Can biocrust moss hide from climate change? Fine‐scale habitat sheltering improves summer stress resistance in Syntrichia caninervis

Abstract

Premise

Mosses provide many ecosystem functions and are the most vulnerable of biocrust organisms to climate change due to their sensitive water relations stressed by summer aridity. Given their small size, moss stress resistance may be more dependent on fine‐scale habitat than macroclimate, but the sheltering role of habitat (i.e., habitat buffering) has never been compared to macroclimate and may have important implications for predicting critical biocrust moss refugia in changing climates.

Methods

We located three populations of a keystone biocrust moss, Syntrichia caninervis, spanning 1200 m of altitude, which comprised three macroclimate (elevation) zones of characterized plant communities in the Mojave Desert. We stratified 92 microsites along three aridity gradients: elevation zone, topography (aspect), and microhabitat (shrub proximity). We estimated summer photosynthetic stress (F _v/F _m) and aridity exposure (macroclimate, irradiance, and shade).

Results

Microsite aridity exposure varied greatly, revealing exposed and buffered microhabitats at all three elevation zones. Moss stress did not differ by elevation zone despite the extensive macroclimate gradient, failing to support the high‐elevation refugia hypothesis. Instead, stress was lowest on northerly‐facing slopes and in microhabitats with greater shrub shading, while the importance of (and interactions between) topography, irradiance, and shade varied by elevation zone.

Conclusions

Fine‐scale habitat structure appears physiologically more protective than high‐elevation macroclimate and may protect some biocrust mosses from the brunt of climate change in widespread microrefugia throughout their current ranges. Our findings support a scale‐focused vulnerability paradigm: microrefugia may be more important than macrorefugia for bolstering biocrust moss resistance to summer climate stress.

Biocrusts (i.e., biological soil crusts) are diverse networks of cyanobacteria, fungi, algae, lichens, mosses, and liverworts that weave together the upper few centimeters of topsoil in arid and semiarid environments globally (Belnap, 2003). Where sufficient moisture or shade exist, mosses are often the larger members of these soil communities and drive many biocrust ecosystem services (Bowker et al., 2013) mediated by their large (typically 1–10 cm²), absorbent colonies or “patches” that can increase water infiltration and retention (Lafuente et al., 2018), buffer temperature (Xiao et al., 2015), increase soil fertility (Belnap, 2003), shelter microbiota (Abed et al., 2019; Fisher et al., 2020), store carbon (Elbert et al., 2012), and prevent erosion (Stovall et al., 2022). However, biocrust mosses are predicted to have more dramatic responses to climate change than most other poikilohydric biocrust species because of their requirements for higher shade and moisture during hydration periods (i.e., “hydroperiods”) when the plants are metabolically active (He et al., 2016; Seppelt et al., 2016; Rodriguez‐Caballero et al., 2018; Ladrón de Guevara and Maestre, 2022). Such sensitivity in biocrust mosses has been documented in multiple field and laboratory experiments simulating extreme climate stress (e.g., rapid drying events), which has caused stress, tissue damage, or lethality (e.g., Brinda et al., 2011; Stark et al., 2011; Reed et al., 2012; Zhang et al., 2016; Greenwood et al., 2019; Coe et al., 2020). Moreover, ecophysiological models predict that climatic change may greatly reduce biocrust moss biomass via reduced productivity and increased mortality (Coe and Sparks, 2014). Collectively, such impacts on biocrust moss resilience and productivity may reduce their diversity and ecosystem services in changing climates (Ladrón De Guevara and Maestre, 2022).

High‐elevation habitat such as mountains and plateaus offer cooler temperatures and higher humidity than lower elevations in drylands and thus are predicted to offer an oasis to many plants as climate warms and becomes drier (Kelly and Goulden, 2008). Such macroclimate gradients along broad altitudinal ranges drive zonation in the dominant vegetation community (elevation zones). High‐elevation communities are already documented to support greater moss abundance and diversity in several drylands with extensive aridity–elevation gradients exceeding hundreds of meters (e.g., Nash et al., 1977; Seppelt et al., 2016; Clark, 2020). Such gradients in elevation may therefore provide high‐elevation refugia for dryland mosses, assuming species have time and sufficient propagules to migrate upward (Zanatta et al., 2020).

An alternative or co‐occurring possibility is that mosses could “hide” from climate change in sheltered microrefugia that offer optimal microhabitat within a species' present range (Ashcroft, 2010). The microscale of dryland moss colonies or “patches” is often less than 5 cm² (Nash et al., 1977; Clark, 2020) such that fine‐scale habitat structure within sites may be equally or more important to moss resiliency than macroclimate gradients. Fine‐scale habitat such as shade vegetation and pole‐facing slopes and rock surfaces have been shown to reduce local extremes in temperature, incident radiation, and evaporative demand adjacent to dryland mosses (Alpert, 1985, 1988; He et al., 2016; Li et al., 2018; Ladrón de Guevara and Maestre, 2022). The habitat buffering hypothesis predicts that such sheltered microhabitats will “buffer”, or reduce, climate stress for many organisms (Williams et al., 2008; Scheffers et al., 2014; Shi et al., 2016). Moreover, mosses are small enough to be sheltered by multiple scales of habitat in additive or interactive ways, which we will call “multiscale habitat buffering”. In drylands, there is a prevalence of topographically complex, multiscaled habitat for biocrust, for example, a habitat that is shaded by fine‐scale vegetation on a north‐facing slope within a drainage basin (e.g., Williams et al., 2013; Pietrasiak et al., 2014; Bowker et al., 2016). There is potential that such scales of habitat structure may interact to create strongly sheltered biocrust habitats in future climates.

Predicting the most important scales of habitat buffering for biocrust mosses in changing climates can begin with studying present‐day stress and mortality patterns across scales of habitat structure relevant to these mosses. Scales known to influence biocrust moss distributions include macro‐, meso‐, and microscale environmental features (Bowker et al., 2016). The mesoscale includes within‐site topographical variation on the scale of meters where changes in aspect, slope, and hydrological position can alter soil‐water dynamics important to biocrust water availability, such as location relative to a drainage or soil evaporation rate. Microscale variation can include microtopographical shading driven by proximity and azimuth to surrounding vegetation and rocks. Such studies will elucidate the types of refugia, macro‐ or microrefugia (Ashcroft, 2010), that are more probable for small photosynthetic organisms. Measuring climate stress estimates using gas exchange or chlorophyll fluorescence will also provide spatially explicit data that can be used to strengthen predictions for future moss abundance distributions in drylands (Coe and Sparks, 2014).

We sought to test the high‐elevation refugia and habitat buffering hypotheses using the biocrust moss, Syntrichia caninervis Mitten, a member of the acrocarpous moss family, Pottiaceae. This keystone biocrust species is understood to be one of the most broadly distributed and ecologically important biocrust mosses in western North America, North Africa, and Asia (e.g., Ros et al., 1999; Coe et al., 2012a; Seppelt et al., 2016; Zhang and Zhang, 2019). With a broad geographic and altitudinal distribution in the American Southwest (FNA, 2007+), S. caninervis provides an ideal model species to study summer stress resistance along multiscaled gradients in habitat structure. The Mojave Desert is the most climatically extreme part of the species' North American range where resident populations likely exist at or near physiological thresholds of precipitation minima and temperature maxima (e.g., Stark et al., 2009). Summer presents the highest risk for moss mortality when combinations of extreme desiccation (i.e., cellular water potentials < –400 mPa) interrupted by small rain events have been shown to prevent S. caninervis from achieving positive carbon balance during summer hydroperiods (Coe et al., 2012b). Such carbon deficits can lead to full‐patch mortality and suggest S. caninervis may be increasingly threatened by continued climate change. Current climate trends and predictions in this most arid North American desert include smaller summer rain events and increased drought intensity and variability (Seager et al., 2007; Seager and Vecchi, 2010; Zhang et al., 2021).

Our specific objectives in the Desert National Wildlife Refuge (DNWR) along a ~1200‐m elevation gradient in Nevada were to (1) locate S. caninervis microsites spanning three nested scales of habitat structure by surveying three elevation zones (macroscale sites), three topography zones per site (mesoscale zones), and three microhabitat types per zone (microscale microsites), (2) estimate moss habitat buffering at each microsite (shade time and potential insolation), (3) use in situ chlorophyll fluorescence to estimate end‐of‐summer moss photosynthetic stress and mortality, and (4) elucidate the most important scale of habitat structure by testing for stress patterns by habitat scale and modeling the relationship between habitat buffering proxies (elevation, potential insolation, and shade) and summer stress in S. caninervis. This natural experiment allowed us to test relationships between three scales of habitat structure and habitat buffering with summer stress resistance in three elevation‐zone populations of a biocrust moss to gather evidence to either support or reject the high‐elevation and habitat buffering hypotheses in one of the harshest environments for which this species occurs globally, the Mojave Desert. These results have important implications for predicting and effectively monitoring the response of this and similar biocrust species to future climate change in drylands by using scale‐appropriate and spatially explicit ecological models (Gignac, 2001; Dunning, 1995).

MATERIALS AND METHODS

Elevation zone sites

The Desert National Wildlife Refuge (DNWR) is a large (6430 km²) topographically and biologically diverse basin and range landscape in the eastern Mojave Desert of southern Nevada (Las Vegas, Nevada, USA). The DNWR is home to the Sheep Range EPSCoR‐NevCAN transect (Established Program to Stimulate Competitive Research, Nevada Ecohydrological Climate Assessment Network; Mensing et al., 2013), a set of five climate stations spanning 2000 m of elevation and located in each of the five Mojave elevation zones (Figure 1A), all of which have been floristically characterized without mention of bryophytes (Ackerman, 2003; NCCP, 2018). Regional soils are limestone‐derived, highly calcareous, with an organic content ranging from low to relatively high organic content at high elevations.

Figure 1.

(A) Elevations (m a.s.l.) surveyed for Syntrichia caninervis in the Sheep Range of the Mojave Desert National Wildlife Refuge (DNWR). EPSCoR‐NevCAN climate towers are shown at each elevation zone site (see section, “Elevation zone sites”). Final sample of 92 microsites (20 × 20 cm quadrats enumerated in (C) spanned three aridity exposure gradients: (B) elevation zones, (C) topography zones (northerly, southerly, or flat transects), and (D) microhabitat shade types. For each topography zone in (C), the slope and microhabitat frequency are shown, and missing squares indicate S. caninervis was not found for a given zone or microhabitat type. See Figure 3 for example quadrats.

The lowest‐elevation desert scrubland site (890 m a.s.l., 36.435345 N, 115.355850 W; hereafter, low scrubland) and surrounding landscape are characterized by an open salt basin interrupted by shallow, calcareous drainages with gentle slopes and occasional steep ravines 1–2 m deep. Excluding drainages, the soil is covered almost entirely by desert pavement (e.g., Pietrasiak et al. 2014) with well‐spaced shrubs >2 m apart (Figure 2A). The blackbrush–Joshua tree (Coleogyne ramosissima and Yucca brevifolia) elevation zone site (1680 m a.sl., 36.51723 N, 115.16191 W; mid shrubland) is situated in the center of an intermountain basin divided by drainages from ~1–3 m deep. The ground is nearly covered by desert pavement, moderately spaced shrubs <2 m apart, and widely spaced succulents (Figure 2A). The pinyon–juniper woodland site (2065 m a.s.l., 36.572808 N, 115.204060 W; high woodland) is at the base of the Sheep Mountains on one of the deeply divided ridges with steep, rocky slopes (>3 m tall, ~10°–15°). The soil is nearly covered by loose gravel with a dense community of short and tall shrubs and dominant well‐spaced pygmy conifers (Pinus monophyla and Juniperus osteosperma; Figure 2A). The montane site (2320 m a.s.l., 36.590255 N, 115.214166 W) has an open canopy of Pinus ponderosa and well‐spaced shrubs >3 m apart. The highest‐elevation subalpine site (3015 m a.s.l., 36.657641 N, 115.200777 W) in the NevCAN transect has a nearly closed canopy of mixed‐conifer forest (Abies concolor, A. lasiocarpa, Picea englemannii, and Pinus longaeva) with calcareous, rocky, organic soils (Ackerman, 2003).

Figure 2.

Microsite annual shade time was quantified using seven photos taken on a smartphone using the Sun Seeker Solar AR Tracker app. The phone was held approximately 1 cm above the center of each microsite quadrat (center photo), and the photos encompass the solar arc from 06:00 to 08:00 hours (red circles), which collectively captures the annual solar window from summer solstice (red lines) to winter solstice (blue lines) for a single microsite. Photos are scored using a 5‐scale percent shade class by assessing the area of all shade objects intersecting each 2‐h solar window (white dashed boxes and triangles). Percent annual shade time is the sum of classes for all seven photos divided by 28, the maximum possible (i.e., for a habitat shaded 75–100% of the year). The microsite shown here is shaded 64% of the year ((4 + 4 + 1 + 2 + 1 + 2 + 4 = 18)/28 = 0.64 × 100). Note: The center photo in this figure is for demonstration only and was not located at the microsite of the illustrated solar window.

Climate metrics

To compare mean annual and summer climate where we located S. caninervis along the NevCAN‐DNWR elevation zone transect, we acquired NevCAN daily means for temperature, humidity, irradiance, and precipitation from 2011 to 2018 (DRI, 2020). We calculated the 7‐yr mean annual air and soil temperature, percent relative humidity (RH), wind speed, and soil moisture. We calculated summer 2017 climate means to include Mojave hot‐season months preceding our moss tissue collection, which took place at the beginning of November (1 June–11 November 2017; Table 1A).

Table 1.

Macroclimate measured by the NevCAN station near each elevation zone site (m a. s. l.) supporting S. caninervis in the Desert National Wildlife Refuge. (A) Annual (2011–2018) and Summer (1 June–6 November 2017) daily climate means ± (SD). The maximum elevation zone buffer (Mean buffer) is the maximum deviance between the three elevation‐zone means for the respective period. (B) Mean ± (SD) potential direct incident radiation (PDIR) for elevation zone sites and topography zones (see Figure 1). The greatest topography zone buffer (Mean buffer) is the greatest mean deviance of all pairwise comparisons between topography zones for each row. Topography zones varied in their sample sizes (see Figure 1C). SD = 0 occurred on two topography zones having nearly identical PDIRs across microsites.

		Elevation zone sites
		All sites	Low scrubland (creosote, 890 m)	Mid shrubland (blackbrush, 1670 m)	High woodland (pinyon‐juniper, 2070 m)
A) Climate metric	Season	Mean (SD)				Mean buffer
Air temperature (°C)	Annual	15.0 (9)	18.9 (9)	13.8 (8)	12.3 (8)	(−) 6.6
	Summer	22.6 (6)	26.9 (6)	21.3 (6)	19.5 (6)	(−) 7.4
Soil temperature (°C)	Annual	17.3 (11)	21.8 (11)	16.2 (10)	13.9 (10)	(−) 7.9
	Summer	26.0 (7)	31.1 (7)	24.9 (7)	22.1 (6)	(−) 9.0
Relative humidity (%)	Annual	32 (19)	29 (17)	34 (19)	35 (20)	(+) 6
	Summer	26 (13)	21 (11)	28 (14)	29 (15)	(+) 8
Wind speed (m/s)	Annual	3.6 (1)	4.2 (2)	3.8 (1)	2.8 (1)	(−) 1.4
	Summer	3.6 (1)	4.2 (1)	3.8 (1)	2.8 (1)	(−) 1.4
Soil moisture (θ)	Annual	6.1 (6)	2.9 (4)	5.8 (6)	14.5 (5)	(+) 11.6
Precipitation (mm)a	Annual	186 (52)	119 (41)	160 (54)	278 (85)	(+) 159
	Summer	92	26	81	169	(+) 143

B) Topography zone		Mean PDIR (kJ cm−2 yr–1)
Topography zones pooled		943	905	975	949	(−) 70
South‐facing zoneb		1022b	1061b	999 (8)	1044 (4)	(−) 45
Flat terrain zone		990	985 (0)	1000 (0)	986 (2)	(−) 15
North‐facing zone		856	826 (16)	926 (8)	817 (29)	(−) 109

Note: Climate metrics were measured at 2‐m height except for soil temperature and moisture, measured at 1 and 4 inches below the soil surface, respectively.

^a The standard deviation of total precipitation is annual variability; thus, summer 2017 total precipitation has no standard deviation.

^b The south‐facing slope at the low‐scrubland site was not included in mean calculations because no mosses were found there.

Aridity exposure survey: elevation zones, topography zones, and microhabitats

To determine the macroclimate exposure of S. caninervis in DNWR across the three habitat gradients, we surveyed for species occurrence within a 1‐km radius of each elevation‐zone climate tower (Figure 1A, B); when we found the species in an elevation zone, we surveyed three topographic exposure zones (40 × 10 m plots): the most northerly‐facing, southerly‐facing, and flat terrain nearest the tower. Selected topography zones varied in their hydrological positions, including association with drainages, uplands, or mountain ridges (Figure 1C). Within each topography zone, we surveyed for moss occurrence in three microhabitat shade classes (hereafter, microhabitat types): (1) high‐shade canopy microhabitats partially or fully under shrub canopies, (2) low‐shade interspace microhabitats ≥0.5 m from the outer edge of the shrub canopy (the canopy dripline), and (3) intermediate‐shade dripline microhabitats located within a 0.5‐m marginal band from the shrub dripline; Figure 1D). Any vegetation to the north (cardinal 315°–45°) of moss microsites was ignored for habitat type assignment because it did not contribute to moss shade.

Moss microsite selection

We systematically selected 12 microsites (20 × 20 cm quadrats) per topography zone, attempting to find four of each microhabitat type per topography zone, but because this was often not possible (see Figure 1C), the resulting proportion of microhabitats at each elevation zone is a coarse measure of habitat frequency (Appendix S1). We selected only undisturbed microhabitats having the shrub canopy intact (i.e., not dead or broken off). We systematically centered over the highest‐density patch of S. caninervis in each microhabitat (>3 cm² of S. caninervis cover), orienting the quadrat with sides parallel to a north–south axis (Figure 1D). Four shoots were sampled from each quadrat for microscopic species verification. The final sample included 92 microsites because two microsites at the high woodland were removed after lab culture revealed two Syntrichia species (caninervis and ruralis s.l.) were unknowingly intermixed and possibly measured in the stress assay (Clark, 2020; Figure 1A). Two additional microsites were not sampled for moss stress at the mid shrubland because of environmental sensor interference, thus N = 92.

Habitat buffering metrics

Sheltered (i.e., “buffered”) microhabitats for dryland mosses are generally thought to be those with higher humidity, lower temperature, and reduced insolation compared to neighboring local habitats (Alpert, 1988). Although such habitat buffering can be measured directly with microenvironmental sensors, large replication and sensor installation in delicate biocrust systems is highly invasive; therefore, we measured buffering proxies (e.g., slope, aspect, and shade) for the three scales of habitat structure (utilizing only NevCAN climate tower sensors at each of the NevCAN sites) as follows.

Macroscale climate buffering

Elevation zone buffering was estimated at the landscape scale by calculating the mean difference in macroclimate relative to the creosote (low) elevation zone where conditions are most stressful for mosses (i.e., most arid and hottest). For example, the elevation zone temperature buffer was calculated as the mean annual decrease in daily temperature between the creosote elevation zone and the higher elevation zones.

Mesoscale topographical buffering

We measured aspect (compass cardinal direction) and slope (via a hand‐held clinometer) of the 3 × 3 m² area surrounding each microsite for use in calculating potential direct incident radiation (PDIR). Within elevation zones, we calculated mesoscale topographical exposure as the mean of microsite PDIR in each topography zone. Potential direct incident radiation was estimated with a complex formula that incorporates microsite elevation, latitude, slope, and folded aspect (McCune and Keon, 2002; McCune, 2007). We observed S. caninervis ground cover to be greatest on the north side of shrubs in this ecosystem and thus modified the PDIR equation for use with a north–south rather than northeast–southwest axis as described by McCune (2007). To estimate PDIR buffering across and within elevation zones, we calculated differences in mean mesoscale topographical exposure. For example, the maximum topographical zone buffer was simply the difference in mean PDIR between topography zones with the highest and lowest mean PDIRs. Mean elevation zone PDIR was simply the average of all topography zone PDIRs at each elevation zone site.

Microscale shade buffering

To precisely measure fine‐scale shade time of microsites, we developed a photographic method using the smartphone app, Sun Seeker Solar Augmented Reality Tracker (OzPDA, Sydney, Australia), which maps onto the camera view the annual solar window from sunset to sunrise for a given location (e.g., moss microsite). Seven photos taken at any time of the year circumscribe all geographic–time‐referenced shade objects from the vantage point of the moss (Figure 2). For each photo, the area of shade objects (i.e., vegetation and topography) inside the solar window was scored using a shade class from 0 (no shade) to 4 (75–100% shaded; see classes in Figure 2). For each microsite, the resulting seven shade classes were divided by the total possible score of 28 shade points to yield an annual shade time percentage. Our novel microscale shade metric estimates the percentage of the year a microhabitat is shaded by vegetation and/or topography; we used this sensor‐free buffering metric to test for differences in mean shade at various scales of habitat structure.

Field sampling and tissue preparation

To measure the summer stress signal of S. caninervis, we collected shoots when patches were fully desiccated in late fall (6–13 November 2017; Permit #84555‐17‐019, U.S. Fish and Wildlife Service National Refuge System Research & Monitoring), which is still considered the Mojave warm season. This dormant, dry tissue preserved the summer/warm‐season stress signal because tissue was collected before any fall precipitation during which some recovery could begin. We collected two to three shoots for every 2‐cm² patch area per quadrat using fine forceps (Figure 3A, B). Collections were stored dark and dry at 20°C with 20–30% relative humidity, conditions that should not exacerbate stress in this species (Guo and Zhao, 2018). Before the stress assay, shoots were hydrated with distilled water on a microscope slide, and the upper 2 mm were cut and retained to target the living apical “green zone” (Stark, 2017) (Figure 3B). Green zones were swirled in a drop of water to remove debris and placed vertically into a rosette formation on a wetted 7‐mm filter paper disc creating a “moss bouquet” (Figure 3B). The number of shoots in a moss bouquet (5–23) depended on shoot size to maximize microsite representation and standardize leaf area for chlorophyll fluorescence.

Figure 3.

Methods for microhabitat selection at three elevation zone sites showing (A) moss abundance gradient, (B) field shoot sampling, (C) moss bouquet assembly, and (D) photosynthetic efficiency assay to estimate summer stress in a small desert moss.

Summer‐stress resistance assay (F _v/F _m)

Efficiency of photosystem II (PSII) in photosynthesis is commonly measured as a stress proxy in plants via the non‐invasive technique, chlorophyll fluorescence (Papadatos et al., 2017), convenient for measurements on small moss species with limited biomass or long‐term monitoring (Proctor, 2009). We refer to F_v/F_m as a stress proxy because in desiccation‐tolerant mosses, F_v/F_m responds sensitively and negatively in a near linear fashion to the level of desiccation and light stress induced in a species in carefully controlled experiments (e.g., Proctor, 2001; Proctor et al., 2007). In almost all cases, the lower the F_v/F_m when such a moss is rehydrated (30–60 min after desiccation), the more stressful were the conditions during the previous hydration/desiccation cycles (e.g., Coe et al., 2020). This reduced F_v/F_m can be viewed as a photoprotective acclimation to light and low water potential both of which stress the electron transport chain and induce a photoprotective response (Papadatos et al., 2017).

F_v/F_m was measured using a Hansatech FMS‐2 modulated fluorometer (Norfolk, England) and vascular plant leaf clips were hollowed to accommodate taller moss shoots (Figure 3B, C). Hydrated moss bouquets were each placed on a folded chemical wipe and carefully positioned into leaf clips, then placed in a tray of distilled water to maintain full turgor throughout the 20‐min dark acclimation (a period needed to close all active PSII reaction centers). Including 10 min to clean and prep bouquets, we measured chlorophyll fluorescence 30‐min post‐rehydration, a standard photosynthetic reactivation timepoint for assessing the photosynthetic stress or vitality of mosses before significant recovery from the preceding desiccation event (e.g., Hamerlynck et al., 2000; Munzi et al., 2019; Ekwealor et al., 2021).

A Hansatech script was used to automate our measurement of the dark‐acclimated metrics, basal (F _o) and maximum (F_m) fluorescence, during a 0.8‐s saturation pulse of 3000 µmol m^–2 s^–1. Variable fluorescence (F_v = F_m – F_o) was used to derive F_v/F_m, which estimates potential maximum quantum efficiency of photosystem II (PSII) photochemistry of dark‐acclimated plants (Baker, 2008). F_v/F_m provides a universal physiological indicator of climate stress in plants that is more sensitive and integrative than chlorophyll content (Murchie and Lawson, 2013) and requires less tissue and time than gas exchange analysis.

Analyses

Analyses were performed in R (Version 1.447 and 4.2.3; R Core Team, 2023) using the tidyverse package (Wickham et al., 2019) and other statistical packages denoted as “(package: function)”. We used an exploratory approach with hypothesis testing in which we considered the ecological significance of statistical patterns when P < 0.05 and opt out of arbitrary family corrections for testing our small set of environmental factors (Gotelli and Ellison, 2013).

Shade buffering and moss stress vs. multiscale habitat structure

Although our design involved three categorical habitat predictors of moss stress and shade buffering (elevation zone, topography zone, and microhabitat type), using three‐way ANOVAs to test patterns in mean shade buffering and moss stress was not possible due to missing factor levels (i.e. no south‐facing zone in the low scrubland and no interspace microhabitats in several topography zones (Figure 1C; Appendix S1). Alternatively, we conducted six univariate ANOVAs: three stress tests and three shade buffering tests, one for each habitat factor. We calculated the effect sizes, Eta‐squared (equivalent to R ² in one‐way ANOVA) and Cohen's f (Lakens, 2013). When heteroskedasticity was present (determined with the Fligner–Killeen test; Fligner and Killeen, 1976), we opted out of transformations, which are known to yield nonsensical predictions of proportion data and hinder interpretability (Warton and Hui, 2011). Instead, we used robust Welch's denominator degrees of freedom corrections appropriate for the approximate normality, unbalanced design, and heteroskedasticity present (stats: oneway.test; Welch, 1951). Post hoc tests for Welch's ANOVAs were nonparametric Games–Howell multiple comparisons (rstatix: games_howell_test; Ruxton and Beauchamp, 2008; Kassambara, 2023).

Habitat buffering proxies as predictors of moss stress

To explore the relationship between summer moss stress and the three scales of habitat buffering (elevation, PDIR, and shade), we fit a linear multiple regression model for F_v/F_m (Appendix S2). Despite small variance inflation factors of 2.07, 1.17, 2.19, for the three predictors, respectively, our model testing procedure revealed instability in parameter estimates and P‐values when elevation was included in the model (and no interaction terms were included). Therefore, the final reduced model included only PDIR, shade, and their interaction, centering the explanatory variables to remove structural multicollinearity. Diagnostic residual plots revealed adequate fit (Appendix S3); however, we acknowledge that beta regression is optimal for modeling such a two‐category ratio (Douma and Weedon, 2019) and should be used for prediction‐focused studies.

Elevation zone models

With elevation zone management in mind, we performed a set of hypothesis tests focused on each elevation zone site to explore whether our spatially efficient (i.e., nested) sampling design (Figure 1) of mesoscale topography zones (within‐site plots) and microscale shade measurements (within‐plot microsite quadrats) could explain significant patterns in moss stress. (Although ideal, testing these relationships in a single ANOVA model including all elevation zones was not possible due to the multicollinearity between elevation zone [elevation] and shade [as discussed in the habitat buffering model above].)

Alternatively, we ran three separate ANOVAs (i.e., ANCOVAs) to test additive and interactive relationships between topography zone and percent shade time (centered before analysis) with moss stress for each elevation zone. ANOVAs were made robust to (1) the unbalanced design (Figure 1C) using Type II sums of squares (Langsrud, 2003; Logan, 2010) and to (2) unequal variance using a Huber–white heteroskedasticity‐corrected covariance matrix (HCCM; White, 1980; Long and Ervin, 2000) for each model (car: Anova; Fox and Weisberg, 2019). Post hoc pairwise comparisons for regression slopes (by topography zone) were not conducted (when the interaction term was significant) because we believe sample sizes greater than 11–12 quadrats per topography zone should be collected to produce more accurate fine‐scale relationships. The OLS coefficient of determination (R ²) is shown for each final model as an effect size metric; the overall F‐test for each HCCM‐corrected model was derived from a Wald test comparing the intercept model to the respective HCCM model including only significant terms (lmtest: waldtest; Zeileis and Hothorn, 2002).

RESULTS

Results are numbered by objective, and statistics are included in figures if not shown here. We refer to our three scales of aridity gradient sampling (i.e., elevation zones, topography zones, and microhabitat types, Figure 1) as the three scales of habitat structure or habitat buffering, depending on context.

Aridity exposure across elevation zones (Objective 1)

After extensive surveying within a 1‐km radius of each NevCAN climate tower, we located S. caninervis biocrust in the lower three sites of the DNWR ecohydrological gradient spanning 1180 m of elevation, from 890 to 2070 m a.s.l. A closely related species, Syntrichia ruralis s.l., was found primarily at higher elevations (Figure 1A). We will refer to these sampling sites as the low scrubland, mid shrubland, and high woodland, respectively (Figure 1B). We located S. caninervis on north‐facing, south‐facing, and flat topography zones at the mid shrubland and high woodland, but it was absent from south‐facing slopes in the low scrubland (also absent from upland flats) and was only found on shallow drainage flats and north‐facing slopes (Figure 1C). In the low and mid elevation sites, S. caninervis occurred in all three microhabitat types, but in the high woodland, the shrub interspace habitat did not support a high density biocrust with S. caninervis. Relative frequency of microhabitat type changed with elevation: Shrub canopy habitat increased, interspace habitat decreased, and shrub‐dripline frequency remained similar from low to high elevation (Appendix S1; Figure 1C). Canopy microhabitats were the most frequent habitat type supporting high‐density S. caninervis biocrust across elevation zones (54/92 microsites; Appendix S1).

Habitat buffering across three scales of habitat structure (Objective 2)

Macroscale climate buffering

Mean daily macroclimate differed substantially by elevation zone (Table 1A) with higher elevations having cooler temperatures, higher humidity and precipitation, and slower wind. Relative to the low scrubland, the high‐woodland site was climatically buffered in four metrics: (1) 6.5 times more precipitation, (2) 7.4°C lower mean daily air, (3) 9.0°C lower soil temperature, and (4) 8% higher mean relative humidity (Table 1A).

Mesoscale topographical buffering

The nine topography zones collectively represent the mesoscale exposure gradient for S. caninervis‐dominant biocrust at each elevation‐zone site and varied in their hydrological position by elevation zone: the low‐scrubland was in an ephemeral drainage, the mid shrubland included a drainage and upland flat, and the high woodland traversed a mountain ridge (Figure 1C). Of the nine topography zones, there was a 22% reduction from the highest mean PDIR of the south‐facing slope of the high woodland (1044 kJ cm⁻² yr^–1) to the lowest mean PDIR of the north‐facing slope of the high woodland (817 kJ cm⁻² yr^–1), which created a maximum mean topography zone buffer of 227 kJ cm⁻² yr^–1 (Table 1B, Figure 1C). Notably, the low scrubland had the lowest average PDIR (of the three elevation zones) due to its relatively steep 16° north‐facing slope.

Microscale shade buffering

Percent annual shade time (“shade” hereafter) across the 92 microsites ranged from 21 to 96% and averaged 64 ± 2% (SE) such that most microsites were shaded over half of the year. Elevation zone significantly explained 48% of shade variation (Welch F _2,52 = 40.6, P < 0.0001); mean shade in the high woodland was 1.9× greater than in the low‐scrubland (Figure 4A). Mean shade differed little by topography zone when pooling microsites across elevation zones (Welch F _2,58 = 2.7, P = 0.074, R ² = 0.05, Figure 4B), while microhabitat type explained 67% of variation in shade, which increased 2.5‐fold from interspace to canopy habitats, on average (Welch F _2,44 = 44.0, P < 0.0001, Figure 4C).

Figure 4.

Boxplots of shade time (top row) and summer stress (bottom row) measured in 92 Syntrichia caninervis microhabitats pooled by scale of habitat structure: (A, D) elevation zone, (B, E) topography zone, and (C, F) microhabitat type. Overlays include raw data (black points) and respective means (white points) with SE bars. Welch's ANOVA results, R ², and Cohen's F (Eff) are shown for each panel. Letters indicate familywise Games–Howell significant differences between groups (P < 0.05) in post hoc testing. Denominator degrees of freedom in F‐statistics do not reflect sample size due to Welch's heterogeneity correction.

Summer stress proxy (F_v/F_m) by elevation zone (Objective 3)

Nearly all S. caninervis patches showed visible signs of stress in their darkened, sun‐pigmented leaves (Figure 5). The sample distribution of F_v/F_m was left‐skewed with a mean of 0.455 ± 0.148 (SE) and ranged from 0.130 at the mid shrubland to 0.737 at the mid shrubland (Figure 6). Of the 92 mosses measured, ~23% were near normal levels (F_v/F_m > 0.6) leaving 77% stressed (see Discussion). Dividing the typical physiological range of F_v/F_m (0–0.85) into four stress categories, we report 7% severely stressed, 25% moderately stressed, and 42% mildly stressed, and 26% unstressed (Figure 6). The seven samples with F_v/F_m < 0.2 (in the lower 25th percentile; Figure 6) were from the flat zone at each elevation zone. These outliers included two interspace, three dripline, and one canopy microhabitat and had moderate annual shading ranging from 43–68%, except for one interspace microsite at the low scrubland that was shaded only 28% of the year. Extreme stress in these samples was also visually evidenced by chlorotic leaf tips (Figure 5B, C).

Figure 5.

Syntrichia caninervis patches (i.e., shoot colonies) by elevation zone site (rows) (A) before sampling, (B, C) in leaf clips during F_v/F_m assay. Healthy and mildly stressed shoots (green arrows) when wet (A, Low‐ and high‐elevation zones), often appeared green (A) or blackish‐red from protective pigments (A, low‐elevation zone). Severe stress can appear orange‐green (A, high‐elevation zone). Dead (chlorotic) shoots appear orange (A, red arrows). (B) Subset of variously stressed shoots (chlorophyll fluorescence stress metric, F_v/F_m, listed for each sample) from topography‐shrub microsites (labels) by elevation zone site. (C) Orange leaf tips suggest stressed shoots, while F_v/F_m reveals healthy photosynthetic efficiency (Figure 6), showing the importance of photosynthetic stress assays over visual stress surveys.

Figure 6.

Frequency histogram of biocrust moss (Syntrichia caninervis) microhabitats with severe, moderate, and mild levels of photosynthetic stress (estimated by the chlorophyll fluorescence metric, F_v/F_m) in 92 microsites colored by elevation zone site (see Figure 1). Each bin is open on its maximum value (i.e., the bin [0, 0.1) does not include 0.1). Four colored bands indicate moss stress categories to aid vulnerability assessment in the Mojave Desert. Healthy mosses vary in their maximum F_v/F_m, so we have broadened “unstressed” to begin at 0.6 (see “Results” for Objective 3).

Moss stress (F_v/F_m) vs. three scales of habitat structure and buffering (Objective 4)

Habitat type ANOVAs

Contradicting our expectations, S. caninervis F_v/F_m did not differ on average by elevation zone (Welch F _2,59 = 1.6, P = 0.204, Figure 4D). When topographical zones were pooled across sites, topography zone explained 21% of variation in F_v/F_m with the north‐facing zone supporting healthier mosses, on average, than the south‐facing or flat zones (Welch F _2,54 = 12.7, P < 0.0001, Figure 4E). Microhabitat type explained 23% of variation in F_v/F_m with mosses under shrub canopies healthier than those in drip lines or interspaces (Welch F _2,45 = 16.6, P < 0.0001, Figure 4F).

Habitat buffering model

The full regression model including all multiscale buffering variables (elevation, PDIR, shade, and their interactions) was statistically significant explaining 55.4% of variance in F_v/F_m (F _7,84 = 14.9, P < 0.0001; note this model should only be used for predictive purposes due to multicollinearity; Appendix S2). The highly significant reduced model (without elevation) had acceptably stable beta estimates for PDIR, shade, and their interaction providing a model that can be reliably used for mechanistic inference (F _3,88 = 23.5, P < 0.0001, R ² = 0.444, Figure 7). Specifically, shade and PDIR were positively and negatively related to F_v/F_m, respectively (Figure 7B); however, the full model indicated the importance of shade increased with site elevation (Figure 7A).

Figure 7.

(A) Relationship of summer moss stress (lower F_v/F_m values indicate higher stress) of 92 Syntrichia caninervis microsites to habitat buffering proxies measured at three spatial scales: macroclimate (elevation zone: point colors), mesoscale potential direct incident radiation (PDIR: circle size), microscale shade time (Shade.Index in B), and their interaction (:). The OLS regression line for shade vs stress is plotted with 95% confidence interval in grey, and the interaction of elevation with shade is illustrated by varying slopes (not tested) of three OLS best fit lines plotted for each elevation zone. Elevation is shown graphically, but could not be included in the final model due to its multicollinearity with shade (see full model in Appendix S2). (B) OLS regression results; residual plots are in Appendix S3.

Elevation zone models (topography zone and shade)

F_v/F_m was not related to shade at the low scrubland, but the two topography zones explained 57% of variation in which mean F_v/F_m was higher on the north‐facing zone than on the flat zone (Wald F _1,22 = 26.9, P < 0.0001, Figure 8C). At the mid shrubland, shade and topography, and their interaction explained 67% of variation in F_v/F_m in which the flat zone appeared more positively related to shade than the north‐ and south‐facing zones (regression slopes not tested; Wald F _5,28 = 12.5, P < 0.0001, Figure 8B). At the high woodland, shade was positively related to F_v/F_m and explained 52% of variation (Wald F _1,32 = 57.6, P < 0.0001; Figure 8A).

Figure 8.

Three elevation zone models and plots illustrate the linear relationship of summer moss stress of Syntrichia caninervis to topography (Topog.), annual shade time (Shade), and their interaction (Intera.). HCCM‐ANCOVA results for each elevation zone (panels) include alpha significance for each term in the full model (bold P < 0.05), R ² for the reduced model (including only significant terms), and best fit lines with 95% CI (grey bands) from OLS regression. Note: No mosses occurred on the south‐facing slope of the low scrubland (see Figure 1C).

DISCUSSION

F_v/F_m summer stress interpretation

Our Mojave Desert stress assay confirmed the extreme resiliency of S. caninervis to present‐day summer climate because we found no patch mortality in the 92 S. caninervis microsites studied (i.e., all F_v/F_m > 0, Figure 6) despite the extreme weather of summer 2017 (Table 1: 2017), which included record‐breaking temperatures (NOAA, 2020), low relative humidity, many small rain events (likely leading to negative carbon balance; Coe et al., 2012b), and over 45 days without rain. Moreover, only 32% of the S. caninervis microsites showed signs of moderate to severe photosynthetic stress as suggested by low F_v/F_m values relative to the maximum potential PSII efficiency of plants, ~0.85 (Figure 6). Interestingly, microsites with a moderate to extreme photosynthetic stress signal were present at all three elevation zones (Figure 6). F_v/F_m can be confidently used as a stress proxy for desiccation‐tolerant mosses and revealed a much larger variation in photosynthetic summer stress and photoprotective acclimation than could be detected visually. The appearance of sampled shoots ranged from evidently stressed with reddish‐orange leaf tips (indicating cell death) to those lacking visible damage but having dark sun‐pigmentation from acclimatory zeaxanthins (Ekwealor et al., 2021; Figure 5).

Reduced F_v/F_m in desiccation‐tolerant mosses may be caused by many stress factors including rapid drying, low water potential, high temperatures, excess light, and UV damage (Takács et al., 1999; Proctor, 2001; Proctor et al., 2007; Greenwood et al., 2019; Ekwealor et al., 2021), all of which are likely acute or chronic stressors in these Mojave mosses especially during hydroperiods when the mosses are metabolically active after rare precipitation events. Part of their impressive resiliency arises from being in a dry, dormant state most of the year during which they can tolerate extreme heat and extended drought (Stark, 2017). When fully desiccated, S. caninervis can tolerate intense solar loading of 120°C (248°F)—two times greater than typical Mojave summer air temperatures (Stark, 2005; Stark et al., 2009).

Historically called photoinhibition, reduced F_v/F_m has been described as a state of compromised efficiency in the light reactions of photosynthesis thought to result from deactivated PSII reaction centers and/or acclimatory changes in nonphotochemical quenching (NPQ; Demmig‐Adams and Adams, 2006). Specifically, there are three possible mechanisms of reduced F_v/F_m in desiccation‐tolerant bryophytes including rapid light acclimation through NPQ, and two longer‐term acclimation responses including PSII repair cycling and slowly reversible NPQ (ql), which Muller et al. (2001) stated are the result of prolonged, severe light stress. Regardless of the cellular mechanism, reduced F_v/F_m is usually interpreted as a photoprotective acclimation response to climate stress in mosses undergoing desiccation–rehydration cycles (Proctor and Smirnoff, 2000, 2011). Severely low F_v/F_m (<0.2) can precede mortality (e.g., Coe et al., 2020); however, some experiments have shown recovery from such low F_v/F_m (e.g., Cruz de Carvalho, 2011; Ekwealor et al., 2021). Further research is needed to determine the climate scenarios in which low F_v/F_m leads to reduced productivity or survival in bryophytes (e.g., Hájek and Vicherová, 2013). Regardless of long‐term resiliency potential, the large variation we observed in biocrust moss stress resistance presented a strong sample for testing the high‐elevation and habitat buffering hypotheses.

High‐elevation refugia hypothesis

We had anticipated that the cooler, wetter climate and mountain topography of the high woodland (Table 1A) would reduce summer moss stress relative to the low scrubland and mid shrubland, but found no such evidence to support high‐elevation buffering that could indicate potential for future high‐elevation refugia in the DNWR (Figures 4A, 6). The lack of signal for high‐elevation stress buffering along a 1200‐m macroclimate gradient is surprising given the evident increase in moss abundance with elevation along this aridity‐stress gradient. We suggest future comparisons in other drylands where floristics and/or community analyses have been conducted on bryophytes to determine if this hypothesis is lacking in other high‐elevation ecosystems where high‐elevation buffering appears to at least support increased moss cover (Nash et al., 1977; Clark, 2012; Seppelt et al., 2016).

In contrast, we found support for the habitat buffering hypothesis at all three elevation zones in which meso‐ and/or microscale habitat structure and buffering were positively related to summer stress resistance in S. caninervis. Mesoscale topography varied primarily by hydrological position and exposure (PDIR), while vegetation shading varied as a function of three microhabitat types related to shade shrub proximity (Figure 1C). To our knowledge, our study is the first to link natural (i.e., nonexperimental) fine‐scale habitat buffering to physiologically significant stress variation in a dryland moss. Future implications of these findings include the potential that the most resilient moss patches may not simply be those at high elevations in drylands, but could be in microhabitats scattered throughout all elevation zones within species' current ranges.

Habitat buffering hypothesis

We next discuss how our fine‐scale, within‐site habitat structure and buffering factors were related to moss stress in univariate and multivariate (i.e., multiscale buffering) frameworks.

Fine‐scale habitat buffering

Supporting the habitat buffering hypothesis, high‐shade canopy microhabitats were the most common habitat type for S. caninervis across elevation zones (Appendix S1) and harbored mosses with the highest F_v/F_m, thereby appearing the most stress resistant. Beneath these shrub canopies, mosses were shaded over 60% of the year regardless of elevation zone (Figure 4C, Figure 4F). Furthermore, our novel shade‐buffering proxy, percent annual shade time (Figure 2), appeared most important to summer‐stress resistance at the mid shrubland and high woodland, where shade was positively related to F_v/F_m within each mesoscale topography zone (Figure 8A, B). However, we found no evidence for a shade‐stress relationship at the low scrubland (Figure 8C). The lack of a shade signal in this most arid and exposed elevation zone contradicted our expectation because even high levels of microsite shade on the north zone did not correlate with higher F_v/F_m values. We interpret this to mean that shade was overpowered by two other habitat buffers. Specifically, on the steep north‐facing zone, topography buffering appears to have overpowered the buffering effect of microhabitat shade; however, there is a slight confoundedness here because steep north‐facing slopes do have increased shading that is contributed by topography of the slope itself. On the flat zone, hydrological buffering by the drainage basin likely overpowered shade buffering in this flat‐lying basin where additional soil saturation may have increased moss recovery time significantly regardless of the microhabitat shade level (Figure 1C). So, we conclude that increased habitat buffering complexity can be introduced by the possible confoundedness of shade and slope, or by hydrology patterns. We also conclude that the F_v/F_m patterns we report in fine‐scale habitat buffering at the two higher‐elevation zones suggest that most shrub microhabitats in the Mojave Desert can provide an impressively high and physiologically meaningful shade buffer for S. caninervis regardless of shrub species and elevation zone.

In addition to microscale habitat buffering, mesoscale buffering also appears important to moss stress resistance in the DNWR (Figure 8). We found small differences in topographical exposure (6–14% deviance in PDIR) are related to moss summer stress resistance and that, overall, northerly exposures appear to reduce stress even if slopes are gentle. For example, topography buffering appeared most important to stress at the low scrubland where a 14% reduction in exposure was linked to healthier mosses on the north‐facing slope compared to mosses on the flat drainage basin (Figures 1C, 8C). Mesoscale exposure may also be important at the high woodland, but topographical exposure was confounded with shade here; the steep, north‐facing zone was also the most shaded zone (Figure 8A). Such ecological confoundedness is often unavoidable—greater shade on northerly‐facing slopes is driven by topographical slope shading and higher vegetation density (Pelletier et al., 2018). However, we suggest that disentangling confounded fine‐scale habitat features is less important for climate change vulnerability monitoring; determining the contribution of fine‐scale habitat structure vs. macroclimate (elevation) to moss stress is perhaps a most critical first step to inform focal scales for monitoring and predicting future impacts on biocrust.

We found relationships that suggest mesoscale topography can drive unique buffering patterns for biocrust mosses that involve aspect and hydrological position, which appear to influence biocrust moss distribution. For example, on the highly exposed south‐facing slopes of the low scrubland a mere 7% PDIR buffer (Table 1B) was associated with an absence of moss on our reference south‐facing slope. In fact, no mosses were found on any south‐facing slopes in this elevation zone, suggesting a niche limitation for S. caninervis on soil surfaces receiving >1000 kJ cm⁻² yr⁻¹ at low elevations in the Mojave Desert (Table 1C, Figure 1C). Niche constraints in response to such small differences in exposure shed light on the sensitivity of these poikilohydric organisms and corroborate prior warnings of their potential vulnerability to even slight shifts in climate (Ladrón de Guevara and Maestre, 2022).

Multiscale habitat buffering

We also found strong support for the multiscale habitat buffering hypothesis using multivariate models, which revealed moss microsites for which two scales of habitat structure and buffering could predict moss stress resistance. At the mid shrubland, 67% of moss stress was predicted by mesoscale topography zone, microhabitat shade, and their interaction (Figure 8B). Similarly, 43% of stress across all elevation zones was related to mesoscale PDIR, microscale shade, and their interaction (Figure 7). There appeared to be an interaction between elevation zone (i.e. macroclimate) and shade in which the importance of shade increases with elevation (Figure 7). Overall, these models suggest that within‐site fine‐scale habitat buffering can have additive and interactive effects on moss stress and that potential exists for interactions with the macroscale environment, which we predict will be driven largely by macroclimate and shading by the dominant vegetation community. Specifically, effects of fine‐scale shade seem likely to vary based on topographical exposure, hydrological position, and elevation zone (or site); thus, we urge future researchers to anticipate such complexities in understanding ecological mechanisms of moss climate stress resistance (Appendix S2).

In the Mojave Desert and other drylands, sheltered fine‐scale habitats such as north‐facing slopes and soil beneath vegetation have been shown to reduce extremes in soil temperature, soil moisture, and moss patch thermal loading, but these studies do not link the habitat types or buffering to moss stress (Breshears et al., 1998; Bowker et al., 2000; Thompson et al., 2005; Kidron. 2009). The only two studies that have linked natural variation to photosynthetic stress in dryland moss also found microhabitat variation to be important, but did not compare its signal to macroscale environmental variation as we have done. In the first study, Alpert (1988) used infrared gas analysis to compare carbon balance (i.e., the carbon offset between respiration and photosynthesis) of mosses cycling in and out of desiccation, and found healthier carbon budgets in north‐facing, high shade habitats compared to south‐facing, low shade habitats for rock mosses in semiarid California. In the second study, Yin et al. (2017) experimentally altered moss shade environments in the temperate Gurbantunggut Desert of China They reported healthier S. caninervis patches (i.e., those having greater Fv/Fm and antioxidant concentrations) to shaded habitats under native shrub canopies vs. exposed patches in a shrub‐removal treatment. Our findings corroborate these studies and emphasize the potential that fine‐scale habitat buffering will be most relevant to these small plants if climate conditions persist or worsen.

Can biocrust mosses hide from climate change?

Many studies predict that climate change will reduce dryland moss biomass and thus the functional roles of these plants in future drylands, which may cause ecosystem‐wide consequences while potentially accelerating desertification (e.g., Coe and Sparks, 2014; Rodriguez‐Caballero et al., 2018; Ladrón de Guevara and Maestre, 2022). Predicting such climate change responses in mosses is more complex than for other plants given their unique poikilohydric physiologies and fine‐scale habitats, which require different temporal (e.g., intermittent hydroperiods) and spatial (i.e., microhabitat) scales of study than for large, homiohydric tracheophytes (He et al., 2016; Ladrón de Guevara and Maestre, 2022). The habitat buffering offered by shrubs and northerly‐facing slopes in our study exemplifies such fine‐scale spatial complexity. Such fine‐scale habitat buffering may have important implications for climate change presenting the likelihood that (1) microrefugia created by shade vegetation and northerly‐facing slopes will protect biocrust mosses in future climates and that (2) such fine‐scale microrefugia will be more prevalent and important to long‐term moss resilience than macrorefugia existing within or outside species ranges.

Increased drought and/or erratic summer precipitation is predicted to increasingly stress sensitive moss hydration cycling (e.g., Coe et al., 2012a). We warn that these hydrological stressors may strengthen the moss‐shrub dependency presenting an additional factor of vulnerability for biocrust mosses because some shrub species may not be resistant to climate change. However, high‐shade microhabitats in our study were not restricted to a particular shrub species by elevation zone, but rather, included over 30 species. This lack of shrub specificity should reduce future vulnerability in Mojave biocrust mosses due to the prevalence of diverse and abundant shrub habitat in this ecosystem. We must note, however, that S. caninervis establishment was most frequent under shorter shrubs with low‐lying canopies (<0.5 m) such as Ambrosia dumosa, Coleogyne ramosissima, and Artemisia spp. at the low scrubland, mid shrubland, and high woodland, respectively. Therefore, there is risk that large patches of biocrust moss may be threatened if short dominant shrub species die back (e.g., Ladrón de Guevara and Maestre, 2022). Nonetheless, the previously mentioned niche shrub diversity may help alleviate impacts on Mojave biocrust moss across the many other shrub microhabitats currently available.

Furthermore, unlike tracheophytes whose distributions are usually constrained to narrower elevational bands and single continents (Patiño and Vanderpoorten, 2018), many moss species, including S. caninervis, have broad ecological amplitude (María Ros et al., 1999; FNA, 2007+; eFloras, 2023). Such broad niches are thought to result from the ability of mosses to exploit similar microhabitats across vastly different environments (Ladrón de Guevara and Maestre, 2022) often resulting in cross‐continental or disjunct distributions (e.g., Burge et al., 2016; Carter et al., 2016; Brooks and Jauregui‐Lazo, 2023). The broad elevational and fine‐scale niche we observed for S. caninervis in the DNWR exemplifies such a pattern in which many shrub species provide similarly buffered microhabitats distributed across three unique elevation zones. Therefore, we foresee potential that the diversity of shade shrubs spanning multiple elevation zones and plant communities in the Mojave Desert may offer widespread and common microrefugia for biocrust mosses in future climates and may serve to minimize climate impacts on moss frequency, abundance, and ecological function in biocrusts.

CONCLUSIONS

We used a natural experiment to test the habitat buffering and high‐elevation refugia hypotheses for a broadly distributed, keystone biocrust moss, Syntrichia caninervis, within a climatically extreme part of its global distribution, the Mojave Desert. By characterizing its diverse microhabitat, aridity exposure, and photosynthetic stress resistance, we found evidence for physiologically significant summer‐habitat buffering in reduced‐exposure microhabitats created by meso‐ and microscales of topography and vegetation sometimes interacting with each other or with macroclimate (elevation). These buffered microsites were distributed across all three elevation zones in the species' current range, failing to support the high‐elevation buffering hypothesis while strongly supporting the fine‐scale (within‐site) habitat buffering hypothesis. We conclude that many of these buffered microsites appear as candidate microrefugia potentially capable of “hiding” this species from future climatic extremes via sheltering by shrubs and northerly‐facing slopes. Contrary to most climate change experiments that predict low moss resiliency to increasing aridity and altered precipitation patterns in drylands, our findings present the possibility that at least one ecologically critical biocrust moss may be more prepared for a changing climate than previously thought, as long as the mortality of the associated shrubs does not significantly accelerate.

AUTHOR CONTRIBUTIONS

T.A.C. conceptualized this dissertation research under advisership by L.R.S. and D.D. T.A.C. performed all analyses, made graphics, and wrote the original draft. T.A.C. and A.R. performed field investigations. T.A.C., A.R., and J.G. performed laboratory work. All authors contributed to manuscript review and editing.

Supporting information

Appendix S1. Microhabitat type relative frequency by elevation zone site.

Appendix S2. Correlation and regression table for full habitat buffering model.

Appendix S3. Residual diagnostic plots for reduced habitat buffering regression model.

Clark, T. A. , Russell A., Greenwood J. L., Devitt D., Stanton D., and Stark L. R.. 2025. Can biocrust moss hide from climate change? Fine‐scale habitat sheltering improves summer‐stress resistance in Syntrichia caninervis . American Journal of Botany 112(2): e16464. 10.1002/ajb2.16464

DATA AVAILABILITY STATEMENT

Data and R code used in this manuscript are openly available on GitHub at https://github.com/TreesaClark/Syntrichia_summer_stress.