An investigation on the estimation of evaporation by combining artificial neural network and dynamic factor analysis

Evaporation is a substantial factor in hydrological circle, moreover a significant reference to the management of both water resources and agricultural irrigation. In general, evaporation can be directly measured by evaporation pan. As for its estimation, the accuracy of traditional empirical equation is not very precise. Therefore, in this study the Dynamic Factor Analysis (DFA) is first applied to investigating the interaction and the tendency of each gauging station. Additionally, the analysis can effectively establish the common trend at each gauging station by evaluating the corresponding AIC (Akaike Information Criterion) values. Furthermore, the meteorological factors such as relative humidity and temperature are also conducted to identify the explanatory variables which have higher relation to evaporation. These variables are further used as inputs to the Back-Propagation Neural Network (BPNN) and are expected to provide meaningful information for successfully estimating evaporation. The applicability and reliability of the BPNN was demonstrated by comparing its performance with that of empirical formula. Keywords: Evaporation, Dynamic Factor Analysis, Artificial Neural Network.