Soil moisture is key in maintaining surface energy balance and the hydrological cycle. However, the patterns of soil moisture change at different altitudes and their response to rainfall are not fully understood. The aim of this study is to analyze the changes in soil moisture at different altitudes in the Mu Us Sandy Land, assess the soil moisture response to various rainfall events, and evaluate the factors affecting this response. To achieve this, we set up two soil moisture and precipitation monitoring systems in the Ejin Horo Banner, located at the northern edge of the Mu Us Sandy Land. Results show that the annual precipitation at the site was 342.6 mm, with most rain falling between July and September. The region experiences soil drought from June to October. At the lower X1 test site, the increase in soil moisture in the 0–90 cm layer during rainfall was greater than the amount of precipitation, suggesting that lateral water flow from higher areas influences soil moisture at lower positions. This study used Grey Relational Analysis to identify the key factors influencing soil moisture changes; the result showed that the initial soil moisture and total precipitation have the strongest correlation with changes in soil moisture, making them the main factors controlling soil moisture response at the site. These findings enhance our understanding of soil moisture dynamics and provide a foundation for vegetation restoration in semi-arid sandy areas and ecosystem rehabilitation.

Accurate daily runoff prediction is essential for flood control, reservoir operation, and scientific water resources management. However, runoff processes are increasingly affected by climate change and human activities, leading to pronounced nonlinearity and nonstationarity that limit the performance of single data-driven models. This study aims to improve the reliability and hydrological credibility of daily runoff prediction by systematically evaluating recurrent neural network (RNN) structures and explicitly modeling prediction residuals. Three commonly used RNN architectures—long short-term memory (LSTM), gated recurrent unit (GRU), and bidirectional long short-term memory (BiLSTM)—are systematically compared for daily runoff prediction in the Andi Reservoir watershed under identical hydrometeorological conditions. Based on the comparative results, BiLSTM is selected as the base model to capture dominant temporal dependencies. To further address systematic prediction errors under complex hydrological conditions, a residual-learning framework is constructed by integrating BiLSTM with extreme gradient boosting (XGBoost), in which XGBoost is employed to model and correct the nonlinear residuals of BiLSTM predictions. In addition, the Shapley Additive Explanations (SHAP) method is applied to interpret the contributions of input variables and to examine the learning mechanisms of both the base model and the residual-correction stage. Results indicate that BiLSTM performs better than LSTM and GRU for daily runoff prediction and that residual correction using XGBoost further enhances prediction accuracy and robustness, particularly under nonstationary conditions and peak-flow scenarios. The contribution of this study lies in providing a systematic modeling framework that combines model comparison, residual learning, and interpretability analysis to support more reliable daily runoff prediction in complex watersheds.