A Framework for Constructing Machine Learning Models with Feature Set Optimisation for Evapotranspiration Partitioning

TL;DR

This paper presents a comprehensive framework for selecting optimal machine learning models and features to improve evapotranspiration partitioning, revealing that no single model or feature set is best across different wetlands, and highlighting methane flux as a potentially important feature.

Contribution

The work introduces a systematic framework for model and feature selection in evapotranspiration modeling, including the novel insight into methane flux's role.

Findings

No single best model or feature set across all sites.

Methane flux may significantly influence evapotranspiration predictions.

Feature importance varies by site and model.

Abstract

A deeper understanding of the drivers of evapotranspiration and the modelling of its constituent parts (evaporation and transpiration) could be of significant importance to the monitoring and management of water resources globally over the coming decades. In this work, we developed a framework to identify the best performing machine learning algorithm from a candidate set, select optimal predictive features as well as ranking features in terms of their importance to predictive accuracy. Our experiments used 3 separate feature sets across 4 wetland sites as input into 8 candidate machine learning algorithms, providing 96 sets of experimental configurations. Given this high number of parameters, our results show strong evidence that there is no singularly optimal machine learning algorithm or feature set across all of the wetland sites studied despite their similarities. A key finding discovered when examining feature importance is that methane flux, a feature whose relationship with evapotranspiration is not generally examined, may contribute to further biophysical process understanding.