AGRIPINN: A PROCESS-INFORMED NEURAL NETWORK FOR INTERPRETABLE AND SCALABLE CROP BIOMASS PREDICTION UNDER WATER STRESS

Revolutionizing Crop Biomass Prediction with Process-Informed AI

AgriPINN is a novel process-informed neural network designed to accurately predict crop above-ground biomass (AGB) under water stress. It integrates a biophysical crop-growth differential equation as a differentiable constraint within a deep learning backbone. This unique approach ensures physiologically consistent biomass dynamics while maintaining scalability for large spatial domains. AgriPINN excels by recovering latent physiological variables (LAI, PAR, RUE, water-stress factors) without direct supervision, offering interpretability beyond traditional black-box models. Benchmarked against state-of-the-art deep learning models and the LINTUL5 process-based model, AgriPINN demonstrates superior accuracy (up to 43% RMSE reduction) and computational efficiency (8x faster inference). Its robustness across varying water-stress conditions and ability to provide biophysically consistent predictions make it invaluable for irrigation planning, yield forecasting, and climate-adaptation strategies in agriculture.