This Common Operating Picture would not be possible without AQUAVIEW — a federated ocean data infrastructure that unifies dozens of disparate oceanographic data sources into a single, standards-based discovery layer.
AQUAVIEW is a geospatial data federation built on the SpatioTemporal Asset Catalog (STAC) standard. It provides a single, searchable API spanning 24+ oceanographic collections from NOAA, IOOS, CoastWatch, and other providers — making it possible to discover and access real-time ocean observations that would otherwise require querying dozens of different APIs, formats, and endpoints.
Ocean data is fragmented across agencies, formats, and access protocols. NDBC buoy data lives on one system, CO-OPS tidal stations on another, glider missions on a third, satellite products on yet another. Building a comprehensive operational picture traditionally requires writing and maintaining separate integrations for each. AQUAVIEW collapses this complexity by cataloging all sources into a uniform STAC interface with standardized metadata, bounding-box spatial queries, and direct links to source ERDDAP servers.
Every 2 hours, our automated pipeline queries the AQUAVIEW STAC API with a Gulf of Mexico bounding-box filter. For each matching item, AQUAVIEW provides aquaview:source_url pointing to the live ERDDAP endpoint and aquaview:variables listing available parameters. This lets us automatically discover and fetch the latest data across all 9 collections powering this COP — without hard-coding a single data source URL.
https://aquaview-sfeos-1025757962819.us-east1.run.app/collections
Query any collection, filter by bounding box and time, and follow aquaview:source_url links to get live data from the original providers.
The ultimate Common Operating Picture for the Gulf of Mexico — bringing together 22 real-time layers across 9 AQUAVIEW data collections into a single visualization. Surface currents, satellite imagery, thousands of in-situ sensors, glider missions, oil spill alerts, and weather observations — all discovered through AQUAVIEW's federated ocean data infrastructure.
Animated particles trace ocean surface currents from satellite altimetry. IDW-interpolated heatmaps show continuous temperature and dissolved oxygen fields from hundreds of sensors. Overlaid: ~500 IOOS sensors, 313 NDBC buoys, 140 CO-OPS tidal stations, 5 USF glider missions with real ERDDAP tracks, pollution incidents, and hurricane monitoring probes.
The Gulf of Mexico hosts the Western Hemisphere's largest recurring hypoxic dead zone. By fusing real-time data from every available source through AQUAVIEW, this COP enables early warning of hypoxic events, oil spill tracking, hurricane monitoring, and ecosystem-scale situational awareness.
The COP opens with an operational baseline: sea surface temperature and dissolved oxygen heatmaps showing water conditions, current flow particles showing circulation, and active glider tracks showing real-time sampling missions. Toggle additional layers from the panel on the right.
All data discovered through the AQUAVIEW STAC API, pre-fetched server-side every 2 hours via GitHub Actions.
Every 6 hours, a Python script queries the AQUAVIEW STAC API with a Gulf bounding box filter across 7+ collections. Each STAC item includes aquaview:source_url pointing to the original ERDDAP server, plus aquaview:variables listing available parameters. The script fetches latest readings from each source and writes static JSON files to docs/data/, bypassing browser CORS restrictions.
5,000 particles with 120-frame lifetime, bilinear velocity interpolation, speed-proportional line width and color, and alpha-fade trails (0.975 opacity per frame). Zoom-responsive scaling maintains visual coherence from zoom 4 to 12.
Leaflet 1.9 HTML5 Canvas AQUAVIEW STAC CoastWatch WMS NDBC Realtime CO-OPS API IOOS ERDDAP GitHub Actions
The NDBC hypoxia stations (yellow markers) are paired with ML models forecasting dissolved oxygen dropping below 2.0 mg/L at 1, 3, 5, and 7-day lead times.
200M-parameter zero-shot time series forecaster. Ingests ~512 hours of DO observations and produces probabilistic forecasts blending quantile crossings (60%) and point forecast hours below threshold (40%).
Traditional ML baseline with engineered features: rolling statistics, temporal features, tidal harmonics, and change-point detection. Trained per-station with temporal splits.
Every 6 hours: fetch ~45 days of DO data from NDBC, preprocess, run TimesFM on CPU, and update prediction JSON. Stations without active sensors use a statistical estimator.