TDAI Catalyst Grant Fuels Breakthrough in Explainable Machine Learning for Agricultural Water Management

A new study published in Agricultural Water Management demonstrates how explainable machine learning (ML) can improve the accuracy and interpretability of latent energy (LE) flux estimation—a critical factor in understanding agricultural water use. The research, led by TDAI Faculty Affiliate Member Darren Drewry, was supported by a TDAI Catalyst Grant, which provided funding to advance the project from concept to peer-reviewed publication.

About the Study

The research team applied an explainable machine learning (eXML) framework to evaluate how different predictor variables and remote sensing data can be used to model field-scale water use. The study developed and cross-validated 64 machine learning model sets for each crop type, using combinations of climate variables and proximal remote sensing indicators such as land surface temperature (LST) and normalized difference vegetation index (NDVI).

By comparing advanced ML architectures—extreme gradient boosting, random forests, and neural networks—the team demonstrated that a small, carefully selected set of variables can accurately predict LE flux while maintaining transparency in model outputs. This addresses a key challenge in agricultural modeling: building models that are not only powerful, but also explainable to researchers, policymakers, and farmers.

Why It Matters

Accurately estimating water use is essential for optimizing irrigation, conserving resources, and adapting agricultural practices to changing climate conditions. By combining interpretability with predictive power, this work lays the foundation for decision-support tools that can guide sustainable water management across diverse cropping systems.

TDAI’s Role

The project was made possible through a TDAI Catalyst Grant, designed to spark new interdisciplinary research collaborations and accelerate high-impact, data-driven science at Ohio State. Catalyst Grants provide seed funding that enables faculty to test innovative ideas, gather preliminary data, and compete for larger external funding opportunities.

📄 Read the full study: Explainable machine learning to quantify the value of proximal remote sensing in latent energy flux estimation (Agricultural Water Management)