Tropical forests are disappearing worldwide mainly due to human-related causes, with seasonally dry tropical forests (hereafter dry forests) particularly suffering from intensified land-use (Miles et al., 2006; Corona-Núñez et al., 2021). Dry forests cover 42% of all tropical habitats, are home to at least one billion people, and maintain the livelihood of millions of people by delivering a wide range of ecosystem services, such as nutrient cycling, primary productivity, food supply, and income generation (Sunderland et al., 2015; Silva and Barbosa, 2017). This dependency, combined with century-lasting human disturbances and intensifying climate change, has made dry forests one of the most impacted ecosystems on the planet (Riggio et al., 2020). To meet the sustainable development goals of the United Nations in these often economically poor environments (Schmerbeck and Fiener, 2015; UN General Assembly, 2015), safeguarding biodiversity and ensuring the continuous provision of ecosystem services are necessary. Both rely on the forest’s ability to regenerate from human disturbance (i.e., forest resilience; Chazdon, 2008).

The ability of dry forests to regenerate is strongly driven by edaphic legacies derived from disturbance, notably a reduced water availability and an increased soil erosion risk (Becknell and Powers, 2014). By influencing vegetation and soil development (Poorter et al., 2019) both edaphic properties are tightly linked to biodiversity, ecosystem functioning, productivity, and ecosystem service provision (Quijas et al., 2019). Water availability is a key limiting factor in dry forests and directly influences seed production, seed bank composition, germination, seedling survival, and plant growth (Vieira and Scariot, 2006). However, the strong dependency of the local population on forest goods and services has intensified water limitations. Conversion of 50% of dry forests worldwide has drastically altered the infiltration capacity of their soils, as agricultural practices compact and seal soils with long-lasting negative effects on water availability (Leite et al., 2018). The subsequent drought stress has aggravated seedling mortality and selected towards a set of resprouting species, leading to biotic homogenization in dry forests (Barros et al., 2021; Vanderlei et al., 2022). The limited infiltrability of soils under disturbance not only reduces the local water availability, including ground-water recharge and soil water holding capacity (Abdallah et al., 2021). It also increases the chances of overland flow run-off, flooding, and erosion (Ludwig et al., 2005; Bradshaw et al., 2007). Soil erosion, which refers to the removal of unstable soil surfaces mostly by wind and water, can cause drastic and potentially irreversible changes for ecological and socioecological systems on the local, regional, and continental scale (Flores et al., 2020). As soil erosion reduces the water holding capacity and therefore water availability of the soil it can directly influence all water-related limitations (Andraski and Lowery, 1992). Soil erosion is also accompanied by dust production and deposition in streams (Belnap et al., 2011), crop damage (Neff et al., 2005), and degraded soil quality (Karlen and Rice, 2015). Based on rainfall duration, magnitude, and intensity, dry forests are facing a moderate risk to rainfall-induced erosion (Panagos et al., 2017). However, the highly erodible soils of many dry forests (Rito et al., 2017), a high degree of leaf deciduousness during the dry season, and a reduced leaf litter cover caused by livestock grazing (Pfister et al., 1988) leave dry forest soils unprotected against the physical impact of raindrops during high-intensity rains, inducing a strong land degradation process (Bruijnzeel, 1990). Furthermore, soil erosion can be directly induced by human land use, such as slash-and-burn agriculture and shifting cultivation, two of the major disturbance agents in dry forests (Blackie et al., 2014; Borrelli et al., 2017). Moreover, rainfall and subsequent soil erosion over deforested areas can remove the seedbank required for forest regeneration (García-Fayos et al., 2010). Even under sufficient seed-supply soil erosion can obstruct forest recovery by creating a negative, self-reinforcing feedback loop, due to a lack in nutrient supply and capture (Flores et al., 2020). Aggravatingly, soil formation rates are very often so low that once the nutrient-rich dry forest topsoil layer has been removed by erosion, opportunities for forest restoration or regeneration are highly limited (Lal, 1990). The disturbance-derived reduced water availability in combination with increased soil erosion represent effective environmental filters selecting for floristic subsets of stress-tolerant plant species and thus modify successional trajectories during regeneration (Flores et al., 2020). Continuing disturbances over decades can lead to severe soil degradation (Celentano et al., 2017), forcing the local population to expand their activities deeper into old-growth forests. This further destabilizes and degrades dry forest soils, also with consequences for soil water availability and plant species composition. As a result, large parts of dry forests are being turned into steppes or deserts (Vieira et al., 2015). Despite the functional links between soil health and forest regeneration, studies on soil and water-related ecosystem services have been largely neglected in dry forests, even though they could be the decisive factor for their resilience (Calvo-Rodriguez et al., 2017; Quijas et al., 2019).

After disturbance, soils are often bare and degraded and are covered by primary colonizers during secondary succession like biological soil crusts (biocrusts) (Pointing and Belnap, 2012; Szyja et al., 2019). Biocrusts are a complex community of organisms, composed of cyanobacteria, green microalgae, bryophytes, lichens, fungi, heterotrophic bacteria, and archaea, as well as representatives of several invertebrate animal groups, living in or on the uppermost millimeters of the soil (Belnap et al., 2016). Poikilohydry and other adaptations have enabled them to survive harsh abiotic conditions like high solar irradiance, extreme temperatures, and water limitations (Bowker, 2007), making them key organisms during primary and secondary succession in drylands, including dry forests (Szyja et al., 2019). Biocrusts are regarded as ecosystem engineers, as many of their soil-related effects influence the establishment, survival, and productivity of vascular vegetation (Havrilla et al., 2019). Specifically, their influence on soil water availability and erosion control could affect dry forest regeneration. More precisely, biocrusts influence soil hydrology via changes in soil porosity, aggregation, organic matter content, and water repellency, thereby affecting soil water infiltration, hydraulic conductivity, and retention (Chamizo et al., 2012; Rodríguez-Caballero et al., 2013). Additionally, by changing soil surface roughness and water storage capacity, they affect the water residence time on the surface and water velocity during runoff (Rodríguez-Caballero et al., 2012, 2015). Erosion-reducing effects result from the fact that biocrusts entrap soil particles via organic exudates (extracellular polysaccharide sheaths; EPS) and filamentous structures. Thereby they stabilize the soil surface (Rossi et al., 2018), promote soil aggregate formation and increase erosion resistance against wind and water (Belnap et al., 2014; Chamizo et al., 2015). The biocrust effect on edaphic properties increases with biocrust developmental stage, which can be related to an increase in biocrust biomass, thickness, and EPS production, which stabilize the soil (Rossi et al., 2018). Another stabilizing factor is the biocrust-contributed soil organic carbon (SOC), which increases with biocrust succession and can be composed of increasingly more water repellent substances (Drahorad and Felix-Henningsen, 2013). As anthropogenic disturbance in biocrusts leads to a retrogressive shift to an earlier successional stage human impact can impair their ecosystem service provision (Kuske et al., 2012). It is plausible to assume that the edaphic ecosystem services provided by biocrusts could shape dry forest resilience. Their impact on soil water infiltration and stability has the potential to influence two of the most important edaphic properties necessary for successful dry forest regeneration, water availability and erosion control. Given that ecosystem engineering impacts vary with environmental context, with a greater effect in more arid or resource poor environments (Wright et al., 2006), biocrusts could be highly significant for semi-arid, often resource poor dry forests (Rodríguez-Caballero et al., 2018), but also highly threatened by the ever-increasing human disturbance in these forests. Despite such potentially far-reaching consequences for forest resilience, the ecosystem engineering effect of biocrusts in dry forests is yet to be investigated, both in their natural state and during regeneration after disturbance.

The most continuous dry forest of the New World, the Caatinga in northeast Brazil, has been transformed by intense human impact since the 16th century and currently suffers from slash-and-burn agriculture and free-ranging livestock farming (Silva and Barbosa, 2017). The remaining forest stands, even if denoted as natural reserves and protected sites, often experience chronic anthropogenic disturbance like firewood, fodder, and timber collection (Arnan et al., 2018). The continuous opening of the Caatinga old-growth dry forest has transformed it into a dynamic mosaic of active and abandoned fields and forests of different successional stages (Tabarelli et al., 2017). Under such conditions, biocrust occurrence is facilitated, e.g., on abandoned fields, with an average cover of 8.1%, and locally over 50% (Szyja et al., 2019). Considering that large parts of the Caatinga are composed of structurally unstable Arenosols with a low water holding capacity, biocrusts may provide essential ecosystem services related to erosion control and soil water availability, especially during times of low vegetation cover, i.e., the dry season. However, the vulnerability of biocrusts to small- and large-scale disturbances together with soils of low aggregate stability might result in an inability of the biocrust to fulfil their pedological and ecological roles (Chamizo et al., 2015). In this context, the Caatinga offers an excellent opportunity to investigate the ecological role played by biocrusts in dry forests.

The aim of this study was to evaluate the impact of biocrusts on soil water infiltration and erosion control in the Caatinga dry forest and discuss its implications for ecosystem resilience and vegetation regeneration. For this, we measured sorptivity, unsaturated hydraulic conductivity, and water repellency as infiltration parameters and soil surface penetration resistance and wet aggregate stability as erosion related parameters. To assess the effect of human disturbance on these biocrust-mediated ecosystem services, the investigations were conducted at two sites with contrasting disturbance levels: an active cashew plantation (grazed by cattle) and an abandoned pastureland with regenerating Caatinga vegetation. We hypothesize that biocrust presence reduces infiltrability compared to biocrust-free soil, with a stronger water-repellent effect in the later successional biocrust due to a higher amount of soil organic carbon. We also expect that biocrusts will increase soil penetration resistance and wet aggregate stability, even under disturbed conditions, as there they can be the main soil stabilizers due to a reduced vegetation cover, but that disturbance will lessen this stabilizing effect due to successional setbacks, which confer weaker erosion resistance (Belnap and Büdel, 2016).

Biocrust presence substantially reduced infiltration, by lowering sorptivity and unsaturated hydraulic conductivity, but increasing water repellency of the soil (Figures 2A–C; Supplementary Table S2, statistical results in Supplementary Table S3). In combination, these values point towards a strong reduction of water input into deeper soil layers by biocrust coverage of the soil surface. This negative biocrust effect on infiltration was consistent between sites, but the disturbance regime exerted a strong impact on all hydrological parameters, with differences already between control soils but also between biocrusts. Comparing the two sites (i.e., disturbance pressure), the lowest absolute values for sorptivity were found in the regenerating site, where they were 54 and 85% lower compared to the disturbed site in both biocrusts and biocrust-free control soils, respectively. The same was true for unsaturated hydraulic conductivity, with values 64 and 84% lower in biocrusts and biocrust-free control soils, respectively (Supplementary Tables S2, S3). This indicates that the biocrust-induced decrease of water uptake into deeper soil layers is larger under regenerating conditions. Within sites, the actual ecosystem impact of the biocrust was largest in the disturbed sites. This is reflected by a reduction of sorptivity by 48% in the disturbed site and by 37% in the regenerating site, compared to the control. Correspondingly, unsaturated hydraulic conductivity decreased by 42 and 35%, respectively. The opposite was found for the repellency index, which was highest in the regenerating site and got significantly increased by biocrust presence only in this site (increase of 115%). Under disturbed conditions biocrusts did not influence the repellency index, which was similar in the early successional biocrust and the biocrust-free control soil (Supplementary Table S2). Soil repellency was strongly and positively related with SOC, for both biocrusts and control soils (Figure 2D).

Dry biocrust presence increased maximum penetration resistance, independent of site (Figure 3; Table 2) and only showed one layer of increased penetration resistance at a depth of ca. 2–3 mm (Figure 4). This is indicative of only one biocrust layer within the soil profile at both sites. The soil stabilizing effect was 4.1-fold at the disturbed site (control_dry: 0.16 ± 0.02 MPa; biocrust_dry: 0.64 ± 0.17 MPa) and 3.9-fold at the regenerating site (control_dry: 0.34 ± 0.09 MPa; biocrust_dry: 1.32 ± 0.26 MPa) and thus independent of the disturbance regime. Wetting reduced penetration resistance values for all sites. In the disturbed site the biocrusts remained harder than the biocrust-free soils, while the difference disappeared in the regenerating site. The observed mean maximum penetration resistance values under wet conditions were similar in biocrusts (0.16 ± 0.05 MPa for disturbed and 0.23 ± 0.07 MPa for regenerating), despite differences in biocrust species compositions and successional stages. It is important to note that the softening of the biocrust under wet conditions does not indicate a lower protection against water-induced erosion, which was tested for with the aggregate stability and raindrop simulation measurements. At the disturbed sites, vertical penetration resistance profiles were more homogeneous showing less variation on penetration resistance at any depth and for any soil surface compared to the regenerating site (Figure 4). The maximum penetration resistance of biocrusts peaked at similar soil depths in both sites (0.22 ± 0.04 cm for disturbed; 0.24 ± 0.05 cm for regenerating).

Aggregate stability, i.e., the protection against water erosion, did not differ between the biocrusts of the disturbed and the regenerating site when analyzing geometric mean weight diameter (Figure 5A). This value describes the stability of biocrust aggregates against mechanic and hydraulic stress and was similar between both sites (9.18 ± 1.05 mm in the regenerating site, 9.00 ± 2.19 mm in the disturbed site; Supplementary Table S3). In contrast, when investigating the effect of raindrop impact, i.e., the vertical kinetic energy transferred onto the biocrust until breakage, a strong difference was observed between the biocrust successional stages (Figure 5B). In the regenerating site, the kinetic energy by raindrop impact necessary to break apart a biocrust aggregate was significantly (almost three times) higher than for the disturbed biocrust (0.54 ± 0.22 J in the regenerating site, 0.19 ± 0.22 J in the disturbed site; Supplementary Table S4).

The low infiltration on biocrusts indicates a reduced water input into the soil under both disturbed and regenerating conditions. The lowest infiltration rates are measured for well-developed dark cyanobacteria-dominated biocrusts found in regenerating vegetation sites, although the biocrust-induced ecosystem impact is largest in more open sites subjected to human disturbance. In constrast, biocrust presence drastically increases protection against water-induced soil erosion not only during regeneration but also under disturbance and for different soil textures. Stability increases considerably with progression of biocrust succession, but early successional biocrusts already form stable aggregates against water erosion on sandy soils even under ongoing disturbance. Despite their low stability against raindrop impact, this result suggests that biocrusts confer protection to dry forest soils even in the worst combination of land-use and soil conditions with a very high soil erosion risk. The findings suggest that biocrusts may take on some of the ecosystem functioning roles otherwise held by vascular plants, influencing dry forest resilience and regeneration.

Despite earlier controversial findings, our results are in accordance with previous meta-analyses (Chamizo et al., 2016; Eldridge et al., 2020). They reinforce the idea that biocrusts limit the infiltration of water to greater depths in sandy soils, and that their infiltration-reducing effect increases with biocrust successional stage (Warren, 2003; Bowker et al., 2008). All three investigated infiltration parameters of the studied Caatinga biocrusts were within the lower range reported for cyanobacteria-dominated biocrusts on sandy soils (Drahorad et al., 2013; Lichner et al., 2018; Guan and Liu, 2021). Water infiltration was consequently strongly reduced by sorptivity and unsaturated hydraulic conductivity, but only moderately by water repellency (with repellency index values in biocrusts in literature ranging from 1.9 to 210; cf. Keck et al., 2016). Our results also confirm that biocrusts protect mobile sediments from water-induced soil erosion (Chamizo et al., 2017), with later successional stages delivering higher soil protection (Bowker et al., 2008). Although their protective effect decreases under disturbance the stability conveyed by early successional biocrusts against soil erosion even under disturbed conditions could explain the unexpected result of Leite et al. (2018), who reported no difference in erosion rates between recently abandoned fields and old-growth Caatinga forest. Our stability values were all within ranges previously reported for biocrusts on sandy soils (Zhao et al., 2014; Chamizo et al., 2018; Felde et al., 2018), while slightly towards the lower end under disturbance. Except for geometric mean weight diameter, which had significantly higher values for biocrusts at both sites than previously reported (2.51 ± 0.11 mm, Kakeh et al., 2018), pointing towards an extraordinarily high aggregate stability against water erosion. This remarkably high stability, however, might be due to differing sampling methods employed, as in our study only the biocrust itself (3–5.5 mm) was investigated while Kakeh et al. (2018) investigated the first 5 cm of soil, introducing more soil material to the sample and “diluting” the effect of the biocrust. All penetration resistance measurements were dominated by a single peak of maximum resistance, which showed that the biocrust top layer did not develop on older buried soil surfaces, as was described in Felde et al. (2018). As expected, penetration resistance was reduced after disturbance, which was also described for biocrusts from South Africa (Dojani et al., 2011). The fact that the disturbed site had more uniform penetration resistance is indicative of the loss of soil structure and a higher structural homogeneity after disturbance. The peak of maximum penetration resistance disappeared under wet conditions, which is in accordance to the results of previous studies, that also showed a decrease of penetration resistance under wet conditions (Lapen et al., 2004; Chamizo et al., 2015). However, since this is the first study with high-resolution depth-dependent penetration resistance data of dry and wet biocrusts, it remains unclear if the pattern of completely disappearing penetration resistance peaks of wet biocrusts will replicate in other ecosystems with different edaphic and climatic conditions.

Much of the effects of biocrusts on soil hydrology can be attributed to EPS production. For example, low infiltration rates on biocrusts are likely caused by the blocking of soil matrix pores upon wetting and the subsequent swelling of EPS (Belnap, 2006; Fischer et al., 2010). Biocrust EPS production could also be responsible for the strong water repellency of the late successional biocrusts at the regenerating site. Biocrust hydrophobicity was related to an increasing SOC content (Figure 2D), and up to 75% of SOC is stored as EPS in biocrusts (Mager, 2010). The ability of cyanobacteria-dominated biocrusts to stabilize sandy soils is also connected to EPS production. It functions as a binding matrix between filamentous cyanobacteria and soil particles, leading to soil particle aggregation (Guhra et al., 2019) and dust capture (Danin and Ganor, 1991), which cements the upper layer of the soil (Mazor et al., 1996) and stabilizes the surface (Hu et al., 2002). Specifically, the species Microcoleus vaginatus and Scytonema sp. have a very high protective effect against wind and water erosion, respectively and only need low biomass to confer protection (Hu et al., 2002). Both were found throughout the Catimbau National Park and were present even under heavy disturbance (Szyja et al., 2019), which explains the erosion protection even of early successional biocrusts under disturbance. The observed increase of biocrust impact on edaphic properties with biocrust succession is related to gains in biocrust biomass, thickness, SOC content and EPS production (Bowker et al., 2008; Rossi et al., 2018). However, biocrust impacts can also be shaped by disturbance, with opposing effects based on biocrust succession and disturbance type (Faist et al., 2017). Acute trampling of the early successional biocrust could increase infiltration compared to untrampled biocrusts, due to the destruction of biocrust integrity and disruption of biocrust-created soil aggregates. In contrast and counterintuitively, trampling of the late-successional cyanobacteria-dominated biocrust could decrease infiltration, due to drastic changes in pore geometry (Felde et al., 2014) and pore clogging by biocrust fragments left in place (Faist et al., 2017). This biocrust succession-based disturbance effect can explain the strong limitation of infiltration of the late successional biocrust compared to the early successional biocrust, although the disturbances at the studied sites were not only acute but also chronic and long-term. Furthermore, disturbance can lead to a loss of fine soil particles to which nutrients are bound, including stabilizing SOC, increasing soil erosion (Hagemann et al., 2017). Additionally, site-specific underlying soil properties (e.g., texture, pore connectivity, bulk density) can significantly shape biocrust effects. They influence biocrust protection against erosion and can change or even override the biocrust effect on infiltration. Firstly, the formation of stabilizing biocrusts is dependent on the soil texture, as coarse textured soils produce thin and less stable biocrusts than fine textured soils (Rozenstein et al., 2014; Chamizo et al., 2015). Secondly, the loamy sand of the regenerating site naturally supports lower infiltration rates than the sandy soil of the disturbed site (Warren, 2003). Thirdly, the higher amount of fine soil particles of the regenerating site is more prone to physical crusting and soil compaction, both reducing infiltration but potentially increasing erosion (Drahorad and Felix-Henningsen, 2013; Chamizo et al., 2015). Consequently, the biocrust and biocrust-free loamy sand of the regenerating site promote a higher erosion protection and lower infiltration. The impact of soil properties is also reflected in the relative infiltration parameters per site, where the reduction in hydraulic properties was stronger in the early successional biocrust of the disturbed site, despite a higher infiltration rate. In combination with the lower water holding capacity and rapidly attained wilting point of the sandy substrate (Souza et al., 2020), the disturbed site may induce a fast onset of drought stress for vascular vegetation. The water availability in the disturbed site is therefore probably lower than in the regenerating site, despite higher hydraulic properties. The present study does not allow a quantitative assessment of texture effects, because controlled laboratory experiments that eliminate all other factors would be necessary for this (Rozenstein et al., 2014).

Biocrusts represent a surface runoff source in dry forests due to their reduced soil water infiltration, although the strength of this reduction is likely dependent on seasonality and soil moisture (Chamizo et al., 2012; Keck et al., 2016). Under the assumption that the reduced water infiltration is a stable ecosystem engineering effect, the patchy distribution of the biocrust community creates small-scale source-sink patterns of water- and nutrient availability within the landscape (Cantón et al., 2020). It is plausible to assume that matter fluxes from biocrusted, vegetation-free spots, e.g., abandoned fields, to deposition sites can generate preferential growth conditions and facilitate regeneration as a higher water availability is directly correlated to increased plant recruitment in dry forests (Vieira and Scariot, 2006). Since water redistribution in the Caatinga is naturally rare (de Figueiredo et al., 2016), this would be a particularly impactful ecosystem service. However, the high water infiltration and low water repellency of the sandy Caatinga soils can lead to biocrust-induced runoff being lost before it reaches vegetated patches, creating a resource deficit within the system (Ludwig et al., 2005). Additionally, the smooth surface of early successional biocrusts can potentially increase runoff velocity due to their higher connectivity of runoff pathways, promoting erosion despite its aggregating function (Belnap and Büdel, 2016). The reduced water intake into soils below cyanobacteria-dominated biocrusts can be counterbalanced by their greater soil moisture retention due to blocking of surface pores (Eldridge et al., 2020) and the strongly reduced water vapor diffusion that leads to lower evaporation rates (Benard et al., 2019). These processes support the existence of hotspots of maximum productivity, nutrient content, and microbial activity in dryland soils (Whitford, 2002), which directly benefit from the biocrust-induced restriction of water percolation into greater depths. Additionally, biocrusts can break up self-reinforcing erosion feedback-loops by reducing erosion directly and by increasing and protecting the local nutrient content (Barger et al., 2006; Flores et al., 2020). Erosion control can further help to sustain the life quality of the local population by preventing dust deposition in water sources (Belnap et al., 2011), increasing cropland productivity (Neff et al., 2005), and protecting the seedbank and seedlings (García-Fayos et al., 2010). Currently large parts of dry forests are threatened by desertification caused by unsustainable anthropogenic land exploitation (Miles et al., 2006; Vieira et al., 2015). However, biocrusts have been proven to be useful tools for nature-based dryland restoration and to combat land degradation and desertification (Dadzie et al., 2022; Maggioli et al., 2022). Their influence could become even more important, since dry forests will face dramatic changes and climate extremes, particularly an increase in drought stress for vascular plants (Torres et al., 2017). The following decline in plant biomass (Vieira and Scariot, 2006) would increase the risk for erosion of unprotected soils. Under such circumstances, biocrusts could be the last barrier against land degradation and the decisive factor for dry forest resilience. Edaphic ecosystem engineering by biocrusts must therefore be considered an underexplored key process for ecosystem functionality in anthropogenically disturbed dry forests.

This study shows that biocrusts reduce water infiltration and at the same time protect against water-induced soil erosion in a human-impacted tropical dry forest. To our knowledge, this ecosystem type was investigated for the first time (Weber et al., 2022). Combined with the high rainfall intensity during the rainy season, the marked decrease in infiltration through biocrusts might be a key factor for surface runoff. Despite showing high runoff parameters, both biocrusts were stable against water-induced erosion, although disturbed biocrusts conferred lower protection against raindrop erosion. The combination of lower erosion protection and reduced water infiltration under disturbance could negatively impact vascular plant establishment and productivity. In contrast, during regeneration biocrusts also decreased water infiltration but protected the most critical dry forest soil layer against soil erosion, which might increase the resilience of this ecosystem. Biocrusts have the potential to reduce land degradation, although their associated ecosystem services can be depleted by disturbance such as grazing. Considering an average biocrust cover of 8.1%, and locally more than 50% (Szyja et al., 2019), biocrusts could be particularly important local ecosystem engineers. Future studies should aim to investigate if biocrust-induced ecosystem service provision is universal across dry forests and independent of the strong seasonality. Related to erosion protection, future studies should focus on how biocrusts can be used as a soil protective agent against erosion for the reversion of the desertification process, and how climate change will impact their ecosystem service provision.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found at: www.figshare.com; DOI to access data directly: https://doi.org/10.6084/m9.figshare.21561183.v1.

MS, RW, BB, and VF conceived and planned the experiment. MS and SL carried out the experiments with help from RW, BB and VF. MS and SL analyzed the data. BB, RW, IL, MT, and VF contributed to the interpretation of the results. MS took the lead in writing the manuscript. RW, MT, IL, and BB supervised the project. All authors provided critical feedback and helped shape the research, analysis, and manuscript.

This study was funded by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES project ID: 88881.030482/2013-01), the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq-PELD project ID: 403770/2012-2) and by the German-Brazilian PROBRAL program (CAPES process 88881.030482/2013-01; DAAD project ID: 57413496) to RW, IL, and MT.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.