The temporal stability of soil moisture (TSSM) was widely applied to optimize the soil moisture sampling scheme on a catchment or even larger spatial scale over wet and dry observational periods. However, the integration of the TSSM feature with specific hydrological response at a small plot scale has not been sufficiently researched. This study analyzed the temporal stability of surface soil moisture (0–10cm) characteristics and corresponding influencing factors of different vegetation types under two typical soil moisture changing processes including wet-to-dry (WTD) and dry-to-wet (DTW), and determined the representative points. A total of 16 microplots (60×60cm each) that were composed of three vegetation types containing Andropogon, Artemisia scoparia and Spiraea pubescens and bare land cover were selected. And the soil moisture in the central point (CP) and four ambient points (APs) of each microplot were measured during the WTD and DTW processes. The results showed that, 1) from DTW to WTD processes, the distribution of the soil water content in different vegetation types indicated a significant difference. Compared with the soil moisture in the AP or CP area of other vegetation types, the soil water content in tall shrub types (S. pubescens) was the lowest. 2) The autocorrelation coefficient indicated that both in the AP and CP areas, the soil moisture of the low shrub types (A. scoparia) had a higher temporal stability than that of other vegetation types. However, the soil water content in bare land had the highest temporal fluctuation from the DTW to WTD processes. Additionally, in the CP area, the TSSM of all the vegetation types tended to decrease during the WTD process. 3) Based on the TSSM analysis system that was derived from the principle of probability and statistics, the soil moisture in the low shrub types (A. scoparia) most likely provides the best representativeness of the spatial average soil water content of heterogeneous vegetation types. The determination of the representative soil moisture point via the hydrological-trait sampling method could be supplementary and significant for a TSSM study of the available soil water resources in an arid and semi-arid ecosystem.