Potential of Support Vector Machine Fed by ERA5 for Predicting Daily Discharge in the High Atlas of Morocco

In the high Atlas Mountains of Morocco, forecasting discharge is of crucial importance for water resource management but it remains a difficult task due to the scarcity of observations. This study presents meteorological re-analysis data as a promising alternative for feeding hydrological model where observation data are absent. Re-analysis data are used here as inputs to a ‘Support Vector Regression’ (SVR) algorithm to simulate the daily discharge over the Rheraya catchment in Tensift, Morocco. The SVR model is calibrated, and its accuracy is assessed using a time series from September 2003 to August 2016. The efficiency of daily predictions of streamflow is compared with the daily-observed discharge. SVR model fed by re-analysis data showed a good performance during the test period with an NSE and RMSE respectively equal to (0.88 and 0.89 m3/s). This accuracy is mainly due to the high correlation between the discharge and the predictors. The SVR model fed by re-analysis data can be considered as a reliable tool to estimate daily discharge.