Ecology, 91(7), 2010, pp. 2141–2150� 2010 by the Ecological Society of America

Spatial niches and coexistence: testing theory with tarweeds

SUSAN HARRISON, HOWARD CORNELL,1 AND KARA A. MOORE

Department of Environmental Science and Policy, University of California, Davis, California USA 95616

Abstract. Competitive coexistence in a spatially heterogeneous environment is tradition-ally attributed to niche differences, but several recent theories have proposed importantadditional roles for propagule limitation and chance (e.g., neutral theory, stochastic nichetheory, spatial storage effect). We tested whether propagule supply and timing of disturbanceaffected the coexistence of three ecologically similar plants that replace one another withpartial overlap along a local soil gradient. We asked what prevents the species that dominatesthe most common habitat (Holocarpha virgata, open hillsides) from invading the habitatswhere the other two species are dominant (Calycadenia pauciflora, rocky hilltops; Hemizoniacongesta, clay-rich bottomlands). We added abundant Holocarpha seeds into Calycadenia andHemizonia habitats that were experimentally disturbed at different times of year. InitialHolocarpha seedling densities in Calycadenia and Hemizonia habitats equaled or exceededthose in unmanipulatedHolocarpha habitat, butHolocarpha survival, adult size, and fecunditywere much lower outside its own habitat. Holocarpha persisted in Calycadenia and Hemizoniahabitats for three years, and springtime disturbance promoted this expansion. However,outside its own habitat Holocarpha showed below-replacement fitness and little competitiveeffect on the other two species. Our results were most consistent with a deterministic view ofspatial niches. Nonetheless, chance events may often cause natural communities to includesome transient populations at any given time, leading them to appear ‘‘unsaturated’’ withspecies.

Key words: Calycadenia pauciflora; competition; disturbance; Hemizonia congesta; Holocarphavirgata; Hutchinsonian niche; neutral theory; seed addition; serpentine; spatial storage effect; stochasticniche; tarweed.

INTRODUCTION

Spatial heterogeneity in the environment clearly

affects competitive coexistence and community assem-

bly, but the precise mechanisms and their larger

implications are debated. In traditional Hutchinsonian

theory, species coexist because they cannot compete

successfully in one another’s niches. In some cases niche

differentiation is spatial, such as when each competitor is

more effective at acquiring the same limiting resource(s)

under a slightly different set of biotic or abiotic

conditions that vary among localities (Amarasekare

2003). Distinct spatial niches lead to the apparent

paradox of competitors that ‘‘coexist’’ at the regional

or metacommunity scale, within which the environment

is heterogeneous, by virtue of being relatively segregated

at the scale of more homogeneous local sites (Levine and

Rees 2002, Amarasekare 2003, Levine and Murrell 2003,

Tilman 2004, Gravel et al. 2006, Kadmon and Allouche

2007). In communities where spatial niche differentiation

is an important mechanism of coexistence, species

composition varies predictably along continuous gradi-

ents of biotic or abiotic conditions that affect relative

competitive abilities (Chave 2004, Tilman 2004). One

weakness of the niche-based perspective, however, is its

failure to address the random or contingent aspects of

community structure. For example, the openness of

many communities to new species has been demonstrated

by seed addition experiments (Tilman 1997, Zobel et al.

2000, Munzbergova 2004, Ehrlen et al. 2006; but see

Turnbull et al. 2000, Wilsey and Polley 2003), linear

local–regional richness relationships (Harrison and

Cornell 2008), and positive native–exotic richness

correlations (Davies et al. 2005, 2007, Sax and Gaines

2008, Stohlgren et al. 2008), evidence that argues against

an overly deterministic view of competitive communities.

Neutral theory (Hubbell 1997) represents an opposite

extreme, in which nonequilibrial coexistence arises from

competitive equality, no niche differences exist, and any

spatial patterning in community composition arises only

from chance and dispersal limitation. Neutral theory is

compatible with the evidence for stochastic aspects of

community structure, and highlights the importance of

random processes in community assembly. However, a

purely neutral perspective is not consistent with the

ubiquitous correlations that are observed between

community composition, species traits, and environ-

mental gradients (Chave 2004, Tilman 2004). In an

attempt to reconcile the niche-based and neutral

viewpoints, Tilman (2004) proposed stochastic niche

Manuscript received 29 April 2009; revised 25 September2009; accepted 19 October 2009. Corresponding Editor: J.Weiner.

1 Corresponding author. E-mail: hvcornell@ucdavis.edu

2141

theory, in which random processes interact with spatial

heterogeneity and niche differences to shape communi-

ties. According to this theory, competition occurs

primarily at recruitment, whereas established adults

are relatively resistant to competitive exclusion. Niche

expansion and community enrichment are possible when

propagule arrival coincides with disturbances that create

pulses of unused resources. Although community

composition should still vary somewhat predictably

along environmental gradients, it is more affected by

chance processes than in the traditional view.

The spatial storage effect (Chesson 2000) has also been

proposed as a general theory of competition in spatially

heterogeneous environments that incorporates niches,

propagule limitation, and stochastic effects. In this

model, spatial heterogeneity permits competitors to

coexist when, first, each species has higher fitness than

the other(s) in some portion of the environment, and

second, each species is most strongly self-limited through

intraspecific competition in the environment where it has

the highest fitness. The second condition is required

because this model implicitly assumes that the most

successful competitor in a widespread habitat (i.e., the

species with the highest overall fitness in the landscape)

could otherwise export unlimited propagules into the

nearby habitats of other species, which would overcome

these species’ local fitness advantages and cause it to

become competitively dominant everywhere. Self-limita-

tion of propagule production by each species within its

optimal habitat is thus necessary for coexistence.

While many empirical studies have demonstrated the

existence of propagule limitation, few have tested

specific theories for how it interacts with spatial

environmental heterogeneity to determine niche bound-

aries, coexistence, and/or the structure of natural

communities (Turnbull et al. 2000, Levine and Murrell

2003, Vandvik and Goldberg 2005). Using common

elements of the neutral, stochastic niche, and spatial

storage effect theories as a conceptual foundation, we

examined coexistence in three ecologically similar

annual tarweeds (Asteraceae, subtribe Madiinae; Carl-

quist et al. 2003) whose distributions are spatially

segregated with partial overlap across a heterogeneous

5003 550-m grassland study site (as well as elsewhere in

the region). Holocarpha virgata is dominant in the most

common habitat, open hillsides, while Calycadenia

pauciflora is most abundant on rocky hilltops, and

Hemizonia congesta is most abundant in clay-rich

bottomlands. We assume that these species are potential

competitors and that their partial segregation contrib-

utes to their coexistence within this heterogeneous

environment. Thus, to better understand their coexis-

tence at this larger scale, we asked what conditions

would allow the widespread Holocarpha to invade and/

or outcompete Calycadenia and Hemizonia within their

more specialized local habitats.

All three theories predict that the arrival of large

numbers of propagules should, at least under some

circumstances, permit Holocarpha to invade and persist

in Calycadenia and Hemizonia habitats. Under stochas-

tic niche theory, an additional prediction is that

propagule arrival must coincide with a pulse of limiting

resources, such as might be provided by an appropri-

ately timed disturbance. Under the spatial storage effect

model, we expect that relaxing dispersal limitation

through experimental propagule addition should allow

Holocarpha not only to persist in Calycadenia and

Hemizonia habitats, but to outcompete these species

within their own habitats (note that this is an underlying

assumption, rather than an explicit prediction, of the

spatial storage effect model). Under neutral theory,

which seems unlikely a priori given the apparent

correlations of these species distributions with environ-

mental gradients, an additional prediction is that

Holocarpha will not only persist in Calycadenia and

Hemizonia habitats but will show equal fitness in all

three habitats. In contrast, under a traditional view of

the spatial niche, even a large influx of propagules will

not allow Holocarpha to invade and persist in Calyca-

denia or Hemizonia habitats, either because it cannot

tolerate the abiotic conditions (fundamental niche) or

because it is competitively inferior in these habitats

(realized niche).

We emphasize that the neutral, stochastic niche, and

spatial storage effect theories are not mutually exclusive,

and that our study is not intended as a definitive test of

all of the assumptions or predictions of any of them,

which would require quantification of model parame-

ters. Our primary focus is on testing the shared

qualitative assumption made by all three of these

theories (as well as others) that relaxation of propagule

limitation can lead to broadening of the spatial niche.

Secondarily, by testing additional predictions of each of

the theories, our goal is to identify the most promising

conceptual framework for better understanding the

controls over spatial distributions in this system.

We used a one-time Holocarpha propagule addition

into Hemizonia and Calycadenia habitats, with and

without disturbance, followed by 3.5 years of observa-

tion of the growth, survival, and recruitment of

transplants and natural controls. As our principal

measure of Holocarpha success under different treat-

ments, we used an index of population-level fitness

(number of mature plants times mean flowers per plant).

We also examined individual fitness components and

yearly population replacement rates.

METHODS

Study system

Our study site at the Donald and Sylvia McLaughlin

University of California Reserve is a heterogeneous

mosaic of vegetation including grasslands, oak wood-

lands, and chaparral (available online).2 The site

2 hhttp://nrs.ucdavis.edu/McL/index.htmli

SUSAN HARRISON ET AL.2142 Ecology, Vol. 91, No. 7

comprises a complex of rocks and soils that includes

serpentine, which is relatively infertile because of low

primary nutrients and Ca, low water availability, and

excess Mg (Alexander et al. 2006). The Ca:Mg ratio is

often considered an indicator of the harshness of

serpentine soils, with ratios �1.0 representing the

harshest soils (Kruckeberg 1984, Harrison 1999). At

our study site, grasslands on shallow rocky serpentine

soils with low Ca:Mg are low in biomass and rich in

native forbs, while those on deeper soils or those with

higher Ca:Mg are dominated by exotic annual grasses

and forbs (Harrison 1999, Davies et al. 2005, Elmendorf

and Moore 2007).

Tarweeds (Madiinae) are a tribe within the Asteraceae

that diversified greatly in the California Floristic

Province and that share a distinctive set of traits,

including resinous glandular foliage, late-summer flow-

ering phenology, and deep roots (Carlquist et al. 2003).

At our study site, as well as elsewhere in California,

tarweeds are among the latest-flowering species (and

often the only abundant native late-flowering species)

within grasslands that are otherwise dominated by

winter-germinating annual grasses and vernal annual

forbs.

Three tarweeds are found at our study site in

grasslands where the soil is partly or completely formed

from serpentine. The most abundant is Holocarpha

virgata, which is widespread on open hillsides and is

sometimes considered a rangeland pest (Perrier et al.

1981, 1982). The others are Calycadenia pauciflora,

which is found on rocky hilltops and is a serpentine

endemic, and Hemizonia congesta, which is found in

clay-rich bottomlands. They flower late in the growing

season (July–September), forming nearly monospecific

stands at a time when few species other than sparse

perennials (e.g., Nassella pulchra, Poaceae and Eriogo-

num nudum, Polygonaceae) are still active. Both

Holocarpha and Hemizonia are known to have the

capacity for multi-year seed dormancy (Perrier et al.

1981, 1982, Gulmon et al. 1983); seed dormancy in

Calycadenia has not been studied. The tarweeds appear

to replace one another, sometimes at sharp boundaries,

along a soil gradient from rocky serpentine hilltops

through hillsides to bottomlands (see Plate 1). Where the

serpentine soil influence ends, Holocarpha and Hemi-

zonia often give way to the exotic yellow starthistle

(Centaurea solstitialis, Asteraceae), which has a similar

late-flowering phenology.

Observational data

To identify niche differences among the three

tarweeds, we used observational data from Davies et

al. (2005; see also Moore and Elmendorf 2006,

Elmendorf and Moore 2007). In a 0.33-km2 area

selected for its substantial heterogeneity, a grid of 132

sampling sites spaced 50 m apart was established in

2000. At each sampling site, presence or absence of all

vascular plant species was recorded annually in four 1-

m2 quadrats from 2001 to 2003. In 2001, one value of

mean soil depth and one bulked soil sample were taken

from each site; samples were analyzed for soil chemistry

at A and L Western Laboratories (Modesto, California,

USA). For our present analyses, we used only the 81

sampling points located in open grassland, excluding the

51 points that were shaded either by dense chaparral or

oak canopy.

Based on previous studies and our observations, we

predicted that soil depth, texture, and Ca:Mg ratio

would distinguish the sites occupied by the three tarweed

species. Specifically, we expected Holocarpha habitats to

be intermediate between the habitats of Calycadenia

(shallow soils, low clay content, low Ca:Mg) and

Hemizonia (deep soils, high clay content, high Ca:Mg).

Of a large suite of other available soil variables,

including NO3-N, NH4-N, Bray-P, cation exchange

capacity, and pH, none had significant predictive power

in this study or others using the same study sites and soil

data (Davies et al. 2005, Moore and Elmendorf 2006,

Elmendorf and Moore 2007; S. Harrison, unpublished

data).

We first used Pearson’s chi-square tests of association

to determine whether occurrences of each pair of the

three tarweeds were significantly negatively associated

with each other. We then used multivariate analysis of

variance (MANOVA) to test for significant differences

in soil depth, percentage clay, and Ca:Mg among the

sites occupied by each species. Replicates were the

occurrence of each species at each site, and species was

the sole independent variable. Separate univariate

ANOVAs were then used to test for differences among

species for each environmental factor. Analyses were

performed in JMP version 5.1 (SAS Institute 2004).

Field experiment

To test whether propagule arrival and/or appropri-

ately timed disturbance would permit Holocarpha to

invade and persist in Calycadenia and Hemizonia

habitats, we conducted a field experiment from fall

2005 to spring 2009. Within the same area studied by

Davies et al. (2005), we established 10 experimental sites

in Calycadenia stands, 10 experimental sites in Hemi-

zonia stands, and 20 natural control sites in Holocarpha

stands. Sites were chosen haphazardly within stands of

relatively uniform and high density of each species, with

the three types of sites well interspersed spatially. At

each of the 40 sites we established one 1 3 2-m plot

comprising eight 0.5 3 0.5-m subplots. At the 20

experimental sites we randomly assigned the eight 0.5

3 0.5-m subplots to two levels of propagule addition

(control, addition) crossed with four levels of distur-

bance (control, fall clearing, fallþ early spring clearing,

fall þ late spring clearing).

The four levels of the disturbance treatment were

designed to ensure that disturbance occurred at what-

ever time it is most effective at promoting Holocarpha

invasion into its competitors’ habitats; we had no

July 2010 2143SPATIAL NICHES AND COEXISTENCE

specific a priori hypothesis about which time of

disturbance would be the most effective. We combined

fall (preemergence) clearing with early and late spring

(postemergence) clearing to ensure that enough Holo-

carpha seedlings would emerge at the beginning of the

experiment (spring 2006) to make further data collection

possible. Fall clearing consisted of clipping and remov-

ing all biomass, which was mostly dead annual grasses,

to within 1 cm of the soil (8 September 2005). Spring

clearing consisted of carefully clipping and removing all

live biomass except forHolocarpha; most of this biomass

was live vernal annuals in the early spring (16 April

2006), and senescing vernal annuals plus live juvenile

tarweeds in the late spring (24 May 2006). Biomass of

vernal annuals consisted predominantly of exotic annual

grasses (e.g., Avena fatua, Bromus hordeaceus, Taenia-

therum caput-medusae).

The propagule addition treatment was designed to

ensure that initial recruitment of Holocarpha in Calyca-

denia and Hemizonia habitats was at least as high as in

its own habitat, to allow an adequate test of treatment

effects. We collected Holocarpha seeds and added them

to the experimental plots by harvesting flower heads

from adjacent Holocarpha stands at a time when seeds

were ripe but had not dropped (27 September 2005). We

added all seed-containing flower heads from 1 3 1 m of

dense Holocarpha to each 503 50-cm experimental plot,

thus producing a seed density approximately four times

higher than a high natural seed rain.

We counted Holocarpha seedlings in March 2006 and

adults in August 2006 in each experimental subplot in

Calycadenia and Hemizonia habitat and in each of four

randomly chosen subplots of the natural control plots

within Holocarpha habitat. In August, we also randomly

chose three Holocarpha adults from each subplot and

measured their heights and numbers of flowers, and

counted and weighed seeds from 10 flowers per plant. In

2007–2009 we again counted seedlings in March. In 2007

and 2008 we again counted and measured adults in

August, and also counted and measured competitor

(Calycadenia and Hemizonia) adults in August (which

we could not do in 2006 because they had been removed

from the disturbed subplots).

Data analyses

To determine the effects of propagule addition and

disturbance on the performance of Holocarpha in

Calycadenia and Hemizonia habitats, we analyzed the

2006 experimental data using a split-plot ANOVA. An

index of population-level fitness (number of adult

Holocarpha 3 mean number of flowers/plant) was the

dependent variable, and disturbance treatment, habitat

(Calycadenia or Hemizonia), and treatment by habitat

interaction were the independent variables.

The performance of Holocarpha was further analyzed

with two additional split-plot ANOVAs using emer-

gence (number of seedlings) and survival (ratio of adult

plants to seedlings) as dependent variables, and the same

independent variables as the population-level fitness

analysis. We could not combine these into a singleanalysis because lack of emergence in some subplots led

to unequal sample sizes (e.g., it is not meaningful tomeasure seedling survival in a plot with no seedlings).

Also, too few plants survived in some treatments toallow a comparison of flowers per plant or othermeasures of adult plant size.

To compare the performance ofHolocarpha in its ownhabitat vs. in Calycadenia and Hemizonia habitats, we

analyzed data from the 20 natural control plots vs. the20 experimental plots in 2006. We used a simple

ANOVA with population-level fitness as the dependentvariable, and habitat (Holocarpha, Calycadenia, or

Hemizonia) as the independent variable.To decompose this second analysis into its demo-

graphic components, we used simple ANOVAs withemergence (numbers of seedlings) and survival (ratio of

adults to seedlings) as dependent variables, and habitat(Holocarpha, Calycadenia, or Hemizonia) as the inde-

pendent variable. We used a simple MANOVA to testthe effect of habitat on measures of plant size (height,

flower number, seeds per flower, seed mass).To determine the persistence of the effects of

disturbance, habitat, and propagule addition, we re-peated the analyses of Holocarpha performance using

the 2007 and 2008 data. We also calculated populationgrowth rate (k2007¼ seedlings 2007/seedlings 2006, k2008

¼ seedlings 2008/seedlings 2007, k2009 ¼ seedlings 2009/

seedlings 2008) as a way to compare populationtrajectories in the three habitats.

To test for competitive effects of introduced Holo-carpha on resident Calycadenia and Hemizonia in 2007

and 2008, we used a simple ANOVA with population-level fitness of the latter two species as the dependent

variable, and population-level fitness of Holocarpha, aswell as its interaction with habitat (Calycadenia or

Hemizonia) as the independent variables.

RESULTS

Observational data

The distributions of the three species overlapped atthe scale of presence or absence in sampling sites (i.e.,

clusters of four 1-m2 quadrats distributed at 50-mintervals across the grassland portion of the 0.33-km2

study area); Holocarpha was present in 45% of samplingsites containing Calycadenia and 32% of sites containing

Hemizonia. Chi-square tests found a significant negativeassociation between Holocarpha and Hemizonia at this

scale (v281 ¼ 0.0246, P ¼ 0.0237), while Calycadenia was

not significantly associated with eitherHolocarpha (v281¼

0.1619, P ¼ 0.1237) or Hemizonia (v281 ¼ 0.0943, P ¼

0.1059).

Soils in quadrats occupied by the three target speciesdiffered significantly in the predicted directions, with

clay content increasing from Calycadenia to Holocarphato Hemizonia sites; also, soil depths were greater inHemizonia sites than Holocarpha or Calycadenia sites

SUSAN HARRISON ET AL.2144 Ecology, Vol. 91, No. 7

(Table 1). Mean soil Ca:Mg ratios also differed in the

predicted direction, with the lowest values (i.e., most

serpentine influence) in Calycadenia sites, but this

difference was not significant.

Effects of experimental disturbance and propagule

addition on Holocarpha success in Calycadenia

and Hemizonia habitats

Throughout the experiment, Holocarpha abundance

was close to zero in subplots in Calycadenia and

Hemizonia habitats to which no Holocarpha seeds were

added, so these control subplots were not used in any

analyses.

Population-level fitness of Holocarpha introduced into

Calycadenia and Hemizonia habitats was significantly

affected by disturbance and habitat in 2006 (Fig. 1).

Fitness was highest in the fall þ early spring clearing

treatment and in Calycadenia habitat, and no interaction

between the habitat effect and the disturbance effect.

This disturbance effect was significant in 2007 but not

2008, while the habitat effect remained significant in

both years (Table 2).

Considering individual fitness components, seedling

emergence in 2006 was not significantly affected by

disturbance or by the interaction between disturbance

and habitat (ANOVA, P . 0.10 for all effects), although

there was marginally higher (P ¼ 0.08) emergence in

Calycadenia than Hemizonia habitat. Survival from

seedling to adult was improved by fall þ early spring

clearing and was higher in Calycadenia than Hemizonia

habitat, with no interaction between these effects

(Tables 3 and 4).

Holocarpha performance in its own habitat vs.

in Calycadenia and Hemizonia habitats

Population-level fitness of Holocarpha in 2006 was

significantly higher in the natural controls in Holocarpha

habitat than in the experimental plots with Holocarpha

seed addition in either Calycadenia or Hemizonia

habitats (Fig. 2; ANOVA, F ¼ 40.1, df ¼ 2, 157, P ,

0.0001; Fisher’s least significant difference test, P ,

0.0001 for comparisons of natural controls and the other

two habitats).

TABLE 1. Site conditions (mean with SE in parentheses) for the three focal species of tarweeds(Asteraceae) at 81 observation sites within the study area at the Donald and Sylvia McLaughlinUC Reserve, California, USA.

Dependentvariable

Holocarpha Hemizonia CalycadeniaF2, 108 P(n ¼ 64) (n ¼ 33) (n ¼ 14)

Ca:Mg 0.99 (0.13) 1.05 (0.27) 0.50 (0.11) 0.0203 0.3372Clay (%) 29.83a (1.58) 35.87b (2.19) 19.43c (1.96) 0.1699 0.0002�Soil depth (cm) 31.23

a(2.36) 41.39

b(3.40) 25.16

a(4.50) 0.0875 0.0108�

Note: Overall variation among site type was significant in a MANOVA (Pillai’s trace F6, 214 ¼4.0383, P¼ 0.0007).

� Rows in boldface highlight factors showing significant variation (P � 0.01) in univariateANOVA models comparing species; pairwise differences are indicated by different superscriptletters.

FIG. 1. Competitive relationships among three ecologicallysimilar annual tarweeds (Asteraceae) on a soil gradient in theDonald and Sylvia McLaughlin University of CaliforniaReserve, showing effects of experimental disturbance atdifferent times of the year on population-level fitness (meanwith 95% CI; number of plants 3 mean flowers/plant) ofHolocarpha planted in Calycadenia and Hemizonia habitat in2006. Different letters above bars indicate significant differencesamong habitats (P , 0.05).

TABLE 2. Effects of experimental disturbance and habitat(Calycadenia or Hemizonia) on population-level fitness ofHolocarpha.

Source SS df F P

2006

Habitat 36.9 1 6.5 0.02Disturbance 20.2 3 6.7 0.03Habitat 3 disturbance 7.9 3 1.3 0.29Plot(habitat) 102.9 18Disturbance 3 plot(habitat) 109.1 54

2007

Habitat 54.8 1 11.7 0.003Disturbance 15.7 3 4.1 0.01Habitat 3 disturbance 7.8 3 2.0 0.12Plot(habitat) 84.4 18Disturbance 3 plot(habitat) 69.4 54

2008

Habitat 65.1 1 18.8 0.0004Disturbance 3.6 3 1.1 0.39Habitat 3 disturbance 3.9 3 1.2 0.35Plot(habitat) 62.4 18Disturbance 3 plot(habitat) 61.2 54

Notes: Sources are from a split-plot ANOVA. Populationfitness is measured as number of plants 3 mean flowers/plant.

July 2010 2145SPATIAL NICHES AND COEXISTENCE

This difference remained significant when comparing

natural controls to just the fall þ early spring distur-

bance plots which exhibited the highest fitness

(ANOVA, F¼ 5.8, df¼ 2, 97, P , 0.005). The difference

between natural and experimental plots remained

significant in 2007 and 2008 (2007, F ¼ 99.7, df ¼ 2,

128; 2008, F ¼ 36.9, df ¼ 2, 96).

Considering individual fitness components, initial

seedling emergence in 2006 was higher in Calycadenia

habitat than in Hemizonia habitat or natural controls,

with no significant difference between the latter two

(Fig. 3; ANOVA, F ¼ 13.7, df ¼ 2, 158, P ¼ 0.0001;

Fisher’s least significant difference test, P , 0.0001 for

comparisons of Calycadenia habitat and the other two

habitats). Survival from seedling to adult was lower in

Hemizonia habitat than in either Calycadenia habitat or

natural controls, with no significant difference between

the latter two (ANOVA, F ¼ 17.2, df ¼ 2, 124, P ¼0.0001; Fisher’s least significant difference test, P ,

0.0001 for comparisons of Hemizonia habitat and the

other two habitats).

All measures of plant size (height, flowers per plant,

seeds per flower, seed mass) were significantly higher in

natural controls than in Calycadenia or Hemizonia

habitats in 2006 (Fig. 4; MANOVA, Pillai’s trace ¼1.7, F ¼ 31.4, df ¼ 412, 303, P , 0.0001, followed by

univariate F tests, all P , 0.0001). The height and flower

number differences remained significant in 2007 (MAN-

OVA, Pillai’s trace ¼ 1.4, F ¼ 80.6, df ¼ 6, 256, P ,

0.0001) and 2008 (MANOVA, Pillai’s trace ¼ 1.5, F ¼128.6, df ¼ 6, 192, P , 0.0001); we did not repeat the

seed number or seed mass measurements in 2007 or

2008.

Long-term persistence of Holocarpha in Calycadenia

and Hemizonia habitats

During the first year following establishment, Holo-

carpha populations increased significantly in the natural

controls (k2007 ¼ seedlings 2007/seedlings 2006 . 1.0),

but decreased significantly in the other two habitats

(k2007 , 1.0; one-sample t tests, P , 0.001). Populationsize change (k2007) was not significantly affected by the

2005 disturbance treatments (ANOVA, N¼ 80 subplots,

F ¼ 0.68, df ¼ 3, 54, P ¼ 0.41). Population size changewas significantly higher in natural controls than in

Calycadenia habitat, and in Calycadenia habitat than

Hemizonia habitat (Fig. 5; ANOVA, N¼ 155, F¼ 30.0,

df ¼ 2, 152, P ¼ 0.0001; Fisher’s least significantdifference test, P ¼ 0.0001 for natural controls vs.

Calycadenia habitat and P ¼ 0.035 for Calycadenia vs.

Hemizonia habitats).

In the second post-establishment year, populationsdeclined significantly in all habitats (k2008 ¼ seedlings

2008/seedlings 2007 , 1.0; one-sample t tests, P ,

0.001). Population change (k2008) was again unaffectedby the 2005 disturbance treatments (ANOVA, N¼ 45, F

¼ 0.055, df ¼ 3, 37, P ¼ 0.39) and was not significantly

different among the three habitats (Fig. 5; ANOVA, N¼124, F ¼ 1.634, df ¼ 2, 121, P ¼ 0.51). Sample sizes arelower than for the first post-establishment year because

populations in some subplots (mainly Calycadenia and

Hemizonia subplots) went permanently extinct.

In the third post-establishment year, populationsincreased significantly in the natural controls (k2009 ¼seedlings 2009/seedlings 2008 . 1.0; one-sample t test, P

, 0.001), and again decreased in Calycadenia andHemizonia habitats, but not significantly so (k2009 ,

1.0; one-sample t tests, P . 0.264 and 0.173, respective-

ly). Population change was significantly higher in

natural controls than in Calycadenia and Hemizonia

TABLE 3. Effects of experimental disturbance and habitat(Calycadenia or Hemizonia) on survival of Holocarpha, 2006.

Source SS df F P

Habitat 4.0 1 15.6 0.0009Disturbance 1.8 3 5.5 0.002Habitat 3 disturbance 0.1 3 0.2 0.87Plot(habitat) 4.6 18Disturbance 3 plot(habitat) 5.9 54

Notes: Sources are from a split-plot ANOVA. Survival ismeasured as number of adults/number of seedlings.

FIG. 2. Population-level fitness (mean with 95% CI; numberof plants 3 mean flowers/plant) of Holocarpha in naturalcontrols in Holocarpha habitat vs. in experimental plots inCalycadenia and Hemizonia habitat in 2006. Different lettersabove bars indicate significant differences among habitats (P ,0.05). For simplicity, all disturbance treatments are combinedhere and in Figs. 3–5.

TABLE 4. Pairwise contrasts of disturbance treatment using Fisher’s LSD test.

Season ofdisturbance

Control Fall Fall þ early spring

Meandifference Probability

Meandifference Probability

Meandifference Probability

Fall �0.09 0.52Fall þ early spring 0.28 0.04 0.37 0.01Fall þ late spring �0.08 0.58 0.01 0.92 �0.36 0.01

SUSAN HARRISON ET AL.2146 Ecology, Vol. 91, No. 7

habitat but was not different between the latter two (Fig.

5; ANOVA, N¼ 113, F¼ 13.94, df¼ 2, 110, P¼ 0.0001;

Fisher’s least significant difference test, P ¼ 0.0001 for

natural controls vs. Calycadenia and Hemizonia habitat

and P ¼ 0.73 for Calycadenia vs. Hemizonia habitats).

Population change (k2009) was again unaffected by the

2005 disturbance treatments (ANOVA, N ¼ 36, F ¼0.397, df¼ 6, 29, P¼ 0.88). Sample sizes are again lower

than for the first and second post-establishment years

because additional populations in some Calycadenia and

Hemizonia subplots went permanently extinct.

Frequencies of Holocarpha seedlings in experimental

Calycadenia plots declined from 100% in 2006 to 98%,

68%, and 70% in 2007–2009, respectively; frequencies of

Holocarpha seedlings in experimental Hemizonia plots

changed from 100% in 2006 to 14%, 10%, and 18% in

2007–2009, respectively.

Competitive effects of Holocarpha on Calycadenia

and Hemizonia

In 2007, there was no evidence for an effect of the

population-level fitness of Holocarpha on the popula-

tion-level fitness of its two competitors, or vice versa

(ANOVA, P ¼ 0.79), and no significant interaction

between the effects of Holocarpha and species (Calyca-

denia vs. Hemizonia). However, a trend toward compe-

tition was suggested by negative correlations between

the numbers of Holocarpha and the mean heights and

flower numbers of its competitors (Pearson r ¼�0.31,�0.33), as well as between the number of Holocarpha

and the mean heights and flower numbers of Holocarpha

(Pearson r ¼�0.24,�0.20).In 2008, there was a significant positive association

between the population-level fitness of Holocarpha and

those of its competitors (ANOVA, F¼19.1, df¼51, 1, P

,0.0001), with no significant interaction between the

effects of Holocarpha and species (Calycadenia vs.

Hemizonia). Numbers of Holocarpha and the mean

heights and flower numbers of its competitors were less

negatively correlated than in 2007 (Pearson r ¼ �0.01

and�0.19, respectively), and the numbers of Holocarpha

and the numbers of its competitors were positively

correlated (Pearson r ¼ 0.44). Negative correlations

persisted between the numbers of Holocarpha and the

mean heights and flower numbers of Holocarpha

(Pearson r ¼�0.22, �0.34).

DISCUSSION

Species occurrence and environmental data supported

our initial observations that the three tarweed species

have different although partly overlapping spatial

niches. Distributions of the two more common species,

Holocarpha and Hemizonia, showed a significant nega-

tive association with one another at the scale of four 1-

FIG. 3. Number of Holocarpha seedlings emerging (meanwith 95% CI) per 0.25-m2 subplot in natural controls inHolocarpha habitat vs. in experimental subplots (approximatelyequal numbers of seeds added in all experimental seed additionsubplots) in Calycadenia and Hemizonia habitat in 2006.Different letters above bars indicate significant differencesamong habitats (P , 0.05).

FIG. 4. Height, number of flowers/plant, number of seeds/flower, and seed mass (mean with 95% CI) of Holocarpha innatural controls in Holocarpha habitat vs. in experimentalplots in Calycadenia or Hemizonia habitat in 2006. Differentletters above bars indicate significant differences amonghabitats (P , 0.05).

July 2010 2147SPATIAL NICHES AND COEXISTENCE

m2 sampling units spaced 50 m apart. The distribution

of the less common species, Calycadenia, was not

significantly negatively associated with those of the

other two species, perhaps because of its small number

of occurrences (N¼ 14). However, soil differences were

found to separate Calycadenia, Holocarpha, and Hemi-

zonia sites. As we expected, soil texture and depth are

key indicators that differentiate the habitats of the three

species along the gradient from rocky hilltops to

bottomlands.

Consistent with recent theories emphasizing the role

of propagule limitation within a spatially heterogeneous

environment, we found that the addition of a large

quantity of seeds allowed Holocarpha virgata to

establish and persist for several generations in habitats

dominated by its competitors Calycadenia pauciflora and

Hemizonia congesta. Consistent with stochastic niche

theory (Tilman 2004), in particular, the ability of

Holocarpha to establish in Calycadenia and Hemizonia

habitats depended on appropriately timed disturbance.

In both of these habitats, the critical disturbance was the

removal of competing vernal annual plants in early

spring when Holocarpha seedlings were small, whereas

there was little effect of removal of either dead biomass

in the fall, or senescing vernal annuals plus live summer-

flowering annuals in the late spring. Under the

framework of stochastic niche theory, these results

suggest that in undisturbed Calycadenia and Hemizonia

habitats, established vernal annuals play a key role in

preventing enough resources from being available to

allow seedlings of Holocarpha to reach maturity. Also

consistent with this theory, the ability of Holocarpha to

establish in these two habitats appeared to be facilitated

by its attainment of reproductive maturity at very small

sizes (Fig. 4).

While the effect of early spring disturbance on

Holocarpha establishment was significant in both

Calycadenia and Hemizonia habitats, there was a trend

toward a weaker effect in rocky, unproductive Calyca-

denia habitat than in more productiveHemizonia habitat

(Fig. 1). These results support the idea that stochastic

aboveground disturbance plays a greater role in

facilitating community enrichment in productive than

unproductive environments, an idea that has received

previous experimental support in the same grasslands

(Elmendorf and Moore 2007) as well as others (e.g.,

Foster 2001).

Our results suggest that coexistence among these

species is not well explained by the spatial storage effect

model (Chesson 2000), in which the dominant compet-

itor in the most common habitat can outcompete all

other species if the propagule limitation imposed by

intraspecific competition within its optimal habitat is

relaxed. We found that, although an abundant propa-

gule influx did allow Holocarpha to establish in

Calycadenia and Hemizonia habitats, there was no

evidence that it suppressed the abundances of either

species. This is not surprising, given that Holocarpha

individuals were considerably smaller at maturity in

Calycadenia and Hemizonia habitats than in Holocarpha

habitat (Fig. 4). For our system, then, only the first

condition of the spatial storage effect model (distinct

spatial niches) appears to be required for coexistence;

the second condition (self-limitation of the dominant

competitor within the optimal habitat) appears not to be

necessary. The spatial storage effect model assumes that

spatial environmental heterogeneity affects the relative

fitnesses of competing species but does not change their

per capita competitive effects on one another (e.g., the

aij’s of the Lotka-Volterra model). However, per capita

competitive effects for plants may depend on relative

plant sizes (Weiner 1985), and thus it may be common

for a species to be a strong competitor in its optimal

habitat where it is large, but a weak one in suboptimal

habitats where it is smaller.

As we expected, the results were inconsistent with

neutral theory (Hubbell 1997) because niche differences

among the three species are clearly present. Despite its

ability to invade and persist in Calycadenia and Hemi-

zonia habitats when seeds were added, Holocarpha was

not very successful in these habitats. The rates of

population replacement were ,1.0 in both a good year

(2006–2007) when the natural Holocarpha population

grew, and a bad year (2007–2008) when the natural

population declined. This evidence for strong spatial

niche differences is reinforced by the lower rates of

survival, growth, and reproduction we found for

Holocarpha in Calycadenia and Hemizonia habitats

compared with its own habitat, and by the significant

differences in soils among the habitats occupied by the

three species.

Holocarpha appears to be excluded from its compet-

itors’ habitats primarily by abiotic factors because its

performance in these habitats was low even when

competitors were excluded, but also by diffuse compe-

FIG. 5. Interannual rate of population change (k, meanwith 95% CI) for Holocarpha in natural controls in Holocarphahabitat (N ¼ 80), vs. in experimental plots in Calycadenia orHemizonia habitat (N ¼ 40 for each): k2007 ¼ seedlings 2007/seedlings 2006, k2008 ¼ seedlings 2008/seedlings 2007; k2009 ¼seedlings 2009/seedlings 2008. Confidence intervals crossing thedashed line indicate changes not significantly different fromstasis (k¼ 1). Different letters above bars indicate significantlydifferent changes among habitats (P , 0.05).

SUSAN HARRISON ET AL.2148 Ecology, Vol. 91, No. 7

tition because its performance in these habitats was even

lower when competitors were present. The apparent

inability of Holocarpha to persist in the habitats

dominated by the other two species appears more

consistent with the traditional Hutchinsonian niche

concept than with any of the more recent theories we

examined. Competitive outcomes and community as-

sembly in this system appear to be largely deterministic

in space and time. Based on these results, we would

expect the distributions of the three tarweeds to remain

fairly constant, although with minor changes over time

due to random dispersal events and recurring distur-

bances such as gopher activity and grazing. These results

are consistent with reviews by Turnbull et al. (2000) and

Levine and Murrell (2003), suggesting that altering the

seed rain within realistic bounds often has little long-

term effect on community composition. Our results also

support the prediction by Chave (2004) that community

structure will be relatively predictable in cases where a

small number of abundant species interact in an

environment with strong spatial heterogeneity.

We conclude that there is a strong case for a

deterministic view of competitive coexistence in het-

erogeneous environments. Nonetheless, our results

also have some implications for the stochastic aspects

of community assembly. The experimental populations

of our annual study species will likely persist for a few

more years (generations), with this period of transient

persistence facilitated by their capacity for long-term

seed dormancy. In communities where disturbance and

propagule movement are frequent and established

individuals are long-lived, species richness may be

enhanced for long periods of time by transients such as

our experimental Holocarpha populations. This pre-

sents a possible explanation for the paradox of

community unsaturation (Tilman 2004, Harrison and

Cornell 2008, Stohlgren et al. 2008): why are native

species absent from communities and yet able to

persist when they are added? A species may be

excluded from a community by its long-term inability

to replace itself, but may still, if added, contribute to

that community’s diversity for an ecologically mean-

ingful length of time.

PLATE 1. Tarweed habitat at the Donald and Sylvia McLaughlin University of California Reserve. Calycadenia paucifloradominates the rocky hilltop, Holocarpha virgata the hillside, and Hemizonia congesta the bottomland in the foreground. Photocredit: S. Harrison.

July 2010 2149SPATIAL NICHES AND COEXISTENCE

ACKNOWLEDGMENTS

For helpful discussions and manuscript comments, we aregrateful to P. Amarasekare, B. Anacker, J. Baty, P. Chesson, K.Davies, S. Elmendorf, J. Levine, B. Melbourne, and S. Veloz.For use of the Grid data we thank B. Inouye and K. Davies.For logistical support we thank the Donald and SylviaMcLaughlin University of California Natural Reserve.

LITERATURE CITED

Alexander, E. A., R. G. Coleman, T. Keeler-Wolf, and S.Harrison. 2006. Serpentine geoecology of western NorthAmerica. Oxford University Press, Oxford, UK.

Amarasekare, P. 2003. Competitive coexistence in spatiallystructured environments: a synthesis. Ecology Letters 6:1109–1122.

Carlquist, S. J., B. G. Baldwin, and G. D. Carr. 2003.Tarweeds and silverswords: evolution of the Madiinae(Asteraceae). Missouri Botanical Garden Press, St. Louis,Missouri, USA.

Chave, J. 2004. Neutral theory and community ecology.Ecology Letters 7:241–253.

Chesson, P. 2000. General theory of competitive coexistence inspatially-varying environments. Theoretical Population Biol-ogy 58:211–237.

Davies, K. F., P. Chesson, S. Harrison, B. D. Inouye, B. A.Melbourne, and K. J. Rice. 2005. Spatial heterogeneityexplains the scale dependence of the native–exotic diversityrelationship. Ecology 86:1602–1610.

Davies, K. F., S. Harrison, H. D. Safford, and J. H. Viers. 2007.Productivity alters the scale dependence of the diversity–invasibility relationship. Ecology 88:1940–1947.

Ehrlen, J., Z. Munzbergova, M. Diekmann, and O. Eriksson.2006. Long-term assessment of seed limitation in plants:results from an 11-year experiment. Journal of Ecology 94:1224–1232.

Elendorf, S. C., and K. A. Moore. 2007. Plant competitionvaries with community composition in an edaphicallycomplex landscape. Ecology 88:2640–2650.

Foster, B. L. 2001. Constraints on colonization and speciesrichness along a grassland productivity gradient: the role ofpropagule availability. Ecology Letters 4:530–535.

Gravel, D., C. D. Canham, M. Beaudet, and C. Messier. 2006.Reconciling niche and neutrality: the continuum hypothesis.Ecology Letters 9:399–409.

Gulmon, S. L., N. R. Chiariello, H. A. Mooney, and C. C. Chu.1983. Phenology and resource use in three co-occurringgrassland annuals. Oecologia 58:33–42.

Harrison, S. 1999. Native and alien species diversity at the localand regional scales in a grazed California grassland.Oecologia 121:99–106.

Harrison, S., and H. V. Cornell. 2008. Toward a betterunderstanding of regional influences on local communityrichness. Ecology Letters 11:969–979.

Hubbell, S. 1997. A unified theory of biogeography and relative

species abundance and its application to tropical rainforestsand coral reefs. Coral Reefs 16:S9–S21.

Kadmon, R., and O. Allouche. 2007. Integrating the effects ofarea, isolation, and habitat heterogeneity on species diversity:a unification of island biogeography and niche theory.American Naturalist 170:443–454.

Kruckeberg, A. R. 1984. California serpentines: flora, vegeta-tion, geology, soils and management problems. University ofCalifornia Press, Berkeley, California, USA.

Levine, J. M., and D. J. Murrell. 2003. The community-levelconsequences of seed dispersal patterns. Annual Review ofEcology and Systematics 34:549–574.

Levine, J. M., and M. Rees. 2002. Coexistence and relativeabundance in annual plant assemblages: the roles ofcompetition and colonization. American Naturalist 160:452–467.

Moore, K. A., and S. C. Elmendorf. 2006. Propagule vs. nichelimitation: untangling the mechanisms behind plant species’distributions. Ecology Letters 9:797–804.

Munzbergova, Z. 2004. Effect of spatial scale on factorslimiting species distributions in dry grassland fragments.Journal of Ecology 92:854–867.

Perrier, G. K., W. A. Williams, and J. W. Menke. 1982.Tarweed an unloved annual-type range plant. Rangelands4(4):149–150.

Perrier, G. K., W. A. Williams, and S. R. Radosevich. 1981.Managing range and pasture to suppress tarweed. CaliforniaAgriculture 33(1):18–19.

Sax, D. F., and S. D. Gaines. 2008. Species invasions andextinction: the future of native biodiversity on islands.Proceedings of the National Academy of Sciences USA105:11490–11497.

Stohlgren, T. J., D. T. Barnett, C. S. Jarnevich, C. Flather, andJ. Kartesz. 2008. The myth of plant species saturation.Ecology Letters 11:313–322.

Tilman, D. 1997. Community invasibility, recruitment limita-tion, and grassland biodiversity. Ecology 58:840–849.

Tilman, D. 2004. Niche tradeoffs, neutrality, and communitystructure: a stochastic theory of resource competition,invasion, and community assembly. Proceedings of theNational Academy of Sciences USA 101:10854–10861.

Turnbull, L. A., M. J. Crawley, and M. Rees. 2000. Are plantpopulations seed-limited? A review of seed sowing experi-ments. Oikos 88:225–238.

Vandvik, V., and D. E. Goldberg. 2005. Distinguishing theroles of dispersal in diversity maintenance and diversitylimitation. Folia Geobotanica 40:45–52.

Weiner, J. 1985. Size hierarchies in experimental populations ofannual plants. Ecology 66:743–752.

Wilsey, B. J., and H. W. Polley. 2003. Effects of seed additionsand grazing history on diversity and productivity ofsubhumid grasslands. Ecology 84:920–931.

Zobel, M., M. Otsus, J. Liira, M. Moora, and T. Mols. 2000. Issmall-scale species richness limited by seed availability ormicrosite availability? Ecology 81:3274–3282.

SUSAN HARRISON ET AL.2150 Ecology, Vol. 91, No. 7