This example runs one external comparison experiment without changing the simulation workflow. The comparison TOML generates two child simulation TOMLs from one shared physical case, runs them through hmp run, then writes metrics, audit files, reports, and figures.

Synthetic shared-mesh run:

python examples/projects/09_comparison_workflow/run_comparison_example.py --case synthetic --show

Natural catchment run:

python examples/projects/09_comparison_workflow/run_comparison_example.py --case natural --show

Reduced natural-mesh Boussinesq run:

python examples/projects/09_comparison_workflow/run_comparison_example.py --case natural-bouss --show

Reduced natural-mesh Boussinesq run with recharge:

python examples/projects/09_comparison_workflow/run_comparison_example.py --case natural-bouss-recharge --show

Reduced natural-mesh transient recharge-pulse run:

python examples/projects/09_comparison_workflow/run_comparison_example.py --case natural-bouss-transient-pulse --show

Nancon transient seasonal recharge run:

python examples/projects/09_comparison_workflow/run_comparison_example.py --case nancon-seasonal --show

Nancon transient seasonal recharge run with observed hydrography loaded:

python examples/projects/09_comparison_workflow/run_comparison_example.py --case nancon-seasonal-hydrography --show

Nancon monthly MF6/Boussinesq run with one precomputed mesh and aligned transient observables:

python examples/projects/09_comparison_workflow/run_comparison_example.py --case nancon-monthly-bouss-comparable --show

Nancon MF6-only hydraulic-conductivity sweep with observed hydrography loaded:

python examples/projects/09_comparison_workflow/run_comparison_example.py --case nancon-seasonal-hydrography-k-sweep-mf6 --show

Run all examples:

python examples/projects/09_comparison_workflow/run_comparison_example.py --case all --show

The same comparison TOMLs can also be executed through the public CLI:

hmp run examples/projects/09_comparison_workflow/compare_dupuit_mf6_bouss.toml
hmp run examples/projects/09_comparison_workflow/compare_vire_natural_mf6_nwt.toml
hmp run examples/projects/09_comparison_workflow/compare_10km2_natural_mesh_mf6_bouss.toml
hmp run examples/projects/09_comparison_workflow/compare_10km2_natural_mesh_recharge_mf6_bouss.toml
hmp run examples/projects/09_comparison_workflow/compare_10km2_natural_mesh_transient_pulse_mf6_bouss.toml
hmp run examples/projects/09_comparison_workflow/compare_nancon_transient_seasonal_mf6_bouss.toml
hmp run examples/projects/09_comparison_workflow/compare_nancon_transient_seasonal_hydrography_mf6_bouss.toml
hmp run examples/projects/09_comparison_workflow/compare_nancon_transient_monthly_mf6_bouss_comparable.toml
hmp run examples/projects/09_comparison_workflow/compare_nancon_transient_seasonal_hydrography_k_sweep_mf6_only.toml

Synthetic outputs are written under:

examples/projects/09_comparison_workflow/outputs/dupuit_mf6_vs_bouss

Natural outputs are written under:

examples/projects/09_comparison_workflow/outputs/vire_natural_mf6_vs_nwt

Reduced natural-mesh Boussinesq outputs are written under:

examples/projects/09_comparison_workflow/outputs/natural_mesh_10km2_mf6_vs_bouss

Reduced natural-mesh recharge outputs are written under:

examples/projects/09_comparison_workflow/outputs/natural_mesh_10km2_recharge_mf6_vs_bouss

Reduced natural-mesh transient-pulse outputs are written under:

examples/projects/09_comparison_workflow/outputs/natural_mesh_10km2_transient_pulse_mf6_vs_bouss

Nancon transient seasonal outputs are written under:

examples/projects/09_comparison_workflow/outputs/nancon_transient_seasonal_mf6_vs_bouss

Nancon transient seasonal hydrography outputs are written under:

examples/projects/09_comparison_workflow/outputs/nancon_transient_seasonal_hydrography_mf6_vs_bouss

Nancon monthly MF6/Boussinesq comparable outputs are written under:

examples/projects/09_comparison_workflow/outputs/nancon_transient_monthly_mf6_bouss_comparable

Nancon MF6-only K-sweep outputs are written under:

examples/projects/09_comparison_workflow/outputs/nancon_transient_seasonal_hydrography_wide_k_sweep_mf6

Key files:

• comparison_manifest.json
• comparison_report.md
• comparison_audit.md
• observables.csv
• comparison_metrics.csv
• comparison_differences.csv
• hydrographic_network_metrics.csv when the compared runs expose both canonical hydrographic networks (reference and generated)
• simulated_active_network_overlap_metrics.csv when the runs expose accumulation_flux, a plottable mesh, and the observed reference network
• comparison_figures/*.png

Open comparison_figures/case_configuration.png first. It summarizes the reference support used by the comparison: mesh geometry or cell centroids, topography when available, detected fixed-head side boundaries, point/outlet observables, and the recharge chronicle.

Then open the *triptych*.png figures. They show the reference head field, the candidate head field, and the candidate-minus-reference difference in one panel.

Once a comparison has been run, the post-run stability checker verifies that the materialized outputs still satisfy explicit thresholds. It does not launch any solver; it reads comparison_manifest.json, comparison_audit.json, comparison_metrics.json, and selected figures.

python examples/projects/09_comparison_workflow/check_comparison_stability.py

The default target file is:

examples/projects/09_comparison_workflow/stability_targets.toml

Run one stabilized case only:

python examples/projects/09_comparison_workflow/check_comparison_stability.py --case natural_mesh_10km2_transient_pulse_mf6_vs_bouss

The first stabilized cases are:

• dupuit_mf6_vs_bouss: compact synthetic shared-mesh MF6/Boussinesq lock.
• natural_mesh_10km2_transient_pulse_mf6_vs_bouss: natural transient pulse lock with known audit warnings kept visible.
• nancon_transient_monthly_mf6_bouss_comparable: controlled Nancon natural benchmark with one precomputed mesh, aligned monthly times, lower initial heads, and stronger top drainage.
• nancon_transient_seasonal_hydrography_mf6_vs_bouss: broad Nancon stress test lock. Its thresholds are intentionally loose because the case still exposes large configuration-sensitive MF6/Boussinesq differences.

The controlled Nancon target is the preferred natural MF6/Boussinesq benchmark. The older seasonal Nancon target remains a broad stress sentinel: it flags sudden regressions while keeping its current audit warnings and large head discrepancies visible.