Abstract: The catchment above Tangnag station is the principal areas for runoff formation in the upper reaches of the Yellow River.For example,the water flows from the Tangnag Station account for 95% of the inflows into the Longyangxia Reservoir,the largest reservoir on the upper Yellow River.However,runoff in the upper Yellow River above the Tangnag has been decreasing recently due to the persisting drought in the basin,which not only greatly influences the economy and people's standards of living in the upper reaches,but also curbs the economic development and ecological environment imporvement within the Yellow River basin.As a result,the accurate prediction for the variation of the runoff at the Tangnag Station is indispensable for the efficient and logical exploitation of the water resources at the catchment scale,as well as determining the amount of water transported from other basin.However,sophisticated methods are not available at present to determine the variability of water flows,due to complexity of their intrinsic evolutions,and close and complicated relationships to climatic changes.In addition,precision of runoff prediction is greatly influenced by the difficulty in long-term weather forecasting.To solve the above problems,a dispersal stochastic process model based on the Markov Chain was presented for forecasting the inflow into the Longyangxia Reservoir in this paper.This forecasting method is mainly applicable to the objects that are of biggish stochastic vibration.The results show that the calculated values can commendably tally with the measured values.Therefore,Morkov forecast model is of obvious advantage.

Key words: dispersal stochastic process, Markov Chain, runoff forecasting, probability transfer matrix