Spatio–Temporal Graph Neural Networks for Dairy Farm Sustainability Forecasting and Counterfactual Policy Analysis

Efficiently steering dairy production toward environmental viability requires models that jointly exploit spatial, temporal, and management information. This study introduces a novel data-driven framework and the first-ever county-scale application of Spatio-Temporal Graph Neural Networks (STGNN) to forecast composite sustainability indices from herd-level operational records.The methodology employs a novel, end-to-end pipeline utilizing a Variational Autoencoder (VAE) to augment Irish Cattle Breeding Federation (ICBF) datasets, preserving joint distributions while mitigating sparsity. A first-ever pillar-based scoring formulation is derived via Principal Component Analysis, identifying Reproductive Efficiency, Genetic Management, Herd Health, and Herd Management, to construct weighted composite indices. These indices are modelled using a novel STGNN architecture that explicitly encodes geographic dependencies and non-linear temporal dynamics to generate multi-year forecasts for 2026–2030.This novel model achieves high predictive skill with a validation coefficient of determination above 0.91, significantly outperforming Gaussian Kernel Regression (GKR), Long Short-Term Memory (LSTM), Recurrent Neural Network (RNN) and Feedforward Neural Network (FFNN) baselines. Furthermore, first-ever counterfactual scenario analyses illustrate how targeted interventions in fertility and culling strategies can realistically shift long-term sustainability pathways in representative counties like Cork and Kerry. This framework successfully links herd-level management decisions to regional sustainability outcomes, offering a novel, transferable methodology for policy design. By enabling first-ever spatio-temporal stress-testing of operational levers, the study provides robust, actionable decision support for improving the environmental footprint of livestock systems.