Objective Precipitation is the primary source of soil water replenishment in the loess hilly region, and its abundance or scarcity directly regulates soil water storage as well as vegetation growth and development. However, as a key factor connecting precipitation, soil water, and vegetation utilization, the inherent patterns of rainfall infiltration and the regulatory mechanisms affecting soil water replenishment efficiency remain unclear. This knowledge gap restricts the efficient conversion of regional rainwater resources and the targeted restoration of vegetation ecosystems. By systematically analyzing the rainfall infiltration characteristics in the loess hilly region, this study aims to clarify the relationship between precipitation, infiltration, and soil water. By optimizing the rainfall infiltration process and enhancing soil water replenishment efficiency, it supports regional vegetation restoration and ecological improvement, thereby providing a theoretical basis for the refined utilization of rainwater resources and the optimization of ecological restoration models in this region.

Methods In this study, grasslands converted from cropland under natural field conditions were selected as the experimental plots. Rain shelter devices were used to simulate different precipitation variations. Seven precipitation treatments were established according to the annual precipitation patterns of the study area and the drought classification criteria of the loess hilly region: 60% rainfall increase, 40% rainfall increase, 20% rainfall increase, natural rainfall, 20% rainfall decrease, 40% rainfall decrease, and 60% rainfall decrease. Time domain reflectometry (TDR) was employed for in-situ fixed-point monitoring of soil water content under vegetation cover. Additionally, the HYDRUS model was used to simulate and validate the soil water infiltration process under natural rainfall conditions.

Results The results showed that: 1) In the 10−20 cm soil layer, the average increase in soil water content under rainfall increase treatments (31.39%) was lower than that under natural rainfall (50.17%). In contrast, the average increase under rainfall decrease treatments (83.11%) was significantly greater than that under natural rainfall. The soil water content in the 20-80 cm layer first increased and then decreased with increasing depth, and this trend gradually slowed down below 100 cm. 2) The 40% rainfall increase, 60% rainfall increase, and 20% rainfall decrease treatments significantly increased the soil water content in this region. 3) In the 0-90 cm soil layer, soil water content under both rainfall increase and decrease treatments showed a significant positive correlation with rainfall amount (P < 0.05). However, in the 90-180 cm soil layer, there was no significant correlation between soil water content and rainfall amount (P > 0.05). 4) Under different precipitation treatments, the simulated and measured soil water contents showed consistent trends over time during individual rainfall events, both increasing rapidly after rainfall and then gradually decreasing. The model validation using the root mean square error (RMSE), Nash-Sutcliffe efficiency coefficient (NSE), and correlation coefficient (R) indicated that the overall simulation results were highly accurate.

Conclusions The findings of this study can provide theoretical support and practical recommendations for vegetation restoration and soil and water conservation in the loess hilly region.