Decision support tool for precision application of soil solutes

Overview

This project develops an AI-driven decision support tool to help farmers and agronomists determine the best days to apply nitrogen and other weakly bound agricultural solutes to soil. Roughly half of applied nitrogen is not taken up by crops, resulting in leaching, water pollution, and greenhouse gas emissions. Rainfall and soil moisture are major drivers of these losses, and shifting precipitation patterns are creating a feedback loop that increases waste and environmental harm. By using forecasted rainfall alongside historical patterns, the tool aims to reduce solute losses, improve fertilizer efficiency, and support more sustainable arable farming under increasingly variable weather conditions.

Technical specifications

Key features:

• Uses validated crop models to generate synthetic datasets across hundreds of real rainfall patterns, overcoming the impractical replication needs of field trials
• Employs a neural network trained on past and forecasted rainfall to predict suitable application days for nitrogen and other solutes
• Extends a previously developed nitrogen fertilization model that captured the relationship between cumulative rainfall and leaching
• Benchmarked against experimental data for each solute to ensure predictive accuracy
• Planned delivery as a web application for practitioner testing, with industry guidance on solute selection, scenario coverage, and usability

Technology readiness level

The underlying fertilization timing model has already been validated in a prior study, demonstrating the ability to determine optimal application timings across multiple growing seasons. The current project advances the technology by broadening coverage to additional solutes, training the neural network on expanded synthetic datasets, and benchmarking predictions against experimental data. The next stage involves building a web application for field testing by practitioners, placing the solution at an early-to-mid development stage with active validation underway.