Improved prediction of daily pan evaporation using Bayesian Model Averaging and optimized Kernel Extreme Machine models in different climates

No Thumbnail Available
Ehteram M.
Graf R.
Ahmed A.N.
El-Shafie A.
Journal Title
Journal ISSN
Volume Title
Springer Science and Business Media Deutschland GmbH
Research Projects
Organizational Units
Journal Issue
Evaporation is one of the most important parameters of meteorological science. Therefore, predicting evaporation is necessary for both water resources and planning management. The present study uses Bayesian Model Averaging (BMA) based on developed and optimized Kernel Extreme Learning Machine (KELM) models for predicting daily evaporation in different provinces of Iran with different climates. The Water Strider Algorithm, Salp Swarm Algorithm, Shark Algorithm, and Particle Swarm Optimization were combined with the KELM to predict daily evaporation in the Hormozgan, Mazandaran, Fars, Yazd, and Isfahan provinces. The models� inputs were average temperature, rainfall, number of sunny hours, wind speed, and relative humidity. The introducing a new hybrid gamma test for determining the adequate inputs, using hybrid and optimized KELM based on developing ELM for predicting evaporation, integrating individual models for predicting evaporation, and quantifying the uncertainty of outputs are the main innovations of the current study. Multiple error indices were used to evaluate the ability of models for predicting evaporation. The standalone and optimized KELM models were used to predict daily evaporation in the first level. In the next level, the BMA based on outputs of standalone and optimized KELM models predicted daily pan evaporation. The general results indicated that the BMA provided the best accuracy among other models in all stations. This study also introduced the new hybrid gamma test (GT-WSA) for choosing the best input combinations. The hybrid GT-WSA gave the best input combination without computing all input combinations (25�?�1). The uncertainty analysis of models also indicated that the uncertainty of BMA and optimized KELM models was lower than that of the KELM model. � 2022, The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature.
Bayesian networks; Climate models; Forecasting; Knowledge acquisition; Machine learning; Particle swarm optimization (PSO); Uncertainty analysis; Water resources; Wind; Bayesian model averaging; Daily pan evaporation; Gamma test; Kernel extreme learning machine model; Learning machines; Machine modelling; Optimization algorithms; Uncertainty; Water planning; Water resources management; Evaporation; algorithm; Bayesian analysis; evaporation; machine learning; numerical model; optimization; stochasticity; uncertainty analysis