Modelling daily reference evapotranspiration in humid locations of South Korea using local and cross-station data management scenarios

Modelling reference evapotranspiration (ET0) is a key issue in hydrology, water resources management and irrigation scheduling. This article aimed at evaluating gene expression programming (GEP) and support vector machine (SVM) techniques for modelling ET0 in humid regions of South Korea. The GEP and SVM results were also compared with the empirical ET0 models with the same input parameters. Daily meteorological data from eight weather stations comprising a period of 10 years were utilized to establish the models. This study evaluated two different data management scenarios of simulations. At the first scenario, the heuristic and empirical models were developed and tested at each location, separately, which confirmed the superiority of heuristic models to empirical models. At the second scenario, two cross-station approaches were developed: modelling ET0 using data from ancillary stations and developing generalized heuristic models for the studied stations. The obtained results revealed that both the GEP and SVM techniques could simulate ET0 in both the scenarios, successfully. Meanwhile, GEP outperformed SVM in cross-station scenario.