Impact statement
Wildfires are becoming more frequent and destructive across the world’s dryland regions, increasing bare ground, triggering erosion, with recovery occurring at decadal scales. In desert ecosystems like the Sonoran Desert, this problem is compounded by invasive grasses and growing human pressures – and many native plants are not fire-adapted, prolonging recovery. The result is a cycle of degradation that threatens water quality, soil health, wildlife habitat and the communities that depend on these landscapes. This research offers a practical, nature-based path forward. By testing two innovative restoration techniques using biological soil crusts – living communities of cyanobacteria, lichens and mosses that form a protective layer on the soil surface – this study demonstrates that burned, eroded desert soils can be actively restored, rather than left to recover on their own timescale. Conducted in the Tonto National Forest in Arizona, the findings show that applying cultivated biocrust material – either as living sods placed across erosion-prone areas or as seed-and-biocrust capsules – measurably increases biocrust recovery, and sods also resulted in stabilizing soils. The implications reach far beyond Arizona. Dryland ecosystems cover roughly 40% of the Earth’s land surface and are home to over two billion people. As climate change drives more frequent fire events globally, scalable and low-impact restoration tools are urgently needed. Biocrust restoration represents exactly that – an ecologically grounded strategy that land managers, conservation practitioners and policymakers can deploy across arid regions worldwide to rebuild soil health, protect watersheds and strengthen ecosystem resilience in the face of changing fire regimes and climate.
Introduction
Land degradation in drylands is accelerating worldwide due to climate change and land-use change (Prăvălie, Reference Prăvălie2016). Wildfire is becoming a new driver due to highly flammable invasive annual plants which enhance fuel connectivity (Fusco et al., Reference Fusco, Finn, Balch, Chelsea Nagy and Bradley2019; Wilder et al. Reference Wilder, Jarnevich, Baldwin, Black, Franklin, Grissom, Hovanes, Olsson, Malusa, Kibria, Li, Lien, Ponce, Rowe, Soto, Stahl, Young and Betancourt2021). Resultant fires accelerate the loss of native species, increase soil erosion and affect downslope water quality (Balch et al., Reference Balch, Bradley, D’Antonio and Gómez-Dans2013; Wilder et al. Reference Wilder, Jarnevich, Baldwin, Black, Franklin, Grissom, Hovanes, Olsson, Malusa, Kibria, Li, Lien, Ponce, Rowe, Soto, Stahl, Young and Betancourt2021). Efforts to stabilize soils in disturbed desert landscapes are often unsuccessful for a variety of reasons. Seeding and planting restoration success is generally poor due to low and variable precipitation and active soil erosion (Abella et al., Reference Abella, Gunn, Daniels, Springer and Nyoka2009; Rowe et al., Reference Rowe, Sprague, Ball, Langenfeld and Rivera2022). Interventions are often implemented outside the optimal timeframe, and such delays can reduce their ability to prevent early soil loss and to support vegetation recovery (Gómez-Rey and González-Prieto, Reference Gómez-Rey and González-Prieto2014). These challenges highlight the need for alternative restoration strategies that enhance native plant establishment and directly and rapidly enhance soil stability and function, even under low-precipitation conditions. Biocrust communities offer a promising tool for reducing erosion in arid regions (Zhao et al., Reference Zhao, Bowker, Zhang, Zaady, Weber, Büdel and Belnap2016).
Biological soil crusts (biocrusts) are thin, stable soil surface layers created by cyanobacteria, lichens, mosses, fungi and other microorganisms (Weber et al., Reference Weber, Belnap, Büdel, Antoninka, Barger, Chaudhary, Darrouzet-Nardi, Eldridge, Faist, Ferrenberg, Havrilla, Huber-Sannwald, Malam Issa, Maestre, Reed, Rodriguez-Caballero, Tucker, Young, Zhang, Zhao, Zhou and Bowker2022). This living skin over soil surfaces is crucial in stabilizing soils, regulating hydrology and enhancing nutrient cycling in drylands (Bowker et al., Reference Bowker, Belnap, Bala Chaudhary and Johnson2008; Delgado-Baquerizo et al., Reference Delgado-Baquerizo, Morillas, Maestre and Gallardo2013; Eldridge et al., Reference Eldridge, Reed, Travers, Bowker, Maestre, Ding, Havrilla, Rodriguez-Caballero, Barger, Weber, Antoninka, Belnap, Chaudhary, Faist, Ferrenberg, Huber-Sannwald, Issa and Zhao2020). Biocrusts cover approximately 12% of Earth’s terrestrial surface and about 30% of drylands (Rodriguez-Caballero et al., Reference Rodriguez-Caballero, Belnap, Büdel, Crutzen, Andreae, Pöschl and Weber2018). However, disturbances including wildfires and other human activities can reduce biocrust cover, increase soil erosion and exacerbate ecosystem degradation (Bowker et al., Reference Bowker, Belnap, Rosentreter and Graham2004a; Palmer et al., Reference Palmer, Hernandez and Lipson2020). While some studies have suggested neutral biocrust responses to wildfire (Bowker et al., Reference Bowker, Belnap, Rosentreter and Graham2004b; Condon et al., Reference Condon, Shinneman, Rosentreter and Coates2023), other studies have observed negative outcomes, including losses in microbial community connectivity, moss and lichen cover and biocrust species richness (Haubensak et al., Reference Haubensak, D’Antonio and Wixon2009; Root et al., Reference Root, Brinda and Dodson2017; Aanderud et al., Reference Aanderud, Bahr, Robinson, Belnap, Campbell, Gill, McMillian and Clair2019; Palmer et al., Reference Palmer, Hernandez and Lipson2020). In cases of propagule loss and post-fire soil erosion, biocrust restoration represents a promising nature-based strategy for rehabilitating degraded drylands and enhancing ecosystem resilience (Antoninka et al., Reference Antoninka, Faist, Rodriguez-Caballero, Young, Chaudhary, Condon and Pyke2020b), while also helping to restore multiple ecological functions, including soil stability, hydrology, nutrient cycling and potentially native plant communities (Bowker, Reference Bowker2007; Bowker et al., Reference Bowker, Doherty, Antoninka, Ramsey, DuPre and Durham2022).
Restoring biocrusts in degraded locations requires the addition of biocrust biota, either obtained from natural biocrusts (preferably prior to a planned disturbance, i.e., salvaged) or cultivated ex situ. Cultivating biocrusts to increase the availability and application of salvaged material is necessary to produce enough inoculum to scale up restoration (Tucker et al., Reference Tucker, Antoninka and Day2020). Greenhouse and field cultivation trials have demonstrated that biocrust can be cultivated and used in restoration (Muñoz-Rojas et al., Reference Muñoz-Rojas, Román, Roncero-Ramos, Erickson, Merritt, Aguila-Carricondo and Cantón2018; Giraldo-Silva et al., Reference Giraldo-Silva, Nelson, Barger and Garcia-Pichel2019; Antoninka et al., Reference Antoninka, Rowe, Weser, Langenfeld, Brady, Sprague and Day2024). However, translating this production into full functional recovery remains challenging. While cyanobacteria may expand rapidly after inoculation, improvements in soil stability are minimal, and late-successional lichen and moss components typically show limited establishment (Antoninka et al., Reference Antoninka, Bowker, Chuckran, Barger, Reed and Belnap2018). The means of field-delivery of inoculum can greatly influence establishment. For example, slurry or broadcast inoculation is easy to apply, but had low establishment, whereas “biocrust sods” provided immediate physical stability and higher survival (Rowe et al., Reference Rowe, Antoninka, Langenfeld, Brady, Sprague, Fastiggi, Kollath, Ramsey and Barker2026). Biocrust sods are intact, high-biomass biocrust mats produced by cultivating salvaged biocrust on a thin soil layer over biodegradable materials, including weed cloth or burlap (Bailey et al., Reference Bailey, Lauria, Jech, Mayorga, Anenberg, Griffen, Condon, Doherty, Tucker, Ramsey, Rowe, Reed, Antoninka, Durham, Cooke, Gibson, Olney, Charley, Covington, Nelson, Gonzalez, Pablo, Bia, Sloan, Yohannes Kebede, Fule, Grover, Dohrenwend and Bowker2026). This approach allows the biocrust community to develop as a cohesive, physically stable unit that can be rolled, transported and installed directly onto disturbed soils with minimal disruption to community structure (Jech et al., Reference Jech, Dohrenwend, Day, Barger, Antoninka, Bowker, Reed and Tucker2025; Rowe et al., Reference Rowe, Antoninka, Langenfeld, Brady, Sprague, Fastiggi, Kollath, Ramsey and Barker2026). Sods are a promising method to stabilize soil while restoring native biocrust communities in targeted areas, like eroding hillslopes. However, we also need methods that are less intensive for broader distribution of biocrusts and seeds, yet hand or mechanized broadcast of seeds and biocrust inoculum generally have poor success for various reasons (Abella et al., Reference Abella, Gunn, Daniels, Springer and Nyoka2009; Doherty et al., Reference Doherty, Grover, Bowker, Durham, Antoninka and Ramsey2020; Rowe et al., Reference Rowe, Sprague, Ball, Langenfeld and Rivera2022). Methods that can retain propagules in place, reduce abiotic stress and aid water retention for propagules are needed to improve success. One option is encapsulation and distribution of biocrust biota and seeds in gel capsules that maximize propagule amounts and allow light to penetrate (Mayorga et al., Reference Mayorga, Charley, Bailey, Nelson, Pablo, Bia, Bowker and Antoninka2026). Capsules are cheap and easy to make and distribute, but have yet to be field tested (Mayorga et al., Reference Mayorga, Charley, Bailey, Nelson, Pablo, Bia, Bowker and Antoninka2026).
Our study builds on these foundations by testing biocrust sod installations and capsule applications directly in post-fire field conditions. This approach represents an alternative to traditional inoculation techniques (e.g., broadcast of biocrust fragments) and remains relatively unexplored in applied field restoration contexts. By evaluating these methods under real post-wildfire conditions, we aim to assess their potential to accelerate soil stabilization and biocrust and native plant recovery in degraded desert ecosystems. At a burned site draining into Lake Roosevelt in the Tonto National Forest, Arizona, we tested (1) biocrust sods arrayed perpendicularly to erosional features compared to controls and (2) biocrust and seed capsules compared to unencapsulated inoculum and controls. This area, previously impacted by construction, was burned in the 2020 Bush Fire that was fueled by invasive grasses and prolonged drought conditions, exacerbating soil erosion. Specifically, we tested the following hypotheses: (1) Biocrust sod treatments will reduce sediment transport, increase soil stability and enhance biocrust establishment compared to untreated controls. We expected sod-treated soils to improve soil compared to controls, consistent with microscale hydrological benefits observed (Rodriguez-Caballero et al., Reference Rodriguez-Caballero, Canton, Chamizo, Afana and Sole-Benet2012; Faist et al., Reference Faist, Herrick, Belnap, Zee and Barger2017). (2) Biocrust sod transplants will facilitate the expansion of biocrust cover beyond sod-treated surfaces into adjacent microsites, resulting in higher biocrust cover in biocrust sod-treated plots compared to controls. (3) Biocrust and seed addition, particularly through capsule application compared to dry broadcast, will promote greater biocrust development and native plant establishment relative to uninoculated control plots.
Methods
Site description
Our research site is located within Tonto National Forest in the Sonoran Desert ecoregion adjacent to Roosevelt Lake, Arizona (33.730361 N, −111.212812 W, Supplementary Figure S1). The site occurs on a gentle slope to a moderate steep alluvial fan remnants, dissected alluvial plains and inset fans (USDA Forest Service, Southwestern Region, 2023). The site was previously affected by decades-old highway construction impacts, then burned in the 2020 Bush Fire, which burned over 193,000 acres in Tonto National Forest, northeast of Phoenix, AZ, USA, making it one of the largest desert-scrub fires in Arizona’s history. The burn severity at the site was low (USDA FS, 2020; USGS and USDA FS, 2026). See Supplementary Figure S1 for details regarding soils, biocrusts and plant community at the site. The mean annual precipitation (30-year normals) at the site is 38.1 cm and the mean annual air temperatures range from 12.7 to 28.4 °C (Supplementary Table S1). The first winter season, when sods were installed, had slightly above normal precipitation, but the following summer and winter were substantially below normal (Supplementary Table S1).
Sod experiment – experimental design
In fall 2023, we established ten plots based on visible rill features to directly assess the effectiveness of biocrust sods in stabilizing soils and reducing erosion. To do this, first we identified ten rills and their surrounding runoff producing zones across the site. Using the ten rills, we paired nearby rills with similar slope, aspect, vegetation and erosion severity resulting in five pairs, 3 m – 13 m apart, located within our 0.5 ha study site. Within each of the five pairs, one rill was randomly assigned as a biocrust sod treatment plot, while the other served as an untreated control (Figure 1). Each pair represented one block to control for local differences in slope and erosion severity in the statistical models.
Photo showing (A) erosion bridge installation below plots; (B) schematic diagram of a sod treatment plot and (C) control (no sod) plot showing fixed subplot quadrat sampling positions. Light blue rectangles represent biocrust sods (3 × 1 m and 2 × 1 m). Green squares indicate on-sod subplot positions (1, 3, 5 and 7) and orange markers indicate between-sod positions (2, 4 and 6). Analyses compared either the on-sod positions on the sod treatment plots with the same positions on the control plots or the between-sod positions on sod treatment and control plots. The erosion bridge was positioned 30 cm downslope from quadrat position 4 to measure soil surface changes over time.
Figure 1 Long description
Panel A is a photograph of a gravelly, reddish, sparsely vegetated slope with dry grass tufts and small colored flags. A level on two thin legs (the erosion bridge) spans the foreground; a grassy hillside, saguaro cacti and blue sky are behind. Panels B and C are schematics with a downward arrow marked “slope”. In B, the sod treatment plot, a light blue rectangle labelled “3 m x 1 m sod” holds green squares 7 (left) and 1 (right). Below it are orange squares 6 (left) and 2 (right). A second light blue rectangle, “2 m x 1 m”, holds green squares 5 (left) and 3 (right). Orange square 4 sits below it, centered, and two lines labeled “Erosion bridge” are at the bottom. Panel C, the control (no sod) plot, has the same layout of squares but no blue sod rectangles.
Biocrust sod cultivation and farm site
We cultivated biocrust sods at Scottsdale Community College (SCC) grounds (33.513788° N, 111.879778° W; Arizona, USA) from November 2022 until December 2023. The average precipitation at this site is 220.15 mm and the temperature averages 22.9 °C with highs reaching over 40 °C in the summer (1991–2020 Normals, PRISM data downloaded 8-11-2025). The temperatures ranged from 0.4 °C to 46.8 °C during the cultivation growing period (PRISM data downloaded 01-30-2026). The site was cleared of debris and leveled for a previous experiment (Antoninka et al., Reference Antoninka, Rowe, Weser, Langenfeld, Brady, Sprague and Day2024).
Sod cultivation followed the general cultivation framework of Antoninka et al. (Reference Antoninka, Rowe, Weser, Langenfeld, Brady, Sprague and Day2024) but instead of small plots, we grew the sods in long continuous rows, 1 m wide and 18–21 m long on top of a heavy plastic 20-year weed-blocker to exclude exotic invasive plants. In November 2022, we installed sods with a biodegradable weed cloth base (sheets of maize-derived poly-lactic acid, Vigoro), a 0.5 cm layer of substrate (2 L Quikrete graded and washed coarse sand m−2), topped by a continuous sheet of finely woven burlap fabric and finally a layer of salvaged biocrust sieved through 0.4 cm mesh and applied evenly over the burlap to achieve 20.2% biocrust surface cover (0.47 ± 0.86% cyanobacteria, 2.47 ± 0.92% moss and 17.65 ± 7.57% lichen) across the full sod length (200 mL/m2). We originally salvaged the biocrust prior to the construction of a trailhead in the McDowell Sonoran Preserve (33.7464, −111.8883, < 100 miles from the study site) in October 2019 and stored in the dark at ambient temperatures. Drip irrigation was installed, with a 50% shade cloth laid on the top to retain moisture and prevent overheating. During the initial cultivation period (November 2022–April 4, 2023), the sods were watered four–five times daily in short (2–3 min) cycles, maintaining continual hydration. Weeding was conducted by hand-clipping to avoid disturbing the developing crust. We irrigated the biocrusts during optimal growing temperatures (< 27 °C) and ceased irrigation in summer to allow natural dormancy during these stressful conditions. We re-established irrigation from mid-October to mid December 2023 (8 weeks), resulting in visible increases in total biocrust cover as seen on baseline measurements at time of installation.
Sod harvest and field installation
In mid-December 2023, we weeded, cut sods into 2 m and 3 m long sections and rolled sods around a PVC pipe for transport. For rolling, the biocrust surface was kept inward. Sods were kept moist during handling with plastic covering and transported to the field site within approximately 24 hours of harvest.
At the study site, we unrolled sods onto sod treatment plots, carefully peeling off the biodegradable weed cloth during the process, leaving the burlap layer as the base. We did so to make these sods permeable to vascular plants, thereby assisting with soil surface stability. Each sod treatment plot received two sods: one upslope 3 × 1 m sod and one downslope 2 × 1 m sod positioned 1 m apart, installed perpendicular to the main slope to intercept runoff and stabilize developing rills (Figure 1). Two sods were applied to create redundant interception opportunities for runoff and sediment, while spacing and length of the two sods were based on the dimensions of the plot.
Biocrust and plant sampling
We marked seven fixed sampling subplots (Figure 1B,C) per plot to guide repeatable quadrat placement. The four “on-sod” subplots were each positioned over the sod-covered areas (two per sod piece, 30 cm from edges; Figure 1B), or in equivalent areas where sods would have been placed in control plots. Three “between-sod” subplots were established to monitor whether sods facilitated biocrust colonization between sods in both sod treatment and control plots (Figure 1B,C).
To sample biocrust and soil coverage, we used the grid-point intercept method (30 cm by 30 cm quadrat) in the subplot locations in March 2024, October 2024 and March 2025. Soil surface categories included: bare soil, cyanobacteria (light and dark), lichen, moss, jute, litter and rock (<2 mm, 2 mm < 10 cm, 10 cm < 30 cm and > 30 cm). Any biocrust categories not intercepted but present in the plot were recorded as an additional 0.5 point. We calculated the percentage of total biocrust cover by summing cover of lichens, mosses, light cyanobacteria and dark cyanobacteria. Late-successional biocrust cover was calculated as the sum of dark cyanobacteria, lichen and moss cover, omitting early successional light cyanobacteria. We visually estimated total plant cover in Spring 2024.
Erosion and soil stability sampling
At the bottom of each plot, but immediately upslope from a developing gully, we installed an erosion bridge to measure distance to the soil surface (Figure 1A) (Shakesby, Reference Shakesby1993) and measured the center 14-pin positions in December 2023, March 2024 and February 2025. To normalize the soil surface elevation changes with baseline for analysis, we calculated the differences for each of the 14 individual pin positions per erosion bridge by subtracting baseline values (December 2023) from the subsequent measurement (March 2024) for the 2023–2024 difference and March 2024 from the February 2025 measurement (2024–2025 difference).
To test our hypothesis that biocrust increases soil stability, we took three complementary field measurements at the beginning (December 2023) and at the end of the experiment (February 2025). We followed established protocols to measure soil stability using the Jornada soil stability test kit (Herrick et al., Reference Herrick, Whitford, De Soyza and Zee2001), soil shear strength, using a Torvane shear device (DGSI-Durham Geo Slope Indicator) (Brunori et al., Reference Brunori, Penzo and Torri1989) and a pocket penetrometer (Forestry suppliers #77114) to measure unconfined compressive strength (Mousavi et al., Reference Mousavi, Abdi, Ghalandarayeshi and Page-Dumroese2021). In December 2023, one measurement per device was collected at each between-sod subplot, totaling 30 measurements per device. In February 2025, we improved precision by collecting three measurements per device at the same subplots, resulting in 90 measurements per device.
Capsule experiment: experimental design and treatments
In December 2023, we marked three 30 × 30 cm plots (each ≥ 30 cm apart) in each of ten bare areas (blocks) lacking visible biocrust at least 7 m away from sod plots at the same study site (30 plots total). Within each block, treatments were randomly applied to one of the three plots: (1) capsule application, (2) broadcast application or (3) untreated control in a randomized block design.
Capsules were formulated with biocrust organisms and seeds and diatomaceous earth to deter ant herbivory (Korunic et al., Reference Korunic, Rozman, Liška and Lucić2016). Each treatment plot received: 6.3 g of diatomaceous earth, 90 mL of salvaged biocrust material and 1.24 g of seeds, spread over the plot evenly either in capsule or broadcast form. To assure that the capsule plots had the same materials, we measured out the materials and filled the capsules for each plot separately. We used a manual capsule-making device (CAPSULE-IT, HerbAffair.com). All plots were covered with biodegradable jute netting to retain soil moisture and added biological materials. The salvaged biocrust material had the same origin as that used for sod cultivation, then cultivated in the Northern Arizona University (NAU) greenhouse following methods by Antoninka et al. (Reference Antoninka, Rowe, Weser, Langenfeld, Brady, Sprague and Day2024), generally resulting in biocrust comprised of 40% moss, 20% lichen and 60% cyanobacteria (Antoninka unpublished). The native seed mix applied to each plot included ~200 viable seeds each of three species: three-awn grass (Aristida purpurea, 0.37 g per plot), sideoats grama (Bouteloua curtipendula, 0.76 g per plot), and desert marigold (Baileya multiradiata, 0.11 g per plot). Seed weights were calculated using species-specific purity adjustments provided by the seed provider (CactusStore.com, Phoenix, AZ).
We followed the same methods described above to measure biocrust cover within 30 × 30 cm quadrats placed within each plot at three-time points: March 2024, October 2024 and March 2025. We also counted individuals of the seeded plant species in March 2024.
Statistical analysis
All statistical analyses and data visualization were conducted using R version 4.3.1 (R Core Team, 2021). Analyses were performed using the following packages: lme4 for linear mixed models (Bates et al., Reference Bates, Mächler, Bolker and Walker2015), glmmTMB for zero-inflated, generalized linear mixed models (Brooks et al., Reference Brooks, Kristensen, van, Magnusson, Berg, Nielsen, Skaug, Maechler and Bolker2017), emmeans for estimated marginal means and post hoc comparisons (Lenth et al., Reference Lenth, Buerkner, Herve, Love, Miguez, Riebl and Singmann2021) and ggplot2 for data visualization (Wickham, Reference Wickham2009). We used linear regressions to calculate predicted values across the range of observed data values using the function ggpredict() from the package ggeffects (Lüdecke, Reference Lüdecke2018). Where appropriate with a significant main or interaction effect, Tukey’s HSD was used for post hoc pairwise comparisons (α = 0.05) using multcomp (Hothorn et al., Reference Hothorn, Bretz and Westfall2008).
We applied 16 linear mixed models across the two experiments (see model details in Supplementary Table S2). For seedling counts collected in March 2024, a zero-inflated, generalized linear mixed model with a Poisson distribution and log link function was used to assess the fixed effect of treatment (three levels: capsule, broadcast and control) with block as a random effect. We examined residuals for each model and improved normality of model residuals by applying transformations (Supplementary Table S2).
Results
On-sod biocrust cover
Season, treatment and their interaction were all significant predictors of both total and late-successional biocrust cover for the on-sod treated plots (Figure 2 and Supplementary Table S2). Sod-treated subplots had significantly higher late-successional biocrust cover than control subplots in all sampled seasons, with the strength of the difference varying among seasons (Figure 2A). At the final timepoint of the study (March 2025), mean late-successional biocrust cover in on-sod positions was almost eight times more on sod plots (39.69 ± 1.69%) compared to on control plots (5.20 ± 2.45%; Supplementary Table S2). For the same on-sod positions, total biocrust cover was almost two-times higher on sod plots (41.08 ± 1.66%) compared with control plots (23.20 ± 4.0%; Supplementary Table S3).
Seasonal and treatment effects on late-successional biocrust in (A) on-sod positions and (B) between-sod positions and total biocrust cover in (C) on-sod positions and (D) between-sod positions. Treated (sod) and control plots were compared within each position. Biocrust cover (%) was measured on four sampling dates: December 2023, March 2024, October 2024 and March 2025. Mean (black dot) and standard error (black bars) are shown for each treatment group; colorful points display the raw data points. Letters indicate differences among treatments based on Tukey’s HSD tests.
Figure 2 Long description
Each panel plots percent cover (0 to 100) for Control and Sod on four dates: Dec 2023, Mar 2024, Oct 2024 and Mar 2025. Black dots show means with error bars; colored dots show raw data. Letters mark statistical groupings. A, late-successional cover at on-sod positions: sod is far higher than control on every date, about 40–45% versus 2–10%, except Oct 2024 (about 17% versus 3%). B, between-sod positions: cover is low (under about 15%) for both treatments, with little difference. C, total cover at on-sod positions: sod is higher than control in Dec 2023 (about 66% versus 6%), Oct 2024 (72% versus 56%) and Mar 2025 (41% versus 23%); Mar 2024 is similar (76% versus 63%). D, total cover at between-sod positions: control and sod are similar; cover is low in Dec 2023 (under 10%), high in Mar and Oct 2024 (55–65%) and lower in Mar 2025 (15–22%).
Between-sod biocrust cover
We found no significant sod treatment effect or significant interaction between treatment and season (Figure 2B,D and Supplementary Table S3). However, there was a significant season effect (Supplementary Table S3 and Figure 2B,D), in which late-successional biocrust was highest in March 2024 and total biocrust cover was highest in March and October 2024.
Plant and biocrust cover relationships
Sod treatment alone was not a significant predictor of plant cover in spring 2024 for either on-sod (F1,34 = 0.22, p = 0.643) or between-sod (F1,24 = 0.50, p = 0.485) positions in linear mixed models. However, the relationship between plant cover, biocrust cover and treatment was more complex. With plant cover as the response variable, linear mixed model regressions indicated a significant negative slope resulting from the interaction of sod treatments and late-successional biocrust cover in between-sod positions (F1,25 = 4.482, p = 0.044; Figure 3B and Supplementary Table S4), indicating that as late-successional biocrust cover between sod plots increased, plant cover decreased. Total biocrust cover, treatment and the interaction of total biocrust cover and treatment were all marginally significant (p < 0.1) predictors of plant cover in between-sod positions (Figure 3D and Supplementary Table S4); however, these marginally significant relationships did not hold when a single outlying point with anomalously high plant cover was removed. There was no statistically significant relationship between plant cover, total biocrust cover, late-successional biocrust cover or treatment for on-sod positions (Figure 3A,C).
Linear regressions of plant cover (%) in relation to (A) late-successional biocrust cover in on-sod positions, (B) late-successional biocrust cover in between-sod positions, (C) total biocrust cover in on-sod positions and (D) total biocrust cover in between-sod positions. Black dots depict raw plant cover data. Lines show predicted estimates from linear mixed models and colored ribbons show the bootstrapped 95% confidence intervals for predicted values with random effect variance set to zero. Significant predictors (p < 0.05) of plant cover as determined from linear mixed models are noted where they exist in panel B; otherwise, the plotted relationships were not statistically significant predictors of plant cover.
Figure 3 Long description
Each panel plots plant cover (0 to 25%) against biocrust cover, with separate Control and Sod sections. Panels A and B use late-successional biocrust cover; panels C and D use total biocrust cover. A and C are on-sod positions; B and D are between-sod positions. Black dots are raw data; most plant cover values are below 10%. In A, C and D, the points are scattered with no fitted line, so no relationship is significant. In B, between-sod positions, fitted lines with shaded 95% confidence ribbons appear. In control plots, the line rises from about 4% to 9.5% plant cover as late-successional cover increases from 0 to 46%. In sod plots, it falls from about 7% to 2.5%. The interaction of biocrust cover and treatment is significant (p = 0.044). One sod point in B and D is an outlier at about 22% plant cover.
Erosion bridge
We detected no differences in distance-to-surface measurements for treatment or the interaction with season, based on erosion bridge data (Supplementary Table S5). However, there was an effect in which in the first year both the treatment and control plots accrued soil, but lost soil in the second year (Supplementary Table S5 and Figure 4A).
Effects of treatment (sod and no sod) and time on (A) erosion bridge measurements – change in distance to soil surface (cm), (B) soil stability class, (C) penetrometer resistance (kg/cm2) and (D) Torvane shear resistance (kg/cm2). Change in distance to soil surface was calculated by subtracting December 2023 erosion bridge measurements from March 2024 (2023–2024) and March 2024 measurements from February 2025 (2024–2025). Negative values indicate erosion, while positive values indicate sediment deposition. Box plots show the median (line), interquartile range (box, 25th – 75th percentile) and outliers (points beyond 1.5× IQR). Lower case letters indicate differences among treatments * time and upper case letters indicate differences among seasonal differences/time based on Tukey’s HSD post hoc analysis for significant interactions.
Figure 4 Long description
Each panel has box plots with overlaid points for No sod (purple) and Sod (green). A, change in distance to soil surface (cm): in 2023–2024, medians for both are about +0.6 (deposition); in 2024–2025, both are slightly negative, about −0.3 to −0.5 (erosion). Periods differ (letters B then A); treatments do not. B, soil stability class (0–6): in Dec 2023, both medians are 3; in Feb 2025, no sod stays at 3 while sod rises to 5, significantly higher (letter b). C, penetration resistance (0–5 kg/cm2): sod medians (about 4.2 and 4.5) exceed no-sod medians (about 2.7 in Dec 2023, 3.3 in Feb 2025). D, Torvane resistance (kg/cm2): low in Dec 2023 (medians about 0.5–0.7) and higher in Feb 2025 (about 2.5 for no sod, 3 for sod); the dates differ (A then B), the treatments are similar.
Soil stability
Differences in soil stability between sods depended on the season sampled; no differences were detected between control and treated plots at the time of installation, December 2023, but by February 2025, sod plots exhibited significantly higher stability scores than controls (F1,88 = 4.71, p = 0.033, Supplementary Table S5 and Figure 4B).
Penetrometer and Torvane
We detected no differences in penetrometer resistance across treatments or season (Supplementary Table S5 and Figure 4C). We detected a difference in season, in which shear strength (Torvane) was greater at the end of the study, compared to initial conditions (Supplementary Table S5 and Figure 4D), but we detected no sod treatment differences.
Capsule experiment biocrust and seed application
Treatment and season were both significant predictors of late-successional biocrust cover in the capsule experiment (Supplementary Table S3). Averaged over all three seasons, the capsules resulted in higher late-successional biocrust cover than the no biocrust control plots and marginally higher late-successional biocrust cover (p = 0.08) compared with the broadcast treatment (Figure 5A and Supplementary Table S6). By March 2025, late-successional biocrust cover was as follows, mean ± SE: 16 ± 3.73% in capsule plots, 10.4 ± 3.33% in broadcast plots, and 4.8 ± 2.22% in control plots. Averaged across all treatments, late-successional biocrust cover was highest in March 2024 and lowest in March 2025; October 2024 values were intermediate between the two (Supplementary Table S3 and Figure 5A).
Differences in broadcast, capsule and control treatment effects on (A) late-successional biocrust cover and (B) light cyanobacteria across all three sampling seasons. Boxes show the median (black line) and interquartile ranges (box); colored dots show the raw data points for each season/treatment combination. Lowercase letters indicate differences in treatment effects averaged over season based on Tukey’s HSD post hoc analyses of linear mixed models. Season was statistically different for both models. Late-successional biocrust was higher in March 2024 than March 2025, but was not significantly different in October 2024 compared to the other sampling periods. Light cyanobacteria was significantly higher in the first two seasons compared to the final season (March 2025).
Figure 5 Long description
Both panels show box plots with points for Broadcast, Capsule and Control, each with three seasons: Mar 2024 (light blue), Oct 2024 (green) and Mar 2025 (orange). Y-axes run 0 to 100%. A, late-successional cover: broadcast medians are about 12%, 5% and 10%; capsule about 14%, 12% and 20%; control about 19%, 0% and 0%, with a wide spread up to 65% in Mar 2024. Treatment letters: broadcast ab, capsule b and control a. B, light cyanobacterial cover: broadcast medians about 85%, 85% and 52%; capsule about 81%, 81% and 58%; control about 78%, 76% and 48%. In all three treatments, cover is clearly lower in Mar 2025.
Season, but not treatment, had a significant effect on light-cyanobacterial cover (Supplementary Table S3 and Figure 5B). Mean light-cyanobacterial cover did not differ between March 2024 and October 2024, however, values declined significantly by March 2025 (Figure 5B). B. multiradiata (desert marigold) seedlings were the only seeded species observed in March 2024. The average number of B. multiradiata seedlings per plot differed among treatments such that broadcast (seedling mean ± s.e.: 7.1 ± 1.83) and capsule (6.7 ± 1.09) treatments were not different from each other, but different from the unseeded control (0.0 ± 0.00) plots (Z = −2.45, P = 0.014).
Discussion
By the end of this 2-year study, we found that we were able to successfully accelerate biocrust recovery using two distinct techniques (sods and capsules) in a degraded, burned system. Consistent with hypothesis 1, biocrust sods increased late-successional biocrust establishment and led to enhanced soil aggregate stability between sods. Contrary to hypothesis 1, sod application had no effect on water erosion, soil resistance to penetration or shear strength. Hypothesis 2 was not supported because we detected no visible evidence that sod application colonized outward from the sods. Hypothesis 3 was partially supported in that encapsulated biocrust addition resulted in more biocrust coverage than controls, while broadcast biocrust addition did not. However, native plant seeding resulted in greater establishment of one of the seeded species regardless of encapsulation.
Biocrust sods successfully establish biocrusts, but colonization did not spread
We found that biocrust sods significantly increased both late-successional and total biocrust cover. The sods themselves maintained late-successional cover of ~40% after 2 years, which is comparable to or better than some of the most successful biocrust restoration practices (Condon and Pyke, Reference Condon and Pyke2016; Doherty et al., Reference Doherty, Grover, Bowker, Durham, Antoninka and Ramsey2020), using either cultivated biocrust inoculum (Wang et al., Reference Wang, Liu, Li, Hu and Rao2009; Doherty et al., Reference Doherty, Grover, Bowker, Durham, Antoninka and Ramsey2020) or field-sourced biocrust inoculum (Condon and Pyke, Reference Condon and Pyke2016). Multiple outcomes of biocrust inoculation have been documented: (1) added biocrust inoculum does not speed biocrust development relative to background rates (Young et al., Reference Young, Bowker, Reed, Duniway and Belnap2019; Faist et al., Reference Faist, Antoninka, Belnap, Bowker, Duniway, Garcia‐Pichel, Nelson, Reed, Giraldo‐Silva, Velasco‐Ayuso and Barger2020), (2) a steady progression of growth or successional development is observed (Lan et al., Reference Lan, Zhang, Wu, Liu, Zhang and Hu2014) and (3) biocrusts fluctuate in abundance or development after inoculation (Condon and Pyke, Reference Condon and Pyke2016; Doherty et al., Reference Doherty, Grover, Bowker, Durham, Antoninka and Ramsey2020). Fluctuations in biocrust abundance during restoration are commonly attributed to weather and climate differences between seasons and years (Doherty et al., Reference Doherty, Grover, Bowker, Durham, Antoninka and Ramsey2020; Antoninka et al., Reference Antoninka, Bowker, Barger, Belnap, Giraldo-Silva, Reed, Garcia-Pichel and Duniway2020a). Here, we observed apparent seasonal fluctuation, most notably as a decline in cover measured after the only summer in our monitoring period, followed by a substantive rebound to the level of cover observed in the previous cool season. Biocrusts are only active when hydrated. In ecosystems that receive cool season precipitation (mid-autumn to mid-spring), it is often observed that periods of biocrust growth or net photosynthetic activity are concentrated in the cool season as opposed to the summer (Darrouzet-Nardi et al., Reference Darrouzet-Nardi, Reed, Grote and Belnap2015; Han et al., Reference Han, Wang, Li and Hu2022). In contrast, warm season hydration periods can be stressful for some biocrust biota because temperatures are supraoptimal and hydration periods too short to allow full recovery from the preceding desiccation event (Coe et al., Reference Coe, Belnap and Sparks2012).
Biocrust sod application promotes soil aggregate stability, but does not affect erosion
Biocrusts are a well-documented soil stabilizer (Huang et al., Reference Huang, Huang, Zhao and Liu2025) and can strongly influence the development of vascular plant communities, which also stabilize soils (Song et al., Reference Song, Li and Hui2017). Thus, a major motivating factor in our research was to determine if biocrust sods could be used to stabilize soils and reduce erosion in a site degraded by multiple disturbances, including fire. Sod treatments provided measurable improvements in soil stability of soil surfaces in between sod placements, in the second year. The sods themselves consist of well-aggregated, fabric-reinforced soil that confer immediate reductions in erodible soil surface, and these results indicate that sod use can continue to improve plot-scale soil aggregate stability over time in surfaces near to but not covered by sods. Though our visible cover data did not clearly demonstrate biocrust colonization or an increase of plant cover, we suggest that sods could have physically reduced movement of sediment into and out of the interspaces among sods, protecting soil intactness of soil aggregates, and increased the propagule supply of biocrust biota, boosting cryptic incipient biocrust formation and creation of new aggregates. These two effects cannot be clearly separated given our data, and either or their combination could have induced the patterns we observed.
Soil shear strength was temporally dynamic, increasing ~3-fold from the beginning of our experiment until the second observation. This finding aligns with the observed temporal increase in soil aggregate stability, suggesting an overall trend of increasing erosion resistance. However, unlike soil aggregate stability, these data did not support an influence of sod placement. Soils with greater soil aggregate stability would generally be expected to also have greater shear strength, but the measurements are based on distinct principles and do not always align (Mallen‐Cooper et al., Reference Mallen‐Cooper, Bowker, Antoninka and Eldridge2020). Penetrometer measurements neither increased through time nor with sod placement. While penetration resistance can be a negative indicator of soil surface wind erosion resistance (Li et al., Reference Li, Okin, Herrick, Belnap, Munson and Miller2010), it may also reflect compaction, density or lower porosity (Mallen‐Cooper et al., Reference Mallen‐Cooper, Bowker, Antoninka and Eldridge2020). Possibly these properties were static during our experiment or the site experienced counter-balancing change in more than one of them.
The effects of sod treatment on soil aggregate stability did not translate into measurable differences in erosion, as measured by the erosion bridges. Erosion bridge data showed that treatments had nearly identical changes in surface elevation: in the first year we detected a loss of surface height (indicating net erosion), but in the second year we detected a nearly equal gain of surface height (indicating net sedimentation). This temporal effect may be linked to increasing soil aggregate stability and shear strength through time or it may be linked to differences in rainfall. The low rainfall in the second year was possibly only sufficient to transport sediment from our plots a short distance, resulting in sedimentation where the erosion bridges were located.
Capsules promote biocrust establishment, but neither help nor hinder seeding success
In the capsule experiment, late-successional biocrust cover showed statistically significant differences across treatments, with capsule applications producing significantly higher cover than control, whereas broadcast biocrust additions did not produce higher cover than controls. These results suggest that capsules may increase the probability of a successful inoculation within 2 years and can establish the presence of taxa typically associated with well-established biocrusts. Similarly, Román et al. (2020) demonstrated that cyanobacteria can be successfully incorporated into extruded soil pellets and redistributed across degraded substrates, supporting their viability and effects on early soil surface properties. This effect was somewhat surprising because capsules only persisted for a short period of time due to heavy rain shortly after application which dissolved the capsules. Without these heavy rains, we would have expected a residence time of weeks to months based on previous controlled trials (Mayorga et al., Reference Mayorga, Charley, Bailey, Nelson, Pablo, Bia, Bowker and Antoninka2026). The short residence time means that the enhanced outcomes were not due to prolonged protection but may instead reflect some combination of the following hypothetical mechanisms: (1) A positive outcome of localized concentration of viable propagules engaging in facilitative interactions, (2) a purely retentive benefit in which the short presence of the capsule helped inoculum fragments bond to each other upon hydration forming larger masses, or in which the soluble, sticky capsule constituents themselves helped adhere the inoculum to the soil (Bailey et al., Reference Bailey, Lauria, Jech, Mayorga, Anenberg, Griffen, Condon, Doherty, Tucker, Ramsey, Rowe, Reed, Antoninka, Durham, Cooke, Gibson, Olney, Charley, Covington, Nelson, Gonzalez, Pablo, Bia, Sloan, Yohannes Kebede, Fule, Grover, Dohrenwend and Bowker2026) or (3) a short-duration amelioration of stress, such as a prolonged hydration event under enhanced physiological conditions (Coe et al., Reference Coe, Belnap and Sparks2012). We cannot discern among these hypothetical explanations with our data but advance them to inform future studies. In addition to biocrust responses, both capsule and broadcast treatments supported robust Baileya radiata emergence, a species that did not emerge in control plots. This finding means that although encapsulation did not improve emergence, seed could be successfully co-distributed with biocrust in capsules, promoting both native plants and biocrusts.
Comparing and contrasting emerging biocrust restoration techniques
We tested two very different methods to improve biocrust establishment in this study: biocrust capsules and sods. Though there was a large difference in their effect on treated areas in terms of biocrust coverage and development, their efficacy should not be compared directly because of the different levels of resource and effort input and, accordingly, different application targets. Sods require intensive input to create and are very effective at rapidly inducing a high degree of biocrust development in small, targeted application zones of several square meters over a shorter time horizon. Capsules are relatively quick to make and meant to be more broadly dispersible to ensure the potential for biocrust initial colonization exists across larger areas.
Despite the differences, biocrust sods and capsules are based on a common principle of inoculation coupled with simultaneous retention and habitat amelioration mechanisms (Bailey et al., Reference Bailey, Lauria, Jech, Mayorga, Anenberg, Griffen, Condon, Doherty, Tucker, Ramsey, Rowe, Reed, Antoninka, Durham, Cooke, Gibson, Olney, Charley, Covington, Nelson, Gonzalez, Pablo, Bia, Sloan, Yohannes Kebede, Fule, Grover, Dohrenwend and Bowker2026). We believe that sods are successful because they simultaneously inoculate biocrust propagules, retain them on site by allowing them to attach to fabric-reinforced soil before use and by stabilizing a sufficiently large patch of soil surface, thereby ameliorate the environment. Likewise, capsules not only contain biocrust inoculum but they also promote retention of the inoculum on site, either through greater mass concentration, adherence to soil surfaces or deterrence of animal pests (Mayorga et al., Reference Mayorga, Charley, Bailey, Nelson, Pablo, Bia, Bowker and Antoninka2026). Both biocrust sods and capsules are newly emergent, thus development of new variants and determination of their efficacy and range of use cases is an ongoing research subject (Bailey et al., Reference Bailey, Lauria, Jech, Mayorga, Anenberg, Griffen, Condon, Doherty, Tucker, Ramsey, Rowe, Reed, Antoninka, Durham, Cooke, Gibson, Olney, Charley, Covington, Nelson, Gonzalez, Pablo, Bia, Sloan, Yohannes Kebede, Fule, Grover, Dohrenwend and Bowker2026). Rader et al. (Reference Rader, Chiquoine, Weigand, Perkins, Munson and Abella2022) found that applying a diverse suite of treatments as a bet-hedging strategy in dynamic restoration environments can help buffer restoration projects from uncertainty and potential failure (Rader et al., Reference Rader, Chiquoine, Weigand, Perkins, Munson and Abella2022). Perhaps strategic combinations of these distinct biocrust restoration approaches with each other and with other dryland restoration techniques, such as seeding, will enable more consistent restoration successes in response to a variety of disturbances including wildfire.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/dry.2026.10054.
Acknowledgments
We are grateful to Scottsdale Community College for providing facilities for the experiment and the McDowell Sonoran Conservancy for providing experimental materials. John Weser, Natalie Case, Nelly Nelson, Seth Charlie, Ryan Yazzie, Michael Sloan, Lydia Bailey, Madeline Mayorga, Jane Brady, Debbie Langenfeld and Mary Fastiggi all assisted in our farm or field site.
AI statement
We used claude.ai (https://claude.ai/new) to shorten and reformat our abstract to the journal’s specifications; the revision was then edited by the authors.
Author contributions
All authors contributed to writing; D.O., K.G. and H.I.R. conducted data analyses, and designed figures; M.B., A.A. and H.I.R. conceptualized the experiment; D.O. managed the project and led data collection; all authors contributed to the field work, revised and approved the final version of the manuscript.
Data availability statement
Data and analysis files will be available on https://github.com/ksgriff/Biocrust-Sod-Erosion-Control upon publication.
Declaration of interests
The authors have no conflicts of interest to report.
Funding statement
This work was funded by the Salt River Project (SRP). Elvy Barton served as our SRP advisor.
Open access
Comments
Thank you for considering this submission to the special biocrust edition of Drylands. This manuscript reports on an early field trial of biocrusts in the field setting to stabilize an eroded site.