Monthly Stream-flow forecasting using the ECMWF model, case study: Sefidrud basin-Iran

Stream flow forecasting on a monthly time scale is essential for optimal water resources management and planning. In this paper using the predictions obtained from the ECMWF climate model, monthly stream flow forecast was made in Shahroud river Subbasin, part of Sefidrood basin northwest of Iran. To achieve this aim, using monthly precipitation forecasts from ECMWF climate model in tandem with SVR data-driven modeling, as a rainfall-runoff model, the stream flow was predicted based on the predicted precipitations. First, the results of precipitation forecast, for the desired historical period, up to a 3-month forecast horizon for the study area were obtained from the Climate Data Store. Then, by using the SVR driven model, a linked Climate-Data-driven model was developed to predict the flow up to a 3-month forecast horizon. The results showed that flow forecasting based on climate forecasting models is more accurate for the forecast horizon of the next month than two and three months. So that the forecast horizon of the next month has the highest Nash-Sutcliffe coefficient, in calibration 0.77 and in validation 0.48. The highest correlation coefficient in calibration 0.87 and validation 0.69, the lowest root mean square error in calibration 6.8 and validation 6.3 million cubic meters and also has the best relative bias value for calibration 0.96 and validation 1.1. Also the results, based on the POD and FAR probabilistic indices, showed that the developed predictive model has a high ability to detect different states of stream flow events, especially for extreme flows event.