基于偏互信息法遴选因子的长江中长期径流预报

为提高流域中长期径流预见期和预报精度,以长江流域代表性水文站为例,研究遥相关气候因子对水文站径流的影响,通过偏互信息方法遴选与逐月径流具有显著相关性的气候因子,并采用多元回归方法建模进行中长期径流预报。研究表明:根据偏互信息法选出的输入因子建立的回归方程在建模和试报阶段的拟合优度值都大于0.6,而且入选的气候因子均具有4个月及以上的预见期。可见利用偏互信息法挑选遥相关气候因子进行中长期径流预报能够延长预见期并提供具有较高精度的预报结果。

Medium and long-term runoff forecasting based on input variables selected by partial mutual information

In order to improve the accuracy and prolong the prediction period of medium and long-term runoff forecasting, we take the representative hydrological stations in the Yangtze River Basin as an example to study the influence of teleconnection climatic factors on the runoff of hydrological stations. Firstly, the partial mutual information (PMI) is introduced to help select the teleconnection climatic factors that are significantly related to runoff. Then, the medium and long-term runoff is forecasted based on the multiple regression method. The results show that with input factors selected by PMI, the fitness of the regression equation is above 0.6 in both modeling and trial forecasting stage. Furthermore, each selected factor has a prediction period more than 4 months. In conclusion, using PMI to select the teleconnection climatic factors can provide high accurate medium and long-term hydrological forecasting.