Based on the ARMA (p

q) model of non-stationary time series analysis

a self-adapting amendment forecasting model is derived for the real-time forecasting of daily runoff hydrograph in the rivers located in arid or semi-arid areas and lack of hydrologic and meteorological data with the constideration of effects of temperature and precipitaton on runoff. This model has been used on the Manasriver in Xinjiang for the continuous simulation of daoiy runoff hydrograph from1984 to 1988 with satisfation

and is proved to be practicable.