The field test sites were located on the northeastern Qinghai–Tibet Plateau (37∘37′ N, 101∘19′ E) in Qinghai Province, China (Fig. 1a). The area has a continental plateau climate with a mean air temperature of -1.7 ∘C and a mean annual precipitation of approximately 562 mm (Dai et al., 2020). It should be noted that approximately 80 % of the precipitation occurs during the growing season (between May and September), and the other 20 % occurs during the non-growing season. The main vegetation type in this region is Kobresia meadow, which is dominated by Kobresia humilis (Fig. 1b). The soil type in the study area is silt loam according to the USDA soil taxonomy system of classification, with a soil thickness of approximately 60–80 cm. The pH and EC are 7.5 ms m-1 and 6.7, respectively, in the study area (Y. Li et al., 2016).

In August 2020, we chose two study sites on the northeastern Qinghai–Tibet Plateau to avoid pseudoreplication, and two types of soil surfaces were selected at each study site, i.e., normal Kobresia meadow (NM, Fig. 1b) and biocrust meadow (BM, Fig. 1c). To reduce the differences caused by spatial heterogeneity, BM was selected adjacent to NM to ensure the soil type and topographic condition were the same. The vegetation cover in BM is usually less than 20 %, with a thick turf but no litter layer in the topsoil, and the BM type is dominated by cyanobacteria crust (ca. 80 %) (Y. Li et al., 2016). In contrast, NM has a dense vegetation cover and is mainly dominated by Kobresia pygmaea, with average plant heights of 1–3 cm. Furthermore, a clear typical turf horizon and litter layer were observed within the topsoil in NM, i.e., the Afe horizon. BM had a higher root biomass than that of NM, owing to its thick turf (Table 1).

We obtained the disturbed soil samples (i.e., non-ring knife soil samples) in NM and BM. Four quadrats (1×1 m) were randomly selected for soil sampling with a depth of 10 cm in each treatment using an earth-boring auger and then brought back to the laboratory to measure and analyze soil organic matter (SOM), soil microbial biomass carbon (MBC), microbial biomass nitrogen (MBN), total carbon (TC), total nitrogen (TN), and soil texture (PSD). Undisturbed cylindrical ring samples (i.e., ring knife soil samples) were also obtained in each treatment to determine the soil bulk density (BD), soil porosity, and soil hydraulic properties (i.e., soil water retention and soil water supply capacity). The soil infiltration rates were measured using a double-ring infiltrometer for each treatment.

First, the disturbed soil samples were sieved through 0.25 and 2 mm soil sieves to remove debris and roots for the analysis of soil properties. SOM was measured based on the Walkley–Black procedure (Nelson and Sommers, 1982), MBC and MBN were measured by the chloroform fumigation direct extraction method (Vance et al., 1987), and TC and TN were measured using an element analyzer (Elementar Vario EL III, Hanau, Germany). PSD was determined using a Mastersizer 2000 (Malvern Instruments, UK). BD was measured as the ratio of the oven-dried soil mass to the core volume (100 cm3). The soil total porosity, soil capillary porosity, and soil non-capillary porosity were measured using the following equations (Dai et al., 2020): 1TP=1-BDds×100%,2CP=CWC×BD,3NCP=TP-CP, where TP, CP, and NCP represent the soil total porosity (%), soil capillary porosity (%), and soil non-capillary porosity (%), respectively; CWC represents the soil capillary water capacity; and ds is the soil particle density, which was assumed to be 2.65 g cm-3.

The soil water retention curves (SWRCs) were measured using a pressure plate apparatus (1500 F1, Soil Moisture Equipment Corp., SEC, USA), and the relationship between soil water content and matric potential was fitted by the Gardner model. The formula of the Gardner model is as follows (Gardner et al., 1970): 4h=Aθ-B, where h is the soil water content (%), θ is the matric potential (kPa), and A and B are the fitting parameters. Higher values of A⋅B and A indicate a higher soil water supply capacity and soil water retention capacity, respectively.

In this study, to compare the differences between BM and NM in soil water retention and soil properties, we conducted one-way analysis of variance (ANOVA) statistical tests to determine differences in plant and soil properties for the same soil layers between BM and NM, and a least-significant-difference test (P<0.05) was conducted when significant differences were detected by ANOVA. To explore the relationship between soil properties and soil water retention and for quantitative evaluation of the effects of soil properties on soil–soil water retention, Pearson's correlation and a variance partition in the analysis were used by R software version 3.4.3 with the hier.part and corrplot packages. Furthermore, structural equation modeling was used to examine the soil properties' direct and indirect effects on soil water retention.

Sand content dominated the soil texture in the 0–40 cm soil layer across the two surface soil types (mean 61.69 %), followed by sand (mean 30.13 %) and clay (mean 8.18 %) (Fig. 2). Specifically, the 0–10 cm clay content in BM was 9 % higher than that in NM, whereas the 10–40 cm clay content in BM was 16 % lower than that in NM, especially for the 10–20 cm soil layer (P<0.001). In contrast, the 0–40 cm silt content in BM was higher than that in NM, especially for the 20–30 cm soil layer (P<0.05). However, no clear pattern was observed for the sand content between BM and NM. Overall, the 0–40 cm clay content (8.62 %) in NM was 11 % higher than that in BM (7.69 %), whereas in 0–40 cm, the silt content (61.24 %) in NM was nearly equal to that in BM (62.13 %).

There were no significant differences for 0–40 cm BD, 0–40 cm TP, 0–40 cm CP, and 0–40 cm NCP (P>0.05) (Fig. 3), but the 0–20 cm BD in NM was 13 % lower than that of BM, and the TP and CP in NM were 7 % and 5 % higher than that of BM. No clear pattern was observed for NCP in NM and BM (Fig. 3). Furthermore, the 0–20 cm TN and SOM in NM were much higher than those in BM and reached a significant level at 0–10 cm (P>0.05), whereas the 30–40 cm TN and SOM in NM were lower than those in BM (Fig 3). Similarly, the 0–10 cm TC and the C:N ratio in NM were significantly higher than those in BM, whereas the 30–40 cm TC and the C:N ratio in NM were lower than those in BM (Fig. 3). Additionally, the 0–40 cm MBC and MBN in NM were higher than those in BM and reached a significant level at 0–10 cm (P<0.05) (Fig. 4).

The soil hydrological processes varied between BM and NM (Fig. 5 and Table 1). Given that parameter A fitted by the Gardner model represents the soil water retention (a higher A value indicates higher soil water retention), the soil water content reduced with decreasing matric potential, reduced sharply at high matric potential, and remained stable at low matric potential (Fig. 5). The 0–30 cm layer's soil water content and soil water retention in NM were higher than those in BM, whereas the 30–40 cm layer's soil water content and soil water retention in NM were lower than those in BM (Table 1 and Fig. 6b). Similarly, the 0–10 and 20–30 cm soil water supply capacities (i.e., A⋅B fitted by the Gardner model) in NM were higher than those in BM, while the 10–20 and 30–40 cm soil water supply capacities in NM were lower than those in BM (Fig. 6a). Furthermore, the surface infiltration rate (Ks) in BM was significantly lower than that in NM (Table 1).

Pearson correlation analysis showed that soil water retention was significantly negatively related to BD but significantly positively related to TP, CP, and SOM (Fig. 7a), whereas soil texture exerted weak soil water retention (Fig. 7a). Furthermore, the variance partition showed that SOM explained the greatest variability in soil–soil water retention (24.40 %), followed by CP (21.24 %), BD (18.22 %), and TP (18.22 %) (Fig. 8b), and structural equation modeling showed that the effect of SOM on soil water retention was achieved by altering CP and BD (Fig. 8).

The effects of biocrust on soil properties have been widely explored in previous studies (Guo et al., 2008; Liu et al., 2019). Compared with non-biocrust and most studies conducted in arid regions, the presence of biocrust could improve soil aggregation and stability (Wu et al., 2020), increase soil fertility (Zhou et al., 2020), and reduce soil erosion (Chamizo et al., 2017). In this study, however, we found that the presence of cyanobacteria crust could improve topsoil texture compared with normal meadow but not that of deep soil. The 0–10 cm clay content in cyanobacteria crust meadow was higher than that in normal meadow, whereas the 10–40 cm clay content in cyanobacteria crust meadow was lower than that in normal meadow, which is in line with previous studies conducted in arid and semi-arid regions (H. Li et al., 2016; Wu et al., 2020). The higher clay content in cyanobacteria crust meadow was attributed to the exudation and cohesiveness of the biocrust, which promoted clay and silt formation and reduced sand content (Wang et al., 2021). Furthermore, we found that the 0–20 cm soil bulk density in normal meadow was lower than that in cyanobacteria crust meadow, thereby leading to a higher soil porosity and total capillary porosity in normal meadow. Such a higher soil capillary porosity in normal meadow was attributed to its higher soil organic matter content, which was also confirmed by the significant positive relationship between soil organic matter and soil capillary porosity (Fig. 7) because it has been well documented that higher soil organic matter could improve soil aggregation and stability and subsequently increase soil capillary porosity (Cui et al., 2021).

Moreover, most previous studies indicated that the presence of cyanobacteria crust can also improve soil nutrient conditions in the process of mobile sand fixation (Belnap et al., 2004; Guo et al., 2008; W. H. Li et al., 2005). However, we found that the presence of cyanobacteria crust reduces the 0–10 cm soil total carbon, total nitrogen, and C:N ratio compared with normal meadow, which is in contrast to most previous studies conducted in arid and semi-arid regions (Chamizo et al., 2012a; Zhao et al., 2010). A possible reason for these differences may be the environmental differences. It is well documented that the formation of biocrust is a changing process from simple to complex in its morphology, and the early cyanobacteria crust was formed only under favorable hydrothermal conditions such as temperature, soil water, solar radiation, and nutrient content (Belnap et al., 2004; X.-R. Li et al., 2005). For instance, biocrust is metabolically active when the external environment is wet, and its metabolically active environment is sensitive to temperature (Belnap et al., 2004; X.-R. Li et al., 2005); otherwise, the biocrust may choose to enter the dormant stage when the external environment is under unfavorable conditions. Therefore, compared to the higher soil temperatures in arid and semi-arid lands, the biocrust in alpine ecosystems may be in a dormant stage owing to its lower temperature and less available nutrients. Moreover, the biocrust in our study was mostly dominated by cyanobacteria crust, which was vulnerable to external disturbances such as grazing activity. Thus, the biocrust may choose dormancy when it is subjected to grazing pressure; this evidence was also confirmed by the significantly lower microbial soil carbon and microbial soil nitrogen content in cyanobacteria crust meadow compared with normal meadow (Fig. 4).

We found that soil water infiltration was greatly reduced in cyanobacteria crust meadow compared with that in normal meadow, which was consistent with the results of a previous study conducted in alpine meadows (Y. Li et al., 2016). However, this is in contrast to other studies conducted in cool desert ecosystems where biocrust increased soil water infiltration and reduced runoff by increasing soil porosity and aggregate stability compared with physical crusts and non-crusted bare soils (Kidron and Benenson, 2014; Wei et al., 2015). These discrepancies were associated with soil texture and biocrust developmental stage. In general, soil water infiltration in coarse-textured soils is higher than that in fine-textured soils owing to its large pores compared with the narrow pores in fine-textured soils, which reduces the movement of water into the soil (Belnap, 2006). However, we found that the establishment of cyanobacteria crust increased clay content and subsequently reduced soil macropores, which hindered soil water infiltration. Therefore, we conclude that the soil in cyanobacteria crust meadow may be more vulnerable to runoff generation and consequent soil erosion owing to its lower soil water infiltration and soil water retention capacity. On the other hand, biocrust can reduce available pore spaces for water to infiltrate by clogging the soil surface conductive pores owing to its higher water absorption and swelling of biocrust (Fischer et al., 2010), and it can consequently reduce soil infiltration. In addition, soil water infiltration was also affected by the developmental stage of the biocrust in homogeneous soil. A previous study found that soil hydraulic parameters differed significantly between cyanobacterial biocrust and moss biocrust (Wang et al., 2017). For instance, Chamizo et al. (2012b) reported that the incipient cyanobacterial crust had a lower soil infiltration rate than that of the cyanobacterial crust, whereas the dark-colored mosses' crust had a higher surface soil infiltration capacity by increasing macroporosity and unsaturated hydraulic conductivity in the grasslands (Jiang et al., 2018). In our study, the biocrust was dominated by incipient cyanobacterial crust, which had low biological activity and low porosity owing to the predominance of vesicle pores, thereby leading to a lower soil infiltration rate.

Furthermore, the soil water retention and soil water supply capacity varied significantly between the biocrust and normal meadows. We found that, in 0–10 cm, the soil water retention and soil water supply capacity in normal meadow were higher than those in cyanobacteria crust meadow, which was not in line with previous studies conducted in drylands in which biocrusts enhanced surface soil water retention capacity and water availability (Sun et al., 2022). We speculate that the lower soil water retention in cyanobacteria crust meadow was related to its lower soil organic matter; this evidence was also confirmed by the lower microbial biomass carbon (Fig. 4a). Furthermore, the structural equation model indicated that the effect of soil organic matter on water retention was mainly achieved by altering soil bulk density and soil porosity (Fig. 8) because higher soil organic matter could reduce soil bulk density and thereby increase soil porosity (Liu et al., 2019), leading to higher soil water retention. This result was also confirmed by the significant positive relationship between soil organic matter and soil water retention (Fig. 7), considering that soil organic matter was derived from vegetation litter and root biomass, whereas the vegetation litter in cyanobacteria crust meadow was lower than that in normal meadow owing to its lower aboveground biomass and vegetation coverage, ultimately resulting in lower soil organic matter in cyanobacteria crust meadow.

Grassland ecosystems cover more than 60 % of the QTP and provide important ecosystem services, such as biodiversity conservation, carbon storage, and water conservation (Qian et al., 2021). However, in recent decades, grasslands in the QTP have suffered from serious degradation due to increasing human activity (Dai et al., 2020). Biocrust is an important surface feature of degraded alpine meadows. It is acknowledged that biocrust has a positive effect on soil nutrient and soil water content retention in arid regions. In contrast, we found that the presence of cyanobacteria crust decreased soil water retention and soil infiltration rate, which did not improve water conservation in alpine meadows. Therefore, the soil in the cyanobacteria crust region may be more vulnerable to runoff generation and consequent soil erosion. Moreover, soil nutrients, such as SOM, TC, and TN, were reduced significantly in the cyanobacteria crust meadow, suggesting that the growth of vegetation in the cyanobacteria crust meadow may be limited by soil nutrients. Considering the negative effects of biocrust on alpine meadows, some steps should be taken to reduce the formation of cyanobacteria crust in degraded alpine meadows, such as reducing grazing intensity. Nevertheless, our study results were only obtained by conducting on a site scale, which may not sufficiently extrapolate the whole QTP owing to its high spatial heterogeneity. Thus, a larger scale or more study sites are necessary to draw a generalizability conclusion regarding the effects of biocrust on hydrological processes in alpine meadows of the QTP.