基于LightGBM模型秋浦河洪水过程模拟与预报研究

Study on Simulation and Forecast of Qiupu River Flooding Based on LightGBM Model

文章利用皖南秋浦河洪水过程资料,以历史径流、降水量等为协变量,新建了LightGBM模型对洪水过程进行模拟,最后运用交叉验证方法对预报精度进行评估。结果表明,LightGBM方法在模拟洪水过程方面的性能良好,其R^(2)达0.96,RMSE和Bias仅为101.50 m^(3)/s、4.57%。另通过对比不同预见期下的预测精度,发现随着预见期增加,对洪水过程预报精度逐渐降低,当预见期超过4时,其预测精度迅速下降。文中展示了机器学习技能在水文过程、径流量序列变化仿真中良好前景;预测结果可辅助决策者制定有效的洪水应对措施以减少灾害损失。

In this research,using the data of Qiupu River flooding in southern Anhui,historical runoff and precipitation were used as covariates to establish a LightGBM model to simulate the flooding process.Finally,the cross-validation method was used to evaluate the accuracy of the forecast.The results show that the performance of the LightGBM method in simulating the flooding process is excellent,with an R^(2) of 0.96,and RMSE and Bias of only 101.5m^(3)/s and 4.57%.Furthermore,by comparing the forecast accuracy under different lead times,it was found that the accuracy of the forecast gradually decreased as the lead time increased,and the forecasting accuracy rapidly declined when the lead time exceeded 4.This research demonstrates the great potential of machine learning skills in simulating hydrological processes and changes in runoff sequences.The predicted results can assist decision-makers in formulating effective flood response measures to reduce disaster losses.