Author(s): Wang Xuemei; Liu Ronghua; Zhai Xiaoyan; Guo Liang

Linked Author(s):

Keywords: Design flood; Mountainous catchment; Random forest algorithm; Terrain and underlying surface characteristic; Rainfall characteristic

Abstract: Design flood provides powerful technical supports for carrying out flash flood prevention and control in mountainous catchments. Based on China National Flash Flood Disasters Investigation and Evaluation, the design flood peak discharge data of 6,914 villages (recurrence interval ranging from 5 to 100 years) in Jiangxi Province, China were collected. Fourteen design flood conditioning factors, combined with Random Forest algorithm, were used to develop design flood regression prediction model for Jiangxi Province. Results showed that model showed satisfactory regression prediction capacity, with rRMSE ranging from 0.248 to 0.254, and R2 ranging from 0.932 to 0.936 for different recurrence intervals. Catchment area was of the greatest influence on prediction results, followed by 6-hour rainfall, altitude, average slope and other conditioning factors. The study is of great significance for evaluating design flash flood.

DOI: https://doi.org/10.3850/iahr-hic2483430201-472

Year: 2024