Species Interactions in a Successional Grassland. II. Colonization of Vegetated SitesAuthor(s): D. R. PeartSource: Journal of Ecology, Vol. 77, No. 1 (Mar., 1989), pp. 252-266Published by: British Ecological SocietyStable URL: http://www.jstor.org/stable/2260928Accessed: 20/08/2010 00:05

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available athttp://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unlessyou have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and youmay use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained athttp://www.jstor.org/action/showPublisher?publisherCode=briteco.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printedpage of such transmission.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact support@jstor.org.

British Ecological Society is collaborating with JSTOR to digitize, preserve and extend access to Journal ofEcology.

http://www.jstor.org

Journal of Ecology (1989), 77, 252-266

SPECIES INTERACTIONS IN A SUCCESSIONAL GRASSLAND. II. COLONIZATION OF VEGETATED SITES

D. R. PEART

Department of Biological Sciences, Dartmouth College, Hanover, NH 03755, U.S.A.

SUMMARY (1) The colonization and inhibition abilities of species were identified as two essential

and distinct components of competitive ability. These components can be quantified to evaluate species interactions and predict population and community change in a community where changes in abundance depend on colonization processes.

(2) Colonization and inhibition abilities were quantified and compared in a patchy Californian grassland community. This was accomplished with an experimental study of the colonization of vegetated sites, using natural seed-rain densities.

(3) The four most abundant perennial species, Anthoxanthum odoratum, Holcus lanatus, Deschampsia holciformis and Rytidosperma pilosum, differed greatly in colonizing ability, as measured by survival, growth and reproduction. All perennial species effectively inhibited colonization of their stands, but annual vegetation could be colonized by Anthoxanthum, and to a lesser degree by Holcus.

(4) Species differences in colonizing ability depended in part on their different densities of seed rain, but there were major differences in intrinsic, per-seed colonizing ability as well.

(5) In contrast, species differences in inhibition ability could be explained by above- ground biomass alone; intrinsic differences among species in their per-unit biomass inhibition effects were not important.

(6) The findings led to definite predictions about successional change in undisturbed sites. These methods and results may have some quite general application to the analysis of population and community dynamics.

INTRODUCTION

The resources required for growth are common to most plant species, and to all stages of the vegetative part of the life cycle. Therefore, there are relatively few opportunities, compared with animal populations, for avoidance of competition by resource partition- ing (Harper 1977). This is probably especially true among canopy-dominant species that coexist in the same stratum. Antonovics & Levin (1980) suggested that it may be because the nature of plant competition seems obvious that there are so few field studies that demonstrate interspecific competition. Although competition in a closed canopy is likely, it cannot be assumed. Further, the simple demonstration of competition does not provide a measure of the influence of one species on another. In this study, several field experiments were performed, with the overall goal of predicting population dynamics and successional change in a closed canopy grassland community.

To measure the influence of competitive effects on population dynamics, it is necessary to focus on those interactions that are quantitatively most important under natural field conditions. Plants respond to competition from near neighbours (Mack & Harper 1977; Weiner 1982, 1984; Silander & Pacala 1985). The neighbourhoods experienced by seeds and seedlings may determine the distributions and abundances of adult plants, as emphasized by Grubb (1977) and demonstrated in the field by many workers (e.g. Gross

252

D. R. PEART 253

& Werner 1982; Platt 1975). In communities that include perennials, seeds and seedlings frequently encounter adult plants in their neighbourhoods, and are usually inhibited by them (e.g. Tripathi & Harper 1973). Whether a population increases or decreases in abundance during succession depends, therefore, not only on the ability of its seedlings to establish and grow in the presence of adult plants, but also on the ability of adults to hold sites by suppressing establishment and growth of seedlings under their own canopies.

Except in communities of annuals with similar phenologies, or those consisting of widely spaced individuals, two distinct components of competitive ability must affect establishment. These will be referred to as colonizing ability (of seeds and seedlings) and inhibition ability (of adults). Colonizing ability is defined here as the number of recruits, leaf area and seed production of a species that has been introduced into quadrats in established vegetation. Changes in the number of individuals indicate survival from the introduced seed population; changes in individual size indicate growth rate. Total leaf area per quadrat integrates survival and growth rate and is a better measure than cover to represent the vegetative development of a colonizing population whose cover is too low to estimate accurately. If two species, A and B, differ in their colonizing abilities, they will differ in survival, growth and seed production in stands of species C (where C may be the same as A or B).

Inhibition ability of a species A is defined here as the amount (or proportion) by which A reduces the colonization success of another species B. This can be measured as the difference between the colonization success of B in a stand of A, and the performance of B when introduced to a similar site where a stand of A has been removed. This quantitative definition is consistent with the qualitative use of the term inhibition by Connell & Slatyer (1977). The measure can accommodate facilitation effects (Connell & Slatyer 1977), which would be expressed simply as negative values of inhibition. Inhibition abilities can be compared among species without evaluating absolute inhibition abilities (i.e. without experimentally removing vegetation). Thus, if two species, A and B, differ in their inhibition abilities, stands of these species will reduce the survival, growth and seed production of a third species by different amounts. Only relative inhibition abilities were evaluated in this study.

Reported here are the results of experiments measuring the colonizing abilities of the four dominant perennial species (all bunchgrasses) and the most abundant annual grass in the intact canopies of a successional grassland. Inhibition by established vegetation was ranked among vegetation types, including nearly-monospecific stands of three of the perennial grasses. Colonizing and inhibition abilities are likely to be very important in this community. Although various forms of disturbance to the canopy, including the death of canopy individuals, can modify colonizing abilities (Peart 1989b), most of the vegetation consists of closed canopy. Therefore, seeds and seedlings encounter perennial or annual vegetation in almost all sites. Moreover, these species of bunchgrasses grow as discrete individuals, most being much smaller than the maximum size of 60 cm in diameter. Increase in cover must depend largely on establishment and subsequent growth of new individuals from seed.

The grassland, which experiences a mediterranean climate, is at Sea Ranch, on the north coast of California, U.S.A. Sheep were removed from the study area in 1968, and post-grazing succession has resulted in an increase in the relative cover of perennial grasses (Foin & Hektner 1986). The most cover-dominant species are Anthoxanthum odoratum L., Holcus lanatus L., Deschampsia holciformis Presl. and Rytidospermapilosum (R.Br.) Connor et Edgar. The annual grass, Vulpia bromoides S. F. Gray, is the next most

254 Colonization of vegetated sites in grassland

abundant species. For brevity, these grasses will be referred to by their generic names only. Because species have very patchy distributions, the grassland can be divided easily into four distinctly different patch types. A 'patch' is an area with recognizable boundaries that is dominated (> 85% cover) by one of three perennial species, Anthoxanthum, Holcus or Deschampsia (referred to as perennial patches), or by Vulpia and other annuals in association with Rytidosperma (referred to as annual patches). The term 'resident' refers to the dominant species in each type of patch, to distinguish it from the species that were experimentally introduced. Peart (1 989a) has reported the spatial pattern of seed rain and recruitment in this community, and included a detailed description of the study site.

METHODS

Experimental design Seeds of Anthoxanthum, Holcus, Deschampsia, Rytidosperma and Vulpia were collected

in the field. Seeds were introduced at natural densities (see below) into the four common vegetation patch types (Table 1). The vegetation was undisturbed except for the possible effects of wild grazers (mainly deer, meadow voles and grasshoppers). Because it is the most abundant species and has increased in cover in recent years (Foin & Hektner 1986), emphasis was placed on Anthoxanthum, both as an experimentally introduced species and as a resident species. Also, because it seemed likely that sites in annual vegetation might be more easily colonized than those in perennial vegetation, treatments involving introduc- tions of seeds into annual patches were emphasized. Three patches of each vegetation type were used in the experiments. Two patches of each type, referred to as typical biomass patches, were similar in biomass and most representative of the grassland. A third patch of each type, referred to as a low biomass patch, was also examined. Although these lower biomass areas are much less common in the grassland, they may be important sites for colonization and establishment.

Accordingly, all treatments involving Anthoxanthum, either as a resident or as an introduced species were included, and each of the five species was introduced into the annual patch vegetation. The ability of each species to establish in its own patch type (i.e. where it is the resident species) was also evaluated (Table 1). Seed introductions were done in all three patches of each vegetation type. The exceptions were Holcus and Deschampsia introductions into annual patch vegetation; their seeds were introduced into only one annual patch of typical biomass. Thus, the experiment was conducted in a total of twelve patches (four types, three of each).

TABLE 1. The treatments (combinations of species and patch types) included in seed introduction experiments in a grassland at Sea Ranch, California.

Patch type Anthoxanthum Holcus Deschampsia

Species introduced odoratum lanatus holciformis Annual

Anthoxanthum odoratum X* X X Xt Holcus lanatus X X* Xt Deschampsia holciformis X X* X Rytidosperma pilosum X X* Vulpia bromoides X X*

* Seeds not introduced for resident species; seed input for resident species was from natural seed fall.

t An introduction done in both 1980 and 1981. All other introductions were done in 1980.

D. R. PEART 255

In each of the twelve patches, quadrats were representative of the undisturbed vegetation and classified into five blocks, according to their position in the patch. For each of the combinations of species introduced and vegetation type in Table 1, seeds were introduced into five replicate 50 cm x 50 cm quadrats (one in each block), in three patches of the appropriate patch type. There were also monitored quadrats of the same size where no seeds were introduced, one per block in each patch. Quadrats were allocated at random to treatments and monitoring, and were permanently marked with metal stakes.

In the summer of 1980, seeds were introduced evenly over a 60 cm x 60 cm area covering each quadrat, providing a buffer zone 5-cm wide to avoid any possible edge effects. Three measures of colonization success per quadrat were taken: number of recruits, total leaf area and seed production. At the end of the 1981 growing season, the numbers of surviving individuals were counted in each quadrat. For each perennial recruit, the length and width of each leaf were measured. Seed production of the same individuals was estimated in 1982 by measuring the length of each seed head, and applying linear regressions of seed head length v. number of seeds per head (correlation coefficients for all species were greater than 0-9). Leaf area was not recorded for Vulpia, because seed output is the ecologically important measure of production for annuals. Furthermore, the leaf area production of an annual is not directly comparable to that of perennials because of the different allocation patterns of annuals and perennials.

Natural recruitment was recorded as number of individuals per quadrat without introduced seed, and reported elsewhere (Peart 1989a). Here the natural recruitment data are extended to include individual size (leaf area) for comparison with the results of the introduction experiments. Anthoxanthum and Holcus introductions into annual patch vegetation were repeated in 1981, to provide some information on yearly variation in the experimental results. Numbers of survivors, leaf areas and seed production were measured for the 1981 introductions in 1982. Thus, seed production was recorded over two years of growth for the 1980 introductions, but over only one year for the 1981 introductions.

The percentage cover of each species present in all experimental quadrats was estimated visually late in the 1980 growing season, with the aid of a 0-5 m x 0-5 m quadrat frame divided into nine equal parts. Mean relative cover values of the resident species in the experimental quadrats were Anthoxanthum 94% (in Anthoxanthum patches), Holcus 92% (in Holcus patches), Deschampsia 88% (in Deschampsia patches), Rytidosperma 33% and Vulpia 23% (both in annual patches). Other species contributing more than 1 % in one or more perennial patch types were Rubus ursinus C. & S., Rumex acetosa L., and Iris douglasiana Herbert. In annual patches, several species of forbs and annual grasses were present (Peart 1989a).

Seed collections and introductions Natural seed-fall densities are spatially variable (Peart 1989a) and differ greatly among

species. The number of seeds introduced for each treatment was chosen to approximate the highest natural levels of immigration by seeds from outside a patch to sites within a patch. At an idealized, sharp boundary between two patches, the average seed fall density of either resident species would be 50% of the seed fall in the interior of the patch where it is resident. Thus, densities for the introductions were calculated as follows. Seed fall densities for the resident species were measured (Peart 1989a) in the interiors of the same twelve patches used in the introduction experiment. A mean density for each resident species was obtained from the ten sampling stations with highest seed-fall densities out

256 Colonization of vegetated sites in grassland

of a total of twenty-one stations in each patch type. An exception was made for Rytidosperma, which, although it occurs almost entirely in the annual patches, is variable in its relative cover in those patches. Ten sampling stations were placed in areas of high Rytidosperma cover, and the mean seed-rain density from these samples were used to estimate the highest natural seed rain of Rytidosperma. For all five species, the densities of seeds introduced in the experiments were 50% of the mean densities calculated as described above. The introduced numbers m-2 were Anthoxanthum 38872, Holcus 49 768, Deschampsia 4424, Rytidosperma 7484 and Vulpia 21 668.

Natural seed fall was not excluded from the experimental quadrats. The densities of seeds introduced were adjusted, where necessary, to take account of any natural seed input of the species introduced. The five species of this study all have very limited dispersal, with few seeds travelling more than a few metres from the parent (Peart 1982). If there were any individuals of the introduced species within dispersal distance of the quadrats where seeds were introduced, their contributions to seed fall in the quadrats was estimated. This was done by measuring their fecundities and distances from the quadrats, and applying empirically derived relations between seed-fall density, the distance and direction from the parent plant (Peart 1982). The densities for introductions were then reduced accordingly, but in no quadrat was more than a 7% reduction necessary.

Seeds for the introduction experiments were collected randomly from plants through- out the habitats occupied by each species. Seeds were thoroughly mixed and random samples of 100 were weighed to estimate seed weights. Mean seed weights (mg) were Anthoxanthum 0 523 (S.E. 0 008), Holcus 0 318 (S.E. 0-005), Deschampsia 0 290 (S.E. 0-005), Rytidosperma 1 079 (S.E. 0019) and Vulpia 1 429 (S.E. 0-024). Seed lots were then weighed for the introductions, which were done in the late summer, near the end of the natural dispersal period.

Analyses Colonization success data were log transformed to homogenize variances. The analyses

were simplified by the lack of any consistent spatial variation in the outcomes within or between the patches of any patch type, except between patches that were chosen a priori at different biomass levels. First, there was no effect of location within patches (i.e. no significant block effect). Second, there was no significant effect of patch in any vegetation type, when the results for the two typical biomass patches of each type were compared with a two-way analysis of variance (ANOVA), grouped by patch and by the identity of the introduced species. Therefore, the data from the two typical biomass patches in each vegetation type were pooled for all further analyses. The low biomass patches were analysed separately.

Colonizing abilities were compared among species in two ways. First, the colonization success data were analysed without adjustment for the different densities of seeds introduced for each species. These data were appropriate for analysing current successional trends in the community, because the absolute and relative densities of natural seed fall differed among species. The outcomes of the introductions can, therefore, be attributed partly to differences among species in their seed rain densities, and partly to differences in species-specific inhibition and colonizing abilities. However, these factors can be separated because the results indicate that density-dependent interactions among the introduced plants were highly unlikely. Therefore, in a second analysis, colonizing abilities were compared among species on a per-seed basis, to see whether differences in seed production alone could account for differences in colonization success.

D. R. PEART 257

Differences in inhibition abilities among vegetation types were evaluated by comparing colonization success of a test species among those vegetation types. Inhibition abilities were compared using two test species, Anthoxanthum and Holcus. Rytidosperma and Deschampsia had low colonization success in all vegetation types. Consequently, they were not used as test species in the inhibition analysis. Total leaf area per quadrat, which incorporates both number of survivors and individual size, was the measure of colonization success used to evaluate inhibition.

Analyses of inhibition, and of interaction effects between patch type and species introduced, were done only on the more comprehensive data from the 1980 introductions. Total leaf area per quadrat, which incorporates both number of survivors and individual size, was the measure of colonization success used for the inhibition and interaction analyses. Inhibition effects were then examined further, to see whether differences among vegetation types could be explained by differences in biomass alone. The effect of vegetation biomass on the colonization success of Anthoxanthum in each patch type was analysed with non-linear regression.

Colonizing abilities were compared among the species introduced, using one-way ANOVA within patch type. Inhibition by resident species was compared among patch types using one-way ANOVA for each introduced (test) species. Where the results of the ANOVAS were significant, Tukey's multiple-comparison test for differences among treatments was applied. To test for interactions between species introduced and vegetation type, two-way ANOVAS were performed on parts of the experimental design. A two-way analysis could not be applied to the whole design because some combinations were omitted (Table 1).

Other statistical methods used are stated in the text. Analyses were run on the SAS computer programs (Statistical Analysis Institute 1986).

RESULTS

Colonizing abilities

1980 introductions: unscaled data Table 2 shows unscaled colonization success as recorded in the quadrats, with the

different field-relevant densities of seed input for each species. Species differed significantly in their colonization (total leaf area per quadrat) of typical biomass annual patches. Anthoxanthum, with a mean of sixty-three individuals per quadrat, averaging about 1 cm2 leaf area per plant at the end of the first growing season, established better than all other species in the annual vegetation (Table 2a). Holcus, with a mean of thirty- nine survivors per quadrat, averaging about 0 7 cm2 leaf area, established less than Anthoxanthum, but much better than either the resident perennial Rytidosperma, with fewer than one survivor per quadrat, or Deschampsia, with no survivors. Anthoxanthum began to reproduce after two years, with a mean of over 600 seeds per quadrat. In contrast, Holcus produced a mean of only two seeds per quadrat. There was no reproduction by Deschampsia or Rytidosperma (Table 2a).

Establishment was relatively poor in the typical biomass perennial patches. There were significant, but small, differences among species in number of survivors and total leaf area after one year: Anthoxanthum and Holcus generally established better than Rytidosperma and Deschampsia, as in the annual patches. However, there was no seed production by any species in any perennial-dominated patch after two years (Table 2a).

258 Colonization of vegetated sites in grassland

TABLE 2. Colonization success (means + 1 S.E.) of five dominant grass species introduced as seeds in a grassland at Sea Ranch, California.

Number of seeds produced

Resident No N.o. Leaarap Total leaf area 025 Vegetation omass Year of Species see s input survivors iviv u per quadrat After After (patch type) (g 0 25 m

- 2) introduction introduced 0 25 m-2 0 25 m-2 (cm) (cm2 0-25 m 2) one year two years

(a) Typical biomass patches Annual 142 + 6 1980 Anthoxanthum 9718 63 2+5 4? 1 0+0 1? 65 3 + 5-6? 0 442 + 186?

1980 Holcus* 12442 39 2+8 3? 0-7?0 1? 26 5+5 611 0 2+21 1980 Rytidospermat 566+85 0 3+0 211 0 6+0 2? 02+0 1** 0 Oil 1980 Deschampsia 1106 Oil 0** 0 0i]

F3,29=32 63 F2,841=1 8 F3,29=2664 F3,29= 1 13 P<00001 P>02 P<00001 P<00001

1981 Anthoxanthum 9718 16 1 + 1 9? 3 6+0 3? 58 6+ 104? 0 45+25? 1981 Holcus 12442 5 1+201I 23+0211 11 8+3211 0 21?7?

t=38,df=8 t=22,df=103 t=29,df=8 t=04,df=8 P<001 P<005 P<005 P>05

Anthoxanthum 299+8 1980 Anthoxanthumt 16170+583 8 2+5 3?11 09+0 l? 7 7+4 9? 01 0 1980 Holcus 12442 4 4?+ 17?11 0 4+0 1II 1 8 +0 7? ?11 0 1980 Rytidosperma 1871 Oil --- ?? ?11 0 1980 Deschampsia 1106 0 1+0111 07+01 0 1+0 1? Oil 0 1980 Vulpia 5417 7 7+3 1? Not measured Not measured 967+34 7? 0

F4,41 = 4 82 F2,89 = 4-3 F3,32 = 3-89 F4,41 = II 72 P< 0005 P< 005 P< 005 P< 00001

Holcus 290+9 1980 Anthoxanthum 9718 1 1+0 8? 0 5+0 1 0 6+0 5? 0 0 1980 Hokust 20597+878 0? 0? 0 0

t=l l,df =14 t=08,d f =14 P>02 P>0 3

Deschampsia 398+19 1980 Anthoxanthum 9718 5 9+1 6? 0 6+0 2 3 6+1 0? 0 0 1980 Deschampsiat 1323 + 145 Oil Oil 0 0

t=3 6, d f = 14 t=3 4, d f =14 P<O01 P<001

(b) Low biomass patches Annual 71+5 1980 Anthoxanthum 9718 131 + 16? 2 3+0 1? 303+37? 0 3038+1774?

1980 Ryvdosperma+ 566+85 2 0+0 611 0 7?0 211 1 3+0 51 0 ?1 t=865, df =6 t=29 1, df =658 t=91 1, df =6 t=5 8,df =6

P< 00001 P<O 0001 P<O 0001 P<001

Anthoxanthum 258 + 16 1980 Anthoxanthumt 16 070 + 583 6 3+5 8?11 1 0+0 3? 6 2 + 5 71 ?1 0 1980 Holcus 12442 182+63? 1 1+0 l? 198+83? 01! 0 1980 Rytidosperma 1871 Oil Oil 0t! 0 1980 Deschampsia 1106 Oil O11 oil 0 1980 Vulpia 5417 5-6+2 7?11 Not measured Not measured 127+66? 0

F4,18=9 38 t= I 0, d.f.= 107 F3,14= 18 41 F4,18=5 05 P<00001 P>02 P<00001 P<001

Holcus 205+8 1980 Anthoxanthum 9718 11 6+2 9? 1 5+0 2? 17 3+9 8? 0 0 1980 Holcust 20597+878 13?+1311 08?03? 1 0+1 0 0

t=53,df =6 t=l 9,df =59 t=28.df =6 P<001 P> 005 P<O 05

Deschampsia 153+23 1980 Anthoxanthum 9718 100+27? 1 1+0l1? 91+24? 0 342+257? 1980 Deschampsial 1323+145 2 3+?13? 1 6+08? 3 6+1 8?1 0 ??

t=47,d.f =6 t=l 1,d.f =498 t=37,df=6 t=l 2,df =6 P<001 P>02 P<005 P>02

For typical biomass patches, colonization success data were pooled from two patches of each type, n 10 quadrats except where noted. For low biomass patches, data are from one patch of each type, n = 5 quadrats except where noted. All quadrats were 0 25 m2. For resident species, no seeds were introduced; seed input was from the natural seed rain, shown as mean+ 1 S.E. Within each patch type, values not significantly different from one another are indicated by the same superscript. 1981 introductions have separate superscripts. Analyses for typical and low biomass patches are separate. Biomass values are aboveground dry weight in 0-25-m2 quadrats; n = 6 for typical biomass patches, n = 3 for low biomass patches. Leaf area and number of survivors were measured after one growing season.

* Data from one patch only, n= 5. t Resident species, n = 6. t Resident species, n = 3.

D. R. PEART 259

In low biomass patches of each vegetation type, colonization success was generally higher, by all three measures, than in the corresponding treatments in typical biomass patches (Table 2b). Two-way ANOVAS, including the effects of species and biomass, were run using all species-vegetation type combinations that were examined at both biomass levels, to examine the effect of biomass level on colonization success (measured as total leaf area per quadrat). The effect of biomass was significant in all patch types except Anthoxanthum, the patch type with the least difference in biomass between typical and low biomass patches (annual, F1,32 = 19 2, P <0001; Holcus-dominated, F1,16 = 59, P < 005; Deschampsia-dominated, F1,16= 18 6, P<0 001). Anthoxanthum had more than double the number of survivors and almost five times the leaf area per quadrat in low biomass than in typical biomass annual patches, although seed production after two years was similar. An even more pronounced biomass effect was observed for Anthoxanthum introductions into Deschampsia-dominated patches, where survivorship and leaf area per quadrat were more than an order of magnitude greater in low than in typical biomass patches. The introduction of Anthoxanthum seeds into low biomass Deschampsia vegetation was also the only instance where an introduced perennial species reproduced in perennial-dominated vegetation. This Deschampsia patch had the lowest biomass of all perennial-dominated patches studied. In spite of the quantitative differences between results in the low v. the typical biomass patches, the rank order of colonization success among species was not altered. Anthoxanthum established better than Holcus, as in typical biomass patches. Establishment of Rytidosperma and Deschampsia was consistently poor in all vegetation types, including the patches they dominated, irrespective of vegetation biomass. Rytidosperma and Deschampsia did not reproduce in any treatment.

The annual, Vulpia, was the only species to produce seed in the first year after introduction into a perennial-dominated patch, with a mean of 97 seeds 0-25 m-2 in typical biomass Anthoxanthum vegetation (Table 2a). However, this represented a severe decline in the seed population introduced; seed production was equivalent to a mean of only 0 018 seeds per seed introduced, and there were no Vulpia recruits in the experimental quadrats in the second growing season after introduction. Seed production was slightly higher in low biomass Anthoxanthum vegetation (Table 2b), but again there were no Vulpia recruits in the second year.

1981 introductions. unscaled data Anthoxanthum had significantly more survivors, with significantly greater total leaf

area, than did Holcus in the 1981 introductions (Table 2a). This parallels the greater success of Anthoxanthum in the corresponding 1980 treatments. Although the results were qualitatively similar, corresponding treatments differed quantitatively between years. There were significantly fewer survivors of Anthoxanthum (t = 6 2, d.f. = 8, P < 0-001) and Holcus (t=4 9, d.f.=8, P<0 01) in the 1981 experiments, but the survivors of both Anthoxanthum (t=8-9, d.f.= 395, P<0 001) and Holcus (t=4-8, d.f.=220, P<0 001) were significantly larger than in the 1980 experiments. Only in the 1981 experiments did Anthoxanthum and Holcus produce seed in the first year of growth. As in the 1980 introductions, Anthoxanthum produced more seeds than did Holcus, but the difference was not significant in the 1981 experiments.

Per-seed colonizing abilities: scaled data As described in the Methods, the densities of seeds introduced were in the order

260 Colonization of vegetated sites in grassland

Holcus > Anthoxanthum > Rytidosperma > Deschampsia. If there were density-dependent interactions among seedlings, the colonization success of a species would not be proportional to the number of its seeds introduced. Then a linear scaling of colonization success, to correspond to a common density of 10000 seeds introduced, would give a biased estimate of colonization success per seed. However, it is highly unlikely that density-dependent seedling interactions were important in this experiment. This assertion is based on three observations. First, even in the treatment with greatest colonization success in typical biomass vegetation (Anthoxanthum introduced into annual vegetation), the percentage cover of the species introduced was less than 5 % at the end of the first year of growth. Second, seedlings of the introduced species were rarely in physical contact above ground. Third, the biomass of the one-year-old introduced plants was negligible compared to that of the established vegetation. The proportion of total biomass contributed by the introduced plants was estimated using the total aboveground biomass values for these patch types (Peart 1 989a). Leaf area was related to biomass by weighing samples of leaves for which leaf area had been measured. On this basis, the contribution of recruits to total biomass was estimated to be less than 1 % in all treatments. Therefore, competitive effects among colonists, if any, were probably negligible in the first year, relative to the effects of the dominant vegetation on the colonists.

The main results of the experiments were unchanged when expressed on a per-seed basis. The greater colonization success of Anthoxanthum relative to Holcus in annual patches was still significant, and the magnitude of the difference was increased. The results of all multiple-comparison tests, for all introductions, in both typical biomass and low biomass patches were unaltered. Therefore, it is very unlikely that seed input densities alone could account for the observed rankings of colonization success.

Inhibition (1980 introductions only)

Differences in inhibition effects among vegetation types were tested by comparing the scaled invasion success (total leaf area per quadrat) of the test species, Anthoxanthum and Holcus, among vegetation types. Scaling was appropriate here, because the seed-rain density of both Anthoxanthum and Holcus was higher, in the patches they dominated, than the densities at which they were introduced in the other patch types. However, the results of the analyses were identical, whether run on the scaled or unscaled data. In the following results, it should be noted that rankings of the inhibition exerted by vegetation types are the reverse of the rankings of colonization success of the test species.

For Anthoxanthum as a test species, there was a significant effect of vegetation type in typical biomass patches (F3,14=198, P<0001). Total leaf area per quadrat of Anthoxanthum in different vegetation types followed the ranking: annual > Deschampsia- dominated - Anthoxanthum-dominated - Holcus-dominated. (The symbol - indicates no significant difference in the multiple-comparison test.) In low biomass patches, the effects of vegetation type were again significant (F3,14=23 16, P<0 001), but the rankings of colonization success among vegetation types were: annual Deschampsia-dominated >Anthoxanthum-dominated Holcus-dominated. The number of treatments involving Holcus introductions was less extensive than for Anthoxanthum (Table 1). Vegetation type had a significant effect on Holcus colonization success in typical biomass patches (F2,10= 50 3, P<0 001), with colonization success ranked by vegetation type as annual >Anthoxanthum-dominated Holcus-dominated. In low biomass patches, the vege- tation was again significant (F, 6= 23 7, P < 0 005), and the ranking was Anthoxanthum-

D. R. PEART 261

dominated > Holcus-dominated. Thus, the test species responded very similarly to the different vegetation types. However, in the low biomass patches they differed slightly, in that Holcus colonization success was higher in Anthoxanthum-dominated than in Holcus- dominated vegetation, while this was not the case for Anthoxanthum colonization.

Overall, inhibition by perennial-dominated vegetation was greater than inhibition by annual-dominated vegetation. The lowest inhibition by perennial vegetation (i.e. the greatest colonization success of the test species) was in the low biomass Deschampsia patch, which also had the lowest biomass of all perennial-dominated patches in the experiments.

Interactions between colonization and inhibition

To test for an interaction between the species introduced and vegetation (i.e. patch type), all possible analyses by two-way ANOVA of the treatments in Table 1 were done. Only typical biomass patches were included, and total leaf area per quadrat was the measure of colonization success used in the analysis. Considering the results for all four perennial species introduced into the Anthoxanthum-dominated and the annual vege- tation types (Table 2a), there was a significant interaction between vegetation type and species introduced (F3,36=3 5, P<0 05). The significant interaction occurred because species differed in their quantitative response to vegetation type, although they responded in the same direction. For example, Anthoxanthum and Holcus had much greater colonization success in annual vegetation than in Anthoxanthum-dominated vegetation, but Rytidosperma and Deschampsia colonized poorly in both vegetation types (Table 2a). The other possible two-way analyses were for Anthoxanthum and Holcus introduced into Anthoxanthum and Holcus patches, and for Anthoxanthum and Deschampsia introduced into Anthoxanthum and Deschampsia patches. These combinations involved perennial- dominated vegetation only; colonization success was poor in all cases, and there were no significant interactions. When the two-way ANOVAS were run on scaled data, the same interactions were significant.

Inhibition as a function of biomass

The relationship between biomass of resident vegetation and colonization success was analysed for Anthoxanthum, using number of survivors, individual size, total leaf area per quadrat and total seed production as measures of colonization success (Fig. 1). The scaled data were used, because of the higher density of seed input in Anthoxanthum patches. Least-squares fits to an exponential model were obtained by nonlinear regression. The exponential model explained 79% of the variance in number of survivors (Fig. I a), 53 % of the variance in individual leaf area (Fig. 1 b) and 91 % of the variation in total leaf area per quadrat (Fig. I c). The model explained only 36% of the variation in seed production after two years (Fig. Id). In all cases, the exponential model provided a better fit than a standard linear model, or simple hyperbolic or polynomial models. It is clear from Fig. Id that there was no reproduction in typical biomass perennial vegetation, but variable reproduction in low biomass vegetation. The best single measure of colonization success for a perennial is probably total leaf area per quadrat (Fig. Ic), incorporating both number of survivors and sizes of recruits. Seed production did not adequately represent the colonization success of perennials; many quadrats with a substantial leaf area of Anthoxanthum produced few or no seeds, probably because the plants were still below the threshold size for reproduction.

Scaling had a minimal effect on these analyses: when unscaled data were used, there was

262 Colonization of vegetated sites in grassland

200- 4-- (a) (b)

110 OD I o 150 _ t 3-

0

00

E 100 E E 2

c\J ~~~~~~~~~~0 (n ~ ~ ~ ?0

> 50 - 1

0 0 8 *

400 - 10 0000 (c) iT(d)

) U) @ 8000 - o 300 0 0 o 0 o 0 o 200 t 6000 - E E ~200 -

o 0 4000 -

E 0

100 a

o~~~~~~~~~~~~~ J ~ ~ ~ ~ ~~ ~2000-

0 100 200 300 400 500 0 100 200 300 400 500

Biomoss of vegetation (g dry wt 0-25 m-2)

FIG. 1. Relationship between biomass of vegetation and colonization success of Anthoxanthum odoratum L. in a grassland at Sea Ranch, California. Quadrats were located in four different vegetation types, and each point represents colonization success data from one quadrat. Lines are least-squares, best-fit curves to exponential model y = a exp(bx). (a) Number of individuals 0 25- m-2 quadrat; (b) mean leaf area per individual; (c) leaf area 025-m-2 quadrat; (d) seed production 0O25-m-2 quadrat; (a), (b) and (c) are for the end of the first growing season after seed

introductions; (d) is for the end of the second growing season.

a difference of no more than 1.5% in the variance explained by the exponential model. This is not surprising, as all Anthoxanthum patches had relatively high biomass, and colonization success in the Anthoxanthum patches was low (Table 2).

DISCUSSION Interpretation of the experiments

Although the emphasis on field relevance introduced some complexity into the experimental design, methods and analyses, the patterns of colonization success were

D. R. PEART 263

quite simple and consistent. There was wide variation in the colonizing abilities of the dominant perennials in this grassland community. Anthoxanthum and Holcus were more successful colonizers than Deschampsia and Rytidosperma. The difference between these two groups of species was most apparent in the annual patch type, the vegetation that all species colonized best. In the annual patches, the experiments showed clearly that Anthoxanthum established better than Holcus, despite the greater number of Holcus seeds introduced. Thus, the denser seed rain of Holcus did not compensate for the greater per- seed colonizing ability of Anthoxanthum in annual vegetation. The poor colonization success of Rytidosperma and Deschampsia was due in part to the low densities of seeds introduced, reflecting their low densities of seed rain, but their per-seed colonizing abilities were also much less than for Anthoxanthum and Holcus. The rankings of species' colonizing ability were consistent in both low and typical biomass vegetation, and in both years that the introduction experiments were done.

Inhibition effects were also consistent. Many combinations of vegetation type and species introduced were tested, and when stands differed in their inhibition of colonization by one test species, they differed similarly in their inhibition of the other test species. Species differences in inhibition were most evident in comparisons between annual and perennial-dominated vegetation. Several species of annual grasses, together with forbs and Rytidosperma make up the annual patch vegetation (Peart 1989a). The inhibition abilities of the annual patch species cannot be separated using the data from this study. However, because the species in annual patches are intermixed on a small spatial scale, and individual plants (except for Rytidosperma bunchgrasses) are small, recruits are likely to experience mixed species assemblages in the neighbourhoods of ecological influence that surround them in the annual patches. Therefore, for analyses of inhibition effects, it is convenient to treat the species mixture in annual patches as a single vegetation unit. In the perennial-dominated patches, which were almost monospecific stands, the inhibition levels that were measured approximated species-specific inhibition effects.

In the typical biomass patches, the mixture of species in the annual patch vegetation inhibited colonization, by all test species, less than any species of perennial. This is not surprising, considering that colonization of perennial-dominated sites requires establish- ment under a well-formed canopy, while in annual-dominated sites, colonists may experience as neighbours other seedlings with roughly similar growth phenologies. However, in low biomass patches, with Anthoxanthum as test species, Deschampsia- dominated vegetation inhibited colonization no more than annual-dominated vegetation. Thus, the ranking of species' inhibition abilities differed slightly between typical and high biomass patches. This apparent complexity is simply explained, because the biomass of the low biomass Deschampsia patch was unusually low for perennial-dominated vegetation. Vegetation biomass, irrespective of the species comprising that vegetation, was the best predictor of inhibition by a stand. There was no evidence for strong intrinsic (i.e. species-specific) differences in inhibition ability.

What species characteristics are associated with high colonizing ability? High seed production obviously produces a denser seed rain and higher seed inputs to colonizable sites. Seed size is often assumed to be an advantage in colonization (e.g. Harper, Lovell & Moore 1970). The effect of seed size has been well demonstrated for some old field species (Gross 1984), but Fenner (1978) found no relationship between seed size and establish- ment success in artificial swards. The results for this community also indicate that other factors may sometimes override seed size. Per-seed colonizing ability was not clearly

264 Colonization of vegetated sites in grassland

related to seed size for the dominant perennials. Rytidosperma had the largest seeds and Deschampsia the smallest, but both species had similarly poor, per-seed colonizing abilities. Holcus seeds were only 10% larger than those of Deschampsia, but Deschampsia colonization success was zero in many treatments, and always much less than that of Holcus. The aboveground growth forms of all the perennial grasses are very similar, but differences in root morphology and physiological responses were not examined.

The results for Vulpia require a different interpretation from the experiments with perennials. While introduced perennials may continue to grow and achieve a position in the canopy after the first year (as occurred with Anthoxanthum plants that reproduced in the second year in annual and low biomass Deschampsia vegetation), introduced populations of annuals can persist only by renewed establishment from the seeds produced by the initial cohort. Vulpia was able to establish and reproduce in perennial (Anthoxanthum-dominated) vegetation, but the seeds produced after one year were only 2% of the numbers introduced, and those few seeds produced no recruits in the second year. Thus, although Vulpia could germinate and grow to maturity in a perennial stand, the attrition rate was too high for local populations to persist. This suggests that a continual, dense seed rain from a nearby source would be needed to maintain a population of Vulpia in a perennial-dominated patch. Vulpia and other annuals were rare in perennial-dominated patches, and mostly restricted to occasional openings near annual patches (personal observation). Foin & Hektner (1986) documented an increase in the cover of Vulpia over the period 1974-78 when annual grasses as a group declined in cover, while perennial grasses as a group increased in cover. This increase in cover occurred in sites previously occupied by other annuals (T. C. Foin and M. M. Hektner, unpublished data). However, the dynamics of annual populations within annual patches were not addressed in the present study.

Because of localized dispersal, seeds from outside larger patches do not reach the interiors of those patches (Peart 1989a), and any colonization of larger patches must occur near the edges. Therefore, many of the important interactions between established vegetation and potential colonizers must, under natural conditions of seed dispersal, occur in the mixed boundary zones between patches. It would be experimentally intractable to measure colonization and inhibition for all species of colonists, in vegetation consisting of all combinations and relative abundances of species. This is not necessary, however, because colonization success was predictable from the species of colonist and the biomass of vegetation experienced by the colonist. The relative colonizing abilities of species did not change depending on the species composition of the vegetation.

Current dynamics of the grassland What then do the results imply for the present dynamics of the grassland and for

successional change in the near future? Inhibition effects were so strong in typical biomass perennial-dominated sites that colonization was relatively ineffective by all species. It is unlikely that changes in species dominance will occur in such sites without mortality of the dominants or some other disturbance that releases resources for growth. However, Anthoxanthum was able to colonize in low biomass, perennial vegetation, reaching reproductive maturity after two years in the lowest biomass, perennial-dominated (Deschampsia) patch. Anthoxanthum was also the most effective colonizer of annual patches, although Holcus had substantial colonization success there as well. These two perennials can reach reproductive maturity after two years in annual-dominated sites.

D. R. PEART 265

The advantage of perennial colonists over annual competitors is likely to increase as their size advantage becomes more pronounced each year.

In contrast, Vulpia, the most abundant annual and the only one that has recently increased, should decline in abundance, along with other annuals, as annual patches are colonized by Anthoxanthum and Holcus. Anthoxanthum, and to a lesser extent Holcus, should increase correspondingly in abundance. The experimental results suggest that both Deschampsia and Rytidosperma should decline. They colonize very poorly or not at all in sites dominated by other species, even in annual patches, and do not recruit well in their own stands. However, all of these conclusions about dynamics from the experiments must be qualified because disturbances can alter some of the infer-red trends (Peart 1 989b).

Implications for the study of plant community dynamics How are colonizing and inhibition abilities related? Are good colonizers good

inhibitors and vice versa? The analyses indicate little or no relationship between the two. Intrinsic (per-unit biomass) inhibition abilities were similar among species, but colonizing ability differed among species, both overall and on a per-seed basis. As a result, inhibition and colonization must be considered as two very distinct components of competitive ability. The dependence of colonization success on biomass, irrespective of species of vegetation (Fig. 1), raises the interesting possibility that colonizing ability might be defined independently of species of vegetation. Thus, colonizing ability could be measured per-unit biomass of vegetation, at least in continuous canopies consisting of species with similar growth form. The presence of forbs in the annual patches of this grassland did not seem to upset the relationship between vegetation biomass and colonization success apparent in Figs la-d. However, in general, plants of different growth forms may have different inhibition effects even at equal vegetation biomass. Goldberg & Werner (1983) suggested that established individuals of any species, providing they are of similar size and growth form, may inhibit colonists equally. If biomass and growth form alone determine the inhibition effects of vegetation, the experimental and theoretical analysis of plant community dynamics would become more tractable. Two more recent studies on plant competition have also explicitly recognized and evaluated two distinct components of competitive ability. Both Goldberg & Fleetwood (1987) and Miller & Werner (1987), examining competitive relations affecting growth, presented measures of 'competitive response' and 'competitive effect'. These are analogous to the measures of colonization success and inhibition which were used in this study to evaluate the competitive relations that influence the colonization of sites.

It is likely that a focus on colonization processes will be very useful in understanding the dynamics of populations within communities, as well as the spread of populations into new habitats. Measurements of colonizing and inhibition abilities lead directly to inferences about mechanisms of interactions and predictions of change. Furthermore, an emphasis on spatial aspects, i.e. cover, space-holding ability (inhibition) and colonization of sites, makes it much easier to deal with patchiness, whether that patchiness is in species' distributions, substrate, or both. Clearly, for some species, colonizing ability would be most appropriately evaluated for rhizomes or stolons rather than seeds. Measures of colonization and inhibition abilities incorporate the size-asymmetric interactions that are characteristic of most plant communities. As a result, they may be more useful for some purposes than measures of competition that are based on interactions among individuals of similar age and size.

266 Colonization of vegetated sites in grassland

ACKNOWLEDGMENTS

Thanks go to T. C. Foin for stimulating discussion and encouragement. T. W. Peart, P. T. Peart, C. L. Folt, K. R. Hopper and J. Galland assisted in the field. C. L. Folt and J. White critically reviewed the manuscript. J. H. Connell, M. Detwieler, K. Dwire, D. Ferguson, T. C. Foin, J. Galland, S. K. Gleeson, D. E. Goldberg, K. L. Gross, K. R. Hopper, F. Kosnitsky, R. N. Mack, J. Major and R. W. Snaydon constructively criticized an earlier draft. I thank A. Mendel and R. Neff for statistical advice, J. Nichols for assistance with data analysis, and D. S. Wilson and the Kellogg Biological Station for the use of facilities. Financial support was provided by Regents' Fellowships and Jastro-Shields Research Fellowships from the University of California, Davis, and from a Sigma Xi Grant in Aid of Research. The Sea Ranch Association generously provided field accommodation and access to field sites.

REFERENCES Antonovics, J. & Levin, D. A. (1980). The ecological and genetic consequences of density-dependent regulation

in plants. Annual Review of Ecology and Systematics, 11, 411-452. Connell, J. H. & Slatyer, R. 0. (1977). Mechanisms of succession in natural communities and their role in

community stability and organization. American Naturalist, 111, 1119-1144. Fenner, M. (1978). A comparison of the abilities of colonizers and closed-turf species to establish from seed in

artificial swards. Journal of Ecology, 66, 953-963. Foin, T. C. & Hektner, M. M. (1986). Secondary succession and the fate of native species in a California coastal

prairie community. Madrono, 33, 189-206. Goldberg, D. E. & Fleetwood, L. (1987). Competitive effect and response in four annual plants. Journal of

Ecology, 75, 1131-1143. Goldberg, D. E. & Werner, P. A. (1983). Equivalence of competitors in plant communities: a null hypothesis and

a field experimental approach. American Journal of Botany, 70, 1098-1104. Grubb, P. J. (1977). The maintenance of species richness in plant communities: the importance of the

regeneration niche. Biological Review, 52, 107-145. Gross, K. L. (1984). Effects of seed size and growth form on seedling establishment of six monocarpic perennial

plants. Journal of Ecology, 72, 369-387. Gross, K. L. & Werner, P. A. (1982). Colonizing abilities of four biennial plant species in various vegetation

patch types: implications for distributions in a successional sere. Ecology, 63, 921 -931. Harper, J. L. (1977). Population Biology of Plants. Academic Press, London. Harper, J. L., Lovell, P. H. & Moore, K. G. (1970). The shapes and sizes of seeds. Annual Review of Ecology and

Systematics, 1, 327-356. Mack, R. N. & Harper, J. L. (1977). Interference in dune annuals: spatial pattern and neighbourhood effects.

Journal of Ecology, 65, 345-363. Miller, T. E. & Werner, P. A. (1987). Competitive effects and responses between plant species in a first-year old-

field community. Ecology, 68, 1201-1210. Peart, D. R. (1982). Experimental analysis of succession in a grassland at Sea Ranch, California. Ph.D. Thesis,

University of California, Davis. Peart, D. R. (1989a). Species interactions in a successional grassland. I. Seed rain and seedling recruitment.

Journal of Ecology, 77, 236-25 1. Peart, D. R. (1989b). Species interactions in a successional grassland. III. The effects of canopy gaps, gopher

mounds and grazing on colonization. Journal of Ecology, 77, 267-289. Platt, W. J. (1975). The colonization and formation of equilibrium plant species association on badger

disturbances in a tallgrass prairie. Ecological Monographs, 45, 285-305. Silander, J. A. & Pacala, S. W. (1985). Neighborhood predictors of plant performance. Oecologia, 66, 256-263. Statistical Analysis Institute (1986). S.A.S. User's Guide. Statistics. S.A.S. Institute, Cary, North Carolina. Tripathi, R. S. & Harper, J. L. (1973). The comparative biology of Agropyron repens (L.) Beauv. I. The growth of

mixed populations established from tillers and from seed. Journal of Ecology, 61, 353--368. Weiner, J. (1982). A neighborhood model of annual plant interference. Ecology, 63, 1247-1251. Weiner, J. (1984). Neighbourhood interference among Pinus rigida individuals. Journal of Ecology, 72,183-196.

(Received 2 September 1987; revision received 2 May 1988)