Descripción
In line with climate change and water crises, the population growth has drawn atten tion to agriculture. The sector is responsible for the use of 70% of the world’s water and a waste of approximately 50% of this total in irrigation processes. Several technological methods are being developed to minimize this impact and collaborate with the UN Sustainable Development Goals, within them, sustainable agriculture. Aiming at optimizing the use of water resources, it is necessary to analyze evapotranspiration, as it is the most active variable in the hydrological cycle and the main component in the water balance of agricultural ecosystems. Based on the hypothesis that Machine Learning analysis can be applied to determine evapotranspiration and aid in decision making in irrigation, allowing assertive estimates and without dependence on a wide variety of data, this article presents the application of Decision Tree Regressor techniques. , Random Forest, Artificial Neural Network and XGBoost for that purpose. Using a dataset from the National Institute of Meteorology (INMET), the models were trained based on widely validated equations for Evapotranspiration calculations. After the testing routine, it was possi ble to obtain satisfactory results, with MAE less than 0.0015, demonstrating the effectiveness of computational techniques for estimating evapotranspiration.
