Comparison of machine learning techniques and spatial distribution of daily reference evapotranspiration in Türkiye

Abstract Reference evapotranspiration (ET 0 ) estimates are commonly used in hydrologic planning for water resources and agricultural applications. Last 2 decades, machine learning (ML) techniques have enabled scientists to develop powerful tools study ET patterns the ecosystem. This investigated feasibility effectiveness of three ML techniques, including k -nearest neighbor algorithm, multigene genetic programming, support vector regression (SVR), estimate daily Türkiye. In addition, different interpolation ordinary kriging (OK), co-kriging, inverse distance weighted, radial basis function, were compared most appropriate maps All developed models evaluated according performance indices such as coefficient determination ( R ), root mean square error (RMSE), absolute (MAE). Taylor, violin, scatter plots also generated. Among applied models, SVR model provided best results determining with = 0.961, RMSE 0.327 mm, MAE 0.232 mm. The model’s input variables selected solar radiation, temperature, relative humidity. Similarly, spatial distribution produced OK method, which estimates.