Application of SVMs Algorithms for Prediction of Evaporation in Reservoirs

Assessment of water resource requires effective procedures for evaporation estimation involving measurable meteorological parameters and such approaches are rarely available in the literature. In the present work, the data of daily evaporation, temperature, solar radiation, relative humidity, wind speed are used to assess the potential and usefulness of SVM based modeling. The performance of the SVM algorithms (Rbf & Polynomial) is compared with the linear regression on the basis of performance parameters having different combinations of input parameters. The comparison of results showed that there is better agreement when more input parameters are considered for model building as compared to a single parameter. The outcome of study suggests that the SVM based modeling can be applied as an alternative approach for estimation of daily evaporation from the reservoirs.