Geostatistical active learning for expanding monitoring networks for environmental decision-making: The case of groundwater nitrate

Expanding environmental monitoring networks is a spatial sampling problem often addressed with pointwise uncertainty heuristics that ignore the neighborhood impact of new sites and the decision tasks motivating monitoring. We recast network densification as active learning and propose model-agnostic acquisition criteria, including a decision-aligned focal logit utility that prioritizes neighborhoods whose exceedance probabilities lie near regulatory thresholds. A look-ahead criterion based on the expected drop in prediction standard error (SE) is also examined. In a groundwater nitrate case study, the focal logit criterion consistently selected more informative sites than dispersion or SE heuristics, yielding up to 58 % greater gains in exceedance-mapping accuracy (Cohen’s κ ). Focal logit and SE criteria outperformed pointwise counterparts by ∼ 45 % on average, while the look-ahead criterion performed well at much higher computational cost. The framework is simple, generalisable to other pollutants, and supports transparent, decision-oriented monitoring design.