Abstract:The ensemble methods to get a better combined result from l arge quantities of numerical forecast model output products are investigated. S ome variable weig ht decision making models based on super ensemble methods are put forward. E vidences suggest that these models are simple, convenient and effective, which c an help forecasters to pick up enough high quality ensemble statistics informat i on from massive numerical products， such as circulation situation, fields of ph ysical elements,　precipitation pre diction, cold air prediction, ensemble matrix, map of character. The ensemble m ethods are more perfect and refined, which can dyn amically reflect the abilities of charactering each numerical model. And this h elps to improve the effect and skill of refined numerical weather forecasting for different area and time to some exten t, and provides valuable reference to the operational prediction.