Modelling reference evapotranspiration using principal component analysis and machine learning methods under different climatic environments

Abstract

Reference evapotranspiration (ETo) is a complex process in the hydrologic cycle that influences several hydrologic parameters. Although several methods have been developed to model ETo, a reliable method that can use limited climatic input parameters for data-limited regions is still limited. This study evaluated four machine learning (ML) methods: M5 pruned (M5P) tree, sequential minimal optimization (SMO), radial basis function neural regression (RBFNreg) and multilinear regression (MLR). The major objective of this study was to identify the best approach to estimate ETo with minimum input data in five stations (Multan, Jacobabad, Faisalabad, Islamabad and Skardu) located in Pakistan. The datasets of these stations comprised maximum and minimum temperatures (T-max, T-min), average relative humidity (RHavg), average wind speed (Ux), and sunshine hours (n) variables. Two scenarios were used for ETo modelling. In the first scenario, five climatic variables were used as inputs to estimate ETo as obtaining full climatic parameters is the biggest challenge in developing countries. Principal component analysis (PCA) was used as a clustering technique in the second scenario to reduce the climatic input parameters. The PCA results indicated that T-max, T-min and n were identified as effective inputs for ETo estimation. Based on statistical indicators, the M5P tree outperformed the other applied ML methods in estimating ETo under various climatic environments. This study recommends focusing on areas with high ETo values and adequate irrigation scheduling of crops to achieve water sustainability.