Classification of Evapotranspiration using the Weather Forecasting Model and Self-Organizing Map Learning

The accurate estimation of evapotranspiration (ETo) is crucial for efficient irrigation management and the sustainable administration of water resources. One of the most widely used approaches for simulating atmospheric variables is the Weather Research and Forecasting (WRF) model, which has proven effective in both climate research and numerical weather prediction. In this work, outputs from the WRF model are integrated with a Self-Organizing Map (SOM), an unsupervised neural network technique in the field of Machine Learning, to create a robust framework for classifying spatial patterns of evapotranspiration. Moreover, to optimize the processing of large volumes of simulated meteorological data, we propose the use of the Somoclu library, an efficient implementation of SOM that significantly accelerates training while preserving the topological structure of high-resolution data. Specifically, the proposed approach is applied to classify climatic patterns associated with ETo in Ecuador's Amazon and Andean regions. The results reveal both stable and dynamic spatial groupings of ETo, providing strategic information that can significantly contribute to improved irrigation practices and water resource planning.