' 4‘14“? 3',“
' .‘r t- ‘

~13": ,

3-51.» :8»)!

 

4 .4-

ohfd

911

  

was:

«it sk'

.
A

_‘%.1j"l

,.
-

T ~.-.-r. L'.)

’ b_.~.u.,A_f‘-
w- .

. r.--‘

...».;'.11 «.g -.—~ w
‘_ v .,.......7. J,

...77...1r

' H‘-
.n... .9 x

n... .-.no <

n.

 

I I
l
f
:1 .11
1 31.
I I.
:3 £14.}
.- u i. "1
.1 i: I
I 1.‘
‘ mm
a 1 ’
‘3‘
1 ,.'-m m.
r ,z‘ .. 41
«! ,
1: ”2‘11. g
, 1
I.
q‘,‘
‘1.

 

 

IHESIS
()1

UIN IVERSITY LIBRARIES

IIIIIIIIIIIIIIIIIIIIIIII II III III I

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

This is to certify that the

dissertation entitled

ROOT RESPONSES TO NUTRIENT HETEROGENEITY:
A COMPARISON OF DOMINANT AND SUBORDINATE
SPECIES FROM OLD FIELDS

presented by

Andrea L. Corbett

has been accepted towards fulfillment
of the requirements for

Ph .D - degree in Botm

“flaw; :r. A;

Major professor

Date _May_lL_19_9_8___

MS U Lt an Affirmative Action/Equal Opportunity Institution 0-12771

 

 

' LIBRARY

. Michigan State
I Unlverslty

 

 

 

PLACE IN RETURN BOX
to remove this checkout from your record.
TO AVOID FINE-3 return on or before date due.

 

DATE DUE DATE DUE DATE DUE

 

 

 

359g 95

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1/” Woes-p.14

ROOT RESPONSES TO NUTRIENT HETEROGENEITY: A COMPARISON OF
DOMINANT AND SUBORDINATE SPECIES FROM OLD FIELDS

By

Andrea L. Corbett

A DISSERTATION

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

DOCTOR OF PHILOSOPHY

Department of Botany & Plant Pathology
and
W. K. Kellogg Biological Station

1998

ABSTRACT

ROOT RESPONSES TO NUTRIENT HETEROGENEITY: A COMPARISON OF
DOMINANT AND SUBORDINATE SPECIES FROM OLD FIELDS

By

Andrea L. Corbett

Many potential mechanisms have been proposed to explain coexistence among plant
species. Spatial heterogeneity in soil resources at scales smaller than individual plants may
promote coexistence by forcing a trade-off between the efficient locating of nutrient patches
(scale foraging) and the efficient exploiting of nutrient patches (precision foraging). Evidence
from Campbell et al. (1991) suggests that dominant species use a scale foraging strategy
whereas subordinate species use a precision foraging strategy.

To test whether dominant and subordinate species differed in foraging strategy under
heterogeneous soil resource conditions, I conducted a series of greenhouse experiments in
which I varied patch size and intensity, using two dominant (Bromus inermis and Solidago
canadensis) and three subordinate (Achillea millefolium, Rumex acetosella, and Silene
latlfolia) herbaceous perennial species commonly found in local old fields. One of the
dominant species (Bromus) consistently had high tissue nitrogen content, but had lower
nitrogen contents as patch size and nutrient intensity decreased. This performance pattern was
consistent with that expected for a scale forager. The other dominant species (Solidago) and
the three subordinate species had lower nitrogen contents overall, but more consistent nitrogen
contents across patch sizes and nutrient intensities. This performance pattern was consistent

with that expected for a precision forager. There appeared to be a trade-off between the

ability to acquire a large amount of nitrogen and the ability to maintain constant acquisition
levels in the plant as nutrient patch sizes changed.

I also predicted that precision and scale-foraging species would differ in the
mechanisms they used to respond to nutrient heterogeneity. Although I found differences
among species with respect to which response mechanisms they used, the differences were not
related to precision vs. scale foraging strategies. Two of the four precision-foraging species,
Achillea and Solidago, increased their root branching density in nutrient patches. All species
were able to selectively allocate root biomass to nutrient enriched pot quadrants. None of
these species changed root system topology in response to nutrient patches. Rama and
Silene, two of the precision-foraging species, adjusted nitrogen uptake so that root nitrogen
concentrations were constant across patch sizes. Bromus (a scale forager), Achillea and
Solidago (precision foragers) adjusted nitrogen uptake so that root concentrations in enriched
pot quadrants were higher than in background quadrants. These five species did not differ in
the patch size or nutrient intensity at which they used a response mechanism.

To fully understand the role that nutrient heterogeneity may play in determining plant
community structure, we must also know the size, intensity and frequency of small scale
patches in the environment. Spatial sampling and geostatistical analysis techniques are useful
tools for quantifying spatial variation of nutrients in the field, but it is unclear what spatial
sampling pattern will be most sensitive to multiple scales of heterogeneity. Using computer
simulated data and semivariance analysis, I determined that a stratified-nested grid sampling
regime would be more sensitive than a random or stratified grid sampling regime when

determining the magnitude and scale of environmental heterogeneity across several sites.

ACKNOWLEDGEMENTS

I would like to thank Kay Gross, my advisor, for the years of mentoring and
partnership. I would like to thank my committee, Frank Ewers, Peter Murphy and Phil
Robertson, for their support and efforts, and for encouraging me to look beyond just my
dissertation project. Thanks to the lab group for many discussions: Laura Broughton, Bryan
Foster, Wendy Goodfriend, Lisa Huberty, Kevin Kosola, Melissa McCormick and Heather
Reynolds. Special thanks to those who gave feedback on the many drafts of the various
chapters: Paul Cavers, Colleen Doucet, Wendy Goodfriend, Lisa Huberty, Rob Jackson,
Kevin Kosola, Melissa McCormick, Jessica Rettig, Heather Reynolds and Jay Veer Hoef.
My undying gratitude to all those who helped with data collection, tubing and storage bin
washing (1. Call, R. Gabrielse, M. Kellogg, M. Kingslight, J. Pittnaro, M. Russell, M.
Simon, R. Smith, S. & T. Zimmerman), especially Mary Kellogg for weighing what seemed
like an infinite number of root samples!

Financial support was provided for this research by the Department of Botany and
Plant Pathology, Kellogg Biological Station, the KBS LTER project, the KBS Graduate
Research Training program, MSU’s Ecology and Evolutionary Biology program, and a
Sigma Xi Grant-in-aid of Research.

I owe a debt of gratitude to the whole KBS community for creating an environment
that nurtures intellectual development and encourages a lot of interaction. Special thanks to
Alan Tessier for advice on the statistical analyses in Chapters 2 and 3, and Dawn

Jenkins-Klus for assistance with the analysis in Chapter 4. Thank you to Kristen McMillen

iv

for drawing the species sketches in Figure 1.1. Thank you to Tom Getty, Alice Gillespie,
Jessica Rettig, Liz Smiley and all the KBS quilters for providing opportunities to vent some
steam. Thank you to Lisa, Heather and Wendy for the hand-holding and reassurance.

To my parents, who nodded sympathetically when I swore after my seventh grade
science project that I would never work with plants again and who contained their laughter,
when I announced 10 years later that I was going to become a plant ecologist! To the rest of
my family, who all tried to understand what I was doing, regardless of how much biology
they knew. To all my friends near and far who provided caring support by phone, in person
and in prayer.

Finally, a million thank-yous to my husband, Dave, without whose love, patience,
humour, faith, computer programming tutorials, photography and pump repairing skills I

would have never made it this far.

TABLE OF CONTENTS

LIST OF TABLES ............................................ viii
LIST OF FIGURES ............................................ ix
CHAPTER 1
INTRODUCTION .............................................. 1
Competition and coexistence ................................... 1
The role of environmental heterogeneity ............................ 1
Foraging strategy .......................................... 2
Species descriptions ......................................... 3
Thesis overview ........................................... 8
CHAPTER 2
WHOLE PLANT RESPONSES TO NUTRIENT HETEROGENEITY: A COMPARISON
OF DOMINANT AND SUBORDINATE SPECIES ........................ 11
Introduction ............................................. 11
Methods ............................................... 16
Study species ...................................... 16
Pre-experiment conditions ............................... 16
The experimental apparatus .............................. 18
Experimental procedure ................................ 18
Statistical analyses ................................... 24
Results ................................................ 25
Total biomass ...................................... 25
Shoot nitrogen concentration ............................. 29
Plant nitrogen content ................................. 29
Discussion .............................................. 32
Conclusions ....................................... 37
CHAPTER 3
MECHANISMS OF RESPONSE TO NUTRIENT HETEROGENEITY ............ 39
Introduction ............................................. 39
Methods ............................................... 42
Root architecture .................................... 42
Biomass allocation ................................... 46
Nutrient uptake rates .................................. 46
Statistical analyses ................................... 48
Results ................................................ 48
Whole root system architectural indices ...................... 48

vi

Resource heterogeneity effects on root architecture ............... 55

Root Biomass ...................................... 60
Biomass allocation within the root system ..................... 60
Nitrogen uptake assessments ............................. 65
Patch vs. background differences in nitrogen uptake .............. 72
Discussion .............................................. 77
Architecture indices .................................. 77
Biomass responses ................................... 79
Uptake rate responses ................................. 79
Summary of species responses ............................ 80
CHAPTER 4
USING SEMIVARIOGRAMS FOR COMPARISON OF SPATIAL PATTERNS IN
MULTIPLE ENVIRONMENTS: CHOOSING A SAMPLING SCHEME .......... 83
Introduction ............................................. 83
Methods ............................................... 85
Results ................................................ 93
Range estimates ..................................... 93
Nugget estimates .................................... 93
Semivariance estimates ................................ 95
Discussion .............................................. 95
Range estimates ..................................... 95
Nugget estimates .................................... 99
Semivariance estimates ............................... 100
Conclusions and recommendations ........................ 100
CHAPTER 5
DISCUSSION ............................................... 102
APPENDIX ................................................ 104
LIST OF REFERENCES ........................................ 108

vii

LIST OF TABLES

Table 1.1 - Biomass abundance of study species in several old fields at the W.K.

Kellogg Biological Station ............................... 9
Table 2.1 - F-values from three-way analysis of variance on total biomass, shoot

percent nitrogen concentration and total plant N content .............. 26
Table 3.1 - F-values from repeated measures analyses of variance ............... 50
Table 3.2 - Summary of species responses by different mechanisms across patch

sizes and concentrations .................................. 78
Table 4.1 - Average estimates of semivariogram parameters (n=15) for the three

sampling regimes ...................................... 94
Table 4.2 - Average number of pairs used to estimate the semivariance at the first

eight lag classes, using a step size of two ....................... 96
Table A.l - Preliminary data for Daucus carota and Poa compressa ............. 104

viii

LIST OF FIGURES

Figure 1.1 - The five species used in this study ........................... 5

Figure 2.1 - Possible patterns of response to heterogeneity by plant of different
species ............................................ 14

Figure 2.2 - a) Side view of a plant in pot, showing how experimental apparatus is
positioned and b) Top view of pot showing the four quadrants that the
apparatus creates in the pot ............................... 20
Figure 2.3 - The four patch-size treatments applied during the experimental period . . . . 22

Figure 2.4 - Log10 total biomass for two dominant and three subordinate species at
a) low and b) high nutrient intensity levels ..................... 28

Figure 2.5 - Nitrogen concentration in shoot tissue for two dominant and three
subordinate species at a) low and b) high nutrient intensity levels ....... 31

Figure 2.6 - Total plant nitrogen content for two dominant and three subordinate
species at a) low and b) high nutrient intensity levels ............... 34

Figure 3.1 - Types of branching possible in root systems .................... 45

Figure 3.2 - Root architecture index a/E(a) in relation to patch size for high and low
fertilizer levels for each species ............................ 52

Figure 3.3 - Mean root segment length in root samples across patch size treatments
at high and low patch intensities for each species ................. 54

Figure 3.4 - Topological index values for patch and background quadrants within
root systems at high and low nutrient intensity levels for each species . . . . 57

Figure 3.5 - Patch and background quadrant values of mean root segment length
within root systems at high and low nutrient intensity levels for each

species ............................................ 59
Figure 3.6 - Average root biomass per plant at high and low patch intensities for

each species ......................................... 62
Figure 3.7 - Rootzshoot ratios at high and low fertilizer levels for each species ....... 64

Figure 3.8 - Root biomass values for patch and background quadrants within root
systems at high and low nutrient intensity levels for each species ....... 67

Figure 3.9 - Uptake rate by excised roots of the five species as determined by
15N assay with excised roots ............................... 69

Figure 3.10 - Average %nitrogen concentration of whole roots systems at high and
low patch intensities levels for each species .................... 71

Figure 3.11 - "N uptake rate in patch and background quadrants within root systems
at high and low nutrient intensity levels for each species ............ 74

Figure 3.12 - Root nitrogen concentration in patch and background quadrants within
root systems at high and low nutrient intensity levels for each species . . . . 76

Figure 4.1 - Examples of the three different environments with different patch sizes
and number of patches .................................. 87

Figure 4.2 - Examples of the three different sampling regimes, including the grid
divisions that were used (if any) ............................ 90

Figure 4.3 - Idealized semivariograms showing the parameters of nugget (Co). range
(A) and sill (C) ...................................... 92

Figure 4.4 - Average semivariograms estimated for the three environmental spatial
structures using the different sampling regimes ................... 98

Chapter 1

INTRODUCTION

COMPETITION AND COEXISTENCE

Understanding the mechanisms that allow for species coexistence and determine
biodiversity is one of the key questions in ecology (Tilman 1988, Schluter and Ricklefs 1993,
Reynolds er al. 1997). Plants provide a unique challenge for understanding mechanisms of
coexistence, because they are sessile and require the same basic resources. Many potential
mechanisms have been proposed to explain coexistence in plant species (e. g. Zobel 1992,
Bengtsson et al. 1994, Reynolds et al. 1997). Reynolds et al. (1997) categorize these
mechanisms into three types: those that emphasize spatial and temporal resource partitioning;
those that emphasize competitive equivalence and thus prolonged times to competitive
exclusion; and those that emphasize factors that interrupt or reverse competitive exclusion,

like herbivory or disturbance.

THE ROLE OF ENVIRONMENTAL HETEROGENEITY

Considerations of how environmental heterogeneity in plant resources will influence
competitive interactions fall into the category of resource partitioning. At scales of resource
heterogeneity larger than individual plants, species may have trade-offs in competitive abilities
at different resource levels, and thus different species will be favored in different areas of a
spatially heterogeneous environment (Tilman 1988, 1994, Tilman and Pacala 1993).

At very small scales of resource heterogeneity (i.e. less than the size of an individual

2
plant), plants will encounter multiple patches, rather than being entirely within one patch. At
this scale, competition between plants will shift from focussing on the exploitation of a single
patch to having roots that both find and exploit multiple patches (Hutchings and de Kroon
1994, Casper and Jackson 1997). Species differences in ability to obtain nutrients from a
patchy soil environment will translate into differences in growth, performance and competitive
ability that will change as the spatial distribution of nutrients changes (Casper and Cahill

1996).

FORAGING STRATEGY

Grime and colleagues (Campbell et al. 1991, Grime er al. 1991, Grime 1994)
hypothesized that there would be a fundamental trade-off between a species’ ability to explore
a large area (or ’scale foraging’) and the ability to efficiently exploit resources within that
area (or ’precision foraging’). This trade-off may lead to different competitive abilities for
species in heterogeneous vs. homogeneous environments, or may promote coexistence of
species with different foraging strategies in heterogeneous environments.

Within the scale of a plant’s root system, both patch size and nutrient intensity (the
magnitude of difference between nutrient concentrations in a patch and the background soil)
could vary widely. Some species may be more responsive to patches across a range of sizes
or intensities, while other species only respond to the largest and most intense patches. In
addition, plants may utilize different mechanisms to respond to nutrient patches such as
changing nutrient uptake rates (Caldwell 1994), changing root branching pattern (Fitter 1994),
or proliferating roots in patches (Robinson et al. 1994). If the trade-off between scale and
precision of foraging is important, then there could be differences among species in how well
they perform in environments that differ in patch size and nutrient intensity (relative to a

"homogeneous " environment). Species with a scale foraging strategy would be expected to be

3
less sensitive to patches, while precision foraging species would be expected to have the
capability of exploiting patches of many sizes and intensities.

Campbell et a1. (1991) observe that dominant species typically used a scale foraging
strategy, while subordinate species adopted a precision foraging strategy. If this is a general
pattern, then the trade-off between scale and precision in resource foraging could provide a
means by which subordinate species can reduce competitive effects of dominant species in
heterogeneous environments and therefore persist. A subordinate species that can more
effectively exploit small patches can maintain or improve its performance in patchy

environments, relative to the dominant species.

SPECIES DESCRIPTIONS

Mid-successional old fields at the Kellogg Biological Station (KBS) are dominated by
herbaceous perennials (Huberty et al. 1998). In productive sites, the biomass dominants are
Solidago species and the perennial grasses Bromus inermis, Phleum pratense, Andropogon
virginicus and Agropyron repens (Burbank er al. 1992, also Goldberg 1987, Goldberg and
Gross 1988, Foster 1997). Centaurea maculosa, Hieracium spp., and Aster pilosus are often
co-dominants with the above species. Other frequently occurring species are Achillea
millefolium, Medicago sativa, Euphorbia corollata, Daucus carota, and Monarda fistulosa
(Burbank er al. 1992).

From the suite of perennial herbaceous species found in local mid-successional old-
fields, I chose five that represent a range of growth form and root systems: Achillea
millefolium, Bromus inermis, Rumex acetosella, Silene alba and Solidago canadensis
(Figure 1.1). These species also differ in their relative abundance in the field. Solidago and
Bromus are generally biomass dominants in these fields, often making up 40 to 80% of the

community biomass, whereas the other three species are less common and generally comprise

Figure 1.1 - The five species used in this study: a) Bromus inermis, b) Solidago canadensis,
c) Achillea milquolium, d) Rumex acetosella, and e) Silene alba. Images not drawn to scale.
Drawings by K. McMillen.

 

less than 10% of the community biomass.

Bromus inermis Leysser (Family: Poaceae) is a perennial cool season grass (C,) that
was introduced to North America from Europe. It grows 50-100 cm tall and can reproduce
by seeds, tillers and creeping rhizomes (Stubbendieck er al. 1986). Individual leaf blades are
15-40 cm long and 0.4—1.5 cm wide. Inflorescences are 7-20 cm long panicles, with 5-13
flowers per spikelet. Bromus has been observed to form mycorrhizal associations (Harley and
Harley 1987). This species has been cultivated as a hay and pasture grass, but is also found
along roadsides and in waste places (Stubbendieck er al. 1986). In southwest Michigan, it can
become dominant in successional old fields (Burbank er al. 1992, Foster 1997).

Solidago canadensis L. (Family: Asteraceae) is a long—lived perennial forb that occurs
in a broad range of habitats having moderately moist to fairly dry soils and moderate to full
sun. Solidago grows 25-200 cm tall with alternate leaves on the stem and begins to flower in
early August. Flowers are yellow, less than 3 mm in size and form terminal inflorescences.
Solidago has been shown to form mycorrhizal associations and can reproduce vegetatively
from underground rhizomes. In the fall, the flowering stems die and plants overwinter as
dormant rhizomes. Solidago is native to North America, and often becomes the dominant
species in the secondary succession of abandoned fields (8-17+ years). It is also found in
waste areas, tallgrass prairies and infrequently grazed pastures (Werner et al. 1980).

Achillea millefolium L. S.L. (Family: Asteraceae) is a long lived perennial that
overwinters as a rosette of dormant leaves. Flowering stems can be 20-100 cm tall, with
finely divided alternating leaves and several 2-10 cm composite inflorescences at the
top,composed of 2-4 mm flowers. Achillea generally flowers in July and August. Clonal
reproduction by rhizomes is possible, but Achillea generally does not propagate
vegetatively (Warwick and Black 1982). Achillea has been observed to have mycorrhizal

associations (Harley and Harley 1987). It can be found in a wide range of habitats, usually in

7
the open and it is capable of growing under poor soil or drought conditions. The species is
native to Eurasia, but is widely distributed through North America (Warwick and Black
1982).

Ram acetosella L. (Family: Polygonaceae) is a dioecious perennial that grows 10-40
cm tall from a basal rosette of leaves (Gleason and Cronquist 1991). Flowers are very small
in leafless racemes at the top of the stem (Newcomb 1977). Rama can reproduce
vegetatively from root buds that are produced at in high numbers (Houssard and Escarré
1995). Rumex is considered an amycorrhizal species (Harley and Harley 1987) and its fine
roots have abundant root hairs (Corbett, pers. obs.). A native of Eurasia, Rumex is found in
dry fields, pastures, roadsides and gardens (Newcomb 1977) but also tolerates sandy and acid
soils (Gleason and Cronquist 1991). Seedlings will establish in recently disturbed areas, but
vegetative propagation dominates in closed communities (Houssard and Escarré 1991).

Silene latifolia Poiret (Family: Caryophyllaceae) (formerly known as Silene alba
(Miller) E. H. L. Krause, Lychnis alba Miller and Melandrium album (Miller) Gracke) is a
short lived perennial or biennial that requires habitats of well drained soil and high light
levels. Silene is dioecious and produces flower shoots 30-100 cm tall from basal rosettes of
leaves from June through October (McNeill 1977). Flowers are white, 2.5-3.0 cm wide and
sepals are fused into a finely-veined bladder-shaped calyx (Newcomb 1977). Plants have a
prominent tap root and no special structures for vegetative reproduction (McNeill 1977).
Silene is an amycorrhizal species (Harley and Harley 1987) and has long dense root hairs on
the fine roots (Corbett, pers. obs.). Introduced from Eurasia, it is found in waste places,
roadsides and field edges, and usually does not persist in closed communities (McNeill 1977).
These five species are all commonly found in successional old fields in southwest Michigan,
but differ in their abundances (Table 1.1). Bromus and Solidago dominate in terms of

biomass when they occur (> 40% and > 20%, respectively). Achillea and Rumex occur

8
frequently, but generally at lower biomass abundance. Silene occurs at only a few sites, and

when it is found it is always in lower biomass abundance.

THESIS OVERVIEW

I used a series of greenhouse experiments to investigate whether these species differ in
the magnitude and/or mechanism of response to nutrient heterogeneity as patch size and patch
intensity vary. I then assessed if there were similarities in the magnitude and mechanism of
response to nutrient heterogeneity among species based on relative dominance, foraging
strategy, mycorrhizal status or how closely related the species are. Chapter 2 summarizes the
magnitude of response to nutrient heterogeneity. I compared the growth performance of these
species in response to different patch sizes and intensities to test the hypothesis that more
precise foragers will have more consistent performance as patch size and nutrient intensity
change. A companion paper in chapter 3 investigates whether the mechanisms used by plants
to adjust their root systems to patchy nutrient conditions differ across patches of different size
and intensity. The greater emphasis on effective patch exploitation by precision-foraging
species leads to several expectations. Precision-foraging species are expected to: 1) use more
response mechanisms than scale foraging species; 2) respond to a wider range of patch sizes
and nutrient intensities; and 3) employ those mechanisms with finer control than scale-
foraging species, such that roots in enriched nutrient patches will be more different from roots
in unenriched areas.

To understand how species responses to spatial heterogeneity in resources relates to
their competitive ability or the pattern of diversity we see in the field, we need to know not
only the potential of different species to respond to small nutrient patches, but also what the
patterns of nutrient heterogeneity are in the field. Only a few studies have documented small

scale nutrient heterogeneity in the field (Robertson et al. 1988, Jackson and Caldwell 1993,

N
889 seam .49

 

8.9 8.8 8.9 8._ 8.9 Zn 8.8 8 8.8 8.89 8.888 8.. 3.82

28:85 838.892 .0 .8 see .88

8.9 8.8 2.9 8.8 8.9 8.8 8.8 8 8.2 8.89 3: N88 83 8888 833

28885 882852 .e .8 s86 .88

8.9 8.8 8.9 2 8.9 S. 2.8 8 8.8 8.3.8 8.3 E 83 8:88 888:

63:93 .3 8 $80 .5

8.9 2 8.9 8._ 8.9 2 8.89 8.5 8.8 8 8.8 Sm a. 88:88; «PS

385:: .3 a... 8880 .5

8.9 88 2.9 8.8 8.9 8.88. 8.89 I: 8.8 8 8.8 888 8850

63:98 8595.5 .5

8.9 8.8 8.9 8.8 8.9 8.2 8.89 8.89 8.8 8 8.8 $8 .22 x3

28888 £85m .5

8.9 8.8 2.9 8.8 8.9 8.8 2.8 8 M84 8.9 SE 83. 8883 3&3
A9 .88 89 see 89 “an A9 .58 A9 sea .88

£8.23— 8=omo§ 83—8058 $888858 £8.85 888885 A8839

one—mm x883— 8=Eo< owmuzom 388m 38,—. .8388..—

.80E8q8 093858 882:: 80:93:93 85 83 E 838:8 388 =< .8888 53m 8 so 8?. nozm a
88 $5383 29298 .5“. 8m28>8 28 825$ 888883 38 .8 panacea 8 88 888.298 8“ 838% ha 88:3. .688 98:88 Log 8888
8 88888 888885 38. donam game—3m 38:3— .Ms? 2: H8 8.8—om Bo 88268 E 860% 8688 no 858838 888885 - fig 038,—.

10
Gross et al. 1995), but we still know little about the size, intensity and frequency of small
scale patches under most field conditions and how heterogeneity might vary across different
communities. Spatial sampling and analysis using geostatistical techniques is a useful way to
quantify nutrient variation under field conditions. Chapter 4 addresses the question of how to
choose a spatial sampling scheme for geostatistical analysis that is sensitive to several scales

of heterogeneity when sampling several sites to test an ecological hypothesis.

Chapter 2

WHOLE-PLANT RESPONSES TO NUTRIENT I-IETEROGENEITY: A COMPARISON
OF DOMINANT AND SUBORDINATE SPECIES.

INTRODUCTION

Heterogeneity of soil resources in space and time can influence community structure
and the coexistence of plant species (Bengtsson et al. 1994, Tilman 1994). Differences
among species in their ability to respond to nutrient heterogeneity can explain variation in
their relative competitive abilities and community structure (Casper and Cahill 1996,
Campbell and Grime 1989a, Grime 1994).

Soil resource heterogeneity can occur over a range of spatial scales (Robertson and
Gross 1994, Ehrenfeld et al. 1997). At large scales of heterogeneity, soil resource patches
are larger than individual plants, and plant responses to heterogeneity may be expressed in
terms of the relative abundance of species in different types of patches, the ability to colonize
patches, or disturbances (Goldberg and Werner 1983, Tilman 1988, 1990, Casper and Jackson
1997). Field studies have documented that soil nutrients can be heterogeneous at small scales
(i.e. less than 1 m, Schlesinger er al. 1996, Jackson and Caldwell 1993, Gross er al. 1995).
At this scale, nutrient patches are often smaller than the size of individual plant root systems.
The ability of the plant root system to make plastic adjustments to compensate for spatial
heterogeneity of nutrients could be important in determining how much resource the plants
can obtain from the soil (Campbell et al. 1991). The spatial arrangement of the nutrients

could be as vital to a plant’s survival as the total amount of nutrients.

11

12

Grime and colleagues (Campbell et al. 1991, Grime 1994) have proposed that plants
in environments with small-scale nutrient heterogeneity experience a trade-off between the
ability to have roots foraging over a large volume of soil ("scale" foragers) and the ability to
forage intensively in a smaller area ("precision" foragers). Many studies have documented
that plant species can selectively proliferate fine roots or adjust uptake in nutrient-enriched
soil patches, but that not all species exhibit this plasticity (see review by Robinson 1994).

These previous studies have usually investigated plant responses to only one size of
nutrient patch, or compared a single heterogeneous environment to a homogeneous environ-
ment. Yet, both patch size and nutrient intensity (i.e. the magnitude of difference between
nutrient concentrations in a patch and in the background soil) can vary widely, while still
remaining within the scale of a plant’s root system. If there is a trade-off between scale and
precision of foraging, as suggested by Grime and colleagues (Campbell et al. 1991, Grime
1994) then one can expect that there will be differences among species in their performance at
different patch sizes and nutrient intensities, based on which foraging strategy they use.
Precision foragers would be expected to take advantage of patches of smaller sizes or
intensities and thus perform equally well under heterogeneous and homogeneous nutrient
conditions. If the pattern of their performance is measured across a range of patch sizes, they
would exhibit either a "responsive" (the same nutrient acquisition whether the same total
amount of nutrients is patchy or is spread uniformly through the soil) or "extra-responsive"
(obtain more nutrients under patchy conditions than uniform conditions; e. g. Borkert and
Barber 1985, Anghinoni and Barber 1980) pattern (Figure 2.1). Scale foragers, on the other
band, would be less efficient at exploiting nutrients in patches, and their performance would
be expected to be poorer under heterogeneous nutrient conditions than under homogeneous
conditions. The pattern of response across a range of patch sizes would be ”partially

responsive”: they respond to fertilization but obtain fewer nutrients when fertilizer is patchy

13

Figure 2.1 - Possible patterns of response to heterogeneity by plants of different species,
showing hypothetical performances of plants as enriched patch size varies from very small to
as large as the whole root system. The performance of an unfertilized control (no patch) is
included for reference.

Measure of plant performance

 

 

Non-responsive

 

 

 

No

 

patch"">Sma|1

Patch size

Figure 2.1

>Large

15
than when it is uniform (see Figure 2.1). If a plant obtained no nutrients from fertilization, it
would exhibit a "non-responsive" pattern.

Precision foragers will be either "responsive" or "extra-responsive" to a range of
patch sizes because the ability to be plastic at small scales allows the detection and
exploitation of nutrients even when patches are small. Scale foragers will be ”partially
responsive" to changes in patch size, because the trade-off between scale and precision
prohibits efficient exploitation of small patches.

Similarly, as the intensity (or magnitude) of a patch decreases, greater precision will
be needed by a plant to detect an enriched patch, relative to background areas. Thus, scale
foragers are expected to be less able to respond to nutrient patches (vs. uniform enrichment)
as nutrient intensity decreases, while precision foragers remain either "responsive" or "extra-
responsive" to nutrient patches.

Campbell et at. (1991) observed that fast-growing, competitively dominant plants
were able to acquire more nutrients in total and thus grow larger (scale foraging). In
contrast, the smaller slow-growing subordinates were more flexible in allocating root biomass
to nutrient patches (precision foraging). Based on this previous work, I hypothesized that,
across a range of patch sizes and intensities, the dominant species in old fields would be scale
foragers whereas the subordinate species in these communities would be precision foragers.

To examine these predictions, I compared five perennial species that are common in
mid-successional old fields in southwest Michigan, but differ in relative abundance. I
classified these species as either dominant or subordinate based on their proportional biomass
abundance in local old fields (see Chapter 1). In a series of greenhouse experiments I
measured the ability of these species to respond to nutrient patches of three different sizes and
two different intensities. I measured their response in terms of total plant biomass, shoot

nitrogen concentration and plant nitrogen content.

16

METHODS
Study species

The five species I selected for these experiments represent a range of growth forms
and rooting characteristics typical of herbaceous perennials. Bromus and Solidago are
generally dominant species in relative biomass abundance in old fields (Table 1.1 and Burbank
et al. 1992). Achillea, Rum and Silene occur in many mid- to late-successional old fields,
but in much lower abundance, usually less than 10% of total community biomass and so I
classified them as subordinates. Bromus inermis Leysser (Poaceae) is a C3 perennial grass
that reproduces clonally by rhizomes and forms mycorrhizal associations (Gleason and
Cronquist 1991, Harley and Harley 1987). Solidago canadensis L. (Asteraceae) is a long-
lived clonal perennial and is the only one of my species native to North America. It forms
mycorrhizal associations and produces rhizomes (Werner et al. 1980). Achillea millefolium L.
S.L. (Asteraceae) is a rosette-forming short-lived perennial or biennial that can reproduce
vegetatively by rhizomes and forms mycorrhizal associations (Warwick and Black 1982,
Harley and Harley 1987). Ram acetosella L. (Polygonaceae), a dioecious perennial that can
reproduce vegetatively, is common in dry, sandy and acidic soils (Gleason and Cronquist
1991). It has not been observed to form mycorrhizal associations (Harley and Harley 1987).
Silene lanfolia Poiret (Caryophyllaceae - formerly Silene alba (Miller) E. H. L. Krause,
Lychnis alba Miller and Melandrium album (Miller) Gracke) is a short-lived dioecious
perennial that has no vegetative reproductive structures (McNeill 1977). It is also

amycorrhizal (Harley and Harley 1987).

Pre-experiment conditions
I collected seeds of all five study species from field populations at the W.K. Kellogg

Biological Station (KBS) of Michigan State University in southwest Michigan. Seeds were

17
collected during the summer of 1995 at the time of seed maturation for each species and
stored dry. I germinated swds of all species on wet sand in petri dishes placed in growth
chambers under 16 hours of light and at a constant temperature of 25°C. Rumex seeds, which
required pre-treatment to enhance germination (Anderson 1968), were soaked for 10 minutes
in 10 ml 1:1 concentrated sulfuric acid: distilled water and rinsed in distilled water before
being placed in the petri dishes.

Two weeks after germination, I transferred individual seedlings of each species to 5
cm diameter containers filled with silica sand, with a total of 10 seedlings transplanted per
species. I grew the seedlings in the greenhouse at temperatures of 20-23°C, under metal
halide lights set to a 14 hour day (mean light intensity of 388 umol rn'2 sec“, measured under
the lights, mid-morning on a cloudless sunny day). Temperatures were maintained by
greenhouse heaters in winter and evaporative coolers in summer. I watered plants daily and
fertilized them once a week with 10 ml of a 1 g/L Peter’s 20-10—20 Peat-lite Special Fertilizer
solution with micronutrients (7.77% ammonium nitrogen, 12.23% nitrate nitrogen, 10%
available phosphoric acid, 20% soluble potash, 0.15% soluble magnesium, 0.02% boron,
0.01% chelated copper, 0.1% chelated iron, 0.056% chelated manganese, 0.01 % molybdate,
and 0.0162% zinc).

After 2-3 weeks, I re-potted single individuals of each species into 20 cm pots filled
10 cm deep with silica sand (6 pots per species). The plants were grown under the same
light and temperature conditions as above, and I continued to water the plants daily and give
them the weekly fertilizer treatments for an additional 3-4 weeks. This allowed the root
systems to spread throughout the volume of the pot, without becoming pot-bound. I then

selected four plants of each species of similar size for experimentation.

18

The experimental apparatus

I created different nutrient treatments, using a continuous drip irrigation system
modelled after that used by Campbell et al. (Campbell and Grime 1989b, Campbell et al.
1991, Hendry and Grime 1993). Each pot had four equally spaced Tygon tubes held in place
by a PVC frame, dividing the pot into four equal sized quadrants (see Figure 2.2). Two
peristaltic pumps pumped fertilizer solutions through the tubing at a rate of 20 ml per hour.
As long as the solution was flowing through all four tubes in a pot at the same rate, there was
no diffusion between quadrants. Therefore, by varying the nutrient concentrations flowing
through each tube I could create a variety of patchy environments (Campbell and Grime
1989b). To verify that the experimental apparatus was generating distinct patches, I ran two
day and week-long trials with potassium iodide (an analog for N03“) as a marker in the
irrigation system before starting the experiments (see van Ommen er al. 1988). I removed the
soil in 2 cm layers, and traced the location of potassium iodide in pots that had either one or
two enriched quadrants. I observed that patches retained their integrity to the bottom of the

pot (i.e. there was no horizontal spread).

Experimental procedure

The experiment was set up in a nested randomized block design. Each experimental
block consisted of all five species exposed to four different patch size treatments with one
plant of each species per treatment. The number of channels on the pumps limited the
number of plants that could be used in an experimental block, so replication was generated by
repeating blocks through time. There were four patch size treatments applied to the plants
(Figure 2.3): an unfertilized control with background fertilizer solution to all four quadrants,
and small, medium or large patch sizes with one, two or all four quadrants enriched with

fertilizer. In the fertilized treatments, the remaining unenriched quadrants received

19

Figure 2.2 - a) Side view of a plant in pot, showing how experimental apparatus was
positioned and b) Top view of pot showing the four quadrants that the apparatus creates in the
pot (based on Grime 1994). Numbers indicate placement of drip tubes.

I /
92‘

 

21

.82ch 3588.838 85 mats—8 82:58 85:58.: 858-583 .58 2F .. QN enema

22

8308502801

698...

 

QN 9...»:—

88860056:

22882

 

mzoocomeofioI

=mEm

mzoocomoEoI

”EoEcocSco
*0 ._>>o_>.. mac“?—

0 .
83:52
ESE—88:8 Co E

63:00
”mum 22mm

 

 

23
background solution. In the medium-sized patch treatment, the two quadrants that were
enriched were opposite each other (Figure 2.3).

I repeated the experiment six times at two nutrient intensity levels between March and
December, 1996. For the high nutrient intensity treatment, a total of 7.5 mg of Peters 20—10-
20 Peat-lite Special fertilizer was added in solution to each pot per hour. I calculated that this
was equivalent to 20 ppm nitrogen in a pot, an amount chosen to approximate the high end of
available nitrogen in soil patches in local old fields (Gross et al. 1995). The total amount of
fertilizer enrichment per pot was kept constant, delivered as either 7.5 mg per hour to one
quadrant, 3.75 mg per hour to 2 quadrants or 1.875 mg per hour to four quadrants. The
lower nutrient intensity level was half that of the high level (i. e. 3.75 mg of fertilizer added in
solution per pot per hour). Unenriched quadrants in all pots (including all four quadrants in
the control treatment) received a solution of 0.0375 mg fertilizer per pot per hour, or 0.4
ppm nitrogen. Consequently, enriched patches were 50-200 times more concentrated in
fertilizer content than background patches (depending on patch size and intensity levels). The
solutions kept the moisture level of the sand saturated, so no further watering was required.
Bromus and Silene had to be excluded from the last replicate of the experiment because of an
equipment malfunction, so there were only 5 replicates for these two species at the lower
nutrient intensity treatment.

I placed 10 plants around each of the two pumps and then randomly assigned two
treatments to each pump. The plants were exposed to the nutrient treatments for two weeks.
Light and temperature conditions in the greenhouse were maintained at the same levels used
during the pre-experimental growth period.

At the end of two weeks, I harvested plants by first severing the shoots at the surface
of the sand and then used two perpendicular metal sheets as cutting edges to separate the sand

and roots into the four quadrants created by the drip irrigation apparatus. I rinsed the sand

24
from the roots in each quadrant over a 2 mm sieve. The shoots and roots were dried at 45°C
to minimize loss of volatile nitrogen compounds. I weighed the dried biomass of shoots and
roots on a Mettler AE260 Analytical Balance (Mettler-Toledo Instruments, 1991, Hightstown,
NJ) to four decimal places, and then ground the dried plant material for elemental analysis.
Nitrogen content of the shoot and root tissues was determined using an elemental analyzer
(Nitrogen Analyser 1500 Series 2, 1990, Carlo-Erba Instruments, Milan, Italy).

To assess plant performance, I evaluated total plant biomass, shoot nitrogen
concentration and total plant nitrogen content. Total plant N content was estimated by
summing the nitrogen in shoot (%N in shoot / 100 * shoot biomass) and root tissues [(%N of
roots in enriched quadrant(s) I 100 * root biomass in enriched quadrant(s)) + (%N of roots in
unenriched quadrants / 100 * root biomass in unenriched quadrants)]. Plant N content can be
used as an index of total N acquired by the plant during the experiment if we assume that all
plants of a given species in a given block had approximately the same initial N content (i.e.
were the same initial size and N concentration). This assumption is reasonable, because all

plants were grown for the same length of time under standard conditions.

Statistical Analyses

All analyses were performed using Systat 5.0 for Windows (Systat Inc. 1992), using a
4-way mixed model nested ANOVA. Nutrient intensity, patch size and species were all fixed
effects, while block was random. Block effects were nested within nutrient intensity level in
the ANOVA model because blocks were replicated through time and the blocks at high
nutrient intensity were run at different times than the blocks at low nutrient intensity. The
model was left unbalanced where data was missing in one block for Bromus and Silene. Total
biomass measures were log transformed to better fit assumptions of normality and

homoscedasticity in the analyses. The log transformation removes inherent size effects and

25
makes the tests within the model proportionate, because log (a) - log(b) = log (a/b).

In these analyses, significant interaction terms indicate that species differed in their
ability to detect and respond to patches of different sizes and intensities. If there are no
significant interaction terms, then the species all exhibited the same pattern of response, even
if there were absolute differences in performance. A significant patch size x species
interaction indicates differences among species in performance as patch size changes. A
significant nutrient intensity x species interaction indicates differences among species in
performance as nutrient intensity changes. A three-way interaction (nutrient intensity x patch
size x species) indicates species differences in pattern of response across both patch size and
nutrient intensity, where some species responses across patch size are more influenced by

nutrient intensity than other species.

RESULTS
Total Biomass

The ANOVA of log10 total biomass gave significant block, species and patch size
main effects, but no significant interactions (Table 2.1). There was sufficient variation in the
size of plants among the experimental replicates within nutrient intensity levels to produce a
significant block effect. The significant species effect was due to differences in plant size
which were consistent across all treatments. Seedlings of Bromus and Silene were the largest,
Solidago were the smallest, and Achillea and Ram were intermediate in size (Figure 2.4).
Although there were differences in size among species, there were no differences among
species in pattern of biomass response to nutrient heterogeneity. Plants grown in treatments
with fertilized patches had much greater biomass than those in the unfertilized control
treatment, but there were no differences in plant size among the three fertilized treatments (1,

2 and 4 quadrants fertilized; Scheffe post-hoe pairwise comparisons, P (0.05; Figure 2.4).

26

.8588 Z 858 5385:8858 z 88 m 858 888805 88 2 28 .28 8.83 82: .8538 z .88 83.8—88.8 283 88—83 2 85 88 a 3:0 .8
6°.on * "8.ch as." ”:5de is."

 

 

888.8 3. 888.8 888.8 a 888 8388 x 82888 8 88825 85882

1.... 888.8 888._ 888.8 a 888 8388 8 82888

888.8 888.8 888.8 8 838 8388 8 88825 82882

t. 82.8 .. 888.8 88_ .8 8 8.88888 8 88835 .5382

3.... 888.88 1.... 89.28 3.... 88.88 8 on... 8388

3.... 888.88 .2. 888.8 .2... 2.8.88 8 82888

8888 .2... 888.8 s... 88.8 .88.: 38825 883828 885

.2. 888.8 888.8 888.8 _ 88835 888882

8:88:80 2 “SE 883858880 2 825 888885 38.5 smog “:8 858m
8835

.838 :83 8:8 898% 58:85 888328 88383 885888.885 85 8:8 8&8 8888— 83888 58:35 85385 mo 8888.88 usages
23:8 2 8:83 :38 858 828.5888 comer—HE 88.88 88:8 888883 :33 no 8:88:28 .8 888.3898 883.8085 89c 8o=_8>.n_ - EN 838,—.

27

Figure 2-4 - Loglo total biomas
, . S i SE. for two domin ' '
(dashed lme) spemes at a) low and b) high nutrient integgigtgls] me) and three SUbordinate

 

28

A) Low Nutrient Intensity

 

 

 

 

 

 

 

 

 

 

 

 

 

 

3 i
‘6
.5.
m o
m
cu
E
.2
.n
2
8’ +Bromus
.J —O—Solidago
- - o - -Achillea
- - a! - -Rumex
1.5 I I I . "fir-Silene
0 1 2 4
Patch size (No. quadrants enriched)
B) High Nutrient Intensity
3 _
’5
E.
m 2.5
on
«s
E
.2
.n
3 2
g! +Bromus
..l —O—Solidago
- - o - -Achillea
- - a! - -Rumex
15 "fir-Silene

 

 

 

 

 

Patch size (No. quadrants enriched)

Figure 2.4

29
Despite the two-fold difference in fertilizer concentration in the high vs. low nutrient intensity

treatments, there was no significant effect of nutrient intensity on biomass (Table 2.1).

Shoot Nitrogen Concentration

The ANOVA for shoot N concentration revealed a significant nutrient intensity x
species x patch size interaction (Table 2.1), which indicates that the pattern of response to
patch size and nutrient intensity differed among species. Unfertilized control plants (0
quadrants enriched) always had lower N concentrations than plants in the fertilized treatments
(1, 2 and 4 quadrants enriched, Figure 2.5). At the high nutrient intensity level, all species,
except Rumex, had increased shoot N concentrations as patch size increased from small (1
quadrant enriched) to moderate (2 quadrants enriched). In contrast, Rumex had a relatively
constant N concentration across the three fertilized patch size treatments. In the low nutrient
intensity treatments, all five species had similar N concentrations at the larger two patch sizes
(2 and 4 quadrants enriched), but Rum and Bromus tended to have lower N concentrations
at the smallest patch size. There was sufficient variation in shoot N concentration among

replicates to produce a significant block effect in the analysis.

Plant Nitrogen Content

Plant nitrogen content is a function of both nitrogen acquistion and growth rate.
Differences among species in growth rates, or in the rate of conversion of nitrogen to
biomass, can result in biomass or N concentrations alone giving an incomplete assessment of
plant response to nutrient treatments. Thus, I also examined the total N incorporated by each
species during the course of the experiments. I used total plant N content as an estimate of
total N acquired. The ANOVA of plant N content showed significant species x patch size and

species x nutrient intensity interactions, indicating that species differed in their pattern of

30

Figure 2.5 - Nitrogen concentration (%N per g dry weight) in shoot tissue 1 SE. for two
dominant (solid line) and three subordinate (dashed line) species at a) low and b) high nutrient
intensity levels.

31

A) Low Nutrient Intensity

  

 

 

 

 

 

 

 

 

 

 

 

 

3 1
7 _
8 6 ~
3
a: 5 ~
‘6
.2 4 -
m
.5 3 -
z +Bromus
o\° 2 ‘ —O—Solidago
- - <> - -Achillea
1 ., nab-Rumex
0 T l l l 1--*..Silene
0 1 2 4
Patch size (No. quadrants enriched)
B) High Nutrient Intensity
8 l
7 4
3 6 -
3’,
:3: 5 -
*6
2 4 -
m
.5 3
2 +Bromus
.\° 2 —O—Solidago
1 - - o - -Achillea
- - at - - Rumex
O - - a - -Silene

 

 

Patch size (No. quadrants enriched)

Figure 2.5

 

32
response to patch size and to nutrient intensity. Bromus acquired the most nitrogen of any
species (Figure 2.6), and it obtained more N at the high nutrient intensity level than the low
nutrient intensity level (2.29 vs 1.16 mg, 2.62 vs 2.31 mg and 3.62 vs. 2.50 mg at small,
medium and large patch sizes respectively). The other four species acquired about the same
amount of total nitrogen at both nutrient intensities (ranging from 0.4-1.8 mg).

The amount of N acquired by Bromus also varied with patch size. Bromus acquired
less N at the small patch size (1 quadrant enriched) in the low nutrient intensity level, and in
the small and medium patch sizes (1 and 2 quadrants enriched) at the high nutrient intensity
level, than at the largest patch size (4 quadrants enriched; Figure 2.6). Solidago acquired the
least nitrogen of any species, but the amount of N it obtained was constant regardless of patch
size (1, 2 or 4 quadrants enriched) or nutrient intensity (Figure 2.6). Rumex, Achillea, and
Silene were intermediate between Solidago and Bromus in their N acquisition, and in general

tended to show a gradual increase in N content as patch size increased.

DISCUSSION

There were differences among these five old field species in the pattern of response to
patch size and intensity. However, the species differences were not consistent with the
hypothesis that dominant species will exhibit a scale-foraging strategy and subordinate species
exhibit a precision-foraging strategy. The magnitude and pattern of response to patch size
depended on which parameter was used to assess the response. There were no differences
among species in the am of biomass response to patch size or nutrient intensity (i.e. no
significant species x patch size or species x nutrient intensity interactions), but there were
differences among species when shoot nitrogen concentration or plant nitrogen content were
used to assess the pattern of response to nutrient heterogeneity. Of the measured parameters,

nitrogen content was the most informative measure of plant performance, as it combined

33

Figure 2.6 - Total plant nitrogen content i SE. for two dominant (solid line) and three
subordinate (dashed line) species at a) low and b) high nutrient intensity levels.

34

A) Low Nutrient Intensity

 

 
  
 
 
   

 

 

 

 

 

 

 

 

  
 
 
 
 
  
  

 

 

4.5 i
4.0 -*
3.5 -
’5 _
£30
E 2.5 -
E 2.0 4
8
z 1'5 I +Bromus
1,0 — ’ —O—Solidago
--0--Achillea
0.5 ~ nab-Rumex
0.0 ' --¢--Silene
O 2 4
Patch size (No. quadrants enriched)
B) High Nutrient Intensity
4.5 7
4.0 4
3.5 —
m
g 3.0 r
E 2.5 -
a:
E 2.0 -
8 ------------ ::-'- """
21’5 H .'._.—"'. ::::::: ............... i +Bromus
1.0 _ —O—Solidago
- - <> - - Achillea
0.5 - ';:,:" ‘i nab-Rumex
" "fir-Silene

 

 

 

2 4
Patch size (No. quadrants enriched)

Figure 2.6

 

 

35
biomass and nitrogen concentration to give an index of the total nitrogen acquired by the
plant.

While I found significant differences among species in biomass and nitrogen acquired,
they were not consistently related to the species’ biomass abundance in old fields at KBS. My
initial expectation was that dominant species would be scale foragers, but only one of the
dominant species (Bromus) exhibited the pattern of performance across changing patch sizes
that was expected of a scale forager. The other dominant species (Solidago) and the three
subordinate species (Achillea, Rumex, and Silene) had patterns of response to patch sizes that
were consistent with precision foraging. The scale forager (Bromus) had a ”partially
responsive” pattern of performance (see Figure 2.1). Bromus acquired the most N of all the
species, but the amount of N acquired declined as patch size decreased. In contrast, the four
species that were precision foragers had a "responsive" pattern of performance (see Figure
2.1), acquiring the same amount of N across all patch sizes, but not acquiring as much N in
total. There were indications that the three subordinate species might not have as responsive
a performance pattern as originally expected. Achillea had slightly lower performance in the
small patch size at both nutrient intensities and Rumex had slightly lower performance when
patch size and nutrient intensity were both lower. Silene’s performance decreased only when
nutrient concentrations in soil were very high (as in the 1 quadrant patch of the high intensity
level). Improved statistical power and/or an expanded range of patch sizes and intensities
would bring these patterns into sharper focus. The trade-off between scale and precision in
foraging appears to be a trade-off between ability to acquire a large amount of nutrients vs.
ability to maintain constant acquisition levels across a range of patch sizes.

I also predicted that precision foragers would be responsive at both high and low
nutrient intensity, while the performance of scale foragers in patches would drop relative to

the performance in the equivalent homogeneous environment at lower nutrient intensities. I

36
did not observe this pattern. Although the pattern of N acquisition across patch sizes did not
vary between nutrient intensity levels (no significant intensity x species x patch size
interaction), Bromus, the scale forager, was the only species to acquire less N in total at the
lower nutrient intensity than the higher nutrient intensity.

Species with similar growth forms (e. g. functional groups), particularly those that are
closely related species, could be expected to be more similar in the traits they possess, and
thus have more similarity in ability to respond to heterogeneous nutrient supply. However, in
this study the two Asteraceae species (Achillea and Solidago) were not more similar in their
pattern of performance than any of the other species. Bromus, which performed very
differently from the other species, is a grass, and grasses can have different patterns of
response than dicotyledons (Grime 1994, Taub and Goldberg 1996). However, preliminary
data from Poa compressa, a grass species found as a subordinate in the same communities,
suggests that its pattern of response to patch size and nutrient intensity is more similar to that
for the other subordinate species in this study than it is to Bromus (see Appendix).

Another factor that may influence a species’ ability to respond to nutrient
heterogeneity is whether or not a plant forms mycorrhizal associations. Species that form
mycorrhizal associations are often less able to adjust their root systems to patches in
heterogeneous environments (Hetrick er al. 1991, Hetrick 1991). My results do not support
the expectation of amyccorhizal species being more precise in their foraging. Both of the
amycorrhizal species I used (Rumex and Silene) were not more responsive than the other
species. Furthermore, Rumex and Silene differed from one another in their responsiveness to

nutrient intensity.

37
Conclusions

This study provides some support for the hypothesis of Campbell et al. (1991) that
there is a trade—off between scale and precision in resource foraging by plants in
heterogeneous environments. In these experiments, I found that species patterns of nitrogen
acquistion fell into only two categories: those species that acquired more nitrogen overall, but
whose nitrogen acquisition dropped as patchiness increased (scale foragers) and those species
that acquired less total nitrogen, but whose nitrogen acquisition stayed constant as patchiness
increased (precision foragers). There were no species that were able to both acquire high
amounts of nitrogen and keep their acquisition constant across the range of patch sizes. This
basic trade-off could lead to differences in competitive ability or performance in the field. If
the precision forager is maintaining a constant or improved performance under heterogeneous
conditions while a scale forager’s performance declines, then the precision forager may
experience less interspecific competitive effects from the scale forager under heterogeneous
conditions. So a trade-off in foraging strategy may provide a means by which less
competitive subordinate species can persist: increased plasticity and precision versus the faster
growing but less precise dominant species.

However, for the species I studied, the differences in response to nutrient
heterogeneity are not related to biomass abundance in the field. I classified species as
dominant or subordinate by using field biomass abundance, whereas Campbell et al. (1991)
used a lab competition experiment to determine a competitive hierarchy. It could be that
some of the species I used have higher or lower abundance in the field for reasons other than
their ability to compete for nutrients, or that some other factor besides abundance influences
whether or not a given species exhibits a scale or precision foraging strategy. The trade-off
may not be exhibited by differences between dominant and subordinate species, per se, but in

large versus small species. In plant communities, dominant species are also usually the larger

38
species, so a strategy that is correlated with large plant size will also appear to be correlated
with dominance. In this experiment, the seedlings of one of the dominant species (Solidago)
were also the smallest, and it was the species with the most precise foraging. It may be that
the precision with which a species can forage will decrease as the species grows larger and
forages over a wider scale. In field conditions, plants of a given species may be of different
sizes (and ages), so the ability to forage with precision while small will help the plant survive
until it grows large enough that precision no longer is necessary. A patch of a small absolute
size will be relatively larger to a small plant than to a large one, so the relative benefit of
exploiting the patch will be larger for a small plant than a large one.

This study also demonstrated the importance of the scale and magnitude of
heterogeneity in determining the response of a species. When patches are larger in size and
higher in concentration, these five species will perform as well as if the nutrients were
homogeneously distributed around the roots. In environments with a coarse scale of
heterogeneity plants could compete with each other to exploit patches in a similar fashion as
they would compete in a homogeneous environment. However, as the size and intensity of
patches decrease, differences among species in ability to detect and respond will become more
important. The trade-off between scale and precision in resource foraging could be key in
reducing competition between species. Large dominant scale foragers would have relatively
poorer performance under highly heterogeneous conditions, while the precision foragers
would be able to maintain their performance and thus be relatively more competitive in

heterogeneous environments.

Chapter 3

MECHANISMS OF RESPONSE TO NUTRIENT HETEROGENEITY

INTRODUCTION

The ability of plants to obtain nutrients in patchy environments is well documented
(Robinson 1994 and references therein). However, species differ in their ability to respond to
nutrient heterogeneity at scales smaller than individual plant root systems (Campbell et al.
1991, also see Chapter 2). Differences in the ability of species to adjust to heterogeneously
distributed resources may influence interspecific competitive interactions and community
structure in spatially heterogeneous environments. A vital component of linking the responses
of individual plants to resource heterogeneity to community structure is understanding the
mechanisms of plasticity used by plants and relating those mechanisms to plant performance
(Casper and Jackson 1997).

Plants can use a variety of mechanisms to increase nutrient acquisition in response to
nutrient heterogeneity. It is well documented that plant species can proliferate roots in
nutrient patches (see review by Robinson 1994 and references therein) and some species adjust
the nutrient uptake rate of roots in enriched patches (Drew and Saker 1975, Robinson and
Rorison 1983, Jackson et al. 1990, Caldwell et al. 1992, Caldwell 1994, Robinson 1994, van
Vuuren er al. 1996). Plants also may change the rate of root turnover (Gross er al. 1993) or
the branching architecture of the root system (Fitter 1994) in response to nutrient
heterogeneity.

It will not be advantageous for a plant to expend energy to exploit a nutrient patch

39

40

unless there is sufficient benefit (on average) in the amount of nutrients acquired to
compensate for the cost of obtaining those nutrients (Robinson 1996, Gleeson and Fry 1997 ).
The construction of new roots, or new cells within a root, likely involves a greater
expenditure of carbon and energy than to adjust the function of cells or roots that already
exist. Physiological responses would be expected to occur before responses involving new
root growth (Robinson 1996). Thus, the type of response mechanism a plant utilizes will
influence the cost of acquiring nutrients. The concentration, size and duration of an enriched
nutrient patch will influence the potential nutrients available to the plant and thus the benefit
to the plant. The net benefit will be determined by how well a plant can maximize nutrient
uptake while minimizing costs (Eissenstat 1992, Robinson 1996). For example, if a plant
allocates energy to increase root biomass, it faces a trade-off between producing short, highly
branched roots that cover a smaller soil volume, and thoroughly spread throughout that
volume, versusproducing longer, less branched roots that exploit a larger soil volume, but
with less intense packing of roots within that volume. Campbell et al. (1991) suggested that
this is a trade-off that characterizes whether a plant will be a precision or scale forager. An
increased scale of foraging allows a plant to encounter more patches in a heterogeneous
environment, while increased precision of foraging is a means of more thoroughly exploiting
those patches that are encountered.

If the trade-off between precision and scale of foraging is important for plants, then
one could expect there to be differences between precision- and scale-foraging species in the
mechanisms they use to adjust to small scale nutrient heterogeneity (Fitter 1994, Grime 1994).
A plant using a highly precise foraging strategy would be expected to have greater plasticity
of response. Thus, it should be more plastic in both the mechanisms it can use to respond to
patches and the range of conditions under which it utilizes those mechanisms. Plants using a

scale foraging strategy would be less likely to detect patches they encounter and instead

41
expand the root system into the largest possible area. Scale foragers would be expected to
have less plasticity in foraging, and so would use fewer types of mechanisms to respond to
nutrient patches and would not respond to patches that are small, relative to the size of the
root system, or of lower nutrient intensity, relative to the nutrient levels in the rest of the soil.

In this study, I compared five perennial species (Achillea millefolium, Bromus inermis,
Rumex acetosella, Silene latrfolia and Solidago canadensis) that commonly occur in Southwest
Michigan old-fields in a series of greenhouse experiments to determine their responses to
varying nutrient patch sizes and intensities. Bromus and Solidago are dominant species in
these communities and Achillea, Rumex, and Silene are subordinate species. Based on the
pattern of nitrogen acquisition across a range of patch sizes (see chapter 2), I classified
Bromus as a scale forager and the other four species (Solidago, Achillea, Rumex and Silene)
as precision foragers. Of these four species, Achillea, Rumex and Silene have a tendency to
be less precise than Solidago.

In this chapter, I investigated three possible mechanisms of response to nutrient
heterogeneity: root branching architecture, biomass allocation and nutrient uptake rates.

Based on the hypothesis that there is a fundamental trade-off between the scale and precision
of root foraging that leads to differences in the plasticity of root systems, I expected that:

1) Scale and precision foragers would differ in the mechanisms used to respond to
heterogeneous nutrients. Precision-foraging species were also expected to use more
mechanisms that scale-foraging species.

2) Precision-type species would respond across all patch sizes and intensities, scale-
type species would only respond at larger patch sizes (or homogeneous enrichment).

3) For a mechanism used by both precision- and scale-foraging species, precision
species were expected to have a finer degree of control within the root system. That is, the

degree to which roots in enriched patches differ from roots in unenriched areas would be

42

greater in precision foragers than scale foragers.

METHODS

Each of the five species was exposed to four patch size treatments: control, small,
medium and large (0, 1, 2 and 4 quadrants enriched, respectively; see Chapter 2, Figure 2.3)
at two nutrient intensity levels. The experiment was repeated six times at each nutrient
intensity level. Chapter 2 contains a detailed description of the growth conditions and

experimental apparatus.

Root Architecture

To assess if root architecture patterns change in response to patch size and nutrient
intensity, I sampled the largest unbroken piece of root harvested from two randomly selected
quadrants in the control and large patch treatments (0 and 4 quadrants enriched, respectively)
and from one patch and one background quadrant for small and medium patch treatments (1
and 2 quadrants enriched, respectively). Roots were dyed using Safranin O and then laid out
on clear acetate sheets. I used forceps to tease apart roots so that their complete branching
structure was visible. A second acetate sheet was placed on top of each sample for protection
and the samples were allowed to air dry. I photographed each sample using black and white
print film (Kodak TMAX 400) and developed the images onto CD-ROM. These digitized
images were processed using Adobe Photoshop version 3.0 (Adobe Systems Inc. 1994).
Architectural indices were calculated using the BranChing program (Bernston 1992, Online).

I used both topological (arrangement of links or branches within the root system,
following Fitter 1987) and geometrical (link size or density of branching) indices of root
branching to quantify architecture. Root system topology can vary between two extremes:

perfectly herringbone and completely dichotomous (see Figure 3.1) with a random pattern of

43
branching as an intermediate. Herringbone root systems are predicted to minimize the
overlap of root depletion zones in the soil, while dichotomous systems have more efficient
nutrient transport within the plant (Fitter 1985 , 1987). As nutrient levels increase, depletion
zones are less critical (especially for mobile nutrients) and producing a root system with better
transport efficiency will benefit the plant (Fitter 1987). This would lead to the expectation
that plants using this response mechanism move from more herringbone to more dichotomous
branching. Fitter (1994) found evidence that faster-growing species are more likely to use
changes in root topology as a response to nutrient patches, so I expected that scale-foraging
species would use this mechanism and precision-foraging species would not.

I used a topological index of a/E(a) where a= altitude (# links in longest path from
exterior link to top of root system) and E(a) is the expected altitude of a root system of that
size (total # links) given random growth (Fitter 1994). A herringbone root branching pattern
gives the highest a/E(a) index and a dichotomous branching pattern gives the lowest (Fitter
1987 , Fitter er al. 1991). After measuring the altitude of a sample, the expected altitude was
calculated by the software program BranChing (Bernston 1995, Online) using formulae based
on Werner and Smart (1973). If a plant responds to nutrient patches by changing its
branching pattern to be more dichotomous, a/E(a) will decrease. Although topological indices
are usually calculated for whole root systems, sub-samples can be assessed if they are cut
randomly from the system and one assumes that topology is consistent across the whole
system (Van Pelt and Verwer 1984, Fitter and Stickland 1992).

I calculated mean segment length (MSL) on the same root samples used for the
topological index, by dividing total root length by the number of segments (or links) as
measured by BranChing. MSL is an index of root system geometry that assesses the density

of branching in the roots; as MSL decreases, the root system becomes more densely

.8388? 88 05 mo m9 2: 8 an: :85on :8 Bob 28:88am 595—53 825588 u 38288 .882:
3585 go 838:: :38 N 833982 .638 33280 888 858,—. Set @8983 8888.88 88 5 28:88.8 $83885 go 833. - fin 952m

45

—.n 0.53%

\
/\
\#

///////
\\\\\\\\

 

 

or u onzzzm w u ouszzm
88 n 8888888. 88 u 8.8888886
mchcSm econmcEmI mchcEm Eoucmm

 

m u ouazzm
3 u onBEmmE
mczocem waoanocoE

45
branched. I expected that precision foraging species would be more likely to use this

mechanism of response to nutrient patches.

Biomass Allocation

The root biomass from each quadrant of the pot was kept separate as it was harvested,
and the total root dry weight plus that of any sub-samples used in other analyses were added
to give total root biomass per pot quadrant. I summed the root biomass from each quadrant
to get the total root biomass for each plant. I calculated root:shoot ratio from dried shoot and
root biomass to test whether allocation between roots and shoots changed across treatments.
To assess whether plants were selectively allocating more root biomass to enriched patches, I
compared the biomass of roots from two different quadrants for each plant. I randomly
selected two quadrants from each pot in the control and large patch treatments (0 and 4
quadrants enriched, respectively). I compared the enriched patch to a randomly selected
background quadrant in the small patch treatment (1 quadrant enriched). In the medium patch
treatment (2 quadrants enriched), I compared a randomly selected patch quadrant to a

randomly selected background quadrant.

Nutrient Uptake Rates

To assess if resource heterogeneity affected nutrient uptake rates, I took a sub-sample
of root from two quadrants when the plants were harvested. For the control and large patch
treatments (0 and 4 quadrants enriched, respectively), I sampled a randomly chosen quadrant
and its counter-clockwise neighbor. For the small and medium patch treatments, I sampled
the enriched patch quadrant (or randomly chose one of the two patch quadrants), and its
neighboring background quadrant in a counter-clockwise direction. These excised roots were

then used immediately to conduct an "N enrichment assay (adapted from Kosola and Bloom

47
1994, 1996 and Jackson and Reynolds 1996 ). I placed the excised roots in an aerated
equilibrating solution (2 mM CaSO4) for 10 minutes, then added labelled K'5N03 (0.5 ml of
0.1 M K‘SNO3 in 500 mL of 2 mM CaSO,) and let the roots soak for 30 min. The samples
were aerated and kept in a 25°C water bath to maintain constant temperatures and keep
solutions well mixed. I then rinsed the root samples in water to remove excess solution, and
soaked them for 2 min in an unlabelled nitrogen solution (0.5 mL 0.1M KNO3 in 500 mL of
2mM CaSO,) to rinse labelled N off the exterior surfaces of the roots. I allowed samples to
air dry before placing them in a drying oven for several days at 45°C. 1 ground the samples
in microfuge tubes with two ball bearings inside using a dental amalgamator and then weighed
out sub-samples of 1-5 mg into tin cups for analysis on an ANCA-MS mass spectrometer
(Harris and Paul 1989). I was only able to analyze two of the six replicates for each nutrient
intensity level, due to budget and time constraints.

As an indirect assessment of uptake rate, I measured root N concentrations in four of
the six replicates at the high nutrient intensity level and in five of the six replicates at the
lower nutrient intensity level. I used the CN protocol described in Chapter 2 to measure root
nitrogen concentration (%N per g dry weight). Again I chose samples randomly where
possible, from two quadrants in the control and large patch treatments, or a patch and a
background quadrant in the small and medium patch treatments. As this was the last set of
sub-samples taken from the root systems, in many cases there were only two quadrants that
had sufficient tissue left to provide a sample for CN analysis. These measures of root N
concentrations are the result of both uptake into the roots and transport from the roots, and so
N concentrations will not provide as robust a measure of uptake. Nevertheless, the data will

allow estimation of general differences between species and treatments.

48
Statistical Analyses

Statistical analyses were performed using Systat 5.0 for Windows (Systat Inc. 1992).
I performed a 4—factor repeated measures mixed model ANOVA. Nutrient intensity, patch
size and species were fixed effects, while block was random. Block effects were nested
within nutrient intensity level, because blocks were replicated through time and the blocks at
high nutrient intensity were run at different times than the blocks at low nutrient intensity.
The model was left unbalanced where data was missing. Root biomass and MSL were log
transformed to improve normality and reduce heteroscedasticity.

I used a repeated measures analysis because every experimental unit (= 1 plant) was
measured twice for each root system variable. I sampled either one patch and one
background for small and medium patch treatments (1 and 2 quadrants enriched) or two
quadrants for control and large patch treatments (0 and 4 quadrants enriched). For each
variable I measured, the between-subjects component of the analysis is the average of the two
quadrants and estimates of differences between species, nutrient intensity and patch size
treatments are based on this average. The between-subjects component thus provides
information as to how the whole root system of these species responded to resource
heterogeneity. The within-subjects components of the analysis measures the variation between
the two samples from different quadrants of a pot. This provides information on how

individual root systems varied across species, patch size and nutrient intensity levels.

RESULTS
Whole Root System Architectural Indices

For both of the architectural indices (altitude/expected altitude (a/E(a)) and mean
segment length (MSL)), the values calculated for the two sampled quadrants in a pot were

averaged for the between-subjects component of the ANOVA, providing an estimate of the

49
mean architectural index for the whole root system. This analysis thus provides information
on any differences in the architectural indices among species, patch size or nutrient intensity.
There was a significant species effect on root systems topology (a/E(a); Table 3.1), but the
effect was driven by Bromus, the scale forager, having a higher index (or more herringbone-
like root systems) than the other species (Figure 3.2). There was also a significant patch size
effect on root topology. The a/E(a) index was slightly lower (indicating a more random
branching topology) in the three fertilized treatments (small, medium and large patch size)
than in the unfertilized control treatment (Figure 3.2). The decrease in a/E(a) index in
response to enriched patches appears to be driven by Silene at the low nutrient intensity
(Figure 3.2a) and Bromus at the high nutrient intensity (Figure 3.2b). However, there were
no significant interactions, indicating that species did not differ significantly in the pattern of
a/E(a) values across patch sizes.

MSL was more variable among species and there were significant species x patch size
interactions, indicating that some species responded differently to the patch size treatments
(Table 3.1). Both Achillea and Solidago had lower MSL, and thus increased branching
density in the fertilized treatments (small, medium and large patch size), while the other three
species did not change in MSL relative to the unfertilized control (Figure 3.3). Achillea and
Solidago also had shorter MSL in the control treatment at the lower nutrient intensity, giving
a significant species x nutrient intensity interaction. Bromus, Rumex and Silene all had shorter
MSL in the unfertilized control treatment than Achillea and Solidago. This suggests that
Bromus, Rumex, and Silene have high root densities that did not allow them to adjust their

roots to decrease MSL further in the fertilized treatments.

50

Table 3.1 - F-values from repeated measures analyses of variance presented with degree of
significance (*** P<0.001; ** P<0.01; * P<0.05). Between-subjects values summarize
differences between root systems of individual plants. Within-subjects values estimate
differences between quadrants within plants. 1’ df =8 for Rep (Nutrient intensity) and
Quadrant x Rep (NI) because only 9 of the 12 experimental replications were analyzed for
nitrogen concentrations.

 

 

 

Variable

Between-subjects df Logw Root alE(a) Log10 MSL Root N

Biomass Concentration?
Nutrient intensity 1 0.816 2.404 8.423 ** 2.217
Rep (Nutrient intensity) 11 4.952 *** 0.899 0.642 3.309 **
Species 4 47.914 *** 34.640 *** 152.196 *** 22.853 ***
Patch Size 3 1.130 3.714 * 7.502 *** 175.818 ***
NI x S 4 3.473 ** 1.011 5.182 ** 1.115
NI x PS 3 0.357 2.058 1.083 6.960 ***
S x PS 12 0.934 0.853 2.080 * 6.633 ***
NI x S x PS 12 0.597 0.872 1.108 1.260
Within-subjects
Quadrant 1 9.744 ** 1.853 2.041 15.574 ***
Quadrant x NI 1 3.061 0.307 0.939 0.010
Quadrant x Rep (N1) 11 0.933 0.891 1.111 0.486
Quadrant x S 4 0.227 3.216 * 2.758 * 3.284 *
Quadrant x PS 3 4.560 ** 0.932 2.659 * 16.875 ***
Quadrant x NI x S 4 0.218 0.878 3.752 ** 0.095
Quadrant x NI x PS 3 2.145 1.753 1.351 4.455 **
Quadrant x S x PS 12 1.417 0.916 0.806 1.730
Quadrant x NI x S x PS 12 1.348 0.889 0.724 1.025

51

Figure 3.2 - Root architecture index alE(a) 1; SE. in relation to patch size for high and low
fertilizer levels for each species. The expected altitude is the altitude that would occur if
branching was random at a given magnitude (number of root links). alE(a) =1 means
random branching did occur. alE(a) > 1 indicates more herringbone branching. alE(a) < 1
indicates more dichotomous branching. Filled symbols with solid lines are dominant species,
hollow symbols with dashed lines are subordinate species.

52

A) Low Nutrient Intensity
2

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

+ Bromus
1.8 —o— Solidago
1.6 ~o - - Achillea
-x -- Rumex
1-4 --a--Silene
A 1.2
.53.
m 1
3 0.8 w
0.6 -
0.4 ~
0.2 J
0 I l l l
0 1 2 4
Patch size (No. quadrants enriched)
B) High Nutrient Intensity
2 —-I—— Bromus
1.8 —0— Solidago
1 6 --<> -- Achillea
' ‘ - - as - - Rumex
1.4 n rm-” ”'A"Silene
A‘L2 —::~.::‘_‘_‘_._.. :;:'.'-’-"—"'- ‘ ..... -—:--:'_~_-_“--- --
a --- _ - -
if 1 """ - """" -—
3 0.8 - ’
0.6 —
0.4 ~
0.2 -
0 I l l l
0 1 2 4

Patch size (No. quadrants enriched)

Figure 3.2

53

Figure 3.3 - Mean root segment length 1 SE. in root samples across patch size treatments at
high and low patch intensities for each species. Filled symbols with solid lines are dominant
species, hollow symbols with dashed lines are subordinate species.

54

A) Low Nutrient Intensity

m)
:5
0'1

l

Mean segment length (m

 

 

 

 

 

 

 

0 1 2 4
Patch size (No. quadrants enriched)

B) High Nutrient Intensity

 

 

 

5
E45
E 4 .
g ‘\“‘.
53.5 i
33 3 .
$2.5 “5 ........................
2 .....................
101.5 -_
g 1 ------- I """"""" I ________________________ I
a:
20.54
O I I I 1

Patch size (No. quadrants enriched)

Figure 3.3

 

+ Bromus
+ Solidago
-- o -- Achillea
- - a: - - Rumex

 

--a -- Silene

 

 

 

——I— Bromus
—o— Solidago
- - o - - Achillea
u-x - - Rumex
--a - - Silene

 

 

 

55

Resource heterogeneity effects on root architecture

The within-subjects component of the ANOVA for the topological index is an estimate
of the amount of variation between the two quadrants sampled in each pot and indicates
whether sub-sections of a root system have different topologies. Within subjects, there was a
significant quadrant x species interaction for alE(a) (Table 3.1), indicating differences among
species in variability of alE(a) between quadrants. In addition to having higher alE(a) index
values, Bromus had more variability in alE(a) within its root systems than the other species
(Figure 3.4a, b). There were no significant interactions between species and patch size or
nutrient intensity, indicating there was no significant pattern in the alE(a) index related to
either patch size or nutrient intensity. Although there was not a significant pattern, both
Bromus and Achillea had lower alE(a) values in enriched vs. unenriched sectors at the high
nutrient level. This raises the possibility that they might be able to selectively adjust their
root architecture in only part of the root system as a response to nutrient heterogeneity.

The within-subjects component of the ANOVA for MSL indicated significant
quadrant x species and quadrant x nutrient intensity x species interactions. These interactions
were driven by Achillea and Solidago, both of which had more variation in MSL within their
root systems than Bromus, Rumex, or Silene (Figure 3.5). Achillea had lower MSL values in
the enriched patches than in background quadrants of heterogeneous patch size treatments (1
and 2 quadrants enriched) at both nutrient intensity levels (Figure 3.5e, t). Solidago also
selectively decreased MSL in enriched patches at the higher nutrient intensity (Figure 3.5d),
however, the reverse was found for the 1-quadrant-enriched treatment at low nutrient
intensity, where MSL was greater in enriched patches than in the background quadrants

(Figure 3.5c).

56

Figure 3.4 - Topological index [alE(a)] values i 5.13. for patch and background quadrants
within root systems at high and low nutrient intensity levels for each species. Expected
altitude is that generated by random branching at a given magnitude (# of root links). alE(a)
=1 means random branching occurred (dashed line). alE(a) > 1 means more herringbone
branching. alE(a) < 1 means more dichotomous branching. Open diamonds are the means
of root samples from unenriched quadrants. Solid squares are means of root samples from
enriched quadrants. All quadrants in the patch size 0 treatment were unenriched. All
quadrants in the patch size 4 treatment were enriched.

Low Nutrient intensity

57
High Nutrient Intensity

 

 

 

 

 

 

{ a) Bromus b) Bromus
1 I i
’ i i E " i I i
c) Solidago d) Solidago

-i¥!i mum-mi]; ......... I"--.

—

l l l l L l I l l

 

 

 

 

 

 

 

 

 

 

 

 

 

 

#e) Achillea .f) Achillea

2min-.-"- ---—------! ----- :----:----I-----.- --------- I -----

bg)Ru1mex I I l l -h) Rumex l l l I

ill? ..... 3i! .......... 2 _____

~i) Silene I 1 l 1 pi) Silene J L L 1

It. ......... i--- L--_-;_---i--_-! .......... 1----
Q. ; ; ’ l I. 7 Q. i 3.

Patch size (No. quadrants enriched)

Patch size (No. quadrants enriched)

Figure 3.4

58

Figure 3.5 - Patch and background quadrant values of mean root segment length 1 SE.
within root systems at high and low nutrient intensity levels for each species. Open diamonds
are the means of root samples from unenriched quadrants. Solid squares are means of root
samples from enriched quadrants. All quadrants in the patch size 0 treatment were
unenriched. All quadrants in the patch size 4 treatment were enriched.

MSL (mm)
M (A) vb 0|

”'3'- (mm) MSL (mm) MSL (mm) MSL (mm)
wwewoiwwemOAMwhmo-swwemo-s

o—l

Low Nutrient intensity

59

High Nutrient intensity

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

a) Bromus Lb) Bromus
P i I l h i i i i
. -
c) Solidago d) Solidago
_ i i h i E I i
f 1 i i t ' I
e) Achillea 1 f I f) Achillea
- F !
I
_ i s 1 . . i
- L
g) Rumex h) Rumex
F -
_ _ I
I s i ' i i ' I
I) Silene J) Silene
. " a I I - ' I I I
0 1 2 4 0 1 2 4

Patch size (No. quadrants enriched)

Patch size (No. quadrants enriched)

Figure 3.5

Root Biomass

The between-subjects component of the repeated measures AN OVA on logl0 root
biomass provides information on whether root system biomass differed among species, or
across patch size or nutrient intensity. There were significant differences in root biomass
between species (Table 3.1); Bromus and Silene had larger root biomass than Rumex and
Silene, and Solidago had the smallest root biomass (Figure 3.6). There was a significant
species x nutrient intensity interaction (see Table 3.1), driven by one species (Achillea) having
less root biomass at high nutrient intensity, and one species (Rumex) having more root
biomass at high nutrient intensity. The other three species did not differ in root biomass at
the two resource intensity levels. There were no differences in total root biomass between
patch size treatments for any of the species (no patch size effect and no species x patch size
interaction). Although the total root biomass of these species did not change between
fertilized and unfertilized treatments, all of the species had larger total plant biomass in the
fertilized treatments (see Chapter 2). All species increased shoot and not root production in
response to nutrient enrichment. Consequently, the R:S ratio decreased in all species with

fertilization (Table 3.1, Figure 3.7).

Biomass allocation within the root system

Although the total root biomass of these species did not vary across the patch size
treatments, there were differences among patch sizes in how root biomass was allocated
among quadrants in a pot. The within-subjects component of the root biomass ANOVA
provides an estimate of the variation between the two quadrants sampled in each pot. There
was a significant quadrant x patch size treatment interaction (see Table 3.1, within-subjects
effects), suggesting that the variation in root biomass between quadrants varied across patch

size treatments. All the species had greater root allocation in the enriched patches of the

61

Figure 3.6 - Average root biomass per plant i SE. at high and low patch intensities for each
species. Filled symbols with solid lines are dominant species, hollow symbols with dashed
lines are subordinate species.

62

A) Low Nutrient Intensity

 

 

 

 

 

 

 

 

 

 

 

 

0.08 -
+ Bromus
- + Solidago
0'07 --o - - Achillea
A - --x -- Rumex
30.06 “A - - Silene
3 0.05
g .................
.2 0.04
.n _________________
0.03 ------
a. __________________________________
0 0.02
0.01
O l I I l 1
1 2 4
Patch size (No. quadrants enriched)
B) High Nutrient Intensity
0.08 7
+ Bromus
0.07 e —e—— Solidago
--o - - Achillea
@006“ ---:I£--Rumex
a 0.05 --a-- Silene
g 0.04 ...............
.9
30.03
0.02 ' '
0.01
O | I I I I

 

1 2 4
Patch size (No. quadrants enriched)

Figure 3.6

63

Figure 3.7 - Rootzshoot ratios 1 SE. at high and low fertilizer levels for each species.
Filled symbols with solid lines are dominant species, hollow symbols with dashed lines are
subordinate species.

A) Low Nutrient Intensity

 

 

 

 

 

 

 

 

2'50 -—I—— Bromus
—e— Solidago
200 -- o - - Achillea
.3 —-x - - Rumex
3 ~13 - - Silene
*5 1.50
O
6
31.00
0
O
a:
0.50 -
0.00 I I I I
0 1 2 4

Patch size (# quadrants enriched)

B) High Nutrient intensity

 

 

 

 

 

 

 

 

 

2'50 + Bromus
—e-— Solidago
2.00 --o - - Achillea
:g ---x -- Rumex
S -- A -- Silene
.5 1 .50
O
.c
,9:- 1.00
O
a?
0.50 —
0.00 l I I l
0 1 2 4

Patch size (# quadrants enriched)

Figure 3.7

65
small and medium patch size treatments (1 and 2 quadrants enriched) and so greater within-
subjects variation (Figure 3.8). There was generally less variation between quadrants in the
control and large patch size treatments than in the small and medium patch treatments for all
five species. The lack of significant quadrant x species x patch size effect (Table 3.1),
indicates that the five species responded the same way to resource heterogeneity by shifting
allocation to the enriched quadrants in heterogeneous treatments (ie. root biomass in
background < patch), but having approximately even allocation among quadrants in

homogeneous treatments (Figure 3.8).

Nitrogen uptake assessments

There are indications that uptake rates changed in response to nutrient heterogeneity.
All species had higher and more variable uptake in the small or medium patch treatments in at
least one level of nutrient intensity (Figure 3.9). A lack of replication in the 15N uptake
analyses, however, limited the power to detect differences across patch sizes or among
species.

Root N concentration increased significantly across patch size treatments (Figure 3.10,
Table 3.1) and was always higher in the fertilized treatments (small, medium and large patch)
than in the unfertilized control, for all species at both levels of nutrient intensity. There was
a significant species x patch size interaction, indicating differences between species in the
pattern of response to the different treatments (Table 3.1). At both levels of nutrient
intensity, Achillea and Bromus had higher root N as the patch size increased (Figure 3.10).
This is an indication that these species are less able to adjust N uptake in response to
heterogeneity, because the plants have less root nitrogen at smaller patch sizes (even though
the total amount of fertilizer in the pot was constant). Both Silene and Rumex were able to

better adjust their uptake to nutrient heterogeneity, and had just as much root N regardless of

Figure 3.8 - Root biomass values i SE. for patch and background quadrants within root
systems at high and low nutrient intensity levels for each species. Open diamonds are the
means of root samples from unenriched quadrants. Solid squares are means of root samples
from enriched quadrants. All quadrants in the patch size 0 treatment were unenriched. All
quadrants in the patch size 4 treatment were enriched.

67

 

 

 

 

 

 

 

Low Nutrient Intensity High Nutrient Intensity

0.08
A a) Bromus b) Bromus

D
:006 — - E

a

50041- i g i

o .
3 5 E

30.02 . —
n:

O 1 1 1 1 1 1 1 1 1 1
0.08
c) Solidago d) Solidago

.0
o
a:

.— 1—

Root biomass (g)
C
' 'o
.5

O
O
N
n—H
iOii-I-l
i-IH

ihigi !

1

 

 

O
.-
_
_
_
_

 

P
o
co

 

e) Achillea f) Achillea

.0
o
a:
I
1

Root biomass (g)
0 O
O O
M as
i—O—H
ii-Q-H
i-O-I-I-i
i-l-i-i
HO!
H-I-I-i
III
i-I-I-i

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0 1 1 1 1 1 1 1 1 1 1
0.08
A g) Rumex h) Rumex
a
' 0.06 - ~
to
E
o 0 04 ~ ~
3 i
§002- I ~
11: I!
0 1 1 1 1 1 1 1 1 1 1
0.08
A 1) S" e J) Silene
3%06~ ¥' E —
3 1
a
5004- I i — i
.. 1
750.02 - -
o
a:
0 1 1 1 1 1 1 1 1 1 1
0 1 2 4 0 1 2 4
Patch size (No. quadrants enriched) Patch size (No. quadrants enriched)

Figure 3.8

68

Figure 3.9 - Nitrogen uptake rate i SE. for the five species as determined by 15N assay with
excised roots. Filled symbols with solid lines are dominant species, hollow symbols with
dashed lines are subordinate species.

69

A) Low Nutrient Intensity

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0.07 —
-- + Bromus
—e— Solidago
--o - - Achillea
- ~11: - - Rumex
~11 -- Silene
0 1 2 4
Patch size (No. quadrants enriched)
B) High Nutrient Intensity

0'07 I + Bromus
—e— Solidago

0-05 “ --o--Achillea
x -- Rumex

0.05 -" --A-- Silene

0.04 l‘

 

 

Uptake rate (uglglhr)

 

O 1 2 4
Patch size (No. quadrants enriched)

Figure 3.9

70

Figure 3.10 - Average %nitrogen concentration of whole roots systems i SE. at high and
low patch intensities levels for each species. Filled symbols with solid lines are dominant
species, hollow symbols with dashed lines are subordinate species.

71

A) Low Nutrient Intensity

 

+ Bromus
—e—— Solidago
-- o - - Achillea
- - at - - Rumex

- - a - - Silene

 

 

 

 

 

 

Patch size (No. quadrants enriched)

B) High Nutrient Intensity

 

7 I —I— Bromus
—e— Solidago
-- o - - Achillea
- - a: - - Rumex

- - a - - Silene

  
 
 
 

C3
1

 

 

 

 

 

 

l

 

 

  
  
   

l

N OD -F 01
L

   

l

%N in root tissues

 

 

O 1 2 4
Patch size (No. quadrants enriched)

Figure 3.10

72
the spatial arrangement of enrichment at both nutrient intensity levels. The highest root N
concentration of Rumex and Silene was lower than those obtained by the other three species
(Figure 3.10). Solidago also adjusted uptake to nutrient heterogeneity, although its ability to
adjust to the smallest patch size was diminished at the higher nutrient intensity. Nitrogen
concentrations for all species were higher in the largest two patch sizes (2 and 4 quadrants
enriched) at the higher nutrient intensity (Figure 3.10), as suggested by a significant patch

size x nutrient intensity interaction.

Patch vs. background differences in N uptake

There were indications that uptake rates between quadrants for some of the patch size
treatments differed. There was generally less variability in uptake in the control and large
patch size treatments (0 and 4 quadrants enriched, respectively) than in the small and medium
patch size treatments (1 and 2 quadrants enriched; see Figure 3.11).

There was a significant quadrant x species interaction for N concentrations. Although
all five species generally had higher root N concentrations in enriched patches compared with
background quadrants, Bromus, Achillea and Solidago had greater magnitudes of difference
between patch and background quadrants than Rumex and Silene (Figure 3.12). There were
also significant quadrant x patch size x nutrient intensity and quadrant x patch size
interactions, driven by greater variation between quadrants in the heterogeneous than in
homogeneous treatments, especially at the high nutrient intensity level. The higher nutrient
concentrations in the roots from unenriched quadrants of the heterogeneous treatments (1- and
2-quadrants enriched) than in roots from the treatment with no quadrants enriched are

indicative of some transport of nitrogen taking place within the root system.

73

Figure 3.11 - 15N uptake rate 1 SE. in patch and background quadrants within root systems
at high and low nutrient intensity levels for each species. Open diamonds are the means of
root samples from unenriched quadrants. Solid squares are means of root samples from
enriched quadrants. All quadrants in the patch size 0 treatment were unenriched. All
quadrants in the patch size 4 treatment were enriched.

0.1

0.08
g 0.06
E004

0.02

0.1
0.08
20.06
E1
g0.04

0.02

0.1
0.08
20.06
E
g004

0.02

0.1
0.08
g 0.06
2
30.04

0.02

0.1
0.08

g 0.06

2

g' 0.04

0.02

74

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Low N utrlent Intensity High Nutrient Intensity
a) Bromus b) Bromus
.. .. T’
- I -
’ i 5 ! g ” ‘ f
c) Solidago d) Solidago
e) Achillea f) Achillea
- L
- L
P s ¥ " i 5 i
’ l l 1 F l l l l -
g) Rumex h) Rumex

_

 

 

 

 

 

 

 

I) Silene j) Silene
- _ 1.
I <1 i
L. _
I I i ! I
0 1 2 4 0 1 2 4

Patch size (No. quadrants enriched)

Patch size (No. quadrants enriched)

Figure 3.11

 

 

75

Figure 3.12 - Root nitrogen concentration 1 SE. in patch and background quadrants within
root systems at high and low nutrient intensity levels for each species. Open diamonds are
the means of root samples from unenriched quadrants. Solid squares are means of root
samples from enriched quadrants. All quadrants in the patch size 0 treatment were
unenriched. All quadrants in the patch size 4 treatment were enriched.

 

N In root (‘4)
N «b O)

CDC

Nin root ($6)
to a. a:

one

Nin root (96)
N1 «5 a:

coo

N in root (7.)
oo 0 N 4h c»

O)

N in root (96)
N A

0

Low Nutrient Intensity

76
High Nutrient Intensity

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Patch size (No. quadrants enriched)

 

 

a) Bromus b) Bromus
- i _ I i
” i i “ I .
- _ I
I
’ 1 g 1 1 1 1 1 1 1
c) Solidago d) Solidago
_ g i _ . I 1
- § - E
J 1 1 l 4 ’ 1 l 4 1
e) Achillea f) Achillea
“ i i I
1’ i i P i
_ g . §
’ l l 1 1 ’ I 1 1 41
g) Rumex h) Rumex
_ i s l _ i i
I s
I) Silene J) Silene
- L i
" g 1 I * g I g
' s
0 1 2 4 0 1 2 4

Patch size (No. quadrants enriched)

Figure 3.12

77
DISCUSSION
This study shows that species differ in which mechanisms they used to respond to
nutrient heterogeneity, although there were no patterns linking response mechanisms to either
foraging strategy or dominance in the field. Species also differed in how many mechanisms
they used and in how well they could use a mechanism across the range of patch sizes or

nutrient intensity in these experiments (Table 3.2).

Architecture indices

All of the species had topological indices indicative of random branching architecture
except for Bromus, which had more herringbone root architecture. Although there was no
significant species difference in the pattern of how branching architecture changed across
patch size or nutrient intensity, there was some indication that Bromus might be able to adjust
its root architecture to less herringbone branching in enriched patches. Fast growing species,
like Bromus, are expected to be more likely to adjust their root architecture in response to a
patch (Fitter 1994). The fact that Bromus was more herringbone than the other species is
probably a general trait of grass species. Preliminary data from another grass species, Poa
compressa, also had higher alE(a) values than dicot species (Corbett, unpubl.). Taub and
Goldberg (1996) also found grasses to have a more herringbone root architecture than dicots.

There were significant differences among species in the pattern of root branching
density in response to nutrient heterogeneity, as measured by mean segment length (MSL).
Three of the species had no change in MSL across patch treatments: Bromus, Silene, and
Rumex had short MSL in both the unfertilized controls and fertilized treatments. Only
Solidago and Achillea exhibited plasticity in MSL and produced roots with larger MSL in

unfertilized soil and short MSL in enriched soil.

78

 

3:05:00: 88:8 37:00.5

:38 8:5. .5 .96

:83 :0 E 1 0:0: 0:0: 0:0: 08:03:: 8.» 0:0: 3553 0:05-

:08qu 5300.5

one :38 6:5. .5 .89

0:0: 8&8: a: 1 0:0: 0:0: 0:0: :83 8» 0:0: 0:08:00: 823%

589:»? 8928 56805

v 3e. BEE: .8 as: .5 .39

0:50:05 :80: 8 v 8» 8:? V 0:: .058 :83 8.» 0:0: 53:03:? 8330‘

3:05:00: 88:8 55005

:38 5 85:5 22 .5 .88

8» u 3.50:8 8» 8:3 v 0:0: :83 8.» 08: 3:303:00 3633“.

82800:: 585.05% 80.28 030m

50:»: V 050. 8: A032 :32 .509

a 8» :80: 8 V 0:0: 0:0: :5 .088 850:2: 8% 0:0: .3505 938$
80:08:06 839$ 808.8%: 889$ 80:08:08 830.»: 80:08am: 80%.?
0:88:00: 88 08:3 3:89—00: 58 08:3 “5:08:00: 82 05:3 0:50:38: 88 05:3

.m> :33 E 030:0 .m> :30“. E 050:0 .m> :83 E 093:0 .m> :30.“ 5 030:0 850%

 

 

 

 

.- 8380 2 so: .-. 50:3 8808 as: Ema mass:
i

.80§§:_ 3:805: 505:: a: 880% 8 05: :80: .3 880%
0:03:87. 083 0:05 :03? :8 8:35, 1 * _§:5o%a«u 0.32 .Efloazfifl 0:2 .80515 58:05.18: 68:65:81 new
48:58:" :80 8:285:88”: :8 85¢. :80: 888 «5:50:88 50:04:: .3 880:8: 880% no Saga - fin 030,—.

 

79

Biomass responses

None of the species increased total root biomass across patch sizes, although this
could be a function of adjusted root birth and death rates, which were not measured in this
study. While root systems had larger biomass at the high nutrient intensity and some species
had larger root biomass than others, there was no difference in the pattern of response via
root biomass that species exhibited across patch sizes. Instead, all five species responded to
nutrient heterogeneity by shifting root biomass allocation to patches. All species had higher
root biomass in patch quadrants than in background quadrants of the heterogeneous treatments
(small and medium patch size). This result is consistent with the pattern observed by
Robinson (1994), where an increase in root biomass in enriched patches resulted in a decrease
elsewhere in the root system. There were no differences among species in their ability to
selectively allocated biomass to patch quadrants, i.e. species did not differ in how much

variation in biomass there was between patch and background quadrants (Figure 3.8).

Uptake rate responses

In spite of the low power of the analysis of the 1"N uptake assay, there was some
indication that these species had higher uptake rates in heterogeneous treatments vs.
homogeneous treatments, and that there was more variability in uptake in heterogeneous
treatments. It was not clear if there were different uptake rates between patch and
background quadrants, and no indication of whether there might be differences among
species.

When root N concentration was used as an estimate of N uptake, there were
differences in how plant species responded to nutrient heterogeneity and in the abilities of
plant species to selectively use this mechanism within portions of the root system. Solidago,

Bromus, and Achillea had lower uptake per root in some of the heterogeneous treatments, as

80
indicated by lower root N concentrations in those treatments, even though total N in the pot
was the same across the enriched patch size treatments. These species, however, were able to
compensate for nutrient heterogeneity by having higher N concentrations in roots from
enriched patch quadrants than roots from unenriched quadrants. Silene and Rumex kept the N
concentrations in their roots relatively constant as patch size changed, but they had less
difference between patch and background quadrants within the root systems of heterogeneous
treatments. These differences in N concentration among species and patch size treatments
could also be due to differing rates of transport from the root. The fact that all species had
higher N concentrations in roots from unenriched quadrants in the heterogeneous treatments (1
and 2 quadrants enriched) than in roots from the unenriched control treatment indicates that
some transport of N is occurring. However, since nitrogen was supplied to enriched
quadrants right up until the moment of harvest, there will be N acquired through uptake that
could not be transported out of the roots before harvest. So the differences between enriched

and unenriched sectors are driven by uptake differences.

Summary of species responses

These five species exhibited differences in the mechanisms they used to respond to
nutrient heterogeneity (Fable 3 .2). However, these differences were not associated with
whether the species was a scale or precision forager, or a dominant or subordinate species in
the field. Instead, mechanisms appear to be more constrained by root system type. The
amycorrhizal species (Rumex and Silene) had lower %N concentrations than the other species,
although they were able to keep concentrations constant across treatments. These two species
and the grass species (Bromus and preliminary data from Poa; see Appendix) were densely
branched under all nutrient conditions. This may have limited them from exhibiting plasticity

in branching density in response to nutrient patches. There was some indication that Bromus

81
might be able to adjust root branching topology to be less herringbone in response to nutrient
patches, but it was not significant in this study.

Fitter (1994) found evidence of a negative correlation between biomass allocation
adjusters and species that changed root architecture and suggested that these might represent
alternative means for species to respond to nutrient heterogeneity. However, I did not find
any evidence for trade-offs between the mechanisms I investigated in this study. I found no
evidence that species used a response mechanism at only certain patch sizes or nutrient
intensities. I also found no evidence of differences in the magnitude of plasticity between
precision and scale foragers as patch size changed. If a species used a given response
mechanism, it was used with equal effectiveness across the range of patch sizes in the
experiments.

There was limited evidence that species differ in how precisely they can allocate a
response mechanism within the root system. There were no species differences in how much
they shifted biomass allocation within the roots, but there were differences in how species
adjusted nitrogen uptake in response to nutrient patches. Contrary to my hypothesis, the scale
forager (as well as two precision foragers) had greater difference in uptake between patch and
background quadrants than two of the precision foragers.

This study shows that linking the mechanisms of response by root systems to overall
plant performance in nutrient patches is a complex process. In my experiments, there was no
consistent pattern of individual response mechanisms being linked to foraging strategy, pattern
of dominance in the field, or overall performance of the species in patchy conditions.
Response mechanisms that were used by a species were determined more by root system
morphology and plant growth form. In addition, all species used more than one mechanism
when responding to nutrient heterogeneity. Further research would be valuable in

determining whether these mechanisms differ in the importance of the contribution they make

82
to the plant adjusting to nutrient heterogeneity or whether the temporal scale of heterogeneity

influences which mechanisms are used to respond to nutrient patches.

Chapter 4

USING SEMIVARIOGRAMS FOR COMPARISON OF SPATIAL PATTERNS IN
MULTIPLE ENVIRONMENTS: CHOOSING A SAMPLING SCHEME
(Manuscript co-authored with K.L. Gross and G.P. Robertson)
INTRODUCTION

Determining spatial patterns in environments is important to many ecological
questions. Many techniques have been developed for assessing spatial patterns (for example,
see Cressie 1991). Of these techniques, geostatistical analyses provide excellent means for
both quantifying the scale and magnitude of spatial structure in a variate (from the
semivariance analysis; Rossi et al. 1992, Robertson and Gross 1994) and for interpolating or
mapping the spatial distribution of a variate in an environment (using kriging algorithms; see
Robertson 1987).

Geostatistical analyses were originally developed to aid geologists in estimating the
locations of ore deposits. Ecologists have adopted geostatistics for many uses, such as
assessing the distributions of insects (e.g. Schotzko and O’Keeffe 1990; Liebhold er al. 1993);
mapping the abundances of migratory song birds (e. g. Villard and Maurer 1996); and
evaluating variability of soils (e. g. Robertson et al. 1988, Schlesinger et al. 1996)

A critical issue in any application of geostatistics is the choice of an apprOpriate
sampling design (Olea 1984, Yfantis et al. 1987), especially when there is insufficient
background information about the variate or relationship being studied. Many people in both
geology and ecology have sought ways to improve sampling designs. In fact, there is a wide

body of applied mathematical literature dedicated to the improvement of geostatistical

83

84
measures and interpretation. Some have assessed the best geostatistical sampling methods to
use for kriging and mapping purposes (Burgess er al. 1981, McBratney and Webster 1983,
Olea 1984, Trangmar et al. 1985, Oliver and Webster 1986, Robertson 1987, Yfantis er al.
1987). These assessments require the researcher to already have a semivariogram on which
to base his or her calculations of optimal sampling. Others have taken a more theoretical
approach, with the aim of minimizing sampling variance (e. g. Cressie 1991; Zimmerman and
Homer 1991; Brus and de Gruiter 1994) A few have assessed different sub-sampling
strategies within one sampled population (e. g. Fortin et al. 1989, Oliver and Webster 1986).
All these studies have focussed on customizing sampling strategies to the particular
environment being sampled such that a maximum amount of information about spatial patterns
in the environment is gained for a minimum of sampling effort.

Even though ecologists have used geostatistical techniques to aid in describing spatial
patterns in the environment, a more recent application of geostatistical methods to ecology is
to compare patterns of spatial structure in multiple environments, or the same environment
over time. This extends the use of geostatistics from the merely descriptive to the testing of
hypotheses. For example, Gross et al. (1995), studied whether the scale of heterogeneity in
soil changed between communities of different successional age. Schlesinger et al. (1996)
compared spatial distributions of soil nutrients between grassland and shrubland desert
ecosystems. Ryel er al. (1996) tracked the changes in soil spatial heterogeneity through a
growing season.

In the situation where we wish to make comparisons, spatial structure is being
determined concurrently in several locations or at the same location through time. The central
question we ask when selecting sampling locations then, is not "how do we customize our
spatial sampling to this one particular site? ", but "what spatial sampling scheme can I use in

all of my sites and still obtain a good estimate of spatial structure at each site?" Will the

85
same set of sampling points need to be used in each environment in order to have
methodological consistency? Given that the location of sampling points may affect our
estimate of spatial structure, we would like to be confident that the sampling set is
maximizing the information that we obtain from the data in all environments.
Our goal was to address this question by creating a simple model that would mimic
potential sampling schemes that an ecologist might use and investigating how those sampling

schemes fared in assessing the spatial pattern in a variety of environments.

METHODS

We chose three types of sampling regimes that we felt were commonly used by
ecologists: simple random (all sampling points randomly chosen); stratified grid (environment
divided into a grid and one sampling point randomly chosen within each grid cell); and
stratified-nested grid (same as stratified grid, except that, in addition, a sub-set of randomly
chosen grid cells will contain multiple sampling points).

We then generated very simple simulated environments that were squares of 50x50
cells. There were environments at three different scales of heterogeneity: 25 patches of 10x10
cells each, 100 patches (5x5), or 625 patches (2x2). Each patch was randomly assigned a
numerical value (0 to 4), with equal numbers of patches of each value in each environment
(see Figure 4.1). Five different replicates of each environment were generated by changing
the random assignment of the patches. and compared the variogram parameters estimated by
three different sampling regimes. We generated replicates of these environments so that we
could statistically compare the results of the semivariance analyses and determine which
sampling regime gave the most consistent results across a range of environments with

different scales of spatial structure.

86

Figure 4.1 - Examples of the three different environments with different patch sizes and
number of patches. Numerical values have been converted to shading for visual effect.
Increasing darkness equals increasing numerical value (i.e. white=0, black=4).

 

 

87

 

A) 10x10 patch size

 

 

 

 

 

 

 

B) 5x5 patch size

C) 2x2 patch size

    
     

 

 

   

II... II... II
Figure 4.1

88

We generated 100 sampling points using three different sampling regimes: simple
random, stratified grid and stratified-nested grid (see Figure 4.2). For the simple random
pattern, 100 points were randomly chosen from within the matrix. For the stratified grid, the
matrix was first divided into a grid of 5x5 cells (100 squares) and a point was randomly
selected from within each square. In the stratified-nested grid design, the matrix was divided
into a grid of 50 rectangular 5x10 cells and one point was randomly selected within each
rectangle. An additional 5 points were randomly selected in 10 of the rectangles to create the
nested structure of smaller scale sampling of the matrix. Three replicates of each sampling
regime were generated and all the environments were sampled by every sampling regime
replicate, for a total of 135 data sets.

We used GS + v2.1 (Gamma Design Software, 1990) to estimate the semivariance
parameters for every data set. Semivariograms were calculated using two-thirds of the
maximum lag and a step size of two. For consistency in comparing the semivariograms, we
fit a spherical model to all the semivariograms:

t(h)=Co+C[1.5(h/Ao)—(h/Ao)3] for h 5 A0
and t(h)=Co+C for h> A0
where h=lag interval, Co=nugget variance (20), C=structural variance (Ace) and

Ao=range (see Figure 4.3). In most cases (86.5%) this model generally gave the best fit (by
lowest residual sum of squares) or had an RSS within 0.1 of the RSS for the best fit model.
The semivariance parameters of the 15 replicates for each environment type x sampling
regime were averaged and used to produce an average semivariogram with standard error
bars.

We used repeated measures AN OVA to compare the estimates of the semivariogram
parameters (sill, range and nugget) with different patch sizes and sampling regimes. The

nugget variance (C0) is expected to be zero if all the variance is spatially structured, otherwise

89

Figure 4.2 - Examples of the three different sampling regimes, including the grid divisions
that were used (if any).

A) Simple Random

 

 

 

 

 

 

 

 

 

 

 

 

 

 

so 40 so
I I I
I
I I I
I
I
II I
I I
I
III I
>- I I
I
I
I I
I: I =
I I
« I II I
I I I
._l_l_lllll J
X
8’ Stratified Grid 0 5 1o 15 20 25 so 35 4o 45 so
>.
C) Stratified-nested Grid
).

 

 

 

 

 

 

 

 

 

 

 

 

 

 

91

Figure 4.3 - Idealized semivario '
sill (C4420). grams ShOng the parameters of nugget (Co). range (A0) and

OODN""“N<"“BGU)

 

 

Range I (Ao-I) Range 11 (Ao-II)

 

Sill (population variance)

 

Distance Interval (h)

Figure 4.3

93

it represents the variance that is not spatially structured or is structured at a scale smaller than
that sampled. The sill (C0+C) is expected to be equal to the population variance. The range
(A) is the distance at which spatial dependence no longer occurs. A0 can be used to
approximate the scale of heterogeneity (or the size of patches) in the environment (see Figure
4.3). All analyses were performed using SYSTAT (Systat Inc. 1992).

To assess whether a sampling regime gave "better" estimates of spatial structure, we
used four criteria: a "good” sampling regime will produce consistent estimates of range (A);
a "good" sampling regime will consistently produce nugget estimates close to zero; a "good"
sampling regime will have little variation in the semivariance values at lag distances smaller
than the A0 distance; as a corollary of this, a "good” sampling regime will have more data

pairs used in the semivariance calculation at small lag distances.

RESULTS
All three sampling regimes gave similar, and accurate, estimates of the true population

mean and variance for patch value (2.001200).

Range estimates

The sampling regimes did not differ significantly in their estimates of the range (A0),
but estimates from the random sampling regime were more variable, especially at the
2x2patch size (see Table 4.1). All three sampling regimes showed a decrease in range as

patch size decreased.

Nugget estimates
If all the variance in an environment is spatially structured and the sampling regime is

at the appropriate scale, the nugget (Co) of the semivariogram will be zero. The estimated

.

 

m: 0: 3..
08.00080 80.00020: 50.00020 .8800
$000800 00285.2 08038.0 882
08.00800 30.3032 83800.0 2:0 0:5:
as." 8 *azw
3.0.3800 82002.0 08.08000 880:
€00.80: 08.0300: 08.0380 882
E00320 E08000 000.880.: 0:0 ca
8 a .2.
08.00.0000 0:000:80: 08.0802 885:
08.0880 80.82.00 83.03.00 382
08.00.0000 03.0030. 000.0880 000 3
3+0: :5 30 005: 000 00002 0802 08800 one :28

.280 V A" is. . 5.0 v mu 5.. .30 v mu L 0::0> :088000008005 0:: .0: 00:08.00 .<>OZ<
880008 080000: 80.: 0088.88: 0:03 80:08:06 0:005:88 .080 :80: 8808:0880 :000 8:03 8880: 085800 :0038: 0008
0:03 0808:0088 8:080:00 :000 :0": .8880: 82:80.: 00:5 0:: :00 Am: "5 80880.00 8080:0383 00 800808 0w0:0>< - u... 030,—.

95
nugget differed among sampling regimes (Table 4.1) and the stratified-nested sampling regime

consistently had the lowest, and least variable, estimate of nugget.

Semivariance estimates

The stratified-nested design has more pairs of points at lower lag classes (see Table
4.2). The random sampling regimes has no sampling points in the first lag class, and the grid
sampling regime has an average of 3 pairs at the smallest lag distance.

The nested sampling regime provided a less variable estimate of the semivariance at
the lowest lag classes (see Figure 4.4). Particularly, at the smallest lag distance (where the
random regime has no sampling points), the magnitude of the standard error from the nested
sampling regime is lower than that of the grid sampling regime at all three patch sizes. In the
second lag class, the standard errors for all three sampling regimes are similar, but the nested
sampling regime is slightly less variable than the grid at 5x5 and 10x10 patch sizes, and

slightly less than the random regime at 2x2 and 10x10 patch sizes.

DISCUSSION
Range estimates

The range of the semivariogram is defined as the lag distance at which the
semivariogram reaches the sill; at distances larger than the range, points are spatially
independent of each other (Yfantis et al. 1987, Robertson and Gross 1994). Thus the range is
equal to the radius of the area over which spatial dependence occurs, and can be considered
an estimate of the mean patch size in the environment. While there were not differences
between sampling regimes in the estimates they produced, the random regime was less
consistent. All three sampling regimes were able to detect the mean patch size in the

environments, as shown by the decrease in range as patch size decreased.

Table 4.2 - Average number of pairs used to estimate the semivariance at the first eight lag

classes, using a step size of two.

 

Lag Class Grid Nested Random
1 3 14 0
2 36 92 54
3 106 1 11 108
4 142 154 142
5 171 191 148
6 195 234 226
7 250 231 197
8 227 246 245

97

 

 

8.0 030:. 8 m0m0:0>0

0:: 80:: 008030 Ao< 000 00+ 0 50: 8308800 0:: :83 _0008 .0808? 0:: 8 0200 008: 0::. 830:0 0:0 8388 000880 :000 :0 80:0
0:0000:m Am: "5 00:00:00: 88—0800 000 880800880 08:00 0:03:00 000880 m0: :000 :0: 000.8000 0008:0888 0:: 80:: 00m0:0>0 0:03
m080> 00880: 880800 80:0080 0:: M80: 00:80:80 3:000 880800880 008: 0:: :0: 00:08:00 0808808800 0M0:0>< - 0.0 0:08.:

98

O _. x00 mN_m IOF<Q

g3
60 a 8 av O
H 1 d 1

 

 

 

 

 

 

 

 

0.0 0::»...—

mxm MNE IOF<Q

83.8 as!
9 00 8 o: 0
I 1 d d

 

 

 

 

 

L.

 

wa MNE IO._.<Q

9
1

8
1

gas.
8 c—
d

 

4

 

 

P

 

 

DUE) pelSGN [MODUBH

EWIEEH SNI—ldWVS

DUE)

99

Note that the range provides an estimate of mean patch size only. In the simulated
environments we sampled for these analyses, all patches are the same size in a given
environment and there are sharp boundaries between patches. So although the range provides
a good estimate of patch size in our examples, they are very simplistic compared to natural
environments. In natural environments, patch sizes can vary considerably, and boundaries
will be much more gradual, such that the range does not wholly represent patch size. For
example, an environment could consist of a mixture of small and large patches which would
result in an estimated "intermediate range" over which there is spatial dependence.
Alternatively, if the sampling is sufficiently fine-scaled, it may be possible to detect this as a
"nested structure" with the semivariogram (c.f. Robertson and Gross 1994; Gross er al.

1995).

Nugget estimates

If all the variance in an environment is spatially structured and the sampling regime
is at the appropriate scale, the nugget (Co) of the semivariogram will be zero. Low and
consistent estimates of the nugget are important in semivariance analysis, because it is used to
estimate the magnitude of spatial structure in a variate (Co/C+Co; e.g. Gross et al. 1995).
The stratified nested sampling regime was able to consistently provide low estimates of the
nugget over a range of patch sizes, while the other two sampling regimes produced higher
nugget values. This indicates that the nested sampling regime is more likely to capture the
spatially-structured variance in environments when the same set of sampling points is used

across several locations.

100
Semivariance estimates

The level of confidence in a semivariogram depends on the number of sample pairs
used in the calculation of each point, particularly at the smaller lag distances, as these
determine the nugget (Co) and the range (A0) over which there is spatial dependence. Based
on this criteria, we are most confident in the estimates produced by the stratified-nested
sampling design. The larger number of pairs at shorter lags generated by the stratified-nested
design (Table 4.2) probably accounts for this design giving a lower, and more consistent,
estimate of the nugget (Co).

The random sampling regime cannot be used to assess semivariance at scales as small
as the other two regimes, because of the lack of sampling points in the first lag class used.
We expected more variation in the semivariance estimate when the number of data pairs used
in the calculation is small. This was observed with the grid sampling regime, where more

variation was observed at the lag class that had only three data pairs.

Conclusions and recommendations

The results of our analyses suggest that a stratified-nested grid sampling regime is
better than a random or grid system of sampling when one wishes to use geostatistics to
generate a set of semivariograms from several environments sampled concurrently, especially
when the scales of heterogeneity are not well known before the sampling takes place . The
greater consistency of the stratified-nested grid sampling pattern in detecting spatial structure
over a broad range of patch sizes is the result of better estimating the nugget term of the
semivariogram. The lower and less variable nugget is due to the increased number of pairs at
short lag distances from the smaller scale sampling and thus less variation in the semivariance

estimates at those lag distances. The lower end of the semivariogram is the critical area of

 

101
interest for semivariance analysis, because it is where the range, nugget and sill are
determined.

In addition to providing lower and more consistent estimates of the nugget, the
stratified-nested regime could be used to examine spatial structure over a greater distance than
the other two sampling regimes. When the spatial scale of greatest interest may not be known
it allows the assessment of heterogeneity at more than one level (e.g. are belowground
resource levels patterned at the scale of individual roots, whole plants, entire fields, or some
combination? Robertson and Gross 1994), using a sampling pattern that incorporates a broad
range of spatial scales (such as the stratified-nested grid sampling pattern).

Other authors have also recommended using a stratified-nested sampling scheme for
sampling ecological variables in a single environment (Oliver and Webster 1986, Fortin et al.
1989). Our investigation shows that this type of sampling regime will also provide more
consistent estimates than a simple random or stratified grid design of the scale and magnitude
of spatial structure when used in a study comparing several different environments. This
should be encouragement for ecologists to adapt geostatistical techniques to ecological
questions, and to expand the uses of these techniques further, from a range of descriptive

studies into the realm of hypothesis testing.

Chapter 5

DISCUSSION

The studies reported in this dissertation demonstrate that there are species differences
in ability to respond to nutrient heterogeneity and in which mechanisms they use when
responding. These experiments also provides some support for the hypothesis that plants
experience a trade-off between scale and precision in foraging strategy. Consistent differences
among species in the ability to exploit nutrients in a heterogeneous environment could
potentially be important in influencing community structure. The next step is to determine
how foraging strategy translates into competitive ability under different conditions. Campbell
et al.(1991) found evidence that competitive dominants were scale foragers, but in my study
only one of the two dominant species exhibited a scale-type foraging strategy. Instead, I
found that foraging strategy was not related to dominance in the field per se, but to the size of
the plant. Large, fast-growing plants were the dominants in Campbell et al.’s (1991) study
and they were scale foragers, but in my experiments one dominant species was a large scale
forager and the other was a small precision forager. In both studies, subordinate species were
smaller and scale foragers.

My initial hypothesis was that scale foragers are competitively more dominant in
homogeneous conditions, pre-empting resources in the soil. Precision foragers can exploit
any small pockets of nutrients left in between the depletion zones created by scale foragers’
roots, but I would also hypothesize that precision foragers competitive ability (relative to scale

foragers) would improve as the environment becomes more heterogeneous. I found that the

102

103
scale forager always obtained more nitrogen than the precision-foraging species. As patch
size decreased, however, the amount of nitrogen obtained by the scale forager decreased,
while, for the precision foragers, the amount of nitrogen acquired was independent of patch
size. These results suggest that when patches become sufficiently small or of low intensity,
then precision forager would acquire more N than the scale forager, resulting in a competitive
advantage.

The next step would be to conduct a series of experiments comparing the competitive
abilities of various scale- and precision-foraging species in environments where soil conditions
have different nutrient patch sizes and intensities. Under heterogeneous conditions with many
smaller patches, I hypothesize a better performance by precision foragers, while in
homogeneous conditions I would expect scale foragers to dominate.

If it is determined that the relative competitive abilities of plant species are influenced
by the scale of nutrient heterogeneity in the environment, then determining the scale of
heterogeneity in an environment will be an important component of understanding community
structure. One way to assess the scale of nutrient heterogeneity is to use geostatistical tools.
The modelling used in chapter 4 indicated that a nested-grid sampling design would provide
the most consistent assessment of spatial scales of heterogeneity across a range of locations.
Consistency in the determination of scale of heterogeneity will become even more important
when the modelling of plant responses to heterogeneity is expanded to include a temporal
component (Ehrenfeld et al. 1997). The same environment would be sampled multiple times
through the growing season and the scale at which heterogeneity is important to the plants

could change drastically through the season.

APPENDIX

APPENDIX

Table A.1 - Preliminary data for Daucus carota and Poa compressa. Data values for two

additional species used in these experiments, listed as mean 1 standard error. Data that are

quadrant means list first the mean of the background quadrant (patch size of 1 or 2

quadrants) or the first quadrant sampled (patch size of 0 or 4 quadrants) and then the mean of

the enriched quadrant (patch size of 1 or 2 quadrants) or the second quadrant sampled (patch

size of 0 or 4 quadrants). T For Daucus at low nutrient intensity, n=1 where no standard

error is given. :1: For Daucus at high nutrient intensity, n=4 for shoot N concentration, total

plant N content, and root N concentration. * N/A = no data available.

 

 

Variable N o. of Daucus carota Daucus carota Poa compressa
quadrants (n=2)‘l' (n=6)i (n=3)
enriched Low nutrient High nutrient Low nutrient

intensity intensity intensity

Total Plant 0 0.171012 01210.02 0.141004

Biomass (g)

1 0.151007 02110.03 03610.07
2 01210.01 0.271006 0.321003
4 0.171001 0.221022 03610.04

Shoot N 0 1.161011 1.661014 0.141010

Concentration

(%) 1 5.511027 3.891031 4.001061
2 6.441020 5.251052 4.881025
4 7.131030 4.571045 4.701046

 

104

105

Table A. 1 (cont’d)

 

 

 

 

Variable N o. of Daucus carota Daucus carota Poa compressa
quadrants (n=2) (n=6) (n=3)
enriched Low nutrient High nutrient Low nutrient

intensity intensity intensity

Total Plant N 0 0.131008 01410.04 0.061002

C te t

on " (mg) 1 0.82 0.7910.07 1.281046
2 0.811008 1.531029 1.641015
4 1.121002 1.201013 1.451012

alE(a) - plant 0 1.301014 1.461011 2.141019

means
1 1.231008 1.421014 1.641019
2 1.241008 1.321014 1.991012
4 1.231015 1.071011 1.531021

- quadrant means 0 1.471015 1.381012 2.011018

1.141020 1.531010 2.271035
1 1.301002 1.401013 1.601033
1.151017 1.451015 1.681028
2 1.221004 1.261013 1.951020
1.251018 1.371017 2.031019
4 1.091025 1.021009 1.361027
1.371020 1.121014 1.701035

Mean segment 0 3.361070 3.421050 1.351007

length (mm) -

plant means 1 2.581040 2.491030 1.551011
2 2.471018 2.191022 1.491005
4 2.731034 2.171020 1.431003

- quadrant means 0 2.701066 3.351063 1.441012

4.021127 3.491041 1.261008
1 3.001078 2.881031 1.451018
2.161000 2.101021 1.671017
2 2.751012 2.431026 1.391005
2.201019 1.941012 1.591009

 

106

Table A.1 (cont’d)

 

 

 

 

Variable No. of Daucus carota Daucus carota Poa compressa
quadrants (n=2) (n=6) (n=3)
enriched Low nutrient High nutrient Low nutrient

intensity intensity intensity

MSL (mm) - 4 2.661073 2.081005 1.441007

quadrant means 2.801037 2.261028 1.441005

Root biomass (g) 0 0.101008 00710.01 0.081002

' plant means 1 0.0710.05 01010.02 01010.01
2 0.051001 0.101002 01210.08
4 0.061000 0.081001 00710.01

- quadrant means 0 0.011001 0.021001 0.031000

00510.05 00210.00 0.021001
1 00110.01 0.011000 00310.01
00110.01 00410.01 00410.01
2 00010.00 00410.01 00210.00
00310.00 00210.01 0.041001
4 0.011000 0.021001 0.041000
00110.01 0.021001 0.031000

15N uptake rate 0 N/A* 0.021001 0.031001

lhr

("3 ) 1 N/A 0.061002 0.031001
2 N/A 00410.01 0.021000
4 N/A 0.031001 0.021000

- quadrant means 0 N/A 0.021001 00210.01

00110.01 00310.01
1 N/A 00510.02 0.021000
00710.05 0.031001
2 N/A 0.03 0.021000
0.041002 0.021000
4 N/A 0.051003 0.031001
0.021001 0.021000

 

107

Table A.1 (cont’d)

 

 

Variable No. of Daucus carota Daucus carota Poa compressa
quadrants (n=2) (n=6) (n=3)
enriched Low nutrient High nutrient Low nutrient

intensity intensity intensity

Root N 0 04910.06 05010.06 07710.00

centration

fig“) _ pm 2.66 4.0910.20 1.9210.39

means 2 5 .6611 .26 5.691055 2.601050
4 61911.07 6.981058 2.711044

- quadrant means 0 0.501007 04710.14 0.761006

0.51 0.521016 0.781006
1 35911.99 2.811031 1.671045
3.73 5.371015 2.171033
2 4.311237 5.141064 2.141063
7.011015 6.231081 3.061040
4 6.341073 6.951074 2.661040
60511.40 7.001091 3.201029

LIST OF REFERENCES

LIST OF REFERENCES

Anderson, R.N. 1968. Germination and Establishment of Weeds for Experimental Purposes.
Weed Science Society of America, Urbana, IL. 236 pgs.

Anghinoni, I. and SA. Barber. 1980. Phosphorus application rate and distribution in the
soil and phosphorus uptake by corn. Soil Sci. Soc. Am. J. 44: 1041-1044.

Bengtsson, 1., T. Fagerstrom and H. Rydin. 1994. Competition and coexistence in plant
communities. Trends in Ecology and Evolution 9: 246-250.

Bernston, G.M. 1992. A computer program for characterizing root system branching
patterns. Plant and Soil 140: 145-149.

Bernston, G.M. 1995. The characterization of topology: a comparison of four topological
indices for root binary trees. Journal of Theoretical Biology 17: 271-281.

Bernston, G.M. [Online] BranChing documentation version 2.13. Available
http://plantecohost.harvard.edu/gmbWWW. Jan 14, 1997.

Borkert, CM. and SA. Barber. 1985. Soybean shoot and root growth and phosphorus
concentration as affected by phosphorus placement. Soil. Sci. Soc. Am. J. 49:
152-155.

Brus, DJ. and 1.]. de Gruijter. 1994. Estimation of non-ergodic variograms and their
sampling variance by design-based sampling strategies. Mathematical Geology 26:
437-453.

Burbank, D.H., K.S. Pregitzer and K.L. Gross. 1992. Vegetation of the W.K. Kellogg
Biological Station, Kalamazoo County, Michigan. Research Report 510 Michigan
State University Agricultural Experiment Station, East Lansing, MI.

Burgess, T.M., R. Webster, and AB. McBratney. 1981. Optimal interpolation and
isarithmic mapping of soil properties. IV. Sampling strategy. Journal of Soil Science
32: 643-659.

Caldwell, MM. 1994. Exploiting nutrients in fertile soil microsites. In: M.M. Caldwell
and R.W. Pearcy (eds.) Exploitation of Environmental Heterogeneity by Plants:
Ecophysiological Processes Above and Belowground. Academic Press, NY. pp.
325-347.

108

109

Caldwell, M.M, L.M Dudley and B. Lilieholrn. 1992. Soil solution phosphate, root uptake
kinetics and nutrient acquisition: implications for a patchy soil environment.
Oecologia. 89: 305-309.

Campbell, BB. and J .P. Grime. 1989a. A comparative study of plant responsiveness to the
duration of episodes of mineral nutrient enrichment. New Phytologist 112: 261-267.

Campbell, BB. and J .P. Grime. 1989b. A new method of exposing developing root systems
to controlled patchiness in mineral nutrient supply. Annals of Botany 63: 395-400.

Campbell, B.D., J.P. Grime and J .M.L. Mackey. 1991. A trade-off between scale and
precision in resource foraging. Oecologia 87: 532-538.

Casper, BB. and J .F . Cahill, Jr. 1996. Limited effects of soil nutrient heterogeneity on

populations of Abutilon theophrasti (Malvaceae). American Journal of Botany 70:
333-341.

Casper, BB. and RB. Jackson. 1997. Plant competition underground. Annual Review of
Ecology and Systematics. 28: 545-570.

Cressie, N .A.C. 1991. Statistics for Spatial Data. John Wiley & Sons, Inc., New York,
New York. 900 pgs.

Drew, M.C. and LR. Saker. 1975. Nutrient supply and the growth of the seminal root
system in barley H. Localized, compensatory increases in lateral root growth and rates
of nitrate uptake when nitrate supply is restricted to only part of the root system.
Journal of Experimental Botany 26: 79-90.

Ehrenfeld, J.G., X. Han, W.F.J. Parsons and W. Zhu. 1997. On the nature of
environmental gradients: temporal and spatial variability of soils and vegetation in the
New Jersey pinelands. Journal of Ecology 85: 785-798.

Eissenstat, D.M. 1992. Costs and benefits of constructing roots of small diameter. Journal
of Plant Nutrition. 15: 763-782.

Fitter, A.H. 1985. Functional significance of root morphology and root system architecture.
In: A.H. Fitter (ed.) Ecological Interactions in the Soil: Plants, Microbes and
Animals. Blackwell Scientific Publishing, Oxford. pp. 87-106.

Fitter, A.H. 1987. An architectural approach to the comparative ecology of plant roots
systems. New Phytologist 106(suppl.): 61-77.

Fitter, A.H. 1994. Architecture and biomass allocation as components of the plastic
response of root systems to soil heterogeneity. In: M.M. Caldwell and R.W. Pearcy
(eds.) Exploitation of Environmental Heterogeneity by Plants: Ecophysiological
Processes Above and Belowground. Academic Press, NY. pp. 305-323.

Fitter, A.H. and TR. Stickland. 1992. Architectural analysis of plant root systems III.
Studies on plants under field conditions. New Phytologist 121: 243-248.

110

Fitter, A.H., T.R. Stickland, M.L. Harvey and G.W. Wilson. 1991. Architectural analysis
of plant root systems. 1. Architectural correlates of exploitation efficiency. New
Phytologist. 118: 375-382.

Fortin, M.-J., P. Drapeau, and P. Legendre. 1989. Spatial autocorrelation and sampling
design in plant ecology. Vegetatio 83: 209-222.

Foster, BL. 1996. Plant competition and diversity in relation to productivity in old-field
plant communities. Ph.D. Dissertation, Michigan State University, East Lansing, MI.

Foster, BL. and K.L. Gross. 1997. Partitioning the effects of plant biomass and litter on
Andropogon gerardi in old-field vegetation. Ecology 78: 2091-2104.

Gamma Design Software. 1990. 68+, version 2.1. Gamma Design Software, Plainwell,
MI.

Gleason, HA. and A. Cronquist. 1991. Manual of Vascular Plants of Northeastern United
States and Adjacent Canada, 2nd Edition. The New York Botanical Garden, Bronx,
NY. 910 pgs.

Gleeson, SK. and J .E. Fry. 1997 . Root proliferation and marginal patch value. Oikos 79:
387-393.

Goldberg, DE. 1987. Neighborhood competition in an old-field plant community. Ecology
68: 1211-1223.

Goldberg, DE. and K.L. Gross. 1988. Disturbance regimes of midsuocessional old fields.
Ecology 69: 1677-1688.

Goldberg, DE. and RA. Werner. 1983. The effects of size of opening in vegetation and
litter cover on seedling establishment of goldenrods (Solidago spp.). Oecologia
60: 149-155.

Grime, J .P. 1994. The role of plasticity in exploiting environmental heterogeneity. In:
M.M, Caldwell and R.W. Pearcy (eds.) Exploitation of Environmental Heterogeneity
by Plants: Ecophysiological Processes Above and Belowground. Academic Press,
NY. pp. 2-19.

Grime, J.P., B.D. Campbell, J.M.L. Mackey and J.C. Crick. 1991. Root plasticity,
nitrogen capture and competitive ability. In: D. Atkinson (ed.) Plant Root Growth:
an Ecological Perspective. Blackwell Scientific Publications, Oxford, England. pp.
381-397.

Gross, K.L., A. Peters and KS. Pregitzer. 1993. Fine root growth and demographic
responses to nutrient patches in four old-field plant species. Oecologia 95: 61-64.

Gross, K.L., K.L. Pregitzer and AJ. Burton. 1995. Spatial variation in nitrogen availability
in three successional plant communities. Journal of Ecology 83: 357-367.

111

Harley, LL. and EL. Harley. 1987. A check-list of mycorrhiza in the British flora. New
Phytologist 105(suppl.): 1-102.

Harris, D. and EA. Paul. 1989. Automated analysis of 15N and 1‘C in biological samples.
Commun. Soil Sci. Plant. Anal. 20: 935-947.

Hendry, G.A.F. and J .P. Grime. 1993. Methods in Comparative Plant Ecology. Chapman
and Hall, New York, NY. 252 pgs.

Hetrick, B.A.D. 1991. Mycorrhizas and root architecture. Experientia 47: 355-362.

Hetrick, B.A.D., G.W.T. Wilson and J.F. Leslie. 1991. Root architecture of warm- and
cool- season grasses: relationship to mycorrhizal dependence. Canadian Journal of
Botany 69: 112-118.

Houssard, C. and J. Escarré. 1991. The effects of swd weight on growth and competitive
ablity of Rumex acetosella from two successional old-fields. Oecologia 86: 236-242.

Houssard, C. and J. Escarré. 1995. Variation and covariation among life-history traits in
Rumex acetosella from a successional old-field gradient. Oecologia 102: 70-80.

Huberty, L.E., K.L. Gross and CL Miller. 1998. Effects of nitrogen addition on
successional dynamics and species diversity in Michigan old fields. Journal of
Ecology (in press).

Hutchings, MJ. and H. de Kroon. 1994. Foraging in plants: the role of morphological
plasticity in resource acquisition. Advances in Ecological Research 25: 159-238.

Isaaks, E.H. and RM. Srivastava. 1989. An Introduction to Applied Geostatistics. Oxford
University Press, New York, New York. 561 pgs.

Jackson, RB. and M.M. Caldwell. 1989. The timing and degree of root proliferation in
fertile-soil microsites for three cold-desert perennials. Oecologia 81: 149-154.

Jackson, RB. and M.M. Caldwell. 1993. Geostatistical patterns of soil heterogeneity around
individual perennial plants. Ecology 81: 683-692.

Jackson, R.B., J .H. Manwaring and M.M. Caldwell. 1990. Rapid physiological adjustment
of roots to localized soil enrichment. Nature 344: 58-60.

Jackson, RB. and H.L. Reynolds. 1996. Nitrate and ammonium uptake for single- and
mixed-species communities grown at elevated C02. Oecologia 105: 74-80.

Kosola, K.R. and A.J. Bloom. 1994. Methylammonium as a transport analog for ammonium
in tomato (Lycopersicon esculentum L.). Plant Physiology 105: 435-442.

Kosola, K.R. and A.J. Bloom. 1996. Chlorate as a transport analog for nitrate absorption by
roots of tomato. Plant Phsyiology 110: 1293-1299.

112

Liebhold, A.M., R.E. Rossi and WP Kemp. 1993. Geostatistics and geographic
information systems in applied insect ecology. Annual Review of Entomology 38:
303-327.

McBratney, A.B., and R. Webster. 1983. How many observations are needed for regional
estimation of soil properties? Soil Science 135: 177-183.

McNeill, J. 1977. The biology of Canadian weeds 25. Silene alba (Miller) E.H.L. Krause.
Canadian Journal of Plant Science 57: 1103-1114.

Newcomb, L. 1977. Newcomb’s Wildflower Guide. Little, Brown and Company, Boston,
MA. 490 pgs.

Olea, RA. 1984. Sampling design optimization for spatial functions. Mathematical Geology
16: 369-392.

Oliver M.A., and R. Webster. 1986. Combining nested and linear sampling for determining
the scale and form of spatial variation of regionalized variables. Geographical
Analysis 18: 227-242.

Reynolds, H.L., B.A. Hungate, F.S. Chapin III, and GM. D’Antonio. 1997. Soil
heterogeneity and plant competition in an annual grassland. Ecology 78: 2076-2090.

Robertson, G.P. 1987. Geostatistics in ecology: interpolating with known variance. Ecology
68: 744-748.

Robertson, G.P. and K.L. Gross. 1994. assessing the heterogeneity of belowground
resources: quantifying pattern and scale. In: M.M. Caldwell and R.W. Pearcy (eds).
Exploitation of Environmental Heterogeneity by Plants: Ecophysiological Processes
Above and Belowground. Academic Press, NY. pp. 237-253.

Robertson, G.P., M.A. Huston, F.C. Evans and 1M. Tiedje. 1988. Spatial variation in a
successional plant community: patterns of nitrogen availability. Ecology 69:
1517-1524.

Robinson, D. 1994. Tansley Review No. 73 The responses of plants to non-uniform
supplies of nutrients. New Phytologist 127: 635-674.

Robinson, D. 1996. Resource capture by localized root proliferation: why do plants bother?
Annals of Botany 77: 179-85.

Robinson, D. and 1.11. Rorison. 1983. A comparison of the responses of Lolium perenne L.,
Holcus lanatus L. and Deschampia flexuosa (1..) Trin. to a localized supply of
nitrogen. New Phytologist 94: 263-273.

Rossi, R.E., D.J. Mulla, A.G. Journel, and EH. Franz. 1992. Geostatistical tools for
modelling and interpreting ecological spatial dependence. Ecological Monographs 62:
277-314.

113

Ryel, R.J., M.M. Caldwell and J .H. Manwaring. 1996. Temporal dynamics of soil spatial
heterogeneity in sagebrush-wheatgrass steppe during a growing season. Plant and Soil
184: 299-309.

Schlesinger, W.H., J.A. Raikes, A.E. Hartley and A.F. Cross. 1996. On the spatial pattern
of nutrients in desert ecosystems. Ecology 77 : 364-374.

Schluter, D. and RE. Ricklefs. 1993. Species diversity: an introduction to the problem.
In: R.E. Ricklefs and D. Schluter (eds.) Species Diversity in Ecological
Communities: Historical and Geographical Perspectives. University of Chicago Press,
Chicago, IL. pp. 1-10.

Schotzko, D.J., and LE. O’Keeffe. 1990. Effect of sample placement on the geostatistical
analysis of the spatial distribution of Lygus hesperus (Heteroptera: Miridae) in lentils.
Journal of Economic Entomolgy 83: 1888-1900.

Stubbendieck, J ., S.L. Hatch and KJ. Hirsch. 1986. North American Range Plants, Third
Edition. University of Nebraska Press, Lincoln, NE. 465 pgs.

Systat Inc. 1992. SYSTAT for Windows Version 5.0. Systat Inc, Evanston, IL.

Taub, DR. and D. Goldberg. 1996. Root system topology of plants from habitats differing
in soil resource availability. Functional Ecology 10: 258-264.

Tilman, D. 1988. On the meaning of competition and the mechanisms of competitive
superiority. Functional Ecology 1: 304-315.

Tilman, D. 1988. Plant Strategies and the Dynamics and Structure of Plant Communities.
Princeton University Press, Princeton, NJ. 360 pgs.

Tilman, D. 1990. Mechanisms of plant competition for nutrients: the elements of a
productive theory of competition. In: J.B. Grace and D. Tilman (eds.) Perspectives
on Plant Competition. Academic Press, New York, NY. pp. 117-141.

Tilman, D. 1994. Competition and biodiversity in spatially structured habitats. Ecology 75:
2-16.

Tilman, D. and S. Pacala. 1993. The maintenance of species richness in plant communities.
In: R.E. Ricklefs and D. Schluter (eds.) Species Diversity in Ecological
Communities: Historical and geographical perspectives. University of Chicago Press,
Chicago, IL. pp. 13-25.

Trangmar, B.B., R.S. Yost, and G. Uehara. 1985. Application of geostatistics to spatial
studies of soil properties. Advances in Agronomy 38: 45-94.

van Ommen, H.C., L.W. Dekker, R. Dijksma, J. Hulshof, and W.H. van der Molen. 1988.
A new technique for evaluating the presence of preferential flow paths in
non-structured soils. Soil Sci. Soc. Am. J. 52: 1192-1193.

114

Van Pelt, J. and R.W.H. Verwer. 1984. Cut trees in the topological analysis of branching
patterns. Bulletin of Mathematical Biology 46: 283-294.

van Vuuren, M.M.I., D. Robinson and BS. Griffiths. 1996. Nutrient inflow and root
proliferation during the exploitation of a temporally and spatially discrete source of
nitrogen in soil. Plant and Soil 178: 185-192.

Villard, M.-A., and BA. Maurer. 1996. Geostatistics as a tool for examining hypothesized
declines in migratory songbirds. Ecology 77: 59-68.

Warwick, SJ. and L. Black. 1982. The biology of Canadian weeds 52. Achillea millefolium
L. s.l. Canadian Journal of Plant Science 62: 163-182.

Werner, C. and J .8. Smart. 1973. Some new methods of topologic classification of channel
networks. Geographical Analysis 5: 271-295.

Werner, P.A., A.K. Bradbury and R.S. Gross. 1980. The biology of Canadian weeds 45.
Solidago canadensis L. Canadian Journal of Plant Science. 60: 1393-1409.

Yfantis, E.A., G.T. Flatrnan, and J.V. Behar. 1987. Efficiency of kriging estimation for
square, triangular and hexagonal grids. Mathematical Geology 19: 183-205.

Zimmerman, BL. and KB. Homer. 1991. A network design criterion for estimating
selected attributes of the semivariogram. Environmetrics 2: 425-441.

Zobel, M. 1992. Plant species coexistence - the role of historical, evolutionary and
ecological factors. Oikos 65: 314-320.

"‘11111111111111‘5