FORECASTING STREAM FLOW OF KADUNA RIVER AT WUYA GAUGING STATION IN NIGER STATE

Abstract

ABSTRACT The Box Jenkins methodology for modeling and forecasting was used to identify suitable models and forecast future stream flow values for average and maximum discharge of Kaduna River at Wuya gauging station in Niger state and to predict extreme events of flooding. The average and maximum monthly stream flow data used for the analysis were obtained from the Nigerian Hydrological Serves Agency from January 1988 to September 2021 (33 years). Autoregressive Integrated Moving Average ARIMA (0,1,4)(0,1,4) and Autoregressive Integrated Moving Average ARIMA (0,1,3)(0,1,3) models were the most suitable models found adequate to forecast the average and maximum stream flow of Kaduna River at Wuya gauging station.. Both models achieved the lowest normalized Bayesian Information Criterion (BIC) which were the criteria for chosen the best model after prognosis. The selected models also have the best Ljung-Box Q statistical significance P Value of all the models identified. Diagnostic checks of the models revealed that the models selected have a residual that was white noise. The models used for forecasting had the highest Coefficient of determination (R2) of 0.795 and 0.813 with the least Mean absolute percentage error (MAPE) of 28.785%, 34.461% respectively for ARIMA (0,1,4) (0,1,4) and ARIMA (0,1,3) (0,1,3). Both model forecast was reasonable having their MAPE between 21% to 50%.

Keywords

• forecasting
• stream flow
• arima model