Comparative Performance of Empirical and Heuristic Daily Reference Evapotranspiration Models in the Lower Donga River Basin, Nigeria

The need to accurately estimate evapotranspiration (ET0) in tropical regions with limited climatic data is prodigious, considering its influence on hydrology, agriculture, and agro-meteorology. Evaluation of the ET0 model is particularly essential in developing countries where meteorological data needed to estimate ET0 using Penman-Monteith FAO-56 (PMFAO-56) model are limited or not available. The purpose of this study is to compare a few empirical ET0 models with the corresponding heuristic data-driven models to test their efficacy in the lower Donga basin, Taraba State, Nigeria. For this purpose, a temperature-based model (Hargreaves) and a combination-based (PMFAO-56) model was evaluated and validated with observed ET0. Data-driven models consisting of Artificial Neural Networks (ANNs) and Gene Expression Programming (GEP) were employed for evaluating models using 32 years of daily meteorological data. The results were compared with the empirical models with respect to coefficient of determination, (R2), Nash Sutcliffe efficiency coefficient (NSE), root mean square error (RMSE), and scatter plots. ANNs and GEP models have the least RMSE with NSE and R2 of up to 1 and 0.95 at the training and testing periods, respectively. However, GEP models produced a set of equations as compared to ANNs and are therefore preferred for engineering applications. The proposed approach produced simple, yet reliable estimates for ET0 evaluations in the basin, which can serve as promising alternatives to the conventional methods.