Support vector regression model optimized with GWO versus GA algorithms: Estimating daily pan-evaporation

Ensuring accurate estimation of evaporation is vital for suitable planning and efficient utilization of water resources, especially in arid and semiarid regions. Thus, the purpose of this study was to explore the applicability of the support vector regression (SVR) model optimized by two-novel nature-inspired algorithms, i.e., Grey Wolf Optimizer (SVR-GWO), and Genetic Algorithm (SVR-GA) to estimate the daily pan-evaporation (EP) process in Punjab State (i.e., Bathinda and Ludhiana stations) and Haryana State (i.e., Hisar station). The optimal input combination was extracted by applying the Gamma test (GT) for the application of hybrid SVR models in estimating daily EP at three study locations. The estimate obtained by the hybrid SVR-GWO and SVR-GA models were compared with the observed one using six performance measures Correlation Coefficient (CC), Nash-Sutcliffe Efficiency (NSE), Willmott Index (WI), Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Scatter Index (SI), and via pictorial examination (radar chart, time-variation plot, scatter plot, and Taylor diagram). The obtained outcomes demonstrate that the hybrid SVR-GWO model exhibits the highest prediction performance at all the study stations for both periods, closely followed by the hybrid SVR-GA model. Also, the SVR-GWO-5 model attained the highest value of NSE (0.715/ 0.812/ 0.833), CC (0.849/ 0.904/ 0.913), and WI (0.912/ 0.949/ 0.951) and lowest value of MAE (1.693/ 0.947/ 0.781 mm/day), RMSE (2.258/ 1.279/ 1.223 mm/day), and IS (0.374/ 0.322/ 0.274) in testing stage at Bathinda, Ludhiana and Hisar stations, respectively. The proposed hybrid SVR-GWO model was found to be more efficient and robust than the SVR-GA model in estimating the daily EP at the study stations.