Generalized Neurofuzzy Models for Estimating Daily Pan Evaporation Values from Weather Data

Abstract

Estimation of daily pan evaporation values is of most importance in water resource system management and planning. This paper presents a study aimed at developing generalized neurofuzzy (GNF)–based evaporation models corresponding to Penman, Stephens-Stewart (SS), and Griffiths methods. A GNF model was also made by using temperature as the sole input parameter to evaluate the application of single-input temperature-based models for estimating the evaporation values. In the first part of the study, neurofuzzy (NF) evaporation models were developed and compared with Penman, SS, and Griffiths models for three weather stations located in Arizona, USA. Five-parameter NF models were generally found to be better than the Penman, SS, and Griffiths models. The NF models were used for estimating evaporations at the Tucson station by using the data from the Phoenix and Flagstaff stations in the second part of the study. It was found that NF models can be successfully used in cross-station applications. In the third part of the study, the GNF models were obtained by calibrating and using the pooled data from the Phoenix, Flagstaff, and Tucson stations located in Arizona and were tested using the data from weather stations in Albuquerque, NM; Tucumcari, NM; Cedar City, UT; and Ahwaz, Iran). Generalized SS and Griffiths models were also obtained and compared with GNF models. The comparison of the results revealed that the GNF models performed better than the Penman generalized SS, and Griffiths models. However, the generalized SS and Penman models were found to be better than the GNF model for the Ahwaz station.