Different regions exhibited different patterns for ΔSM_{p_0-10}, ΔSM_{p_10-20}, ΔSM_{a_0-10}, and ΔSM_{a_10-20} ( Figure 1 ). Spatially mean values of ΔSM_{p_0-10}, ΔSM_{p_10-20}, ΔSM_{a_0-10}, and ΔSM_{a_10-20} were −0.01 (from −0.46 to 0.54 g g⁻¹ year⁻¹), −0.06 (from −0.42 to 0.27 g g⁻¹ year⁻¹), 0.02 (from −0.42 to 0.51 g g⁻¹ year⁻¹), and −0.02 g g⁻¹ year⁻¹ (from −0.56 to 0.49 g g⁻¹ year⁻¹), respectively. There were 39.97%, 22.29%, 56.60%, and 34.71% areas showing increasing trends for ΔSM_{p_0-10}, ΔSM_{p_10-20}, ΔSM_{a_0-10}, and ΔSM_{a_10-20}, respectively. In contrast, there were 60.03%, 77.67%, 43.40%, and 65.29% areas showing decreasing trends for ΔSM_{p_0-10}, ΔSM_{p_10-20}, ΔSM_{a_0-10}, and ΔSM_{a_10-20}, respectively. Longitude had higher influences on ΔSM_{p_0-10} and ΔSM_{a_0-10} than latitude and elevation ( Figure S1 ). In contrast, latitude had higher influences on ΔSM_{p_10-20} and ΔSM_{a_10-20} than longitude and elevation ( Figure S2 ). Compared to GST and GSRad, GSP had higher influences on ΔSM_{p_0-10}, ΔSM_{p_10-20}, and ΔSM_{a_10-20} ( Figures S3 , S4 ). In contrast, compared to GST and GSP, GSRad had a greater effect on ΔSM_{a_0-10} ( Figure S3 ). The ΔGSP had the greatest effects on ΔSM_{p_0-10} and ΔSM_{a_0-10}, and the ΔGST had the lowest effects on ΔSM_{p_0-10} and ΔSM_{a_0-10} among the three variables of climate change ( Figure S5 ). In contrast, compared to ΔGST and ΔGSP, ΔGSRad had higher influences on ΔSM_{p_10-20} and ΔSM_{a_10-20} ( Figure S6 ). The ΔGST, ΔGSP, and ΔGSRad had higher influences on ΔSM_{p_10-20} than ΔSM_{p_0-10} ( Figures S5 , S6 ). The overall effect of the three variables of climate change on the ΔSM_{p_10-20} was greater than that on ΔSM_{p_0-10} ( Figure S7 ). The ΔGSP had greater exclusionary influences on ΔSM_{p_0-10} and ΔSM_{a_0-10}, but the ΔGSRad had higher exclusionary influences on ΔSM_{p_10-20} and ΔSM_{a_10-20} among the three variables of climate change ( Figures S7 , S8 ). The GSNDVI and ΔGSNDVI significantly affected ΔSM_{a_0-10} and ΔSM_{a_10-20} ( Figure S9 ). Climate change had the greatest exclusionary effects on ΔSM_{p_0-10}, ΔSM_{a_0-10}, ΔSM_{p_10-20}, and ΔSM_{a_10-20} than geographical position and mean climate conditions ( Figure 2 ). Mean climate conditions had higher exclusionary influences on ΔSM_{p_0-10} and ΔSM_{p_10-20} than geographical position ( Figure 2 ). Mean climate conditions and geographical position had the same exclusionary effects on ΔSM_{a_0-10} and ΔSM_{a_10-20} ( Figure 2 ).

Spatially mean values of R _{SM_0-10} and R _{SM_10-20} were 0.64 (from 0.28 to 1.45) and 0.79 (from 0.33 to 1.97), respectively ( Figure 3 ). There were only 2.82% and 30.03% areas where R _{SM_0-10} and R _{SM_10-20} values were greater than one, respectively. In contrast, there were 97.18% and 69.97% areas where R _{SM_0-10} and R _{SM_10-20} values were lower than one, respectively. Longitude had the strongest effects on R _{SM_0-10} and R _{SM_10-20}, but latitude had the least effects on R _{SM_0-10} and R _{SM_10-20} among the three geography variables ( Figure S10 ). The GSP had the strongest effects on R _{SM_0-10} and R _{SM_10-20}, but the GST had the least effects on R _{SM_0-10} and R _{SM_10-20} among the three variables of mean climate conditions ( Figure S11 ). Compared to the three variables of mean climate conditions, the GSNDVI had higher influences on R _{SM_0-10} and R _{SM_10-20} ( Figure S11 ). The ΔGSP had the strongest effects on R _{SM_0-10} and R _{SM_10-20}, but the ΔGSRad had the least effects on R _{SM_0-10} and R _{SM_10-20} among the three variables of climate change ( Figure S12 ). The ΔGSNDVI also had some effects on R _{SM_0-10} and R _{SM_10-20} ( Figure S12 ). Longitude had the greatest exclusionary effects on R _{SM_0-10} and R _{SM_10-20}, but elevation had the least exclusionary effects on R _{SM_0-10} and R _{SM_10-20} among the three geography variables ( Figure S13 ). Compared to the three variables of mean climate conditions, the GSNDVI had the greater exclusionary influences on R _{SM_0-10} and R _{SM_10-20} ( Figure S13 ). The GST had the greatest exclusionary effects on R _{SM_0-10} and R _{SM_10-20}, but the GSRad had the lowest exclusionary effects on R _{SM_0-10} and R _{SM_10-20} among the three variables of mean climate conditions ( Figure S13 ). The ΔGST had the strongest exclusionary influences on R _{SM_0-10}, but the ΔGSP had the strongest exclusionary influences on R _{SM_10-20} among the three variables of climate change ( Figure S13 ). The ΔGSRad had the lowest exclusionary effects on R _{SM_0-10} and R _{SM_10-20} among the three variables of climate change ( Figure S13 ). The ΔGSNDVI also had some exclusionary effects on R _{SM_0-10} and R _{SM_10-20} ( Figure S13 ). Mean climate conditions + GSNDVI had the greatest exclusionary effects on R _{SM_0-10} and R _{SM_10-20}, but the geographical position had the lowest exclusionary effects on R _{SM_0-10} and R _{SM_10-20} ( Figure 4 ).

Spatially mean values of ΔR _{SM_0-10} and ΔR _{SM_10-20} were 0.0011 (from −0.0287 to 0.0263) and 0.0013 (from −0.0590 to 0.0564), respectively ( Figure 3 ). There were 63.78% and 53.36% areas showing increasing trends for ΔR _{SM_0-10} and ΔR _{SM_10-20}, respectively. In contrast, there were 36.22% and 46.64% areas showing decreasing trends for ΔR _{SM_0-10} and ΔR _{SM_10-20}, respectively. The three geography variables and GSNDVI had some effects on ΔR _{SM_0-10} and ΔR _{SM_10-20} ( Figures S14 , S15 ). The GSP had the strongest effects on ΔR _{SM_0-10} and ΔR _{SM_10-20}, but the GST had the lowest effect on ΔR _{SM_0-10} and ΔR _{SM_10-20} among the three variables of mean climate conditions ( Figure S15 ). Compared to the three variables of climate change, the ΔGSNDVI had the higher influences on ΔR _{SM_0-10} and ΔR _{SM_10-20} ( Figure S16 ). Compared to the three variables of climate change, the ΔGSNDVI had the higher exclusionary influences on ΔR _{SM_0-10} and ΔR _{SM_10-20} ( Figure S17 ). Climate change + ΔGSNDVI had the greatest exclusionary effects on ΔR _{SM_0-10} and ΔR _{SM_10-20}, but the geographical position had the lowest exclusionary effects on ΔR _{SM_0-10} and ΔR _{SM_10-20} ( Figure 4 ).

Inconsistent with the hypothesis, this study implies that climate change resulted in soil drying trends across the entire TP grassland region during the past 20 years. The finding is likely attributed to the elevated evapotranspiration triggered by climate warming offsets and even exceeds the water increase triggered by increased precipitation (Berg et al., 2017). These findings contradict previous studies that suggested a general increase in wetness across the TP grasslands in recent years (Lu and Liu, 2010; Diffenbaugh and Field, 2013). Moreover, this study also showed that the proportion of soil drying (≥43.40%) outweighed the proportion of drought trend based on precipitation (22.57%) in the TP grassland regions during the last 20 years. These discoveries indicated that previous assessments of climate humidity trends based solely on precipitation may have underestimated the extent of drying. It is important to note that precipitation does not always effectively support plant growth, and SM may be more associated with plant growth (Fu et al., 2018; Rigden et al., 2020).

Climate warming, changes in precipitation, and radiation can significantly affect SM in the TP grasslands. This finding is consistent with previous research (Yu et al., 2019; Fu and Shen, 2022; Han et al., 2023). Notably, the exclusionary effects of radiation change on SM emphasize the necessity of considering the impacts of radiation change on grassland ecosystems in addition to climate warming and decreased/increased precipitation, particularly on the TP. These exclusionary effects of radiation change on SM also align with some previous findings that demonstrated its notable impact on vegetation carbon use efficiency, plant physiological characteristics, forage nutritional quality, and storage on the TP (Fu et al., 2016; Fu et al., 2022; Li and Fu, 2023).

This study shows that climate change had non-linear effects on SM in the TP grassland areas. This aligns with previous studies (Shen et al., 2015; Fu et al., 2018) and can be attributed to several underlying mechanisms. First, it is essential to consider that SM holding capacity is not infinite and that there exists a state of soil saturation where the soil cannot hold more water. Although increased precipitation can initially lead to an increase in SM, the point of saturation will eventually be reached, and further increases in precipitation will not result in additional gains in SM (Shen et al., 2015; Fu et al., 2018). The level of soil saturation is largely influenced by soil porosity, which, in turn, can be altered by factors such as plant growth and soil organic matter decomposition. These processes are directly or indirectly influenced by climate change (Fu et al., 2018; Zhang et al., 2022). Second, the influences of climate change on SM can be closely associated with the magnitude of climate change (Xu et al., 2013), local climate conditions (Fu and Sun, 2022), and geographical location (Zhang et al., 2021). Third, the influences of climate change on SM depend on their relative intensity of the influences of climate warming, increased/decreased precipitation, and dimming/brightening on SM. However, it is crucial to note that the influences of these factors on SM are not simply additive or subtractive effects in nature (Yu et al., 2019; Xiao et al., 2023).

This study shows that climate change had stronger negative effects on ΔSM_{p_10-20} than ΔSM_{p_0-10}. This is similar to previous studies (Rui et al., 2011; Berg et al., 2017; Dai et al., 2022), and several potential reasons contribute to this phenomenon. First, the dominant climate change factors affecting SM were different between soil depths under climate change scenarios ( Figure S7 ). Second, the influences of climate change on SM were related to local water availability and warming magnitude (Rui et al., 2011; Xu et al., 2013). Both local SM and increase in soil temperature generally decrease with the increasing soil depth (Rui et al., 2011). Third, plant roots interspersed in soil pores draw water from water-bearing soil pores and generally decline with increasing soil depth (Berg et al., 2017). Although climate warming generally stimulates plant root growth and increases root biomass, such a boosting effect varies with soil depth (Fu et al., 2015).

This study shows that climate change resulted in the spatial relocation of SM across the TP grassland areas. This finding is consistent with previous studies (Wang et al., 2021a; Fu and Sun, 2022) and can be explained by one or more of the following reasons. First, significant spatial variations in the direction and amplitude of climate change have directly led to the spatial relocation of air temperature, precipitation, radiation, and wind (Huang and Fu, 2023). Moreover, because of such spatial heterogeneity of climate change, soil temperature and vapor pressure deficit can indirectly be relocated (Wang et al., 2021b; Han et al., 2022). Increases in wind speed, soil temperature, and vapor pressure deficit can generally elevate evapotranspiration, leading to a reduction in SM (Xu et al., 2013; Fu et al., 2018). Second, different species or function groups of plants may have diverse physiological water requirements, root growth depths, and phyllosphere microbes, indicating the change in plant growth, α-diversity, and community composition can also affect SM (Lian et al., 2020; Huang et al., 2023). Climate change can result in spatial relocation of plant growth and community structure (Wang et al., 2021a; Fu and Sun, 2022; Huang and Fu, 2023).

In line with the hypothesis, this study shows that human activity strengthened the sensitivities of SM at 0–10 cm but weakened the sensitivities of SM at 10–20 cm to climate change. This is similar to previous studies (Rui et al., 2011; Zhang et al., 2020), and the underlying mechanism for this phenomenon is likely rooted in the fact human activity may have the capacity to either amplify or attenuate the influences of climate change on critical factors, including soil temperature, vapor pressure deficit, plant growth, and community structure (Wang et al., 2012; Fu and Shen, 2016; Li and Fu, 2023).

Furthermore, this study demonstrated that the influence of human activity on SM was related to soil depth. This is in agreement with previous studies (Sun et al., 2019). Several factors may account for this relationship. First, the growth of the human population has led to an increasing demand for groundwater, which serves as a vital supply of SM and is more proximate to the 10–20 cm soil layer. Second, the effects of human activity on soil temperature, bulk density, and root density are dependent on soil depth (Rui et al., 2011; Sun et al., 2019; Lai and Kumar, 2020).

In addition to soil depth, our study also indicated that the influence of human activity on SM varied from year to year ( Figure 3 ), which aligned with findings from some previous studies (Sun et al., 2021; Zhang and Fu, 2021). This variability is likely attributable to at least one of the following factors. First, human activity can have cumulative influences on plant growth and community structure, etc. (Fu and Shen, 2016; Zhao et al., 2016; Gao and Carmel, 2020). Second, climate conditions can regulate the influence of human activity on SM ( Figure S11 ), and climate conditions vary from year to year. Third, the intensity and extent of human activity also varied from year to year.

This study shows that human activity resulted in the spatial relocation of SM across the TP grassland ecological systems. This is in agreement with some previous studies (Zhang et al., 2020; Zhang and Fu, 2021). The underlying mechanisms for this phenomenon are likely rooted in at least one of the next causes. First, both the scope and intensity of human activity are spatiotemporal heterogeneous (Sun et al., 2019; Zha et al., 2022). For example, warm- and cold-season grazing may have diverse influences on SM (Sun et al., 2021; Zhang and Fu, 2021), and they exhibit spatial heterogeneity. Second, human activity may result in the spatial relocation of vapor pressure deficit, soil temperature, porosity, and compactness (Fu et al., 2012; Fu and Shen, 2016). Third, influences of human activity on SM were related to plant growth ( Figures S11 , S12 ), and human activity may result in spatial relocation of plant growth (Fu et al., 2022; Zha et al., 2022; Li and Fu, 2023). Moreover, human activity can also result in spatial relocation of plant α-diversity and community composition in TP grassland ecological systems (Sun et al., 2021; Zha et al., 2022).