Based on soil moisture and meteorological elements data at four pasture observation stations in typical steppe of Inner Mongolia from 1981 to 2015, the correlations between multi-time-scales meteorological drought indexes and soil relative moisture at different depths in spring, summer and autumn were comparatively analyzed, and the applicability of these drought indexes to drought monitoring was explored in typical steppe of Inner Mongolia. And on this basis, the prediction models of soil relative moisture at different depths were built by using multiple linear regression method in different seasons. The results show that the soil relative moisture at 0-20 cm depth in three seasons was affected by atmospheric water balance in the past two months, while the timescales influencing soil relative moisture at 0-50 cm and 0-100 cm depths were different in typical steppe of Inner Mongolia. In spring, the soil relative moistures at 0-50 cm and 0-100 cm depths were significantly regulated by previous annual precipitation. In summer, the soil relative moisture at 0-50 cm depth was closely related to atmospheric moisture balance in the past two months, while the soil drought at 0-100 cm depth was mainly controlled by precipitation deficit in the past 2-6 months. In autumn, the soil relative moisture at 0-50 cm depth was significantly influenced by previous precipitation for 3-6 months, while that at 0-100 cm depth was closely correlated with the balance between precipitation and evapotranspiration in the past three months, and the influence of meteorological droughts in the past six months and above was obvious. CI, MCI and PDSI had higher correlation with soil relative moisture at different depths than other meteorological drought indices in different seasons due to taking into account the comprehensive effects of long-term and short-term atmospheric water deficit. The established prediction models of soil relative moisture based on previous meteorological drought indexes could better capture the change of soil moisture in typical steppe, and they could support the drought monitoring and predication of pasture to some extent.

Adequate understanding of water stress influences on the temporal-spatial variations of soil moisture is crucial for studying water consumption and drought identification in grasslands. Based on soil volumetric water content data monitored in multi-layer soil under two water supply conditions during 2016-2017, soil water status, the seasonal variations and the vertical distribution in the typical steppe in Xilinhot of Inner Mongolia were investigated. The results show that soil water showed distinct seasonal variations in a normal year with dual peaks curve. That was, soil volumetric water content reached the peak in the spring green-up and the rainy season (July), and then it decreased and reached the minimum value in the late of May and the early of September. Soil water content of top 50 cm layer also exhibited marked variation, especially for 0-30 cm layer. The water stress changed the seasonal dynamics of soil moisture (i.e. reduction in the amplitude). More importantly, serious water stress caused that the dual peaks curve failed (becoming smooth), and the water consumption happened in the deeper layer. Moreover, the sensitive area of soil water change extended to the 60-80 cm soil layer. The results highlight that seasonal variations in soil moisture and soil depth are two valuable factors analyzing the interactions between water stress and ecosystem as well as soil drought index.