基于超定量取样的场次洪水自动分割方法OA

In hydrological forecasting,the accurate identification and segmentation of independent flood events from continuous hydrological processes are crucial for ensuring parameter accuracy and improving flood forecast precision.Traditional manual methods for selecting flood events suffer from low efficiency and a lack of universal standards,leading to strong subjectivity in flood event extraction.This paper proposes a simple method for automatic identification of flood events from continuous flow process.The method fully considers the attributes and characteristics of flow data itself,and determines the flood peak and starting and ending points of flood events based on conditions such as flood peak flow threshold,start-stop flow threshold,rising slope threshold.The uniqueness of floods was ensured by rationality test,and the influences of peak flow threshold,start-stop flow threshold and rising slope threshold on flood segmentation were further analyzed.Taking the Muma River Basin as the study area,the proposed method and other flood segmentation methods were applied to segment the continuous flow process from 1980 to 1990.The impact of different flood segmentation methods on forecast accuracy was evaluated by employing the XAJ Model and the TOPMODEL for hydrological forecasting.Results indicate that the flood peak flow threshold controls the number of floods and the duration of floods,the start-stop flow threshold and rising slope threshold mainly affects the duration of floods,while the influence of the rising slope threshold being relatively limited.The proposed method can recognize the flood segmentation process in long series hydrological data based on the objective judgment criteria,and quickly split the flood segmentation of different flow processes with simple principle,high computational efficiency and high accuracy.Compared with other flood segmentation methods,it achieves more precise event segmentation,higher forecasting accuracy,and lower peak error,providing relatively accurate input data for flood forecasting research.