GeoAI-QGIS-Asset-Refinement-Lab

AI-assisted urban resilience site selection and geospatial asset QA for a QGIS specialist portfolio.

This repository is a complete, runnable geospatial analysis project. It creates deterministic synthetic urban GIS data, validates and repairs spatial assets, runs a transparent site suitability model, exports QGIS-ready layers and styles, and produces cartographic PNG maps and Markdown reports.

Reviewer Snapshot

If you only have two minutes, review these artifacts:

• outputs/maps/04_top_candidate_sites.png for the final cartographic product.
• outputs/maps/00_portfolio_review_map.png for the fastest visual review.
• outputs/tables/top_candidate_sites.csv for ranked recommendations and QA status.
• outputs/tables/site_selection_rationale.csv for plain-English site rationales.
• outputs/reports/fellowship_reviewer_guide.md for the 60-second approval path.
• outputs/reports/ai_asset_model_card.md for the AI asset workflow framing.
• outputs/asset_manifest.json for reproducibility metadata, hashes, layer counts, CRS, and scoring configuration.
• qgis/GeoAI-QGIS-Asset-Refinement-Lab.qgs for QGIS interoperability.

Why This Fits A QGIS Specialist Role

The project demonstrates the practical work expected from a geospatial analyst supporting AI research digital assets:

• CRS-aware vector data preparation in a projected meter-based CRS.
• Geometry validation, repair, clipping, buffering, overlay, spatial joins, and proximity analysis.
• Grid-based suitability scoring with explainable weighted factors.
• GeoPackage and GeoJSON exports that open cleanly in QGIS.
• QML style files and a QGIS project file for cartographic review.
• Asset QA reporting for schema, geometry, ID, topology, and bounds issues.
• Asset lineage gates so known source defects are retained for audit but excluded from recommendations.
• Sensitivity analysis across alternate weighting schemes.
• Reproducible command-line automation and tests.

Map Outputs

Technical Stack

• Python 3.11+
• GeoPandas, Shapely, PyProj, Pyogrio
• Pandas, NumPy
• Matplotlib
• Pytest
• QGIS 3.x for optional desktop inspection

Data Model

All source data are synthetic and license-safe. The pipeline generates:

• area_of_interest
• neighborhoods
• roads_transit
• flood_risk_zones
• green_spaces
• protected_areas
• utilities
• restricted_zones
• candidate_parcels
• suitability_grid
• top_ranked_sites

The main QGIS-ready package is:

data/processed/geoai_qgis_asset_refinement_lab.gpkg

Methodology

The analysis selects candidate AI research field lab sites using normalized criteria:

Factor	Purpose	Weight
Accessibility	Near transit corridors or major roads	0.24
Utility access	Near power, fiber, or water infrastructure	0.18
Flood safety	Away from high and moderate flood-risk zones	0.22
Green proximity	Near green space, but not too close or inside protected land	0.14
Area suitability	Sufficient parcel area for field operations	0.12
Neighborhood balance	Avoids over-concentrating recommendations	0.10

Restricted zones and protected areas are treated as exclusion constraints. Flood exposure and proximity measures are computed in EPSG:26918, so distances and areas are meter-based.

Parcels with duplicate IDs or known source geometry defects are not recommended even if their repaired geometry scores well. This is intentional: an AI research asset workflow should preserve questionable source features for audit while preventing them from entering final recommendations.

Run The Pipeline

Create an environment and install dependencies:

python -m venv .venv
.venv\Scripts\activate
pip install -r requirements.txt

Run the complete workflow:

python scripts/run_pipeline.py

Or run stages individually:

python scripts/00_generate_sample_data.py
python scripts/01_preprocess.py
python scripts/02_run_suitability_analysis.py
python scripts/04_asset_qa.py
python scripts/03_export_maps.py
python scripts/05_build_qgis_project.py
python scripts/06_generate_asset_manifest.py
python scripts/validate_outputs.py

If QGIS is installed locally, generate the polished .qgz project:

"C:\Program Files\QGIS 3.44.10\bin\python-qgis-ltr.bat" scripts\07_build_qgis_project_pyqgis.py

Run tests:

pytest tests/test_geoai_pipeline.py

Open In QGIS

1. Open qgis/GeoAI-QGIS-Asset-Refinement-Lab.qgz if available. Otherwise open qgis/GeoAI-QGIS-Asset-Refinement-Lab.qgs, or open QGIS and add data/processed/geoai_qgis_asset_refinement_lab.gpkg.
2. Load layers in this order: AOI, suitability grid, flood risk, green spaces, roads/transit, candidate parcels, top ranked sites, utilities, restricted zones, protected areas.
3. Apply QML files from qgis/ to the matching layers.
4. Inspect suitability_score, site_rank, recommendation, and QA fields in the attribute table.
5. Export or modify map layouts using the generated PNG maps as reference outputs.

See docs/qgis_workflow.md for detailed QGIS steps.

Outputs

• Maps: outputs/maps/*.png
• Reports: outputs/reports/*.md
• Tables: outputs/tables/*.csv
• Asset manifest: outputs/asset_manifest.json
• GeoPackage: data/processed/geoai_qgis_asset_refinement_lab.gpkg
• GeoJSON exports: data/processed/geojson/*.geojson
• QGIS styles/project: qgis/*.qml, qgis/*.qgs, qgis/*.qgz

QA Checks

The asset QA module checks for missing CRS, invalid and empty geometries, duplicate IDs, required schema fields, out-of-bounds features, suspicious areas or lengths, naming issues, and parcel overlaps. Findings are exported to:

• outputs/reports/geospatial_asset_qa_report.md
• outputs/reports/ai_asset_model_card.md
• outputs/reports/fellowship_reviewer_guide.md
• outputs/tables/asset_qa_findings.csv
• outputs/tables/sensitivity_scenarios.csv
• outputs/tables/site_selection_rationale.csv

What This Demonstrates

For hiring reviewers, this project shows that the candidate can move beyond map screenshots and deliver a reproducible geospatial asset workflow: data engineering, spatial analysis, cartographic communication, QGIS interoperability, QA discipline, and clear reporting for AI research operations.

Limitations

The data are synthetic and designed to demonstrate method quality. A production study would add authoritative parcel, zoning, flood, utility, and mobility data; stakeholder constraints; community review; and formal uncertainty reporting.

Future Improvements

• Add raster cost-distance analysis for evacuation or service accessibility.
• Add QGIS print layout XML generation through PyQGIS when QGIS is installed.
• Add CI checks for GeoPackage layer presence and map artifact generation.