Forecasting of rainfall and subsequent river runoff is important for many operational problems and applications related to hydrology. Modeling river runoff often requires rigorous mathematical analysis of vast histori...Forecasting of rainfall and subsequent river runoff is important for many operational problems and applications related to hydrology. Modeling river runoff often requires rigorous mathematical analysis of vast historical data to arrive at reasonable conclusions. In this paper we have applied the stochastic method to characterize and predict river runoffofthe perennial Kulfo River in southem Ethiopia. The time series analysis based auto regressive integrated moving average （ARIMA） approach is applied to mean monthly runoff data with 10 and 20 years spans. The varying length of the input runoff data is shown to influence the forecasting efficiency of the stochastic process. Preprocessing of the runoff time series data indicated that the data do not follow a seasonal pattern. Our forecasts were made using parsimonious non seasonal ARIMA models and the results were compared to actual 10-year and 20-year mean monthly runoff data of the Kulfo River. Our results indicate that river runoff forecasts based upon the 10-year data are more accurate and efficient than the model based on the 20-year time series.展开更多
文摘Forecasting of rainfall and subsequent river runoff is important for many operational problems and applications related to hydrology. Modeling river runoff often requires rigorous mathematical analysis of vast historical data to arrive at reasonable conclusions. In this paper we have applied the stochastic method to characterize and predict river runoffofthe perennial Kulfo River in southem Ethiopia. The time series analysis based auto regressive integrated moving average （ARIMA） approach is applied to mean monthly runoff data with 10 and 20 years spans. The varying length of the input runoff data is shown to influence the forecasting efficiency of the stochastic process. Preprocessing of the runoff time series data indicated that the data do not follow a seasonal pattern. Our forecasts were made using parsimonious non seasonal ARIMA models and the results were compared to actual 10-year and 20-year mean monthly runoff data of the Kulfo River. Our results indicate that river runoff forecasts based upon the 10-year data are more accurate and efficient than the model based on the 20-year time series.
