Performance Evaluation of Taiwan’s Rainfall-Based Debris Flow Warning System: A 20-Year Operational Analysis (2005–2025)

Taiwan is highly susceptible to frequent debris flow disasters due to its steep terrain, fragile geological structures, and intense rainfall from typhoons. To mitigate these risks, the Agency of Rural Development and Soil and Water Conservation (ARDSWC) under the Ministry of Agriculture (MOA) established a rainfall-based debris flow warning system in 2005, adopting the critical rainfall model proposed by Jan (2004). This system utilizes Effective Accumulated Rainfall (EAR)—combining current event rainfall and 7-day antecedent rainfall—as the primary indicator for debris-flow warning. Through statistical analysis of historical rainfall events, specific warning criteria have been set for townships prone to debris flows across Taiwan, classified into 9 levels ranging from 200 to 600 mm at 50-mm intervals. Consequently, warning issuance is triggered when the EAR exceeds these designated criteria. Based on a dataset of 252 officially recorded debris flow events and their associated warning records spanning from 2005 to 2025, this study assesses the effectiveness of warning operations using four performance indices: Capture Rate of Warning Issuance (C1, reflecting operational effectiveness), Capture Rate of Warning Criteria (C2, evaluating the appropriateness of criteria setting), Disaster Occurrence Rate within Issued Warnings (C3, indicating occurrence probability in warned areas), and Disaster Warning Coverage (C4, evaluating if events occurred within designated potential zones). The results show that both C1 and C2 achieved 67.5%, C3 was 8.4%, and C4 reached 90.8%. Overall, the debris flow warning system has proven to be operationally mature and highly effective in disaster mitigation. Approximately 70% of debris flow events were preceded by warnings that facilitated preemptive resident evacuation. Since the implementation of the system, casualties have been significantly reduced, notably with zero debris flow-related fatalities recorded over the past decade. Furthermore, the majority of events occurred within identified potential risk zones. Nevertheless, facing climate-change-induced extreme rainfall and post-seismic slope instability, continuous optimization of warning protocols and criteria remains essential to ensure robust early warning capabilities.

Keywords: debris flow, early warning system, performance evaluation