A dynamic groundwater vulnerability assessment framework for agricultural nitrate leaching: Integrating non-equilibrium transport and neural operators for precision groundwater protection.

Static groundwater vulnerability assessments often neglect transient soil-moisture dynamics, potentially underestimating nitrate leaching during extreme rainfall. This study developed a dynamic framework integrating a spatial-fractional solute transport model, a data-driven neural reaction operator, and multi-epoch geophysical resistivity measurements. A fixed 48-h standardized stress test with a 100-mm rainfall input was used to quantify acute nitrate leaching across contrasting antecedent moisture states. The fractional derivative order characterized regional departures from Fickian transport, while the neural operator represented apparent nitrate attenuation conditioned by moisture, temperature, and soil texture. Held-out laboratory evaluation showed that the operator explained 74% of the observed variation in apparent decay rates. Subsurface structure was represented by an Integrated Electrical Conductance index, and the resulting indicators were combined using moisture-conditioned entropy weights. Across the complete 39-month monitoring record, the dynamic vulnerability index was strongly associated with observed nitrate concentrations, with Pearson r values of 0.90-0.95. Independent concentration-scale validation over the final 27 months yielded an R² of 0.82, compared with 0.57 for the static baseline. Under High antecedent moisture, extending the simulation window to 120 h increased cumulative leaching by 11.1-20.0%, while sequential rainfall produced a further increase of 5.0-14.8% without changing the relative ordering of the three representative regions. The framework therefore characterizes transient nitrate-leaching vulnerability, although the fixed 48-h stress test should not be interpreted as a predictor of cumulative seasonal leaching.