Mohawesh, O. (2010) 

 Abstract

Accurate estimation of reference evapotranspiration (ETo) is of paramount importance for many studies, such as hydrologic water balance, irrigation system design and management, crop yield simulation, and water resources planning and management. The Penman–Monteith (PM) method is recommended by FAO as the sole method to calculate ETo wherever the required input data are available Unfortunately, the weather input data for the PM equation is expensive and often difficult to obtain for practical applications. An alternative is the application of mathematical models like artificial neural networks (ANNs). ANNs are very appropriate for the modeling of non-linear processes, i.e. the case of evapotranspiration. This study is to examine the potential of artificial neural networks (ANNs) for estimating monthly ETo and to compare the ANN models with local calibrated Hagreaves equation (HarLC) under arid and semi arid environments of agricultural irrigated lands. Considering all test locations for ANNs, the RMSE values ranged from 0.001 to 0.121 mm day^-1. ME ranged from -0.137 to 0.070 mm day^-1. R² ranged from 0.95 to 0.99. HarLC values of RMSE, ME and R² ranged from 0.394 to 0.4754 mm day^-1, -0.2842 to -0.196 mm day^-1, and 0.94 to 0.96, respectively. The results suggested that the monthly ETo could be computed from air temperature and R_a or R_s using the trained ANNs at another location. Based upon statistical evaluation, it can be concluded that extraterrestrial solar radiation (R_a) and or solar radiation (R_s)are the most effective variable and is highly recommended to be into the ANNs model. Therefore, in the specific climate of Jordan valley, R_a or R_s have to be regarded as a necessary variable for the modeling the ETo with high accuracy. More over, ANNs performed better than local calibrated Hargreaves equation (HarLC) in this study area.

A case study, Jordan valley. (Submitted to Water Resources management)