摘要
地表径流量的非线性预测是水资源优化管理与开发的先决条件。文章以泰国春武里水文站2016年日降雨与径流量数据为基础,采用相关时滞、随机森林(RF)的方法构建了日径流预测模型。结果表明,该地区降雨与径流之间存在1-3d的显著时滞尺度;以前置的降雨时间序列变量数据能够很好辅助建立径流预测模型,RF模型的R2达0.9982,MAE为12.53,RMSE为7.79。该方法能够较好地拟合降雨-径流之间的关系,具有一定应用性。
Nonlinear prediction of surface runoff is a prerequisite for optimal management and development of water resources.Based on the data about daily rainfall and runoff at Chunwuli hydrological station of Thailand in 2016,a daily runoff prediction model was established by using the correlate time-lag,random forest method.The results show that there is a remarkable time-lag dimension of 1-3d between rainfall and runoff in this area;the data about rainfall time series variable can assist better to construct the runoff prediction model,R2 of RF model is close to 0.9982,MAE is 12.53 and RMSE is 7.79.The method can fit better the relationship between rainfall and runoff,and has certain applicability.
作者
唐大伟
TANG Da-wei(Changzhou Sub-bureau,Jiangsu Provincial Hydrology and Water Resources Investigation Bureau,Changzhou 213000,China)
出处
《黑龙江水利科技》
2019年第12期6-9,共4页
Heilongjiang Hydraulic Science and Technology
关键词
随机森林
时滞相关性
预测
R语言
random forest
time-lag correlation
forecast
R language
