Spatial variability and forecast of soil water in the ultra‐deep loess profile across a south–north transect of the Chinese Loess Plateau.

Knowledge of the spatial variability and forecast of soil water in the ultra‐deep (>21 m) loess profile are important for understanding the chemical, physical, and biological processes in the CZ. In this study, we regularly monitor soil water content (SWC) in deep soil profile along regional transect on the Loess Plateau. Descriptive statistical analysis found that the coefficient of variation (CV) of mean SWC in Ansai and Shenmu were 16.036% and 13.606%, respectively, indicating moderate variability. The CV of mean SWC in Yangling, Changwu, and Fuxian were 4.111%, 7.951%, and 6.117%, respectively, showing low variability. Geo‐statistical analysis indicated that mean SWC showed strong spatial dependence. Wavelet analysis showed that the approximative trend of mean SWC in five sampling sites showed an increased trend along depth series. In addition, a good fit line equation (R2 = 0.329) was established by using observed values and ANN‐forecast of three sites (Yangling, Fuxian, and Shenmu). And the RMSE values of five sample sites, respectively, were 1.302, 4.546, 2.662, 6.231, and 4.293. This study fills the gap in research' ultra‐deep (>21 m) soil water changes and forecast, evidence from soil borehole data. At the same time, our research provides valuable information for the vegetation restoration and modelling of deep soil water. [ABSTRACT FROM AUTHOR]