Journal list menu
Niche evolution across spatial scales: climate and habitat specialization in California Lasthenia (Asteraceae)
Corresponding Editor: J. Cavender-Bares. For reprints of this Special Issue, see footnote 1, p. S1.
Abstract
Ecologists and evolutionary biologists perceive the ecological niche as a multidimensional relationship between an organism and its environment. Yet, we know little about the degree to which multiple niche axes evolve in concert across various spatial scales to explain differences in distribution patterns and habitat specialization among lineages. Here we used contemporary phylogenetic approaches to analyze the evolution of species' distributions across multiple spatial scales in Lasthenia, a young and ecologically diverse plant clade largely occurring within the California Floristic Province, USA. Lasthenia species and subspecies range from widely distributed taxa that occupy a diversity of habitat types to locally restricted habitat endemics, including several lineages that are strongly associated with isolated ephemeral wetlands called vernal pools. We quantified the niche of Lasthenia species and subspecies at three different spatial scales: the range-wide climate niche, the habitat niche, and the within-habitat depth niche for those taxa occupying vernal pools. We incorporated phylogenetic uncertainty into our analyses by reanalyzing previously described DNA sequences in a Bayesian context and conducting all subsequent comparative analyses over the resulting posterior distribution of ultrametric phylogenetic trees. Using a biogeographic approach for ancestral habitat reconstruction, we estimated that Lasthenia lineages have undergone up to four independent transitions from strictly terrestrial habitats to a niche that incorporates semiaquatic habitats, and one of these transitions led to the subsequent proliferation of vernal pool species and subspecies. We found that the local niche axis, corresponding to the depth distribution of taxa within pools, was more phylogenetically conserved than the large-scale axes representing climatic associations. Furthermore, we did not find evidence that niche breadth estimates along different axes were consistently correlated, indicating that ecological specialization may be specific to certain niche axes rather than an overall characteristic of a species.
Introduction
The concept of the ecological niche provides a fundamental framework in ecology for understanding where species live and which species can stably coexist. Since its original formulation by Grinnell (1917), countless studies have examined the niche from a variety of biological, spatial, and temporal perspectives to explore the reciprocal relationships between organisms and their environments. Theoretical, empirical, and comparative approaches to studying the niche have been closely intertwined with the study of ecological and evolutionary mechanisms that limit species' distributions (Lenormand 2002, Holt 2003, Bridle and Vines 2007, Kawecki 2008, Sexton et al. 2009) and drive ecological specialization in heterogeneous environments (Futuyma and Moreno 1988, Kassen 2002, Sultan and Spencer 2002). Distribution patterns are the realization of a species' niche on the landscape, filtered by the available environment and refined by dispersal patterns and biotic interactions (Pulliam 2000); in turn, environmental heterogeneity, intrinsic constraints, and patterns of gene flow shape the evolution of the ecological niche (Holt and Gaines 1992, Lenormand 2002, Gomulkiewicz and Houle 2009). The balance between these interacting processes ultimately determines the trajectory of niche conservatism and diversification within and among species to generate patterns of genetic, functional, and taxonomic biodiversity.
A basic, yet empirically elusive property of the niche is that it is multidimensional and associated with environmental variation across spatial and temporal scales (Hutchinson 1957). The “N-dimensional” nature of the niche is universally acknowledged but generally understudied, and is rarely incorporated into studies exploring niche evolution (Harmon et al. 2005, Nosil and Sandoval 2008, Evans et al. 2009). Ecologists and evolutionary biologists have typically focused on understanding niche axes that are considered to be important drivers of distribution and abundance at specific spatial scales, such as climate variables at large spatial scales (e.g., Evans et al. 2009, Smith and Donoghue 2010) and species interactions at smaller scales (e.g., Losos et al. 2003, Nosil and Sandoval 2008), or have studied functional traits as proxies for niche relationships (e.g., Ackerly et al. 2006). At the scale of the geographic range, species distributions are largely thought to be driven by climatic tolerances, and recent work has integrated phylogenetic information and climate data to examine patterns of niche conservatism and divergence in climatic associations across speciation events (Graham 2004, Kozak and Wiens 2006, Evans et al. 2009, Smith and Donoghue 2010). Within the geographic range, species are usually restricted to certain habitat types; in plants, this patchiness is often attributed to species' responses to edaphic factors (Kruckeberg 2002). Finally, at the most local scales, species can be restricted to particular microhabitats within individual communities (i.e., the “alpha niche,” sensu Pickett and Bazzaz 1978, Silvertown et al. 2006). Community ecologists have emphasized a central role of niche differences to explain patterns of community structure and species coexistence at these local scales (e.g., Chesson 1991, 2000, Silvertown 2004). Plant ecologists have identified particularly important roles for hydrological niches in defining plant distributions and facilitating species coexistence within communities (Silvertown et al. 1999, Cavender-Bares and Holbrook 2001, Cavender-Bares et al. 2004, Silvertown et al. 2006, Araya et al. 2011, Savage and Cavendar-Bares 2012). To date, few of these studies have simultaneously examined multiple niche dimensions that determine species' distributions at different spatial scales to explore the multidimensional nature of niche evolution over the evolutionary history of a clade.
A multidimensional-niche perspective is particularly critical for understanding the evolution of ecological specialization and the ecological and evolutionary consequences of limited niche breadth. While some have argued that niche breadth should be similar across all axes and positively related to overall abundance (Brown 1995), others have emphasized that we should not expect a species that is narrow along one niche axis to be narrow along all other axes (Futuyma and Moreno 1988, Devictor et al. 2010), and similarly, that plasticity in one character does not always correlate with plasticity in other traits (Clausen et al. 1948, Bradshaw 1965). Yet, some of our most fundamental explanations of the processes shaping species distribution patterns are based on the dichotomous classification of species as either specialists or generalists (e.g., MacArthur 1972, Brown et al. 1995), and different interpretations have emerged about the relationship between specialization and diversification rates (e.g., Simpson 1944, Armbruster and Muchhala 2009). The relative breadths of different niche axes can also play a key role in determining the potential for adaptive evolution. Ackerly (2003) describes a hypothetical situation where the fundamental niche of a species includes two environmental axes (e.g., temperature and precipitation) that are correlated across the landscape, and the species' fundamental niche is narrower along one axis than the other. When the available environmental space changes along either axis, the niche axis with the broader tolerances is most likely to undergo adaptive evolution, while the more constrained niche axis will drive changes in distribution patterns. Today, obtaining a more general understanding about the relative lability of multiple niche axes, and the correlations in niche breadth along different axes, is particularly urgent in the context of predicting which species are most vulnerable to extinction in the face of rapid environmental change (Colles et al. 2009, Devictor et al. 2010).
Here, we investigated the trajectories of niche evolution among species within a single clade to understand the evolution of habitat affinities and to test if niche specialization coevolves across niche axes that explain plant distribution patterns at multiple spatial scales. We focused on the genus Lasthenia (Asteraceae), a young, rapidly evolving clade of predominantly annual plants that occurs mostly within the California Floristic Province (Ornduff 1966, Chan et al. 2001). The 21 species and subspecies in Lasthenia range from widespread taxa to locally restricted habitat endemics (Rajakaruna 2004; see Plate 1), and the distributions of many Lasthenia lineages naturally fall into a hierarchical spatial structure corresponding to spatially nested niche dimensions. At the largest spatial scales, species and subspecies are distributed across climatic gradients of temperature, precipitation, and seasonality. Within the broad climate zones in California, 12 Lasthenia species and subspecies are primarily found in or near isolated ephemeral wetlands known as vernal pools. We consider these vernal pool wetlands to be “environmental islands” that represent abrupt discontinuities in the environmental space within a landscape (Ackerly 2003), and it is at this “habitat” scale that we classified species and subspecies according to their affinity for these semiaquatic habitats. Finally, within vernal pool wetland communities, Lasthenia lineages span small-scale gradients in pool depth. Individual pool gradients rarely exceed 50 cm in vertical relief, yet correspond to dramatic differences in hydrological conditions driven by annual inundation patterns (Bauder 1987, Zedler 1987, Holland and Dains 1990).
A dense floral display of Lasthenia burkei (Burke's goldfields) intermixed with a few individuals of Layia chrysanthemoides (larger, white-tipped inflorescences) near Santa Rosa, California, USA. L. burkei is a state and federally endangered species that is restricted to vernal pools in the Inner North Coast Ranges bioregion of the California Floristic Province. Photo credit: D. D. Ackerly.
In this study we adopted a clade-based approach to examine the multidimensional nature of niche evolution across multiple spatial scales across the diversification of Lasthenia. Previous work examining niche evolution within clades has provided important insights into the evolution of large-scale niche axes such as climate (Evans et al. 2009, Smith and Donoghue 2010), but can rarely examine niche axes operating at smaller spatial scales due to the challenge of obtaining fine-scale environmental data for all species within entire clades. At these local scales, we can explore niche evolution by pruning the clade of interest to those taxa that co-occur in a local area: the scale at which it is possible to collect detailed data on the distribution, abundance, ecophysiology, and environmental correlates of the lineages of interest (e.g., Cavender-Bares et al. 2004, Slingsby and Verboom 2006, Graham et al. 2012, Savage and Cavender-Bares 2012). An alternative to the clade-based approach is a community-based methodology that examines the phylogenetic structure of a group of distantly related, but closely interacting species within a community or community type (e.g., Silvertown et al. 2006, Swenson et al. 2007, Dinnage 2009; Eaton et al. 2012, Lamarre et al. 2012). Studies assuming a community-based approach have illuminated the relative importance of phylogenetic and niche differentiation in structuring communities, but provide less power to investigate how niche axes evolve across speciation events throughout the diversification of a clade.
The relatively small, young, and rapidly evolving Lasthenia clade has several properties that collectively offer a particularly powerful context for investigating niche evolution across multiple axes that span different spatial scales. First, the phylogenetic relationships among all Lasthenia species and subspecies have been rigorously characterized using multiple accessions of all minimum-rank taxa sampled from across their geographical ranges (Chan 2000, Chan et al. 2001), providing the critical foundation for comparative analyses. Second, all but one species in the clade is endemic to the California Floristic Province, a relatively restricted geographic region that nonetheless encompasses a large diversity of climatic and edaphic variability. The rapid ecological diversification and contemporary distribution of Lasthenia within the California Floristic Province makes it feasible to quantify range-wide niche differences for nearly every lineage in the clade, which can span drastically different climatic conditions despite relatively small geographic ranges. Finally, the availability of extensive information on habitat associations of all Lasthenia species and fine-scale vegetation surveys in species associated with vernal pools provide a relatively unique opportunity to examine the nature of niche evolution and ecological specialization across the various spatial scales that together shape species' distribution patterns. Here, we addresses the following specific questions: (1) Do transitions into specific derived habitat types (here, vernal pools) occur multiple times during the diversification of the clade? (2) Are niche axes associated with range-wide distribution patterns more or less evolutionarily labile than axes associated with microhabitat distributions? (3) Does specialization (i.e., narrow niche breadth) at one spatial scale predict specialization at other spatial scales? Similarly, do we observe correlations in niche breadth between axes that may be functionally related, but operate at different spatial scales?
Methods
The macrohabitat niche axis: climate
We characterized the climate niche of each Lasthenia species and subspecies using precipitation and temperature GIS data obtained for known georeferenced locations of each taxon. Locality information was obtained from online specimen records at the Consortium of California Herbaria (available online),5 California-wide vernal pool vegetation sampling conducted by Barbour et al. (2007), our own vegetation surveys (see Appendix A), the California Natural Diversity Database (CNDDB), and Sloop and Ayres (2011; L. burkei only). Lasthenia kunthii (a vernal pool species in Chile) and L. maritima (a coastal specialist on avian guano deposits) were excluded from our analysis due to insufficient locality information to accurately estimate climate niche values (zero and five points, respectively). GPS locations with unknown or questionable datums were dropped from the data set.
We restricted our analysis of the Lasthenia climate niche to the California Floristic Province (CA-FP) within the boundaries of the state of California, USA, as delineated in the second edition of The Jepson Manual (Baldwin et al. 2012). The GIS polygon of the CA-FP was originally produced by the Information Center for the Environment (ICE) GIS laboratory at the University of California at Davis and is archived at the Cal-Atlas Geospatial Clearinghouse (available online).6 We extracted high-resolution monthly precipitation and temperature (maximum and minimum) data from the PRISM Climate Group (Oregon State University, created 3 December 2009, available online; Daly et al. 2008)7 and calculated five climate indices for the CA-FP: (1) mean annual precipitation, (2) mean annual temperature, (3) the coefficient of variation (CV) of monthly precipitation, (4) the standard deviation (SD) of monthly temperature, and (5) the correlation of monthly temperature and precipitation. We used different estimates of variability for precipitation and temperature because the CV is appropriate for estimating variation in ratio scale data with a nonarbitrary zero value (e.g., precipitation), while the SD is the better measure of variability for interval scale data (e.g., temperature; Zar 1999:40). Estimates of variation in monthly precipitation and temperature provide indices of seasonality, and a negative correlation between temperature and precipitation reflects the degree to which the sampled climate represents mediterranean-type conditions of cool, wet winters and hot, dry summers. We accounted for correlations among these five indices by calculating five orthogonal principal component axes that were used in all subsequent climate analyses.
We quantified niche breadth in the overall climate niche, as well as along individual climate niche axes. Overall climate niche breadth was estimated as the five-dimensional climate volume of each species by taking the convex hull of their points in climate space (Cornwell et al. 2006). Climate variables were centered and scaled to unit variances prior to principal components transformation, so the convex hull volume of points in the original five-dimensional climate space was identical to the convex hull volume of points along the five principal component axes. We calculated the geographic range for each species or subspecies as the convex hull of their distribution in geographic space (latitude and longitude). Climate volume and range size data were linearized using a natural-log-transformation. The natural log of climate volume and the natural log of range size were positively correlated (R2 = 0.772), so we also calculated niche breadth corrected for range size (hereafter the “residual climate volume”) as the residuals of the regression between ln(convex hull volume) and ln(geographic range). Niche breadth of each individual climate axis was estimated as the standard deviations of each species' component scores for each of the first three principal components of the climate niche [hereafter PC1(SD), PC2(SD), and PC3(SD)]. To separate the potentially correlated effects of species range size on the estimates of niche breadth along these individual axes, we also calculated the standard deviations of the residuals from a regression between the geographic range and each principal component.
The habitat niche axis: vernal pool and terrestrial affinity
We determined the “habitat affinity” of each Lasthenia lineage using previous vernal pool plant classifications (Keeler-Wolf et al. 1998) and species descriptions (Ornduff 1976, Chan 2000, 2012), and our own field surveys of vernal pool plant communities (Appendix A). In the CA-FP, vernal pools predominantly occur in grasslands underlain by a restrictive soil horizon or geological surface (often called a “hardpan” or “duripan,” depending on the geomorphological origin) that prevents the downward percolation of water during the winter rainy season, generating a standing water table in naturally occurring depressions in the landscape. The Mediterranean climate drives ponding in these basins during the wet winter months, followed by a heavily waterlogged phase and ultimately drought-like conditions as the rain recedes and temperature rises through the spring and into the arid summer months (Holland 1978, Holland and Dains 1990, Smith and Verrill 1998).
All species and subspecies were assigned to one of three categories: vernal pool, aquatic/terrestrial, or terrestrial (Table 1). Keeler-Wolf et al. (1998) identified all plant taxa associated with vernal pool wetlands. In their classification system, “indicators” are restricted to vernal pools and not known from other habitats, “associates” are species that regularly occur in vernal pools as well as other similar wetland habitats, and “generalists” are species that are found near vernal pools but also occupy a variety of other habitat types. Lasthenia species and subspecies that were not listed by Keeler-Wolf et al. (1998) and are described as occupying nonaquatic, non-vernal-pool habitats by taxonomic experts (Ornduff 1966, Chan 2000, 2012, Chan et al. 2001) were classified as having a terrestrial-habitat affinity. Lasthenia species and subspecies described as either “vernal pool indicator” or “vernal pool associate” species by Keeler-Wolf et al. (1998), that had vernal pools listed as their primary habitat by taxonomic experts (Ornduff 1966, Chan 2000, 2012, Chan et al. 2001), and that fell within pool boundaries in our vegetation surveys (Fig. 1; see Appendix A) were classified as having vernal pool habitat affinities. Some of these species occur in other semiaquatic habitats in addition to vernal pools (e.g., marsh edges, ephemeral ditches, pond margins); thus, here we used “vernal pool” affinities to broadly refer to all relatively isolated, ephemeral wetland community types. Finally, species and subspecies that did not have vernal pools listed as a primary habitat type by taxonomic experts, yet fell at or above pool edges in our vegetation surveys (Fig. 1; Appendix A) were distinguished as “aquatic/terrestrial.” These aquatic/terrestrial taxa were classified as either vernal pool associates or habitat generalists by Keeler-Wolf et al. (1998).
Characterization of habitat and microhabitat affinities of Lasthenia taxa sampled in vernal pool vegetation surveys in the California Floristic Province, USA. The solid circles represent the mean depth of maximum abundance across pools (i.e., mean depth position); error bars indicate standard errors. Boxes correspond to the habitat affinity classification in Table 1: the black box identifies vernal pool species and subspecies, and the gray box identifies aquatic/terrestrial taxa. The horizontal lines below taxa names indicate independent transitions into vernal pool habitat (1) or terrestrial/aquatic habitat (2–4), and the circled numbers correspond to the transition points labeled in Fig. 2. Lasthenia lineages that are not illustrated were classified as terrestrial species and subspecies (with the exception of L. kunthii, a Chilean species not included in this analysis). These depth data were used in all subsequent comparative analyses involving the local niche axis, though L. chrysantha was not included because its depth estimate was limited to a single site (Appendix A). Climate data were acquired for all taxa listed in Table 1.
A representative ancestral state reconstruction of habitat affinities in Lasthenia using Lagrange (Ree and Smith 2008). The presented phylogeny was randomly selected from all trees containing the most commonly observed habitat transition rates (Table 3). Circled numbers represent independent transitions from terrestrial habitat into either vernal pool habitat or terrestrial/aquatic habitat (see Fig. 1).
The microhabitat niche axis: depth within pools
The annual cycles of flooding and drought in vernal pools establish sharp gradients in many abiotic variables over relatively shallow depth gradients (Lathrop 1976, Holland and Dains 1990), and plant species segregate across these gradients to occupy specific depths with respect to the underlying hydrologic regime (Bauder 2000, Barbour et al. 2003, Emery et al. 2009). We quantified the position and breadth of the within-pool (depth) niche of Lasthenia lineages using the abundance and microelevation data collected in our vernal pool surveys (Appendix A). Within each pool, we selected the plot (or plots) with maximum abundance observed for each species or subspecies. The relative depth (i.e., the vertical distance above or below the pool edge) associated with this plot (or plots) was averaged across pools to obtain a single niche optimum and standard error for each Lasthenia species or subspecies. The breadth of the pool depth niche (hereafter “depth range”) was estimated from the range of pool depths occupied by each species or subspecies, calculated as the difference in microelevation between the deepest and shallowest plots occupying the middle 90% of the microelevational distribution for the species in each pool, averaged across all pools.
Phylogenetic uncertainty in Lasthenia
As is expected for a young, rapidly evolving clade, there is considerable uncertainty in the estimate of the Lasthenia phylogeny. We accounted for this uncertainty by (1) reanalyzing the sequence data from Chan et al. (2001) using Bayesian techniques, and using the posterior distribution of trees in all subsequent phylogenetic analyses; and (2) using replicated randomization procedures for each tree in this distribution to account for uncertainty associated with pruning trees with all accessions down to species-level trees for ancestral state reconstructions and comparative analyses.
Chan et al.'s (2001) sequence data includes concatenated nuclear ribosomal ITS and ETS and chloroplast (trnK intron) DNA sequences for all 21 Lasthenia species and subspecies and two outgroup taxa (Amblyopappus pusillus and Eriophyllum congdonii). Each named species or subspecies of Lasthenia is represented by a minimum of three accessions collected from geographically separated populations spanning the known range of each taxon. The alignment was partitioned by gene region and the following model for each partition was chosen using the likelihood ratio test in ModelTest (Posada and Crandall 1998): trN+G for ITS, HKY+G for ETS, and F81+G for the cpDNA partition.
A Bayesian analysis of Lasthenia relationships was conducted in BEAST (Drummond and Rambaut 2007). The models recommended by ModelTest (Posada and Crandall 1998) were implemented in BEAST by manual modification of the XML output from the BEAUti graphical interface. Lasthenia was constrained to be monophyletic. The prior for the height of the most recent common ancestor of the Lasthenia clade was fixed to 1.0 to force the trees in the posterior distribution to have a uniform height. We used the relaxed clock model with a lognormal distribution in conjunction with a Yule process speciation model so that all trees in the posterior distribution would be ultrametric.
The posterior distribution was sampled from three independent Markov chain Monte Carlo (MCMC) analyses run to completion in BEAST. The length of each chain was set to 100 million generations and one tree was sampled every 20 000 generations. The output from each chain was analyzed in Tracer version 1.4 (Rambaut and Drummond 2007) to evaluate stationarity of tree likelihood scores (for the compound set of all partitions, as well as each individual partition) and convergence of the accumulated posterior distribution using effective sample size (ESS) scores for each parameter of the MCMC. Visual inspection of the graphical output confirmed that all the tree likelihoods had achieved stationarity well before 10 million generations (10%), so a 10% burn-in was determined to be very conservative and the resultant distribution had ESS scores much greater than 200. Using LogCombiner (in the BEAST software package), the three distributions were first resampled at a frequency of once for every 200 000 generations of the original chain (only 1 tree in every 10 was retained) and combined to create a single final distribution of 1350 trees, which was examined in TRACER v.1.4 to confirm that ESS scores were still >200 for all relevant parameters for the final distribution.
Each tree in the Bayesian distribution contained a minimum of three accessions per named species or subspecies, yet our habitat affinities and niche characterizations were assigned to species or subspecies and not individual accessions. Thus, it was necessary to first collapse the accession tree to a species tree prior to phylogenetic comparative analysis. When a named species or subspecies consisted of a monophyletic group of accessions, we simply collapsed the multiple branches leading to each accession down to a single species (or subspecies) branch. However, two well-supported clades in the Lasthenia phylogeny contained accessions that did not form monophyletic groups associated with the named species or subspecies (L. californica subsp. bakeri, L. californica subsp. californica, and L. californica subsp. macrantha in sect. Amphiachaenia; and L. chrysantha, L. ferrisiae, L. glabrata subsp. coulteri, and L. glabrata subsp. glabrata in sect. Hologymne). In these clades, we resolved each tree in the Bayesian posterior distribution by randomly selecting a single accession for each named species and dropping the remaining accessions. This randomization step was repeated 10 times for each of the 1350 trees in the posterior distribution, generating a final distribution of 13 500 trees for the subsequent analyses of niche evolution.
Evolutionary transitions to vernal pool habitat
We examined the evolution of vernal pool, aquatic/terrestrial, and terrestrial habitat affinities in Lasthenia using a likelihood DEC (dispersal–extinction–cladogenesis) ancestral range reconstruction in Lagrange (Ree and Smith 2008). This approach incorporates a traditional likelihood framework (Pagel 1994, Schluter et al. 1997), but differs from character reconstruction by allowing cladogenesis events to be coincident with range splitting or expansion (i.e., dispersal events), rather than requiring two daughter lineages to inherit the character state of their common ancestor (Ree et al. 2005, Ree and Smith 2008). In our analysis, this allowed each node or tip to occupy vernal pool habitat, terrestrial habitat, or simultaneously occupy both habitats (aquatic/terrestrial). We assigned each extant taxon its respective habitat affinity (Table 1). For the results presented here, we classified the branch subtending the root node to a terrestrial habitat affinity because the habitat affinities of outgroup taxa indicate that the ancestor of the Lasthenia clade occupied terrestrial habitat; relaxing this constraint did not significantly alter the results. The ancestral state reconstruction was conducted for each of the 13 500 species-level trees in the posterior distribution.
Local and range-wide niche evolution
We tested for phylogenetic conservatism and lability in the local (pool depth) and large-scale (climate) niche axes using Blomberg's K (Blomberg et al. 2003). Blomberg's K compares the distribution of trait values across a phylogenetic tree to the expected distribution under a random walk model of evolution. We incorporated phylogenetic uncertainty by estimating K for each of the 13 500 species-level trees and then calculating the mean K of this distribution. We tested if this mean K value was significantly different from the expectations of K under two evolutionary models: (1) the absence of phylogenetic signal and (2) the Brownian motion expectation (see Ackerly 2009). First, we used a Monte Carlo permutation test (Evans et al. 2005) to create a null distribution of K values by randomizing the niche parameters across the tips once on each of the 13 500 trees. Second, we generated the distribution of K values expected under Brownian motion evolution by simulating trait evolution once along each tree using the sim.char function in the R software package GEIGER (Harmon et al. 2008). In both cases, we considered a trait to be significantly more conserved than expected under a given model if the mean observed K exceeded 97.5% of values in the distribution, and significantly divergent if it fell below 97.5% of the values.
We evaluated the hypothesis that specialists along one niche axis are also specialists along other niche axes by testing for correlations in niche breadth estimates between different axes. We tested for significant correlations between each pairwise combination of PC1(SD), PC2(SD), and PC3(SD) using all California Lasthenia species (n = 18), and correlations between pool depth range and PC1(SD), PC2(SD), PC3(SD), range-wide precipitation variation (estimated as standard deviation of ln(mean annual precipitation)) and the natural log of climate volume in the subset of Lasthenia that were categorized as vernal pool and aquatic/terrestrial lineages (n = 11; L. chrysantha, a vernal pool species, was dropped because only a single population had been sampled for its depth niche). We also conducted a separate analysis to test the specific hypothesis that the breadth of precipitation conditions across the range of each vernal pool taxa is correlated with the range of depth positions spanned within pools. All correlations described to this point (hereafter “TIP correlations”) were performed without controlling for phylogenetic nonindependence. To investigate the effect of common ancestry on these correlations, we repeated all of these analyses using independent contrasts (hereafter “PIC correlations”; Felsenstein 1985). We accounted for phylogenetic uncertainty by conducting independent contrasts across each of the 13 500 trees. We used correlation through the origin to test for a significant relationship between niche breadth estimates for each of the 13 500 trees to generate a distribution of the correlation coefficients for each comparison (Garland et al. 1992). To evaluate the cumulative significance of these correlations, we report the mean correlation over this distribution (and its significance), as well as the fraction of trees (out of 13 500) with correlation coefficients that were significantly positive or negative. The significance of each TIP and PIC correlation was assessed without correction for multiple comparisons because we had prior expectations that specialization would not be correlated across axes and we did not want to bias our tests toward acceptance of the null hypothesis of no correlation.
Results
Lasthenia species and subspecies span climatic gradients in California that vary in temperature, precipitation, and seasonality. The first three principal component axes explain 85.7% of the variation in the climate variables for Lasthenia distributed in California (Table 2). The first axis (PC1, 46.5% of the total variation) captured an approximately north–south gradient from cooler and wetter conditions in the North Coast Ranges and northern Sierra Nevada to hotter and drier conditions in central and southern California. The second principal component axis (PC2; 22.9% of the total variation) captured variation between a coastal, maritime climate with high rainfall and moderate seasonality in temperature and the arid conditions associated with inland locations in the Central Valley. The third axis (PC3; 16.3% of the total variation) primarily reflects the correlation between temperature and precipitation, a measure of the degree of mediterranean-like seasonal conditions (cold, wet winters and warm, dry summers).
Our vegetation surveys quantified microhabitat differences in distributions across the inundation gradient within pools for vernal pool and aquatic/terrestrial taxa (Fig. 1). Vernal pool species L. glaberrima, L. conjugens, L. fremontii, and L. burkei consistently occupied microhabitats below the pool edge, while the distributions of L. glabrata subsp. coulteri, L. glabrata subsp. glabrata, L. ferrisiae, and L. chrysantha were all typically found around the edges of the pools. Lasthenia aquatic/terrestrial lineages (L. californica subsp. californica, L. gracilis, and L. platycarpha) varied in the degree to which they encroached into vernal pools. These species are most abundant in the uplands surrounding the vernal pools, and thus, our vernal pool sampling transects only captured the lower ends of their distributions. As a result, our depth estimates for these species are downwardly biased. Yet, we did observe that L. californica subsp. californica and L. gracilis extend closer to the edges of vernal pools than L. platycarpha, and in several cases we observed L. gracilis populations overlapping the pool edge.
Ancestral state reconstructions provide strong support for four transitions from solely terrestrial habitats to an affinity for both aquatic and terrestrial habitats (97% of all trees; Table 3). One of these transitions preceded the diversification of the nine lineages classified as vernal pool species and subspecies (including the Chilean vernal pool species L. kunthii). Two or three independent transitions from aquatic/terrestrial to vernal pool habitat affinities occurred in 30% and 70% of the trees, respectively. A majority of trees (65%) included two reversals from aquatic/terrestrial to strictly terrestrial (the L. debilis/L. microglossa clade and the L. minor/L. maritima clade). Relaxing the constraint of fixing the root node to a terrestrial-habitat affinity resulted only in minor shifts in the timing of transitions, but the number and topological positions of transition events were equivalent to those observed in the constrained model. A single phylogeny representing a common tree topology and habitat affinity reconstruction scenario is provided in Fig. 2. In this example (and in >99% of all reconstructions; Table 3), the most recent common ancestor of the strictly vernal pool taxa is reconstructed to have an aquatic/terrestrial-habitat affinity. As illustrated in Fig. 2, this reconstruction occurred even though no extant taxon in that clade exhibits the aquatic/terrestrial state. The initial persistence of the vernal pool species and subspecies in aquatic/terrestrial habitat is a byproduct of implementing the biogeographical model to reconstruct habitat states: Allowing for the simultaneous occupation of both aquatic and terrestrial habitats permits species to persist in this polymorphic state for a longer period of time before transitioning to strictly vernal pool habitats.
Phylogenetic niche conservatism in Lasthenia was observed for the local pool depth niche axis but not range-wide climate axes (Table 4). Mean depth position within vernal pools showed significantly more phylogenetic signal than expected under a null model of evolution generated using tip-swap randomization and was not significantly different than the amount of change expected under a phylogenetically dependent Brownian motion model. By comparison, large-scale niche axes (i.e., climate axes) were relatively labile and consistently showed more divergence than expected under Brownian motion (Table 4; Appendix B). We also did not detect significant phylogenetic signal in pool depth range or the convex hull of the residual climate volume, indicating that these estimates of small-scale and large-scale niche breadth, respectively, have also been relatively labile throughout the diversification of this clade.
Niche breadth estimates along individual environmental axes were not consistently correlated, indicating that specialists along one niche axis were not predictably specialized along all other axes (Fig. 3, Table 5). Two of three TIP correlations and PIC correlations found significant positive relationships between niche breadth estimates along different climate axes, but only one remained significant after correcting for range size (PC1(SD) and PC3(SD); Table 5). Although the breadth of the pool depth niche was positively correlated with the overall climate volume, the only significant relationship between depth range and individual climate axes was a positive PIC correlation with PC2(SD). We identified a marginally significantly negative TIP correlation between pool depth range and range-wide precipitation variation, indicating that species with relatively broad distributions across local inundation gradients tend to occupy narrower precipitation ranges. However, this correlation was clearly driven largely by patterns of shared ancestry because the corresponding PIC correlations were not significant (Table 5, Fig. 4).
A subset of TIP correlations (see Methods: Local and range-wide niche evolution for clarification) between niche breadth estimates for microhabitat (vernal pool depth) and macrohabitat (overall climate) niche axes. The microhabitat niche breadth of each species is the difference in microelevation (m) between the deepest and shallowest plots bracketing the middle 90% of each population's microelevational distribution in each pool, averaged across all sampled pools containing that species. Residual climate volume is a unitless quantity that captures the convex hull volume of all five principal components used to describe the climate niche of each species, corrected for the correlation between climate volume and geographic range. Range-wide precipitation variation and individual principal components were estimated as the standard deviations (SD) of the values or scores, respectively, among locations within each species. Precipitation data were ln-transformed prior to analysis. Points represent niche breadth estimates for each of 11 vernal pool and aquatic/terrestrial taxa and eight terrestrial taxa [included in panels (c) and (d) only]. The figures illustrate TIP correlation tests between (a) pool depth range (microhabitat niche breadth) and climate volume (overall climate niche breadth); (b) pool depth range and range-wide precipitation variation (SD[ln(mean annual precipitation)]); (c) SD in the PC scores for two independent climate niche axes (PC1 and PC2); and (d) SD in the PC scores for two independent climate niche axes (PC1 and PC2) after correcting for the potentially confounding effects of range size variation among taxa. None of the correlations presented here was statistically significant (Table 5).
Testing for phylogenetically independent contrasts (PIC) correlations (see Methods: Local and range-wide niche evolution for clarification) between niche breadth axes. Data are the independent contrasts of pool depth range (m) and the standard deviation in mean annual precipitation (precipitation breadth) in vernal pool and aquatic/terrestrial Lasthenia lineages. (a) A representative PIC correlation between niche breadth estimates for one of the 13 500 trees in the Bayesian posterior distribution. (b) Frequency distribution of PIC correlation coefficients from all 13 500 trees in the Bayesian distribution. The dashed lines indicate the proportion of the distribution with correlation coefficients that were significantly negative (left of dashed line on the left) or significantly positive (right of dashed line on the right). Although the vast majority of the correlation coefficients were weakly positive, most were not statistically significant, indicating that niche breadth across depth gradients within pools is not correlated with mean annual precipitation niche breadth after accounting for shared ancestry. The corresponding TIP correlation was marginally negatively significant (Table 5).
Discussion
Our results indicate that the ecological niche in Lasthenia has evolved to encompass vernal pool habitat four separate times during the diversification of this clade, but only one of these transitions led to the subsequent proliferation of vernal pool endemic lineages. Furthermore, evolution in Lasthenia has occurred relatively independently along axes spanning different climatic dimensions and different spatial scales. In this young and ecologically diverse clade, range-wide climate niche axes are rapidly evolving and are relatively labile, while the local pool depth axis is comparatively conserved among vernal pool and aquatic/terrestrial lineages. The degree of specialization (measured as the variation along individual niche axes) was also labile at both local and range-wide spatial scales, as neither residual climate volume nor pool depth range showed a tendency to remain phylogenetically conserved among close relatives. Similarly, the breadths of individual niche axes were not consistently correlated with one another, indicating that specialization along one axis did not predict specialization along other axes.
Over the past decade, numerous studies have integrated species habitat or trait data with species-level phylogenies to investigate niche evolution across speciation events. At large spatial scales, georeferenced occurrence information and high-resolution climate data have been examined in a phylogenetic context to suggest that divergence (e.g., Graham 2004, Evans et al. 2009) and conservatism (e.g., Peterson et al. 1999, Kozak and Wiens 2006) in the climate niche can each promote speciation by different mechanisms. Along more local niche axes, competitive interactions among co-occurring relatives are understood to promote niche diversification, which in turn, may promote rapid evolution and adaptive radiation of lineages colonizing island-like habitats (Schluter 1996, Losos et al. 2003, Givnish et al. 2009). In the recent, rapidly evolving clade of Lasthenia, we found that the large-scale climate axes have been highly labile throughout the history of the genus, while the microhabitat niche axis associated with positions along local depth gradients was relatively conserved among lineages, suggesting that the pool depth niche is evolving in a phylogenetically dependent manner (Table 4). Because competitive interactions are likely to be strongest among close relatives (Darwin 1859, Elton 1946, Vamosi et al. 2009, Burns and Strauss 2011) and at the most local spatial scales (Weiher and Keddy 1999, Cavender-Bares et al. 2006), conservatism in the local pool depth niche suggests that either (1) close relatives in Lasthenia rarely occur in sympatry and thus have not experienced divergent selection for microhabitat associations, and/or (2) these interactions have led to competitive exclusion and phylogenetic evenness within local communities (Webb et al. 2002). Although many Lasthenia species (and several of the vernal pool species in particular) have largely overlapping geographic ranges (Ornduff 1966, Chan 2000) and sympatry can be extensive at scales of <1 km (e.g., up to eight taxa at the Arena Plains site), niche conservatism in depth associations is reflected by phylogenetic overdispersion at the scale of individual pools (N. C. Emery, E. J. Forrestel, and D. D. Ackerly, unpublished data). These patterns are consistent with the results of Ackerly et al. (2006), where conservatism in traits is associated with the “α niche” (within-habitat niche) compared to “β niche” (among-habitat niche) in the woody California plant group Ceonothus (but see Silvertown et al. 2006 for a counterexample in a meadow plant community). As a corollary, increasing levels of phylogenetic evenness with narrowing phylogenetic and spatial scales have been demonstrated in other plant clades (Cavender-Bares et al. 2006, Swenson et al. 2006) and may be an emerging general pattern in the phylogenetic structure of communities.
The multidimensional analysis of niche breadth in Lasthenia reinforces the assertion that specialization along one niche axis does not necessarily reflect an overall restricted ecological niche for a species, challenging the assumption that edaphic endemics are characterized by narrow ecological niches that may make them vulnerable to extinction under rapidly changing climatic conditions (Futuyma and Moreno 1988, Devictor et al. 2010). We tested the hypothesis that “universal specialists” exist in Lasthenia by comparing all estimates of niche breadth along the local pool depth axis and principal component axes describing precipitation, temperature, seasonality gradients, and the overall volume of the climate niche envelope (Table 5). In one case, a pair of statistically significant TIP and PIC correlations was no longer significant once the correlated effects of range size on climate breadth were taken into account (PC1(SD) and PC2(SD) in Table 5). This indicates that widespread species may be broadly distributed along each of these climate axes, but do not occupy a greater or lesser range of conditions than expected given their geographic range. At a mechanistic level, it would be useful to examine “null geographic ranges” constructed on the California landscape to determine if the relationship between geographic and climatic range simply reflects the spatial structure of climate variability, or if there is evidence of climatic factors shaping species' geographic ranges (see Gotelli et al. 2009).
Our multidimensional, multi-scale approach to studying niche evolution in Lasthenia suggests several potential hypotheses regarding the evolutionary history of this clade that are not readily apparent from examining niche axes independently. The majority of California's endemic flora are considered to be recently evolved neoendemics (Stebbins and Major 1965) whose diversification is associated in part with the rise of the novel mediterranean-type climate combined with climatic variation throughout the region since the late Miocene to early Pliocene (7 to 4 million years ago; Millar 2012). Many botanists have argued that edaphic factors play an equal or larger role in explaining patterns of diversity in Lasthenia (Rajakaruna 2003) and vernal pool endemics in general (Jain 1976, Stebbins 1976, Kruckeberg 2002). While vernal pool lineages are obviously adapted to tolerate flooding stress, many of the non-vernal pool Lasthenia species are similarly restricted to osmotically stressful, patchily distributed habitats (Rajakaruna 2003), ranging from serpentine outcrops to bird guano deposits (Ornduff 1966, 1976, Chan 2000, Rajakaruna 2003), so the hydrological history of the CA-FP may have played a particularly important role in the diversification of this clade. Geological data and fossil evidence indicate that the Great Central Valley of California was periodically inundated during the Pleistocene as glaciers melted during warming periods (Raven and Axelrod 1978, Millar 2012), and the large shallow inland lake known as Lake Clyde (also Lake Corcoran) likely reached its maximum extent throughout the Central Valley between 600 000 to 700 000 years ago (Harden 2004). Thus, the current island-like distribution of vernal pool habitat is markedly different from the ecological context in which much of the diversification of Lasthenia likely occurred. Our results for Lasthenia indicate that, within this historical context, the proliferation of lineages restricted to vernal pool wetlands likely followed a single transition into ephemeral wetland habitats (Fig. 2, Table 3) and, subsequently, that close relatives have been more likely to remain in similar depth positions along wetland gradients than to occupy similar climatic niche envelopes (Table 4; Appendix B). These patterns are consistent with the hypothesis that the initial invasion into the “vernal pool niche” may have taken place when wetlands were much more widespread throughout California (Stebbins 1976). As these larger, more contiguous wetlands receded into smaller, relatively isolated pools and aquatic archipelagoes, conservatism in traits associated with osmotic tolerances may have facilitated the early stages of divergence by restricting populations to increasingly isolated habitats and limiting gene flow among previously connected subpopulations. Thus, niche conservatism in traits associated with adaptations to microhabitat variation in osmotic stress may have facilitated allopatry, genetic divergence, and ultimately speciation among Lasthenia lineages (Wiens 2004). If the contemporary climatic associations of these species reflect different climatic adaptations, niche conservatism and specialization along local axes (e.g., specialization to vernal pool habitats) may have limited the ability of species to track shifts in climate and promoted in situ adaptation to climate (Ackerly 2003). The lack of fossil calibration for the Lasthenia phylogeny makes it difficult to precisely examine the degree to which divergence events align with these geological landmarks in the history of the CA-FP. Furthermore, experimental tests of environmental tolerances will be critical to predict the responses of Lasthenia (and other edaphic specialists) to the climatic axes investigated here, as observed field distributions do not provide direct evidence of physiological tolerances. Even broadly distributed species may be composed of specialized populations or sublineages that each exhibit relatively narrow environmental tolerances (e.g., Kelly et al. 2011). Consequently, it will be important to collectively consider the phylogeographic structure found in some taxa of Lasthenia (e.g., L. gracilis, Chan et al. 2002), the contemporary population structure in all taxa, and the spatial distribution of climatic responses and gene flow to fully evaluate the impacts of local and climatic variation on speciation patterns, and the potential responses of Lasthenia lineages to future climate change.
Acknowledgments
The authors thank Jennifer Buck for her generous and critical contributions to this work, including field assistance, plant identification, facilitating access to multiple field sites, and general expertise in the flora of California vernal pools. R. Solan also provided invaluable field assistance during the spring 2007 sampling period. We acknowledge the following individuals, agencies, and institutions for providing georeferenced locality information for Lasthenia: M. Barbour and colleagues, the Consortium of California Herbaria, the California Natural Diversity Database, and C. Sloop and the Laguna de Santa Rosa Foundation. We are very grateful for the cooperation of many landowners and property managers for granting us permission to sample vernal pool vegetation, including the California Department of Fish and Game (Stone Ridge Preserve, North Table Mountain Preserve), the United States Department of Fish and Wildlife (Sacramento National Wildlife Refuge, San Jose National Wildlife Refuge, Arena Plains unit of the San Luis National Wildlife Refuge), the University of California Natural Reserve System (Jepson Prairie Preserve), the Sacramento Department of Economic Development (Mather Field), the Nature Conservancy (Vina Plains Preserve, Howard Ranch), the United States Department of the Interior Bureau of Land Management (Fort Ord, Carrizo Plain National Monument), and the United States Air Force (Travis Air Force Base). We also acknowledge the generous assistance of several individuals who facilitated the identification and accessibility of various field sites: Christina Sloop and Gene Cooley (Santa Rosa Plain), Jenny Marr (Stone Ridge Preserve), Sharon Collinge (Travis Air Force Base), and Joe Silveira (Sacramento National Wildlife Refuge). Permission to sample populations containing state-listed species was granted under CDFG Research Permit 07-02-RP. This work was funded by NSF DEB-06213 to D. D. Ackerly and B. G. Baldwin.
Supplemental Material
Appendix A
Sampling protocol, site names, and GPS coordinates for vegetation sampling that estimated the depth of occurrence for Lasthenia lineages occupying vernal pools (Ecological Archives E093-183-A1).
Appendix B
Extended table summarizing the estimates of phylogenetic signal (Blomberg's K statistic) and tests of significance for niche parameters not included in Table 4 (Ecological Archives E093-183-A2).
