Classifying floods by quantifying driver contributions in the Eastern Monsoon Region of China

Abstract

Understanding the generation mechanisms of floods is important to estimating future flood hazards by improving flood frequency analysis and flood trend interpretation. From the hydrological perspective of flood classification, flood types are determined by the seasonality or magnitudes of hydrological drivers (e.g., rainfall, snowmelt, catchment wetness). Few studies attribute floods to the causative drivers by quantifying the contributions of each driver to the flood peaks. Therefore, we propose the Quantification of Driver Contributions for Flood Classification (QDC-FC) method, which combines a data-driven hydrological model, the Shapley value, which quantifies driver contributions to flood peaks, and a classification tree for dividing floods into five types: long-rain floods (>1 day), rain-on-snow floods, snowmelt floods, short-rain floods (≤1 day), and antecedent-wet floods. In the Eastern Monsoon Region (EMR) of China, the method is applied in 225 catchments where the data-driven model has high Nash–Sutcliffe efficiency (NSE > 0.6) in daily runoff simulation and a high R-squared value (R2>0.6) in flood peak simulation. Under the impact of the eastern monsoon, short-rain floods and long-rain floods are the most frequent types in 143 and 69 catchments, respectively. In 213 catchments, all extreme floods (with return periods larger than 10 years) are caused by short rain or long rain. Antecedent wetness causes more than 20% of floods in 26 out of 66 catchments in northeastern China and northern China with cumulative catchment wetness in summer. Although snowmelt-induced floods occur in April in some catchments of northeastern China, only one catchment has more than 20% of floods caused by snowmelt. As a new method for flood classification, the QDC-FC sheds light on the mixed generation mechanisms of floods and possible drivers of flood trends in the EMR of China.