Multi-scale time-series framework for piping-induced levee failure using a piping potential index

River levee failures can occur from not only overtopping but also piping, a progressive internal erosion mechanism driven by sustained hydraulic gradients and accumulated moisture. This study proposes a time-series framework integrating river level, groundwater level, rainfall, and Soil Moisture Active Passive (SMAP) root-zone soil moisture to investigate six piping-induced levee failures in South Korea since 2020. The Piping Potential Index (PPI) is defined as the sum of the standardized indicators that represent the driving head difference (D), accumulated rainfall (R), time-averaged soil moisture (SM), and incremental groundwater rise (GW). Hydro-environmental data were reconstructed at hourly resolution over a −14 to +1 day event window, and 27 accumulation-window combinations were evaluated using lead time, duration, and prominence metrics. Short-term rainfall and soil moisture windows (24 h) produced stronger and earlier anomaly peaks, whereas longer rainfall windows (72 h) prolonged the persistence of critical conditions. A configuration integrating short-term rainfall and soil moisture with medium-term groundwater accumulation demonstrated a balanced trade-off between early warning performance and signal strength, highlighting the multi-scale nature of piping-related levee instability.