mcp

Use when the user asks about Argo floats — phrases like "Argo profiles near …", "find floats that reached … dbar", "Core / Deep / BGC Argo", "subsurface temperature near {location}", or specific WMO float IDs. Encodes how to query the GADR collection with depth thresholds via CQL2 and how to distinguish Core / Deep / BGC programs.

Use whenever the user mentions a named ocean, sea, coastal region, or marine area — e.g. "Gulf of Mexico", "Mid-Atlantic Bight", "Sea of Cortez", "Florida Keys", "California Current System", "Arctic", "Hawaii EEZ", "Bering Sea", "Mediterranean". Provides the bbox to pass to AQUAVIEW MCP tools so the agent doesn't have to guess coordinates (which models do poorly for marine regions).

Use when the user wants to filter AQUAVIEW search results on numerical thresholds for ocean/atmospheric variables — phrases like "waves over 6 m", "SST above 28°C", "min pressure under 10 dbar", "salinity between 34 and 35", "wind speed greater than 25 m/s". Encodes CQL2 syntax for STAC item statistics so the agent emits valid filters on the first try.

Use when the user asks about ROV dives, deep-sea video, NOAA Okeanos Explorer, SeaTube, submersibles, hydrothermal vents, abyssal exploration, or specifies very deep targets (>1000 m, "deep sea", "midwater", "benthic"). Encodes how to query HYPERION (Okeanos Explorer) and SEATUBE (Ocean Networks Canada video archive), the dive metadata model, and depth filtering via CQL2.

Use when the user asks about fisheries data — phrases like "bottom trawl survey", "stock assessment", "ichthyoplankton", "groundfish", "longline survey", "{species} catch records", or any NMFS science center mention (AFSC, NEFSC, NWFSC, PIFSC, SEFSC). Encodes which science center covers which region and what survey types live where.

Use when the user asks about underwater gliders — phrases like "active glider missions in …", "Slocum / Spray / Seaglider", "glider profiles near {location}", "glider data for {variable}", or any IOOS regional glider question. Encodes which collection holds which regional glider feed and how to filter live vs archived deployments.

Use when the user asks about surface currents — phrases like "HF radar", "ocean surface currents", "drift forecast", "tidal circulation", "{location} surface flow", or "where would a buoy / debris drift". Encodes the IOOS_HFRADAR query patterns, the regional networks behind it, and resolution / coverage tradeoffs.

Use when the user is exploring what AQUAVIEW data is available for a region, variable, or time period — phrases like "what data does AQUAVIEW have for…", "what's available in…", "is there data for…", "show me sources covering…". Encodes the discovery progression so the agent doesn't burn tokens on overly broad searches.

Use when the user asks about oil spills, vessel groundings, marine pollution, hazardous spill incidents, or trajectory forecasts — phrases like "oil spill in …", "vessel grounding at …", "trajectory of a slick from …", "incident reports near …", "Deepwater Horizon-style analysis". Encodes the cross-domain workflow combining INCIDENT_NEWS / NOAA_ORR (incident reports), Sentinel-1 SAR (slick detection), MARINECADASTRE_AIS (nearby vessels), and IOOS_HFRADAR + HRRR (drift forecast).

Use when the user asks for current / recent / live observations — phrases like "current conditions at …", "right now", "latest reading from …", "this week's tides at …", "live wave heights off …", "real-time wind speed near …". Encodes which collection has which kind of in-situ feed (NDBC for offshore, COOPS for tide gauges, IOOS_SENSORS for regional aggregator, CDIP for waves).

Use when the user wants satellite scenes — phrases like "Sentinel-2 imagery of X", "least cloudy scene over Y", "SAR coverage of an oil spill", "Landsat-Sentinel timeseries", "land cover classification". Encodes which AQUAVIEW collection matches which question (optical / SAR / harmonized / land cover) and the right CQL2 patterns for cloud cover and SAR polarization filtering.

Use when the user asks about sea surface temperature, chlorophyll, ocean color, sea ice extent, or harmful algal blooms — phrases like "SST near …", "is there a HAB in …", "sea ice anomaly in …", "weekly chlorophyll over …", "warmest waters last week". Encodes which AQUAVIEW collection holds which variable, the canonical variable names, and how to assemble a multi-temporal stack.

Use when the user wants to reconstruct a hurricane, tropical storm, atmospheric river, bomb cyclone, or extreme wave/wind event from multiple data sources — phrases like "reconstruct Hurricane {Name}", "buoys during {Storm} landfall", "satellite imagery for {Event}", "storm surge at {Location} during {Storm}". Encodes the multi-source composition pattern (in-situ + satellite + track + surge + atmospheric) and the time/space windowing heuristic.

Use when the user asks about AIS / vessel traffic — phrases like "vessel traffic in …", "shipping density near …", "fishing vessels off …", "port approaches to …", "tracks of MMSI {N}", "container ships in {region}". Encodes the MARINECADASTRE_AIS query patterns, the geohash-aggregate-first rule for density questions, and vessel-type filtering.

Use when the user asks about World Ocean Database, WOD, historical T/S casts, ocean climatology, or specific cast types (CTD, XBT, MBT, PFL, OSD, GLD, DRB, MRB, SUR, UOR, APB). Encodes WOD's cast-type taxonomy, quality-flag handling, and how WOD complements GADR/Argo for the pre-1980s historical record.