Factors driving distribution limits in an annual plant community

Authors

  • Nancy C. Emery,

    1. Center for Population Biology and Department of Evolution and Ecology;
    2. Present address: Department of Biological Sciences and Department of Botany and Plant Pathology, Purdue University, 915 West State Street, West Lafayette, IN 47907-2054, USA
    Search for more papers by this author
  • Maureen L. Stanton,

    1. Center for Population Biology and Department of Evolution and Ecology;
    Search for more papers by this author
  • Kevin J. Rice

    1. Department of Plant Sciences and Center for Population Biology, University of California, Davis, CA 95616 USA;
    Search for more papers by this author

Author for correspondence:
Nancy C. Emery
Tel: +1 765 496 6931
Fax:+1 765 494 0876
Email: nemery@purdue.edu

Summary

  • • Studies examining plant distribution patterns across environmental gradients have generally focused on perennial-dominated systems, and we know relatively little about the processes structuring annual communities. Here, the ecological factors determining local distribution patterns of five dominant annual species distributed across micro-topographic gradients in ephemeral California wetlands are examined.
  • • Over two growing seasons in three vernal pools, patterns of inundation and above-ground biomass were characterized across the microtopographic gradient, population boundaries for five dominant species were documented and a reciprocal transplant experiment and neighbor removal treatment were conducted to test the relative effects of within-pool elevation, competition and seed dispersal on plant performance.
  • • Despite large differences in inundation time between growing seasons, above-ground biomass and the elevation of population boundaries remained consistent. The predicted ‘optimal’ depth for each species shifted between years, but competition and recruitment limitation restricted species’ abilities to track these conditions.
  • • The distributions of the focal taxa are primarily driven by differential responses to environmental conditions associated with different microtopographic positions along pool inundation gradients, and are reinforced by competition and dispersal constraints. The relative importance of competition, other environmental factors and dispersal patterns appear to contrast with results obtained in systems dominated by perennial plants.

Introduction

Understanding the processes that determine the distribution and abundance of organisms has been a core focus of ecology even before its origination as a scientific discipline (Parmesan et al., 2005). Predicting the effects of biotic and abiotic conditions on species distributions has become increasingly urgent as ecologists are being challenged to anticipate the consequences of global change associated with biological invasions (Mack, 2000), species decline (Margules & Austin, 1994), habitat loss and fragmentation (Didham et al., 1998; Cushman, 2006), and climate (Theurillat & Guisan, 2001; McKenney et al., 2007).

Hypotheses that provide a general explanation for species distributions can generally be grouped into models that invoke neutral processes and those that emphasize an important role of the ecological niche. Neutral models predict that species distributions reflect the collective effects of dispersal probabilities and the abundance structure of the local and regional communities (Volkov, 2003). The simplest hypothesis incorporating ecological niche theory expects species distributions to reflect adaptations to specific abiotic conditions within physically heterogeneous environments (Lenssen & De Kroon, 2005). Alternative models assume that biotic factors interact with the abiotic environment to set species limits. One prominent hypothesis argues that trade-offs among life-history strategies interact with abiotic heterogeneity to restrict stress-tolerant species to unproductive habitats, ruderal species to disturbed habitats, and competitive species to benign habitats (Grime, 1979; Huston, 1979; Keddy, 1990). A contrasting theory predicts that competitive processes operate across all abiotic conditions and that heterogeneity in the relative abundance of limiting resources (and the abilities of species to compete for those resources) determines community structure (Tilman, 1982, 1988; Austin, 1990). Other biotic interactions that can influence species distributions across heterogeneous environments include facilitation (Bertness & Callaway, 1994), predation (Connell, 1961; Lubchenco, 1980) and herbivory (Louda, 1989; Maron & Crone, 2006).

Plant communities that show changes in vegetation across naturally-occurring environmental gradients provide an arena in which to test these various hypotheses. Studies involving the careful manipulation of environmental variables and species interactions have been conducted across a wide variety of environmental gradients, including marsh (Grace & Wetzel, 1981; Bertness, 1991; Pennings & Callaway, 1992), desert (Gurevitch, 1986; Kadmon, 1995), lakeshore (Keddy, 1983; Wilson & Keddy, 1985), sea cliff (Goldsmith, 1973) and restored prairie pothole (Budelsky & Galatowitsch, 2000) plant communities. Collectively, these studies reveal that species’ environmental tolerances often interact with biotic factors to determine their distribution patterns within communities. Transplant experiments at larger spatial scales have also revealed that a variety of processes can operate to drive regional or geographic distributions (Gaston, 2003), for example dispersal limitation (Eriksson, 1998), predation (Louda, 1982), plant–plant interactions (Callaway, 1998) and environmental constraints (Pierson & Mack, 1990). Thus at both local and regional scales, the degree to which specific hypotheses are supported or rejected appears to vary among species, habitats, and the nature of the ‘stress gradient’ under examination.

To date, the large majority of research on the processes driving plant distribution patterns across gradients has been limited to perennial communities, and as a result we know remarkably little about the processes operating in communities dominated by annual plants (Brose & Tielboerger, 2005; Cousens et al., 2006). There are several reasons to expect that the mechanisms driving species’ distributions within annual communities may be different from those observed in communities where longer-lived species dominate. First, processes that influence germination rate and timing have been shown to play an important role in annual seed plant communities (Brewer et al., 1998; Rand, 2000; Houle & Valery, 2003), but may be less important for perennial species that recruit primarily by vegetative spread. Second, the distribution limits (and their underlying mechanisms) for annual species may be particularly sensitive to short-term environmental variation (Cousens et al., 2006). Third, patterns of seed dispersal and dormancy may have relatively larger effects on population borders, and particularly the stability of those borders over time, in annuals compared with perennials. Finally, competition may be somewhat less important in annual than in perennial plant communities, since competitive effects in perennial communities often result from the accumulation of above-ground biomass owing to vegetative growth and litter accumulation across generations (Keddy, 1990). As a result of the empirical focus on perennial communities, we do not know the degree to which the large amount of current ecological theory applies to annual communities.

California vernal pools are ephemeral wetlands dominated by annual plant species, and so they provide an opportunity to explore the processes driving plant distribution patterns in a novel community type using a classic experimental approach. Different species are restricted to different microelevation ranges along pool side-slopes (Bauder, 1987; Barbour et al., 2003), and are often visible as concentric rings of contrasting flowers in the spring (Kopecko & Lathrop, 1975; Kruckeberg, 2006). The clearly defined distributions of species over small-scale environmental gradients make it relatively easy to conduct reciprocal transplant experiments across population boundaries. High seed germination rates enabled us to perform transplants using seed additions rather than transplanting juvenile or adult plants; as a result, we can observe the effects of treatments on the entire life cycle of individual plants, and make inferences about the effects of dispersal constraints on plant distribution patterns. Finally, many vernal pool plant species (and all of the species examined in this study) germinate in the autumn, so neighbor-removal treatments are easily implemented by removing neighbors at the beginning of the growing season and only periodically maintaining the treatment throughout the lifecycle of the plants.

Here, we use observations and a reciprocal transplant experiment to investigate the processes driving the distribution patterns of the dominant plant species in three vernal pool plant communities. To our knowledge, this is the first such study in a system dominated by annual native plant species. Furthermore, the results presented here contribute much needed experimental data on ecological processes operating in vernal pool plant communities, which have historically been studied almost exclusively with observational and correlational approaches. Specifically, we addressed the following questions in two consecutive years:

  • • What are the distributions of the dominant plant species across the vernal pool microtopographic gradient?
  • • How do inundation time and above-ground neighbor biomass change with topographic position in the three focal pools?
  • • What is the relative importance of within-pool elevation, competition and limited seed dispersal in determining the distribution patterns of the dominant species in these pools?

Materials and Methods

Study system

Vernal pool wetlands occur throughout the California floristic province, and redevelop each year when winter rain accumulates in natural depressions in a landscape underlain by an impervious layer of soil (Keeler-Wolf et al., 1998). Standing water recedes when temperatures rise and precipitation declines in the spring, at which point the waterlogged habitat rapidly transitions to drought-like conditions. The annual cycling of the accumulation and recession of standing water establishes a sharp gradient in the timing and duration of inundation, moisture predictability, soil temperature, soil depth, soil texture and pH with subtle changes in elevation along the pool side-slopes (Lathrop, 1976; Holland & Dains, 1990). Above-ground vegetative biomass also increases with increasing elevation within pools (Holland & Jain, 1981; Gerhardt & Collinge, 2003; Bauder, 2005).

Study location and conditions

This study was conducted at Mather Regional Park, Sacramento County, CA, USA. Mather Field is located in the Southeastern Sacramento Valley vernal pool region (Keeler-Wolf et al., 1998). The pools at this site are characterized as Northern Hardpan vernal pools, which occur on terrace landforms and are underlain by an acidic iron–silica cemented hardpan (Keeler-Wolf et al., 1998). Red Bluff–Redding complex soils occur at the site, which consist of loam, clay loam, gravelly loam and gravelly clay loam soil profiles (http://websoilsurvey.nrcs.usda.gov). Experiments and observations were conducted between November and June of each water-year (hereafter ‘year’) in 2000–01 and 2001–02. These years exhibited dramatic differences in rainfall patterns. In 2000–01, heavy winter rains did not begin until early February, and total precipitation was approx. 28 cm (as measured at Sacramento International Airport, http://www.wunderground.com/), roughly 12 cm below average. In 2001–02, heavy rains arrived in late December, and a total of approx. 41 cm of precipitation fell through June of 2002. The length of time that pools retained water was below average in 2000–01 and above average in 2001–02.

Selection of pools and focal taxa

All aspects of this study were repeated in three different vernal pools at the study site, hereafter referred to as 397, 385 and 246, in accordance with the numbering system implemented by a wetland delineation prepared by Jones & Stokes Associates, Inc. environmental consulting firm, in 1997. Mather Field contains over 200 vernal pools, which vary in size, depth, floristic composition and steepness of pool side-slopes. This study focused on medium-sized vernal pools because small pools lacked enough area for replicated transplant experiments, while larger pools are partially dominated by perennial wetland species to the point of being classified as freshwater marshes. Medium-sized pools at Mather Field are dominated by annual plants, yet are large enough to conduct fully replicated transplant experiments. Our three experimental pools were deliberately chosen to sample across observed variation among medium-sized pools in (a) the pool bottom dominants and (b) the steepness of pool side-slopes. The nonrandom method of pool selection was accounted for by treating pool as a fixed effect in all statistical analyses (see below). Although not randomly selected, we made an effort to select these pools as a representative sample of medium-sized pools at the site with similar species composition.

Within each focal vernal pool, we divided the continuous gradient into two to three segments (hereafter ‘habitats’) based on floristic, topographic and inundation characteristics. Similar designations have been implemented in other vernal pool studies (Lin, 1970; Schlising & Sanders, 1982; Zedler, 1987; Bauder, 2005). In this study, the ‘bottom’ habitat was classified as the relatively deep, flat portion of the focal pools that is more or less consistently inundated following initial submersion each winter, and that is dominated by vernal pool specialists. The ‘edge’ habitat, topographically located at the shoulder of these pools, is only briefly inundated after heavy rainstorms in particularly wet years and tends to be dominated by exotic grasses and forbs (although some vernal pool endemics still occur). The ‘transition’ habitat spans the side-slopes between the pool bottom and pool edge. Transitional habitat can be exposed and resubmerged multiple times throughout the winter and is partially occupied by plant species restricted to intermediate elevations, but also includes low densities of some species found more abundantly in bottom and edge habitats. The proportion of total pool area occupied by each of the three habitats varies with pool depth and geometry.

Focal species were selected based on their relative abundance in the local vernal pool flora and their distribution with respect to the three habitat types. Lasthenia fremontii (A. Gray) and Navarretia leucocephela ssp. leococephela (Benth.) are common dominants of pool bottoms at this site, and were the focal species of the bottom habitat in pools 385 and 397. Pogogyne douglasii (Benth.), a species that is relatively rare at the site but locally abundant within a few pools, was the focal species in the bottom habitat of pool 246. Pools 397 and 246 had a ring of the composite Layia fremontii (Torrey and A. Gray) at intermediate, transitional elevations. Pool 385 lacked Layia and other species often found in the transition community. Finally, Limnanthes alba (Benth.) was the primary vernal pool endemic occurring along the edge of all three pools, although its relative density is often lower than that of invasive annual grasses in the edge habitat. The five species selected are relatively common vernal pool taxa, produce seeds that can be easily collected in large numbers and have high germination rates in the field.

Submergence time and above-ground biomass along the vernal pool gradient

The above-ground vegetation biomass and the duration of inundation during the 2000–01 and 2001–02 growing seasons at the bottom, transition and edge habitats in each pool were characterized. One 15 × 15 cm reference plot was placed in each habitat along each of six randomly-oriented transects spanning the inundation gradient for a total of six reference plots/habitat/pool. Replication was doubled in the transition habitat of pool 397 in anticipation of increased variance due to patchy community composition in that area. The mean ± SE relative elevation of reference plots in each habitat, averaged across years, were as follows:

  • Pool 397: bottom = 9.3 ± 1.0 cm, transition = 21.0 ± 0.6 cm, edge = 32.3 ± 1.5 cm;

  • Pool 385: bottom = 11.4 ± 1.1 cm, edge = 32.8 ± 1.4 cm;

  • Pool 246: bottom = 3.7 ± 0.7 cm, transition = 10.7 ± 1.0 cm, edge = 26.4 ± 2.0 cm.

Inundation status (submerged or exposed) was documented weekly throughout the wet season in each reference plot, as well as in all plots used in the transplant experiment (see below). After the focal species senesced in the late spring of each growing season, the above-ground vegetation in each reference plot was harvested, dried to a constant weight and weighed to the nearest 0.01 g.

Submergence time and above-ground biomass data were analysed using a mixed linear model. Specifically, we implemented a mixed-model anova using PROC MIXED in SAS v. 8.02 (SAS Institute, CARy, NC, USA) using the METHOD = TYPE3 option to prevent terms with zero or negative estimated variance from being dropped from the model. Transect was nested within pool and year. Transect and the habitat (i.e. bottom, transition, or edge) × transect interaction term were treated as random effects. Year, pool, habitat and their interactions were treated as fixed effects. Within each pool, paired t-tests among bottom, transition and edge habitats were conducted to identify differences among these positions in submergence time and above-ground biomass production. The P-values were corrected for multiple comparisons using sequential Bonferroni adjustments (Rice, 1990).

Elevation limits of focal taxa

The elevation of upper and/or lower population boundaries were documented for each focal species, in all three pools, in spring 2001 and 2002. During the peak flowering period, the edges of focal populations were flagged at approx. 3-m intervals around the circumference of each population, and the elevation of these population borders, relative to the deepest point in each pool, were measured to the nearest centimeter with an autolevel.

Distribution and abundance of focal taxa

The distribution and abundance of focal taxa across the inundation gradient in pools 246, 385 and 397 were documented in the spring of 2005. During the peak flowering period, four randomly-oriented transects were placed spanning the inundation gradient in each pool. Along each transect, the percent cover of focal taxa was documented every 0.5 m in a 0.5 × 0.25 m quadrat (long side placed perpendicular to the inundation gradient), and the elevation of each sample point was measured using an autolevel.

Transplant experiment

Reciprocal transplant experiments, coupled with neighbor-removal treatments, were conducted among the five focal species during the 2000–01 and 2001–02 growing seasons in three pools. Seeds of all five species were collected from the pools in the spring before each experimental season. Seeds were collected in bulk by walking through a population, collecting inflorescences with mature seeds from maternal plants, and placing them together in a paper bag. Seeds were never mixed across pools or years. Although a single inflorescence typically contained multiple seeds, the total number of inflorescences collected far exceeded the number of seeds used in the subsequent experiment, so it is very unlikely that multiple seeds from the same maternal plant were used in the experiment. Thus, we considered each seed to be independent from all other seeds within a particular pool and growing season. Seeds were stored under laboratory conditions until the autumn of the year they were collected.

In early December of 2000 and 2001, when vernal pool plants were naturally emerging in these pools, a pair of plots was established for every species in each habitat (bottom, transition, and edge) along each of six reference transects spanning the inundation gradient in each pool (see above). The mean ± SE relative elevation of all plots in each habitat, averaged across species, treatment and years, were as follows:

Pool 397: bottom = 9.7 ± 0.5 cm, transition = 22.0 ± 0.3 cm, edge = 31.5 ± 0.6 cm;

Pool 385: bottom = 11.1 ± 0.6 cm, edge = 31.0 ± 0.7 cm;

Pool 246; bottom = 3.4 ± 0.3 cm, transition = 10.7 ± 0.4 cm, edge = 25.1 ± 0.3 cm.

One plot within each pair was randomly selected to receive a neighbor-removal treatment, which involved clipping all vegetation before planting the focal species and regularly removing any background vegetation that emerged over the course of the experiment. In 2000, 10 seeds were planted in each plot, with the intention of using plot means as the response variable. Fruits (usually consisting of three or four seeds) were planted instead of individual seeds for Navarretia because seeds remain attached to one another in a capsule fruit until submerged, at which point they separate within a thick gelatinous coating. High emergence and survival rates in 2000 prompted us to reduce within-plot replication to four seeds (fruits for Navarretia) per plot in 2001 for all species except Lasthenia, which had experienced high levels of seedling mortality during the previous year's experiment. Individual seeds (or fruits) of all species were glued to toothpicks with water-soluble glue before planting to ensure that target seedlings could be distinguished from background emergent seedlings. Seeds were randomly selected from the bulk field collections, and because our transplants were designed to mimic naturally produced, dispersing seeds, no attempt was made to standardize maternal environmental effects (e.g. seed size). All seeds were planted into the same pools from which they were collected.

Seedling emergence was monitored from the planting date until the first plot was submerged by winter rain, which occurred in early February in the first year and in late December in the second year. For Navarretia transplants, emergence (yes/no) was recorded on a per-fruit basis because of the difficulty in distinguishing between fruits with one vs multiple emergent seedlings.

After the water receded in the spring, surviving plants were monitored weekly and harvested upon senescence. Fecundity was estimated for each species in the laboratory. Lasthenia and Layia are both composites with numerous florets (and seeds) per inflorescence, so total inflorescence weight was used to estimate fecundity. Since individual Navarretia seeds could not be counted, total fruit weight was used as the fecundity estimate. When multiple Navarretia individuals survived per fruit planted, one individual was randomly selected for the fecundity estimate. Pogogyne and Limnanthes seeds are large and easily removed, so seed counts were used to estimate fecundity for these species.

Results for each species were analysed separately with a split-plot mixed model anova using PROC MIXED with the METHOD = TYPE3 option in SAS (SAS Institute). Transect was treated as a random block effect nested within pool and year. Pool, year, and transplant destination (i.e. the bottom, transition or edge habitat) were fixed whole-plot effects, and competition treatment was the subplot factor. The mean fecundity for each plot was the response variable (n = 6 for each species/pool/year/zone/treatment). Individuals that did not emerge, survive or produce any seeds were given a fecundity value of zero. To reduce heteroscedasticity and the nonnormal distribution of errors, all fecundity data were loge(x + y) transformed, where x was the response variable and y was a small nonzero value equal to or less than the smallest nonzero value observed in the data set (whichever best increased the normality of residuals). Overall destination effects for each species were further examined with Tukey's pairwise comparisons among destination habitats and across competition treatment, pool and year. The effects of competition treatment within each habitat type were examined by conducting paired contrasts between control and neighbor-removal plots within each pool, year, and destination habitat. Sequential Bonferroni-adjusted P-values were used to assess significance for competitive effects in these post hoc comparisons (since this adjustment is conservative, all P values that were significant before adjusting for multiple comparisons are also reported below). When significant interactions were found between pool and other independent variables, additional paired t-tests (with sequential Bonferroni-adjusted P-values) were conducted within pools. Least-square means were calculated using the LSMEANS statement in PROC MIXED (Littell et al., 2006). The LS means accounting for year, transplant destination, and competition treatment effects were calculated for all species; when significant interactions with pool were found (as was the case for Layia and Limnanthes), LS means by year, transplant destination, competition and pool were also calculated.

Results

Submergence time and above-ground biomass along the vernal pool gradient

In all pools and in both years, the total length of time reference plots were under water decreased sharply with increasing elevation within the pool (Fig. 1). There was also significant pool-to-pool variation in the length of time that each pool retained standing water. This effect was driven primarily by pool 246, which held water approximately half as long as pools 397 and 385. Absolute differences in submergence time between 2000–01 and 2001–02 were greater at the pool bottom than at higher elevation habitats.

Figure 1.

Inundation and above-ground biomass in three vernal pools during 2000–01 (black bars) and 2001–02 (hatched bars) in three topographic habitats along the vernal pool gradient. Data are mean ± SE (n = 6 for all means except the transition zone of pool 397, where n = 12; see the Materials and Methods section). Letters indicate results from pairwise comparisons among habitats within each pool. Habitats with different letters are significantly different from one another. Asterisks indicate significant differences between years within each habitat. The characteristic ‘transition’ habitat was not present in pool 385 (see the Results section).

Within pools, above-ground vegetation biomass increased with elevation along the topographic gradient (Fig. 1). Despite dramatic year-to-year differences in precipitation and inundation, above-ground biomass was relatively consistent across years in the three pool habitats.

Elevation limits of focal taxa

The population boundaries of each species were very consistent across the 2000–01 and 2001–02 growing seasons (Fig. 2); most shifts observed were only of the order of a few centimeters in elevation change. The upper borders of pool bottom species (Lasthenia, Navarretia and Pogogyne) were usually a few centimeters higher in 2001–02 than 2000–01, with the exception of Navarretia in pool 397, which maintained the same upper elevational limit in both years. Layia shifted its lower edge slightly downward in 2001–02 relative to 2000–01, while its upper edge was similar among years. The lower boundary of Limnanthes populations shifted upwards in 2001–02 in pool 385, but stayed at similar elevations in the other two pools, compared with its location in 2000–01. The elevation limits of each species were consistent around the circumference of each pool, as indicated by relatively small error in population boundary estimates. The largest boundary shift observed was the lower limit of Limnanthes in pool 385, which shifted approx. 7 cm upland in 2001–02.

Figure 2.

Locations of focal plant species boundaries along within-pool elevation gradients in three vernal pools in 2000–01 (closed bars) and 2001–02 (hatched bars). Population boundaries represent the mean elevation of observed population borders ± SE (n = 5–20, depending on the size of the population). Boundaries were documented by marking the upper and lower limits of each species around the circumference of each population, and measuring the elevation relative to the pool bottom using survey equipment. Topographic habitats are indicated below the x-axis of each pool.

Distribution and abundance of focal taxa

The abundance distributions of focal species across the inundation gradient reveal species-specific pool and depth preferences (Fig. 3). Lasthenia, Navarretia and Pogogyne reached highest densities in the bottom zones of the focal pools. However, within the bottom zone, Navarretia had highest densities at slightly lower microelevational positions than Lasthenia, while Pogogyne only occurred in the pool where Navarretia and Lasthenia were absent. Layia reached peak abundance in the transition zone of two pools, but was absent from the pool with the steepest pool side-slopes (see Fig. 3, pool 385, inset). Limnanthes consistently reaches peak abundance in the edge zone of all pools, but its distribution extends at low densities into deeper positions in all pools.

Figure 3.

Percent cover of focal taxa along within-pool elevation gradients in three vernal pools at Mather Regional Park in spring 2005. Data are derived from surveys along four randomly-placed transects in each pool. Points are means of plots within each elevation class. Topographic habitats delineated for the purposes of this study are indicated below the x-axis of each pool. Insets show the change in elevation with distance from pool center along each survey transect.

Transplant experiment

Transplants of Lasthenia, a pool bottom species, expressed significant levels of variation in response to year, transplant destination and neighbor-removal treatment (Table 1, Fig. 4a). Overall fecundity was higher in 2001–02 than in 2000–01. Fecundity was consistently lower in the edge habitat relative to the transitional and bottom habitats. Neighbor removal increased mean fecundity overall, but the magnitude of this effect varied among transplant destinations and years. Specifically, transplants into the transition region exhibited a strong positive response to neighbor-removal treatment in both growing seasons, while transplants to the bottom and edge habitats showed a significant increase in fecundity in response to this treatment in the wetter 2001–02 season only.

Table 1.  Mixed-model anova results for lifetime fecundity of each species in the transplant experiment
SourceLastheniaNavarretiaPogogyne*
FnDF, dDFPFnDF, dDFPFnDF, dDFP
Year13.111,21.70.00150.61,22.10.44842.411,100.1516
Pool0.421,37.40.5190.391,34.110.5368
Year × pool0.591,37.40.44620.371,34.110.5462
Destination27.292,28.5<0.000136.22,29.6<0.00013.532,200.0486
Year × destination1.742,28.50.1947.492,29.60.00230.332,200.7216
Pool × destination0.041,32.80.85220.251,41.30.6201
Year × pool × destination0.111,32.80.73790.461,41.30.5028
Competition46.11,66<0.000113.571,640.00058.571,300.0065
Year × competition2.41,660.12630.631,640.42897.671,300.0096
Pool × competition0.431,660.51320.251,640.6207
Year × pool × competition0.231,660.63410.271,640.6071
Destination × competition2.452,660.09431.052,640.35520.112,300.8958
Year × destination × competition1.032,660.36340.692,640.50430.292,300.7523
Pool × destination × competition 01,660.99961.081,640.3023
Year × pool × destination × competition0.321,660.57430.221,640.6431
Transect (year, pool)0.7820,30.30.71511.320,33.40.24350.8510,200.588
Destination × transect (year, pool)2.1629,660.00511.1430,640.31891.420,300.199
SourceLayiaLimnanthes
Fndf, ddfPFndf, ddfP
  • F-values, degrees of freedom (ndf, numerator; ddf, denominator), and P > F are reported.

  • *

    Source effects in italics were not included in the anova for Pogogyne, which only occurred in one pool.

Year0.11,20.20.758718.91,33.40.0001
Pool17.911,20.20.00042.462,36.80.0997
Year × pool0.241,20.20.63123.962,36.80.0276
Destination2.162,41.30.128331.092,49.4<0.0001
Year × destination3.852,41.30.02949.572,49.40.0003
Pool × destination6.482,41.30.00362.093,50.60.1133
Year × pool × destination1.992,41.30.14986.13,50.60.0013
Competition9.091,760.00358.171,940.0052
Year × competition3.551,760.06339.761,940.0024
Pool × competition0.511,760.47811.182,940.3127
Year × pool × competition1.581,760.21320.752,940.4767
Destination × competition6.892,760.00180.272,940.7651
Year × destination × competition2.422,760.09560.622,940.5394
Pool × destination × competition1.672,760.19561.283,940.2865
Year × pool × destination × competition0.432,760.65170.513,940.6795
Transect (year, pool)1.2820,41.10.24510.8330,49.00.7047
Destination × transect (year, pool)1.3940,760.1081.9448,940.0031
Figure 4.

Results from transplant experiments revealing the effects of transplant destination and competition for (a) Lasthenia fremontii, (b) Navarretia leucocephela, (c) Pogogyne douglasii, (d) Layia fremontii, (e) Limnanthes alba by year. Control, open bars; neighbors removed, tinted bars. Data are back-transformed least-square means ± SE from a split-plot mixed-model anova. Transplant destinations with different letters were significantly different from one another in pair-wise comparisons. Asterisks indicate significant effects of neighbor-removal treatment within each destination. The sample sizes associated with least-square means ranges from n = 6 (for Pogogyne) to n = 18 (for Limnanthes), depending on the number of pools occupied by each species.

The pattern of fecundity variation for experimental transplants of Navarretia, another pool bottom species, differed between years (Table 1, Fig. 4b). In 2000–01, Navarretia transplants had highest fecundity at the pool bottom. In contrast, transplant performance in the pool bottom was similar to that observed in the transition area in 2001–02. In both growing seasons, mean transplant fecundity was lowest in the edge habitat. Removing neighbors had an overall positive effect on transplant performance, but paired t-tests identified significant responses only in the transition habitat in the drier 2000–01 year (Fig. 4b). Before sequential-Bonferroni adjustment, the effect of neighbor removal was also significant in the edge zone in 2000–01 (P = 0.018) and 2001–02 (P = 0.023).

Pogogyne only occurred in the bottom areas of pool 246, which retained water for shorter periods of time than the other two pools (Fig. 2). Like Lasthenia and Navarretia, Pogogyne transplants tended to have lowest fecundity at the pool edge, even in the absence of neighbors (Table 1, Fig. 4c). However, despite an overall significant effect of destination, pairwise comparisons between habitat positions were not significant after correcting P-values for multiple comparisons. Positive responses to the neighbor-removal treatment were observed in 2000–01, particularly in the bottom and transition habitats, where plants in control plots produced an average of < 1 seed per seed planted. Neighbor-removal treatment had little to no detectable effect on Pogogyne transplant fecundity in 2001–02. In 2000–01, neighbor removal had a significant effect on plant fecundity in the transition habitat only. Before sequential Bonferroni correction of P-values, the effect of neighbor-removal treatment was also marginally significant in the bottom (P = 0.060) and edge (P = 0.051) zones.

For Layia, a species found most often at transition elevations, transplants achieved a lower mean fecundity in pool 397 than in pool 246 (Table 1, Fig. 5). Larger differences in mean fecundity were observed across transplant destinations in pool 397 than in pool 246, where transplant performance was more consistent across destinations. Transplants consistently exhibited reduced fecundity in the bottom habitat compared with higher locations in pool 397, and there was evidence of facilitation (i.e. a negative response to neighbor removal) in the bottom of pool 246 in 2001–02 (Fig. 5). In pool 397, transplant fecundity was highest in the transition region in 2000–01 and at the pool edge in 2001–02. Significant increases in mean fecundity in response to the neighbor removal treatment were observed in the transition habitat of pool 397 in 2000–01, and at the edge of pool 246 in 2001–02. Before adjusting significance levels to control for multiple comparisons, the neighbor removal treatment was also significant in the transition zone (P = 0.053) and the edge zone (P = 0.022) of pool 246 in 2000–01. Transplants were able to germinate, survive and reproduce outside the transition zone at the edge of pool 397 in 2001–02 and the bottom of pool 246 in both years (Fig. 5).

Figure 5.

Transplant experiment results illustrating the effects of transplant destination, competition and pool on Layia fremontii performance in each year. Results were analysed at the pool level because significant pool interactions were detected in the split-plot mixed-model anova. Control, open bars; neighbors removed, tinted bars. Data are back-transformed least-square means ± SE from a split-plot mixed-model anova. Transplant destinations with different letters were significantly different from one another in pair-wise comparisons. Asterisks indicate significant effects of neighbor removal treatment within each destination. n = 6 for all estimates, except n = 12 in the transition zone of pool 397.

Transplants of Limnanthes, a species primarily found at pool edges, had higher average fecundity in 2000–01 than in 2001–02, but the magnitude of that effect varied among pools, transplant destinations, and neighbor-removal treatment (Table 1, Fig. 6). Transplants generally had lowest fecundity in bottom habitats, regardless of neighbor removal treatment (Fig. 4e), but in the drier 2000–01 season, this pattern was less pronounced in pools 397 and 385, and disappeared entirely in pool 246 (Fig. 6). The average fecundity of Limnanthes transplants in the transition habitat was most similar to edge transplants in 2000–01 and to bottom transplants in 2001–02. Overall, competitive suppression was stronger in 2000–01 than in 2001–02 (Table 1, Fig. 4e) but paired t-tests only found a statistically significant positive response to neighbor-removal treatment across pools for transplants in the edge habitat in 2000–01 (Figs 4e and 6). Before correcting for multiple comparisons with sequential Bonferroni-adjusted P-values, neighbor removal also had a significant effect on plant fecundity at the edge of pool 246 in 2001–02 (P = 0.022).

Figure 6.

Transplant experiment results illustrating the effects of transplant destination, competition and pool on Limnanthes alba performance by year. Control, open bars; neighbors removed, tinted bars. Data are back-transformed least-square means ± SE from a split-plot mixed-model anova. Transplant destinations with different letters were significantly different from one another in pair-wise comparisons. Asterisks indicate significant effects of neighbor-removal treatment within each destination. There were no transplants placed in the transition zone of pool 385 because this pool lacked the transition habitat. n = 6 for all estimates, except n = 12 in the transition zone of pool 397.

Discussion

Species-specific environmental tolerances largely determine plant performance across the elevation gradient in the focal vernal pools, while competition and dispersal patterns further define population borders and may stabilize the location of boundaries across years. The vernal pool gradient is characterized by sharp differences in flooding duration and productivity over a relatively small-scale environmental gradient (i.e. 25–35 cm in elevation; Fig. 1). Inundation time varies drastically between years with different precipitation regimes, while productivity remains relatively constant (Fig. 1). Year-to-year variation in precipitation patterns caused the inundation conditions along pool topographic gradients to change (Fig. 1), but the populations of the dominant species did not accurately track this variation (Fig. 2). Instead, the elevations of population boundaries were remarkably similar in both growing seasons. The transplant experiment revealed that the consistency of species boundaries across spatially and temporally variable environmental conditions results from a dynamic interaction among environmental factors associated with elevation, plant competition and dispersal limitation (Table 1, Figs 4–6).

Previous studies in vernal pool plant communities have suggested that physiological stressors associated with inundation (e.g. reduced availability of carbon dioxide, soil anoxia and fluctuating pH levels) (Zedler, 1987; Keeley & Sandquist, 1991) or drought associated with the drying phase (Linhart, 1974; Bauder, 1989) primarily determine the lower and upper limits, respectively, of plant species within vernal pools. The results of the transplant experiment are consistent with the hypothesis that environmental tolerances largely prevent bottom-adapted species (i.e. Lasthenia, Navarretia, and Pogogyne) from colonizing the pool edge, and the edge-adapted species (i.e. Limnanthes) from invading the pool bottom. Many environmental factors covary with elevation in vernal pools, including hydrology, soil characteristics and other physical attributes (Lathrop, 1976; Holland & Dains, 1990). Without individually manipulating each abiotic factor, along with other biotic factors including herbivory and interactions with the soil biota, it is not possible to identify the exact mechanism underlying the elevation-specific environmental tolerances of each species. However, from examining plant performance patterns across microtopological and year-to-year variation within pools (Figs 4–6), we suspect that drought tolerance may set the upper boundary, and inundation tolerance may determine the lower limit of the species examined in this study. In the absence of neighbors, pool-bottom species (Lasthenia, Navarretia, and Pogoyne) tolerated the hydrological conditions across their natural distribution (i.e. at the pool bottom and into the transition zone). However, these species could not tolerate the conditions at the extreme upper end of the vernal pool gradient, regardless of neighbor-removal treatment. Layia is naturally centered in the transition habitat; it persisted only at the pool edge in the second, relatively wet year, and consistently had reduced fecundity at the pool bottom where transplants experienced longer periods of inundation (Fig. 5). Consequently, this species may be susceptible to both inundation stress at lower positions and drought stress and higher elevations. Finally, Limnanthes transplants only persisted in the transition or bottom zones under relatively dry conditions (i.e. the 2000–01 growing season and pool 246), strongly suggesting that inundation tolerance is largely responsible for restricting this species to pool edges.

The role of competition in determining vernal pool plant distribution patterns has received little attention, even though vegetation surveys that consistently document a positive correlation between biomass and elevation within pools (Holland & Jain, 1981; Gerhardt & Collinge, 2003; Bauder, 2005) suggest variation in the competitive environment along the vernal pool inundation gradient. In our transplant experiment, plant competition had a significant overall impact on the performance of all species examined, but the effect of competition varied significantly among zones for only one species (Layia, Table 1). Although high within-treatment variation may have restricted our ability to compare effects of competition across zones, competition was not consistently stronger beyond species borders than within them, indicating that competitive interactions alone do not dictate vernal pool plant population boundaries.

Our results suggest that effects of competition on these vernal pool endemics may increase under dry conditions. Competitive effects were relatively rare (although not completely absent) in the bottom habitat in both years (Fig. 4), yet were detected for all species in the transition zone in 2000–01 (this effect was significant in Limnanthes only before adjusting P values for multiple comparisons). The frequency of significant competitive effects in the transition zone indicates that in dry years, competition can play a role in reducing plant fecundity at the upper and lower boundaries of the pool bottom and edge species, respectively (Fig. 4a–e in 2000–01). The tendency for the presence of neighbors to influence plant performance under drier conditions suggests that in dry years, strong competition for water develops as water recedes in the spring, possibly exacerbating the effects of drought stress at the upper limits of species across the inundation gradient. This interpretation is consistent with the work of Bauder (1989), who found that drought tolerance primarily determined the upper limit of a vernal pool endemic restricted to deeper portions of pools (Pogogyne abramsii), while competition had relatively minor effects on plant performance at the edge of its population.

Our results also demonstrate that local dispersal can restrict the ability of these populations to track year-to-year fluctuations in precipitation and inundation conditions. We are able to infer the effects of dispersal constraints because we planted seeds (rather than juveniles or adults), which simulated dispersal to different habitat types and competitive environments. We observed several instances where seed transplants moved beyond current population boundaries, survived and reproduced in control plots (e.g. Limnanthes transplants to the bottom zone in 2000–01; Figs 4 and 6), indicating that species’ borders could shift in response to seasonal conditions if viable seeds were present in those peripheral sites. Since we observed only subtle changes in population borders over the two growing seasons (Fig. 3), we can conclude that a lack of dispersal beyond current boundaries restricts year-to-year shifts in population boundaries. The majority of vernal pool specialists exhibit no apparent mechanism to promote dispersal, and many species exhibit traits that restrict dispersal, such as parental retention of seeds and direct seed placement near the parental plant, presumably as an adaptation to island-like habitat structure or lack of selection for increased dispersal (Zedler, 1990). While dispersal limitation has been shown to limit species’ distributions at larger spatial scales (Eriksson, 1998), its role in structuring communities at smaller scales has not often been observed, and may be particularly important in systems with high year-to-year variation and small seed banks.

This study provides insight into the degree to which models generally tested in perennial communities are relevant to a system dominated by annual plant species. While neutral dispersal processes can constrain the ability of some populations to track environmental variation across years, species-specific adaptations to different ecological niches clearly plays the predominant role in determining species’ distributions in the communities examined here. Consequently, we conclude that neutral models play a limited role in explaining plant community structure in vernal pools at these local scales. Alternative models that incorporate niche differences among species have different expectations for the nature of competitive interactions across environmental heterogeneity, predicting either (1) trade-offs between competitive ability and the ability to tolerate environmental stress (Grime, 1979; Huston, 1979; Keddy, 1990), or (2) consistent levels of competition across varying resource ratios (Tilman, 1982, 1988; Austin, 1990). Our results do not clearly support either hypothesis. First, the direction of the realized stress gradient varies among species. That is, conditions that are stressful for one species are favorable for another, and one end of the gradient is not universally benign for all species. Second, competition occurs primarily within and at the edge of species’ distributions, but plays a minor role beyond population boundaries, where physical factors have the predominant effects on species performances. Trade-offs between tolerance to different stressors (e.g. inundation vs drought), rather than a trade-off between stress tolerance and competitive ability, or the ability to compete for different resources, appear to determine plant performance in these vernal pool species at different elevations.

It is not possible to determine if our results, which differ from those observed in perennial-dominated communities, arise from the annual habit of the plant community in vernal pools, since similar studies in other annual systems are lacking. Experiments have demonstrated that the distributions of resident annual species may be highly sensitive to abiotic conditions in heterogeneous environments (Noe & Zedler, 2000, 2001; Rand, 2000; Houle & Valery, 2003; Reynolds & Houle, 2003). Seed availability has also been identified as a contributing factor to the maintenance of population distributions in an annual salt-marsh species (Rand, 2000). While these studies focus on competitively subordinate annuals that are restricted to marginal habitats in communities of perennial species, the important roles of physical factors and seed availability in determining species distributions are consistent with our results. In a contrasting case, Brose & Tielboerger (2005) performed seed bank transplants and a neighbor-removal treatment in a temporary wetland in an East German farmland. Similar to vernal pools, this habitat is dominated by annual plant species, although the persistence of annuals is facilitated by annual plowing. Their study demonstrated that competitive exclusion prevented wetland annuals from occupying the terrestrial edge of the gradient, while environmental stress restricted weed species from moving into the deeper areas of the wetland (Brose & Tielboerger, 2005). One explanation for the different findings of their study and ours may be that vernal pool wetlands are dominated by endemic species that have adapted to the seasonal wet-and-dry cycles characteristic of vernal pools (Zedler, 1990) and, as shown here, even specific elevations within pools (Barbour et al., 2003). More work is needed on other endemic annual communities distributed across environmental gradients to test whether our results are representative of other annual systems.

The vernal pool plant communities studied here appear to operate by different ‘rules’ from better-studied perennial communities distributed across local environmental gradients: the locations of the focal taxa along the vernal pool inundation gradient are primarily driven by differential responses to site-specific environmental conditions, and their distribution patterns are made more distinct and consistent by competition and dispersal constraints at population edges. Importantly, the roles of competition and dispersal constraints can vary among years as pool hydrology changes with precipitation patterns, underscoring the fundamental importance of physical factors in structuring this community. Owing largely to the annual nature of this system, we were able to observe how the dynamic relationship between dispersal constraints, competitive interactions and species-specific niche differences yields surprisingly consistent distribution patterns across year-to-year variation in the physical environment. The ability to see shifts in ecological processes across short-term temporal environmental variation, and to measure the effects of these shifts on the lifetime fitness of organisms, are major strengths of studying annual communities that can significantly contribute to our understanding of the mechanisms underlying community structure.

Acknowledgements

The authors thank D. Plass, M. Goedde, S. Ratay, T. Schubert, E. Baack, R. Scherson, J. Wright and C. Neu for many hours of laboratory and/or field assistance associated with the transplant experiment, and M. Bertness, S. Pennings and P. Zedler for valuable feedback on earlier drafts of this manuscript. Funding for this study was provided to N.E. by the UCD Botanical Society G. Ledyard Stebbins Research Award, a UC Davis Center for Population Biology Research Award, an NSF Graduate Research Fellowship, an NSF Doctoral Dissertation Improvement Grant #0309006, and NSF DEB-0621337 (to D. D. Ackerly and B. G. Baldwin).

Ancillary