Streamflow forecasting for operational water management in the Incomati River Basin, Southern Africa

If the future availability of water is uncertain to water managers, dam operators and water users, then an effective allocation among competing uses can be difficult. The difficulties can partly be alleviated by including streamflow forecasting as a tool for informed decision making. The Incomati basin in Southern Africa frequently experiences water shortages, and here streamflow forecasting can contribute to an improved water management. This paper explores the skill of streamflow forecasting and its usefulness in decision making in the Incomati basin. The study applies correlation and regression methods to forecast streamflow, and standard verification scores to evaluate the skill of the forecasts. Suitable statistical forecasting techniques were analysed and tested. The data used for forecasting include Sea Surface Temperature (SST), El Niño Southern Oscillation (ENSO), rainfall and streamflow.Results show that there is some scope for streamflow forecasting that can support water management decision making in the basin. The rainfall and streamflow of the previous months and/or season can be used to predict the streamflow in the next month and/or season with reasonable to good results. Results obtained during low flow periods (May-September) were found to be better than those obtained for the high flow periods (October-April). However, inclusion of ENSO and/or SST as an explanatory variable enhanced forecast skill, particularly during high flow periods. Forecasts were conducted for streamflow being in the below normal, above normal or normal terciles, with the forecasts for the extremes found to have better skill than forecast for the streamflow being in the normal tercile. Forecasts for low flows demonstrated the best skills, with these being of most use to the allocation of the scarce water resources.