Endemic plant communities on special soils: early victims or hardy survivors of climate change?


Correspondence author. E-mail: damschen@wisc.edu


1. Predicting and mitigating climate change effects on ecological communities is a tremendous challenge. Little attention has been given to how endemic-rich communities on isolated patches of low-nutrient soil (e.g. serpentine) will respond to climate change.

2. To address spatial factors (the isolated nature of outcrops), we incorporate habitat patchiness into species distribution models under climate change. The degree of overlap between current and future suitable habitat does not change when patchy habitats are incorporated, probably because serpentine occurs in mountainous regions where climatically and edaphically suitable regions geographically coincide. The dispersal distances required to move to newly suitable habitat are large, however, making successful migration unlikely.

3. To address how non-spatial factors affect the climate change responses of serpentine plant communities (e.g. the impacts of nutrient limitation and stress-tolerant functional traits), we conduct a literature review. Some studies suggest that serpentine communities may be at less risk than ‘normal’ soil communities due to their stress-tolerant functional traits, but there is also evidence to the contrary.

4.Synthesis. Assessing climate change risk for the world’s diverse edaphic floras requires determining impacts on both special and ‘normal’ soil communities. Studies are needed that use functional traits, evaluate the role of evolutionary and ecological plasticity, examine responses across spatial and temporal scales and assess the efficacy of managed relocation efforts.


One of the greatest challenges that ecologists and land managers face today is anticipating how climate change will affect the diversity and composition of ecological communities to develop effective strategies for adaptation and mitigation (e.g. Burkett et al. 2005; Glick, Stein & Edelson 2011; Klausmeyer et al. 2011). The direct effects of climate change on communities via changes in temperature and precipitation have been the focus of many studies (e.g. Beckage et al. 2008; Lenoir et al. 2008; Moritz et al. 2008). However, one aspect that remains little studied is how the direction and magnitude of climate change effects may vary among communities (i.e. ‘ecological contingency’ sensu Harrison, Damschen & Grace (2010)) based on differences in factors such as limiting resources, species functional traits and the spatial distribution of suitable habitats.

Plant communities found on patches of unusual bedrock and soil types, such as serpentine, gypsum, limestone, dolomite and shale (i.e. ‘special soil’, ‘azonal’, or ‘low nutrient’ communities), are an interesting case in point. These communities contain many endemic species and make large contributions to regional floristic diversity. For example, in California, 35% of the state’s 1742 rare plant species occur on special substrates (Skinner & Pavlik 1994). Because the plant communities associated with these habitats contribute disproportionately to global diversity, it is important to ask whether they will be especially at risk or relatively protected from the direct and indirect impacts of global climate change.

Plant communities on special soils have two distinctive attributes that may cause them to respond uniquely to climate change. First, they are often found in discrete outcrops making them more spatially isolated from one another than ‘normal’ soils that tend to be more contiguous (hereafter, we refer to this set of considerations as ‘spatial factors’). For example, serpentine outcrops across the world are seldom found in outcrops of more than a few hundred square kilometres, with some notable exceptions including New Caledonia; southern Oregon, USA; Sulawesi; and eastern Cuba (Brooks 1987). This spatial isolation may make it much more difficult for species to successfully migrate under climate change. Second, because these communities are on unproductive substrates, they may differ from communities on ‘normal soils’ in terms of limiting resources, functional traits, and the relative importance of disturbance, competition and other ecological processes (hereafter, we refer to these as ‘non-spatial factors’). Plants in these special soil habitats often have traits associated with tolerance of drought and nutrient limitation [e.g. small stature, low-specific leaf area (SLA), high allocation to roots relative to shoots] because nutrient availability is limited, water can be scarce, and soils may have additional unusual chemistries (e.g. presence of heavy metals or particularly acidic or basic pH). Special soil communities are more strongly water limited than others; therefore, they may be especially responsive to changes in available precipitation. On the other hand, because plants on special soils already have adaptations for stress tolerance, they may be particularly well suited to withstand climatic changes. These features may also create unique feedbacks with other processes such as fire, herbivory and nutrient deposition.

Here, we ask how plant communities on serpentine (ultramafic) soils will be affected by climate change, relative to those on less extreme soils, using two complementary approaches. To assess the importance of spatial factors, we incorporate the configuration of serpentine habitats into species distribution models and determine whether spatial isolation alters the likelihood of persistence for edaphic endemic plant species under future climate change scenarios. To address the role of non-spatial factors, we conduct a primary literature review to find the available evidence for how plant communities on serpentine and non-serpentine soils may differ in their responses to climate change. We focus on serpentine plant communities because most of the limited evidence on climate change and special soils comes from serpentine. In the discussion, we return to the applicability of our findings from serpentine for other soil substrates.

Serpentine study system

Serpentinite and peridotite (collectively called ultramafic or simply serpentine) rocks are found throughout the world, primarily where oceanic crust and mantle have been exposed on continents (Alexander et al. 2006). Soils weathered from these rocks are extremely magnesium-rich and calcium-poor compared with most other soils and are also typically low in macronutrients (especially P and K) and sometimes also high in heavy metals (Ni, Cr, Co) (Alexander et al. 2006). In many cases, these soils have very high rock fragment content, leading to a scarcity of available water (Alexander et al. 2006). Sparse canopy cover may also contribute to high temperature and low moisture near the ground surface. Serpentine vegetation is nearly always lower in biomass than surrounding ‘zonal’ vegetation, but in some parts of the world, serpentine supports a wealth of endemic plant species (Kruckeberg 2005). Serpentine soils and their associated plant communities (Harrison and Rajakaruna 2011) are distributed globally (Brooks 1987) and have been studied in many places, including (but not limited to): Asia (Nakata & Kojima 1987; Rajakaruna & Bohm 2002); Australia (Specht, Forth & Steenbeeke 2001); Cuba (Borhidi 1988; Iturralde 2001); Europe (Rune 1953; Kram et al. 2009); New Caledonia (Proctor 2003; Pillon et al. 2010); New Zealand (Robinson et al. 1997); South Africa (Smith, Balkwill & Williamson 2001; Williamson & Balkwill 2006); South America (Reeves et al. 2007); and California and Oregon, USA (Kruckeberg 1984).

Spatial factors


Plants that are soil specialists face unique challenges under climate change relative to soil generalists because suitable habitat patches are spatially isolated rather than contiguous. Species distribution models have been used to determine species’ current distributions and project future distributions under climate change. Such models, however, have typically used climate data alone to define and project species distributions. While these models have generally predicted upward elevational and latitudinal species shifts under climate change that match empirical evidence (e.g. Parmesan et al. 1999; Lenoir et al. 2008), they overlook factors besides climate that define species’ niches and may more poorly predict distributions for species with spatially disjunct habitats such as edaphic endemics.

Relative to habitat generalists, habitat specialists such as edaphic endemics face two unique problems under climate change (Fig. 1). First, species distribution models have assumed that in the absence of dispersal, species will survive as long as there is geographical overlap between their present and future climate envelopes (e.g. Thomas et al. 2004; Schwartz et al. 2006; Loarie et al. 2008). This may be true for a soil generalist that, in the absence of ecotypic variation, can survive anywhere within the area of overlap between its current and future ranges. However, it is certainly not true for a soil specialist that depends on suitable soils to exist within the area of climatic overlap. For limestone soil specialists inhabiting the southern United Kingdom, for example, their projected future climatic envelopes in the northern UK are lacking in limestone (Berry et al. 2003). Second, a generalist can move through continuous habitat to reach newly suitable climates, but a specialist must make large jumps to disperse from one suitable habitat patch to another, crossing areas of unsuitable habitat and decreasing the likelihood of successful dispersal. Both of these issues make it less likely that soil specialists will be able to find or reach suitable habitat under future climate change scenarios.

Figure 1.

 Conceptual model of how present and future climate changes will affect (a) soil generalist and (b) soil specialist species differentially. Blue represents the current range of a species, Green is the future range of a species and Red is the area of overlap between the current and future ranges. Patchy suitable habitat for the soil specialist creates fewer colonization opportunities among the current and future ranges than for the soil generalist with contiguous suitable habitat.

Species distribution modelling for edaphic endemics (and other habitat specialists) is challenging, because models typically assume that species distributions are entirely dictated by climate, and this is clearly not the case for species that have additional requirements such as particular soils (Schwartz et al. 2006). Effectively, the geographical distribution of an edaphic endemic conveys less information about its climatic tolerances than does the geographical distribution of a soil generalist, as the species’ distribution will not fill its potential, climatically suitable range (Schwartz et al. 2006). The technical problems of how to incorporate additional requirements such as soils into the model-building and projection phases of species distribution modelling have yet to be surmounted. Still, to provide a general illustration of the spatial problems faced by soil specialists, we built on existing modelling approaches and focus on outcomes that are relatively robust to model assumptions: first, the relative amounts of suitable serpentine habitat in the hypothetical future climatic range vs. in the present range, and second, the mean distances that species must travel to reach new serpentine outcrops within their hypothetical future climatic ranges.


We used maximum entropy (Maxent) species distribution models that were created for Californian endemic plant species by Loarie et al. (2008). From these, we selected 12 of the species that are considered by Safford, Viers & Harrison (2005) to be strict endemics to serpentine. These 12 species are Ceanothus jepsonii, Cryptantha hispidula, Hesperolinon disjunctum, Linanthus ambiguus, Phacelia breweri, Phacelia corymbosa, Polystichum lemmoni, Salix breweri, Salix delnortensis, Streptanthus barbiger, Streptanthus drepanoides and Streptanthus polygaloides. We compared the area of overlap between present and predicted future climatic envelopes under two assumptions: (i) the species can only live on serpentine (thus, the present and predicted climate envelopes are ‘masked’ using a map of serpentine outcrops in California), and (ii) the species can live anywhere that the climate is suitable, as is normally assumed in species distribution models. Future climate envelopes were generated for these 12 Californian serpentine endemics 100 years into the future, under two climate and two dispersal scenarios. The first climate scenario projected a warmer and wetter future using the climate model CCMA CGCM3 1.1 (Flato et al. 2000; McFarlane et al. 2005). The second climate scenario projected a warmer and drier future using the climate model GFDL CM2 1.1 (Delworth et al. 2006). For both of these, we used the A1b emissions scenario from the Intergovernmental Panel on Climate Change (IPCC 2000), which depicts an integrated world and balanced emphasis on all energy sources [for further information see Ackerly et al. (2010) and Maurer et al. (2007)]. To evaluate the role of dispersal ability in ameliorating climate change responses, we ran both above-mentioned climate scenarios under two extreme conditions for dispersal: (i) no dispersal (the species persists only where its present and future envelopes overlap) and (ii) unlimited dispersal (the species persists throughout its future climate envelope).

Results and Conclusions

A surprising result of our models (See Fig. 2 and Appendix S1 in Supporting Information for model results) under the no-dispersal scenario was that species do not disproportionately suffer habitat loss as a result of being confined to serpentine. That is, the ratio of the area of the future predicted range to the present range was not higher (or lower) when the species was assumed to be a habitat generalist than when it was assumed to be restricted to serpentine. This implies that serpentine soils are just as abundant at higher elevations and higher latitudes where suitable climatic conditions are predicted to occur in the future, as in the geographical regions where the species exist now. This likely reflects the fact that serpentine tends to occur throughout California’s mountains, where variability in climate is buffered by the existence of rugged topography (Loarie et al. 2009). Therefore, climatically and edaphically suitable regions tend to geographically coincide; a result that may have some degree of generality for bedrock geology substrates such as serpentine, which are frequently exposed in mountainous regions (Anderson, Fralish & Baskin 1999; Kruckeberg 2005).

Figure 2.

 The (a) current range, (b) future range assuming a warmer and wetter climate and (c) future range assuming a warmer and drier climate for Ceanothus jepsonii, one of 12 species examined. In panel a, Red is the species range at present. In panels b and c, Red is the area of overlap between present and future ranges, Blue is the area that is lost due to climate change and Green is the area that is gained due to climate change. Serpentine outcrops and county boundaries are outlined in black. See Appendix S1 for maps of the remaining 11 species.

Under the scenario that included dispersal, we found that to colonize the nearest serpentine patch in their newly suitable range from a patch within their current climate envelope, the 12 species had to make minimum initial dispersal distances averaging 596 m for the warmer–wetter and 1891 m for the warmer–drier climate projection models. Subsequently, they must make minimum dispersal jumps averaging 663–8275 m to colonize each patch within their newly suitable ranges (i.e. these are the mean shortest edge-to-edge distances to each patch in the newly suitable range; Fig. 3).

Figure 3.

 The average minimum dispersal distance for soil specialist species under the two different climate model scenarios (CCMA = CCMA CGCM3 1.1, GFDL = GFDL CM2 1.1; see ‘Spatial factors’ for additional model details). One asterisk indicates that there are no new suitable habitat patches in the future. Two asterisks indicate that there is no current suitable habitat that remains suitable in the future and that only one patch becomes suitable in the future.

Several lines of evidence suggest that plant dispersal on the order of kilometres across unsuitable intervening habitat is unlikely over relatively short timescales. A recent review of dispersal studies found that plants disperse on average between 10–100 m and that average dispersal distances >1 km are uncommon (Kinlan & Gaines 2003). The exceptionally small ranges of most serpentine endemic species in California suggests that geographical spread has been minimal since they evolved restriction to serpentine (Harrison et al. 2008). Other studies have found that plants in spatially isolated habitats have evolved limited dispersal abilities (Cheptou et al. 2008; Riba et al. 2009). On serpentine soils, it appears that plant dispersal modes are shaped more by nutrient-poor soils than by the spatial isolation of their habitat patches (Spasojevic, Damschen & Harrison, in press). Therefore, rare long-distance dispersal events (Nathan 2006) are likely to be especially critical for the persistence of soil specialists relative to soil generalists.

Our analyses also suggested that three species (Linanthus ambiguus, Polystichum lemmoni and Salix delnortensis) will have no climatically suitable serpentine outcrops in the future under one or both climate projection models, although we note the above caveat about the reliability of any such absolute predictions using current species distribution modelling methods.

Non-spatial factors


Previous studies by Grime et al. (2000, 2008) and Matesanz et al. (2009) have suggested that soil fertility may alter plant community responses to climatic variation. On limestone soils, Grime et al. (2000, 2008) experimentally manipulated temperature and precipitation and measured the effects on grassland communities in two settings: Buxton, UK, an unproductive limestone grassland, and Wytham, UK, an early successional productive limestone grassland. Over a 5-year period (Grime et al. 2000), the composition of the unproductive grassland changed considerably less in response to treatments than the productive grassland, and even after 13 years, changes in the unproductive grassland were strikingly small (Grime et al. 2008). This response was attributed in part to a suite of stress-tolerant functional traits shared by the species on the unproductive site and in part to the potential for individuals at the unproductive site to move among microhabitats determined by soil depth, allowing them to persist through climatic changes (Fridley et al. 2011).

On gypsum soils, Matesanz, Escudero & Valladares (2009) experimentally manipulated rainfall and measured the effects on a specialist shrub species, Centaurea hyssopifolia, across habitats that varied in, among other factors, their ‘quality’ as measured by plant cover and soil nutrient content. In poor-quality habitat, reduced rainfall led to greater advancement of flowering and dispersal times, greater reduction in growth rate and greater increase in the fractions of senescent leaves, compared with individuals under similar rainfall treatment in higher-quality habitat.

Taken together, the above two studies suggest that climate change effects may be buffered by the community-level properties of plants on low-nutrient soils, but individual species may suffer stronger effects of climate change on low-nutrient soils than high-fertility soils.


To assess the available evidence for how climate change responses differ in plant communities on serpentine vs. non-serpentine soils, we conducted a quantitative literature search. We used the following key words to search the primary literature on Thompson Reuter’s Science Citation Index: serpentine, special soil, edaphic, non-serpentine, climate change and plant* (* = wildcard character used to find plural forms and variant spellings of word). To be selected for review, papers had to (i) use climate variation as a key predictor variable, (ii) include responses of plant communities on both serpentine and non-serpentine soils and (iii) include analyses of primary data.

Results and Conclusions

Using all combinations of our key words, 207 papers were returned. After applying the criteria above, only three studies remained: Harrison (1997), Damschen, Harrison & Grace (2010) and Briles et al. (2011). We also added one unpublished study from serpentine (Fernandez-Going, Anacker & Harrison, in press) that we were aware of from our own research.

Evidence for higher resistance

Three of the four studies indicate that serpentine communities may be less sensitive to climate change than communities on ‘normal’ soils. In a study of chaparral and oak woodland communities in northern California, Harrison (1997) found that elevational and coast-to-inland gradients had no effect on local diversity on serpentine soils. However, local diversity on non-serpentine sedimentary soils decreased with distance inland, increased with elevation and slope interacted with both of these effects. These results indicate that local diversity on sedimentary soils is greater in favourable climates and is more variable along climatic gradients than on serpentine soils. Other qualitative observations from the literature on the biogeography of serpentine plants likewise support the interpretation that the composition of non-serpentine communities may show greater variation over climate gradients than serpentine communities (Rune 1953; Whittaker 1960; Kruckeberg 1984; Brooks 1987; Borhidi 1991; see review in Harrison, Damschen & Going 2009).

Using the fossil pollen record in six lakes, Briles et al. (2011) compared woody vegetation change during the Holocene on serpentine and non-serpentine soils (granitic) in the Klamath-Siskiyou Mountains. They found that shrub and tree abundances were less variable on serpentine in comparison to granitic substrates in response to the past 15 000 years of climatic variability. On serpentine, the relative abundances of the dominant species were altered, but there was little change in species composition. Briles et al. (2011) conclude that trees and shrubs on serpentine soils were able to persist under a range of past climate conditions for the same reasons that they can tolerate nutrient deficiencies and high heavy-metal concentrations. They caution against generalizing these results to all plant species responses to future climate change in that the paleorecord lacks the ability to detect many species-level responses, particularly for herbs.

Fernandez-Going, Anacker & Harrison (in press) compared temporal variability in diversity over 10 years for serpentine and non-serpentine grasslands in northern California, USA. They found that variability in species richness and composition in response to annual variation in precipitation was lower on serpentine than on non-serpentine soils. They also found that serpentine communities were less functionally diverse and had greater numbers of species with stress-tolerant traits (short stature, low-SLA, low foliar C/N ratio) than non-serpentine communities. Mean foliar C/N ratio was a significant additional correlate of community variability over time. Of the species examined, 41 occurred on both serpentine and non-serpentine soils. On serpentine soils, these soil generalist species had lower variability in frequency of occurrence and in cover when compared with non-serpentine soils, which suggests that soil type is both directly affecting variability and indirectly affecting variability by selecting for stress-tolerant species traits.

Evidence for higher sensitivity

Damschen, Harrison & Grace (2010) examined long-term vegetation change in the Klamath-Siskiyou Mountains, Oregon, USA, using a historical data set originally collected by Robert Whittaker from 1949 to 1951 (Whittaker 1960). Whittaker sampled vegetation to determine how community composition changed along environmental gradients including elevation, topography and soil type. Since that time, the region has shown an increase in mean annual temperatures of c. 2 °C (Damschen, Harrison & Grace 2010). In 2007, these communities were resampled to examine change over time on serpentine and non-serpentine soils. Cover of nearly all herb species either declined or remained the same on both serpentine and non-serpentine soils. Species with functional traits associated with cool and moist habitats (i.e. high specific leaf areas (SLA), northern biogeographic affinities) declined more than those with opposite traits. As a result, species composition of a given site today resembles that of a warmer (more southerly) topographical position in Whittaker’s time. The observed shifts in plant species richness and cover were of greater magnitude for serpentine vs. non-serpentine soils, but changes in functional traits and species composition were similar between soils. Serpentine endemics also showed greater declines in cover than soil generalists. These results suggest that herbaceous communities on serpentine have been as strongly affected by a warming climate as communities on non-serpentine soils. Interestingly, there was little observed change in tree species composition over time, which was attributed to the longer generation times for these species (ca. 100+ years).

Reconciling opposing results

Of the four studies, we found that evaluate the relative sensitivity of serpentine plant communities to those on non-serpentine soils, three studies (Harrison 1997; Briles et al. 2011; Fernandez-Going, Anacker & Harrison, in press) suggest that serpentine plant communities are less variable under climate change while one study (Damschen, Harrison & Grace 2010) indicates that they are at least as vulnerable as those on ‘normal’ soils.

There are several potential explanations for the conflicting results. First, the results of Damschen, Harrison & Grace (2010) came from a data set in which beta diversity was extremely high, with >200 species of which very few were found in >10 plots or had >5% mean cover. This required the focus to be on the aggregate responses of communities in multivariate space, while other studies (Briles et al. 2011; Fernandez-Going, Anacker & Harrison, in press) were able to examine the responses of individual species or taxonomic groups in much greater detail. Second, the serpentine plant communities in Damschen, Harrison & Grace (2010) had considerably lower tree and shrub canopy cover than the non-serpentine communities, especially on the warm south-facing slopes where many serpentine endemic herbs are found. Therefore, it is possible that the serpentine herb communities actually experienced greater temperature increases than the non-serpentine forest herb communities, counteracting any tendency for the serpentine herbs to be more resistant. Since Briles et al. (2011) and Harrison (1997) examined tree and shrub communities and Fernandez-Going, Anacker & Harrison (in press) examined grasslands, this effect would not have pertained to their studies. Third, the study in the Klamath-Siskiyous contained larger numbers of serpentine endemics than Briles et al. (2011), Harrison (1997) or Fernandez-Going, Anacker & Harrison (in press), and these species may be at greater risk of extinction under temporal variability due to their small population sizes, small ranges or for other reasons (Harrison et al. 2008).


We have identified both spatial and non-spatial factors that may contribute to differential responses of serpentine plant communities relative to their ‘normal’ soil counterparts. Spatially, we found that serpentine endemics in California are not likely to suffer disproportionate losses of suitable habitat, since serpentine appears to be equally abundant in their present and projected future ranges; however, serpentine endemics would have to disperse unrealistically large distances between outcrops to colonize newly suitable habitat under climate change. Our review of other (non-spatial) factors suggests that serpentine plant communities may be more resistant to climate change, but conflicting evidence demonstrates the importance of understanding the roles of species turnover, overstorey composition and species endemism.

Relevance to Other Special Soils

Serpentine soils are one of many special soil types across the globe. The limestone grasslands of Europe, the shale barrens of Appalachia, and dolomite, limestone and sandstone glades across the Ozarks all support communities rich in endemic species (Anderson, Fralish & Baskin 1999). Our observations from serpentine may also hold relevance for plant community responses to climate change on other special soils. However, the available evidence for plant community responses to climate change is even scarcer for other soil types, with the notable exception of the work of Grime et al. (2000, 2008) and Fridley et al. (2011) on limestone grasslands and Matesanz (2009) on gypsum soils. After conducting a similar quantitative literature search for other special soils (keywords: limestone, gypsum, dolomite, limestone, sandstone, edaphic, climate change and plant*), no other studies were found.

When generalizing the climate change findings from serpentine to other soil types, there are several factors that need to be considered. First, not all special soils are spatially isolated. The limestone grasslands that were the subject of Grime et al. (2000), for example, were probably a relatively contiguous habitat until recent fragmentation. Second, soils may be ‘harsh’, ‘unproductive’ or ‘nutrient-poor’ in different ways, not all of which may have the same interactions with climate. Some special soils may simply be limited in primary nutrients (e.g. siliceous soils in the Arctic; Eskelinen 2008), others in water retention capacity (e.g. sandstone and shale barrens; Anderson, Fralish & Baskin 1999), others may have excessive levels of cations (e.g. Mg++-rich serpentine and dolomite; Kruckeberg 2005) and others may have additional unique factors (e.g. gypsum soils form a surface crust that is markedly different from other soil types (Pueyo & Alados 2007)). The degree of potential generality in climate change responses across these variable soil situations has yet to be explored.

Feedbacks with Other Processes

Plant communities on special soils may respond uniquely to climate change not only directly, but indirectly through feedbacks with other climate change impacts. For example, the frequency and severity of fire is increasing in response to climate change in some regions (Westerling et al. 2006). Serpentine plant communities in both the tropics and temperate zone may be more sensitive to fire than non-serpentine communities (Proctor 2003; Pillon et al. 2010). In Cuban serpentine vegetation, fire caused degradation of the vegetation and invasion by fire-prone grasses (González-Torres 2010). In California chaparral, serpentine plant communities recover considerably more slowly from fire than sandstone chaparral communities (Safford & Harrison 2004). However, some ‘special soil’ communities such as the Ozark glades are maintained by fire and are threatened by fire suppression (Baskin & Baskin 2000).

Herbivory may also differentially affect the responses of special soil and ‘normal’ soil communities to climate change. Heavy grazing has been shown to offset the effects of climatic warming on plant community composition (Post & Pedersen 2008), but it may be deleterious to communities on low-fertility soils where plants lack tolerance to heavy grazing (Proulx & Mazumder 1998). On gypsum soils, livestock grazing has been shown to cause community homogenization and the decline of endemic species (Pueyo et al. 2008).

Atmospheric N deposition can be as great as 20 kg N ha−1 year−1 (Phoenix et al. 2006) and may also interact with climate change responses. Amelioration of nutrient limitation may reduce the resilience of low-productivity communities to climate change, as has been shown on limestone grasslands (Grime et al. 2000) and in nutrient-poor arctic tundra (Klanderud & Totland 2005). Nutrient enrichment may also facilitate invasion by non-native species, which may be more responsive to climate change because of their ‘ruderal’ traits. Serpentine grasslands near San Francisco, USA, have been invaded by the exotic grass Lolium multiflorum as the result of nitrogen deposition (Weiss 1999). While infertile soils are generally more resistant to invasion, they may be highly susceptible to climate-driven change if nutrient enhancement accompanies the arrival of fast-growing invaders (Grime et al. 2000, 2008).

Below-ground feedbacks may interact with climate change and have large impacts on special soils. One study has shown that the effects of climate change in low-nutrient communities may be slowed or altered by the chemical composition of the litter, which makes it recalcitrant to decomposition (Wardle, Walker & Bardgett 2004). Other studies have demonstrated climate change effects on soil cation availability (Holzinger et al. 2008) and the composition of physical soil crusts (Pueyo & Alados 2007; Pueyo et al. 2007), which may feed back to alter plant community composition.

Finally, climate change may not only affect the location of suitable habitat, but also dispersal trajectories themselves. For example, wind dynamics and dispersal distances for several wind-dispersed plant species in England are predicted to decrease under climate change, although with a high degree of uncertainty (Bullock et al. 2012). Wind-dispersed species are particularly prevalent in serpentine plant communities (Spasojevic, Damschen & Harrison, in press), suggesting that climate change may have unique effects on serpentine via alterations in dispersal trajectories.

Importantly, studies that address these interactions must explicitly include treatments on both special and ‘normal’ soils to be able to determine whether the interactions are stronger or different on special soils.

Future directions

Our single most important recommendation for future research is that studies compare climate change effects among different soil types. Without including both ‘special’ (serpentine, limestone, gypsum, shale, sandstone, etc.) and ‘normal’ soil types, it is impossible to determine whether climate change affects these community types differently.

Another important recommendation is to measure species functional traits, which allow comparisons between communities that do not share many species (e.g. those on different soil substrates). Both Fernandez-Going, Anacker & Harrison (in press) and Damschen, Harrison & Grace (2010) were able to assess climate change responses across multiple soil substrates using species traits. We specifically recommend measuring key functional traits of plants related to climatic and soil tolerances and comparing the community-level means of these traits. Traits that are relatively easy to measure and have been shown to relate to climatic tolerances and responses to experimental climate manipulations include: specific leaf area (SLA) (Garnier et al. 2001; Cornelissen et al. 2003; Cornwell & Ackerly 2010; Anacker et al. 2011), leaf water content (Cornelissen et al. 2003), leaf toughness (Cornelissen et al. 2003), foliar N and P concentrations (Cornelissen et al. 2003; Westoby & Wright 2006; Sandel et al. 2010), seed size (Cornelissen et al. 2003; Sandel et al. 2010), height (Cornelissen et al. 2003; Cornwell & Ackerly 2010) and phenology (Cornelissen et al. 2003). In addition to measuring traits related to stress tolerance, functional traits associated with dispersal ability should also be obtained to quantitatively assess the dispersal abilities of species on special vs. ‘normal’ substrates. Useful traits that are relatively simple to measure and are associated with dispersal ability include: dispersal mode (Cornelissen et al. 2003; Damschen et al. 2008), seed size (Cornelissen et al. 2003), seed terminal velocity (Soons et al. 2004) and seed release height (Soons et al. 2004; Thomson et al. 2011).

In addition to the factors above, future studies would benefit from conducting experiments across multiple spatial scales (e.g. from individuals to plots to regions). Such studies could not only reveal whether the observed community patterns on different soil types are consistent across spatial scales, but could also determine potential mechanisms underlying such patterns. For example, Fridley et al. (2011) found that heterogeneous microhabitats led to greater stability of communities at larger scales. Similarly, studies across varying temporal scales are also needed to determine whether the effects seen over a few years correspond to those over decades or across the paleoecological record. It is possible that while Briles et al. (2011) found less Quaternary change in forests on serpentine than non-serpentine soils, the answer over similarly long time-scales might be very different in the tropics (e.g. Stevenson & Hope 2005). However, we currently lack the comparative studies on serpentine and non-serpentine soils to address this.

Other valuable areas for future work include studies of the degree of evolutionary and ecological plasticity in the requirement of edaphic endemics for their special soils (i.e. whether they will become either more or less soil-restricted under altered future climates) and improved species distribution models that incorporate soils and other niche-determining factors in addition to climate. Finally, if dispersal distances are prohibitively long, and if ‘stress tolerant’ traits are insufficient to confer resistance to climate change, the preservation of narrowly distributed edaphic endemic species under climate change may possibly require conservation strategies such as managed relocation (McLachlan, Hellmann & Schwartz 2007). Studies should be undertaken to evaluate the effectiveness, risks and efficacy of such actions, including studies of impacts on population genetic structure, local adaptation and the effects on surrounding communities.


We thank Anu Eskelinen, James Grace and Scott Loarie for invaluable feedback and discussion and Nate Roth and Will Cornwell for assistance with analyses. National Science Foundation (DEB-0947432, DEB-0947368, DEB-0542451) provided funding for this work (http://www.nsf.gov).