of surrounding vegetation on edge-related tree mortality in Amazonian forest fragments. Biological Conservation 91:129-134. Mesquita, R. C. G., K. Ickes, G. Ganade, and G. B. Williamson. 2001. Alternative successional pathways in the Amazon Basin. Journal of Ecology 89:528-537. Mesquita, R. d. C. G. 2000. Management of advanced regeneration in secondary forests of the Brazilian Amazon. Forest Ecology and Management 130: 1 3 1-140. Molina Colon, S. 1998. Long-term recovery of a Caribbean dry forest after abandonment of different land uses in Guanica, Puerto Rico. Ph.D. University of Puerto Rico, Rio Piedras. Moreira, M. Z., L. d. S. L. Stemberg, and D. C. Nepstad. 2000. Vertical patterns of soil water uptake by plants in a primary forest and an abandoned pasture in the eastern Amazon: an isotopic approach. Plant and Soil 222:95-107. 278 Morrison, L. W. 2002. Determinants of plant species richness on small Bahamian islands. Journal of Biogeography 29:931-941. Mtiller, F., R. Hoffmann-Kroll, and H. Wiggering. 2000. Indicating ecosystem integrity - theoretical concepts and environmental requirements. Ecological Modelling 130:13-23. Murcia, C. 1995. Edge effects in fragmented forests: implications for conservation. Trends in Ecology and Evolution 10:58-62. Murphy, L. S. 1916. Forests of Porto Rico: past, present and future, and their physical and economic environment. USDA. Murphy, P. G., and T. M. Burton. 1993. Islands of biodiversity on altered tropical landscapes. Research Proposal submitted to the USDA Forest Service International Institute of Tropical Forestry, Rio Piedras, Puerto Rico. Murphy, P. G., and A. E. Lugo. 1986a. Ecology of tropical dry forest. Annual Review of Ecology and Systematics 17 :67-88. Murphy, P. G., and A. E. Lugo. 1986b. Structure and biomass of a subtropical dry forest in Puerto Rico. Biotropica 18:89-96. Murphy, P. G., and A. E. Lugo. 1990. Dry forests of the tropics and subtropics: Guanica Forest in context. Acta Cientifica 4:15-24. Murphy, P. G., and A. E. Lugo. 1995. Dry forests of Central American and the Caribbean. Pages 9-34 in S. H. Bullock, H. A. Mooney, and E. Medina, editors. Seasonally dry tropical forests. Cambridge University Press, Cambridge. Murphy, P. G., A. E. Lugo, A. J. Murphy, and D. C. Nepstad. 1995. The dry forests of Puerto Rico's south coast. Pages 178-209 in A. E. Lugo and C. Lowe, editors. Tropical Forests: Management and Ecology. Springer-Verlag, New York. Nason, J. D., P. R. Aldrich, and J. L. Hamrick. 1997. Dispersal and dynamics of genetic structure in fragmented tropical tree populations. Pages 304-320 in W. F. Laurence and R. O. Bierregaard, editors. Tropical forest remnants: ecology, management and conservation of fragmented communities. University of Chicago Press, Chicago. National Oceanic and Atmospheric Administration. 2001. Monthly station normals of temperature, precipitation, and heating and cooling degree days 1971-2000: 66 Puerto Rico. National Oceanic and Atmospheric Administration, US. Department of Commerce, Asheville, North Carolina. 279 Nepstad, D., G. Carvalho, A. C. Barros, A. Alencar, J. P. Capobianco, J. Bishop, P. Moutinho, P. Lefebvre, U. L. Silva, Jr., and E. Prins. 2001. Road paving, fire regime feedbacks, and the future of Amazon forests. Forest Ecology and Management 154:395-407. Nepstad, D. C., C. R. Carvalho, E. A. Davidson, P. H. Jipp, P. A. Lefebvre, G. H. Negreiros, E. D. Silva, T. A. Stone, S. E. Trumbore, and S. Vieira. 1994. The role of deep roots in the hydrological and carbon cycles of Amazonian forests and pastures. Nature 372:666-669. Nepstad, D. C., A. Verissimo, A. Alencar, C. A. Nobre, E. Lima, P. Lefebvre, P. Schlesinger, C. Potter, P. R. S. Moutinho, E. Mendoza, M. Cochrane, and V. Brooks. 1999. Large-scale impoverishment of Amazonian forests by logging and fire. Nature 398:505-508. Neve, G., B. Barascud, R. Hughes, J. Aubert, H. Descimon, P. Lebrun, and M. Baguette. 1996. Dispersal, colonization power and metapopulation structure in the vulnerable butterfly Proclossiana eunomia. Journal of Applied Ecology 33: 14-22. Ortiz-Pulido, R., J. Laborde, and S. Guevara. 2000. Frugivorr'a por aves en un paisaje fragmentado: consecuencias en la dispersion de semillas. Biotropica 32:473-488. Ozorio de Almeida, A. L. 1992. The colonization of the Amazon. University of Texas Press, Austin. Palik, B. J ., and P. G. Murphy. 1990. Disturbance versus edge effects in sugar- maple/beech forest fragments. Forest Ecology and Management 32: 187-202. Parrotta, J. A., O. H. Knowles, and J. M. Wunderle, Jr. 1997. Development of floristic diversity in 10-year-old restoration forests on a bauxite mined site in Amazonia. Forest Ecology and Management 99:21-42. Patterson, B. D., and W. Atmar. 1986. Nested subsets and the structure of insular mammalian faunas and archipelagos. Biological Journal of the Linnean Society 28:65-82. Patterson, B. D., and W. Atmar. 2000. Analyzing species composition in fragments. Pages 9-24 in G. Rheinwald, editor. Isolated Vertebrate Communities in the Tropics; Proc. 4th Intl. Symp. Bonn. zool. Monogr. 46, Bonn. Pennington, R. T., D. E. Prado, and C. A. Pendry. 2000. Neotropical seasonally dry forests and Quaternary vegetation changes. Journal of Biogeography 27:261-273. Pither, R., and M. Kellman. 2002. Tree species diversity in small riparian forest fragments in Belize, Central America. Biodiversity and Conservation 1121623- 1636. 280 Pittman, N. C. A., J. Terborgh, M. R. Silman, and P. Nufiez V. 1999. Tree species distribution in an upper Amazonian forest. Ecology 80:2651-2666. Plotkin, J. B., M. D. Potts, N. Leslie, N. Manokaran, J. LaFrankie, and P. S. Ashton. 2000. Species-area curves, spatial aggregation, and habitat specialization in tropical forests. Journal of Theoretical Biology 207:81-99. Preston, D. 1989. Too busy to far: underutilization of farmland in central Java. Journal of Development Studies 26:43-57. Preston, F. W. 1962. The canonical distribution of commonness and rarity. Ecology 43:185-215. Quevedo, V., S. Silander, and R. O. Woodbury. 1990. Plantas criticas en peligro de extincién en el bosque de Guénica. Acta Cientifica 4: 137-1 50. Quigley, M. F. 1994. Latitudinal gradients in temperate and tropical seasonal forests. Ph.D. Louisiana State University, Baton Rouge. Quifiones, F. 1992. History of hurricanes in Puerto Rico 1502-1989. Acta Cientifica 6:3- 14. Ramjohn, K., F. B. Lucas, C. L. Ramjohn, and I. A. Ramjohn. 2002a. Environmental Impact Assessment for proposed exploratory drilling, and expansion, modification and maintenance of existing production wells: Morne Diablo F armout Block, Siparia. EMA Ref. CEC 0058/2001. Environmental Management Agency, Port- of-Spain, Trinidad & Tobago. Ramjohn, K., F. B. Lucas, C. L. Ramjohn, I. A. Ramjohn, and S. Manickchan. 2002b. Environmental Impact Assessment for proposed drilling, work-over and production operations: Blocks PS-l, WD-l, WD-S and WD-6, Palo Seco. EMA Ref. CEC 0019/2001. Environmental Management Agency, Port-of-Spain, Trinidad & Tobago. Ramjohn, K., F. B. Lucas, C. L. Ramjohn, I. A. Ramjohn, A. Seepersad, and A. Braxton. 2003. Environmental Impact Assessment: Exploration Drilling, Production and Maintenance Operations: South Quarry Farmout Block. EMA Ref. CEC 0421/2003. Environmental Management Agency, Port-of-Spain, Trinidad & Tobago. Ramos Gonzalez, O. M. 2001. Assessing vegetation and land cover changes in northeastern Puerto Rico: 1978-1995. Caribbean Journal of Science 37:95-106. 281 Ranta, P., T. Blom, J. Niemela, E. Joensuu, and M. Siitonen. 1998. The fragmented Atlantic rain forest of Brazil: size, shape and distribution of forest fragments. Biodiversity and Conservation 7 :385-403. Ray, G. J. 1993. The ecological restoration of a Caribbean dry forest in the island of St. John, US. Virgin Islands. Ph.D. University of Wisconsin, Madison. Ray, G. J ., and B. Brown. 1995. The structure of five successional stands in a subtropical dry forest, St. John, US. Virgin Islands. Caribbean Journal of Science 31:212- 222. Ray, G. J ., and B. J. Brown. 1994. Seed ecology of woody species in a Caribbean dry forest. Restoration Ecology 2: 1 56-163. Ricart Morales, C. M. 1999. Frugivory by birds on two subtropical dry forest species (Almacigo: Bursera simaruba and Guayacan: Guaiacum oflicinale) in Guanica, Puerto Rico. Ph.D. Dissertation. University of Colorado at Boulder, Boulder. Rosenzweig, M. L. 1995. Species diversity in space and time. Cambridge University Press, Cambridge. Rosenzweig, M. L., and Y. Ziv. 1999. The echo pattern of species diversity: pattern and process. Ecography 22:614-628. Ross, K., B. J. Fox, and M. D. Fox. 2002. Changes to plant species richness in forest fragments: Fragment age, disturbance and fire history may be as important as area. Journal of Biogeography 29:749-765. Roth, L. C. 1999. Anthropogenic change in a subtropical dry forest during a century of settlement in Jaiqur’ Picado, Santiago Province, Dominican Republic. Journal of Biogeography 26:739-759. Roth, L. C. 2001. Enemies of the trees? Subsistence farmers and perverse protection of tropical dry forest. Journal of Forestry 99:20-27. Rudel, T. K., D. Bates, and R. Machinguiashi. 2002. A tropical forest transition? Agricultural change, out-migration, and secondary forests in the Ecuadorian Amazon. Annals of the Association of American Geographers 92:87-102. Rudel, T. K., M. Perez-Lugo, and H. Zichal. 2000. When fields revert to forest: development and spontaneous reforestation in post-war Puerto Rico. Professional Geographer 52:386-397. Santiago, C. E. 1992. Labor in the Puerto Rican economy: postwar development and stagnation. Praeger, New York. 282 Sarmiento, G. 1972. Ecological and floristic convergences between seasonal plant formations of Tropical and Subtropical South America. Journal of Ecology 60:367-410. Savill, P. S. 1983. Silviculture in windy climates. Forestry Abstracts 44:473-488. Scheiner, S. M., S. B. Cox, M. Willig, G. G. Mittelbach, C. Osenberg, and M. Kaspari. 2000. Species richness, species-area curves and Simpson's paradox. Evolutionary Ecology Research 2:791-802. Schimper, A. F. W. 1903. Plant geography upon a physiological basis, Oxford. Schoener, T. W. 1974. The species-area relationship within archipelagos: models and evidence from island land birds. Proceedings of the International Ornithological Congress 16:629-642. Shmida, A., and M. V. Wilson. 1985. Biological determinants of species diversity. Journal of Biogeography 12:1-20. Silander, S. 1999. Endangered and threatened wildlife and plants: determination of endangered status for Catesbaea melanocarpa. Federal Register 64: 131 16-13 120. Sim, J. W. S., H. T. W. Tan, and I. M. Turner. 1992. Andinandra belukar: an anthropogenic heath forest in Singapore. Vegetatio 1021125-137. Simberloff, D. S., and L. G. Abele. 1979. Island biogeography theory and conservation practice. Science 191:285-286. Simberloff, D. S., and L. G. Abele. 1982. Refuge design and island biogeography theory: the effects of fragmentation. American Naturalist 120:41-56. Skole, D. L., and C. Tucker. 1994. Tropical deforestation and habitat loss fragmentation in the Amazon: satellite data from 1978-1988. Science 260:1905-1910. Smith, A. P. 1997. Deforestation, fiagmentation, and reserve design in Western Madagascar. Pages 415-441 in W. F. Laurance and R. O. Bierregaard, Jr., editors. Tropical forest remnants: ecology, management, and conservation of fragmented communities. University of Chicago Press, Chicago. Snow, D. W. 1962. The natural history of the Oilbird, Steatorm's caripensis, in Trinidad, W.I. Part 2. Population breeding ecology and food. Zoologica 47:199-221. Steininger, M. K., C. J. Tucker, P. Ersts, T. J. Killeen, Z. Villegas, and S. B. Hecht. 2001a. Clearance and fragmentation of tropical deciduous forest in the Tierras Bajas, Santa Cruz, Bolivia. Conservation Biology 15:856-866. 283 Steininger, M. K., C. J. Tucker, J. R. G. Townshend, T. J. Killeen, A. Desch, V. Bell, and P. Ersts. 2001b. Tropical deforestation in the Bolivian Amazon. Environmental Conservation 28: 1 27-1 34. Thomas, S. C. 2004. Ecological correlates of tree species persistence in tropical forest fragments. E. Losos and E. Leigh, editors. Forest diversity and dynamism: results from large-scale demographic plots. University of Chicago Press, Chicago, in press. Tilman, D., C. L. Lehman, and C. Yin. 1997. Habitat destruction, dispersal, and deterministic extinction in competitive communities. American Naturalist 149:407-435. Tilman, D., R. M. May, C. L. Lehman, and M. A. Nowak. 1994. Habitat destruction and the extinction debt. Nature 371 :65. Toh, 1., M. Gillespie, and D. Lamb. 1999. The role of isolated trees in facilitating seedling recruitment at a degraded sub-tropical rainforest site. Restoration Ecology 7 :288-297. Tole, L. 2002. Habitat loss and anthropogenic disturbance in J amaica's Hellshire Hills area. Biodiversity and Conservation 11:575-598. Tosi, J ., and R. F. Voertman. 1964. Some environmental factors in the economic development of the tropics. Economic Geography 40:189-205. Trejo, 1., and R. Dirzo. 2000. Deforestation of seasonally dry tropical forest: a national and local analysis in Mexico. Biological Conservation 94:133-142. Trejo, I., and R. Dirzo. 2002. F loristic diversity of Mexican seasonally dry tr0pical forests. Biodiversity and Conservation 11:2063-2084. Turner, B. L., 11, S. Cortina Villar, D. Foster, J. Geoghegan, E. Keys, P. Klepeis, D. Lawrence, P. Macario Mendoza, S. Manson, Y. Ogneva-Himmelberger, A. B. Plotkin, D. Pérez-Salicrup, R. Roy Chowdhury, B. Savitsky, L. Schneider, B. Schmook, and C. Vance. 2001. Deforestation in the southern Yucatan peninsular region: An integrative approach. Forest Ecology and Management 154:353-3 70. Turner, 1. M. 1996. Species loss in fragments of tropical rain forest: a review of the evidence. Journal of Applied Ecology 33:200-209. Turner, 1. M., K. S. Chua, J. S. Y. Ong, B. C. Soong, and H. T. W. Tan. 1996. A century of plant species loss from an isolated fragment of lowland tropical rain forest. Conservation Biology 10: 1229-1 244. 284 Turner, 1. M., and R. T. Corlett. 1996. The conservation value of small, isolated fragments of lowland tropical rain forest. Trends in Ecology and Evolution 1 1:3 30-3 3 3. Turner, 1. M., Y. K. Wong, P. T. Chew, and Ali bin Ibrahim. 1997. Tree species richness in primary and old secondary tropical forest in Singapore. Biodiversity and Conservation 6:537-543. Turton, S. M., and H. J. Freiburger. 1997. Edge and aspect effects on the microclimate of a small tropical forest remnant on the Atherton Tableland, northeastern Australia. Pages 45-54 in W. F. Laurance and R. O. Bierregaard, Jr., editors. Tropical forest remnants: ecology, management and conservation of fragmented communities. University of Chicago Press, Chicago. US. Fish and Wildlife Service. 1994a. Aristida chaseae, Lyonia truncata var. proctorii, and Vernonia proctorii recovery plan. US. Fish and Wildlife Service, Atlanta, Georgia. US. Fish and Wildlife Service. 1994b. Aristida portoricensis recovery plan. US. Fish and Wildlife Service, Atlanta, Georgia. Uhl, C., and R. Buschbacher. 1985. A disturbing synergism between cattle-ranch burning practices and selective harvesting in eastern Amazon. Biotropica 17:265-268. Uhl, C., R. Buschbacher, and E. A. S. Serrfio. 1988. Abandoned pastures in Eastern Amazonia. 1. Patterns of plant succession. Journal of Ecology 76:663-681. Uhl, C., H. Clark, K. Clark, and P. Maquirino. 1982. Successional patterns associated with slash-and-burn agriculture in the Upper Rio Negro region of the Amazon Basin. Biotropica 14:249-254. Uhl, C., K. Clark, H. Clark, and P. Murphy. 1981. Early plant succession after cutting and burning in the upper Rio Negro region of the Amazon Basin. Ecology 69. Van Bloem, S. J ., S. Molina Colon, A. E. Lugo, P. G. Murphy, R. Ostertag, M. Rivera- Costa, 1. Ruiz-Bemard, and M. Canals Mora. in press. Hurricane effects on subtropical dry forest. Biotropica. Van Bloem, S. J ., P. G. Murphy, and A. E. Lugo. 2003. Subtropical dry forest trees with no apparent damage sprout following a hurricane. Tropical Ecology 44:1-9. Vandermeer, J. H., D. Boucher, I. Perfecto, and I. Granzow de la Cerda. 1996. A theory of disturbance and species diversity: evidence from Nicaragua after Hurricane Joan. Biotropica 28:600-613. 285 Vandermeer, J. H., D. H. Boucher, I. Granzow de la Cerda, and I. Perfecto. 2001. Growth and development of the thinning canopy in a post-hurricane tropical rain forest in Nicaragua. Forest Ecology and Management 148:221-242. Vane-Wright, R. 1., C. J. Humphries, and P. H. Williams. 1991. What to protect? - Systematics and the agony of choice. Conservation Biology 55:235-354. Vazquez, O. J ., and D. A. Kolterman. 1998. Floristic composition and vegetation types of the Punta Guaniquilla Natural Reserve - Cabo Rojo, Puerto Rico. Caribbean Journal of Science 34:265-279. Vélez Rodriguez, L. L. 1995a. Land use and land cover 1936 Guanica Commonwealth Forest. USDA. Forest Service International Institute of Tropical Forestry, Rio Piedras, Puerto Rico. Vélez Rodriguez, L. L. 1995b. Land use and land cover 1950-1951 Gua'mica Commonwealth Forest. USDA. Forest Service International Institute of Tropical Forestry, Rio Piedras, Puerto Rico. Ve'lez Rodriguez, L. L. l995c. Land use and land cover 1963 Guénica Commonwealth Forest. USDA. Forest Service International Institute of Tropical Forestry, Rio Piedras, Puerto Rico. Vélez Rodriguez, L. L. 1995d. Land use and land cover 1971 Guanica Commonwealth Forest. USDA. Forest Service International Institute of Tropical Forestry, Rio Piedras, Puerto Rico. Vélez Rodriguez, L. L. l995c. Land use and land cover 1983 Guanica Commonwealth Forest. USDA. Forest Service International Institute of Tropical Forestry, Rio Piedras, Puerto Rico. Vélez Rodriguez, L. L. 1995f. Land use and land cover 1989 Guénica Commonwealth Forest. USDA Forest Service International Institute of Tropical Forestry, Rio Piedras, Puerto Rico. Viana, V. M., and A. A. J. Tabanez. 1996. Biology and conservation of forest fragments in the Brazilian Atlantic moist forest. Pages 151-167 in J. Schelhas and R. Greenberg, editors. Forest patches on tropical landscapes. University of Chicago Press, Chicago. Viana, V. M., A. A. J. Tabanez, and J. L. F. Batista. 1997. Dynamics and restoration of forest fragments in the Brazilian Atlantic moist forest. Pages 351-365 in W. F. Laurence and R. O. Bierregaard, editors. Tropical forest remnants: ecology, management and conservation of fragmented communities. University of Chicago Press, Chicago. 286 Wadsworth, F. H. 1950. Notes on the climax forests of Puerto Rico and their destruction and conservation prior to 1900. Caribbean Forester 11:38-47. Wadsworth, F. H. 1990. Plantaciones forestales en el Bosque Estatal de Guanica. Acta Cientifica 4:61-68. Walter, H., and H. Lieth. 1967. Klimadiagramm-Weltatlas. Gustav Fischer, Jena. Watson, D. M. 2001. Mistletoe - a keystone resource in forests and woodlands worldwide. Annual Review of Ecology and Systematics 32:219-249. Webb, N. R. 1989. Studies on the invertebrate fauna of fragmented heathland in Dorset, UK, and the implications for conservation. Biological Conservation 47: 1 53-165. Whitmore, T. C., and J. A. Sayer. 1992. Deforestation and species extinction in tropical moist forests. Pages 1-14 in T. C. Whitmore and J. A. Sayer, editors. Tropical deforestation and species extinction. Chapman and Hall, London. Whittaker, R. H. 1965. Dominance and diversity in land plant communities. Science 147:250-260. Whittaker, R. H. 1975. Communities and ecosystems. MacMillan, New York. Whittaker, R. J ., K. J. Willis, and R. Field. 2001. Scale and species richness: towards a general, hierarchical theory of species diversity. Journal of Biogeography 28:453- 470. Wickham, J. D., K. B. Jones, K. H. Riitters, T. G. Wade, and R. V. O'Neill. 1999. Transitions in forest fragmentation: implications for restoration opportunities at regional scales. Landscape Ecology 14:137-145. Wickham, J. D., R. V. O'Neill, and K. B. Jones. 2000. Forest fragmentation as an economic indicator. Landscape Ecology 15: 171-179. Wilcox, B. A., and D. D. Murphy. 1985. Conservation strategy: effects of fragmentation on extinction. American Naturalist 125:879-887. Williams, M. R. 1996. Species-area curves: the need to include zeroes. Global Ecology and Biogeography Letters 5:91-93. Williams-Linera, G. 1990. Vegetation structure and environmental conditions of forest edges in Panama. Journal of Ecology 78:356-373. Willson, M. F., and F. H. J. Crome. 1989. Patterns of seed rain at the edge of a tropical rain forest. Journal of Tropical Ecology 5:301-308. 287 Wilson, E. O. 1985. The biological diversity crisis. BioScience 35:700-706. Wilson, E. O. 2002. The future of life. Knopf, New York. With, K. A., and T. O. Crist. 1995. Critical thresholds in species' responses to landscape structure. Ecology 76:2446-2459. With, K. A., R. H. Gardner, and M. G. Turner. 1997. Landscape connectivity and population distributions in heterogeneous environments. Oikos 78: 151-169. Zapata, T. R., and M. T. K. Arroyo. 1978. Plant reproductive ecology of a secondary deciduous tropical forest in Venezuela. Biotropica 10:221-230. Zweifler, M., M. A. Gold, and R. N. Thomas. 1994. Land use evolution in hill regions of the Dominican Republic. Professional Geographer 46:39-54. 288 llllllllllllllllllllllll