AI-driven evapotranspiration modeling for field-scale agricultural water and crop management
Theory-guided AI and machine learning platform for high-resolution evapotranspiration modeling that integrates physics-based simulations with deep learning super-resolution techniques. Enables near real-time assessment of drought and heat stress impacts at the farm scale to support agricultural water management and climate adaptation.
Overview
This solution addresses the growing challenge of managing agricultural water resources and crop production under increasing hydrologic extremes such as drought and heat stress. By combining physics-based hydrologic models with artificial intelligence and deep learning super-resolution techniques, the platform delivers high-fidelity, high-resolution estimates of evapotranspiration (ET) at the individual farm scale. The approach enables near real-time prediction of water, crop, and soil conditions, providing farmers and water managers with actionable information to mitigate climate impacts and adapt agricultural practices to shifting precipitation and temperature regimes.
Technical specifications
Key features:
- Integration of theory-guided Artificial Intelligence and Machine Learning (AI/ML) with established physics-based hydrologic models
- Deep learning-based super-resolution models that convert low-resolution model outputs into field-scale estimates of evapotranspiration and related variables
- Hybrid physics-based and machine learning methodology designed for generalizable and scientifically meaningful modeling solutions
- Incorporation of satellite-derived ground-truth data for model training and validation
- Multi-scale modeling framework operating across different spatial and temporal resolutions
- Application of recent advances in image super-resolution deep learning to environmental and hydrologic prediction
Technology readiness level
The research team has an established track record in evapotranspiration modeling, including evaluation of WRF-model irrigation methodologies and development of the VIC-MODFLOW coupled model applied to major watersheds. Current work is focused on developing and testing deep learning super-resolution models in the Mekong River basin, with ongoing integration of satellite ground-truth data. Future validation will expand hybrid physics-AI modeling approaches to produce generalizable solutions for farm-scale water and crop management. The technology is at an advanced research and validation stage, ready for collaborative refinement and pilot deployment with agricultural and water management partners.
About Virginia Tech
Virginia Polytechnic Institute and State University (Virginia Tech) is a comprehensive public land‑grant research university with broad graduate and professional programs and a strong industry orientation. An adjacent research and technology park links companies with faculty, shared labs, and prototyping resources, while facilities in the National Capital Region create a direct connection to federal partners and supply‑chain collaborators. Integration with a regional health system supports clinical research and translational pathways, and the statewide extension network enables field deployment and validation with industry and communities. Research is supported by competitive federal funding from agencies such as NSF, NIH, DOE, USDA, and the Department of Defense. A dedicated technology transfer office provides IP management, licensing, startup formation, and industry contracting support.
delivered to your inbox.