Depths of soil water depleted by plants and of infiltration after precipitation under eight typical vegetation types were studied in the Liudaogou watershed, which is located in the water-wind erosion crisscross region. Results showed that soil moisture under land of bare, farmed, abandoned, planted with grasses or shrubs (alfalfa, Caragana korshinskii and Astragalus adsurgens), or typical of indigenous grasslands (wild grasses and Sitpa bungeana) was reduced in all cases during the normal (2010) and dry (2011) years; soil moisture in some plots increased during the wet year (2012). In the normal and dry years, depths of soil moisture depleted by plants were in the order: C. korshinkii > abandoned > A. adsurgens > alfalfa ≈ S. bungeana ≈ wild grass > crop > bare, while that of the depth of infiltration was crop > bare > abandoned > wild grass > S. bungeana > A. adsurgens > alfalfa > C. korshinkii. The depth of soil water depletion was always deeper than that of infiltration. In the year of high rainfall (2012), the water storage balance was positive in bare soil, alfalfa, wild grass and A. adsurgens, but soil moisture under C. korshinkii was depleted between the soil surface and the 260 cm depth, and to a lesser extent under the other vegetation types in the order of abandoned > crop > S. bungeana; depths of soil water infiltration after precipitation were: crop > bare > abandoned = C. korshinkii > wild grass = alfalfa > S. bungeana > A. adsurgens. The depths of soil water depletion and infiltration occurred within the studied 0-120 cm layer, and soil water was replenished in the wet year. Soil water responses to vegetation-soil interactions under farmland were similar to its behavior in bare soil, but the depths of soil water depletion and infiltration were deeper. The depths of soil water depletion and infiltration under abandoned land were determined by the vegetation; the depth of soil water depletion tended to be deeper with increases in vegetation abundance but the soil water depletion depth was less than the infiltration depth. Not only the infiltrating rainwater but also the stored soil water had been consumed rapidly during the managed land conversion from farmland to grassland or shrubland. As a result, we might get a false impression that the depths of soil water depletion were shallower from our observations of the studied soil layers because of the lowered soil water content of the layers below 120 cm. Plots with planted grasses and shrubs, which had high biomasses, also had high evapotranspiration and deeper depths of soil water depletion. Consequently, we should fully consider the annual exchange depth of soil water during the management of land conversion from farmland to grassland or shrubland, and then take measures to reduce the depths of depletion of soil water and to increase infiltration.