Comparative assessment and spatial prediction of the reference evapotranspiration through geostatistical methods

Abstract

The determination of the reference evapotranspiration (ETo) is essential when carrying out agricultural irrigation planning. One of the most recommended methodologies for this is that of FAO Penman-Monteith; however, its use is limited to the need to have the necessary variables for its estimation, so it is possible to resort to the spatial prediction of the ETo from known values of it. In this study, Ordinary Kriging (KO) and Universal Kriging (KU) were compared and evaluated in the spatial prediction of the ETo on a monthly scale in the Venezuelan plains, estimated by FAO Penman-Monteith, which in turn allowed visualizing its spatial behavior. It was evaluated by cross-validation, using the coefficient of determination R², the mean squared prediction error (MSPE) and the root mean square error (RMSE). The KO was determined as the best predictor, attributed to the stationarity of the mean in most months, except March and October, even when the difference in errors is reduced between the methods, as well as a low and moderate spatial dependence of the ETo for the KO, and high for the KU. The ranges ranged from 100 km for October and November to 290 km for July according to KO, while for the KU, they fluctuated between 35 km for July and October and 290 km for January. There was a greater demand for ETo towards the eastern plains than the western plains, as well as its seasonality, due to the behavior of the variables that influence its estimation.