Lake County Road Network Editor

Local-only prototype for road-closure scenarios. This is a standalone project: it does not import Python/JavaScript from DaedalMap or County Map. It accepts their prepared/source artifacts as explicit Docker mounts, so it can be moved to its own repository without code surgery.

Scope

• study origins: Lake County, California
• road context: Lake County only, plus a small (~9 km) cross-border buffer so routes do not stop artificially at the county line
• road source: the local California Geofabrik PBF
• shelter source: the local NSS shelter parquet

The regional graph is never overwritten. Each scenario stores only a mapping of segment_id -> state:

open -> forward_only -> reverse_only -> closed -> open

forward and reverse refer to the direction in which the segment geometry is stored and drawn. The map highlights the active state after every click.

First run

Read Download and prepare data before running the commands below. It explains exactly which road and shelter inputs are needed, where to get them, how the static FEMA inventory differs from live open shelters, and how to create the small standalone input folder.

From the workspace root:

Set-Location road-network-editor
$env:ROAD_EDITOR_OSM_DIR = "C:\path\to\geofabrik"
$env:ROAD_EDITOR_SOURCE_DATA_DIR = "C:\path\to\county-map-data"

docker compose build
docker compose run --rm editor python scripts/prepare_lake_network.py
docker compose up

Then open http://localhost:8123.

The prep command writes these local, gitignored artifacts under data/baseline/:

• lake_region.osm.pbf - trimmed regional road extract
• road_tiles/ - viewport-sized clickable physical road chunks
• graph.json - directed graph used by the scenario routing API
• scope.json - selected counties and bounds

This means a clone contains the application and reproducible preparation code, but not the generated road graph, road tiles, or shelter snapshot. See Download and prepare data for the required two-file prepared data bundle and the California road download.

For a collaborator who only needs to run the already-prepared Lake County prototype, send them a ZIP of data/baseline/. After extracting it into their clone, they can start the editor with no source-data mounts:

docker compose -f docker-compose.prepared.yml up --build

The mounted inputs are read-only. For this workspace, use:

$env:ROAD_EDITOR_OSM_DIR = "C:\Users\Bryan\Desktop\global map\county-map-raw\source_data\osm\geofabrik"
$env:ROAD_EDITOR_SOURCE_DATA_DIR = "C:\Users\Bryan\Desktop\global map\county-map-data"

The map cannot zoom or pan outside this small Lake County scope. Its widest view shows only major roads; zoom in for the full local editable network.

Batch scenario analysis

The interactive map edits one scenario at a time. For Lake County batch analysis—multiple homes/origins, scenario comparison, and first-pass shelter placement recommendations—start with Scenario Optimizer.

To let a local LLM run those operations through narrow, structured tools, see Local MCP setup.

California-wide population, shelter, and risk context used for planning is documented in California reference data. It explains the baseline-versus-scenario model, origin/candidate inputs, and how the scripts support future local MCP tools. Command details are in scripts/README.md.

Deliberate build order

1. Now: base map tiles and an editable road-network layer. Clicking a segment proves that the UI and persisted graph overlay agree on its stable ID and its four-state availability cycle.
2. Now: display the scoped FEMA NSS inventory as grey pins and query FEMA Open Shelters live for green pins.
3. Next: add a map click / geocoder origin control and enable travel-time calculations against the active road scenario and selected shelters.

The first graph is intentionally a transparent prototype graph: OSM road ways are split at their existing OSM nodes, travel times use OSM maxspeed where available and conservative highway-class defaults otherwise, and turn restrictions are not yet modeled. It is suitable for editing and scenario UX, not an evacuation-direction authority. The next calibration step is comparing its baseline times against the existing Valhalla results.

API

• GET /api/network?zoom=...&bbox=min_lon,min_lat,max_lon,max_lat - clickable road GeoJSON for the current map viewport, with its scenario state

• POST /api/scenarios/{scenario_id}/segments/{segment_id}/cycle - advances the four-state cycle

• POST /api/scenarios/{scenario_id}/routes - snaps a start point and shelter points to the editable graph, then returns travel minutes and the count reachable within a requested time limit

• GET /api/shelters - scoped NSS inventory snapshot used for gray pins

• GET /api/shelters/live - current FEMA Open Shelters for the editable road scope; cached by the editor for two minutes and refreshed on the map every five minutes

• GET /api/scenarios/{scenario_id}/shelters - added and hidden shelter changes for a scenario

Scenario state is persisted as JSON in data/scenarios/ and remains after a container restart. Road edits and shelter additions/removals are saved in the scenario; the selected start point remains a browser-session control.