Regresar

Evaluation of Evapotranspiration Classification Using Self Organizing Maps and Weather Research and Forecasting Variables

Abstract:

Remotely sensed data and Artificial Intelligence provide an avenue to Standard Evaporation (ETo) classification. Hence, we developed a methodology for classifying ETo using Self-Organizing Maps on data from the Weather Research and Forecasting (WRF) model obtained from March to October, 2020, at 3 km resolution over Ecuador. This is a preliminary effort to provide a visual comparison of ETo classified maps under different scenarios: a 'shallow' learning that takes only data from one time slice at 12:00 noon for one and three months; a 'deeper' learning that considers hourly slices for three months; raw and pre-processed WRF variables; and different numbers of classified classes. The results of these exercises show overall stable mountainous classification, while the Amazonia and Western regions appear more volatile. Moreover, the 'deeper' learning produces a more stable ETo classification in all regions with a decreased migration of classes over time, as presented at http://www.yachay.openfabtech.org/somETo/description.php.