基于逐步回归—LMBP算法的大通站旬径流与月径流预报

Ten-days and Monthly Runoff Forecasting in Datong Station Based on Stepwise Regression and LMBP Algorithm

为提高长江干流大通站旬径流与月径流预报精度,选取大通站1980~2012年各旬、各月径流观测资料及国家气候中心同期发布的72项大气环流资料,采用逐步回归法—LMBP算法对大通站的旬平均径流序列进行模拟和预报,并与月尺度径流序列的计算结果做了对比。结果表明,预测值与原序列的趋势基本相同,旬尺度的径流预报精度高于月尺度的预报精度,表明时间尺度的选择影响径流预报的精度。

In order to investigate the precision of ten-days and monthly runoff forecasting at Datong Station in the main stream of the Yangtse River, the runoff data of Datong Station and 72 circulation indices issued by the National Climate Center from 1980 to 2012 were selected to simulate and forecast ten-days average runoff series by using stepwise regression and LMBP algorithm. Compared with the monthly runoff series, the results show that the trend of forecasting is basically consist with the original runoff series; ten-days runoff forecasting precision is higher than that of monthly runoff forecasting, which demonstrates that the selection of time scale has impact on the runoff forecasting precision.