HYDROMETEROLOGICAL MODELING APPROACHES USING SUPPORT VECTOR REGRESSION (SVR) AND GENETIC PROGRAMMING (GP): Fid-Bau Portal

HYDROMETEROLOGICAL MODELING APPROACHES USING SUPPORT VECTOR REGRESSION (SVR) AND GENETIC PROGRAMMING (GP)

Maity, Rajib / Kashid, S. S. / Bhatnagar, Ashish

Hydrometeorology deals with the problems involving hydrologic cycle, water budget, and rainfall statistics of storms and streamflows. These systems are quite complex and require complete understanding in order to analyze and manage water resources of a region. The rainfall and streamflow over the region are significantly influenced by large-scale coupled atmospheric-oceanic circulations patterns, such as El Niño-Southern Oscillation (ENSO), Equatorial Indian Ocean Oscillation (EQUINOO) etc. The local meteorological parameters like Outgoing Longwave Radiation (OLR), temperature, pressure, etc., also affect rainfall and basin scale streamflow. To model the complex relationship between these variables and basin scale stream-flow, artificial intelligence tools—Support Vector Regression (SVR) and Genetic Programming (GP) have been employed. The application of Support Vector Machine (SVM) to general regression problem to forecast future streamflow potentially improves the performance of monthly basin-scale streamflow prediction and gives better results than traditional methodologies such as ARIMA models. For multivariate inputs, GP-derived streamflow forecasting models are used. The GP model captured the performance of weekly basin-scale streamflow prediction and successfully improved upon existing methodologies. The observed and predicted streamflows, using SVR and GP, were found to correspond well with each other with a correlation coefficient of 0.78 and 0.81 respectively, which is reasonably good for such a complex system. This may however be noted that Support Vector Regression (SVR) used only single variable input and thus holds still greater promise to arrive at even better results.