u

 

\‘k‘t‘i‘%

--unu.-.

.nvuu an: vuu‘v n+

 

 

 

 

 

 

 
 
 
 

 

 

 

THESIS

/ I

LIIEBRAR

III IIII3IIIIIII

 

 

 

 

IIIIIIIIIIIIIIIIIIIIIIIIIIIII III

3 1293 01770

This is to certify that the

dissertation entitled

Spatial Environmental Variation: Within and
Among Population Effects in Danthonia spicata

 

presented by

Melissa Kay McCormick

has been accepted towards fulﬁllment
of the requirements for

Ph . D . degree in . Z_9_nggy_and
Ecology, Evolutionary Biology and Behavior

 

“(am 1 Am)

Major professor

Date 3/’<"/ 5i?
I I

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

 

 

LIBRARY

Michigan State
University

 

 

 

 

PLACE IN REFURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

moogigeggs

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1m mac/0.0mm“

SPATIAL ENVIRONMENTAL VARIATION: WITHIN AND AMONG POPULATION
EFFECTS IN DANTHONIA SPICA TA
By

Melissa Kay McCormick

A DISSERTATION
Submitted to

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

DOCTOR OF PHILOSOPHY

Department of Zoology
Program in Ecology, Evolutionary Biology, and Behavior
and
W. K. Kellogg Biological Station

1 999

ABSTRACT
SPATIAL ENVIRONMENTAL VARIATION: WITHIN AND AMONG POPULATION
EFFECTS IN DANIHONIA SPICA TA
By

Melissa Kay McCormick

Understanding the role of spatial environmental heterogeneity on the ecological
and evolutionary dynamics of plant populations is a major challenge for ecologists and
evolutionary biologists. Spatially structured variation in resources is commonly
measured, but its temporal persistence and relation to plant performance have rarely been
assessed. Spatial structure in resources can affect ecological and evolutionary processes
at all levels of organization, but the effect of resource variation on these processes will
depend on the spatial and temporal scales at which it varies.

I used a series of ﬁeld and greenhouse experiments to determine whether variation
in soil resources (soil moisture, ammonia, and nitrate) was spatially structured and
temporally consistent at scales relevant to the native, perennial grass Danthonia spicata.
I then related these patterns of variation in soil resources to ecological and evolutionary
differences within and among populations of D. spicata. Speciﬁcally, I tested whether D.
spicata plants responded to measured differences in the spatial structure of soil moisture
and nitrogen availability in different sites. I also related differences in dispersal distance
among ﬁve populations of D. spicata to the amount of spatially structured variation that
occurred in these sites. Interactions with other species may affect the responsiveness of a

species to resource heterogeneity. 1 extended the population-level analyses of spatially

structured variation in soil resources to the community level by testing whether the
interaction between D. spicata and the epiphytic fungus Atkinsonella hypoxylon was
affected by differences in soil moisture and nitrogen.

I found that my ﬁve study sites differed in amount and consistency of spatially
structured variation in soil moisture, ammonia, and nitrate. More importantly, I found
that survival of D. spicata plants was affected by differences in spatially structured
variation in soil resources among populations. This suggests that D. spicata plants
"perceive" differences in spatially structured variation in soil resources similar to those I
had measured.

In sites where soil resources are spatially structured, plants can take advantage of
spatial structure and target their seeds to good environments by decreasing their dispersal
distance. I found that D. spicata populations in sites with more spatially structured soil
resource variation had shorter dispersal distances than those in less spatially structured
sites. This ﬁnding was consistent with the hypothesis that spatially structured
environmental variation selects for decreased dispersal distance.

In three populations of D. spicata infected by A. hypoxylon, I found that infected
plants were distributed differently than uninfected plants with respect to soil moisture and
ammonia supply. In greenhouse studies, infected plants performed less well in dry, low
fertility conditions than uninifected individuals. These differences in the performance of
infected and uninfected plants in response to soil resources structures the interaction

between these two species and inﬂuences their distribution in the ﬁeld.

ACKNOWLEDGMENTS

I would like to thank my advisor, Kay Gross, for her years of encouragement and
advice. I would like to thank my committee, Tom Getty, Alan Tessier, and Jeff Conner,
and previous committee members, Steve Tonsor, Sue Kalisz, and Bryan Epperson for
their suggestions and guidance. Thank you also to the KBS community and particularly
to the lab group, Laura Broughton, Andrea Corbett, Brian Foster, Wendy Goodfriend,
Kevin Kosola, and Heather Reynolds for feedback on research presentations and papers.
I would especially like to thank Keith Clay, Wendy Goodﬁ'iend, Heather Reynolds, Sam
Scheiner, and Robin Smith for comments on earlier manuscripts from this dissertation
and Carol Baker, Jessica Call, Justin Pittinaro, Adrienne Purdy, Robin Smith, Carrie
Stolt, Lisa Tansey, and Karen Tindall for extensive ﬁeld, greenhouse, and lab help.
Special thanks also go to Robin Smith for providing the drawing ofDanthom’a spicata in
Figure 3.1, in addition to other drawings.

Financial support for this research was provided by a National Science
Foundation Dissertation Improvement Grant, the KBS Graduate Research Training
Grant, the Department of Zoology, the College of Natural Science, MSU‘s Ecology,
Evolutionary Biology, and Behavior Program, and the Kellogg Biological Station.

Thanks also go my parents, who instilled me with an appreciation of plants and
the natural environment and who encouraged me to always do my best, with the
assurance that they would love me no matter how things turned out. Finally, I would like
to thank my husband, Mark, whose love, patience, tolerance of long working hours, and

humor supported me throughout my graduate career.

iv

TABLE OF CONTENTS

LIST OF TABLES ......................................................................................................... vii
LIST OF FIGURES ....................................................................................................... viii
CHAPTER 1
INTRODUCTION ............................................................................................................ 1
Dissertation overview ........................................................................................... 5
Literature Cited ..................................................................................................... 8
CHAPTER 2
SPATIAL STRUCTURE IN SOIL RESOURCES: TEMPORAL CONSISTENCY AND
"PERCEPTION" BY DANTHONIA SPICA TA .............................................................. 1 1
Introduction ......................................................................................................... l 1
Materials and Methods ........................................................................................ 14
Study species ........................................................................................... 14
Quantifying spatial variation ................................................................... 17
Soil sampling and processing .................................................................. 19
Data analysis (spatial structure) .............................................................. 21
Plant perception of environmental variation-variable scale planting ..... 23
Comparing plant performance with environmental variables ................ 25
Plant perception of environmental variation-dispersal scale planting 26
Results ................................................................................................................ 29
Site differences in soil resources ............................................................ 29
Spatial structure of soil resources ........................................................... 29
Temporal consistency of spatial structure in soil resources ................... 38
Plant perception of environmental variation-variable scale planting ..... 45
Plant perception of environmental variation-dispersal scale planting 47
Discussion ........................................................................................................... 50
Structure of soil resources ...................................................................... 50
Plant "perception" of environmental structure ....................................... 52
Literature Cited .................................................................................................. 55
CHAPTER 3
THE ROLE OF SPATIAL ENVIRONMENTAL HETEROGENEITY IN THE
EVOLUTION OF GENE FLOW DISTANCE IN DANYHONIA SPICA TA ................ 57
Introduction ........................................................................................................ 57
Materials and Methods ....................................................................................... 59
Study species .......................................................................................... 59

Population and site characterization ....................................................... 60

Measuring dispersal distance .................................................................. 62
Dispersal simulation ............................................................................... 64
Combining dispersal and environmental variation ................................. 65
Calculations ............................................................................................ 68
Predictions ............................................................................................... 73
Interpretation testing ............................................................................... 74
Results ................................................................................................................. 75
Population and site characterization ........................................................ 75
Measuring dispersal distance ................................................................... 77
Dispersal simulation ................................................................................ 79
Combining dispersal and environmental variation .................................. 79
Interpretation testing ............................................................................... 82
Discussion ........................................................................................................... 85
Literature Cited ................................................................................................... 89

CHAPTER 4
PLASTICITY FOR NITROGEN AVAILABILITY IN DANTHONIA SP1 CA T A:
PHENOTYPIC PLASTICITY IN RESPONSE TO SPATIAL ENVIRONMENTAL

HETEROGENEITY? ...................................................................................................... 91
Introduction ......................................................................................................... 91
Materials and Methods ........................................................................................ 93

Study system ........................................................................................... 93
Greenhouse experiment .......................................................................... 95
Results ................................................................................................................. 99
Discussion ......................................................................................................... 104
Literature Cited ................................................................................................. 108

CHAPTER 5

DAN TH ONIA SP1 CA TA AND A TKINSONELLA H YPOXYL 0N : ENVIRONMENTAL

DEPENDENCE OF A SYMBIOSIS ............................................................................ 110
Introduction ....................................................................................................... l 10
Materials and Methods ...................................................................................... 112

Study species ......................................................................................... 112
Study sites ............................................................................................. 112
Field patterns ......................................................................................... 1 13
Statistical analysis ................................................................................. 1 15
Greenhouse experiment ........................................................................ 1 16
Statistical analysis ................................................................................. 1 18
Common garden experiment ................................................................. 118
Results ............................................................................................................... 120
Field patterns ......................................................................................... 120
Greenhouse experiment ........................................................................ 124
Common garden .................................................................................... 131

vi

Discussion ......................................................................................................... 133

Literature Cited ................................................................................................. 137
CHAPTER 6
CONCLUSIONS .......................................................................................................... 140
Literature Cited ................................................................................................. 144
Appendix ............................................................................................................ 145

vii

LIST OF TABLES

Table 4.1: ANOVA tables presenting variation in biomass, rootzshoot and shoot
nitrogen ....................................................................................................... 100

Table 5.1: Survival of infected and uninfected plants in the greenhouse treatments .. 129

Table A. 1: Composition of modiﬁed Hoagland's solution used in this experiment 145

viii

LIST OF FIGURES

Figure 1.1: Danthonia spicata ........................................................................................ 4
Figure 2.1: Population characteristics ........................................................................... 16
Figure 2.2: Soil sampling design .................................................................................. 18
Figure 2.3: Four site planting design ............................................................................ 27
Figure 2.4: Moisture and nitrogen levels across years at four study sites .................... 31
Figure 2.5: Semivariance and covariance of soil moisture over two years .................. 33
Figure 2.6: Semivariance in relative nitrogen supply ................................................... 35
Figure 2.7: Comparison of spatial structure in initial availability and resin bag

nitrogen measures ....................................................................................... 37
Figure 2.8: Covariance of ammonia and nitrate among seasons .................................. 40
Figure 2.9: Covariance of initial nitrogen pools and spring 1998 nitrogen supply ...... 42
Figure 2.10: Spatial structure in nitrogen across years in two sites ............................. 44
Figure 2.11: Spatial structure of culm survival in two ﬁeld sites ................................. 48
Figure 3.1: Danthonia spicata chasmogamous and cleistogarnous seeds .................... 61
Figure 3.2: Characteristics of study populations .......................................................... 67
Figure 3.3: Relating seed dispersal to environmental variation ................................... 71
Figure 3.4: Simulated seed dispersal distances ............................................................ 76
Figure 3.5: Dispersal of CH and CL seeds ................................................................... 78
Figure 3.6: Relationship between CL dispersal distance and percent of CL seeds ...... 80
Figure 3.7: Effect of CL seed allocation on variation encountered by seeds ............... 81
Figure 3.8: Relationship between population %CL and environmental structure ........ 83
Figure 3.9: Comparison to Schoen and Lloyd's rates of encounter with good sites ..... 84
Figure 4.1: Characteristics of ﬁve study sites and populations .................................... 94

ix

Figure 4.2:
Figure 4.3:
Figure 4.4:
Figure 5.1:
Figure 5.2:
Figure 5.3:
Figure 5.4:
Figure 5.5:

Figure 5.6:

Percent cleistogarny ofDanthonia spicata families ................................... 96

Reaction norms across two nitrogen levels .............................................. 102
Reaction norms of response to nitrogen treatment across three levels ..... 103
Characteristics of three study populations ................................................ 121
Field distribution of infected and uninfected Danthonia spicata plants .. 123
Growth ofDanthom'a spicata in greenhouse treatments .......................... 126
Performance of Danthonia spicata in greenhouse treatments .................. 128
Final size distributions ofDanthonia spicata plants ................................ 130
Biognass and nitrogen concentration of plants grown in the common ...... 132
gar en

Chapter 1

INTRODUCTION

Environmental variation occurs at a variety of spatial and temporal scales in
ecological systems. The challenge to ecologists is to determine the importance of
environmental heterogeneity at different scales, and the linkages between scales, to
develop a more complete understanding of how environmental variation can affect the
distribution, abundance, and ﬁtness of an organism. Spatial variation in soils is
ubiquitous (e.g. Grime 1979, Tilrnan 1988, Schlesinger 1996, Gross et al. 1995), but the
role of soil resource variation in determining plant-“perceived” resource heterogeneity is
largely unknown (Robertson et a1 1988, Ehrenfeld 1997). The effect of environmental
variation on plants depends on the magnitude, scale, and temporal consistency of that
variation (Miller et al. 1995, StrattOn and Bennington 1998). It also depends on the
proportion of variation that is spatially structured (the degree to which two points that are
close together are more similar than two points chosen at random). These aspects of
environmental variation interact with plant size, longevity, and dispersal characteristics to
determine the potential evolutionary response of a population to its environment
(Antonovics et al. 1987).

If environmental conditions are predictable (structured in time and space) at the
scale of dispersal and individual longevity, then selection can result in locally adapted
subgroups (Hoffman and Parsons 1991). Locally adaptive responses among populations
to differences in mean environmental conditions have been well documented (e. g.

Bennington and McGraw 1995, Schoen et al. 1986, Antonovics et al. 1987, Platenkarnp

1990), but differences among populations in response to differing magnitudes of
environmental variation have only rarely been examined (Sultan 1987, Stark 1994, but
see Miller and Fowler 1994). The scale at which local adaptation to environmental
conditions occurs depends on the strength and predictability of selection, the available
genetic variation, and the interaction between gene dispersal distance and the scale of
environmental structure.

Although dispersal distance inﬂuences the evolutionary impact of environmental
variation at different scales, dispersal itself can also respond adaptively to differences in
the scale of spatially structured environmental variation (Harper 1977, Venable and
Brown 1988). The spatial scale of dispersal also affects the amount of genetic diversity
that can be maintained within populations by determining the scale at which local
adaptation can occur. Maintenance of genetic diversity by spatial variation has been
theoretically demonstrated to be rare because dispersal distance and the scale of
environmental variation must be precisely matched (e.g. Spieth 1979, Gillespie 1981).
However, if populations can adjust dispersal distance to the scale at which environmental
variation is structured, then conditions may be less restrictive (Hedrick 1986).

When environmental variation occurs at scales smaller than dispersal distance or
is temporally inconsistent, populations cannot respond by local adaptation. Instead,
populations must be phenotypically plastic in their response to enviromnental variation.
Plastic responses to individual-scale environmental variation have been frequently
demonstrated (e.g. Jackson and Caldwell 1993, Fitter 1994, Gross et a1. 1993). However,
the amount of phenotypic plasticity displayed by organisms in a population may be

locally adapted to the amount of environmental variation that occurs at small scales.

Although local adaptation in phenotypic plasticity has occasionally been related to the
temporal consistency of spatially structured environmental variation (Sultan and Bazzaz
1993), it has rarely been related to either the amount or scale of spatially structured
environmental variation (but see Lotz et a1. 1990, Miller and Fowler 1994).

The response of plants to environmental variation can also be modiﬁed by
interactions with other species (Chesson 1985, Reader and Best 1989). For example, the
growth of plants infected by mycorrhizae is often limited by light, while the growth of
uninfected plants in the same community may be limited by soil nutrients (Harley 1969,
Harley and Smith 1983). This suggests that plants with mycorrhizae "perceive" high
nutrient, low light habitats as unsuitable, while plants without mycorrhizae "perceive"
low nutrient, high light habitats as similarly unsuitable. This sort of environmentally-
mediated tradeoff response of a species interaction may produce relatively discrete spatial
patches of plants with and without mycorrhizae in a heterogeneous environment.
Environmentally-dependent species interactions, such as this example, are most likely to
be apparent between intimately associated species.

In this dissertation, I used a combination of ﬁeld and greenhouse studies to study
the response of Danthonia spicata to spatial environmental heterogeneity in soil
resources. Danthonia spicata is a native, perennial, C3 bunchgrass that grows in dry, low
fertility old-ﬁelds and oak savannas throughout the northern and eastern United States
and southern Canada. Danthonia spicata produces dimorphic seeds with different
dispersal abilities (Figure 1.1). Allocation to the two seed types has a strong genetic
basis and is extremely variable within and among D. spicata populations (Clay 1982),

making it a good indicator of population differences in dispersal. I examined differences

 

Figure 1.1: Danthonia spicata
Diagram of Danthonia spicata showing the location of cleistogamous (CL) and
chasmogamous (CH) seeds.

among populations in dispersal distance, an explicitly spatial trait that could be related to
differences in the spatial scale of resource heterogeneity among sites, and in three traits
associated with nitrogen acquisition that were likely to be strongly inﬂuenced by
temporal variation (i.e. be phenotypically plastic). I predicted that D. spicata plants from
environments with more spatially structured variation in soil resources would have
shorter dispersal distances and more phenotypic plasticity than plants from less variable
environments. I then also examined the effect of spatial heterogeneity in soil resources
on the interaction between D. spicata and the epiphytic fungus Atkinsonella hypoxylon.
Because A. hypoxylon has been shown to affect the growth and competitive ability of D.
spicata (Clay 1984, Kelley and Clay 1987), I predicted that infection by A. hypoxylon

would affect plant "perception" of spatially variable soil resources.

DISSERTATION OVERVIEW

Chapter 2 is an analysis of the spatial structure and temporal consistency of
variation in soil resources among ﬁve populations of Danthonia spicata. I measured
heterogeneity in soil moisture and nitrogen availability; for nitrogen I measured inital
pool as well as the supply rate of nitrate— and ammonia-nitrogen over 1-2 years. I then
related these measures of soil resource heterogeneity to performance of D. spicata in a
phytometer study. I used both direct measurements of soil moisture and nitrogen
availability and plant performance in these sites to examine differences in the magnitude
and scale of spatial environmental variation within populations. I predicted that soil

resource heterogeneity would vary to differing degrees among sites and that plant

performance (survival, growth and tillering of D. spicata) would reﬂect these measured
differences in soil conditions.

In chapter 3 I examined the relationship between dispersal distance and the scale
of soil resource variation for ﬁve D. spicata populations. I used dispersal measurements
and computer simulations to estimate the average population dispersal pattern and the
amount of environmental variation (relative to parental conditions) that would be
encountered within each population by dispersing offspring. I then related among-
population differences in dispersal pattern to among-population differences in the scale of
spatial heterogeneity in soil resources. I predicted that populations in sites with less
spatially structured variation in soil resources would have primarily chasmogamous (far-
dispersing) seeds and populations in sites with more spatially structured variation would
have a large proportion of cleistogamous (near-dispersing) seeds.

The results of greenhouse experiments examining the relationship between
phenotypic plasticity and spatial variation in soil nitrogen are presented in chapter 4. I
examined differences in phenotypic plasticity of three traits related to nitrogen
acquisition among ﬁve populations. I also compared the amount of phenotypic plasticity
for nitrogen acquisition among differently-dispersing families within each population. I
predicted that plants from populations with more spatial variability in nitrogen supply
rate would have greater phenotypic plasticity than those from less variable poplations and
that individuals from far-dispersing (low CL) families would be more plastic than those
from near-dispersing (high CL) families.

Because interactions with other species can affect the responses of an organism to

its environment, I also examined the impact of the interaction between D. spicata and the

epiphytic fungus Atkinsonella hypoxylon on plant perception of spatially heterogeneous
soil resources. These results are summarized in Chapter 5. I used ﬁeld experiments to
examine the spatial distribution of A. hypoxylon in ﬁeld populations of D. spicata in
relation to soil moisture and nitrogen supply. I then used greenhouse experiments to
explicitly address the effects of soil moisture and fertility on growth of plants infected
and uninfected by A. hypoxylon. I predicted that plants infected by A. hypoxylon would
be affected differently by soil moisture and fertility than uninfected plants.

I discuss the implications of these results for studies of evolution in spatially and
temporally heterogeneous environments in a concluding chapter (6). I speciﬁcally relate
my results to studies of measured environmental heterogeneity and to theoretical
expectations for the joint evolution of dispersal (local adaptation) and phenotypic

plasticity in response to spatially heterogeneous environments.

LITERATURE CITED

Antonovics, J ., K. Clay, and J. Schmitt. 1987. The measurement of small-scale
environmental heterogeneity using clonal transplants of Anthoxanthum odoratum
and Danthonia spicata. Oecologia 71:601-607.

Bennington, C. C. and J. B. McGraw. 1995. Natural selection and ecotypic differentiation
in Impatiens pallida. Ecological Monographs 65:303-323.

Chesson, P. 1985. Coexistence of competitors in spatially and temporally varying
environments: A look at the combined effects of different sorts of variability.
Theoretical Population Biology 28:263-287.

Clay, K. 1984. The effect of the fungus Atkinsonella hypoxylon (Clavicipitaceae) on the
reproductive system and demography of the grass Danthonia spicata. New
Phytologist, 98: 165-175.

Clay, K. 1983. The differential establishment of seedlings from chasmogamous and
cleistogamous ﬂowers in natural populations of the grass Danthonia spicata.
New Phytologist, 98:165-175.

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 85:185-798.

Fitter, A. H. 1994. Architecture and biomass allocation as components of the plastic
response of root systems to soil heterogeneity. Pages 305-324, in M. M. Caldwell
and R. W. Pearcy (eds.), Exploitation of Environmental Heterogeneity By Plants.
Academic Press, Inc., San Diego, California.

Gillespie, J. H. 1981. The role of migration in the genetic structure of populations in
temporally and spatially varying environments. III. Migration modiﬁcation. The
American Naturalist 117:223-233.

Grime, J. P. 1979. Plant Strategies and Vegetation Processes. John Wiley and Sons,
Ltd, New York.

Gross, K., A. Peter, and KS. Pregitzer. 1993. Fine root growth and demographic

responses to nutrient patches in four old-ﬁeld successional plant species.
Oecologia 95:61-64.

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

 

Harley, J. L. (1969). Fungal symbiosis. Transcripts of the British Mycological Society 51,
1-1 1.

Harley, J. L., and S. E. Smith. (1983). Mycorrhizal Symbiosis. Academic Press,
London, England.

Harper, J. L. 1977. Population Biology of Plants. Academic Press, New York, NY.

Hedrick, P. W. 1986. Genetic polymorphism in heterogeneous environments: A decade
later. Annual Review of Ecology and Systematics 17:535-66.

Hoffman A. A. and P. A. Parsons. 1991. Evolutionary Genetics and Environmental
Stress. Oxford University Press, New York.

Jackson, R. B. and M. M. Caldwell. 1993. The scale of nutrient heterogeneity around
individual plants and its quantiﬁcation with geostatistics. Ecology 74:612-614.

Kelley, S. E., and K. Clay. 1987. Interspeciﬁc competitive interactions and the
maintenance of genotypic variation within the populations of two perennial
grasses. Evolution, 41 :92-103.

Lotz, L. A. P., H. 0le and P. H. van Tienderen. 1990. Within-population variability in
morphology and life history of Plantago major L. ssp. pleiosperma Pilger in
relation to environmental heterogeneity. Oecologia 84:404-410.

Miller, R. E., and N. L. Fowler. 1994. Life-history variation and local adaptation within
2 populations of Bouteloua—rigidiseta (Texas-grarnma). Journal of Ecology
82:855-864.

Miller, R. E., J. M. VerHoef, and N. L. Fowler. 1995. Spatial heterogeneity in eight
central Texas grasslands. Journal of Ecology 83:919-928.

Platenkarnp, G. 1990. Phenotypic plasticity and genetic differentiation in the demography
of the grass Anthoxanthum odoratum. Journal of Ecology 78:772-788.

Reader, R. J. and B. J. Best. 1989. Variation in competition along an environmental
gradient: Hieracium ﬂoribundum in an abandoned pasture. Journal of Ecology
77:673-684.

Robertson, G. P., M. Huston, and F. Evans. 1988. Spatial variability in a successional
plant community: Patterns of nitrogen availability. Ecology 69:1517-1524.

Schlesinger. 1996. On the spatial pattern of soil nutrients in desert ecosystems. Ecology
77:364-374.

Schrnitt, J. and S. E. Gamble. 1990. The effect of distance from the parental site on
offspring performance and inbreeding depression in Impatiens capensis-A test of
the local adaptation hypothesis. Evolution 44: 2022-2030.

Schoen, D. J., S. C. Stewart, M. J. Lechowicz, and G. Bell. 1986. Partitioning the
transplant effect in reciprocal transplant experiments with Impatiens capensis and
Impatiens pallida. Oecologia 70: 149-154.

Spieth, P. T. 1979. Environmental heterogeneity: a problem of contradictory selection
pressures, gene ﬂow, and local polymorphism. The American Naturalist, 113:
247-260.

Stark, J. M. 1994. Causes of Soil nutrient heterogeneity at different scales. Pages 255-
284, in M. M. Caldwell and R. W. Pearcy (eds), Exploitation of Environmental
Heterogeneity By Plants. Academic Press, Inc., San Diego, California.

Stratton, D. A., and C. C. Bennington. 1998. F inc-grained spatial and temporal variation
in selection does not maintain genetic variation in Erigeron anuus. Evolution, 52:
678-691.

Sultan, S. 1987. Evolutionary implications of phenotypic plasticity in plants. Pages 127-
178 in Hecht, M. et al. eds. Evolutionary Biology, volume 21. Plenum Press,
New York.

Sultan, S., and F. A. Bazzaz. 1993. Phenotypic plasticity in Polygonum persicaria II.
Norms of reaction to soil moisture and the maintenance of genetic diversity.
Evolution 47: 1032- 1 049.

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

Venable, D. L., and J. S. Brown. 1988. The selective interactions of dispersal,

dormancy, and seed size as adaptations for reducing risks in variable
environments. The American Naturalist 13 1 :360-3 84.

10

Chapter 2

SPATIAL STRUCTURE IN SOIL RESOURCES: TEMPORAL CONSISTENCY
AND "PERCEPTION " BY DAN THONIA SPICA TA
(Manuscript co-authored with K.L. Gross)
INTRODUCTION

Spatial structure in soil resources can have an important role in generating
patterns of genetic variation (Venable and Brown 1988, Bell and Lechowicz 1991,
Stratton 1995) and structuring interactions in plant populations and communities. Spatial
structure in soil resources has been measured in a number of systems (e.g. Grime 1979,
Tilrnan 1988, Schlesinger 1990, Gross et al. 1995), but whether this measured spatial
structure is relevant to plants is largely unknown (Robertson et a1 1988, Ehrenfeld 1997).
To be relevant to plants, resource heterogeneity must occur at appropriate spatial and
temporal scales, determined by plant size, dispersal distance, and longevity. Temporal
inconsistency may cause measured spatial heterogeneity to appear irrelevant to plants
over the time scales at which ecological and evolutionary responses are produced.

In a spatially structured environment, plants and their dispersing offspring
encounter environmental variation over a range of scales. Individual plants can average
over temporal variation within their lifetimes and families can average over variation at
longer time scales, but the ecological and evolutionary impact of environmental variation
on plant populations may depend on how individuals and families experience that
variation. Plant species that differ in growth phenology may be disproportionately
affected by environmental variation that occurs during different times of the year or

during different seasons. For example, a plant that grows, reproduces and goes dormant

11

in the spring may be little affected by environmental variation during the fall, while a
plant that grows throughout the year may be affected by differently structured variation
during different seasons. Patterns of environmental variation that were consistent among
seasons or during a single season among years would affect these two plant species
differently. Consequently, the temporal scale at which spatial structure would need to be
consistent to produce ecological and evolutionary effects would differ among species

with different phenologies. Temporal persistence of measured soil spatial structure has

been assumed at many temporal scales, but rarely measured (but see Goovaerts and
Chiang 1993). Several recent studies of evolution in heterogeneous environments have
questioned the persistence of measured spatial environmental structure within
populations (e. g. Sultan and Bazzaz 1993, Stratton and Bennington 1998).

Spatial structure in soil resources occurs over a wide range of scales from
millimeters to hundreds of kilometers (e.g. Robertson and Gross 1994, Gross et al. 1995,
Miller et al. 1995, Ehrenfeld et al. 1997). Individual size, offspring dispersal distance,
and population size deﬁne scales at which spatial structure is likely to generate different
evolutionary responses in plant populations (Miller et a1. 1995). An individual plant may
only experience a few cm2 of the soil surface, but the spatial scale at which a plant
experiences environmental structure can change as it grows and/or with life stage. A
plant also "perceives" its environment through differential success of its offspring, which
may be distributed over several m2 or more (Bell and Lechowicz, 1994). The amount of
spatial variation in soil resources that is encountered by individuals or dispersing
offspring is determined by the spatial structure of environmental variation. The

proportion of environmental variation that is spatially structured and the range over

12

which it is structured determine the extent to which environmental variation will increase
with increasing distance between sample points (or between parent and offspring) and the
distance over which it will increase. Spatial structure that occurs at scales smaller than
dispersal distance must be tolerated or exploited through phenotypic plasticity, while
spatial heterogeneity at larger scales may result in local adaptation. For measured spatial
structure to affect the structure of plant populations and communities, it must be of a
magnitude that affects plant performance, occur at spatial scales greater than dispersal
distance, and persist over temporal scales longer than an individual lifetime.

Demonstrating environmental structure at spatial and temporal scales that are
relevant to plants is an important step in relating measured variation to "perceived"
variation. However, the deﬁnitive test of the relevance of ﬁeld-measured soil
heterogeneity to a particular plant species is to grow that species in intact ﬁeld
populations where variation in soil variables has been measured (Bell and Lechowicz
1991, Miller and Fowler 1997) and to relate plant performance to environmental
measurements.

In this study, we combined data on single-time measures of soil moisture and
extractable nitrogen, with longer-term measures of nitrogen supply rate and growth of
phytometers to examine the spatial structure of soil moisture and nitrogen availability, the
temporal persistence of these patterns, and their relevance to the growth and survival of
Danthonia spicata in ﬁeld populations. Nitrogen availability limits plant performance in
many terrestrial plant communities (Chapin 1980) and variation in nitrogen availability
has frequently been demonstrated on spatial scales ranging from centimeters to meters

and much larger (Gross et al. 1995, Ehrenfeld et al. 1997). Although the impact of

13

uniformly increasing nitrogen availability on plant performance has often been
demonstrated, the relevance of measured spatially structured variation in nitrogen to
plants in the ﬁeld is largely unknown (Miller and Fowler 1994). Competition for spatial
heterogeneous soil resources may be particularly important for plants in low productivity
environments (Tihnan 1988). If so, then we might expect that plant populations in low
productivity environments would be responsive (on both evolutionary and ecological
time scales) to resource heterogeneity. We studied ﬁve ﬁeld sites with populations of D.
spicata and examined differences among the sites in the proportion, range and temporal
consistency of spatial structure in soil moisture and nitrogen. We then related soil
moisture and nitrogen measurements to growth and survival of D. spicata planted into the

intact communities of each site where we had measured soil moisture and nitrogen.

MATERIALS AND METHODS
Study species

Danthonia spicata (Poaceae) is a native, perennial, C3 bunch grass that commonly
occurs in dry, nutrient-poor oak savannas, old-ﬁelds, and forest clearings throughout the
eastern and northern United States and Canada (Darbyshire and Cayouette 1989).
Danthonia spicata reproduces vegetatively by tillering, with the tillers forming a discrete
clump. These semi-autonomous tillers can be separated, producing independent clonal
replicates that we refer to throughout this paper as cuhns. Danthonia spicata produces
two ﬂower types. Potentially outcrossed, wind-pollinated chasmogamous (CH) ﬂowers
are produced at the tip of the reproductive stalk and disperse an average of 60 cm

(Chapter 3), with possible additional secondary dispersal. Obligately self fertilized

14

cleistogamous (CL) ﬂowers are produced in the axils of the leaf sheaths along and at the
base of the reproductive stalk and disperse an average of 10 cm or less (Chapter 3).

There is substantial variation within and among D. spicata populations in percent
allocation to shorter dispersing, CL, seeds (Clay 1982). Allocation to CL seeds affects
mean dispersal distance, and thus the spatial scale over which offspring encounter the
environment (Chapter 3). We measured spatial variation in soil resources and its
consistency over time in ﬁve ﬁeld sites across the state of Michigan that had populations
of D. spicata (selected from a larger sample of eight sites; Chapter 3) that differed in
average allocation to CL seeds. Plant populations in Sites A2 and L1 produced a large
percent of CH seeds and so were far-dispersing populations. Populations in Sites D and
L2 had moderate proportions of CL seeds. Population R had a large proportion of CL
seeds and was a short-dispersing population (Figure 2.1; Chapter 3).

The ﬁve sites in which we examined the relationship between environmental
heterogeneity and D. spicata performance all had communities that were dominated by
perennial, herbaceous vegetation, especially Andropogon virginicus, Schizachirium
scoparium, and Carex spp.. None of the sites had been plowed or tilled for at least 45
years. Site A2 was located in a remnant oak savanna in the Allegan State Game Area
near Allegan, Michigan. Sites L1 and L2 were located in old-ﬁelds that had been
abandoned from agriculture at the WK. Kellogg Biological Station. Site R, also an old-
ﬁeld site, was located in the Rose Lake Wildlife Research Area near East Lansing,
Michigan. Site D was located in an alvar grassland in the Maxton Plains Preserve on

Drummond Island, Michigan.

15

 

% CL seeds

 

Density (plants/m2)

 

 

 

Population or Site

Figure 2.1: Population characteristics

Percent of cleistomous (CL) seeds : 1 s.e. (a.) and plant density (b.) in selected
study populations of Danthonia spicata. Population A1 was only used as a source
of phytometers for the two population planting experiment.

 

Quantr'ﬁzing spatial variation:

We used a stratiﬁed, nested sampling design to measure soil moisture, current
nitrate and ammonia availability (standing pool sizes) and relative supply of nitrate- and
ammonia-nitrogen in a 4 x 6 m permanent plot in four of the ﬁve study sites (A2, L1, L2,
R). Using a stratiﬁed, nested sampling design allowed us to detect spatially structured
environmental variation over a wide range of scales (Corbett 1998). Additional points
were nested at 5 and 10 cm to provide additional power to detect spatial variation within
D. spicata's dispersal distance.

To establish this sampling design, we randomly chose 43 points to sample within
each plot. We then randomly chose 13 of these 43 points to potentially receive one or
more of 20 randomly distributed additional sample points 5 cm away. When an initial
point was chosen more than once, we located the second additional point 5 cm from the
ﬁrst additional point (10 cm from the initial point). This design produced 20 pairs of
points 5 cm apart and eight pairs 10 cm apart. To generate a total of 15 pairs of points
separated by 10 cm, we randomly selected seven of the 13 initial points to receive an
additional point 10 cm away. Again, when a point had already received additional points,
this additional point was placed 10 cm from the most recently added point, so that all
additional points were in a line (see Figure 2.2). This produced a total of 70 sample

points within each permanent plot across a range of scales.

l7

 

 

 

 

 

600
= f m
A
i A
500- 3 ‘A A
I
A
3 A A ‘
A A A A
400- 4 ‘
A 4 ‘
E m A AA
8 A
4(1) 300' A
iii ' A ‘ A
>' A
_ m
2004 ‘ A
A A :43
A
100-
« A A
0 ' T T I I ‘ T l I ' ' I ' ' '
O 100 200 300 400
X-axis (cm)

Figure 2.2: Soil sampling design
Diagram of sample point locations in the permanent plot in each population. Of 43

base points, 13 (A) were chosen to receive one or more additional points at 5 or 10
cm away. Total N = 70 points.

18

. i
I

Soil sampling and processing

To estimate environmental conditions at a single point in time, we sampled soil
moisture and current availability of nitrate- and ammonia-nitrogen in four of these ﬁve
sites (A2, L1, L2, R). In Sites L1 and L2, we removed a soil core 2.5 cm in diameter and
10 cm deep from each sample point in June 1996. We immediately placed the soil cores
in plastic bags in an ice-ﬁlled cooler, where they remained until further processing (~ 12
hours). Each soil core was then passed through a 2 mm sieve to remove large rocks and
root masses.

We weighed a 10 g subsample from each sieved soil sample and dried it for 72

hours at 65°C. Dried samples were then reweighed to determine the percent of soil
moisture in g H2O g dry soil". We extracted available nitrogen from an additional 10.00
g subsample of soil with 50.0 ml of 1M KC]. Extracted samples were shaken for 1 min.

and allowed to incubate at room temperature for 24 hours before ﬁltering through a 1 pm

Gelrnan glass ﬁber ﬁlter. Filtered extracts were stored at ~ 3°C for 5-8 days before
analysis for nitrogen concentration. We analyzed extracts for nitrate- and ammonia-
nitrogen using segmented ﬂow colorimetry (Alpkem). Nitrogen concentration of each
sample was then corrected for initial soil moisture and expressed as g N g dry soil".

Extractable nitrogen in soil cores gave us an estimate of nitrogen availability at a
single point in time. We used ion exchange resin bags placed at the same sampling
points to estimate the relative supply rate of nitrate and ammonia (a longer-term, or more
integrated measure of nitrogen availability). After removal of each soil core, we inserted
a mesh bag (PeCap 120 mm polyester mesh, Tetko, Inc., Briarcliff Manor, NY) ﬁlled

with 1-2 tablespoons of mixed bed ion exchange resin (Dowex MR-3) so that each resin

19

bag was positioned 5 cm beneath the soil surface. These resin bags remained undisturbed
in the soil for 12 months (until June 1997), when they were removed and replaced with

new resin bags. We rinsed the removed resin bags with double distilled water to remove

soil particles and dried them at 40°C. We weighed out a 2.00 g subsample of dry resin
from each bag and extracted nitrogen bound to the resins with 30 ml 1M KC]. Resin bag
extracts were ﬁltered and analyzed for nitrate and ammonia using the same protocol as
for soil extracts.

We sampled soil moisture and extractable nitrate- and ammonia-nitrogen in Sites
A2 and R in June 1997 and placed resin bags in each sample location, as we had done for
Sites L1 and L2. In October 1997, we removed the resin bags that were placed in the ﬁeld
in June 1997 (from all four sites) and replaced them with new resin bags. The shorter
sample period used for resin bags in 1997 divided the year into two seasons (summer
1997 and spring 1998) corresponding to D. spicata's two growing periods and allowed an
examination of temporal nitrogen dynamics among seasons. In May 1998 we removed
the resin bags that we placed in the ﬁeld in October 1997 (spring 1998). At the same
time, we resampled soil moisture in all four sites by removing a second soil core
immediately adjacent to each initial sampling location.

In the ﬁfth site (D), the soils were too shallow to allow coring or resin bag
installation (mean depth 4 cm), so we used soil depth as an indication of soil moisture.
We expected soil depth would have a strong effect on microhabitat suitability in this site
because it largely determined the availability of water, with deeper locations retaining
pools of water long after shallow locations had become extremely dry (McCormick,

personal observation). We used the same sampling design that was used for soil cores in

20

the other four sites to sample soil depth in this site in June 1997. We did not resarnple
soil depth, because it was largely the result of topography of the underlying bedrock,

which did not change appreciably among years (McCormick, personal observation).

Data analysis (spatial structure)

We used data from our 70 sample points to calculate mean soil moisture, nitrogen
pools, and nitrogen supply in each of the four sites. To examine site differences in mean
environmental conditions, we tested for differences in each factor among the four sites
using an AN OVA.

We used semivariance analysis (GS+, v. 3.11.6, GammaDesign, Inc., Plainwell,
MI) to calculate semivariograms, which describe the similarity between pairs of points as
a function of how far apart they are, for each environmental variable (Robertson and
Gross 1994). We examined each soil variable over the range 0 to 350 cm (active lag)
using a 20 cm interval (step size). We then used a linear regression of semivariance on
distance (Systat 8.0 for windows, 1997. SPSS, Chicago IL.) to determine whether
semivariance increased with increasing distance between sample points at distances
within and beyond D. spicata’s dispersal distance. These analyses provided an estimate
of whether spatial structure was present over the range 0 to 350 cm, but could not detect
spatial structure present only at very small scales (e. g. 10 to 15 cm).

We examined the temporal consistency of spatial structure in soil resources in
each site by calculating a semivariograrn for the sum of conditions at the two dates to be
compared and testing for the presence of spatial structure in the covariogram calculated

using the equation Variance(A + B) = Variance(A) + Variance(B) — 2*Covariance(A+B)

21

and Semivariance = Variance / 2. The covariogram, composed of the covariance
calculated for each distance class, describes spatial structure that is consistent across the
sample periods compared. We used the signiﬁcance of the linear regression of
covariance on separation distance to test for signiﬁcant temporal consistency of spatial
structure in each soil variable. If there is consistent spatial structure, we expect that the
covariance will increase with distance. Because the covariance will not increase with
distance if there is no spatial structure in the two samples being compared, the
covariogram can only reﬂect presence of consistent structure, not a consistent lack of
structure.

We examined the temporal consistency of spatial structure in soil nitrate and
ammonia over two time periods. First we tested for signiﬁcant spatial structure among
seasons within a year. This compared the June to October 1997 (summer 1997) resin bag
measurements to the October 1997 to May 1998 (spring 1998) measurements. We could
not calculate summer 1997 spatial structure for ammonia or nitrate supply in Site A2
because foraging animals unearthed 50 of the 70 resin bags we installed in June 1997.
We also examined the relationship between spatial structure of standing pool nitrogen
availability (from soil cores) and longer-term measures of nitrogen supply (from resin
bags). This comparison may indicate consistency of nitrogen availability in a season
among years. Because soil cores were sampled in the spring (1996 or 1997), we
compared the spatial structure of standing pool nitrogen to the spatial structure of spring
1998 resin bag measurements. For populations L1 and L2 we also compared 1996 full

year measurements to 1997 measurements (summer 1997 + spring 1998). For soil

22

moisture, we examined the spatial structure in the covariogram comparing the two

sample times (1996 or 1997 with 1998 measurements) as described above.

Plant perception of environmental variation-variable scale planting

Plant performance is the best indication of whether measured spatial
environmental variation is "perceived" by plants in a population. To determine whether
D. spicata plants could perceive differences in the measured scale of spatial variation in
two sites (L1, L2), we conducted a phytometer experiment, using D. spicata cuhns as our
measurement tools. We used individual D. spicata plants collected from three
populations as phytometers: A1 (5 individuals), L2 (5) and L1 (6) in April 1995.
Population Al was an oak savanna site that was located approximately 5 km away from
Site A2. Site A; was very similar to Site A2 in soil moisture, nitrate and ammonia
availability (McCormick, unpublished data). The D. spicata population in sites A] and
A2 had a similar average proportion of CL seeds (Figure 2.1).

We divided each of the 16 parent plants into 26 to 30 culms, each of which was
individually potted into 10 cm square pots ﬁlled with silica sand and grown at 24 to 32°
under ambient light in the greenhouse. Plants were grown without fertilizer for 13
months, and then planted in the ﬁeld (May 1996). Between 22 and 26 culms from each
parent were randomly assigned to one of two sites (L1 or L2). We planted the
phytometers (cuhns) into randomly selected points in one half of each permanent plot in
which we measured the spatial variation of soil moisture and nitrogen, with the
stipulation that each source population had to be represented in each plot half. Planting

culms into only two sites allowed us to have greater power to detect differences in plant

23

performance across a range of scales within each site. Using plants from three source
populations allowed us to determine whether plants from different populations perceived
measured spatial variation differently. The cuhns were grown in the ﬁeld for 25 months
(May 1996 - June 1998). We then harvested all above-ground biomass from each

surviving cuhn, noting the number of tillers each had produced and whether they had

produced seeds. Harvested biomass was dried at 45°C for 48 hours and then weighed (i
0.01 g) to determine above-ground dry weight.

We established a separate common garden in a tilled 2 x 2 m (tilled 20 cm deep)
plot adjacent to the L1 site in May 1996 to distinguish differences in cuhn growth
resulting from genetic differences among source populations from differences in culm
growth resulting from differences in “perceive ” spatial variation in the intact ﬁeld sites
(L1 and L2). We planted four culrns from each parent into randomly selected locations on
a 10 x 10 cm grid established in the central 1 x 1 m of the tilled area. We planted
additional, randomly selected, cuhns around the perimeter of the common garden to
reduce edge effects on the study plants. We measured the initial size of planted culms by
counting the number of living leaves at the time of planting.

The common garden was regularly weeded and culrns monitored until June 1997
(13 months), when we harvested above-ground biomass from all surviving culms. We
determined ﬁnal plant size by measuring above-ground dry weight and by counting the
number of tillers each culm produced. We used an ANOVA to compare the growth (dry
weight and number of tillers) of culms from different source populations across both
intact sites (L1 and L2) and in the tilled common garden. Both dry weight and number of

tillers were log-transformed to improve normality. We used a second AN OVA to

24

 

compare growth of plants from all source populations in the common garden and the two
recipient sites (L1, L2).

To determine whether variation in culm ﬁnal weight was spatially structured, we
conducted a semivariance analysis of plant weight as a function of planting location in
each of the two intact recipient sites (L1, L2). However, this analysis only allowed us to
examine spatial structure in growth variation among surviving culrns and did not take
into account spatial structure in cuhn mortality. We used j oincount analysis (Joincounts,
Epperson 1992), a measure of spatial structure in discrete variables, to determine
whether, and at what scales, cuhn survival was spatially structured in these two sites. In
joincounts, a signiﬁcant, negative, standard normal deviate statistic (SND < -1.96)

indicates that survival was signiﬁcantly spatially structured.

Comparing plant performance with environmental variables

To compare plant growth and survival in these two sites with our measures of
environmental variables, we used kriging analysis (block kriging, GS+, v. 3.11.6,
GammaDesign, Inc., Plainwell, MI) to estimate values of sol moisture and ammonia for
each culm planting location. Kriging analysis uses the eight sample points nearest the
point for which an estimate is desired and weights them according to their distance from
the estimate location and the spatial structure of that variable in that site (Robertson and
Gross 1994). In a site with substantial spatial structure, sample points near the estimate
location would be weighted much more than points farther away, while in a site with little
spatial structure there would be little difference in weighting, regardless of distance from

the estimate location. We examined the signiﬁcance of regressions of growth of the

25

 

planted culms on soil moisture, and nitrate and ammonia supply estimates. To examine
environmental effects on cuhn survival we used t-tests to compare estimates of soil

variables for locations where planted culms survived and where they died.

Plant perception of environmental variation-dispersal scale planting

We conducted a second phytometer experiment to expand the range of sites in
which we measured spatial structure in plant growth and survival. We collected four
randomly chosen adult D. spicata from each of four populations (A2, D, L1, R) in
February 1997. We separated these plants into individual culms, potted each cuhn in a
separate pot, and fertilized potted culms twice per week with Peter's Peat lite (20-20-20)
fertilizer to encourage vegetative growth. In late March 1997, all cuhns that had
produced additional tillers (reproduced vegetatively) were further divided and repotted.
In mid-April we ceased fertilizing all plants in preparation for planting.

We designed this phytometer experiment to determine whether differences in the
spatial structure of soil resources among sites would affect plants. We speciﬁcally
focussed on spatial scales representative of CH and CL seed dispersal distances (10 cm
for CL seeds and 60 cm for CH seeds). Two of the sites used in this phytometer
experiment had D. spicata populations with a high or moderate proportion of short-
dispersing CL seeds (D, R) and two had populations with primarily far-dispersing CH
seeds (L1, A2).

We established four randomly located subplots (43 cm x 43 cm) within the
permanent plot in each study area. We placed a 7 x 7 cm quadrat at each comer of each

subplot and planted a culm of D. spicata at each comer of each quadrat (Figure 2.3). We

26

 

 

 

 

 

quad 9&2

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

L4 A1 D2 R3 A1 R2 D2
Subp lot 60 cm

D"’ R3 A1 D 2 A R2

R3 D2 A1 L4 L2 D2 R2 Al
U A3 D R3 A3 L2 D1

A3 L4 R3 D1 L2 A3 D1 R2
D R3 A1 L4 D1 R2 A3 L2
R3 D1 L4 A3 R2 D1 L2 A3

Figure 2.3: Four site planting design

Diagram of design used to plant cuhns in each population in the four
population phytometer experiment. Shown are four subplots located within a
plot. Diagonal distances across quadrats (10 cm) and across subplots (60 cm).
Distances were chosen to reﬂect the distances dispersed by CL and CH
seeds, respectively. Letters in the quadrat corners represent the source
population of each cuhn. A refers to culms from Population A2, D and R

refer to culms from Populations D and R, respectively, and L refers to cuhns
from Population L1. The number after the population letter refers to the

parent number in each population (1-4).

27

planted each subplot with cuhns from four parent plants, one from each source
population. Within each subplot, cuhns from a single parent were planted in two
opposing comers of a quad in each of two opposing quads in the subplot (see Figure 2.3).
Cuhns from each parent were represented in two randomly chosen subplots (of the four
possible) in one site for a total of 8 culms from each of eight parental plants. Eight culms
from each parent were planted into two sites; one site with a high proportion of short-
dispersing seeds and one with predominantly far-dispersing seeds). This gave us a total
of 256 plants across four sites.

We planted culms in Sites A2, L1, and R over a period of two days in early May
1997. We watered the cuhns three times during the ﬁrst week and once the following
week to ease transplant shock. Cuhns were planted into Site D one week later,
immediately after a heavy rain and transplants into this site were not watered aﬁer

planting. We measured culm survival and growth (number of tillers produced) in early

June 1998, when all cuhns were harvested. Plants were dried at 40°C, and weighed to
determine ﬁnal above-ground biomass. The number of tillers at the time of planting was
used as a measure of initial size.

We tested for differences in number of tillers and biomass (both log transformed
to improve normality) among source populations and among sites using ANOVAs. All
cuhns planted into Site A2 died within one month of planting, so it was not included in
any analyses. We used logistic regression (Systat 8.0 for windows, 1997. SPSS, Chicago
IL.) to test for differences in cuhn survival among the four source populations and the
three remaining sites. Because culm survival was low, the two culms planted into a

quadrat rarely both survived, so we could not use semivariance analysis to examine

28

spatial structure in culm growth at that scale. However, we used joincount analysis

(J oincounts, Epperson 1992) to compare similarity in cuhn survival within versus among
quads in a subplot. A signiﬁcant, negative, standard normal deviate statistic (SND < -
1.96) at the within—quadrat scale indicates that survival was signiﬁcantly more similar

within than among quadrats and, therefore, that it was spatially structured.

RESULTS
Site diﬁ"erences in soil resources

There were signiﬁcant differences among sites in soil moisture (1998), and in
nitrate and ammonia supply rates in both summer (1997; Figure 2.4e, g) and spring
(1998; p < 0.001 for each; Figure 2.4f, h). Sites L1 and L2 had higher soil moisture and
nitrate supply than Sites A2 and R in the spring (1998; Figure 2.4b, 0. Site A2 had higher
soil nitrate and ammonia supply than L1, L2, and R in the summer (1997; Figure 2.4e, g).
The four sites had similar ammonia supply in the spring (1998; Figure 2.4h). We could
not directly compare initial nitrogen availability (from soil extracts) or soil moisture
across sites because these measures were made in different years in the four sites.
Comparing within a measurement year, Site R was consistently more moist and had

lower nitrate and ammonia than A2 (Figure 2.4).

Spatial structure of soil resources
Using semivariance analysis, we found that the spatial structure of soil moisture
and nitrogen within these sites was often spatially structured (Figure 2.5, 2.6, 2.7). As

expected, soil resources were more spatially structured in some sites others. For example

29

r———|

Figure 2.4: Moisture and nitrogen levels across years at four study sites

Mean (1 1 s.e.; N = 58 to 68 samples each) soil moisture (a, b), the availability of
nitrate (c) and ammonia (d) and relative supply of nitrate (e, f) and ammonia (g,
h) nitrogen across four populations of Danthonia spicata. Soil moisture and
initial pools of nitrate and ammonia were sampled in June 1996 in Populations L1
and L2 (I), and in June 1997 in Sites A2 and R (I). Soil moisture was resampled
in all four sites in 1998. Summer nitrogen supply measures were sampled from
June to October 1997. Spring nitrogen supply measures were sampled from
October 1997 to May 1998.

30

udy sites

I, the availability of
and ammonia Q,
moisture and

3 in Populations LI

ure was resampled
re sampled from
ampled from

 

 

 

as 0.2 be b

 

1996, 1997 1998
= 0.15% —
2 8
3
.‘é’ 5‘ 0.1— —
0 on c
E \ a
= O 005- —
63 5" '
on
0- .4
c. 1 d.

 

 

Initial nitrate Initial ammonia

0.75—

  
  

N03 or NH 4 Availability
(ug N / g dry sorl)
O
L, O
‘1” ”1‘

 

 

e.

(relative sup ply)

 

 

 

 

 

 

g' 8 summer 11' spring
A 6- _
>5
a _ a
<1- ; 4- —
a .2 -
‘5 2— b
FE b b
0—
A2 L1 L2 R
Site

 

Figure 2.4: Moisture and nitrogen levels across years at four study sites

31

.Eﬂmetgmﬁom PE .80» 05 9 155?. Seton 05 was aeE>E0m 0:0 .30.» 05 we 0052536

05 9 0.6.3» 755$ QB 05 .§&o§>g0m £000 com .on Pan 0 as 05.03% 33% “gamma—mmméez ._0>0_ 36 W a 05 am 05250
330% “enemas 03065 EEwetgeO .8 -Eom :80 me .5500 Emma 032 05 5 AL 00—383 .05 :80 5 358030008 93 was
0:0 80> $88 E03680 003 85 0:50.50 Human 26% $0530 £03» .58 :0 E 33 S 380:3 0003 mEEwotggm no.“
0:: “003% {V 358050008 9.3 30.» .m 28 ~< mozm com 33 E 98 NA 93 5 Sam no.“ 003 E 880:8 203 agar—gang
com 3:: 0:8 .Ov 030805038 0:0 30> .87. 33% £000 5 £00» 95 new 003038 :8 com agar—«>00 new 38035385
800% 9.3 026 05588 :8 we 005530 28 00§5>mﬁ0m AWN 083m

32

 

 

a N0 5 8 0%
980 85000 35085000 0508500 A3833
02.. SN 2: o 8m 08 2: O 8m 8m 2: o 8m 03 2: o
L L r _ _ _ _ _ . _ . _ . _ . _ _ _ . _ . O
... - o f o - ...
r I 1 Ito m0
0 . f . m
I I T r m6 w.
m
- I I I S
0 . .

 

 

 

 

 

 

 

 

 

aouapmguras

 

 

 

 

 

 

0.60» 25 0020 0.50038 :00 00 005530 Ea 00550,; “N 2&3

33

Samar—02800 PE .30» 05 8 15:50 Boson 05 E3 gem—3380.».

0:0 .30» 05 mo 0030cawmm 05 8 £000“ 398% a8 05 .aotngm 5030 00m .on 80N a as 0.533% 33% «gomawmm

.8: 23 =me 3 358 3 .3 33:30.00 mm Aoﬁov a v no.8 “305:3,“ 3083 33 35 0.5383 330m ._0>0_ mod W a 05 3 05083
33% #305333 030%.: Eﬂwotazﬁg £000 .00 30500 Ewt 830— 05 5 AL 0.me34 AM .3 :6 8% >030 005.: 5 $8300

25 macaw baa «30:53 23 03.53 0332 BM 333033800 6:: 00300 4v mam“ macaw 93 3:: 0:00 .0v 3% 30885
$996 Swept: 0&3—8 5 0033283 6.N 059m

34

_

M

350 35in
com

h

com

_

»

o3

_

o

_

N

u—

E 8&3
com

h

8N

p

oo—

_

5 ”3a

35 85g
0 8m SN 2: o
_ _ . _ _ _ _ O

 

*

 

 

 

O

 

 

 

 

escape/muss

 

 

 

 

 

 

 

 

 

gag

bang mumps:— 02300 E occur—038$ 6N 03$

35

.Eﬁwoﬁggom 030 000% 05 00 805$ 80000

05 0:0 E05288 000 000% 05 .00 000000Ewmm 05 00 000.00.— 395.00 000 05 .800w0t0380m 0000 00m .on 0000 0 3 00300.5
30000 0000082000: 0:0 0me 3 0:00.: 0 03 000009000 mm 83V 0 v 3.8 E00036 0000.00 003 0000 00300.00 30% ._0>0_ 3.0
W 0 05 0 00300.00 030% 00005ch 000305 E00mot0>mE0m £000 .«0 000000 Emu 0030— 05 5 9.9 820002 A00: 00000 .dv wom—
wﬁam 000 A00: 0:00 Kb m 0:0 ~< 08% .80 33 E 000 «A 0:0 5 000mm 00.0 003 5 £000 00050 000 0000:0000 5 0.80250 Bram
0030008 EmobE w0n £000 000 3:30:03 325 5 00300.5 300% 00 00050800 ”EN 000me

36

080 88000
8...

A800 00:00me A800 800005

A83 38000
80

 

 

 

*

*

 

 

 

 

 

 

 

 

 

 

*
*

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

9003112011193

00.50008 0959::— w0: 8000 0:0 5:30:00 003:: 3 00300.50 003000 .00 0005 0800 ”EN 03$:

37

Site A2 (the oak savanna) had signiﬁcant temporally consistent spatial structure in soil
moisture (Figure 2.5) and also in initial (soil core) nitrate and ammonia levels (Figure
2.7). All three old-ﬁeld sites (L1, L2, and R) had some signiﬁcant spatial structure in soil
moisture or ammonia. However, Site L1 had no temporally consistent spatial structure.
In contrast, Site L2 (spring ammonia and nitrate among years) and Site R(soil moisture
and spring nitrate among years) had signiﬁcant temporally consistent spatial structure. In
Site D (the alvar grassland) soil depth (a surrogate for soil moisture) was not signiﬁcantly

structured (p = 0.650).

Temporal consistency of spatial structure in soil resources

Soil moisture showed consistent spatial structure among years in both Site A2 and
Site R, but not in Sites L1 or L2 (Figure 2.5). Neither ammonia nor nitrate was
consistently spatially structured among seasons in any site (Figure 2.8). However, spatial
structure in initial availability of nitrate and ammonia was frequently consistent with
spatial structure in nitrogen supply detected either one (A2, R) or two (L1, L2) years later
from spring 1998 resin bag measures. Sites A2 and L2 had consistent spatial structure in
both initial and relative supply for both nitrate and ammonia (Figure 2.9). In Site R, soil
cores and resin bags detected similar (consistent) spatial structure for nitrate, but not for
ammonia (Figure 2.9). The greater consistency between soil core and resin bag measures
of nitrogen than among resin bag measurements in different seasons (summer 1997 and
spring 1998) might suggest consistency among years but not among seasons. In Sites
Lland L2, nitrogen supply was measured over two full years to allow patterns that

persisted throughout the year to be compared among years. We summed 1997 summer

38

 

0000 0 03 0000000030 0000053000000 000 0E0 3 000000 0 03 00000w0000 00 32V 0 v no.8 0000::0w_0 00000—0

003 0005 000000.00 3:005 ._0>0_ 3.0 W a 0:0 00 000000000 3:000 0000::0w00 00000000 E00w0000>00 :000 .00 000000 0%: 0030— 05 0: AL
00.00.030.40 .2 .0..— .JV 0000 >030 000: 00 0000000 900000 000000.00 3:000 0000000000 305000 0000000 000 00000000 00.0 00000000000060
0000000 w00000 0000000 000 00000000 .00 0000000060 ”QN 003E

39

com

_

M

008 800005

_

com

L

_

2:

o com

N

A

003 005005

—

com

02

o

 

can

_

0

A

080 38000

com

0

OS

.00
0000

 

 

O

O

 

 

 

O

 

I vd
I wd

IN;

0.

 

0.0

 

 

0 3202

 

 

 

 

 

 

 

 

swam/x03

o

I 0.0
I wd
0

r N;

r

 

 

o. 0
00000002

000080 w00000 0000000 0000 000000000 .00 0000009000 ”wd 000mE

40

.on 0000 0 03 0000000300 0000000090000 000 0E0 3 000000 0 an 00000w0000 00 806v 0 v no.8 000000300

00000—0 00? 005 00000000 330m ._0>0_ mod W 0 05 00 000000000 30000 0002.000w00 00000000 0000w0000>00 0000 .00 000000 Emu 0030—
05 00 AL 0x00000040 .000 >030 0000 E 30000 000.000 000 000000000 0300—00 0000 900000 0000 0000000000 _000 5ng a 000 ~< 00009
300 M00000 00 ANA 000 J 000$ 003 w00000 0003000 000000.000 000000 0000000000 $003000 0000000 0000 000000000 00.0 00000w0000>00
b00000 00w00000 wag w00000 0000 £000 00ng 0000000 .00 0000000060 ”0N 000wE

41

M

080 85000

com

com

_

80

_

N

A

Aﬁov 8500mm—

8m

com

_

80

o

I

008 80805

com

_

com

_

80

_ L

o

N<

050 3530

com

_

com

03

_

000

o

 

 

§

 

 

*

 

 

O

T

 

 

.0

T vd
I wd

INA

 

0.0

 

2202

gamma/mg)

 

 

 

 

 

 

 

 

o

r
I to
0
I .06

IN;

 

o;

 

00.000000?

30000 E00000 30— w00000 000 0.000 0&00000 30000 .00 0000009000 ”QN 000w0m

42

.0000w0_00>00000

030 000% 000 00 00000000 0000000 00.. 000 0000300030000 000 000» 000 .00 000000300 000 00 000.000 00000000 000 000 .0000w0000>00000

0000 00 m .on 0000 0 >0 000000.000 000000 000000030000 00000 0E0 3 000000 0 >0 000009000 00 82V 0 v 3.8 00000300 0000000

003 0000 00300000 0000005 00>0. mod w 0 000 00 00300000 000000 00000300 00000000 0000w0000>00 00 -0000 0000 ‘00 000000 00m: 0300
000 00 AL 0000000040 .0000 0000 00 00000000000000 300 000 0000 000000 0000000000 003 0000 00300000 000000 3000 00000w0000>00 .3 000
:0 200 :0 0.0 0208 .4 38.0 + £30 32 05 020 208 .00 090 a 00003. 3200: 05 «soaps 03022 80 8000.052800
0000 030 00 0000» 000000 00ng 00 000000000 300m ”SN 000wE

43

 

 

 

 

 

 

 

00 00 .00 00 000
03085000 0508500 08035000 08080300
8m CON 80 0 COM com 80 o com com 80 o can CON oo— o
_ 0 _ L — 0 r F L 0 h _ _ 0 _ F _ 0 _ r _ 0 _ L o
O I O I O I O T
Ivd My
A
m.
o
9
N.
m
M
m.
o
9

 

 

 

 

 

 

 

 

 

 

 

3202 4

0000 030 00 0000» 000000 00w00000 .0 000000000 30% ”o— .N 000E000

44

and 1998 spring resin bag measures to get data comparable to 1996 full year
measurements. Comparisons between 1996 and 1997 measures of relative nitrate and
ammonia supply showed no temporally consistent spatial structure at the whole year time
scale in Sites Lland L2 (Figure 2.10). However, there was no signiﬁcant spatial structure
in Site Llor L2 in either nitrate or ammonia supply during 1997 (summer + spring), so it
was not possible to detect signiﬁcant temporally consistent spatial structure in these two

sites between 1996 and 1997.

 

These results show that four of the ﬁve study sites (excluding Site D) had spatial
structure during some times. Site A2 temporally consistent spatial structure among years
in soil moisture, ammonia and nitrate. In contrast, Site L1 had no consistent spatial
structure among years. Site L2 had somewhat less spatial structure that was consistent
among years than A2 (ammonia and nitrate were consistently structured, but soil moisture
was not). Site D had no signiﬁcant spatial structure in soil depth (a surrogate for soil
moisture). Similar to Site L2, Site R had less consistent spatial structure than A2, but
more than Site L1 or D (soil moisture and nitrate were consistently structured, but
ammonia was not). The real test of whether these differences among sites in amount of
temporally consistent spatial structure are relevant to plants is to use phytometers to

measure differences in plant “perceived” spatial structure.

Plant perception of environmental variation-variable scale planting

As might be expected, there were strong differences in grth between the

common garden and ﬁeld sites (L1, L2) for both biomass and number of tillers (each p <

45

0.001), but plants grew similarly in Sites L1 and L2 (p = 0.142). Source populations
differed in the number of tillers produced in the common garden and ﬁeld sites, with
plants from Population A1 producing more tillers than plants from Populations L1 and L2
(p < 0.001). But populations had similar biomass (p = 0.271). Differences in tiller
production among the source populations were similar across the common garden and
ﬁeld sites (p 2 0.170). Initial culm size was not a signiﬁcant determinant of ﬁnal size or
survival (P > 0.8) and was not used in analysis of this experiment.

Three of the four parent plants collected from Population L1 were infected with
the epiphytic fungus, Atkinsonella hypoxylon. Plants infected with this fungus often have
higher rates of tiller production (Clay 1984, Chapter 5). To determine whether parent and
source population differences in tillering were a result of infected culms, which were only
collected from Population L1, we removed infected plants from the analysis and tested for
differences among source populations and parents using only uninfected plants.

However, removing infected plants ﬁ'om the analysis did not substantially affect
population differences in tiller production or biomass.

Neither plant dry weight nor the number of tillers produced by culms (regression,
p > 0.210 in all cases) were signiﬁcantly related to estimates of soil moisture and relative
supply rates of nitrate and ammonia nitrogen for individual planting locations in the two
sites. However, in Site L2, the estimates of soil moisture in 1996 and 1998 for locations
where planted cuhns survived were signiﬁcantly higher than estimates for locations
where culms died (t-test, P 5 0.041). Culms that survived in Site L1 showed the same
tendency, as in Site L2, to be in higher moisture locations than culrns that died, but this

relationship was not signiﬁcant in this site (t-test, P 2 0.320). Average soil moisture was

46

lower in Site L2 than in L. (Figure 2.4), and so moisture may have been a more important
determinant of culm survival. Estimated nitrate supply for 1996 and spring 1998 (t-tests,
P S 0.035) and ammonia supply rates in all sampling periods (t-tests, P S 0.03 5) differed
among locations where cuhns survived and locations where they died in Site L2.
Surprisingly, culms that died were in locations with higher estimated nitrate and
ammonia supply than those that lived. Culms that survived in Site L1 were similarly
associated with lower nitrogen locations, but this difference was not signiﬁcant.

Culm survival was spatially structured in Site L2 at distance intervals less than
~150 em but was not spatially structured in Site L1 (Figure 2.11). Thus, in Site L2, where
similar spatial structure was detected by soil cores and resin bags for both nitrate and
ammonia, culrn survival was spatially structured, while in Site L, where we found very
little spatial structure in any soil variable, culm survival was spatially independent.
Neither plant dry weight nor the number of tillers produced by cuhns were signiﬁcantly

spatially structured in either site (p _>_ 0.090 for both variables in both sites).

Plant perception of environmental variation-dispersal scale planting

We conducted a second phytometer experiment that was designed to examine
differences among sites in spatially structured environmental variation “perceived” by
plants at the spatial scales encountered by dispersing D. spicata seeds. In this
experiment, initial plant size, measured as number of tillers, was a signiﬁcant factor in
determining ﬁnal dry weight and number of tillers produced in this experiment (both log
transformed to improve normality; P = 0.012 and 0.028, respectively), so it was included

as a covariate in all analyses. No culms survived in Site A2, so we could not include it in

47

8:000 85000
80 80 80 80 80 80 8 o 80 80 80 80 80 02 8 o

r _ _ _ _ H P _ _ _ _ _ VI

50 0. N

0 , .

1.. 00000,.-.
0 . 0

.0 00m

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

00 000

0000000000

00000000 0 00000000 00.0- 000 00.— 0003000 0.97% 0000000000003 000050.000 000000.00 000+ 0000 000000w .0 0.97% 00000000
0000000 00005000 00000000 00. 0- 0000 0000 .00 0.2 000000000000 000000000000 0000 000000003 000 00 00003000 0000 00 000000000
00000 00 .00 0000000000 000 00000000 00 0000 000 0000000 000000000m 0 0000 0000000000 0000000000 :0va 000300 .00000 055%
8.0 000 95 :0 025... 53 0.. 2325... 00000 0 0.0 2.000

48

any analyses. Final dry weight and number of tillers produced differed among the three
other sites (p = 0.006 and p = 0.010, respectively). Plants from different source
populations did not differ in ﬁnal dry weight and, unlike in the variable scale planting,
source populations also did not differ in the number of tillers they produced (p = 0.433).

However, cuhn survival differed among the three sites (Logistic regression, P <
0.001; Systat 8.0 for Windows, SPSS Inc. Chicago, Il.). Cuhns that survived were
signiﬁcantly larger initially than culms that died, so initial size (initial number of tillers)
was included in this and subsequent logistic regressions. Source populations also
differed in survival, accounting for initial size (P = 0.002), suggesting that culms from
some source populations were more robust than others were (Table 1). However, there
was not a signiﬁcant interaction between source population and planting site (p = 0.646),
so culms from different source populations did not differ in their “perception” of habitat
quality.

Culm survival was more similar within than among quadrats in Site D
(joincounts, SND = -2.47), indicating signiﬁcant spatial structure in culm survival. Culm
survival was not signiﬁcantly structured in either Site L1 (SND = 0.06) or R (SND =
0.93). However, power to detect spatial structure in discrete variables is highest when
both categories, in this case, survival and death, are equally probable. We had equal
power to detect structure in survival in Sites D and L, with 24 of 64 culms surviving in
each. However, we had very little power to detect spatial autocorrelation in Site R,
because only six of 64 cuhns survived in that site.

The results of the variable scale phytometer experiment suggest that D. spicata

plants “perceive” spatially structured soil resources over a range of scales in Site L2 but

49

not Site L1. The dispersal scale phytometer experiment further supports the presence of
differences in spatial structure among sites at the speciﬁc scales encountered by
dispersing CL and CH seeds (see Chapter 3). The results of both phytometer experiments
suggest that the effects of soil resource variation act through survival variation, rather

than through differences in growth.

DISCUSSION
Structure of soil resources

Spatial structure of soil resources is well documented within and among
populations (0. g. Robertson and Gross 1994, Gross et al. 1995, Miller et al. 1995,
Ehrenfeld et al. 1997). However, few studies have explicitly assessed the temporal
consistency of measured spatial structure of soil resources (but see Goovaerts and Chiang
1993) or examined its relevance to plant performance (but see Miller and Fowler 1997).
In this study, we examined the spatial scale and temporal persistence of heterogeneity in
soil moisture, nitrate, and ammonia across ﬁve herb-dominated communities and related
this to spatial structure in the performance of D. spicata phytometers. We speciﬁcally
focused on differences among sites in the magnitude of spatial structure in these
resources and how this was “perceived” by D. spicata.

There were signiﬁcant differences among sites in all of the soil variables we
measured and also in the degree of spatial structure in soil resources and its temporal
consistency. Each soil variable we measured was spatially structured in at least one site,
during at least one sample time. Contrary to the expectation put forth by Grime (1994)

that very unproductive habitats would be dominated by temporal rather than spatial

50

variation in resource conditions, the two driest sites that we measured (A2, R), which had
the lowest plant biomass production (McCormick, unpublished data), tended to have
more spatial structure in soil resources than the two moister, more productive sites (L1,
L2). However, Grime's expectation was based on shorter-term (within individuals in both
space and time) nutrient dynamics than we examined. The disparity between his
predictions and our results underscores the importance of assessing the spatial and
temporal dynamics of soil nutrient variation across the multiple spatial and temporal
scales relevant to plants. It seems possible that ephemeral plant effects on resource
availability may dominate productive habitats while differences in soil type or long term
plant effects dominate less productive habitats.

Environmental factors may differ in the temporal and spatial scales at which they
are structured (Grime 1994). In this study, spatial variation in soil moisture was
temporally consistent in the driest two sites we studied (A2, R). Ammonia and nitrate
supply were inconsistent among seasons in all three old-ﬁeld Sites (L1, L2, and R).
Goovaerts and Chiang (1993) found similar inconsistency in patterns of ammonia
availability among seasons in a fallow agricultural ﬁeld. This suggests that seasonal
differences in spatial pattems of nitrogen availability may be common, at least in old-
ﬁeld sites.

Ammonia (Sites A2 and L2) and nitrate (Sites A2, L2 and R) both had consistent
spatial structure among years in the spring in some sites. This suggests that consistent
spatial structure during one season may be hidden in full year measurements by
inconsistency among seasons. If so, then it is important to measure nitrogen supply

during the time of year most important to plants of interest (0. g. during periods of intense

51

growth) to relate spatial environmental structure to plant performance. If soil resource
structure varies among seasons, then plant species that differ phenologically may also
“perceive” different amounts of resource structuring in the same community.

Plant species effects on the supply rate of both nitrate and ammonia can be
substantial (e. g. Wedin and Tilrnan 1990, Wardle 1998) and could produce differences in
spatial structure among seasons. Our study sites were dominated by long-lived perennial
plants and if plant species affected the spatial structure of nitrate and armnonia, then they
could produce spatially structured variation in soil resources that was consistent over
many years. Strong species effects on soil nutrients have been detected in experimental
studies with other grass species (Tilrnan and Wedin 1991) and also for nitrogen-ﬁxing
species (e. g. Vitousek et al. 1987, Wardle 1998). If species composition affects soil
processes, and species occur in predictable locations over time, then spatial structure in
soil nutrients may be more temporally consistent in communities that are dominated by
long-lived species. Therefore mature grasslands with long-lived species, such as the
habitats in which D. spicata grows (although D. spicata is relatively short-lived;
Darbyshire and Cayouette 1989), may have more consistent spatial structure than habitats

dominated by ephemeral species, like early successional ﬁelds.

Plant ”perception” of environmental structure

Our phytometer experiments revealed similarities between measured spatial
structure of soil resources and spatial structure “perceived” by D. spicata plants. The
major effect of soil resource variation on D. spicata was to affect plant survival. We

found little effect of spatial environmental structure on growth of plants that survived. In

52

the variable scale phytometer experiment, survival of culms planted into Site L2, which
had more spatial structure in soil nitrogen than L1, was signiﬁcantly spatially structured at
scales less than ~150 cm, but culm survival was not spatially structured in Site L1 (Figure
2.11).

Similarly, in our dispersal scale phytometer experiment, culm survival was
signiﬁcantly structured in Site D, but not in Site L1, which had little spatial structure in
any factor. Our analysis of spatial variation in Site D was limited to soil depth because
soils in this site were too shallow to allow soil core removal or resin bag installation. For
this variable, we found little evidence of spatial structure, but other factors may have
affected plant survival in this site and cuhns planted in site D “perceived” the
environment as spatially structured. Also, the randomly located sample points and
phytometers in this study evaluated the way that environments would be sampled by
randomly dispersing offspring. But when dispersal is non-random (as it is in D. spicata),
other ways of measuring spatial environmental structure, and thus the spatial
predictability of environmental conditions encountered by plants and their offspring may
be more powerful (e. g. using dispersal curves to determine sampling designs; see Chapter
3). Spatially structured cuhn survival in Site D and unstructured survival in Site L1 was
consistent with predicted differences among the sites based on dispersal distance (see
Chapter 3). However, extremely low survival of plants in Populations A2 and R limited
our ability to make general conclusions about the relationship between measured and
“perceived” spatial environmental structure among populations from the dispersal scale

phytometer experiment.

53

Our study, like many other studies (0. g. Sultan and Bazzaz 1993, Ehrenfeld 1997),
documented that there is substantial temporal variation in soil nitrogen. Temporal
variation in nitrogen in these sites was evident from the inconsistency of nitrate and
ammonia supply (from resin bags) and from inconsistency among seasons (Figure 2.8).
However, spatial structure of soil variables may be temporally inconsistent among
seasons but consistent among years for a given season, as is suggested by the consistency
between nitrogen pool and supply measurements (Figure 2.9). If a plant species is
affected by soil resource variation during a speciﬁc time of year (determined by its
phenology), then consistent spatial structure during that season may have a consistent
effect on that species' population and community dynamics despite inconsistent resource
structure among seasons. The differences in spatial structure of culm survival and their
relation to soil moisture and nitrogen conditions (variable scale phytometer experiment)
despite seasonal inconsistency of spatial structure in nitrogen supply supports this
interpretation. Similarity between spatial structure in soil resources and spatial structure
in phytometer survival suggests that temporally consistent spatial structure in soil
resources among years is an important determinant of D. spicata performance, but
experiments in additional sites and with other species are needed to expand the generality

of these results.

54

LITERATURE CITED

Bell, A. G., and M. Lechowicz. 1991. The ecology and genetics of ﬁtness in forest
plants. I. Environmental heterogeneity measured by explant trials. J oumal of
Ecology 79:663-686.

Bell, A. G., and M. Lechowicz. 1994. Spatial heterogeneity at small scales and how
plants respond to it. In: Caldwell, M. M., and R. W. Pearcy. Exploitation of
Environmental Heterogeneity by Plants: Ecophysiological Processes Above- and
Belowground. Academic Press, San Diego, CA. pp. 391-414.

Chapin, F. S. 1980. The mineral nutrition of wild plants. Annual Review of Ecology
and Systematics, 11:233-260.

Clay, K. 1984. The effect of the fungus Atkinsonella hypoxylon (Clavicipitaceae) on the
reproductive system and demography of the grass Danthonia spicata. New
Phytologist 98:165-175.

Corbett, A. L. 1998. Root responses to nutrient heterogeneity: a comparison of dominant
and subordinate species from old ﬁelds. Ph. D. Dissertation, Michigan State
University.

Darbyshire, S. J ., and J. Cayouette. 1989. The biology of Canadian weeds. 92.
Danthonia spicata (L.) Beauv. in Roem. and Schult. Canadian Journal of Plant
Science 69: 121 7-1233.

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, 185-798.

Goovaerts, P., and C. N. Chiang. 1993. Temporal persistence of spatial patterns for
mineralizable nitrogen and selected soil properties. Soil Science Society of
America Journal, 57:372-381.

Grime, J. P. 1979. Plant Strategies and Vegetation Processes. John Wiley and Sons,
Ltd, New York.

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

Miller, R. E., and N. L. Fowler. 1994. Life-history variation and local adaptation within

2 populations of Bouteloua rigidiseta (Texas-gramma). Journal of Ecology
82:855-864.

55

Miller, R. E., J. M. Ver Hoef, and N. L. Fowler. 1995. Spatial heterogeneity in eight
central Texas grasslands. Journal of Ecology 83:919-928.

Robertson, G. P. and K. L. Gross. 1994. Assessing the heterogeneity of belowground
resources: quantifying pattern and scale. In: Caldwell, M. M. and R. W. Pearcy,
eds. Exploitation of environmental heterogeneity by plants: ecophysiological
processes above- and belowground. Academic Press, San Diego, CA. pp 237-
252.

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

Schlesinger, W. H., J. F. Reynolds, G. L. Cunningham, L. F. Huenneke, W. M. Jarrell, R.
A. Virginia, and W. G. Whitford. 1990. Biological feedbacks in global
desertiﬁcation. Science 247: 1043-1048.

Stratton, D. A. 1995. Spatial scale of variation in ﬁtness of Erigeron annuus. The
American Naturalist 146:608-624.

Stratton, D. A., and C. C. Bennington. 1998. Fine-grained spatial and temporal variation
in selection does not maintain genetic variation in Erigeron anuus. Evolution, 52:
678-691.

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

Tilrnan, D. and D. Wedin 1991. Plant traits and resource reduction for 5 grasses growing
on a nitrogen gradient. Ecology 72:685-700.

Venable, D. L., and J. S. Brown. 1988. The selective interactions of dispersal,
dormancy, and seed size as adaptations for reducing risks in variable
environments. The American Naturalist 131:360-384.

Vitousek, P. M., L.R. Walker, L. D. Whiteaker, D. Mueller-Dombois, P.A. Matson.
1987. Biological invasion by Myricafaya alters ecosystem development in
Hawaii. Science 238:802-804.

Wardle, D. A., G. M. Barker, K. I. Bonner, K. S. Nicholson. 1998. Can comparative
approaches based on plant ecophysiological traits predict the nature of biotic
interactions and individual plant species effects in ecosystems? Journal of
Ecology 86:405-420.

56

Chapter 3

THE ROLE OF SPATIAL ENVIRONMENTAL HETEROGENEITY IN THE
EVOLUTION OF GENE FLOW DISTANCE IN DANTHONIA SPICA TA
INTRODUCTION

The scale of gene ﬂow within and among populations is critically important in
governing the scale of local adaptation, the maintenance of genetic diversity, and the
cohesiveness of populations. The evolution of gene ﬂow distance in response to spatial
heterogeneity has received theoretical, but little experimental attention (e. g. Spieth 1979,
Gillespie 1981, Hedrick 1986, Venable and Brown 1988). Theory predicts that gene ﬂow
distance should evolve to match the spatial scale of variation in natural selection (Tonsor
1990, Balkau and F eldman 1973, Christiansen and F eldman 1975).

Gene ﬂow in plants includes both pollen and seed dispersal. In many plants, gene
flow occurs predominantly through pollen dispersal (Price and Waser 1979). In others,
especially selﬁng plants, it is primarily through seed dispersal. It has been argued that
some plants may have little ability to ﬁne tune gene ﬂow distances because of reliance on
vectors over which they have little control (Waddington 1983). However, it is easy to
imagine how genetic traits could inﬂuence the movement of pollen and seeds that are
wind or ballistically dispersed (Okubo and Levin 1989).

Increased gene ﬂow is favored by many factors including temporal variation in
site quality among generations and competition with parents and siblings. These factors
selecting for non-zero dispersal are likely always important, however, restricted gene
ﬂow may evolve when environmental heterogeneity is spatially structured (Tonsor 1990).
The balance of the relative strengths of selection to increase and decrease dispersal
determines the optimal dispersal distance. Although temporal variation among
generations (e. g. Stratton and Bennington 1998) and the strength of competition between

parents and offspring or among siblings undoubtedly differs among environments, the

57

major factor selecting for different dispersal distances among relatively similar
populations is likely to be the strength of selection by spatially structured environmental
variation to decrease dispersal distance.

Temporal variation at very small time scales (e. g. within a growing season) may
be substantial at individual points within a population without affecting the mean value of
a location relative to other locations in a site. Some traits, such as seed dispersal, have a
strong spatial component (i.e. may directly affect the spatial distribution of plants or their
offspring) and may evolve relatively independent of smaller scale (within generation)
temporal variation. If traits affecting gene ﬂow can adapt to spatial environmental
variation, then gene ﬂow evolution will be an important factor in tailoring population
structure to the scale of environmental variation, thus affecting the process of local
adaptation and the maintenance of genetic variation.

When environmental variation is temporally consistent and spatially structured,
increasing seed dispersal results in offspring encountering environmental conditions that
are increasingly less likely to resemble their maternal parent's environment. When spatial
variation is consistent over time and most of the microhabitats in a site are unsuitable for
growth, then a maternal environment will be better than a random location in the site. If a
maternal location is better than an average location for seedling growth, then increasing
similarity between maternal and offspring conditions may represent an increasing
probability of offspring encountering good environments.

In populations where the difference between maternal and random locations is
very strong, the evolution of dispersal may be dominated by spatial environmental
variation and may be quite short (although still non-zero). However, in sites where there
is either little spatial environmental variation or where spatial variation is not structured,
the similarity between offspring and maternal conditions will not depend on offspring
dispersal distance. In these populations, spatial environmental variation will not select

for decreasing dispersal. Selection by other factors, such as competition with siblings and

58

parents, acting to increase dispersal may dominate the evolution of dispersal distance in
environments with little spatially structured environmental variation.

One way to approach the question of whether spatial environmental variation can
affect the evolution of gene ﬂow distance among populations is to examine existing
differences in seed dispersal distance among populations in sites that differ in amount of
spatially structured environmental variation. I examined the correlation between spatial
environmental variation and gene ﬂow distance in Danthonia spicata. Danthonia spicata
produces two differently-dispersing seed types and previous work has shown that there is
substantial genetic variation within and among D. spicata populations for proportional
allocation to these two seed types (Clay 1982). In plant species with heteromorphic seeds
that disperse to different distances, the ability to modify dispersal distance may be
particularly strong because dispersal distance can be strongly altered by allocating
different proportions of seeds to the differently-dispersing structures (Lloyd 1984).

I used simulated dispersal distributions and the spatial structure of environmental
variation to estimate the similarity of offspring to maternal conditions in ﬁve D. spicata
populations. I then determined the effect that decreasing dispersal distance would have
on the similarity between maternal and offspring conditions and compared that effect to

the proportional allocation to short-dispersing seeds in these populations.

MATERIALS AND METHODS
Study species

Danthonia spicata (Poaceae) is a perennial, C3 bunch grass that is native to North
America and commonly occurs in dry, nutrient-poor oak savannas, old-ﬁelds, and
openings throughout the eastern and northern United States and Canada (Darbyshire and
Cayouette 1984). Danthonia spicata plants produce two ﬂower types (chasmogamous
and cleistogamous) that produce similar-sized seeds (Clay 1983). Potentially outcrossed,
wind-pollinated chasmogamous (CH) ﬂowers are produced at the tip of the reproductive

59

stalk. Obligately self-fertilized cleistogamous ﬂowers are produced in the axils of leaf
sheaths along and at the base of the reproductive stalk (Figure 3.1). Cleistogamous (CL)
seeds, up to ten per leaf node, remain within the leaf sheath, attached to the reproductive
stalk until germination, while CH seeds are dispersed individually by wind. These
differences in dispersal mechanism mean that the two seed types are likely to have
different dispersal potentials. Chasmogamous seeds of D. spicata are estimated to be less
than 10% outcrossed and there is little quantitative genetic difference between the two
seed types (Clay 1982). The relatively small genetic difference between the two seed
types means that gene ﬂow distance in D. spicata depends largely on seed dispersal
distance. Genetic variation for the proportion of the two seed types among populations
allowed me to investigate the relationship between spatial environmental heterogeneity

and dispersal distance in this species.

Population and site characterization

I located eight sites where D. spicata occurred, all of which were dominated by
perennial, herbaceous vegetation. None of these sites had been plowed or tilled for at
least 45 years. Six of the survey sites were located in southwest Michigan; three were
remnant oak savannas in the Allegan State Game Area near Allegan, Michigan (A,, A2,
A3); and three were old-ﬁelds near the WK. Kellogg Biological Station (L1, L2, LA). A
fourth old-ﬁeld site was located in the Rose Lake Wildlife Research Area in central
Michigan near East Lansing (R). The eighth survey site was an alvar grassland in the
Maxton Plains Preserve on Drummond Island, in northern Michigan (D).

I established a 4 x 6 m permanent plot in each of the eight sites during the spring
of 1996. I measured the percent of seeds that were cleistogamous (CL) in each
population by collecting two reproductive stalks from each of 30 randomly selected D.
spicata plants within each plot in June 1996. The location of each sampled plant was

noted and the plant was marked with an aluminum tag so it could be relocated. In ﬁve of

60

CH seeds

Node 2

 

mania"

Figure 3.1: Danthonia spicata chasmogamous and cleistogamous seeds
Chasmogamous (CH) seeds in a terminal inﬂorescence and cleistogamous (CL) seeds
within leaf sheaths at each node along the reproductive stalk. Nodes are numbered from
the base of the plant towards the top, just below the CH inﬂorescence.

61

the populations (A2, L,, L2, D, R), I resampled reproductive stalks from all tagged plants
that reﬂowered in 1997 to verify that percent allocation to cleistogamy was consistent
among years. Three of the eight populations (L,, L2, R) contained plants infected by the
epiphytic fungus Atkinsonella hypoxylon. Because A. hypoxylon causes abortion of all
CH seeds in infected plants (Clay 1984), I only sampled reproductive stalks from
uninfected plants. In ﬁve of the eight sites (A2, L,, L2, D, R), I measured plant density
(plants m'z) along two randomly located transects (25 cm x 12 m long).

On each sampled reproductive stalk, I counted the number of CH seeds and
measured total reproductive stalk length to use in CH dispersal simulations. I also
counted the number of CL ﬂowers and initiated seeds at each node along the reproductive
stalk, denoting the position of each node by a node number (Figure 3.1). Percent
cleistogamy was calculated for each individual as the total number of CL ﬂowers x (the
total number of CH seeds and CL ﬂowers)“. I used ﬂowers and initiated CL seeds rather
than mature seeds to estimate CL seed production, because D. spicata's CH seeds mature
and disperse before CL seeds mature. Very few individuals reﬂowered in 1997, so I was
only able to assess the consistency of percent CL seeds in those few individuals in each
population. I used an AN OVA assessing the effect of Population and 1996 percent
cleistogamy on 1997 percent cleistogamy to determine the temporal consistency of

allocation to CL seeds among populations.

Measuring dispersal distance

To estimate the primary dispersal distance of D. spicata seeds, I collected three
plants with multiple (20 to 40) reproductive stalks and mature, but not yet dispersing, CH
seeds from two of the study populations, two from Population LI and a third from L2.
Each plant was excavated in early June 1996 and placed in a 10 x 10 cm square pot. I

then transferred each plant (and pot) to a dispersal arena in a ﬁeld location at the W.K.

62

Kellogg Biological Station more than 500 m from the nearest naturally-occurring D.
spicata plant.

Each dispersal arena consisted of a 3 m x 3 m sheet of plywood, the central 1.2 m
x 1.2 m area of which was overlaid by an egg crate panel light diffuser. I sprayed the
entire arena with Tanglefoot (Tanglefoot Company, Grand Rapids, MI) so that it would
trap dispersed seeds (as per Thiede and Augspurger 1996). The potted experimental plant
was placed in a 10 x 10 cm hole in the center of the diffuser and plywood. The three
dispersal arenas were separated by at least 5 m. Using these dispersal arenas, I was only
able to estimate primary dispersal as it might occur with no surrounding vegetation.
However, D. spicata often grows in very open habitats, with little tall vegetation, so these
estimates probably provide a good estimate of the primary dispersal distance of CH and
CL seeds. Secondary dispersal of CL seeds in the ﬁeld appears to be very limited,
evidenced by the presence of persistent reproductive stalks near and attached to parent
plants, from which CL seeds germinate (McCormick, personal observation).

I determined the distance traveled by dispersed seeds in early July 1996, after the
reproductive stalks had senesced. I noted the distance each chasmogamous (CH) seed
had traveled from the base of the parent plant. I also noted the lengths of all reproductive
stalks. Reproductive stalks generally remained intact along their length, allowing stalk
length to be easily measured. I ﬁt a regression of average CH dispersal distance on
average stalk length across the three sets of dispersal measures (N = 3 plants for CH
dispersal distance) to determine whether stalk length could be used to predict dispersal
distance.

To estimate the dispersal distance of cleistogamous (CL) seeds, I measured the
distance from the base of the parent plant to each leaf node on each reproductive stalk
(21-37 per plant). I collected each reproductive stalk and counted the number of CL
seeds at each node. Nodes were numbered relative to their position along the

inﬂorescence (Figure 3.1). The total number of leaf nodes on a reproductive stalk ranged

63

from 2 to 5. All 65 reproductive stalks had at least two leaf nodes, but only 12 stalks had
a ﬁfth node, so the sample size for estimating the distance traveled by each node number
ranged from 12 to 65. The dispersal distance of CL seeds was considered to be the
distance ﬁ'om the base of the parent plant to the position of the leaf node where they
occurred. I used these measurements to determine the mean dispersal distances of CH
and CL seeds and the relation to plant height. I then incorporated this information into a
dispersal simulation to investigate how average dispersal differences changed as the
proportion of CL seeds changed, so that I could compare dispersal in other populations

based on traits I could easily measure.

Dispersal simulation

Because CH seeds are dispersed individually and CL seeds at a node are dispersed
together, I used separate simulation models to predict the dispersal of CH and CL seeds.

I used the relationship between CH seed dispersal distance and reproductive stalk length,
determined in my dispersal experiment, to estimate average dispersal distances for CH
seeds in each of my eight populations. In each of the eight populations I determined the
average stalk length of 32 plants in each population, and the regression of average CH
dispersal distance on stalk length in the dispersal experiment.

To estimate CL dispersal distances, I used the relationship between the node
location dispersal distance of CL seeds from my dispersal experiment. I wrote a
FORTRAN program to randomly select a reproductive stalk sampled from a ﬁeld
population (on which I had counted the number of seeds at each node) that determined
the distribution of seeds along the reproductive stalk. The program then randomly chose
a reproductive stalk from my dispersal experiment (for which I had measured the distance
each node along the stalk dispersed) that determined the distance each node would

disperse. I assumed that all CL seeds at a leaf node dispersed to the same location.

64

This random selection process was carried out 200 times, with replacement, for
each simulation run. CL seed dispersal simulations were run ﬁve times for each
population. Each pOpulation was simulated separately. For each run, I calculated the
average CL dispersal distance. I also generated a frequency distribution for the
proportion of seeds dispersing to different distances using 5 cm intervals. I used these
simulations to determine whether different proportional allocation to CL seeds produced

different dispersal patterns among these eight populations.

Combining dispersal and environmental variation

In environments where resources are highly spatially structured, seeds that
disperse a short distance experience enviromnental conditions more similar to the
environment of their mother than seeds that disperse a long distance. To examine the
relationship between D. spicata dispersal distance and the amount of environmental
variation offspring experience relative to maternal conditions, I measured the distribution
of spatial variation in soil resources in the ﬁve study sites in which I had simulated
dispersal distance. These ﬁve sites were selected to encompass the range of proportional
allocation to CL seeds measured in my site surveys (Figure 3.2a). Two of the selected
populations had a low percent of CL seeds (A2, L,), two had an intermediate percent of
seeds that were CL (L1, D), and one had a high percent of seeds that were CL (R; Figure
3.2a). Each of these ﬁve populations also had substantial variation in percent of CL seeds
within the population (See ranges, Figure 3.2a).

I chose to examine soil moisture and the relative supply rate of ammonia-nitrogen
as estimates of environmental quality because they were relatively consistent over time in
these sites (Chapter 2) and often limit plant growth in the dry, nutrient-poor environments
where D. spicata occurs. Soil moisture and ammonia supply were not correlated within
these sites (Chapter 2), so I have treated them as independent measures of environmental

variability. In four of these ﬁve sites (A2, L,, L2, R), I measured soil

65

Figure 3.2: Characteristics of study populations

Percent of cleistogamous (CL) seeds (mean i l s.e.; a.) and mean reproductive stalk
length (i 1 s.e.; b.) of each of the eight study populations, and density of D. spicata
plants in ﬁve of the study populations. .Diamonds in a. indicate the highest and lowest
percent of cleistogamous seeds measured among the 32 plants sampled in each
population.

66

 

%CL

.5'

 

Stalk Length (cm)

 

 

Density (plants / m2)?

 

 

 

 

07 I I I I
LA L

1
Population

Figure 3.2: Characteristics of study populations

67

moisture (gravimetrically) and relative supply rate of ammonia-nitrogen (using resin
bags). The supply rate of nitrate was also measured in these populations, but it was not
consistent over time (Chapter 2), so its rate of supply at a maternal location could not be
used to predict offspring environmental conditions. In the ﬁfth site (D) soils were too
shallow for resin bag installation or for soil cores (avg. 4.5 cm deep), so I used soil depth
as an indicator of soil moisture.

I measured soil moisture and ammonia at 70 points in a 4 x 6 m plot in each site
using a stratiﬁed, nested sampling design. I measured soil moisture gravimetrically in
Sites A2, L,, L2, and R in early June 1998, during the period of peak D. spicata growth. I
sampled soil moisture using 10 cm deep, 2.5 cm diameter soil cores. A ten gram
subsample of each soil core was passed through a 2mm sieve and dried at 65° for 72
hours before being reweighed (Chapter 2). In the ﬁfth site (D), I measured soil depth in
June 1997 as an indicator of moisture. I measured the relative supply rate of arnmonia-
nitrogen using bags of mixed bed ion exchange resin (Dowex MR-3, Sigma Corp.). The
resin bags were placed in the ﬁeld in October 1997 in the same locations I had sampled
for soil moisture. These resin bags were buried under 5 cm of soil (as per Lajthe 1988) in
four sites (excluding Site D) and collected in May 1998. The resin bags were rinsed,
dried at 40° for 5 days, and extracted with 1M KCl (see Chapter 2) to provide an index of

ammonia supply.

Calculations

I used semivariance analysis to quantify the relationship between environmental
variation and distance separating sample points for soil moisture and ammonia in each
site. Semivariograms summarize the average variance between pairs of points as a
function of their distance apart and can be used to estimate the extent of spatial structure
in variable (Rossi 1992, Robertson and Gross 1994). I calculated the semivariance

(standardized by the overall level of variation) of both soil moisture and ammonia supply

68

using 5 cm step sizes (distance classes) from 0 to 150 cm (active lag). This active lag
encompassed spatial structure at distances beyond the farthest seed dispersal distance I
measured. I used standardized semivariances so that I could assess the effect of different
patterns of allocation to CL seeds among populations, independent of the amount of
environmental variation in each site.

I used dispersal distributions generated from my simulations of CH and CL seeds
in each population as weights on the semivariograrns calculated ﬁom environmental
samples to estimate the impact of the percent of cleistogamy on the average amount of
environmental variation encountered by dispersing seeds (Figure 3.3). On average, the
seeds dispersing to a given 5 cm interval would encounter conditions whose difference
from maternal conditions was predicted by the semivariance at the corresponding 5 cm
step. Using the proportion of seeds dispersing to a given 5 cm interval to weight the
semivariance calculated for environmental samples separated by that same interval
distance allowed me to calculate an average variation from maternal conditions that
would be encountered by dispersing offspring. This allowed me to estimate
environmental variation as dispersing seeds would encounter it.

I calculated the average variation from maternal conditions in soil moisture and

ammonia encountered by all seeds (CL and CH), in each site (SCL + SCH):
29
(SCL + SCH): 2(550 x PCL.5n)+(SSn x PCH.5n) [1]
n-l

29
Wrerz Pcr.5n + Pans" =1

.01
Where S5,, is the standardized semivariance calculated for the nth distance and
Per,» and Pens“ are the proportion of CL and CH seeds, respectively, dispersing in
distance class 11, 5n cm from parent plants. I used equation [1] to calculate variation from
maternal conditions that would be encountered by offspring of plants allocating low,

moderate, high, and the population average proportion of their seeds to cleistogamy for

69

000000000 0000000000000 $000.00 000 00000000 0003000 300000000 05 .00

00000000 003000 00000000300000 00w003 00 000—0 00000000 000 0000.0 000000000 000000.00 w000000000 00000 w000D 000000 .00 00000 0003000
30000000 000 000000000 0000000000000 000. 000000000 0000000000300 0000000000 0000.0 0000000.» 00 000w003 00 0030000 000 m 0000 00
00000 0000 00% ”EB 000000m0000000 000 00000 0003 SUV 000000w0000000 .00 000000000 0000—0000 00.0 .00 000 w000000000000 80005
000000000» 0000000000300 00 000000000 0000 w00000m ”0m 000w00

7O

A008 0000005
mm 2. 3 mm 9. mm mm 2 m

 

spoas jo uonrodord

 

‘9.
o

 

 

------.--*--

 

 

 

 

 

KJINIIIIIISSICI [00010003

000000000, 0000000000000 00 00000 0000 0000 M00300 ”Wm 000000

71

each population. I used the seed distribution simulated for Population A2 to estimate the
amount of variation from maternal conditions that would be encountered by seeds from a
parent with low allocation to CL seeds, Population D for a parent with moderate
allocation to CL seeds, and Population R for a parent with high allocation to CL seeds.
These calculations allowed me to determine how changing allocation to CL seeds would
affect the average similarity to maternal conditions encountered by dispersing seeds
across all ﬁve populations.

I also calculated an equal weighting of the similarity to maternal conditions
encountered by the two seed types ($100%CL , 3100%CH)I

29
Srooescr = 265.. x Pens") [2]

1:81

29
8...... =20... x P.. [3]

ntl

29 29
Where; Perm/z PCH.5n = 1

n=1 "=1
Where S51, is the semivariance calculated for the nth distance and PCLSn and
PC3151, are the proportion of CL and CH seeds, respectively, dispersing in distance class n,
5n cm from parent plants. Each equation is summed over all distance intervals to which
simulated seeds dispersed. For all populations, the farthest distance class containing
seeds was class 29 (145 cm). This equal weighting of the two seed types was used to
indicate differences in the amount of variation from maternal environmental conditions
encountered by CH and CL seeds as a ﬁinction of differences in their dispersal distance,
independent of the proportion of seeds allocated to each seed type across different

populations.

72

Predictions

If allocation to CL seeds in D. spicata is a result of the scale of spatially
structured environmental variation, then populations with different amounts of allocation
to CL seeds would occur in environments in which CL seeds were differently effective at
reducing the amount of variation from maternal conditions that offspring encountered.
Speciﬁcally:

I) If shorter dispersing CL seeds increase the average similarity between maternal
and offspring environmental conditions, then increasing the percent of CL seeds would
decrease the average variation encountered by dispersing seeds across populations,
increasing similarity to parent conditions. This would produce a negative correlation
between the average environmental variation encountered (SCL + SCH) and simulated
allocation to CL seeds (high, moderate, or low).

2) Because CL seeds disperse shorter distances than CH seeds, increasing
allocation to CL seeds decreases average dispersal distance. Decreased dispersal distance
will only increase similarity to maternal conditions, if environmental variation is spatially
structured. If increasing similarity to maternal conditions is beneﬁcial, then I would
expect spatially structured environmental variation to select for increased allocation to
CL seeds. I predict that populations of D. spicata in environments where decreased
dispersal distance increases the similarity between maternal and offspring conditions
would have higher allocation to CL seeds than those occurring in environments where
decreased dispersal distance was less effective at increasing the similarity between
maternal and offspring conditions (i.e. environments with little spatial environmental
structure). This hypothesis predicts a negative correlation between population average
percent of CL seeds and the effectiveness with which a population can increase the
average similarity between maternal and offspring conditions by shifting allocation
between CH and CL seeds (8,00%CH - SWACL). I calculated the extent to which changing

allocation to CL seeds increases similarity between maternal and offspring conditions

73

(8,00%CH - Slow/a) for each population under simulated conditions of allocation to CL

seeds (low, 4.6%; moderate, 16.7; and high, 25.9%).

Interpretation testing

Using the similarity between maternal and offspring conditions as an estimate of
the amount of spatial environmental variation encountered by dispersing seeds assumes
that the parent location is a good microhabitat relative to random locations in a site. This
may not be a bad assumption for D. spicata, which grows in low productivity habitats
where there is often substantial open ground. If a maternal location were better for D.
spicata performance than a location chosen at random in a site, then decreasing the
variance between offspring environment and maternal environment would be analogous
to increasing the rate at which seeds encounter good environments. Schoen and Lloyd
(1984) speciﬁcally predicted that a difference in the rates at which CH and CL seeds
encounter good environments would determine the proportional allocation to the two seed
types (q/(l-q)) that would be favored:

[(e2-e1) x (1 --1—)l
i=3 x 2N2 [4]

1" q 32 {e1 x[1—-271\I—]-[e2x(1--13;;j

 

In equation [4], modiﬁed from Schoen and Lloyd (1984), I have simpliﬁed the
terms el and e2 to the rates at which CL and CH seeds, respectively, encounter good
environments or 'safe sites' by assuming that factors other than the environment (e. g.
differential provisioning of the two seed types, differential maternal effects) did not differ
among populations. N1 and N2 are the number of other maternal plants within the
dispersal range of CL and CH seeds, respectively. In Schoen and Lloyd’s (1984)
equation, the number of plants reﬂects the number of other plants producing seeds with
which seeds of the target plant must compete for good sites. I used plant density

measures from surveys of each population (see Figure 3.2c) and average CL and CH

74

dispersal distances from my dispersal simulations (Figure 3.4) to estimate N, and N2. N1

= average plant density x 1! x (average CL dispersal distance)’. N2 = average plant

density x 1! x (average CH dispersal distance)2.

The proportional resource investment in CL seeds was represented by q. The
proportional resource investment in CH seeds was represented by 1 - q. Although the
amount of resource needed for D. spicata to produce a CL seed was not necessarily the
same as amount needed to produce a CH seed (i.e. equal resource investment may not
produce equal numbers of CH and CL seeds), I assumed that CL seeds required the same
resource investment, relative to CH seeds, in all ﬁve of my populations. Because I did
not measure actual investment in the two seed types, this assumption allowed me to use
the number of CL and CH seeds as an estimate of differences in allocation among
populations. I solved equation [4] for e1 in terms of e2 and calculated e1 / e2, the ratio of
the rate at which CL seeds encounter good environments to the rate at which CH seeds
encounter good environments.

I used this ratio as an estimate of the difference in the amount of environmental
variation encountered by CL and CH seeds that would be needed to explain the
differences in proportional allocation to CH (q) and CL (l-q) seeds that I measured
among my populations. If increased allocation to CL seeds increased the rate at which
seedlings encountered good environments, then the ratio of environmental variance
encountered by CL seeds to variance encountered by CH seeds (SIOO%CL / SIOO%CH)
would be negatively correlated with Schoen and Lloyd's ratio of CL to CH rates of

encounter with good environments (e1 / e2).

RESULTS
Population and site characterizations

The eight populations of Danthonia spicata that I surveyed for this study had
different percentages of CL seeds (Figure 3.2a) but had little difference in their

75

 

U: 0\ \l
O O O
I l l 4

A
O
l

N

O
l .
"h

Dispersal distance (cm)
8
l

—n
O
l
N
h)

 

 

O
l

Population

Figure 3.4: Simulated swd dispersal distances

Simulated dispersal distances (mean i 1 s.e.) of cleistogamous (black bars) and
chasmogamous (gray bars) seeds. Letters above cleistogamous dispersal bars
indicate signiﬁcant differences among populations for simulated CL seed dispersal.
Dispersal distances of CH seeds did not differ signiﬁcantly amongpopulations.

76

reproductive stalk lengths (Figure 3.2b). In the ﬁve populations that were resampled for
percent allocation to CL seeds, few of the plants that were sampled in 1996 reﬂowered in
1997 (6-18 of 32 plants in each population). Therefore I had limited power to assess the
consistency of percent allocation to CL seeds by individual plants. However, both
Population and 1996 %CL were signiﬁcant predictors of 1997 %CL in an ANOVA (each
p < 0.001), suggesting that the ﬁve populations differed consistently in their percent

allocation to CL seeds.

Measuring dispersal distance

The dispersal distances of CL and CH seeds in the dispersal experiment were
markedly different; on average CH seeds dispersed ~4.S times farther than CL seeds (CL
distance = 12.25 i 0.74 cm; CH distance = 56.58 i 0.39 cm). However, the magnitude of
the difference in dispersal distances of the two seed types varied across replicate plants
(Seed Type x Replicate, p < 0.001). Much of this variation was attributable to differences
in reproductive stalk length and the distribution of CL seeds among nodes. Across the
three replicate plants, average reproductive stalk length was a good predictor of average
CH dispersal distance (Figure 3.5a). Although stalk length was a good predictor of
average CH seed dispersal distance, there was substantial variance about the mean in the
distance traveled by individual CH seeds (Figure 3.2b). Because of the variance about
the mean CH dispersal distance populations did not differ in simulated CH dispersal
distance (Figure 3.5a).

CL seed dispersal distance was strongly correlated with the node number at which
CL seeds occurred (Figure 3.5b). Taller reproductive stalks had more nodes at which CL
seeds could be produced, so variation in stalk length was included in node number and
was not an independent variable. Although the plants used in the dispersal experiment

produced most of their CL seeds at lower leaf nodes, plants in other populations had

77

 

9)
O\
O

1_

M
(It
1

1—

Average dispersal distance (cm)
& lJt
Ur O
l l

A
O

p=0.023, R2=O.999

l ' l T l ' l '

40 45 50 55 60
Average stalk length (cm)

 

 

 

DJ
M

 

9‘
ON
C

 

Node dispersal distance (cm)

 

 

 

 

p<0.001, R2=O.756
0 I ﬂ I I I

O 1 2 3 4 5 6
Node number

Figure 3.5: Dispersal of CH and CL seeds

Relationship between CH seed dispersal distance and reproductive stalk length (N =
3 plants, each mean i 1 s.e.; a.) and the relationship between CL seed dispersal
distance (mean i l s.e.) and node number at which the seeds were produced (N = 12
to 65 reproductive stalks per node; b.). P-values indicate the signiﬁcance of each

regression and R2 values indicate regression ﬁt.

78

many CL seeds at leaf nodes higher up on the reproductive stalk. Thus I used node

number in the dispersal simulation rather than actual CL dispersal distance measures.

Dispersal simulation

There was no difference among the eight populations in the simulated dispersal
distance of CH seeds (Figure 3.4). The distribution of individual CH seed dispersal
distances about the mean distance (measured in the dispersal experiment) was so large
that a 95% conﬁdence interval (: 22 cm) for the average dispersal distance of any of the
eight populations encompassed the full range of average CH dispersal distances.
However, the simulated average dispersal distances of CL seeds were signiﬁcantly
different among populations (Figure 3.4). Differences among populations in CL seed
dispersal distance were largely explained by percent allocation to CL seeds. Increases in
the percent of CL seeds occurred predominantly by increasing the number of CL seeds at
the farther-dispersing, upper leaf nodes. Consequently, populations that had a larger
proportion of CL seeds also dispersed their CL seeds further (Figure 3.6).

Combining dispersal and environmental variation

Prediction 1: Increasing allocation to CL seeds signiﬁcantly decreased the
variation from maternal conditions encountered by dispersing seeds (SCL + SCH) across
the ﬁve populations (Figure 3.7). This result indicates that there is the potential for
increased allocation to cleistogamous seeds to reduce environmental variation

encountered by dispersing seeds in these populations.

79

 

H H F—i
N b O
0 L 1

u—s
O
l

Average CL dispersal
distance (cm)
°f’

 

 

 

 

O I I T I I I I I I I T I I I I r Ff I I I I T T T ﬁT I I

0 5 10 15 20 25 30

Percent Cleistogamous Seeds

Figure 3.6: Relationship between CL dispersal distance and percent of CL seeds

P-value and R2 value indicate signiﬁcance and ﬁt of the regression of simulated
averag CL dispersal distance for each population on the population average

percent of CL seeds. N = 8 populations.

80

 

 

 

 

 

1.3 ;
O
1.2-
, I
O
I O
0’ ’ '
. p=0.045
E 0.90 I .
V) . I
0.8-
I
I I
0.7-
0.6 I
Low Moderate High

Simulated percent of cleistogamous seeds

Figure 3.7: Effect of CL seed allocation on variation encountered by seeds

The relationship between average variance from p arental conditions encountered by
dispersing seﬂs and simulated proportional allocation to CL seeds. Variance from
maternal conditions was estimated for each of ﬁve populations based on measured

variation in soil moisture (O) and ammonia supply (0). P-value denotes signiﬁcance
of the regression.

81

Prediction 2: In four of these ﬁve populations (L,, L2, D, R), increased allocation to CL
seeds was associated with increased spatial environmental structure. In sites with more
spatially structured variation in soil resources, CL seeds were effective at increasing the
similarity between parent and offspring environments (high Scu'ScJ- In contrast, lower
allocation to CL seeds was associated with sites that had little spatial environmental
structure, in which increased allocation would be ineffective at increasing the similarity
between parent and offspring environments (low SCH-Sm). The positive correlation
between SCH-SCL and percent of CL seeds was signiﬁcant for high and moderate, but not

low, allocation to CL seeds in these four populations (Figure 3.8). Population A2,

differed substantially from all other populations in the relationship between percent of CL
seeds and SCH-SCL. Soil resources in Population A2 were strongly spatially structured
(high SCH-SCI), but the population had very few CL seeds. This may be due to the effect
that a very short growing season has on the allocation to later-manning CL seeds in this

site.

Interpretation testing

In four of ﬁve populations (L1, L2, D, R) there was a signiﬁcant negative
correlation between my estimate of similarity to maternal conditions at CL and CH
dispersal distances (3100%CI/3100%CH) and Schoen and Lloyd's (1984) predicted rates of
encountering good sites by CL and CH seeds (el/e2). This result indicates that variation
(i.e. average semivariance) from maternal conditions was an accurate indicator of
variation from good environmental conditions. The negative correlation between my
index (3100%CL/ 3100%CH) and Schoen and Lloyd's index (el/e2) was signiﬁcant for
both soil moisture and relative ammonia supply at high and moderate, but not low,

allocation to CL seeds among these four populations (Figure 3.9).

82

 

p
A
O

p = 0.008

.o
w
u:

1

p = 0.003

.o
b)
lllll
FDIO

ix
l

0.05%

 

 

 

 

o..:....01,....
0 5 10 20 25 30
Population % of CL seeds

Figure 3.8: Relationship between population %CL and environmental structure
Correlation between spatial environmental structure in each ﬁeld site and the actual
percent of cleistogamous seeds in each population. Spatial environmental structure
(81000/0 CH - SlOO%CL) refers to the extent to which changing allocation to CL
seeds (changing dispersal distance) changes the variation in soil moisture (solid
symbols) and ammonia supply (open symbols), relative to maternal environmental
conditions, encountered by dispersing chasmogamous (CH) and cleistogamous (CL)
seeds. In environments with substantial spatial structure the difference between CH
and CL conditions is large. In unstructured environments the difference is small.
Spatial structure (as encountered by dispersing seeds) was simulated in each
population for low (solid line, I), moderate (dotted line, 0), and high (dashed line,
A) percent allocation to CL seeds. As %CL increases, average CL dispersal distance
approaches CH dispersal distance, so the difference between the two seed
environments decreases. Gray symbols indicate measurements for Population A2,

which were excluded from regressions. P-values to the right of the graph indicate
signiﬁcance of each regression of effectivenss on population % CL.

83

 

1.3

1.25 ‘

N 1.2‘
0
05—1
1.15 ‘
1.1"
l p <0.001
p =0.029
1.05 '

p = 0.075

 

 

 

 

0.4 0.5 0.6 0.7 0.8 0.9 1 1.1

S100%CL / S100%CH

Figure 3.9: Comp arison to Schoen and Lloyd's rates of encounter with grod sites
Relationship between the ratio of variance from parental conditions encountered
by cleistogamous (81000/0 CI) and chasmogamous (SIOO%CH) seeds with Schoen
and Lloyd's (1984) ratio of the predicted rate of encounter with g>od conditions
by cleistogamous (el) and chasmogamous (e2) seeds for a low (solid line, I),

moderate (dotted line, 0), and high (dashed line, A) percent of simulated allocation
to CL seeds. Soil moisture is indicated by solid symbols and ammonia supply is

indicated by open symbols. P-values indicate signiﬁcance of the regression of
e1/e2 on SlOO%Cl./SIOO%CH' The signiﬁcance of this relationship tests whether

it is reasonable to interpret similarity to maternal conditions as similarity to good
environmental conditions. Gray symbols indicate values for Population A2,

which were not included in regressions.

84

DISCUSSION

Spatial environmental variation has often been proposed to inﬂuence gene flow
evolution (e.g. Spieth 1979, Gillespie 1981, Hedrick 1986, Venable and Brown 1988),
but has rarely been examined in natural systems (but see Venable and Brown 1993). In
this study, I found that the amount of spatial environmental structure in soil resources was
correlated with differences in population gene ﬂow distance among populations of D.
spicata.

In four of the ﬁve populations of D. spicata that I surveyed, the percent allocation
to CL seeds was higher in populations where the environment close to parent locations
was more similar to maternal conditions. In populations with very little spatially
structured environmental variation, increasing the proportion of short-dispersing (CL)
seeds would not be very effective increasing the similarity between offspring and
maternal conditions. These patterns are consistent with those predicted of selection by
spatially structured environmental variation on allocation to CL seeds, where allocation
to CL seeds serves to increase the number of dispersing seeds that encounter locations
similar to their mother’s.

In one population that I surveyed (A2), the percent allocation to CL seeds was
very different than what I would predict based on the pattern of spatial environmental
variation. Soil moisture was strongly spatially structured in Population A.,, which meant
that shorter dispersing seeds would effectively increase the similarity between offspring
and maternal conditions, but plants in this population produced very few CL ﬂowers (and
hence few CL seeds). CL seeds in D. spicata are fertilized before, but matured after, CH
seeds, and so seasonal truncation of the growing season by early season drought could
limit the ability of D. spicata in to mature CL seeds. Danthonia spicata commonly

occurs in dry environments, but grows and matures seed during the spring, when moisture

85

is available (a second growth period occurs in late fall). Population A2 was the most
chronically dry population studied and plants in this population were often unable to
complete seed maturation in the spring, before the soil became too dry and caused the
plants to senesce (McCormick, personal observation). In the long term, energy allocated
to the development of CL seeds would be wasted and thus be disadvantageous. This may
explain the limited CL production by plants in Population A2. Although the other four
populations I studied were all quite dry, they were not subject to the early season
droughts that occurred predictably in Population A2. In these four populations, percent of
CL seeds did not correlate with soil moisture.

In three of the study populations (L,, L2, R) some D. spicata plants were infected
by the epiphytic fungus Atkinsonella hypoxylon. Because A. hypoxylon causes abortion
of CH seeds (Clay 1984), infected plants do not have the potential to alter their
proportional allocation to CL seeds (all seeds produced are CL). However, while this
fungal infection could retard a population's ability to respond to selection for altered
dispersal distance, it does not appear to act as a direct selective force on allocation to CL
seeds in these populations. Plants with different patterns of allocation to CL seeds do not
differ in the number of seeds produced when infected (McCormick, unpublished data).

Quantifying spatial environmental variation at scales that are perceived by plants
and dispersing seeds is notoriously difﬁcult. The correlation between measured and
perceived variation is generally unknown and even measured variation cannot be
measured at all points in a population. McCormick and Gross (Chapter 2) found that
differences in soil moisture and nitrogen supply rate of the magnitude measured in these
ﬁve ﬁeld populations can affect the performance of D. spicata plants in the ﬁeld. These
factors are also likely to be major factors limiting grth in natural populations of D.
spicata, so it is reasonable to interpret measured differences in soil moisture and nitrogen

supply rate as impacting plant growth and ﬁtness.

86

I used semivariance analysis to obtain estimates of the similarity of pairs of points
based on their distance apart from relatively few sample points in each population.
Although interpreting the similarity of offspring to maternal conditions as an estimate of
the quality of environments encountered by dispersing offspring is unconventional, it
may be reasonable in environments that are largely unsuitable for plant grth (Levin et
al. 1984). This interpretation of predicted similarity between offspring and maternal
conditions was consistent with independent estimates of relative environmental quality
that would allow the observed patterns of allocation to two differently dispersing seed
types to be an evolutionarily stable strategy for D. spicata (Schoen and Lloyd 1984).
Other studies of local adaptation have also found that offspring performance decreases
with increasing distance from the maternal location (e. g. Schmitt and Gamble 1990,
Miller and Fowler 1995).

Spatial environmental heterogeneity may have particularly strong selective effects
on species like D. spicata, which produces heteromorphic seeds with different dispersal
potentials. Species with heteromorphic seeds may be especially powerﬁil tools for
detecting the effects of spatially structured environmental variation on dispersal distance
(V enable 1984). Schoen and Lloyd (1984) argued that spatially structured environmental
variation can be an evolutionary force favoring differently-dispersing dimorphic seeds.

However, many plants have genetic variation for seed dispersal distance that is
not related to seed heteromorphism. Venable and Brown (1988) suggested that spatially
structured environmental variation may select more generally on dispersal distance
polymorphisms, of which dimorphic seeds are only one example. In plants with a
continuous dispersal distribution, selection for increased dispersal distance by density-
dependent effects, such as competition with siblings and parents, may be stronger than in
a bimodal disperser. Thus continuous dispersers may have longer mean dispersal
distances than bimodal dispersers in similar environments. However, in environments

where decreasing dispersal distance could substantially increase the probability of

87

offspring encountering good environments (such as Populations D and R studied here), I
would predict that spatial environmental variation would also select for decreased
dispersal distance in species with unimodal dispersal, forming a balance with the density
dependent forces selecting for increased dispersal distance.

In this study, I found that the pattern of allocation to short-dispersing CL seeds
among populations of D. spicata was consistent with predictions based on the degree to
which variation in soil moisture and ammonia in those populations was spatially
autocorrelated. Speciﬁcally, allocation to CL seeds was higher in populations where
short dispersal would be most effective at reducing spatial environmental variation
encountered by dispersing seeds. Because gene ﬂow distance in D. spicata is largely
determined by seed dispersal, these measured differences in seed dispersal distance also
reﬂect differences in gene ﬂow distance. Even though the pattern of variation in soil
moisture and ammonia within these sites was relatively consistent over years, they were
still seasonally variable. Although temporal environmental variation may have strong
effects on the evolution of many traits (e. g. Sultan and Bazzaz 1993, Stratton and
Bennington 1998), these results suggest that a trait that explicitly affects the spatial
distribution of plants or their offspring (e. g. seed dispersal) might be responsive to
spatially structured environmental conditions, despite substantial temporal variation. If
traits like gene ﬂow distance can respond evolutionarily to spatial environmental
heterogeneity despite temporal variability, then the effect of spatial environmental
variation in maintaining genetic variation and structuring interactions within and among

populations may be greater than previously thought.

88

Literature Cited

Balkau, B. J ., and M. W. F eldman. 197 3. Selection for migration modiﬁcation.
Genetics, 74: 171-174.

Christiansen, F. B., and M. W. F eldman. 1975. Subdivided populations: A review of the
one- and two-locus deterministic theory. Theoretical Population Biology, 7: 13-
38.

Clay, K. 1982. Environmental and genetic determinants of cleistogamy in a natural
population of the grass Danthonia spicata. Evolution 36: 734-741.

Clay, K. 1983. The differential establishment of seedlings from chasmogamous and
cleistogamous ﬂowers in natural populations of the grass Danthonia spicata (L.)
Beauv. Oecologia, 57: 183-188.

Clay K (1984) The effect of the fungus Atkinsonella hypoxylon (Clavicipitaceae) on the
reproductive system and demography of the grass Danthonia spicata. New
Phytologist 98:165-175.

Darbyshire, S. J ., and J. Cayouette. (1989). The biology of Canadian weeds. 92.
Danthonia spicata (L.) Beauv. in Roem. and Schult. Canadian Journal of Plant
Science 69: 1217-1233.

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.

Gillespie, J. H. 1981. The role of migration in the genetic structure of populations in
temporally and spatially varying environments. III. Migration modiﬁcation. The
American Naturalist, 117: 223-233.

Lajthe, K. 1988. The use of ion-exchange resin bags for measuring nutrient availability
in an arid ecosystem. Plant and Soil, 105:105-111.

Lloyd, D. G. 1984. Variation strategies of plants in heterogeneous environments.
Biological Journal of the Linnean Society, 21: 357-385.

Okubo, A., and S. A. Levin. 1989. A theoretical ﬁamework for the analysis of data on
wind dispersal of seeds and pollen. Ecology 71: 329-338.

Price, M.W., and NM. Waser. 1979. Pollen dispersal and optimal outbreeding in
Delphim'um nelsoni. Nature, 277:294-298.

Roberston, J .P., and KL. Gross. 1994. Assessing the heterogeneity of belowground
resources: quantifying pattern and scale. In: Caldwell, M.M. and R.W. Pearcy,
eds. Exploitation of environmental heterogeneity by plants: ecophysiological
processes above- and belowground. Academic Press, San Diego, CA. pp 237-
252.

Rossi, R.E., D.J. Mulla, A.F. Journal, and EH. Franz. 1992. Geostatistical tools for

modeling and interpreting ecological spatial dependence. Ecological
Monographs, 62: 277-314.

89

Schmitt, J., and S. E. Gamble. 1990. The effect of distance from the parental site on
offspring performance and inbreeding depression in Impatiens capensis: a test of
the local adaptation hypothesis.

Schoen, D. J ., and D. G. Lloyd. 1984. The selection of cleistogamy and heteromorphic
diaspores. Biological Journal of the Linnean Society, 23: 303-322.

Spieth, P. T. 1979. Environmental heterogeneity: a problem of contradictory selection
pressures, gene ﬂow, and local polymorphism. The American Naturalist, 113:
247-260.

Stratton, D. A., and C. C. Bennington. 1998. Fine-grained spatial and temporal variation
in selection does not maintain genetic variation in Erigeron anuus. Evolution, 52:
678-691.

Thiede, D. A., and C. K. Augspurger. 1996. Intraspeciﬁc variation in seed dispersion of
Lepidium campestre (Brassicaceae). American Journal of Botany, 83: 856-866.

Tonsor, S. J. 1990. Spatial patterns of differentiation for gene ﬂow in Plantago
lanceolata. Evolution 44: 1373-1378.

Venable, D. L. 1984. Using intraspeciﬁc variation to study the ecological signiﬁcance
and evolution of plant life-histories. In: Dirzo, R., and J. Sarukhan, eds.
Perspectives on Plant Population Ecology. Sinauer Associates, Inc., Sunderland,
MA.

Venable, D. L., and J. S. Brown. 1988. The selective interactions of dispersal,
dormancy, and seed size as adaptations for reducing risk in variable environments.
The American Naturalist, 131: 360-384.

Waddington, K. D. 1983. Pollen ﬂow and optimal outcrossing distance. The American
Naturalist, 122: 147-151.

90

Chapter 4

Plasticity for nitrogen availability in Danthonia spicata: phenotypic plasticity in
response to spatial environmental heterogeneity?
INTRODUCTION

Spatial environmental heterogeneity is well known to affect plants ecologically
(e. g. Grime 1994, Fitter 1994, Chapter 5), but it has received less attention for its ability
to affect them evolutionarily. Phenotypic plasticity has been primarily considered an
evolutionary response to temporal variation (e. g. Sultan 1987 and refs. therein), while
local adaptation has been considered the predominant response to spatial variation (e. g.
Bell, Lechowicz and Schoen 1991, Bell and Lechowicz 1994). In addition to being the
only mechanism by which sessile organisms, such as plants, can respond to temporal
environmental heterogeneity that occurs within generations (Sultan 1987, Schlichting and
Levin 1990, Schlichting 1993), adaptive phenotypic plasticity, or appropriate phenotypic
changes in speciﬁc traits in response to particular environmental conditions (Sultan
1987), may also be important means by which plants can respond to spatial
environmental variation encountered among generations. However, few studies have
looked for an explicit relationship between amount of spatial enviromnental variation and
amount of phenotypic plasticity (but see Miller and Fowler 1994).

Prevailing wisdom suggests that temporal variation in environmental conditions
often greatly exceeds spatial variation and thus should dominate selection on phenotypic
plasticity (e.g. Scheiner and Teeri 1986, Sultan and Bazzaz 1993). However, if the range
of conditions experienced within a generation differs consistently among locations, then
that range of conditions will select on a trait’s norm of reaction (Galloway 1995). For
example, one location might be consistently drier than another, but wet conditions at a
dry location could overlap dry conditions at a wet location. Plants in dry locations would

experience conditions that were much drier than those in wet locations. Plants in both

91

locations would need to be able to withstand temporal ﬂuctuation in moisture levels, but
plants in dry locations would experience greater selection for ability to withstand long-
terrn dry conditions than plants in wet locations.

Local adaptation to spatial environmental variation (e. g. Bell et al. 1991, Schmitt
and Gamble 1990) and differences in phenotypic plasticity (e. g. MacDonald and
Chinnappa 1989, Schlichting and Levin 1990, Scheiner 1993) have repeatedly been
demonstrated, but the two have rarely been explicitly examined together (but see Miller
and Fowler 1994). I used a greenhouse experiment to test for differences in
phenotypically plastic responses to nitrogen availability among Danthonia spicata
seedlings from near- and far-dispersing families from ﬁve populations in Michigan.
Because the ﬁve sites had different mean nitrogen availabilities (Chapter 2), populations
from different sites might have different mean nitrogen acquisition traits. However,
because differences in population mean dispersal distance can equalize the amount of
spatial environmental variation encountered, plants from different populations would not
necessarily encounter different amounts of spatial nitrogen variation (Chapter 3).

Within these populations, offspring from far-dispersing families encounter more
spatial variation in nitrogen availability (Chapter 3). Thus, I predicted that offspring from
far-dispersing families would have greater phenotypic plasticity than those from near-
dispersing families in response to varied nitrogen supply in all ﬁve source populations.

1 measured plasticity in three different modes of response to nitrogen levels,
changes in grth rate (biomass accumulation), shifts in allocation (rootzshoot), and
physiological changes in nitrogen uptake (nitrogen concentration). In the low nutrient,
nitrogen limited, habitats where D. spicata grows, nitrogen acquisition is likely to be an
important determinant of plant ﬁtness. If differences in spatial nitrogen variation selected
for differences in amount of phenotypic plasticity, then the plasticity of these three traits

should differ among near- and far-dispersing families.

92

MATERIALS AND METHODS
Study System

Danthonia spicata (Poaceae) is a native, perennial, C3 bunch grass that

commonly occurs in dry, nutrient-poor oak savannas, old-ﬁelds, and openings throughout
the northern and eastern United States and southern Canada (Clay 1982). Danthonia
spicata produces two seed types with different dispersal potential. Obligately self-
fertilized cleistogamous (CL) seeds remain within leaf sheaths and are dispersed near the
parent plant (avg. 10 cm; Chapter 3). Potentially outcrossed, wind-pollinated
chasmogamous (CH) ﬂowers are produced at the tip of the reproductive stalk and
disperse farther from the maternal parent (avg. 60 cm; Chapter 3). Chasmogamous seeds
may also have secondary dispersal that further increases their average dispersal distance
(Clay 1983a). Although these seed types differ in potential for outcrossing, they differ
little in genetic variation and inbreeding depression is not apparent (Clay 1983b, Clay and
Antonovics 1985).

The proportion of seeds that are allocated to cleistogamy in D. spicata is variable
both within and among populations. Clay (1982) found that there is a strong genetic
basis for the proportion of cleistogamous seeds (heritability 52.6 %), producing genetic
variation for dispersal distance. In habitats where there is substantial temporally
consistent spatially structured environmental variation, families that have different
proportions of short-dispersing CL seeds differ in how much spatial variation is
encountered by dispersing offspring. Families that disperse all of their seeds far from the
maternal plant (high CH) would encounter more spatial environmental variation than
families that disperse many of their seeds close to the maternal plant (high CL; Chapter
3).

The ﬁve sites I selected for these studies had D. spicata populations that differed
in their average proportion of short-dispersing, CL seeds (Figure 4.1; Chapter 3). Within

the sites where these populations occurred there was spatially structured variation in

93

 

a- 50

4e
3;
,Q
‘9 l
0- r 1
A L1 L2 D R

Population

Percent cleistogamy
N (a) -b
O O O
J l l

.3
O
l

 

P'
or

 

Nitrogen (pg/g resin)

 

 

 

Figure 4.1: Characteristics of ﬁve study sites and populations
a. Average percent of cleistogamous seeds (: l s.e.) produced in each population (A2,

L

Diamonds (6) indicate highest and lowest allocation to cldstomom seeds measured
in each population. Soil nitrogen was not measured in Population D, because soils

were too shallow. N = 32 plants for %CL; N = 58 to 68 for nitrogen (see Chapter 2).

1, L2, D, R) and b. average nitrogen supply (nitrate + ammonia nitrogen; : 1 s.e.).

94

nitrate and ammonia supply rates (Chapter 2), producing a correlation between parent and
offspring conditions that depended on dispersal distance (Chapter 3). Danthonia spicata
plants have a median lifespan of 2 to 2.5 years (Scheiner 1987) and spatial environmental
variation at this time scale was temporally consistent in these sites (Chapter 2).
Allocation to CL seeds among these populations differed in a way that suggested
population average seed dispersal distance was adapted to the scale of variation in soil
moisture and ammonia supply rates (Chapter 3). Study sites differed in both the mean
availability of nitrogen and the degree to which it was spatially variable (Chapter 2).
Differences in population average dispersal distance somewhat equalized the amount of
nitrogen variation encountered by dispersing seeds among populations (Chapter 3), but
variation in dispersal distance among families within populations produced differences in
the amount of spatial variation in nitrogen their seeds would encounter. Dispersal
distance differences among populations that are concordant with selection by the
distribution of spatial environmental variation suggest that differences in spatial
environmental variation can select on

population characteristics. I hypothesized that spatial variation in nitrogen supply rates

could also select on the amount of phenotypic plasticity in these populations.

Greenhouse experiment

In July 1997, I collected CH seeds ﬁom four maternal parents that differed in their
percent allocation to short dispersing CL seeds, two with many CL seeds (near-dispersing
families) and two with few (far-dispersing families), in each of these ﬁve populations
(Figure 4.2). The seeds were stored at room temperature for ﬁve months and then treated
with 70% sulfuric acid for 35 minutes, rinsed and placed on moist, sterile sand in petri
dishes under ambient light in the greenhouse. Seeds began to germinate after seven days.
Two families from Population A2 and one from Population R had no seed germination

(Figure 4.2). To keep sample sizes among dispersal types as balanced as possible, I

95

 

 

 

 

50
o
40—
E
330- O... z
0
8 20‘ * e
o i 0*
a? .. ’
10— <>+
- * * .
*+ 3* 3* o
O A? r L ' L ' I)ﬂ R
2 1 2
Population

Figure 4.2: Percent cleistogamy of Danthonia spicata families

Percent allocation to short dispersing (cleistogmous) seeds by families used as high
(0) and low (0) CL families from each population. Asterisks (*) designate families
that also received medium (80 ppm) nitrogen fertilizer. Plus signs (+) designate
families with additional seeds used to replace families with no germination.

96

increased the number of seedlings in the experiment ﬁ'om the surviving family in the
dispersal type. Aﬁer one month of growth, I transplanted eight to ten seedlings from each
family (14 - 18 from increased families) into 6 x 6 x 20 cm paper pots ﬁlled with silica
sand. Pots were placed in 6 baskets of 36 pots each. Each seedling was randomly
assigned a basket and pot location within the basket, with two seedlings from each family
in each of four or ﬁve of the six baskets. I fertilized all seedlings with a moderately high
nitrogen (90 ppm N) fertilizer solution for two weeks after transplanting to encourage
establishment.

After the two week establishment period, I washed water through each pot to rinse
remaining fertilizer out of the sand and established the nitrogen level treatments. Within
each block, I randomly chose one of each pair of seedlings per family to receive high
(150 ppm N) or low (30 ppm N) nitrogen fertilizer (modiﬁed Hoagland's solution;
Appendix A. 1). All seedlings received the same background levels of other nutrients.
Fertilizer was applied to each individual seedling at a rate of 10 ml per day, three times
per week. I chose these nitrogen levels to approximate the minimum and maximum of
nitrogen supply rates measured in these sites (Chapter 2).

Because of uncertainty of the shape of the reaction norms of response to nitrogen
level, I established an additional treatment with a total of 20 seedlings from three
populations (A2, L,, L2) that were fertilized with an intermediate level of nitrogen (80
ppm N). This third treatment allowed me to determine whether the reaction norms of
these nitrogen access traits (biomass accumulation, rootzshoot ratio, and shoot percent
nitrogen) to increasing nitrogen were non-linear, and whether my high nitrogen treatment

was too high relative to other nutrients in the fertilizer solution. All seedlings were

grown for four months at 18 to 30°C with 12 hours of supplemental light (762 mmol m2
s‘1 PAR).
At the end of the experiment, all plants were harvested, and I measured total

biomass (indicative of growth rate), rootzshoot ratio, and shoot nitrogen concentration of

97

each plant. I harvested plants by peeling away the paper pot and washing sand from the

roots with water. I separated the root and shoot tissue, dried them for 72 hours at 38°C,

and weighed them. A subset of the dried green shoot tissue from each plant was then

clipped very ﬁnely and analyzed for tissue nitrogen concentration (g N g plant tissue‘l)
using an elemental analyzer (Nitrogen Analyser 1500 Series 2, 1990, Carlo-Erba
Instruments, Milan, Italy).

I examined differences in phenotypic plasticity (biomass, root:shoot ratio, shoot
nitrogen concentration) in response to nitrogen treatment (high vs. low) among
populations and differently dispersing families using an ANOVA for each trait (Systat 8.0
for Windows, 1997. SPSS, Chicago Il.). Biomass and shoot nitrogen concentration were
log-transformed to improve normality. There was a signiﬁcant negative correlation
between log biomass and log shoot nitrogen concentration (-0.198; p < 0.028), but all
other pairs of traits were not correlated. Nitrogen treatment, Population, and Dispersal
type (near or far) were tested as main effects. The presence of plasticity for each trait was
tested by the signiﬁcance of the nitrogen treatment main effect. My hypothesis
speciﬁcally predicted that there would be differences among populations and differently
dispersing families in the degree of plasticity in response to nitrogen treatment. I tested
this hypothesis by examining the signiﬁcance of the Population x Nitrogen treatment and
Dispersal type x Nitrogen treatment interactions, respectively. Both biomass and
nitrogen concentration were log-transfonned to improve normality for these tests.

Plants in three of the baskets became infected with an unknown ﬁrngus, differed
signiﬁcantly in all traits measured from those that were not infected. The unidentiﬁed
fungus invaded over 90% of the pots in these three baskets, but less than 5% of pots in
the other three baskets. Because there were no signiﬁcant differences in plant traits
among baskets within the infected or uninfected baskets, I grouped the six baskets into

two blocks, deﬁned by presence or absence of fungus. To account for effects of the

98

fungal infestation in my analyses, I tested for Block main effects and interactions in the
AN OVA for each trait.

I visually examined the linearity of plastic responses to three levels of nitrogen
treatment in the three populations that received all three nitrogen treatments. Substantial
non-linearity of reaction norms would necessitate caution in interpreting differences in
responses among populations in plasticity to nitrogen level. These three populations (A2,
L,, L2) and differently dispersing families were not signiﬁcantly different in any of the

traits examines and were combined for this analysis.

RESULTS

Plants from different populations had similar mean biomass and shoot nitrogen
concentration, but signiﬁcantly different root:shoot ratios (Table 4.1). Biomass and shoot
nitrogen concentration were phenotypically plastic in response to nitrogen availability,
but root:shoot ratio was not (Table 4.1). Plants from all ﬁve populations had similar
amounts of plasticity in the three traits I examined. Among—population differences in
mean root:shoot ratio and a lack of plasticity in this trait suggest that populations differed
genetically in root:shoot ratio, although difference resulting from maternal effects cannot
be ruled out. Rootzshoot ratio may respond to mean nitrogen availability, which differed
among the sites, rather than to its spatial variation.

Not all plants produced sufﬁcient leaf tissue to analyze for shoot nitrogen
concentration (3 to 5 mg). Some combinations of Block x Population x Dispersal type
had no plants large enough for nitrogen analysis. The Population main effect and
interactions were substantially non-signiﬁcant when analyzed in an ANOVA without
Dispersal type (p 2 0.391). Because the test of my hypothesis speciﬁcally involved tests
of the Dispersal type x Nitrogen interaction and because the Block effect was very

substantial, I tested for differences in shoot nitrogen concentration among nitrogen levels,

99

Table 4.1: ANOVA tables presenting variation in biomass, root:shoot and shoot nitrogen
Effects of Population (P), Dispersal class (high or low allocation to CL seeds; DC),
Nitrogen treatment (30 or 150 ppm N; N), and Block (B) and their interactions on
biomass (a.), root:shoot ratio (b.), and shoot nitrogen concentration (c.) in the greenhouse
experiment. Populations used as seed sources were A2, L1, L2, D, and R.

a. Biomass (log transformed)

 

N = 165 Squared multiple R = 0.646

' - tio P
Population 4 1.275 1.703 0.153
Dispersal class 1 0.630 0.842 0.361
Nitrogen 1 61.871 82.673 < 0.001
Block 1 35.132 46.943 < 0.001
P x DC 4 0.398 0.532 0.712
P x N 4 0.634 0.848 0.498
P x B 4 0.799 1.067 0.376
DC x N 1 0.286 0.383 0.537
DC x B 1 0.459 0.613 0.435
N x B 1 2.837 3.790 0.054
P x DC x N 4 2.407 3.216 0.015
P x DC x B 4 0.327 0.437 0.782
P x N x B 4 0.523 0.698 0.595
DCXNXB 1 0.315 0.421 0.518
P X DC x N x B 4 0.394 0.527 0.716
E59: 125 0.748

 

100

Table 4.1 (cont’d)

b. Root : Shoot ratio

 

N = 163 Squared multiple R = 0.285
Warm 1
Population 4 0.392 2.919 0.024
Dispersal class 1 0.202 1.509 0.222
Nitrogen 1 0.121 0.904 0.344
Block 1 0.741 5.524 0.020
P x DC 4 0.085 0.631 0.641
P x N 4 0.071 0.528 0.716
P x B 4 0.186 1.385 0.243
DC x N 1 0.007 0.049 0.824
DC x B 1 0.044 0.326 0.569
N x B 1 0.014 0.105 0.747
P x DC x N 4 0.058 0.431 0.786
P x DC X B 4 0.088 0.659 0.622
P x N x B 4 0.099 0.741 0.565
DC x N x B 1 0.049 0.367 0.546
P x DC x N x B 4 0.058 0.434 0.784
Error: 123 0.134

 

c. Shoot nitrogen concentration (log %)

 

N = 123 Squared multiple R = 0.512
- ' p
Dispersal class 1 0.016 0.867 0.354
Nitrogen 1 0.074 3.985 0.048
Block 1 1.602 86.172 < 0.001
DC x N 1 0.001 0.065 0.799
DC x B 1 < 0.001 0.017 0.898
N x B 1 0.002 0.083 0.774
DC x N x B 1 0.001 0.066 0.797

Error 115 0.019

101

 

3
e

.0
M

 

 

Biomass (log(g))

 

 

 

 

 

 

P
p—a
in
F1.

 

“ 7‘ ‘ p
0.5
1.5 f.

0.75 l l
30 150 30 150
Fertilizer N (ppm) Fertilizer N (ppm)

Rootzshoot

- \
I

I

 

 

5"

[N] (10g %)

 

 

 

 

 

 

Figrre 4.3: Reaction norms across two nitrogen levels

Mean (i 1 s.e.) biomass (a., d.), root:shoot ratio (b., e.), and shoot nitrogen
concentration (c., f.) norms of reaction across two nitrogen levels by families with
high (a.-c.) and low (d.-f.) allocation to cleistogamous seeds. Each symbol represents

adifferent population: A2 =I , Ll =0 , L2 = A, D = O, R = *.

102

 

3- 1.4

 

 

 

 

 

 

 

 

 

 

—Ntrt.p<0.001
12f Block p < 0.001
3:; liR2=0.4ll
«0 0.8—
g -
g 0.6j
'35 0.4:
0.2—
OrI‘ITTYlT'Tr"'lw'rl'r‘lifﬁl"'
b.
2'5, “F Ntrt. p=0.090
2_ Block p =0.065
3 2
_g - E R =0.052
m 15-
8 1‘ i\r
‘34 a
0.5—
0"'l"Tl"‘l'r'lrﬁ'l'"l"‘lf"
1 Ntrt. p<0.001
~Block p=0.010
0.5—R2=o.663e + ”i

 

 

Shoot nitrogen (log %) .0

 

 

 

I
.O
(II

‘1

.1
q

1

30 80 150
Nitrogen treatment (ppm N)

Figure 4.4: Reaction norms of response to nitrogen treatment across three levels
Norms of reaction across three nitrogen levels in a. biomass, b. root:shoot ratio, and

c. shoot nitrogen concentration in the block group uninfected (I) and infected (D) by
an unidentiﬁed fungus.

103

near and far dispersing families, blocks, and the interactions of these effects without
Population as a factor.

Within populations, differently dispersing families had similar total biomass,
root:shoot ratio, and shoot nitrogen concentration (Figure 4.3). Contrary to my
predictions, differently dispersing fanrilies also had similar amounts of phenotypic
plasticity in the three traits I examined (Figure 4.3, Table 4.1). For biomass, the Nitrogen
x Dispersal type x Population interaction was signiﬁcant (p < 0.015), indicating that high
and low CL
seedlings had different amounts of phenotypic plasticity in some populations. However,
seedlings from low CL families did not have consistently greater plasticity than those
ﬁ'om high CL families (Figure 4.3). No other interactions were signiﬁcant for any of the
traits I examined.

Reaction norms for the populations receiving all three nitrogen treatments
appeared linear for both the infected and uninfected blocks (Figure 4.4). This suggests
that the two nitrogen levels that I used to measure phenotypic plasticity were adequate to
describe the reaction norms of biomass, root:shoot ratio, and shoot nitrogen concentration
in response to nitrogen levels over the range from 30 to 150 ppm N, which is comparable

to the range of nitrogen supply in these ﬁeld sites.

DISCUSSION

Studies of phenotypic plasticity in plants have rarely attempted to explicitly
examine plastic responses to spatially variable environmental factors. In this study, I
tested for differences in amount of phenotypic plasticity in response to nitrogen
availability among and within ﬁve populations of Danthonia spicata. These ﬁve study

sites differed in

104

mean availability of nitrogen and also in amount of spatially structured variation in
nitrogen availability (Chapter 2). Consequently, within each population, differently
dispersing families were expected to encounter different amounts of variation in nitrogen
availability (Chapter 3), which I expected would select for differences in plasticity of
nitrogen acquisition traits. Although population differences in average dispersal distance
equalized the average amount of spatial variation encountered by dispersing seeds among
populations, I expected that plants from sites with higher nitrogen levels would respond
differently to variation in nitrogen levels than plants from sites with low nitrogen
availability.

Despite differences in nitrogen availability among ﬁeld sites, I found that D.
spicata plants from different populations responded similarly to greenhouse nitrogen
treatments. Plants from different populations had similar mean biomass and shoot
nitrogen concentration and also had similar amounts of plasticity in these traits. The
similarity in plasticity among populations with very different mean nitrogen availabilities
suggests that biomass and shoot nitrogen concentration in D. spicata may respond to
factors other than spatial nitrogen variation or that other factors, such as temporal
variation, may act to maintain uniformly high levels of plasticity in these traits. Plants
from these ﬁve populations had signiﬁcantly different root:shoot ratios, but did not
exhibit plasticity for this trait.

Within each population, I predicted that plants from far-dispersing families would
be more phenotypically plastic than plants from near-dispersing families. Differently
dispersing families had similar mean biomass across nitrogen treatments, but some
populations had different amounts of biomass plasticity. However, far-dispersing
families were not consistently more plastic than near-dispersing families. Plants from
differently dispersing families had similar shoot nitrogen concentrations, root:shoot

ratios, and plasticity for both of these traits.

105

Substantial block effects in this experiment were most likely due to invasion of
many pots in one block by an unidentiﬁed fungus. Plants in uninfected pots responded
substantially differently to nitrogen treatments than those in pots infected by ﬁrngus.
However, block did not interact signiﬁcantly with any of the design factors.

One possible explanation for the similarity of plasticity in two of the three
nitrogen acquisition traits I examined is a very small or negligible cost of maintaining
plasticity for nitrogen acquisition. Sultan and Bazzaz (1993) proposed that because the
mechanisms contributing to phenotypic plasticity in response to nutrient availability may
allow individual genotypes to perform well in a wide range of soil conditions, there may
never be sufﬁcient costs of maintenance to outweigh the beneﬁts of plasticity. Grime
(1994) suggested that nutrient acquisition in low fertility habitats consistently require an
ability to access nutrients available in short-lived pulses. If this is true, then plants that
typically occur in low fertility habitats, such as D. spicata, may be especially subject to
temporal variation in soil nutrients, which may select for high levels of phenotypic
plasticity. If there were a consistent beneﬁt of plasticity for nutrient access, then
populations would be ﬁxed for uniformly high levels of plasticity. The generality of
plasticity costs and constraints is currently a subject of substantial research, but the
likelihood of very limited costs in nutrient access traits is still unknown (Dewitt 1998).

Selection by temporal variation in environmental conditions has often been
proposed to dominate selection for phenotypic plasticity, equalizing phenotypic plasticity
among even extremely different, genetically isolated populations (e. g. Sultan and Bazzaz
1993, and references therein). Scheiner and Goodnight (1984) found high, but invariable,
amounts of phenotypic plasticity in multiple biomass allocation traits in response to soil
moisture and light availability in ﬁve D. spicata populations of different successional age
in northern Michigan. Scheiner and Teeri (1986) proposed that selection for plasticity in
D. spicata was dominated by short-term temporal, rather than spatial, variation in

moisture and light conditions.

106

However, in these ﬁve ﬁeld sites in Michigan, the spatial patterns of nitrate and
ammonia availability were relatively consistent within the lifetime of D. spicata plants,
despite substantial temporal variation within years in all sites (Chapter 2). Hence in these
sites, selection by nitrogen conditions should be dominated by spatial variation.

Despite different amounts of spatial heterogeneity in nitrogen supply (Chapter 2)
and possible locally adaptive responses in dispersal distance (Chapter 3), I found little
evidence for differences in phenotypically plastic responses to nitrogen supply rate
among differently dispersing families from these ﬁve populations of D. spicata in
Michigan. Similarity among populations of nitrogen acquisition plasticity could not be
fully explained by the comparative strengths of temporal and spatial variability within the
average lifespan of D. spicata, but might be explained by limited costs or consistent

beneﬁts of plasticity for nutrient access among and within these ﬁve populations.

107

LITERATURE CITED

Bell, A. G., and M. Lechowicz. (1994). Spatial heterogeneity at small scales and how
plants respond to it. In: Caldwell, M. M., and R. W. Pearcy, eds. Exploitation of
Environmental Heterogeneity by Plants: Ecophysiological Processes Above- and
Belowground. Academic Press, San Diego, CA. pp. 391-414.

Bell, G., M. J. Lechowicz, and D. J. Schoen. (1991). The ecology and genetics of ﬁtness
in forst plants. III. Environmental variance in natural populations of Impatiens
pallida. Journal of Ecology 79:697-713.

Chapin, F. S. (1980). The mineral nutrition of wild plants. Annual Review of Ecology
and Systematics 11:233-260.

Clay, K. (1982). Environmental and genetic determinants of cleistogamy in a natural
population of the grass Danthonia spicata. Evolution 36:734-741.

Clay, K. (1983a). Variation in the degree of cleistogamy of within and among species of
the grass Danthonia. American Journal of Botany, 70:835-843.

Clay, K. (1983b). The differential establishment of seedlings from chasmogamous and
cleistogamous ﬂowers in natural populations of the grass Danthonia spicata (L.)
Beauv. Oecologia, 57:183-188.

Clay, K., and J. Antonovics. (1985). Quantitative variation of progeny from
chasmogamous and cleistogamous ﬂowers in the grass Danthonia spicata.
Evolution, 39:335-348.

Dewitt, T. J ., A. Sih, and D. S. Wilson. (1998). Costs and limits of phenotypic plasticity. .
Trends in Ecology and Evolution 13:77-81.

Fitter, A. H. (1994). Architecture and biomass allocation as components of the plastic
response of root systems to soil heterogeneity. In: Caldwell, M. M., and R. W.
Pearcy (eds.), Exploitation of environmental heterogeneity by plants:
ecophysiological processes above— and belowground. Academic Press, San Diego,
CA. pp. 305-322

Galloway, L. F. (1995). Response to natural environmental heterogeneity: maternal
effects and selection on life-history characters and plasticities in Mimulus guttatus.
Evolution 49: 1095-1 107.

Grime, J. P. (1994). The role of plasticity in exploiting environmental heterogeneity. In:
Caldwell, M. M., and R. W. Pearcy (eds.), Exploitation of environmental
heterogeneity by plants: ecophysiological processes above- and belowground.
Academic Press, San Diego, CA. pp. 2-16.

MacDonald, S. E. and C. C. Chinnappa. (1989). Population differentiation for
phenotypic plasticity in the Stellaria longipes complex. American Journal of Botany
76: 1627-1637.

Miller, R. E., and N. L. Fowler. (1994). Life-history variation and local adaptation

within 2 populations of Bouteloua-rigidiseta (Texas-gramma). Journal of Ecology
82:855-864.

108

Scheiner, S. M. (1987). Size and fecundity hierarchies in an herbaceous perennial.
Oecologia 74:128-132.

Scheiner, S. M. (1993). Genetics and evolution of phenotypic plasticity. Annual Review
of Ecology and Systematics 24:35-68.

Scheiner, S. M. and C. J. Goodnight. (1984). The comparison of phenotypic plasticity
and genetic variation in populations of the grass Danthonia spicata. Evolution
38:845-855.

Scheiner, S. M. and J. A. Teeri. (1986). Phenotypic ﬂexibility and genetic adaptation
along a gradient of secondary forest succession in the grass Danthonia spicata.
Canadian Journal of Botany 64:739-747.

Schlichting, C. D. (1986). The evolution of phenotypic plasticity in plants. Annual
Review of Ecology and Systematics 17:667-693.

Schlichting, C. D. and D. A. Levin. (1990). Phenotypic plasticity in Phlox. III.
Variation among natural populations of P. drummondii. Journal of Evolutionary
Biology 3:41 1-428.

Schmitt, J ., and S. E. Gamble. (1990). The effect of distance from parental site on
offspring performance and inbreeding depression in Impatiens capensis: A test of the
local adaptation hypothesis. Evolution 44:2022-2030.

Sultan, S. E. (1987). Evolutionary implications of phenotypic plasticity in plants.
Evolutionary Biology 21:127-176.

Sultan, S. E. and F. A. Bazzaz. (1993). Phenotypic plasticity in Polygonum persicaria.

III. The evolution of ecological breadth for nutrient environment. Evolution
47:1050—1071.

109

Chapter 5

DANTHONIA SPI CA TA AND A TKINSONELLA H YPOXYLON :
ENVIRONMENTAL DEPENDENCE OF A SYMBIOSIS
(Manuscript co-authored with K.L. Gross and RA. Smith)

INTRODUCTION

Much attention has been focused recently on the role of ﬁmgal symbiotes in
structuring interactions within and among plant populations and communities (Clay
1990b, Dobson and Crawley 1997, Thrall and Burdon 1997). Many plant species,
particularly grasses, commonly form symbiotic associations with epiphytic or endophytic
fungi that have the potential to strongly alter the host plant's morphology, physiology,
and response to environmental conditions (e. g., Bacon et al. 1986, Read and Camp 1986,
Belesky et al. 1987, Carroll 1988, Marks and Clay 1990). These fungi can be beneﬁcial
to their host plant in some environments, but may be parasitic in others (Clay 1990a).

Plants that are infected by otherwise advantageous fungi often show growth
disadvantages when available nutrients are very low, because nutrient requirements of
active fungal biomass and toxin production may outstrip nutrient availability (Bacon
1993). Soil fertility-dependent grth advantages of fungal infection have been
demonstrated for endophyte-infected Lolium perenne and F estuca arundinacea (Cheplick
et al. 1989, Marks and Clay 1990, Cheplick 1997). In both of these species, the growth
advantage of infected plants in the greenhouse was either small or negative at the lowest

fertility levels and increased with fertility.

110

Most plant species that have epiphytic or endophytic fungal associations have
both infected and uninfected individuals mixed within and among populations (Bradshaw
1959, Clay 1990a). In mixed populations, infected and uninfected plants are often
patchily distributed, with some areas predominantly infected and other areas
predominantly uninfected. Soil resources can also vary spatially within and among
environments at scales ﬁ'om plant populations to communities (Robertson and Gross
1994, Gross et al. 1995) and differential growth of infected and uninfected plants under
these heterogeneous conditions could produce patches of infected and uninfected plants.

If infected and uninfected plants respond differently to soil conditions that are
patchily distributed, then patches of infection would correspond to patches of
environmental conditions that are favorable to infected plants. To determine whether
patchy distributions of soil nitrogen and percent water were related to observed spatial
variation in a plant-fungal symbiosis, we examined the correlation between Atkinsonella
hypoxylon infection and soil environmental factors (relative soil ammonia supply and
percent water) in three populations of Danthonia spicata in southern Michigan. We then
conducted a greenhouse experiment to test whether growth and survival of D. spicata
plants infected by A. hypoxylon differed ﬁ'om that of uninfected plants under high and
low moisture levels and high and low fertility conditions. A common garden planting of
infected and uninfected culrns was used to determine how the growth and survival of

infected and uninfected plants differed in the ﬁeld.

111

MATERIALS AND METHODS
Study Species:

Danthonia spicata (Poaceae) is a native perennial C3 bunch grass that commonly
occurs in dry, nutrient-poor oak savannas, old-ﬁelds, and openings throughout the eastern
and northern United States and southern Canada (Clay 1982). Throughout much of its
range, D. spicata is infected by the epiphytic fungus Atkinsonella hypoxylon (Balansiae),
which is speciﬁc to the genus Danthonia (Clay 1994). Uninfected D. spicata produce
two ﬂower types. Potentially outcrossed, wind-pollinated chasmogamous ﬂowers are
produced at the tip of the reproductive stalk. Obligately self-fertilized cleistogamous
ﬂowers are produced in the axils of the leaf sheaths along the reproductive stalk. In
infected plants, a ﬁrngal sclerotium, or 'choke', is produced at the initiation of host plant
ﬂowering and causes abortion of all but a few infected cleistogamous seeds at the base of
the much-reduced reproductive stalk. Because plants infected by A. hypoxylon oﬁen have
higher growth rates than uninfected plants, this symbiosis is often considered mutualistic
(Diehl 1950, Clay 1984, Clay 1990a). The proportion of plants infected by A. hypoxylon
varies among populations in Michigan (McCormick, unpublished data). Some
populations contain a high proportion of plants infected by A. hypoxylon, but others are

entirely free of infection (Scheiner 1989, McCormick, unpublished data).

Study Sites
The three populations of D. spicata and A. hypoxylon we examined in this study
were growing in low productivity old-ﬁelds in central and southern Michigan that had

been abandoned at least 45 years ago. Populations L1 and L2 were located near the W.K.

112

Kellogg Biological Station in Hickory Comers, MI. Population R was located in the
Rose Lake Wildlife Research Area near East Lansing, MI. These three sites were all
dominated by herbaceous perennials, simplifying the identiﬁcation of limiting resources.
They also had similar proportions of infected plants. Density (number of D. spicata
plants m'z) was measured along two randomly located belt transects (12 x 0.25 m) in each

population.

Field Patterns

In each population, we established a 4 by 6 m permanent plot where we quantiﬁed
soil moisture and nitrogen supply. Seventy soil samples were taken from each plot, using
a 10 cm deep, 2.5 cm diameter soil core, in a stratiﬁed, nested sampling design. At each
location, where a soil core was removed, an ion exchange resin bag (Dowex MR-3,
Sigma Corp.) was inserted and buried under 5 cm of soil to measure the relative supply of
ammonia-nitrogen in the soil. Soil moisture and relative supply of armnonia were chosen
to characterize environmental quality because these resources often limit plant growth in
the dry, nutrient-poor environments where D. spicata occurs. We also sampled nitrate-
nitrogen in these samples, but found that the availability of nitrogen in these sites was
dominated by ammonia and nitrate supply was rarely consistent among years (Chapter 2).
Because relative nitrate-nitrogen supply was not spatially consistent, we felt that it could
not be used as a reliable determinant of micro-habitat conditions to be related to infection

distribution.

113

We measured gravimetric soil moisture in all three populations (L; and L2) in
mid-May 1998, corresponding to the time of maximum growth by D. spicata. The
relative ammonia supply was estimated in all three populations for four months after
sampling for infection (June 1997-October 1997). Soil moisture and ammonia supply
were also measured at other times using identical methods (see Chapter 2), but in
reference to the distribution of infected and uninfected plants measurements from other
times were used only to assess temporal consistency in infected and uninfected quadrats.
We chose to use 1998 measures of soil moisture and 1997 measures of ammonia supply
because they were taken at all sites, thus allowing comparison across sites, and were the
closest in time to when we measured the distribution of infected and uninfected plants.

To determine whether the pattern of infection incidence in each population was
correlated with these two environmental factors, we surveyed the incidence of A.
hypoxylon infection in each population after the plants had bolted in June 1997. Infected
plants are easily distinguished from uninfected plants at this stage by the presence of a
gray fungal sclerotium ('choke') on aborted reproductive stalks. We randomly selected 75
quadrats (25 x 25 cm) from a 16 x 24 quadrat grid within each of our three permanent
plots to survey for incidence of infection. This small quadrat size was chosen to allow
assessment of plant density without integrating across substantial variation in spatially
heterogeneous environmental conditions. In each quadrat we noted the number and
location of infected and uninfected D. spicata plants. We then used semivariance
analysis and kriging interpolation (GS+ V3.1 1.6. 1999. Gamma Design Software,
Plainwell, M1) to estimate percent water and ammonia supply for the center of each

quadrat that we sampled for infection incidence. Kriging interpolation uses the variance-

114

 

distance relationship, summarized in a semivariogram, to assign weights to sample points
as a function of their distance ﬁom a point for which an estimate is desired. Kriging
allowed us to use the spatial structure of variation in each factor (moisture and ammonia)
in each population to estimate the levels of soil moisture and ammonia for each quadrat
in which we sampled plant infection.

Additional measurements of soil moisture and ammonia supply were taken in
each site using the same methodology as described for the 1998 samples above. To
assess the temporal consistency of estimated soil moisture and ammonia conditions in the
quadrats, we examined the correlation between 1998 and 1996 (L1, L2) or 1997 (R)
estimates of soil moisture and between 1997 and 1998 estimates of ammonia supply for
each quadrat. We assessed the signiﬁcance of temporal consistency using an ANOVA
for each factor, where Population and the earlier soil estimate (1997 for ammonia and
1996 or 1997 for soil moisture) were used as the predictors of later soil estimates (1998

soil moisture and ammonia).

Statistical Analysis

We designated each quadrat as infected if it contained at least one infected plant.
We then used logistic regression to analyze the distribution of infected quadrats over the
range of soil moisture and ammonia conditions across the three populations (Systat 8.0
for Windows. Systat Inc. 1998. Evanston, 11.). Because of the possible confounding of
spatial structure with point of infection introduction or other processes within a
population, we can only draw conclusions about the pattern of association between

environmental factors and infection incidence across all three populations. A signiﬁcant

115

population effect in the regression would indicate that patterns were not the same across
the three populations.

Because more favorable environments might promote higher plant densities and
plant density could affect contagious spread of the fungus, we needed to address the
possibility of indirect environmental effects on the distribution of infection acting
through plant density. To evaluate this effect, we estimated plant density across all
infected and uninfected quadrats. We used a t-test to compare the average plant density

in infected and uninfected quadrats.

Greenhouse Experiment

To determine whether the variation observed in the ﬁeld in soil fertility and
moisture levels could affect the performance, and thus inﬂuence the differential
distribution of infected and uninfected plants, we conducted a 2 x 2 factorial greenhouse
experiment in which watering regime and fertilizer were manipulated. We collected
cleistogamous seeds ﬁ'om 21 infected and 28 uninfected individuals selected randomly
from two D. spicata populations. The two populations from which we collected seeds
were Population L1 from the ﬁeld patterns survey and a second, unsurveyed, population,
which was dominated by woody vegetation and may have experienced very different
environmental conditions.

Infected and uninfected seeds were surface sterilized with bleach and ethanol
according to the methods of Leuchtrnann and Clay (1988) and nicked with a sterile razor

blade to stimulate germination. Seeds were then germinated in moist, sterile sand in a

growth chamber with 14 hour days at 29°C and 10 hour nights at 24°C. Of the 21

116

infected and 28 uninfected plants ﬁom which we collected seeds, 12 infected and 17
uninfected plants had sufﬁcient germination for experimental replication. After nine
days, 109 seedlings from infected and uninfected families were each divided into four
treatment groups, assigned at random to positions in 70-hole ConetainersT'“I (each hole
was 2.5 cm diameter by 15 cm deep) ﬁlled with sterile silica sand and placed in the
greenhouse under ambient light. Each infected family was represented by nine or ten
seedlings, two in each treatment with the additional one or two seedlings randomly
assigned to treatments. Each uninfected family was represented by six or seven
seedlings, one per treatment with the other two or three randomly assigned to treatments.
After a four day stabilization period with daily watering to saturation, we established the
four treatments (2 x 2 factorial) in which fertility and watering regime were varied.
Seedlings from each infected and uninfected family were grown under all combinations
of high and low fertilizer and high and low moisture.

The moisture and fertility levels used in the greenhouse experiment were chosen
to represent the range of conditions observed in the ﬁeld, without imposing extremely
high mortality. Fertilizer levels consisted of 0.015 g L'1 (low fertility) or 0.500 g L'1
(high fertility) of Peter's Peat-lite Special Fertilizer (20-10-20) applied at a rate of 4.5 ml
per planting location 2-3 times per week. The nitrogen levels in these fertility treatments
corresponded to nitrogen mineralization rates of 0.01 and 0.45 mg N g'1 dry soil day], the
approximate range of fertility found in ﬁeld populations of D. spicata (Chapter 2). Plants
in high moisture treatments were watered daily, while plants in the low moisture
treatments were watered every second or third day when approximately half of the plants

showed leaf rolling, symptomatic of water stress. Average greenhouse temperatures were

117

approximately 30°C from June-September and approximately 24°C from October-April
with ambient light. The positions of the 70-hole ConetainersTM on the greenhouse bench
were rotated randomly each week to minimize position effects.

Plants were monitored at regular intervals over nine months for size and survival.
Plant size was measured as the number of living leaves on each plant. After six months,
plants in high fertility treatments were sufﬁciently large that individuals in adjacent
locations began to shade plants each other. Therefore, we transplanted the plants in these
treatments into new 70-hole Conetainersl" and placed individuals in staggered locations

with one empty location on all sides to prevent shading.

Statistical Analysis

Plant growth data were log transformed to minimize heterogeneous variances
produced by substantial growth differences between the high and low fertility treatments.
Transformed data were analyzed using a repeated measures ANOVA (Systat 8.0 for
Windows. Systat Inc. 1998. Evanston, IL.). We also calculated performance by using
the percent survival as a weight on the size of each plant in each treatment (log number of
leaves). We compared performance among treatments using a repeated measures

ANOVA.

Common garden experiment
To assess differences in growth of infected and uninfected plants under ﬁeld
conditions, we established a common garden adjacent to our permanent plot in

Population L. We collected nine uninfected and three infected adult D. spicata plants

118

 

from ﬁeld Populations L1 and L2 in April 1995. The plants were divided into culms and
each culrn was planted into a 10 x 10 cm pot ﬁlled with sterile sand in the greenhouse.
Cuhns that propagated were again divided into individual culrns and planted into new 10
x 10 cm pots. The culms were grown without fertilizer until June 1995, when we used
them to establish the common garden experiment.

We established the common garden by tilling a 2 x 2 m area near Population L1.
We planted four culms of each genotype (individual parent) into randomly selected
locations on a 10 x 10 cm grid established in the central 1 x 1 m of the tilled area.
Remaining planting locations in the grid were occupied by D. spicata culms ﬁ'om another
experiment. We planted additional randomly selected culms around the perimeter of the
common garden to avoid edge effects on the study plants. We measured the initial size of
planted cuhns by counting the number of living leaves at the time of planting.

Planted culms were allowed to grow in the common garden for two years. In June
1997 (peak seed production) we harvested above-ground biomass from all surviving

culms. From the harvested biomass we measured ﬁnal plant size by counting the number

of tillers, leaves, and inﬂorescences produced. All clipped material was dried at 45°C for
48 hours and then weighed. We ground random sub-samples of leaf material ﬁ'om each
planted cuhn to determine the tissue nitrogen concentration. We also ground
reproductive stalks, with all seeds removed, from all infected culrns and from three
randomly selected uninfected culms, each from a different parent. For infected
reproductive stalks we included the ﬁmgal sclerotium, which was intimately associated
with the inﬂorescence, in tissue to be ground so we could assess the overall nitrogen

content of the inﬂorescence. Percent nitrogen content of this tissue was analyzed using

119

an elemental analyzer (Nitrogen Analyser 1500 Series 2, 1990, Carlo-Erba Instruments,

Milan, Italy).

RESULTS
Field patterns

Plant density and relative supply of ammonia were not statistically different
among the three sites, but Population R had signiﬁcantly lower soil moisture than
Populations L1 and L2. The three sites also had similar percentages of infected plants and
soil moisture (Figure 5.1). Over all three sites, plant density in infected quadrats was not
signiﬁcantly different from that in uninfected quadrats (40.8 i 0.79 plants rn‘2 versirs 35.6
i 3.65 plants m'z, respectively; [mean i ls.e.] p>0.2 from t-test). Within each site, 1996
or 1997 soil moisture estimates were signiﬁcantly predictive of 1998 soil moisture (p <
0.04) and 1997 ammonia estimates were signiﬁcantly predictive of 1998 ammonia supply
in two of the three populations (L, R, p < 0.004; L2, p = 0.504), suggesting that soil
conditions where infected or uninfected plants were growing were generally temporally
consistent among years.

Across the populations, infected quadrats were drier and more ammonia-rich than
quadrats with no infected plants (Figure 5.2). There was a signiﬁcant interaction between
soil moisture and ammonia supply in determining infection prevalence (logistic
regression; p < 0.001). Because our three populations did not overlap in their mean
conditions, they reveal substantial information about the distribution of A. hypoxylon
across a wide range of conditions, but they can not be treated as replicates. Populations

L1 and L2 had substantial overlap in their soil moisture and ammonia conditions and both

120

 

 

 

 

 

 

 

 

 

 

 

 

 

4o 50
'0 NA #
33 30g g 40
d.)
_‘g E 30-
£13 20‘ e
E; e 20‘
._4 U)
c.\" 10 5 10—
Q
o- o-
20—
E
.a 15*
O
E 10-
1':
O
(I) 5._
x
0_.
8 12
O
7“ 9 A 10- 6
’5‘ 6— .5
§ 5— § 8‘
DD 4_ GI) 6.d
3 3- E ’
<1- 2_ O Om 4‘ Q
E 1_ 2 2i
0— 04
L1 L1 R L1 L1 R

Population Population

Figure 5.1: Characteristics of three study populations
Percent of plants infected, mean plant density, 1998 soil moisture, 1997 arnrnonia-
nitrogen, and 1997 nitrate-nitrogen (+ 1 s.e.). N = 2 for density measures, and

ranges from 58 to 68 for soil moisture and nitrogen measures. Diamonds (9)
indicate minimum and maximum soil moisture and nitrogen measures.

121

 

.=8 .5 m \ o~m m 26 9 3 ace 8.: as ”Baas :8

05 E x85 mg 05. Adv M noun—seem can AIV 3 :owﬂseom ADV J nowﬂneom 5 bag 3:955 98 0.5688 :8 2 033.2
838%: $38.33;. 3 E0253 :omov 3885:: use Eon—Pam 233 @283 $53 88.3“. #50355 .«o cotanmbmae BE.
95% case. s§s§e 8.825 e5 Essence 8:335 22a ”2 03E

122

98 E 3%: 3 03408 mom

 

 

Nd _no mod

n? o
Imd W
4 4 m
D 4 4 4 44 4 m.
4 4 l a
$444444 m,
,4 4 @ﬁdd Q Q M
‘ 4 q .7-
. 42. -2 m
G 44 m
4 ....
44 4 w

4 4 um

4 4 4
g q [m N
4 4 4
4 H m

 

 

 

3524 38.4% 434$qu @825 Ba 4283 we 42.35% %E ”a.“ 2&3

123

had infected plants only in lower moisture, higher ammonia conditions (Figure 5.2). In
contrast, Population R had somewhat higher ammonia and substantially lower soil
moisture than Populations L1 or L2, placing the entire population in conditions suitable
for infected plants. Accordingly, infected plants were distributed throughout Population

R without regard to soil moisture or ammonia conditions (Figure 5.2).

Greenhouse Experiment

In the greenhouse experiment, the direction and magnitude of growth and survival
differences between infected and uninfected plants depended on both the fertility and
moisture treatments. All plants grew more slowly at low fertility than at high fertility
(Figure 5.3), resulting in a signiﬁcant Fertility effect (p < 0.001). However, no other
between subject effect was signiﬁcant. Within subjects, Time x Fertility and Time x
Infection x Fertility effects were signiﬁcant (p < 0.001), indicating that infected and
uninfected plants grew differently in response to fertility treatment. In contrast to high
fertility, where high moisture uninfected plants grew the least, at low fertility, the low
moisture infected plants had lower growth rates than other low fertility plants (Figure
5.3). This suggested that the differential growth of infected and uninfected plants
depended on both fertility and watering regime, but the Time x Fertility x Water x
Infection interaction was not signiﬁcant (p = 0.165).

Plant performance, calculated as the product of average treatment group survival

and plant size (log number of leaves) was signiﬁcantly different among the experimental
treatments (Figure 5.4). Using performance as an index of relative group success,

uninfected plants under high moisture, high fertility conditions outperformed all other

124

.< >0 Z< 8.5308 3338

035538 a 80¢ 88%? Enema—ma 886.: 82a?& .o.m _ H 823— mo Baas: m2 588 can 83m> Sm x83 3 3 8 Z 9

0 x83 3 mm 8 om 89a can; Hosanna bate.“ 32 05 .8.“ moumm 295m Sm x33 3 mm 8 3 3 o x83 3 mm 8 mm Bob 49; bate.“
:42 é 8e... 295m .eéeoa e8: 8:84 .2853 >23 $2 23 98: £8 .2853 case :3: a $538: bate A: 32
EH AOV :ME 05 5 23498.43 ezmzoneﬁaw 43 Eons—Sm :33 “.28.“? 98 338% come—0v 380%: 353 83.5.4 .Qmo £380
$5.585 8:085on E 88.4% 3.39335 no 5386 ”0m 035

125

 

wﬁnsq 3:7. 383

 

 

  

 

a mm mm 8 8. 2 2 w e o
_ _ — _ r _ _ _ _ O
wad
-.\ H
t .11.. L. e
4‘ “10.0
555 33 wad m.
m a
,L M
h w
w? m
H ,9»
vi
3&5 @m I
, we;
....... 4 . ”8.ch uaaigsﬁ
9‘11‘ . I .
<.\|J-.\.-...\ wood" a tumxuxoﬁa. 1Tw_
.. - 83" a mebaaxoﬁe,
N

 

3:258: 3:058on E 88.3». SzoﬁauQuo 5380 “mi. onE

126

.<>OZ< $5808 @0638 a Bow $950 80.3%. 50389 Enomawmm 032?: moswgi

.o.m _ + ooggofoa :38 08 8:15 Sm Moo? 8 3 8 Z 9 a x33 “a R 3 ca Bob b?» #5585 >2:qu 32 2: you monmm £95m
Hm x33 as mm 8 2 2 o #33 3 mm 9 .3 88w b? mum—Emu :wE 8m maﬁa 295m 65638 Goa: 3:93 £38? >83 32 93
a8: 38 .2853 x85 $2 3 “Egg beta 32 Ea $2585 bzpé E 32 Ba A: ﬁe 2: a 8383 §m§§€w
an E85? :33 wouoomaﬁs Ea Eon—Em 38.8 8603 8:29 88.3w. .Qmo @262 MO 8:85: me. x 33>va ouggotom
$5.535 03025on E 883w 5:32:53 ooﬁgotom ”Wm 8:me

127

 

wanna—q 33m £83

 

 

 

S mm mm mm cm 8 «P m w o
b _ _ g _ _ b _ _ O
\ .. r
pg
1 -
. , . . . ....... #6
bag 33 M
[ad
86
j
bag ﬁE n
/ III nll‘ IIII . .|| . r

 

sod" a 55533: -

J.
((SOABG[#) 801 x lawns) aounuquad

 

83 v a H 1:

Sod v a 535 m

sod V a tum T
3

 

3:08.85 8:05.on 5 883m Sneiqumo éombm ”in 2&5

128

high fertility plants (Figure 5.4). These uninfected plants were the smallest high fertility
plants (Figure 5.3), but had substantially higher survival than other treatment groups.
Infected plants under low fertility, low moisture conditions performed significantly less
well than all other low fertility treatment groups. These infected plants were the smallest

low fertility treatment group (Figure 5.3) and also had the lowest survival (Table 5.1).

Table 5.1: Survival of infected and uninfected plants in the greenhouse treatments
Survival (%; N = 26 to 28 plants) of infected (I) and uninfected (U) Danthonia spicata

plants under high and low fertility and high and low moisture treatments in the

 

greenhouse.
FERTILITY
High Low
Infection

High U 79 48

MOISTURE I 56 46

Low U 59 52

I 59 39

 

Survival in the greenhouse appeared size-dependent. Plants that died often had
few living leaves in the previous months, although their maximum size may have been
quite large. In the high fertility, high moisture treatment many more small uninfected

plants survived than infected plants (Figure 5.5a). It is possible that, small uninfected

129

 

Numberofplants
w «b M 0\
ll Llrllllillllil

N
lllll

 

 

 

h
L l

N

Number of plants
DJ
1 14 l L 1_l 1 l

p—b
l

 

C
It“

5

 

 

 

T l l l T

15 25 35 45 55 65 75 85 95 105115125135145
Number of living leaves

Figure 5.5: Final size distributions of Danthonia spicata plants

Final size distributions of Danthonia spicata plants infected (black bars) and
uninfected (gray bars) by Atkinsonella hwoxylon in the high fertility, high
moisture treatment (a.) and in the high fertility, low moisture treatment (b.) in
the greenhouse experiment. The number of living leaves is the median value of
the categiry (e.g O to 10 leaves is written as 5).

130

plants were able to survive in the relatively benign high fertility, high moisture growing
conditions, while the nutrient demand by A. hypoxylon on small infected plants caused
them to die. In contrast, in the low moisture treatment where neither survival nor plant
size differed, the size distributions of infected and uninfected plants were similar (Figure
5.5b). Thus, the small average size of uninfected plants may be due to their higher

survival under high fertility and moisture conditions compared to infected plants in this

 

same treatment (Figure 5.3). Because differences in mean size are counteracted by
differences in survival, the larger size and higher growth rates of infected plants in the

high fertility, high moisture treatment does not result in a performance advantage under

 

these conditions.

Common Garden

In the common garden, size at planting had no signiﬁcant effect on ﬁnal size
(Pearson correlation, p>0.3), so only ﬁnal size was considered in these analyses. Infected
and uninfected plants did not differ signiﬁcantly in total biomass in the common garden
(p = 0.310). Because reproductive stalks of infected plants were aborted by the fungal
sclerotium at an early stage of growth, they were much smaller than reproductive stalks
of uninfected plants. As a consequence, infected plants had signiﬁcantly more vegetative
biomass, but less reproductive biomass than uninfected plants (Figure 5.6a).

Infected plants had signiﬁcantly higher nitrogen concentrations in their vegetative
tissues than uninfected plants. Inﬂorescences of infected plants (with the fungal
sclerotium) had almost twice the nitrogen concentration of uninfected plants (Figure

5.6b). The substantial increase in nitrogen concentration in infected inﬂorescences

131

 

3' 3.5
3-1
2.5 -

1.5“

Biomass (g dry matter)

   

 

1.5—

  

 

 

    

Nitrogen concentration (percent) 9'
l

Vegetative Reproductive
Plant material sampled

Figure 5.6: Biomass and nitrogen concentration of plants gown in the common garden
Plant biomass allocation (a.) and tissue nitrogen concentration in vegetative and
reproductive tissue (b.) for Danthonia spicata plants infected (gay bars) and
uninfected (black bars) gown in a common garden. Tissure nitrogen of infected plants
included fungal hyphae on vegetative surfaces and the fungal stroma on the
reproductive tissue. All values : l s.e.; N = 3 genotypes for infected plants and
uninfected reproductive tissue. N = 10 genotypes for uninfected biomass and
vegetative tissure nitrogen measurements. Infected and uninfected plants were
sigriﬁcantly different in biomass distribution and nitrogn concentration of both
vegetative and reproductive tissue at p: 0.002.

132

relative to vegetative material was probably due to the much larger concentration of
fungal biomass in the sclerotium. This result suggests that the higher nitrogen
concentrations of infected plants was due to the presence of nitrogen-rich fungal biomass,

rather than from the plant tissue having a higher nitrogen concentration.

DISCUSSION

In two of three ﬁeld populations, quadrats where D. spicata was infected by A.
hypoxylon had lower soil moisture and higher relative ammonia supply. This suggests
that in these Michigan sites soil resource heterogeneity inﬂuences the distribution of A.
hypoxylon infected D. spicata plants. It is unknown how the fertility of their ﬁeld sites
compares to our sites.

Several studies have addressed the role of ﬁmgal symbiotes in altering growth
requirements of infected plants (e.g., Antonovics et al. 1987, Marks and Clay 1990,
Bacon 1993, Latch 1993). However, none of these studies assessed the importance of
environmental heterogeneity in structuring plant-fungal interactions (but see Thrall and
Burdon 1997). Our greenhouse results demonstrate that infected and uninfected plants
respond differently to fertility and moisture. We also found that ammonia supply was
correlated with the incidence of fungal infection in our ﬁeld populations. The uniqueness
of our study is that we linked growth differences in the greenhouse and common garden
to the distribution of fungal infection in natural ﬁeld populations.

Results from our greenhouse study suggested that infected plants had a
performance disadvantage when grown under low moisture, low fertility and high

moisture, high fertility conditions (Figure 5.4). Both watering regime and fertility

133

differentially affected the performance of infected and uninfected plants. Consequently,
there was no overall performance (survival x fecundity) advantage for infected plants
under our greenhouse conditions.

In our ﬁeld surveys, we found that infected plants were most common in areas
with high ammonia supply, while areas with high soil moisture and low ammonia supply
had very few infected plants. This result is in contrast with studies in Indiana and North
Carolina that have found that infected plants grew consistently better than uninfected
plants in both ﬁeld plantings and in the greenhouse (Clay 1984, Kelley and Clay 1987,
Leuchtmarm and Clay 1988). The soil used in these other experiments was quite fertile
compared to our ﬁeld sites and this difference may explain the dominance of infected
plants in their greenhouse studies.

In the greenhouse experiment infected plants grew less well than uninfected
plants in the low moisture treatments. In the greenhouse, soil moisture in the low
frequency watering treatments was maintained at approximately 6%, while in the high
frequency watering treatments soil moisture was approximately 25%. In our three ﬁeld
populations, Danthonia was rarely found in sites where soil moisture was above 20%.
Consequently, the low moisture treatments in the greenhouse were more relevant to
understanding the distribution of infected and uninfected plants in our three ﬁeld
populations than the high moisture treatments were. Further comparisons among D.
spicata populations with and without A. hypoxylon may yield a better understanding of
the role played by soil moisture in mediating the effect of A. hypoxylon on D. spicata.

When both fertility and moisture were low in the greenhouse experiment, infected

plants grew and performed less well than uninfected plants. A possible reason for the

134

poor performance of infected plants in the low fertility, low moisture treatment was
suggested by the nitrogen content data from plants grown in the common garden
experiment. Higher nitrogen concentration in fungal biomass may put a high nitrogen
acquisition demand on infected plants. An increased nitrogen demand on infected plants
could limit them to areas of the ﬁeld with high relative supply of ammonia. This may
also have been what caused them to grow less well than uninfected plants in the low
fertility, low moisture treatment in our greenhouse experiment. In southern Michigan
ﬁelds, D. spicata rarely occurs in areas of the ﬁeld with both high moisture and high
ammonia, as these sites are dominated by other plant species (McCormick, unpublished
data). In a greenhouse study, Kelley and Clay (1987) found that D. spicata was a
relatively poor competitor for light with other grass species, suggesting that other plant
species might exclude D. spicata ﬁ'om richer areas of the ﬁeld. Thus if D. spicata only
occupies areas of the ﬁeld with either high moisture OR high ammonia, then a high
nitrogen demand might also limit infected D. spicata to lower moisture areas of the ﬁeld.
Infected D. spicata plants are common and often dominate low moisture, high
ammonia areas of ﬁelds in southern Michigan. However, in the greenhouse they did not
have a clear performance advantage under these or any conditions. The high incidence of
infection in some areas of the ﬁeld, despite no indication of advantage in the greenhouse
or common garden, suggests that other factors may be inﬂuencing the ﬁeld distribution of
A. hypoxylon-infected D. spicata in these sites. Competition (intra- and inter-speciﬁc)
and herbivory were not included in our greenhouse study, but could inﬂuence the
distribution of infected plants in the ﬁeld. Kelley and Clay (1987) found that infected D.

spicata were better interspeciﬁc competitors than uninfected D. spicata. Infected D.

135

spicata may also be less susceptible to herbivory, as has been shown for several
endophyte-infected species (e.g., Prestidge et al. 1982, Bacon et al. 1986, Read and Camp
1986, Clay et al. 1993). We saw no evidence of herbivory in our common garden study
even though herbivores (e. g. rabbits, deer, mice) are common in this ﬁeld and were not
excluded from the plots. Infected plants also did not differ from uninfected plants in total
biomass, butthey did have greater leaf biomass. Increased competitive ability or
herbivory resistance could convey a performance advantage to infected plants under
conditions of sufﬁcient fertility.

We have found that D. spicata plants infected by A. hypoxylon performed less
well under low moisture, low fertility and high moisture, high fertility conditions than
uninfected plants. These performance differences and the spatial heterogeneity of soil
resources in the ﬁeld may explain some of the variation in the distribution of infected and
uninfected plants we observed in the populations we surveyed. The absence of infected
plants under low ammonia conditions in the ﬁeld may have resulted from reduced
survival or grth caused by a high nitrogen demand by A. hwonzlon from infected
plants. Our observation that infected plants have higher nitrogen concentration may also
suggest a mechanism for the exclusion of infected plants ﬁom dry, low fertility ﬁeld
locations. Additionally, if fungal tissue has a higher nitrogen content than plant tissue, it
may leak nitrogen to host plant leaf tissue. The increased leaf nitrogen content could
increase photosynthetic efﬁciency and possibly the competitive ability of infected plants
under high nitrogen conditions, allowing them to dominate uninfected plants in higher

fertility areas of the ﬁeld. However, this hypothesis remains to be tested.

136

LITERATURE CITED

Antonovics, J ., K. Clay, and J. Schmitt. 1987. The measurement of small-scale
environmental heterogeneity using clonal transplants of Anthoxanthum odoratum
and Danthonia spicata. Oecologia 71 :601-607.

Bacon, CW. 1993. Abiotic stress tolerances (moisture, nutrients) and photosynthesis in
endophyte-infected tall fescue. Agriculture, Ecosystems and Environment 44:123-
141.

Bacon, C.W., P.C. Lyons, J .K. Porter, and J .D. Robbins. 1986. Ergot toxicity from
endophyte-infected grasses: A review. Agronomy Journal 78:106-116.

Belesky, D.P., OJ. Devine, J .E. Pallas Jr., and W.C. Stringer. 1987. Photosynthetic
activity of tall fescue as inﬂuenced by a fungal endophyte. Photosynthetica 21 :82-
87.

Bradshaw, A.D. 1959. Population differentiation in Agrostis tennis Sibth. II. The
incidence and signiﬁcance of infection by Epichloe typhina. New Phytologist
5 8:3 10-3 1 5.

Carroll, G.C. 1988. Fungal endophytes in stems and leaves: ﬁ'om latent pathogen to
mutualistic symbiont. Ecology 69:2-9.

Cheplick, GP. 1997. Effects of endophytic fungi on the phenotypic plasticity of Lolium
perenne (Poaceae). American Journal of Botany 84:34-40.

Cheplick, G.P., K. Clay, and S. Marks. 1989. Interactions between infection by
endophytic fungi and nutrient limitation in the grasses Lolium perenne and F estuca
arundinacea. New Phytologist 111:89-97.

Clay, K. 1982. Environmental and genetic determinants of cleistogamy in a natural
population of the grass Danthonia spicata. Evolution 36:734-741.

Clay, K. 1983. Variation in the degree of cleistogamy within and among species of the
grass Danthonia. American Journal of Botany 702835-843.

Clay, K. 1984. The effect of the fungus Atkinsonella hypoxylon (Clavicipitaceae) on the
reproductive system and demography of the grass Danthonia spicata. New
Phytologist 98:165-175.

Clay, K. 1990a. Fungal endophytes of grasses. Annual Review of Ecology and
Systematics 21:275-297.

137

Clay, K. 1990b. The impact of parasitic and mutualistic fungi on competitive
interactions among plants. In: Grace JB, Tihnan D (eds.) Perspectives On Plant
Competition. Academic Press, San Diego, California, pp 391-412.

Clay, K. 1994. Hereditary symbiosis in the grass genus Danthonia. New Phytologist
126:223-231.

Clay, K., S. Marks, and GP. Cheplick. 1993. Effects of insect herbivory and fungal
endophyte infection on competitive interactions among grasses. Ecology 74:1767-
1777.

Diehl, W.W. 1950. Balansia and the Balansiae in America. US. Department of
Agriculture, Agricultural Monographs 4: 1-82.

Dobson, A. and M. Crawley. 1997. Pathogens and the structure of plant communities.
Trends in Ecology and Evolution 9:393-397.

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

Kelley, SE. and K. Clay. 1987. Interspeciﬁc competitive interactions and the
maintenance of genotypic variation within the populations of two perennial grasses.
Evolution 41 :92-103.

Latch, G.C.M. 1993. Physiological interactions of endophytic fungi and their hosts.
Biotic stress tolerance imparted to grasses by endophytes. Agriculture, Ecosystems
and Environment 44:143-156.

Leuchtrnann, A. and K. Clay. 1988. Experimental infection of host grasses and sedges
with Atkinsonella hypoxylon and Balansia cyperi (Balansiae, Clavicipitaceae).
Mycologia 80:291-297.

Marks, S. and K. Clay. 1990. Effects of C02 enrichment, nutrient addition, and fungal
endophyte-infection on the grth of two grasses. Oecologia 84:207-214.

Prestidge, R.A., R.P. Pottinger, and GM. Barker. 1982. An association of Lolium
endophyte with ryegrass resistance to Argentine stem weevil. Proceedings of the
New Zealand Weed Pest Control Conference, 35th pp. 199-222.

Read, J .C. and B.J. Camp. 1986. The effect of fungal endophyte Acremonium

coenophialum in tall fescue on animal performance, toxicity, and stand maintenance.
Agronomy Journal 78:848-850.

138

 

Robertson, GP. and KL. Gross. 1994. Assessing the Heterogeneity of belowground
resources: quantifying pattern and scale. In: Caldwell MM, and Pearcy RW (eds)
Exploitation Of Environmental Heterogeneity by Plants. Academic Press, San
Diego, California, PP 237-253.

Scheiner, SM. 1989. Variable selection along a successional gradient. Evolution
43:548-562.

Thrall, PH. and J .J . Burdon. 1997. Host-pathogen dynamics in a metapopulation
context: the ecological and evolutionary consequences of being spatial. Journal of
Ecology 85:743-753.

Id

 

139

Chapter 6

CONCLUSIONS

Although it has long been recognized that spatially structured variation in soil
resources has the potential to affect plant populations and communities ecologically and
evolutionarily, population and community effects of spatially structured resource
variation have rarely been examined in ﬁeld populations (Robertson et al. 1988,
Ehrenfeld 1997). This limited attention is partly because of difﬁculty in relating
measured heterogeneity in resources to plant-perceived heterogeneity, and partly because
it was unknown whether patterns of spatial variation would persist long enough to
produce consistent directional selection on plant populations (e. g. Stratton and
Bennington 1998). The studies presented in this dissertation demonstrate that spatially
structured soil resource variation can have strong ecological and evolutionary effects on
plant populations and communities.

The ﬁve ﬁeld sites that I studied differed in the degree to which soil resources
were spatially structured and temporally consistent. More importantly, I found that D.
spicata populations in these sites responded ecologically to the measured differences in
spatially structured soil resources within sites. Plant response to these differences in
spatially structured variation in soil resources was largely through differences in survival.
There was little evidence for growth differences among surviving plants that could be
attributed to spatial variation in soil resources. This suggests that spatial variability in
soil resources inﬂuences the recruitment distribution of D. spicata plants in these sites,

but not their growth and production. The lack of effect of small scale heterogeneity in

140

soil resources on the growth of surviving plants may result ﬁ'om the very high levels of
phenotypic plasticity in D. spicata. In greenhouse studies, I found that there were similar
levels of growth plasticity to variation in nitrogen availability in plants ﬁ'om all ﬁve study
populations.

Differences in average population dispersal distance among these ﬁve D. spicata
populations suggest that this trait may be responding to selection to reduce the amount of
spatial environmental variation encountered by offspring. In spatially structured
environments, decreasing dispersal distance decreases the average environmental
variation encountered by offspring, while decreasing dispersal distance in unstructured
environments has little effect on the amount of environmental variation encountered by
offspring. I found that populations in sites with structured variation in soil resources
allocated a greater proportion of their seeds to short dispersing seeds than those in
unstructured environments. However, the observation that these study populations have
similar, high, amounts of phenotypic plasticity in three traits that affect nitrogen
acquisition suggests that dispersal and nitrogen acquisition traits respond differently to
selection by spatially structured environmental variation.

Selection for phenotypic plasticity of many traits may be inﬂuenced more by
temporal variation in environmental conditions than spatial variation (e. g. Sultan 1987,
Sultan and Bazzaz 1993). Although I found signiﬁcant temporal consistency of spatially
structured variation for soil moisture, nitrogen availability was less consistent in my
study sites. For plants such as D. spicata that are limited to low resource enviromnents,
the beneﬁts of having high plasticity in traits that affect resource acquisition may

outweigh the cost of maintaining that plasticity (Sultan and Bazzaz 1993). However, to

141

clearly distinguish the relative importance of selection by temporal versus spatial
environmental variation on any plant species would require long-term experiments that
followed the success of genotypes with different dispersal distances and levels of
phenotypic plasticity under different levels of independently manipulated temporal and
spatial environmental variation.

I also examined the way that resource heterogeneity inﬂuenced the interaction
between D. spicata and A. hypoxylon. Soil moisture and nitrogen conditions affected the
performance of D. spicata plants infected by A. hypoxylon differently than they affected
uninfected plants. Uninfected plants out performed infected plants under low moisture,
low fertility and high moisture, high fertility conditions in the greenhouse. Because of
these different effects, I also expected that spatially structured soil resources would alter
the spatial distribution of infected and uninfected plants in the ﬁeld. I found that infected
and uninfected D. spicata were distributed differently with respect to soil moisture and
ammonia conditions in three ﬁeld populations. Field conditions in these three sites
corresponded largely to the low moisture treatment in the greenhouse and, consistent with
the results of the greenhouse experiment, infected plants were not found in dry, low
fertility areas of the ﬁeld.

The results from these studies with D. spicata suggest that spatial environmental
structure within plant populations may be more consistent and have greater ecological
and evolutionary effects than previously thought. The effect of spatially structured
environmental variation on population characteristics and species interactions has wide-
ranging implications for processes at levels of ecological organization from genes to

ecosystems. If population evolution in response to spatial heterogeneity is a general

142

phenomenon, then additional attention to the spatial structure of environmental variables
may be warranted in studies of ecology and evolution at all levels of ecological

organization.

143

 

LITERATURE CITED

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 85:185-798.

Robertson, G. P., M. Huston, and F. Evans. 1988. Spatial variability in a successional
plant community: Patterns of nitrogen availability. Ecology 69:1517-1524.

Stratton, D. A., and C. C. Bennington. 1998. Fine-grained spatial and temporal variation
in selection does not maintain genetic variation in Erigeron anuus. Evolution, 52:
678-691.

Sultan, S. 1987. Evolutionary implications of phenotypic plasticity in plants. Pages 127-
178 in Hecht, M. et al. eds. Evolutionary Biology, volume 21. Plenum Press,
New York.

Sultan, S., and F. A. Bazzaz. 1993. Phenotypic plasticity in Polygonum persicaria H.
Norms of reaction to soil moisture and the maintenance of genetic diversity.
Evolution 47:1032-1049.

 

 

144

APPENDIX

Table A.l: Composition of modiﬁed Hoagland’s solution used in this experiment

 

 

CHEMICAL Concentration (g/L)
Micronutrients:
KCl 0.003728
H3BO3 0.001546
MnSO4-H20 0.000338
ZnSO4-7H20 0.000575
CuSO4-5H20 0.000125
MoO3 0.00007197
FeEDTA 0.00742
Macronutrients:
CaSO4-H20 0.3443
MgSO4 0.2465
K2SO4 0.1046
K2HPO4 0.0871
KH2PO4 0.0476
Nitrogen:

NH4NO3 (low N) 0.0429
NH4NO3 (mod N) 0.1144

NH4NO3 (high N) 0.2135

 

145

 

”11111011111111111“