Architecture Overview · Park Whisperer Platform

Park Operations Weather Intelligence API

GCP data pipeline that ingests METAR observations, upper-air radiosonde data, and NWS forecasts into BigQuery, scores weather impact per ride, and publishes JSON to a public GCS bucket every 15 minutes for cross-cloud consumption.

● Production · GCP + BigQuery15-Min Data Freshness$35/mo Max Cost

Click any component to explore it in the full case study.

Input A
METAR Surface Observations
aviationweather.gov API — hourly surface weather from Central Florida stations: temperature, humidity, precip, wind, pressure, visibility, wx codes (TS/FG/BR)
aviationweather.gov
Input B
Upper-Air Radiosonde Data
University of Wyoming soundings (Station 72210 Tampa Bay) — CAPE, CIN, Lifted Index, K-Index, Total Totals, Showalter, PWAT. 7 atmospheric indices for ML scoring.
UWY Soundings · Station 72210
Input C
NWS Forecast API
api.weather.gov point forecasts — 24hr hourly forecast periods. Feeds the forecast view and ride impact scheduling.
NWS · api.weather.gov
GCP Data Pipeline — 3 parallel ingestion jobs + daily ML scoring
Job 1
METAR Ingest
weather-updater-v6: parse METAR strings → BigQuery weather_updater_v3 (DAY partition, station_id cluster)
›
Job 2
Atmospheric Ingest
Sounding parser → BigQuery atmospheric table (CAPE/CIN/LI/K/TT/SI/PWAT per day)
›
Job 3
Forecast Ingest
NWS point forecast → BigQuery forecast_24h view
›
Job 4
ML Scoring
Daily insights job: 5 component impact scores per ride from BigQuery views
BigQuery Knowledge Store
5 Tables + 5 Views
5 raw BigQuery tables (DAY-partitioned, clustered). 5 curated pre-joined views: station_latest, region_latest, forecast_24h, atmosphere_latest. Cost-controlled via partition pruning.
5 tables · 5 views · $35/mo max
Impact Scoring Engine
Per-Ride Weather Sensitivity
Ride-specific thresholds map weather component scores (0–1) to operational states: OPERATE / MONITOR / CLOSE. Spatial storm context boosts lightning risk when upwind stations trigger.
Sensitivity profiles · OPERATE/MONITOR/CLOSE
Output A
Per-Station Nowcast JSON
Published to GCS public bucket every 15 minutes. CORS-enabled for cross-cloud consumption. Contains current conditions, 24hr forecast, and atmospheric indices.
GCS public · 15-min cycle
Output B
Ride Impact Recommendations
Per-ride OPERATE/MONITOR/CLOSE state + reason. Consumed by Azure Park Agent dashboard, content pipelines, and operations staff directly.
Azure SWA consumption · JSON API
3
Data sources
5
BigQuery tables + views (each)
15m
Data freshness cycle
$35
Max monthly cost
StackGCP BigQueryCloud Run JobsCloud SchedulerTerraformPythonGCS public bucketNWS APIaviationweather.govUniversity of Wyoming soundings