Objective  The dried soil layer (DSL) is a special phenomenon of soil moisture deficiency in the Loess Plateau. In order to reveal the spatial-temporal distribution of soil water storage (SWS), sandwiched dried soil layer (SDSL) and DSL, and to predict their status.

  Method  We layouted 15 observational sites in a typical fragmented area in Liudaogou catchment and obtained long-term data of soil moisture from October, 2014 to July, 2017. Using temporal stability method, we estimated SWS, the thickness of DSL and DSL-SWC with the representative sites, and verified the reliability of the simulation equations.

  Results  The results show that: ① The dynamics of SDSL was characterized with somewhat complexity in the fragmented terrain. The level of SDSL was severe, the thickness and soil moisture of sandwiched dried soil layers varied within the range of 20-760 cm and 6.43%-10.73%, respectively. ② The distribution of SWS and DSL were temporally stable. The Spearman rank correlation coefficients of SWS, DSLT and DSL-SWC were 0.94, 0.95 and 0.90, respectively, which indicating that the characteristics of permanent SWS and DSL. ③The prediction results of SWS and DSLT were more accurate than DSL-SWC. Based on the analysis of the time stability characteristics of SWS and dried soil layers, the A5, C1 and C2 observation points could better represent the average levels of SWS, the thickness of DSL and DSL-SWC in the study area (coefficients of determination were 0.88, 0.74 and 0.56), respectively.

  Conclusion  It was feasible to predict the spatial-temporal distribution of soil moisture by using representative sites in the fragmented area of the Loess Plateau. Scientific regulation of soil moisture could alleviate the formation and development of dried soil layer to a certain extent, and provide the possibility for the restoration of dried soil layer.