Modeling reference evapotranspiration in island environments: Assessing the practical implications

Among the reference evapotranspiration (ETo) equations, temperature-based and radiation-based equations have been accepted as promising approaches when the available meteorological data are limited. The present paper aimed at assessing this general implication in island environments, which might be generally volatile than the interior/coastal zones. Using data from five island locations of Iran, daily ETo values were calculated by the benchmark FAO-Penman-Monteith (FAO56-PM) model, then the commonly used temperature-, radiation- and mass transfer-based ETo equations were evaluated against the benchmark values. An additional heuristic modeling of ETo was also performed through employing gene expressions programming (GEP), evaluated by k-fold testing. The obtained outcomes showed that the calibrated mass transfer-based equations and their corresponded GEP models generally surpassed the other applied models in the studied locations. This criticized the commonly accepted implication about applying temperature/radiation-based models in case of data scarcity for different locations. Nevertheless, analysis of the temporal variations of models performance among test years showed that fixing the minimum test size as 1 year for local k-fold testing (that is the case in similar studies) might be misleading as it could not show the fluctuations in modeling accuracy well. A minimum affordable size of at least one growing season was instead selected, tested and recommended in the current research.