To address the challenge of balancing food production, groundwater protection, and economic sustainability in agricultural systems, this study developed a novel multi-objective optimization framework integrating an Extreme Learning Machine (ELM)-based surrogate model. An irrigation–fertilization–economy (IFE) model was established to evaluate region-specific agricultural adjustment strategies under four scenarios: baseline adjustment, flexible restructuring, production reduction and profit-oriented expansion scenarios. The optimization results were subsequently used to assess the impacts of agricultural production regulation on groundwater nitrate concentrations in the study area. The results showed that the IFE model consistently favored the conversion from wheat to maize cultivation across all regions, because maize requires less irrigation and fertilizer inputs while maintaining relatively high net profits. However, the optimal strategies varied among regions: Pingdu and Jiaozhou showed greater economic potential for production expansion, whereas regions with lower economic potential were more suitable for reducing high-input crop cultivation. During the 2020–2030 prediction period, nitrate concentrations exhibited substantial interannual variations. Reducing the overall production scale shifted nitrate concentrations approximately 6% closer to the 30–50 mg/L range over the decade, but this limited environmental benefit was accompanied by substantial short-term economic losses. Considering both environmental impacts and economic viability, the optimization results suggest that moderate agricultural expansion could be acceptable provided that environmental burdens are effectively managed.

Phosphogypsum (PG) stacks are traditionally assessed as sources of legacy inorganic contaminants, but the behavior of emerging contaminants in these chemically complex systems remains poorly understood. This opinion article proposes that PFAS, if present in PG stacks, may not be adequately described by partitioning concepts derived from dilute groundwater or ordinary soil porewater systems. Instead, the low-pH, high-ionic-strength, and calcium–sulfate-rich conditions of PG leachate may promote hydrochemistry-mediated repartitioning of PFAS. Under such conditions, PFAS may exhibit reduced apparent aqueous stability, enhanced association with PG particles or colloids, retention on particle surfaces, and enrichment at air–water interfaces, forming potential hidden reservoirs with the potential for delayed release and episodic remobilization. Consequently, dissolved concentrations alone may underestimate total PFAS storage and long-term groundwater risk in and around PG stack systems. Overall, this study highlights the need to shift from conventional dilute-system assumptions toward a hydrochemically mediated multiphase framework for PFAS occurrence assessment, monitoring design, and risk evaluation in phosphogypsum environments and other chemically complex industrial waste systems.