IAHR Document Library

« Back to Library Homepage « Proceedings of the 40th IAHR World Congress (Vienna, 2023)

Exhaustive Optimized Machine Learning, Modeling, and Analysis of Evapotranspiration

Author(s): Mohammad Zeynoddin; Silvio Jose Gumiere; Hossein Bonakdari

Linked Author(s): Hossein Bonakdari

Keywords: SVM; GMDH; ET; Lag forecast; Teacher learner; GEE; Holt-winters

Abstract: The impact of climate change on water resources management is a significant concern, trending in recent years. Evapotranspiration (ET) is an important process in the water balance of watersheds and a critical factor in water management studies. However, it is not widely measured due to technical and financial constraints. This study uses model-based and remote-sensed data to overcome the challenge of acquiring ET data. It examines two modeling approaches, time series-based and variable-based, using Support Vector Machine (SVM) and Group Method of Data Handling (GMDH) models. To estimate ET data, exogenous (Exo) variables like precipitation (PrpT) and air temperature (ATemp), and different lag inputs of ET are used. The sensitivity analysis and other numerical/graphical accuracy measures of selected inputs revealed that The SVM (Lags=[1,2, 3,4, 5,23]) model was more successful in capturing the extremes of the ET with R = 0.874, RMSE = 36.953, MARE = 0.201, ubRMSE = 36.953. The SVM ([PrpT, ATemp]) overestimated the majority of ET values, while the residuals of the SVM (Lags) are mostly uniform around the baseline and independently dispersed. The results of GMDH are also very close to the SVM (Lags) model. The GMDH and SVM (Lags) more successfully estimated the real data's mean, median, interquartile area, and peaks. However, the SVM (Exo) model was slightly more successful in evaluating the lows of ET.


Year: 2023

Copyright © 2024 International Association for Hydro-Environment Engineering and Research. All rights reserved. | Terms and Conditions