Estimation of Evapotranspiration Using Artificial Neural Networks end Spectral Data: Application in a Protected Area

Abstract

The mapping and spatial analysis of land use and land cover (LULC) changes are essential for monitoring environmental dynamics and managing water resources. Studies that evaluate evapotranspiration (ET) considering LULC changes help quantify the anthropogenic impacts on the hydrological cycle. Artificial Neural Networks (ANN) have been used to estimate ET but rely on meteorological data series. Therefore, the objective of this study is to estimate actual ET using a Multilayer Perceptron Artificial Neural Network (MLP-ANN) for the Itupararanga Environmental Protection Area, a region of hydrological importance for the Metropolitan Region of Sorocaba, utilizing LULC data and spectral bands from the Landsat satellite. The 2021 ET was calculated using the SEBAL algorithm and used as the basis for training and supervised classification in the ANN-MLP model of the TerrSet software. The ANN estimates were evaluated using TerrSet criteria, highlighting visual differences that indicate the need for parameter adjustments in the MLP to better align with the SEBAL estimates, which had been previously validated. This innovative approach complements the use of ANN in ET modeling, offering new perspectives for analyzing environmental dynamics.

Keywords
Itupararanga EPA; Itupararanga Reservoir; machine learning; remote sensing