Spatial and temporal variability of 0- to 5-m soil–water storage at the watershed scale

Dynamic relationships among rainfall patterns, soil water distribution, and plant growth are crucial for sustainable conservation of soil and water resources in water-limited ecosystems. Spatial and temporal variation in deep soil water content at a watershed scale have not yet been characterized adequately due to the lack of deep soil water data. Deep soil–water storage (SWS) up to a depth of 5 m (n = 73) was measured at 19 sampling occasions at the LaoYeManQu watershed on the Chinese Loess Plateau (CLP). At a depth of 0–1.5 m, the annual mean SWS was highly correlated with rain intensity, and the correlation decreased with depth, but within the layers at 1.5–5.0 m, the changes in SWS indicated a lag between precipitation and the replenishment of soil water. Geostatistical parameters of SWS were also highly dependent on depth, and the mean SWS presented similar spatial structures in two adjacent layers. Temporal stability of SWS as indicated by mean relative difference, standard deviation of the relative difference (SDRD), and mean absolute bias error (MABE) was significantly weaker at the shallow than at deeper layers. Soil separates and organic carbon content controlled the spatial pattern of SWS at the watershed scale. One representative location (Site 57) was identified to estimate the mean SWS in the 1- to 5-m layer of the watershed. Semivariograms of the SDRD and MABE were best fitted by an isotropic spherical model, and their spatial distributions were depth-dependent. Both temporal stability and spatial variability of SWS increased over depth. This study is helpful for deep SWS estimation and sustainable management of soil and water on the CLP, and for other similar regions around the world.