argo-profile-explorer

name	argo-profile-explorer
description	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.

Argo profile explorer

The Global Argo Data Repository is exposed as the GADR collection. Argo provides ~4,000 active autonomous profiling floats globally, drifting at 1000 m parking depth and surfacing every 10 days.

Argo programs

Program	Depth	Variables	Float density
Core Argo	0–2000 m	T, S, P	~3,800 active floats globally
Deep Argo	0–6000 m	T, S, P	~250 active floats; concentrated in select basins
BGC-Argo	0–2000 m	T, S, P + O₂, chlorophyll-a, NO₃, pH, suspended particles, irradiance	~600 active floats globally

To filter for one of these programs, combine bbox/time with a CQL2 depth or variable filter.

Pattern: profiles near a region within a time window

search_datasets(
  collections=["GADR"],
  bbox=<bbox from `bbox-from-region`>,
  datetime="<start>/<end>",
  limit=25
)

The default returns up to 25 most-recent profiles within the scope.

Pattern: depth-filtered (Deep Argo selection)

search_datasets(
  collections=["GADR"],
  bbox=<bbox>,
  datetime="<start>/<end>",
  filter="properties->>'cube:variables'->>'pres'->>'max' >= 2500",
  filter_lang="cql2-text",
  limit=25
)

pres >= 2500 dbar is a strong proxy for a Deep Argo float (Core Argo tops out near 2000 dbar).

Pattern: shallow profiles only

For boundary-layer / mixed-layer studies:

filter="properties->>'cube:variables'->>'pres'->>'max' < 200"

This catches truncated profiles that didn't dive (often during ice-cover season at high latitudes — useful for sea-ice avoidance algorithms).

Pattern: BGC-Argo only

The BGC subset reports oxygen, chlorophyll, etc. To filter:

filter="properties->>'cube:variables'->>'oxygen'->>'mean' IS NOT NULL"

Or explicitly:

filter="properties->>'cube:variables'->>'chla'->>'mean' IS NOT NULL"

Pattern: minimum-pressure threshold (e.g., "reached at least 1000 m")

filter="properties->>'cube:variables'->>'pres'->>'max' >= 1000"

Looking up a specific float by WMO ID

If the user provides a float WMO (e.g., "float 5905028"), search by it directly:

search_datasets(
  collections=["GADR"],
  q="5905028",
  datetime="<wide range>",
  limit=200
)

The float ID typically appears in the item ID and in the wmo property.

Argo vs WOD vs IOOS gliders

Want	Collection	Skill
Real-time autonomous T/S profiles	GADR	this skill
Historical / reprocessed archive	WOD (cast type PFL)	wod-profile-explorer
Coastal underwater glider deployments	IOOS, MARACOOS, OOI, SPRAY	glider-mission-tracker

Variable cheatsheet for cube:variables filters

Variable	Units	Notes
temp	°C	In-situ temperature
psal or salinity	PSU	Practical salinity
pres	dbar	Pressure (proxy for depth, ~1 dbar per meter)
oxygen or doxy	μmol/kg	BGC-Argo only
chla	mg/m³	BGC-Argo only
nitrate	μmol/kg	BGC-Argo only, very sparse
ph_in_situ_total	unitless	BGC-Argo only

(Verify with one item's properties.cube:variables keys before authoring filters.)

Worked example

"Find Argo float profiles within 200 km of Hawaii in March 2026 with min pressure under 10 dbar."

1. Get Hawaii bbox from `bbox-from-region` skill: [-160.5, 18.5, -154.5, 22.5]
2. search_datasets(
     collections=["GADR"],
     bbox=[-162.5, 16.5, -152.5, 24.5],   # ~200 km buffer
     datetime="2026-03-01/2026-03-31",
     filter="properties->>'cube:variables'->>'pres'->>'min' < 10",
     filter_lang="cql2-text",
     limit=20
   )
3. For each item, get_item to retrieve asset URLs, then download NetCDF