RadarB

RadarB is a focused, scan‑friendly weather operations board for the Cleveland/Lake Erie corridor. It fuses live station data, ODOT sensors, camera feeds, radar tiles, and DWML/NDFD forecast grids into a single, modern dashboard optimized for quick decisions.

Highlights

• Current conditions + wind card (Ambient Weather: temp, feels‑like, wind/gust/max gust)
• Compact 5‑day strip with Today integrated (today high/low + current PWS data, days 1–4 to the right)
• Sun‑track bar (midnight → midnight) with dawn/dusk + sunrise/sunset hatches and cloud‑aware sky gradient
• Sun‑track countdown to the next light phase (dawn/sunrise/sunset/dusk)
• Hourly forecast strip (DWML):
  • Temperature + feels‑like (wind chill / heat index overlays when applicable)
  • Wind + gusts
  • Precip potential + sky cover
  • Weather type bands (legend only shows active precip types)
  • Accumulated snowfall line (NDFD time‑series, with GFS tail extension)
• Accumulated rainfall line
• ODOT sensors and camera tiles (OHGO)
• Alerts + advisories panel (DWML hazards)
• School closings (Lakewood City Schools only; shown only when a closure exists)
• OHGO incidents (nearby traffic incidents; cached)
• Image grids: GOES16 satellite, NWS/WJW/CLE radar, AccuWeather mosaic (proxied), GLERL ice, NWS storm total snow, CPC 8–14 day outlooks, and Pivotal maps

Data Sources

• Ambient Weather (station data)
• NWS DWML (digital forecast grids)
• NDFD XML time‑series (snow accumulation)
• OHGO (ODOT cameras + sensors)
• Sunrise‑Sunset API (civil twilight + sunrise/sunset)
• Spectrum News closings feed (school closings)
• CPC 8–14 day outlooks (temp/precip)
• AccuWeather mosaic (proxied via Cloud Function)
• Pivotal Weather (model imagery)
• Open‑Meteo (GFS point forecast for snow tail extension)

Runtime

• Cloud Functions (2nd gen) target Node.js 24 via functions/package.json.

Project Layout

• public/ — frontend HTML/CSS/JS
• functions/ — Firebase Cloud Functions (CORS proxies, secrets)
• docs/ — implementation notes (see docs/sun-track-line.md)

Tooling

• Lint: npm run lint
• Format: npm run format

Functions & endpoints

Primary endpoints consumed by the dashboard (see public/scripts/config.js):

• getAmbientWeatherDatav2
• getCameraDatav2
• getSensorDatav2
• getOhgoIncidentsv1
• getSchoolClosingsv1
• getFlightDelaysv2
• getDwmlForecastv1
• getNdfdSnowv1
• getTwilightTimesv1
• getRadarProxyv1
• grabPivotalHRRR6hQPFv2

Other functions exist but are currently unused by the UI:

• getWeatherDatav2 (OpenWeather One Call)
• getTomorrowMinutelyPrecipv1 (Tomorrow.io minutely precip cache)
• getGroundStopInfov2, getCityNamev2

Secrets & APIs

Functions read secrets from Google Secret Manager:

• openweathermap (OpenWeather)
• tomorrow-io-api-key (Tomorrow.io)
• ambient-weather-application-key, ambient-weather-api-key
• ohgo-api
• aeroapi
• google-maps-api
• vertex (Gemini fallback for closings parsing)

Local Development

• Serve public/ with Firebase hosting or any static server.
• Use Firebase emulators for Functions if you need local API testing.
  • Ports (from firebase.json): Functions 5005, Hosting 5400, Emulator UI enabled.

Deploy

• Hosting:
  • firebase deploy --only hosting
• Functions:
  • firebase deploy --only functions

Notes

• The NDFD snow feed is proxied via getNdfdSnowv1 to avoid browser CORS.
• DWML hazards are de‑duplicated and displayed as alert cards.
• Closings are hidden unless Lakewood City Schools is explicitly listed as closed/remote/virtual.
• Closings lookups are skipped between April 15 and December 1 (seasonal gate).
• AccuWeather mosaic is served through getRadarProxyv1 to avoid ORB/CORS issues.
• Firestore is not used; rules deny all reads/writes by default.
• Sun‑track sky uses exact civil‑twilight boundaries (dawn/sunrise/sunset/dusk) with cloud‑aware color stops.
• Snow accumulation tail (days 4–7) is extended using a scheduled GFS point forecast job via Open‑Meteo.
• Accumulation charts are shown only when their precip types are present (>0).

Understanding caching

RadarB uses layered caching to keep the dashboard fast without losing freshness. There are three levels:

1. Server‑side (Cloud Functions, in‑memory)

   • Caches reset on cold start or new instances.
   • Current TTLs:
     • getDwmlForecastv1: 30 min + ETag revalidation
     • getNdfdSnowv1: 1 hour
     • getTwilightTimesv1: 6 hours
     • getOhgoIncidentsv1: 5 min (per map bounds)
     • getSchoolClosingsv1: 6 hours
     • getTomorrowMinutelyPrecipv1: 4 min (shared Firestore cache, prewarmed every 3 min)

2. Client‑side (localStorage)

   • forecastData: JSON snapshot for the 5‑day strip (used immediately on load)
   • Legacy cleanup: the old forecastHtml/forecastHtmlTime keys are removed on startup
   • dwmlForecast: raw DWML XML (used immediately on load)
   • twilightTimes: dawn/dusk/sunrise/sunset (6 hours)
   • incidentsCache: OHGO incidents (5 min)
   • closingsCache: school closings (10 min)

Deferred media loading

• Large image tiles (radar/satellite/Pivotal) are loaded after initial paint via data-src + requestIdleCallback.
• Panels use a .panel-loading skeleton and remove it after image load.

Image refresh cadence

• OHGO camera tiles: ~6s
• Radar tiles (CLE/WJW/AccuWeather): 3 minutes
• Slow tiles (GOES16, GLERL, storm total snow, CPC, Pivotal): 1 hour

How to control caching

• Server TTLs live in the handlers under functions/lib/handlers/.
• Client TTLs live alongside each module (see public/scripts/modules/).
• To bypass caches during debugging, clear localStorage keys or add a cache‑busting query param to the function URL.

Snow tail (GFS) pipeline

RadarB extends snow accumulation beyond the NDFD/DWML window using a scheduled GFS point‑forecast job.

• Job code: jobs/gfs_snow_tail/
• Output: gs://radarb-forecast-358874041676/snow_tail.json (public, cached 30m)
• Schedule: every 4 hours 6:30–22:30 ET during Nov–May

The frontend blends NDFD/DWML accumulation first, then appends the GFS tail.

Tomorrow.io minutely precip cache

RadarB can proxy a shared Tomorrow.io minutely precipitation forecast for 41.48,-81.82 using:

• Endpoint: getTomorrowMinutelyPrecipv1
• Upstream fields: precipitationIntensity, precipitationProbability
• Cache store: Firestore document weatherCache/tomorrowMinutelyPrecipV1
• Refresh cadence: every 3 minutes from 6:00 AM ET through 11:57 PM ET (*/3 6-23 * * *)

The endpoint serves the cached payload first and only refetches Tomorrow.io when the cache is missing or older than 4 minutes.

TODO

• Add DWML hazard time‑layout parsing to display effective windows.
• Consider adding observed cloud cover (METAR/GOES) for earlier‑day sun‑track segments.