ADR1D-ML: Identifiable Parameter Inference for One-Dimensional Reactive Transport

ADR1D-ML is a reproducible machine-learning software release for inverse estimation of identifiable transport parameters from noisy six-sensor ADR1D concentration histories. Four scikit-learn pipelines estimate effective velocity, effective dispersion, decay resolvability, and a conditional first-order decay rate. The final models were fitted with 255 development scenarios and evaluated once on 45 locked test scenarios without post-test tuning. The package includes the serialized model bundle, training tables, inference and validation scripts, locked metrics, and simulation-space diagnostics. It does not claim separate recovery of velocity, dispersion, and retardation from concentration alone.