A Data Analytics Framework for Rainfall-driven Water Quality and Nutrient Retention Risk Assessment in Farm Systems

Purpose: Agricultural systems face increasing pressure from rainfall variability disrupting nutrient dynamics in managed soils and surface water, reducing fertil- izer efficiency and increasing the risk of nutrient export to drainage systems. This study aims to develop, implement and evaluate a Nutrient Retention and Dilu- tion Index (NRDI), a data analytics framework that integrates satellite-derived environmental data from Google Earth Engine with high-frequency surface water quality measurements from the North Wyke Farm Platform (NWFP), Devon, UK, to produce a daily nutrient transport risk classification for farm management decision support. Methods: The NRDI was computed as a weighted composite index: NRDI = 0.4 × rain norm + 0.4 × nitrate load norm + 0.2 × ndvi retention, applied to 1,522 daily observations spanning 2019–2024. The framework was validated through Pearson correlation analysis, K-Means clustering, event-based validation of 14 High-risk events and sensitivity analysis confirming classification agreement across weight perturbation scenarios. Results: NRDI correlated strongly with 3-day rainfall (r = 0.922, R2 = 0.849 for combined inputs) and the NOx-N load proxy (r = 0.795). K-Means clus- tering achieved a 0.463 silhouette score and 95.3% accuracy, isolating all 14 High-risk events into a single coherent cluster, whereas a rainfall-only baseline identified only 14.3%. Classification agreement exceeded 89.3% (κ ≥ 0.782) across perturbations, while a precise 7/7 event split empirically confirmed the equal component weighting. Conclusion: This framework demonstrates the first composite index integrating satellite-derived rainfall, vegetation activity, and catchment NOx-N measure- ments into a validated daily nutrient transport risk framework, providing a transferable, interpretable framework for evidence-based fertilizer application.