Cities for Children: Beer Sheva streetscape accessibility

View the live map

This project maps how easy it is to reach everyday services and child-relevant features on foot from residential buildings in Beer Sheva. It is a collaboration between Urban95 and NUR under the “Cities for Children” initiative.

---

What you get

• Interactive map (docs/) — MapLibre GL, with optional deck.gl clusters for dense points and Chart.js for dashboards.
• Two scoring models — Urban95 uses equal-mean traffic-light statuses; Amenities Focus (internally expanded) uses a broader amenity-count classification.
• Three scales — Single building (with walking-time area), neighborhood comparison, and citywide summary.

---

Run the map locally

From the repository root (serves the whole repo so ./data resolves next to docs/index.html):

pip install -r requirements.txt
npm install
npm run start

Open http://localhost:8080/docs/index.html

To serve only the docs/ folder (paths like ./data will break unless you mirror files), use npm run start:docs only if you adjust asset URLs accordingly.

---

Data layout

Processed layers for the website live under docs/data/ (GeoJSON plus neighborhood_charts.json and citywide_stats.json). The app loads them with BASE = "./data" in docs/app.js.

Typical files include building footprints with accessibility fields, amenity layers (amenities_new.geojson, amenities_all.geojson), trees, parks, street lights, precomputed walking isochrones, and neighborhood boundaries. Exact filenames must match docs/app.js.

Building fields used by the app

• Urban95 overview status: u95_status_*.
• Urban95 category status: u95_status_<category>_*.
• Urban95 direct indicator status: u95_status_sub_<category>_<indicator>_*.
• Urban95 diagnostic status: u95_status_detail_<category>_<parent>_<child>_*. School and Kindergarten are Education diagnostics; Clinic and Tipat Halav are Health diagnostics. They are visible in drill-downs but excluded from means.
• Amenities Focus score: score_expanded_*.

* is one of _5min, _10min, _15min; the frontend currently consumes _10min. Urban95 publishes status strings, raw explanation measurements such as summer_si, and no numeric overview total.

---

Scoring methodology

Urban95 (equal-mean traffic-light statuses)

Urban95 preserves the existing spatial indicator rules but presents every public Urban95 result as Disappointing, Functioning, Thriving, or Unknown. It does not publish, display, rank, or reconstruct a numeric Urban95 overview total.

The hierarchy is Environmental Quality (Shade, Trees, Distance from fast roads), Nature (Parks, Urban nature areas), Play (Playgrounds), Safety & Mobility (Street lights, Bicycle access, Bus stops, Shelters), and Family Services (Education, Community centers, Business centers, Health). Direct indicators retain their project rules and internal 0/0.5/1 attainment levels. Sibling direct indicators contribute equally to their category; the five categories contribute equally to the overview. Thus categories with more indicators do not have extra influence.

Category and overview means use the following status cutoffs: attainment below 0.25 is Disappointing, 0.25 through below 0.75 is Functioning, and 0.75 through 1.00 is Thriving. Exact midpoint ties classify upward. If any required direct child is Unknown, its category is Unknown; if any category is Unknown, the overview is Unknown. Unknown is never replaced with zero or omitted from a denominator.

Missing, unreadable, schema-invalid, or failed source evidence produces Unknown for dependent results. A valid empty source or valid empty subtype remains available and follows the normal indicator rule. Per-building calculation failures produce Unknown only for that building. Area summaries count all four statuses and use the uniquely predominant building status only; ties and unsupported areas publish Unknown with a summary reason rather than a hidden average.

Environmental Quality / Shade — Beer Sheva BDAR Spatial Shade Index (summer_SI) comes from the Derech Tzel shading metrics guide. Values are used as-is (SI, not SAI; not recalculated). Each building gets a 300 m area-weighted mean summer_SI around the building footprint (near-edge buffer), then stored/displayed summer_si is rounded to 1 decimal place with standard half-up ties before output and status calculation (0.15 → 0.2, 0.35 → 0.4). The separate official SI interpretation remains <0.10 severe lack, 0.10–<0.20 significant lack, 0.20–<0.40 needs improvement, 0.40–<0.60 good shade, and ≥0.60 excellent shade. Urban95's status rule is project-defined: rounded SI <0.20, 0.20–<0.40, and ≥0.40 map to low, middle, and high attainment respectively.

Urban95 status columns retain _5min/_10min/_15min for compatibility even though its direct rules are generally fixed-distance near-edge rules from building footprints. Amenities Focus walking isochrones still originate from building centroids.

Amenities Focus (internally expanded)

Amenities Focus score is calculated from broader amenity coverage and is the model that most directly responds to amenity-category composition and filtering:

• Amenities Focus score = num_amenities + 0.25*num_trees + 0.25*num_street_lights within the selected walking isochrone.

Amenities Focus map coloring and percentile displays are rank-based relative to all buildings for the selected walking time.

---

Regenerating outputs (Python)

1. Prerequisites

• Raw GIS under data/raw/ (see src/core/paths.py / python -m pipeline check for expected layers — e.g. buildings.geojson.gz, amenities_clean.geojson, trees.geojson, parks.geojson).

• data/raw/arcgis_shade/bsv_street_summer_shade_index.geojson and data/raw/arcgis_shade/bsv_open_spaces_summer_shade_index.geojson — raw ArcGIS summer SI layers for shade scoring and the web map layer.

• data/raw/amenities_clean.geojson (published as docs/data/amenities_new.geojson by export_web) and optionally data/raw/street_lights.geojson for Urban95/clean-manifest inventory views.

• .env in the repo root with a Mapbox token used only for isochrone generation:

  mapbox_access_token=YOUR_TOKEN
  

Seed provisional raw inputs when needed:

$env:PYTHONPATH="src"
python scripts/seed_provisional_raw.py
python -m pipeline check

2. Shade SI layers (before building accessibility)

Prepares metric scoring layers under output/shade_si/ and a simplified web-only display layer at docs/data/shade_si.geojson (+ .gz):

$env:PYTHONPATH="src"
python -m pipeline run shade

Scoring reads the prepared layers in output/shade_si/; the simplified docs/data/shade_si.geojson is for map display only.

3. Building-level metrics and web GeoJSON

Full pipeline (shade → isochrones → amenity metrics → Urban95 score → publish → neighborhoods):

$env:PYTHONPATH="src"
python -m pipeline run all

Or stage by stage:

$env:PYTHONPATH="src"
python -m pipeline run isochrones
python -m pipeline run amenity_metrics
python -m pipeline run score
python -m pipeline run export_web

export_web owns publication of the building, lookup, amenity, tree, park, and isochrone layers under docs/data/ (including their specified .gz companions). The shade stage owns shade_si.geojson (+ .gz); the neighborhoods stage owns its four neighborhood/statistics files. See docs/data/README.md for the complete ownership table.

Layer-oriented execution model

The commands and browser-facing output contracts above are unchanged. In run all, the pipeline now prepares and reuses named source frames, computes amenity and scoring relations as exact layer-by-layer/chunked reductions, uses exact threaded local street-light unions, reuses one neighborhood IDW geometry plan, and publishes each web layer in one serialization pass. Standalone stages retain their disk-backed fallbacks. The neighborhoods stage intentionally rereads the rounded building publication so its aggregates use the same geometry and values served to the map.

For acceptance-only warm-run checks, set PIPELINE_FORBID_MAPBOX=1. A guarded run validates the complete (building_id, minutes) warm aggregate and aborts before token/session/network work when it is incomplete; ordinary unguarded runs retain their existing Mapbox/cache behavior.

To recompute Urban95 status fields (including shade SI) on existing buildings without Mapbox or isochrones:

$env:PYTHONPATH="src"
python -m pipeline run rescore

This refreshes docs/data/buildings_accessibility.geojson (+ .gz), buildings_lookup.json (+ .gz), and companion publish layers via export_web — amenities_new.geojson, street_lights.geojson, amenities_all.geojson, trees.geojson, parks.geojson, isochrones.geojson (when raw/cache inputs exist). Run python -m pipeline run neighborhoods immediately afterward so docs/data/neighborhoods.geojson, neighborhood_charts.json, and citywide_stats.json stay in sync with the rescored building fields.

Partial updates → what the app actually gets

You can refresh one concern without Mapbox or a full rebuild — but only if that command chain writes the published files the map reads (especially buildings_lookup.json, not only a local scored GeoJSON).

You changed…	Run (no Mapbox unless noted)	Must refresh for the live app
Urban95 status rules / shade SI mapping	run shade (if SI inputs changed) → run rescore → run neighborhoods	buildings_lookup (+ gz), buildings .gz, neighborhoods / charts / citywide
Clean amenities / trees / lights inventory (raw)	run amenity_metrics → run score → run export_web → run neighborhoods	Same publish set + amenities_new / trees / lights sync; isochrones reused from cache
Isochrone geometry / walk sheds	run isochrones (Mapbox if cache cold) → run amenity_metrics → run score → run export_web → run neighborhoods	Isochrones + lookup + aggregates
Neighborhood chart logic only	run neighborhoods	neighborhoods / surface / charts / citywide (buildings unchanged)
Everything	run all	Full publish set

Rule: a stage that changes building scores is incomplete until export_web (lookup) and, when aggregates matter, neighborhoods. run rescore already ends in export_web.

4. Neighborhood and citywide aggregates

Run after buildings_accessibility.geojson (and related layers) exist in docs/data/:

$env:PYTHONPATH="src"
python -m pipeline run neighborhoods

This updates docs/data/neighborhoods.geojson, neighborhood_charts.json, and citywide_stats.json.

5. Spatial syntax layer (street network)

Builds segment and zone-level spatial syntax layers from docs/data/roads.geojson:

$env:PYTHONPATH="src"; python -m optional.generate_spatial_syntax

Outputs:

• docs/data/spatial_syntax_segments.geojson
• docs/data/spatial_syntax_zones.geojson

---

Repository layout (short)

urban95/
├── docs/                    # Static site (GitHub Pages root)
│   ├── index.html           # Map UI + in-app help
│   ├── app.js
│   ├── style.css
│   ├── data/                # GeoJSON + JSON consumed by the map
│   └── icons/
├── output/                  # Full preprocessing output (optional archive)
├── data/                    # Source GIS for preprocessing
├── src/
│   ├── pipeline/            # `python -m pipeline check` / `run`
│   ├── core/                # paths, preflight, geo_io, geojson_utils
│   ├── stages/              # shade, isochrones, amenity_metrics, score, export_web, neighborhoods, rescore
│   ├── lib/                 # shade_si, urban95_weights, buildings_*, amenity_layers
│   └── optional/            # filter, download_osm_roads, generate_spatial_syntax, export_urban_nature_areas
├── requirements.txt
├── package.json
└── README.md

---

GitHub Pages

Publish from the docs/ folder. Commit the contents of docs/data/ so the deployed site has the same files as local ./data, or host assets elsewhere and update URLs in docs/app.js.

---

Optional: filter source layers

$env:PYTHONPATH="src"; python -m optional.filter

Writes distance-filtered layers under filtered/ when configured. Note: filtered/ is deprecated as a pipeline input; resolve layers from data/raw/ via core/paths.py.