wod-profile-explorer

name	wod-profile-explorer
description	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.

WOD profile explorer

The World Ocean Database (WOD collection) is the canonical archive for in-situ oceanographic profile data going back to the late 1700s, with the bulk of data from 1950 onward. It is broader and deeper in time than the Argo network (which starts ~2000) but heterogeneous in instrumentation and quality.

When to prefer WOD

User intent	Best collection
Long-term climatology (pre-1980s)	WOD
Real-time / recent autonomous profiles	GADR (Argo)
Glider deployments	IOOS, MARACOOS, OOI, SPRAY (see glider-mission-tracker)
US coastal stations	NDBC, COOPS, IOOS_SENSORS
Both historical + modern combined	Use both WOD AND GADR and merge in post-processing

WOD cast types

WOD partitions data by instrument (cast type). This matters because quality and depth coverage vary dramatically across types.

See cast-types.md for the full taxonomy. Quick reference:

Code	Instrument	Typical depth	Notes
CTD	Conductivity-Temp-Depth	0–6000 m	Highest accuracy; modern shipboard
OSD	Ocean Station Data (bottle)	0–6000 m	Discrete-depth Niskin/Nansen casts; pre-1990s standard
XBT	Expendable Bathythermograph	0–760 m typ.	T only, no S; common for navy/fisheries surveys
MBT	Mechanical Bathythermograph	0–270 m	Pre-1970 surface T profile
PFL	Profiling Float (Argo-style)	0–2000 m typ.	Subset overlaps with GADR
APB	Autonomous Pinniped Bathy.	0–1000 m	Seal-borne sensors; gappy spatial coverage
GLD	Glider	0–1000 m	Deployment-scale; superseded by IOOS glider feeds
DRB	Drifting Buoy	0–~150 m	Surface drifters with thermistor strings
MRB	Moored Buoy	Fixed depths	NDBC-style; check NDBC first
SUR	Surface only	0 m	TSG underway data; useful for SST-along-track
UOR	Undulating Oceanographic Recorder	0–500 m	Towed-undulating CTD; sparse

Pattern: long-term temperature climatology

search_datasets(
  collections=["WOD"],
  bbox=<bbox>,
  datetime="1950-01-01/1980-12-31",
  filter="properties->>'cube:variables'->>'temp'->>'min' IS NOT NULL",
  filter_lang="cql2-text",
  limit=50
)

For coverage stats first (better than pulling 50 years of items), use aggregate:

aggregate(
  collections=["WOD"],
  bbox=<bbox>,
  datetime="1950-01-01/2024-12-31",
  group_by="datetime",
  interval="1Y"
)

Pattern: cast-type filtered query

If the user wants only "modern, high-quality" casts:

search_datasets(
  collections=["WOD"],
  bbox=<bbox>,
  datetime="2000-01-01/2024-12-31",
  filter="properties.wod:cast_type = 'CTD'",
  filter_lang="cql2-text",
  limit=50
)

(The exact property name wod:cast_type may vary — inspect a sample item with get_item first if the filter returns empty.)

Combining WOD + Argo for full record

For a continuous T/S profile timeseries:

1. Query WOD for 1950-01-01/1999-12-31 (pre-Argo era)
2. Query GADR for 2000-01-01/<today> (Argo era)
3. Merge by depth/time/lat-lon

Don't query both for the overlap period (2000-present) — you'll double-count.

Quality flag awareness

WOD uses WOA-2018 quality-control flags. When reading a downloaded WOD item, filter:

• T/S values where the QC flag is 0 (good) or 1 (acceptable)
• Drop values with QC flag 2 or higher (questionable / bad)

This is a per-asset post-processing step — AQUAVIEW returns the full item, but the user (or the agent's downstream pandas/xarray code) needs to apply the QC filter.

Variables to expect

• temp — temperature (°C)
• salinity — practical salinity (PSU)
• oxygen — dissolved oxygen (μmol/kg)
• nitrate, phosphate, silicate — nutrients (μmol/kg) — sparse outside CTD/OSD
• chlorophyll, pH, alkalinity — even sparser; mostly modern CTD
• pressure — dbar (depth proxy)