Modeling of Pan Evaporation Based on the Development of Machine Learning Methods

TL;DR

This paper explores the use of various machine learning models, especially the weighted K-nearest neighbor, to accurately estimate monthly pan evaporation in different Iraqi regions, aiding water resource management.

Contribution

It introduces and evaluates new ML models for evaporation estimation, demonstrating improved accuracy over existing models in diverse climatic conditions.

Findings

Weighted K-nearest neighbor outperforms other models in accuracy.

ML models effectively capture non-linear evaporation processes.

The approach aids water resource planning in drought-prone regions.

Abstract

For effective planning and management of water resources and implementation of the related strategies, it is important to ensure proper estimation of evaporation losses, especially in regions that are prone to drought. Changes in climatic factors, such as changes in temperature, wind speed, sunshine hours, humidity, and solar radiation can have a significant impact on the evaporation process. As such, evaporation is a highly non-linear, non-stationary process, and can be difficult to be modeled based on climatic factors, especially in different agro-climatic conditions. The aim of this study, therefore, is to investigate the feasibility of several machines learning (ML) models (conditional random forest regression, Multivariate Adaptive Regression Splines, Bagged Multivariate Adaptive Regression Splines, Model Tree M5, K- nearest neighbor, and the weighted K- nearest neighbor) for modeling the monthly pan evaporation estimation. This study proposes the development of newly explored ML models for modeling evaporation losses in three different locations over the Iraq region based on the available climatic data in such areas. The evaluation of the performance of the proposed model based on various evaluation criteria showed the capability of the proposed weighted K- nearest neighbor model in modeling the monthly evaporation losses in the studies areas with better accuracy when compared with the other existing models used as a benchmark in this study.