Hydrochemical data were utilized in this study to elucidate the hydrogeochemical characteristics and genesis of high-arsenic groundwater in the North China Plain, employing both traditional hydrogeochemical analysis and unsupervised machine learning techniques. The findings indicate that the predominant hydrochemical types of groundwater in the study area are HCO₃-Ca·Na and SO₄·Cl-Na·Ca. The primary mechanism influencing groundwater chemistry has been identified as rock weathering. The unsupervised machine learning framework incorporates various methods, such as principal component analysis (PCA), non-negative matrix factorization (NMF), machine learning models (gradient boosting trees and random forests), and cluster analysis to explore the characteristics and genesis of groundwater hydrochemical types within the study area. This study demonstrated that the formation mechanism of high-arsenic groundwater results from multiple interacting factors.

Shallow aquifers in intensively managed alluvial plains worldwide are increasingly impacted by inorganic nitrogen, yet the simultaneous occurrence and interconversion of nitrate (NO₃⁻–N), nitrite (NO₂⁻–N) and ammonium (NH₄⁺–N) often confound source attribution when single indicators are used. Here, we present a transferable, process-linked framework for diagnosing “tri-nitrogen” (tri-N) pollution that integrates hydrogeochemical evolution, data-driven pattern discovery and receptor-model apportionment. We analyzed 409 shallow-groundwater samples from Shijiazhuang City (central North China Plain) for major ions and tri-N species, interpreted within Piper facies and salinization gradients, and then applied a Gaussian mixture model (GMM) to resolve multivariate hydrochemical–nitrogen end-members. Six clusters (I–VI) depict an interpretable progression from background Ca–HCO₃/Ca·Mg–HCO₃ waters to agricultural NO₃⁻–N enrichment under oxic conditions and a distinct NH₄⁺–N-rich point-source end-member under reducing conditions. An attention-based attribution model indicates that total tri-N, Na⁺, NO₃⁻–N, the NO₂⁻ fraction and SO₄²⁻ are the primary discriminators of cluster structure. Species-resolved positive matrix factorization (US EPA PMF 5.0) quantifies dominant controls, with agricultural leaching–nitrification explaining most NO₃⁻–N (Factor 6, 87.9%) and sewage/manure inputs dominating NH₄⁺–N (Factor 3, 95.3%), while NO₂⁻–N reflects mixed contributions consistent with redox-interface transitions. Beyond this case study, the combined GMM–interpretability–PMF workflow provides a general template for separating non-point versus point tri-N inputs and for prioritizing management actions in shallow aquifers where isotope or tracer data are limited.

Fluoride (F⁻) contamination in deep groundwater threatens drinking water security, yet its enrichment is commonly governed by coupled nonlinear hydrogeochemical feedbacks that are difficult to resolve with linear diagnostics alone. Here, we integrate an explainable deep learning framework (HydroAttentionNet + SHAP) with thermodynamic and mass-conservative inverse modeling (PHREEQC) to quantitatively link data-driven thresholds to mineral water processes in a multi-aquifer system. Using 258 deep-well samples, we delineate a robust evolution pathway from background to ultra-high-fluoride (Ultra-High F⁻, ≥1.5 mg/L) waters. HydroAttentionNet achieves strong predictive skill (R² = 0.77) and reveals a clear mechanistic tipping behavior: alkalinity (HCO₃⁻/CO₃²⁻) is the primary trigger for F⁻ activation, while progressive Na+ enrichment and Ca²⁺ depletion act as amplifiers by suppressing a(Ca²⁺) and weakening fluorite precipitation capacity. PHREEQC simulations confirm a coupled “salinization–decalcification–fluoridation” loop in which (i) evaporite dissolution elevates ionic strength (salt effect) and supplies Na⁺ to promote Na–Ca exchange, and (ii) carbonate re-equilibration drives calcite precipitation as an efficient Ca sink, offsetting ~45.8% of Ca²⁺ inputs; together, these processes maintain fluorite undersaturation and sustain net fluorite dissolution, contributing 56.6% of newly added dissolved F⁻ in evolved end-members. Monte Carlo health risk assessment (10,000 iterations) indicates substantial intergenerational inequity: 67.9% of children exceed the non-carcinogenic risk threshold (HQ > 1), compared with 29.3% of adults. Sensitivity analysis identifies source-water fluoride concentration as the dominant driver (Spearman r = 0.93), implying that supply-side interventions (defluoridation, well-screen optimization, and blending with low-F sources) are substantially more effective than behavioral measures.