| Abstrakt: |
Reference evapotranspiration (ETo) simulation is of great importance for various procedures of the hydrological cycle, irrigation, agronomic management, and planning and management of water resources. Agricultural water management and irrigation scheduling require accurate estimation of crop water requirements. The reference evapotranspiration (ETo), which is used to calculate the evapotranspiration of each crop, is estimated first. The purpose of this study was to develop a new coupled model for estimating daily ETo. In this research, differing artificial intelligence (AI) approaches including support vector regression (SVR), algorithm of innovative gunner (AIG), and hybrid algorithm of innovative gunner-support vector regression (AIG-SVR) models were used in two different climatic conditions to estimate reference evapotranspiration, one station in the dry region (Marree Aero) and one station in the humid region (St Helen Aerodrome) in Australia. The results of the AIG-SVR model were compared with those of the conventional support vector regression (SVR) model using several performance evaluation methods comprising the statistical criteria including correlation coefficient (R), root mean square error (RMSE), Nash–Sutcliffe coefficient (NS), and RMSE-observations standard deviation ratio (RSR)). Evaluation results showed that the developed coupled models yielded better results than the classic SVR, with the AIG-SVR outperforming the SVR. The results showed that ETo values predicted by all AIG-SVR models agreed well with the corresponding observed values, with R, RMSE (mm day−1), NS, and RSR = 0.945, 1.124, 0.894, 0.325 respectively in Marree Aero station and 0.951, 0.476, 0.905, and 0.307 respectively in St Helen Aerodrome station in testing data sets. In the local scenario, RMSE and RSR reduced by up to 15.805% (1.335 to 1.124) and 15.803% (0.386 to 0.325), respectively, and NSE and R increased by up to 5.176 (0.850 to 0.894) and 2.383% (0.923 to 0.945), respectively, in testing data sets at Marree Aero station whereas RMSE and RSR reduced by up to 19.594% (0.592 to 0.476) and 19.633% (0.382 to 0.307), respectively, and NSE and R increased by up to 6.096 (0.853 to 0.905) and 2.922% (0.924 to 0.951), respectively, in testing data sets at St Helen Aerodrome station. The findings reveal that the AIG-SVR model performs better than the SVR model. As a result of this study, AIG-SVR and SVR models can both provide important insights into how ETo simulating can be improved. Based on the results obtained from the hybrid model in this research. ETo in Australia is possibly considered an application system in Australia. [ABSTRACT FROM AUTHOR] |
Full text from SpringerLink