# MODFLOW problem-cell map Measured by `scripts/map_modflow_problem_cells.py`: 3,408 G2 upper-layer cells exceed terrain by >0.01 m at least once during 2003–2019; peak excess 8.509 m. The separate historical layer contains 655 cells at its final saved time (2923 days). These are different run configurations and time summaries, not a controlled before/after experiment. P01: later G2 modeled head/terrain exceedance. The source interpretation describes finite-resistance seepage boundaries; exceedance is not observed ponding depth or demonstrated artesianism. Physical applicability remains unresolved. P02: historical standalone high heads. P03: detailed historical r140,c197 diagnosis: perimeter neighbour to the north is inactive; missing northern connection is a constraint, not proof of the repair required. See the original [geometry finding](../../../docs/records/model-domain-review-2026-09-23/lowland_mf6_candidate/high_head_geometry_probe/findings.md) and [later interpretation](../../../docs/records/model-domain-review-2026-09-23/production_adoption/coupling/g2_head_interpretation/g2_vs_g1_interpretation.json). Open index.html for the interactive map; internet is required for Leaflet and street/terrain tiles. The PNG/PDF and GeoPackage work offline. In QGIS add the GeoPackage; IDs and descriptions are attributes in the problem layers. Row/column labels are zero-based and layer labels are one-based. Grey grid is the active upper layer; unhighlighted cells are not proven physically valid. Reproduce from the repository root: `PYTHONPATH=diagnostics/coupled-improvement-2026-09-14/scratch/runtime .venv/bin/python scripts/map_modflow_problem_cells.py` Input hashes and recomputed counts are in map_receipt.json. No model was rerun or changed. ## Steady-state and border diagnostic addition The generator also runs `scripts/diagnose_modflow_border.py`. See [border diagnosis](border_diagnosis.md), [its measured receipt](border_diagnosis_receipt.json), and [head/drain comparison PDF](steady_head_border_comparison.pdf). Both native layers have GeoTIFF head exports. The GeoPackage active grid now contains head, terrain, drainage and hydraulic-property attributes, plus a separate contour layer. These heads are the synthetic mean-stress initial solution; they are not observed piezometry.