Evaluating Machine Learning Approaches for Evapotranspiration Estimation in the Pampean Region of Argentina

TEYSEYRE, A.; CARMONA, F.; HOLZMAN, M.; RODRÍGUEZ, J.M.; SCHIAFFINO, S.; RIVAS, R.; GODOY, D.

Congreso; V IEEE Biennial Congress of Argentina; 2020

Evapotranspiration is an important indicator for themanagement and planning of water resources. The estimation ofevapotranspiration is usually done trough bio-physical modeling,which requires the observation of multiple variables as well asthe definition of the corresponding equations. In this paper, weevaluate the use of supervised machine learning as a strategy toget estimates of evapotranspiration from data observed in multiplemeteorological stations in the Pampean region of Argentina.Particularly, we evaluate and compare regression methods forestimating the evapotranspiration from data collected during anextensive period of time, more than 40 years, from 24 stationsplaced in the region under study. The results obtained thus farare promising as they show the feasibility of applying a machinelearning approach for obtaining accurate evapotranspirationestimates. The practical implications of these findings are relevantfor the design of more efficient water monitoring systems in thecountry