In order to realize the mid-long term runoff prediction of upper Yangtze River, we select characteristic factos relat-ed to the monthly runoff from 74 indices, which reflect the features of the sun, atmosphere, ocean, and employ the optimal subset regression prediction method to establish prediction model. The return check on the monthly runoff in recent 32 years was per-formed. The results showed that the predication result of Pingshan Station is better than that of Yichang Station. In June 2013, the above method was used to predict the monthly runoff in July 2013 at Pingshan Station and Yichang Station and the satisfactory prediciton accuracy was obtained.%为了实现对长江代表水文站流量的中长期(月尺度及以上)预测，从表征太阳、大气、海洋特征的74项指数中，挑选与长江上游月流量有较好相关关系的特征指数作为备选因子，采用最优子集回归预测方法建立了预测模型，并对近32 a的月流量进行回报校验。结果表明，屏山站的最优子集回归预测效果略好于宜昌站。2013年6月，利用该方法预测了2013年7月屏山、宜昌站月流量，预测精度满足要求。
展开▼
