Increased nonstationarity of stormflow threshold behaviors in a forested watershed due to abrupt earthquake disturbance

Abstract. Extreme earthquake disturbances to the vegetation of local and regional landscapes could swiftly impair the former hydrologic function, significantly increasing the challenge of predicting threshold behaviors of rainfall–runoff processes as well as the hydrologic system's complexity over time. It is still unclear how alternating catchment hydrologic behaviors under an ongoing large earthquake disruption are mediated by long-term interactions between landslides and vegetation evolution. In a well-known watershed affected by the Wenchuan earthquake, the nonlinear hydrologic behavior is examined using two thresholds with intervening linear segments. A lower rising threshold (THr) value (210.48 mm) observed in post-earthquake local landslide regions exhibited a faster stormflow response rate than that in undisturbed forest and grassland–shrubland regions, easily triggering huge flash-flood disasters. Additionally, an integrated response metric pair (integrated watershed average generation threshold THg−IWA and rising threshold THr−IWA) with areas of disparate land use, ecology, and physiography was proposed and efficiently applied to identify emergent catchment hydrologic behaviors. The interannual variation in the two integrated hydrologic thresholds before and following the earthquake was assessed to detect the temporal nonstationarity in hydrologic extremes and nonlinear runoff response. The year 2011 was an important turning point along the hydrologic disturbance–recovery timescale following the earthquake, as post-earthquake landslide evolution reached a state of extreme heterogeneity in space. At that time, the THr−IWA value decreased by ∼ 9 mm compared with the pre-earthquake level. This is closely related to the fast expansion of landslides, leading to a larger extension of variable source area from the channel to neighboring hillslopes, and faster subsurface stormflow, contributing to flash floods. Finally, we present a conceptual model interpreting how the short- and long-term interactions between earthquake-induced landslides and vegetation affect flood hydrographs at event timescale that generated an increased nonstationary hydrologic behavior. This study expands our current knowledge of threshold-based hydrologic and nonstationary stormflow behaviors in response to abrupt earthquake disturbance for the prediction of future flood regimes.