Introduction

Biological communities change over time due to factors such as disturbance (Pickett et al. 2008; Platt and Connell 2003), changes in climate (Tylianakis et al. 2008), and species invasion (Mack et al. 2000). Understanding background rates of change is necessary when evaluating human impacts (Magurran et al. 2010). Arid lands are important ecosystems because they cover ~ 40% of Earth’s land surface and are rapidly changing due to biophysical and social factors such as climate, cropland expansion, and overgrazing (Geist and Lambin 2004).

Biological soil crusts are communities of photosynthetic organisms that live on the soil surface of arid lands. Crusts perform many roles in the ecosystem; the presence of mature crust is associated with reduced wind and water erosion and increased carbon and nitrogen fixation (Belnap and Lange 2001). The structure and composition of crust communities are related to precipitation levels, topographic position, soil type, and disturbance history (Belnap and Lange 2001; Ponzetti et al. 2007).

Changes in crust composition over time depend on site and vegetation characteristics, disturbance history, and characteristics of the crust taxa (Rosentreter et al. 2001). Sites with higher resource availability, such as moist microsites, can support more species (Garcia-Pichel and Belnap 2001) and can have higher productivity and more stability (Hooper et al. 2005). Invasive species such as the annual grass Bromus tectorum can decrease crust cover and species richness (Belnap et al. 2006). This grass may prevent crust from recovering (Dettweiler-Robinson et al. 2013) by competing for resources and/or reducing the fire return interval (Brooks et al. 2004). Disturbances such as fire, intense grazing, or drought can directly remove or damage crust. Disturbances can also indirectly affect crust by triggering abiotic changes, such as loss of soil stability (Bowker et al. 2004) or invasion by non-native plants (Chambers et al. 2007). The intensity, timing, recurrence, and type of disturbance can affect the rate of crust recovery even up to 30 years post disturbance (Bowker et al. 2004; Ford and Johnson 2006; Lalley and Viles 2008; Marble and Harper 1989; Zaady et al. 2007). For example, the first fire that burns through intact shrub cover following a long fire-free period would be of high intensity and therefore should have larger effects on the crust than subsequent fires that are less intense because of the reduced fuel loads. Crust taxa differ in susceptibility to damage and removal by different disturbances (Ponzetti et al. 2007). For example, organisms that live under plant canopies would be less likely to survive a high-severity fire than organisms that live in plant interspaces (Hilty et al. 2004). Succession may proceed differently based on timing and severity of disturbance, what portion of the community survived the disturbance, and the spatial arrangement of disturbed and undisturbed patches (White and Jentsch 2001).

Because some crust taxa are small, cryptic, and have slow growth rates, community change can be difficult to detect within short time scales. These changes reflect differences among taxa in growth rate and in response to environmental factors. Cyanobacteria and other single-celled components can colonize a site and expand rapidly in cover (Belnap 1993). In a scraping disturbance experiment, cyanobacterial cover quickly recovered to the level of the control treatment within 8 months and did not differ up to 32 months after treatment (Dojani et al. 2011). Bryophyte cover can increase rapidly in the absence of disturbance (Antos et al. 1983; Bates et al. 2009), and young mosses can grow more rapidly than older mosses (Barker et al. 2005). Lichens generally have slow growth rates that may decline with age (Armstrong and Bradwell 2010). Crustose lichens can invade soon after disturbance (Johansen et al. 1984) but may take a long time to increase in cover. Foliose and fruticose lichens may colonize later in succession because they require stable substrates (Belnap and Eldridge 2001). Dark cyanobacterial, moss, and lichen crusts responded more slowly and variably than light cyanobacteria in the scraping experiment (Dojani et al. 2011), suggesting that longer time scales are needed to identify cover and composition changes in lichen- and bryophyte-dominated crust communities following disturbance.

Few studies have examined cover and composition dynamics within crusts (e.g., Anderson et al. 1982; Hilty et al. 2004; Kaltenecker et al. 1999). In addition, these studies used a space-for-time approach, tended to focus on total crust cover rather than individual taxa, and occurred in sites dominated by cyanobacterial crust. In this study, we monitored permanent plots (Ponzetti et al. 2007) to document changes in a bryophyte- and lichen-dominated crust community 11 years apart. We assessed changes in crust taxa by comparing cover, frequency of occurrence, and indicator species between years. We quantified changes in community composition, total cover, covers of bryophytes and of lichens, and taxon richness, as well as proportional taxon loss, gain, and turnover. We examined the relationship between these variables and potential abiotic and biotic explanatory factors including exclosure from large herbivores, elevation, heat load index (HLI), time since fire (TSF), initial presence of B. tectorum, and change in B. tectorum cover. We expected (1) more change in cover and composition in recently burned quadrats because communities would be reassembling, (2) less change at higher elevation and cooler HLI as these areas have more moisture available to support high cover and diverse crust communities, and (3) more change in quadrats that initially contained B. tectorum or that had a large increase in B. tectorum cover.

Methods

Site description

The Horse Heaven Hills (HHH) complex (46.2°N, 119.5°W) in south-central Washington is managed by the Bureau of Land Management. The climate is semiarid, with an average of 188 mm of precipitation per year, mostly falling between October and April. Heavy fog is common in the winter; we have noted that crusts can be active (green, indicating photosynthetic activity) on these foggy days. Mean daily temperatures are 0°C in January and 22°C in July (Western Region Climate Center 2012). Soil types include Ritzville silt loams and Kiona very stony silt loams (Soil Survey Staff and Natural Resources Conservation Service 2011). Plots are located on elevations from 250–550 m with generally northerly aspects.

The dominant disturbances in the area are fire and grazing. Fires occurred in the area in 1986, 1998, 2002, and 2007; fire history before 1980 is uncertain due to lack of spatially explicit data. Much of the area was grazed from the 1960s to the 1980s. There have been no active grazing leases since 1986, though there are scattered reports of trespass by cattle from nearby pastures.

Vegetation types are dominated by sagebrush (Artemisia tridentata ssp. wyomingensis, A. tripartita), bunchgrasses (Pseudoroegneria spicata, Hesperostipa comata, Poa secunda), or invasive annuals (B. tectorum) (NRCS and USDA 2012). Artemisia spp. are characteristic of mature communities, but A. tridentata is removed by fire and does not recover quickly due to the low dispersal distance and short longevity of its seed (Young and Evans 1989). Bromus tectorum is able to colonize and expand in disturbed areas (Davies et al. 2012).

Previously, Ponzetti et al. (2007) surveyed 350 quadrats in this region and showed that crust diversity was highest on warm slopes and draws, crust richness and cover were inversely related to Bromus tectorum, and taxa differed in their tolerance of disturbance. The most abundant taxa were short mosses, Syntrichia ruralis, black cyanolichens, Trapeliopsis bisorediata, Diploschistes muscorum, and Cladonia spp. (Ponzetti et al. 2007).

Design

For this remeasurement, we focused on 48 permanently marked quadrats from Ponzetti et al. (2007), which are distributed across 12 plots spanning ~2,000 ha. A livestock exclosure (4.88 × 9.75 m) was erected on each plot in 1998. Two quadrats were located in diagonal corners of the exclosure, and two quadrats were diagonally adjacent to one another 50–100 m outside of the exclosure. Each quadrat was 0.5 × 4 m. Quadrats were monitored between October 1998 and May 1999 (hereafter “1999”) and again between November 2010 and March 2011 (hereafter “2010”). Live crust taxa were identified (Esslinger 2011; Flowers 1973; McCune and Rosentreter 2007) and cover recorded using five classes: 0, trace to 1%, >1–10%, >10–50%, and >50–100%. Taxa that were difficult to identify in the field or that had questionable identity across years were aggregated by genera or morphological group (see Additional file 1). Aggregating information to larger groups loses information about individual species but is a robust method for surveying small, cryptic organisms such as crusts (Eldridge and Rosentreter 1999) and preserves important trends in composition (McCune et al. 1997). This also ensured that we were conservative in assessing composition changes that may have occurred between the two monitoring periods. Bromus tectorum cover class was recorded on each quadrat.

Elevation, aspect, and slope were calculated from 10 m digital elevation models (DEMs; U.S. Geological Survey USGS 2011) to the scale of exclosures within plots. Heat load index (HLI) was calculated from latitude, aspect, and slope as per McCune and Keon (2002). Time since most recent fire (TSF) was calculated relative to 2010. Five plots burned prior to 1980 (TSF assigned a value of 30), two burned in 1986, two burned in 1998, two burned in 2002, and one burned in both 2002 and 2007.

Statistical analysis

We averaged each taxon’s cover across the 48 quadrats and tallied the number of quadrats on which each taxon occurred. We used these data to evaluate differences in cover and frequency between 1999 and 2010 for each taxon. We used Indicator Species Analysis to identify BSC taxon groups that were strongly associated with each sampling year. The Indicator Value (IV) is calculated independently for each taxon group i (1999 vs. 1010) (Bakker as the product of its relative abundance and relative frequency in each year j2008). Indicator values range from 1 to 100. Significance was assessed by permuting group identities 99 times and recalculating IV ij . IV s were calculated using the function given in Appendix S1 of Bakker (2008). Significant indicators were those with P < 0.05 and IV ij > 25 (Dufrêne and Legendre 1997). Significant indicators were taxa that changed in abundance from 1999 to 2010.

Community change was assessed in several ways. Cover classes were converted to the mean percent cover in the class as follows: 0% = 0%, trace to 1 = 1%, >1–10% = 5%, >10–50% = 30%, and >50–100% = 75%. We calculated the Bray-Curtis dissimilarity between each quadrat in 1999 and in 2010. The covers of all bryophytes and lichens were summed for each quadrat in each year. We calculated the change in total, bryophyte, and lichen cover by subtracting the value in 1999 from the value in 2010.

Community change is also a result of taxa gains and losses. We calculated the change in taxon richness as S2010 – S1999, where S1999 and S2010 are the taxon richness in 1999 and 2010, respectively. We also calculated the proportional taxa gain (Gp), proportional loss (Lp), and total proportional turnover (Tp) for each quadrat. All rates were expressed as proportions of the average number of taxa present during the measurement period: Gp = G/(0.5[S1999 + S2010] ), Lp = L/(0.5[S1999 + S2010]), Tp = (G + L)/(S1999 + S2010), where G is the number of taxa absent in 1999 but present in 2010, and L is the number of taxa present in 1999 but absent in 2010 (Anderson 2007).

The various change metrics were response variables in separate linear mixed models. Plot was included as a random block effect in all analyses to account for variability across the landscape. Exclosure was a fixed treatment nested within plots. Quadrats were nested within exclosure × plot combinations to account for microsite variability. Potential predictors included three discrete factors. Exclosure had two levels, “in” and “out.” Time since fire (TSF) was binned into two classes: quadrats that burned between the 1999 and 2010 measurements (“recent”), and quadrats that burned prior to the 1999 measurement (“late”). Although B. tectorum cover was recorded by cover class, we converted these data to presence/absence because they were strongly right-skewed, with over half of all quadrats containing no B. tectorum in 1999. We also included three continuous variables: elevation, HLI, and change in B. tectorum cover from 1999 to 2010.

All analyses were run in R (version 2.14.0; R Core Team 2012) using the lme function in the package “nlme” using backwards stepwise regression. Significance was assessed by comparing t-values to a random normal distribution. Delta AIC values of the final model were compared to those of a model with only the plot random block effect. Delta AIC values >2 indicated that the models had some support, and values > 10 indicated that the models had considerable support.

Results

We recorded 34 lichen taxa and 3 bryophyte taxa, for a total of 37 taxa (Table 1). Black crusts, Cladonia squamules, Trapeliopsis spp., and Diploschistes muscorum were dominant and common lichen groups in both years. Trapeliopsis steppica, white crust, and Leptochidium albociliatum were significant lichen indicators in 1999, while Aspicilia spp., cf. Trapeliopsis bisorediata, and cf. Endocarpon spp. were indicators in 2010. Although two bryophyte taxa, Syntrichia spp. and small moss, consistently had higher cover than other taxa, they were significant indicators in 1999 because they declined in cover by 2010.

Table 1 Biological soil crust taxa (see Additional file 1 ), functional type (McCune and Rosentreter 2007 ), percent cover (mean ± SD), frequency of occurrence (out of 48 quadrats), Indicator Value (IV), and P -value in 1999 and 2010
Full size table

Bray-Curtis dissimilarity was significantly different than 0 (mean = 22.4), indicating composition changed significantly from 1999 to 2010. However, none of the potential predictors were significantly related to Bray-Curtis dissimilarity (Table 2).

Table 2 Significant predictors included in the final models for Bray-Curtis dissimilarity; change (Δ) in total covers of BSC, lichens, and bryophytes; change in species richness; and species gain, loss, and turnover between 1999 and 2010
Full size table

Total crust cover and bryophyte cover declined significantly outside the exclosures (Figure 1, Table 2). Lichen cover did not change significantly during this period, nor did total crust and bryophyte cover change significantly inside exclosures.

Figure 1
figure 1

Changes in (A) total crust cover and (B) bryophyte cover between 1999 and 2010 differed significantly between exclosure treatments ( t = −3.0, p = 0.006; t = − 2.8, p = 0.007) whereas (C) lichen cover did not. Each box denotes the first and third quartiles of the data with the median as a thick line near the center and whiskers extending to 1.5 interquartile distances. Observations beyond 1.5 interquartile distances from the median are shown as individual points. The gray horizontal line indicates zero change.

Full size image

Change in taxa richness was significantly positively related to presence of B. tectorum in 1999 (Figure 2, Table 2). Plots with no B. tectorum in 1999 did not significantly change in taxa richness.

Figure 2
figure 2

Change in species richness between 1999 and 2010 was significantly related to the presence of Bromus tectorum in 1999 ( t = 2.2, p = 0.03). Each box denotes the first and third quartiles of the data with the median as a thick line near the center and whiskers extending to 1.5 interquartile distances. Observations beyond 1.5 interquartile distances from the median are shown as individual points. The gray horizontal line indicates zero change.

Full size image

There was significant proportional loss, gain, and turnover of taxa. Quadrats that burned between the two monitoring dates lost significantly more taxa and had higher proportional turnover than those that did not burn (Figure 3, Table 2). Quadrats inside exclosures gained significantly more taxa and had higher proportional turnover than those outside exclosures (Table 2, Figure 3C). Change in B. tectorum cover was significantly positively related to proportional loss and negatively related to proportional gain.

Figure 3
figure 3

Proportional loss, gain, and turnover of taxa between 1999 and 2010. Proportional loss was significantly affected by (A) time since fire (TSF) (t = 6.6, P < 0.001) and by (B) change in Bromus tectorum cover between 1999 and 2010 (t = 3.6, P < 0.001). Proportional gain was significantly affected by (C) change in B. tectorum cover between 1999 and 2010 (t = 3.2, P = 0.002) and by (D) exclosure treatment (t = − 2.5, P = 0.016). Proportional turnover was affected by (E) time since fire (t = − 6.1, P < 0.001) and by (F) exclosure treatment (t = − 3.0, P = 0.007). Relationships are not shown here if they were not statistically significant; all relationships are shown in Additional file 2. Each box denotes the first and third quartiles of the data with the median as a thick line near the center and whiskers extending to 1.5 interquartile distances. Observations beyond 1.5 interquartile distances from the median are shown as individual points. The gray horizontal line indicates zero change.

Full size image

Elevation and HLI had no significant effect on any of the change metrics.

Discussion

This is one of the first studies using repeated sampling of permanent plots to assess change in lichen- and bryophyte-dominated crust communities and provides a baseline for assessing change in similar crust communities. The significant changes in indicator species, crust composition, and proportional turnover metrics support the finding that crust communities are more dynamic over short time scales than previously thought (Belnap et al. 2006). Even quadrats that did not burn within 11 years and that were protected inside exclosures could show high rates of change (Figures 1,2,3, Table 2). Changes in cover and taxa richness metrics responded to different abiotic and biotic factors, with cover responding more strongly to exclosure treatment, and taxa richness and turnover responding to disturbance and annual grass invasion. Changes in cover and taxon richness were less variable than proportional turnover metrics, indicating that dominant and common species have strong effects on total production and do not change in cover or distribution as much as minor and uncommon species. Repeated, long-term sampling is desirable because one can detect the changes in rates of expansion and turnover (Magurran et al. 2010), but with two time points, we detected overall changes in the community in that interval.

Though there were differences between years in each taxon’s cover and distribution, the landscape generally contained high-quality, diverse crust communities. We found fruticose-foliose morphological group lichens such as Cladonia spp., Trapeliopsis spp., Diploschistes muscorum, and Aspicilia spp. in most quadrats in both years (Table 1). These groups tend to be found in undisturbed areas (Hilty et al. 2004; Ponzetti et al. 2007; Dettweiler-Robinson et al. 2013). Lichens such as Rhizocarpon diploschistidina, which parasitizes Diploschistes muscorum, indicate even longer undisturbed time (Ponzetti, personal observation); this species occurred in at least a quarter of our quadrats (Table 1). In addition, Texosporium sancti-jacobi, a Washington state threatened species (Washington Natural Heritage Program 2010) was found on two quadrats in 2010 (Table 1). Either this species dispersed to and established on these quadrats between the two measurements, or it was present but not identifiable on these quadrats in 1999 and subsequently developed identifiable traits. Together, these taxa indicate that these quadrats currently have a relatively high-quality crust.

The changes documented here provide a baseline from which to assess future changes due to disturbance or invasion. However, with high-quality crust cover across this study site, recovery from small-scale disturbances such as individual fires, presence of invasive species, or mechanical disruption from mammals may have proceeded more rapidly than if a larger area was severely degraded; recovery rates of soil lichens are positively related to nearby undisturbed lichen cover (Lalley and Viles 2008).

Fragmenting, wind-dispersed taxa such as mosses and Caloplaca spp. were common on the landscape. Because these taxa are associated with disturbed sites (Kleiner 1983, Lalley and Viles 2008, Dettweiler-Robinson et al. 2013), their presence suggests that a mosaic of disturbance histories, including fire, burrowing animal activity, and trampling, maintains a species pool of both early- and late-successional species in the region.

Bryophyte cover and total crust cover declined outside the exclosures with no change inside, and proportional gain and turnover were higher inside exclosures. Similarly, Anderson et al. (1982) found an increase in crust from <5% cover to >15% cover with 14–18 years of protection from cattle grazing and found that moss had more dramatic increases than lichens. However, although the experimental exclosures were built in 1998 to test the effects of cattle grazing on crusts, no livestock grazing occurred between sampling dates. This suggests that the effects of exclosures on crusts are due to other factors, such as their effect on access or behavior of other animals. For example, in one exclosure, we found extensive small mammal burrows as well as signs of a coyote attempting to dig under the fence. Deer were also present in the HHH. Areas inside exclosures would have been protected from trampling.

Significant exclosure treatment effects may also indicate spatial autocorrelation. We did not account for smaller-scale variation in soil characteristics and topography, such as texture, pH, calcium carbonate concentration, or water availability, which are strongly related to crust composition (Eldridge and Tozer 1997). These characteristics may have been similar between adjacent quadrats in one exclosure treatment but may differ from others.

Although recent fire had no effect on crust cover nor change in taxa richness, it resulted in higher proportional loss and turnover. Previous work showed that the number of moss and lichen species was lower in burned plots but that there was no difference in total moss or total lichen cover (Bowker et al. 2004). Presumably, species that are more sensitive to fire are removed while surviving species can expand in cover, leading to overall similar cover.

The metrics of change in taxa richness and proportional turnover metrics differ in that proportional turnover uses an average of taxa richness between the two years, so if there was low species richness in both years and one species was lost, it would be a high percent turnover but the absolute change in species richness would not be dramatically different.

Contrary to our initial hypothesis that B. tectorum would decrease cover and species richness, sites with B. tectorum present showed an increase in species richness. Because this site had low B. tectorum cover (average 6 ± 12%), the presence of B. tectorum cover may have indicated a previous disturbance that affected both B. tectorum and crust (see Additional file 2). For example, following disturbance such as fire or burrowing mammal disturbance, there would be an initial increase in crust taxa richness as they colonize, and B. tectorum could also invade in the disturbed site. This is further reflected by the significant effect of change in B. tectorum on proportional gain and loss (Table 2).

Based on a larger survey of quadrats throughout the region, Ponzetti et al. (2007) determined that crust composition was related to topographic position. We found no significant relationships between elevation or HLI and composition change. We sampled a much smaller range of topographic positions (see Additional file 3), and the 10 m DEMs did not account for microtopography in the quadrats.

Crust dynamics can be related to climate because Belnap et al. (2006) found that the aboveground biomass of crust organisms responded to seasonal differences in temperature and moisture. There was little seasonal difference in composition detected at these sites (Ponzetti et al. 2007). However, the trend of decreased bryophyte cover from 1999 to 2010 (Table 2) may relate in part to variation in annual precipitation, because annual precipitation in the 2 years prior to the 1999 sampling period was well above average (> 225 mm per year) but was below average (< 180 mm per year) in the 2 years prior to the 2010 sampling (Western Region Climate Center 2012). However, we had limited spatial and temporal extent and did not test this.

Changes in crust composition may affect ecosystem functioning by, for example, reducing soil stability (Lalley and Viles 2008), nitrogenase activity (Yeager et al. 2004), and chlorophyll content (Belnap 1993; Rychert 2002). The full recovery of visual cover of cyanobacteria did not correspond with recovery of chlorophyll content, lichen species number, and bryophyte and lichen cover up to 5 years following disturbance in Utah (Belnap 1993), indicating that ecosystem structure and function do not recover as rapidly as visual cover.

Conclusions

With increased anthropogenic disturbances, the diversity and function of crust communities may be at risk. Repeated disturbances may degrade crust communities so that late-successional species are lost from the community or cover declines. We have shown that crust communities are dynamic through time in terms of composition and proportional turnover, and disturbance and invasive species have major effects on crust composition.