Estimativa da evapotranspiração de referência utilizando modelos preditivos e seleção de atributos

Abstract

In Brazil, irrigated agriculture is the sector responsible for the greater water consumption demand. Therefore, it is necessary to develop techniques that allow the use of this water in a sustainable way. Evapotranspiration is the simultaneous occurrence of evaporation and transpiration processes on a vegetated surface, and designates the amount of water lost by a crop. By estimating this loss, researchers and farmers can efficiently manage the water consumption of their crops. This work proposes the creation of prediction models for reference evapotranspiration, based on climatic data. The multidimensionality of the data can generate models that need many climatic variables, so a solution that employs the execution of feature selection techniques before the construction of the models was also proposed. All in all, it was possible to conclude that models with high accuracy can be generated by using M5’ trees and feature selection techniques.