Reference evapotranspiration estimated with simplified models for the state of Mato Grosso, Brazil

The objective of this work was to evaluate the performance of 12 simplified models for the estimation of reference evapotranspiration (ETo) for the state of Mato Grosso, Brazil. The data were collected from automatic weather stations (AWS) of the Instituto Nacional de Meteorologia, located in 28 municipalities of the state. The following simplified estimation models were evaluated: Hargreaves-Samani, Camargo, Makkink, Linacre, McGinness-Bordne, Romanenko, Turc, Holdridge, Solar Radiation, Jensen-Haise, Hansen, and Caprio. The Fao 56 Penman-Monteith method (FPM) was used as reference for assessing the simplified estimates. Statistical performance was evaluated through relative mean error (RME), root mean square error (RMSE), Willmott's d index, and according to the numerical order of models for each index. The Makkink model overestimated ETo by 2.0 to 3.0 mm per day, with scattering values of 2.75 mm per day and 0.40 d index, which represented the worst results among models, regardless of the municipality evaluated. The Turc and McGinness-Bordne models showed the best performances for estimating ETo in 57.1 and 25% of the AWS, respectively. The Romanenko, Makkink, and Holdridge models are not recommended for the state of Mato Grosso, Brazil.