Development of Generalized Higher-Order Neural Network-Based Models for Estimating Pan Evaporation

Abstract

It is common practice to develop artificial neural network models using location-based single dataset for both the training and testing. Based on this procedure, the developed models may perform poorly outside the training location. Therefore, this study aims at developing generalized higher-order neural network (GHNN) models for estimating pan evaporation (E p) using pooled climate data of different locations under four agro-ecological regions in India. The inputs for the development of GHNN models include different combinations of daily climate data such as air temperature, relative humidity, wind speed, and solar radiation. Comparisons of developed GHNNs were made with the generalized first-order neural network (GFNN) and generalized multi-linear regression (GMLR) models. It is concluded that the GHNNs along with GFNNs performed better than the GMLR models. Further, GHNNs were applied to model development and model testing locations to test the generalizing capability. The testing results suggest that the GHNN models have a good generalizing capability.