Methods & reproducibility

The current analytical pipeline is a deliberately simplified scaffold for testing the logic of (L_d(\varepsilon)) before empirical WUI datasets are brought fully into the workflow. Rather than treating synthetic output as a substitute for real landscape analysis, the pipeline uses toy geometry to make the structure of the problem explicit and reproducible.

This matters because the core claim of the project is methodological as much as geometric. Boundary length must be interpreted in relation to how the boundary object was defined and how the measurement scale was chosen. A small, transparent, synthetic pipeline is therefore useful not because it resolves the empirical question, but because it demonstrates the analytical wiring needed to ask that question clearly.

What follows documents the current execution pathway, the published outputs, and the local and continuous-integration routes through which figures and summary data are regenerated. The purpose of this page is practical, but it should be read in continuity with the rest of the site: reproducibility here serves interpretation, not just software maintenance.

Install dependencies and run baseline checks:

pip install -e .
pip install -r requirements.txt
python -m unittest discover -s tests -v
python scripts/run_minimal_boundary_scaling.py --include-satellite-demo

The package dependencies declared in pyproject.toml include the geospatial runtime stack required by the streaming pilot (geopandas, rasterio, shapely, pyproj, pyogrio, numpy, pandas, matplotlib, and scipy). Install the package (pip install -e .) before running scripts/run_streaming_wui_scaling.py so those imports resolve in normal execution environments.

Run the website validation gate for docs and UI-facing changes:

scripts/pre_pr_site_review.sh

Streaming real-data pilot (OSM + NLCD, no keys)

The repository now includes scripts/run_streaming_wui_scaling.py, a first empirical WUI-like pilot that pairs OSM building footprints (Overpass API) with a streamed NLCD raster subset. The workflow requires network access but no API keys or secrets.

Run a small-area pilot and publish docs-facing assets:

python scripts/run_streaming_wui_scaling.py \
  --bbox "-105.292,40.004,-105.236,40.047" \
  --outdir outputs/real_data_demo \
  --adj-buffer 250 \
  --epsilons "5,10,20,30,60,120" \
  --resolutions "30,60,90,120,150" \
  --veg-classes "41,42,43,52" \
  --publish-doc-assets

Primary run outputs are written under outputs/real_data_demo/:

fixed_boundary_scaling.csv
fixed_boundary_scaling.png
resolution_rebuild_scaling.csv
resolution_rebuild_scaling.png
run_summary.json

When --publish-doc-assets is supplied, the script refreshes docs-facing publication assets:

docs/assets/figures/real_fixed_boundary_scaling.png
docs/assets/figures/real_resolution_rebuild_scaling.png
docs/assets/data/real_fixed_boundary_scaling.csv
docs/assets/data/real_resolution_rebuild_scaling.csv

The local analysis run refreshes the published synthetic assets used by the manuscript pages:

docs/assets/figures/boundary_scaling_plot.svg
docs/assets/figures/satellite_resolution_scaling_plot.svg
docs/assets/data/boundary_scaling_summary.csv
docs/assets/data/satellite_resolution_scaling_summary.csv

For exploratory sweeps that should not overwrite website assets, use --skip-doc-publish:

python scripts/run_minimal_boundary_scaling.py \
  --output-subdir epsilon_dense \
  --min-epsilon 1 \
  --max-epsilon 40 \
  --n-steps 20 \
  --vegetation-threshold 0.45 \
  --neighborhood-radius-m 200 \
  --adjacency-rule intersects \
  --skip-doc-publish \
  --include-satellite-demo

Additional CLI options are available through:

python scripts/run_minimal_boundary_scaling.py --help
python scripts/run_streaming_wui_scaling.py --help

In pull requests, generated artifacts are attached for review in CI. Publication to GitHub Pages still occurs through the repository’s manual dispatch release workflow when maintainers decide to promote a validated revision.

Playwright-backed local review remains part of the expected validation path for website-facing changes, and scripts/pre_pr_site_review.sh is the canonical entry point for that check.

For prompt-level reconstruction of the workflow narrative and implementation sequence, see Reproducible prompts. For the synthetic-to-empirical handoff now scaffolded in code, see Real-data experiments. The staged development context for both is summarized in the Project roadmap.