Initial Soil Moisture Threshold Influences Rainfall–Runoff Processes of Mixed-Forest Catchment: A Case Study of the Dabie Mountains

Initial soil moisture (ISM) represents the soil moisture condition before a rainfall event and is an important factor associated with rainfall and runoff responses in forested catchments. However, it remains unclear whether this nonlinearity can be characterized by distinct ISM thresholds and whether the dominant controls of runoff coefficient (RC) and specific peak discharge (SPD) differ between pre-threshold and post-threshold ISM states. To address this issue, this study analyzed 27 rainfall events monitored in mixed-forest catchments of the Dabie Mountains from 2024 to 2026. A random forest model combined with SHapley additive exPlanations (SHAP) and partial dependence plots (PDPs) was used to simulate RC and SPD, identify ISM thresholds, and compare state-dependent dominant factors and interactions. The results showed that the SHAP-derived ISM thresholds were 18.3% for RC and 19.2% for SPD. For RC, ISM remained the dominant factor in both ISM states. In the pre-threshold state, rainfall intensity (RI) and net precipitation (NP) were the main secondary factors, with RI × ISM and NP × ISM as the leading interactions. In the post-threshold state, temperature and evaporation (Evap) became more important, and the leading interactions shifted to Evap × ISM and ISM × atmospheric pressure. For SPD, the dominant factors in the pre-threshold state were RI, NP, and ISM, with NP × RI and RI × ISM as the main interactions. In the post-threshold state, ISM became the leading factor, wind speed and RI also remained important, and ISM × wind speed and RI × ISM were the main interactions. These findings indicate that event-scale rainfall–runoff responses in mixed-forest catchments are strongly dependent on ISM state, and that ISM thresholds can provide useful boundaries for distinguishing state-dependent runoff-control patterns.