Soil moisture has strong spatial heterogeneity in both vertical and horizontal directions.Considerable interests have been devoted to the assessment of horizontal direction whereas less of that are known in the vertical variation.In this study,we used classical statistics,geo-statistics and fractal methods to characterize the temporal-spatial variability of 6 m depth profile soil moisture in the vertical direction,on the typical slope land on the Loess Plateau. Profile soil moisture at different measurement times followed a log-normal distribution and showed a moderate variability.Soil moisture increased with the increasing of soil depth and behaved temporal stability during measurement periods. Profile soil moisture demonstrated a strongly spatial dependence as reflected by the nearly none nugget ratio,which did not change with measurement time,and the experimental variogram of soil moisture was fitted well with an exponential model.Profile soil moisture at different measure times showed a limited selfsimilarity at an entire study scale.It has a stable self-similarity at distances less than 2.00 m and weak self-similarity at distances larger than 2.00 m.The value of fractal dimension was highly related to the main processes of control soil moisture.At each measurement time,the spatial auto-correlation of the profile soil moisture was the same,that is,it demostrated spatially positive correlation at distances less than 3.20 m whereas spatially negative correlation at distances larger than 3.20 m.The profile soil moisture at different measure times showed a significantly positive correlation.