The Longyangxia Gorge Key Water Control System is the first of the stairstep power sations along the Longyangxi- a-Qingtongxia river section. It has been playing an very important role in providing power, protecting flood and ice run supplying and irrigation etc. in the northwestem China. Therefore, the study on trend prediction, variation on the now into the Longyangx- ia Reservoir are of the great social and economic benefits. In the medium- and-long-range runoff forecast, all kinds of regression equation are often used for predicting future hydrologic regime. However, these regression models aren' t appropriate to super long -range runoff forecast because of the restricting on weather data and so on. So a new super long-range runoff forecast model don' t depend on Real-time weather data and called“ Period correcting for residual error series GM(1,1 ) model” is presented based on analyzing for the relational hydrologic data and the variation on the now into the Longyangxia Reservoir, and the fore- cast model was applied successfully to predict the recent and super long -term trends of the flow into the Longyangxia Reservoir. The results indicate that the annual now into the Longyangxia Reservoir is in the ending minimum period of the runoff history. The runoff increasing is expected in for the coming years. The Longyangxia Gorge Key Water Control System is the first of the stairstep power sations along the Longyangxi- a-Qingtongxia river section. It has been playing an very important role in providing power, protecting flood and ice run supplying and irrigation etc. in the northwestem China. Therefore, the study on trend prediction, variation on the now into the Longyangx- ia Reservoir are of the great social and economic benefits. In the medium- and-long-range runoff forecast, all kinds of regression equation are often used for predicting future hydrologic regime. However, these regression models aren 't appropriate to super long -range runoff forecast because of the restricting on weather data and so on. So a new super long-range runoff forecast model don' t depend on Real-time weather data and called “Period correcting for residual error series GM (1,1) model ” is presented based on analyzing for the relational hydrologic data and the variation on the now into the Longyangxia Reserv The results indicate that the annual now into the Longyangxia Reservoir is in the ending minimum period of the runoff history. The runoff increasing is expected in for the coming years.