Neural networks and genetic algorithm approach for nonlinear evaporation and evapotranspiration modeling

The purpose of this study is to develop and apply the generalized regression neural networks model (GRNNM) embedding the genetic algorithm (GA) in order to estimate and calculate the pan evaporation (PE) and the alfalfa reference evapotranspiration (ETr), Republic of Korea. Since the observed data of the alfalfa ETr using lysimeter have not been measured for a long period, Penman–Monteith (PM) method was used to estimate the observed alfalfa ETr. The COMBINE-GRNNM-GA (Type-1) was developed to calculate a reasonable PE and the alfalfa ETr. The COMBINE-GRNNM-GA (Type-1) was evaluated through the training, the testing, and the reproduction performances, respectively. An uncertainty analysis was used to eliminate the climatic variables of the input layer nodes and to construct the optimal COMBINE-GRNNM-GA (Type-1). The climatic variable with the lowest smoothing factor during the training performance was eliminated from the original COMBINE-GRNNM-GA (Type-1). The climatic variable with the lowest smoothing factor implies the most useless input layer node for the model output. Therefore, the optimal COMBINE-GRNNM-GA (Type-1) can estimate and calculate the PE, which is missed or ungaged, and the alfalfa ETr, which is not measured, with the least cost and endeavor. Furthermore, it is possible to derive a linear regression statistical model between the measured PE and the corresponding alfalfa ETr. Finally, the PE and the alfalfa ETr maps could be constructed to provide the reference data for a drought analysis and an irrigation networks system using the optimal COMBINE-GRNNM-GA (Type-1), Republic of Korea.