AI, Space and Agriculture

Topic: AI, Space and Agriculture

When: 12 February at 12.00-13.00

Where: Online

Speaker: Alexandros Sopasakis, Senior lecturer, Numerical Analysis and Scientific Computing, Lund University

Spoken language: English

Abstract

Predicting crop yield with precision is critical to enhancing agricultural productivity, safeguarding food supplies, and strengthening economic stability. We introduce a Graph Neural Network (GNN) framework that predicts yield variations in winter wheat fields by integrating multi-spectral Sentinel satellite data with soil, weather, and harvest information. The GNN embeds temporal dynamics within its graph architecture, addressing challenges like cloud cover variability and high-dimensional inputs, while delivering detailed yield predictions at a 10m × 10m resolution empowering precision interventions.

Though absolute yield forecasting remains challenging due to weather uncertainties, the model excels in early prediction of yield patterns, identifying high- and low-yield zones. This supports sustainable farming through better resource allocation, reduced fertilizer waste, and minimized environmental impact. By combining advanced AI with practical applications, this scalable solution transforms agriculture promoting sustainability.