Acid mine drainage (AMD) poses significant environmental and health risks due to its high acidity and elevated metal and sulfate contents. Previous studies have primarily focused on short-term AMD monitoring, with limited attention paid to long-term, spatially resolved datasets and predictive modeling. In this 3.5-year study, six wells down-stream of a mine waste rock pile were monitored, and 132 sets of associated water quality (AWQ), geological (GEO), and climate history (CH) parameters were compiled to develop predictive models for Fe, Cu, and Zn concentrations. Random forest (RF), extreme gradient boosting (XGBoost), and support vector machine (SVM) algorithms were applied using different combinations of input variables. The combined AWQ-GEO-CH dataset achieved the best overall performance, with XGBoost yielding the highest R² values for Fe (0.81) and Cu (0.77), and SVM performing best for Zn (0.94). CH variables, particularly precipitation and evaporation over 60-day periods, strongly influenced metal concentrations by driving hydrological and solute redistribution processes. AWQ parameters, especially F⁻ and S²⁻, were key predictors for Fe and Zn and ranked second for Cu, likely due to shared upstream sources and coupled geochemical processes such as FeF₃ dissolution. The most impactful GEO factor was the installation of a vertical barrier, which reduced metal concentrations by 73–80%. These findings highlight the value of integrating multi-source datasets with ML for long-term AMD prediction and management.

Carbonate rock formations constitute common hydrocarbon reservoirs and are often considered as candidates for geological CO2 storage, where acid-driven carbonate dissolution may occur in the near-wellbore region. Calcite dissolution can substantially reconfigure pore networks, altering permeability and influencing storage efficiency and long-term containment integrity. While carbonate dissolution has been extensively studied experimentally and numerically, its detection and characterization using non-invasive monitoring tools remain challenging. Nuclear magnetic resonance (NMR) is a particularly promising tool as it is sensitive to pore geometry and fluid distribution. However, a quantitative framework that links regime-dependent pore-scale dissolution patterns to NMR observables remains underdeveloped. In this work, we establish such structure–signal mapping by coupling pore scale reactive transport simulations of calcite dissolution with forward modeling of low-field NMR responses, generating synthetic observables from dynamically evolving pore geometries due to calcite dissolution. By varying the relative timescales governing advection, diffusion, and surface reaction rates, we analyze the evolution of three representative dissolution patterns: uniform face dissolution, conical channeling dissolution, and wormholing dissolution. To capture the spatial heterogeneity of these features, we segment the pore geometry along the flow axis and derive an NMR T2 relaxation time distribution for each section, constructing flow-direction T2 profiles. In contrast, bulk T2 distributions derived from the entire pore volume tend to average out the spatial heterogeneity of dissolution patterns. Furthermore, to capture the propagation of reaction fronts and characterize the permeability of emerging channels, we formulate specific NMR-based metrics: a pore-enlargement index Ei(t), a heterogeneity index H(t), and a connectivity index C(t). Dissolution breakthrough, defined by k/k0 ≥ 10, occurs at PV10 ≈ 314 for face dissolution, 138 for channeling, and 144 for wormholing. While H(t) consistently evolves non-monotonically, breakthrough is governed by the emergence and strengthening of an inlet-to-outlet pathway. Accordingly, C(t) closely tracks breakthrough during channeling, whereas in wormholing it indicates early connectivity without an immediate permeability increase. Our weighted pore network connectivity by cumulative enlargement yields a single metric that correlates with permeability growth across regimes. This structure–signal framework provides a workflow for using spatially distributed NMR signals to identify pathway formation and provide an early indication of permeability surges. The framework for mapping structures to signals enhances the interpretation of NMR signals in dissolution reactive settings and provides a quantitative foundation for interpreting NMR monitoring signals and informing risk assessment for geological CO2 storage in settings where carbonate dissolution may alter flow pathways.

Mineral dissolution reorganizes pore space through spatiotemporally variable pathways, but non‐invasive tracking remains difficult. We test whether nuclear magnetic resonance (NMR) signatures can diagnose carbonate dissolution by integrating pore‐scale reactive transport simulations with forward modeling of distributions for face dissolution, wormholing, and channeling regimes. Simulated pathways reflect evolving pore coupling between residual matrix pores and macropores formed by dissolution. Face dissolution evolves from single peak to separated peaks and back to single peak as coupling weakens and the matrix is consumed. Wormholing maintains a broad single peak as competing pathways merge, whereas channeling shows transient bimodality followed by rapid peak merging as a dominant conduit expands. A matrix–macropore connectivity index derived from spectra, together with flow‐field tortuosity, shows that earlier matrix–macropore coupling coincides with faster tortuosity reduction and earlier permeability breakthrough at lower porosity. These results support quantitative NMR signatures as indicators of dissolution dynamics.