hf-radar-currents

name	hf-radar-currents
description	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.

HF radar surface currents

The IOOS HF Radar network produces gridded ocean surface currents derived from coastal HF radar stations. Coverage is patchy globally but dense along most US coasts and select international sites.

Primary collection

IOOS_HFRADAR is the unified national feed. Regional aggregator collections also expose HF radar:

Collection	Region
IOOS_HFRADAR	Unified US national feed
MARACOOS	Mid-Atlantic regional grids
CENCOOS	Central California (San Francisco approach, Monterey)
CALOOS	Southern California Bight
SECOORA	US Southeast
NANOOS	Pacific Northwest
CARICOOS	Caribbean (Puerto Rico, USVI)

Default to IOOS_HFRADAR unless the user specifies a region's RA.

Resolution tradeoffs

The network produces grids at three resolutions; pick by use case:

Grid	Use for
6 km	Regional / synoptic circulation (basin-scale flow patterns)
2 km	Coastal current studies (continental shelf, alongshore flow)
1 km	Bay / estuary / harbor entrances (highest detail, smallest footprint)

Items in the catalog are typically tagged with the resolution. Filter to your scale of interest.

Pattern: "current surface flow off X"

search_datasets(
  collections=["IOOS_HFRADAR"],
  bbox=<bbox>,
  datetime="<24-hours-ago>/<now>",
  q="6km",   # or "2km", "1km"
  limit=10
)

For an animation / timeseries, expand the time window:

datetime="<7-days-ago>/<now>"
limit=50

Pattern: drift trajectory forecast

For "where would something released at point P drift over Δt":

1. Pull a sequence of HF radar grids (6 km is usually enough)
2. Combine with surface wind from HRRR (wind drift component)
3. Use a Lagrangian particle integrator downstream (OceanDrift / OpenDrift)

This is a multi-tool workflow — call out to the user that you'll need both currents and winds:

hf = search_datasets(collections=["IOOS_HFRADAR"], bbox=..., datetime=..., limit=20)
wind = search_datasets(collections=["HRRR"], bbox=..., datetime=..., limit=20)

Variable conventions

Variable	Meaning
u (or u_water)	Eastward velocity component (m/s)
v (or v_water)	Northward velocity component (m/s)
current_speed	Magnitude (m/s) — derived
current_direction	Direction (degrees) — derived
qc_flag	Per-cell quality flag (NaN = bad / no data)

Most HF radar files distribute u/v components, not speed/direction. To compute:

speed = np.sqrt(u**2 + v**2)
direction = (np.degrees(np.arctan2(v, u)) + 360) % 360

Coverage and gap-filling

HF radar coverage gaps are real:

• Outside the radar arc: typically 30-200 km offshore depending on frequency
• Behind headlands: geometric blanking
• During radio interference: time-varying

Many products include a qc_flag raster — drop cells where qc_flag != 0 before any analysis.

For users who need "every cell filled," recommend an OMA (Open-boundary Modal Analysis) or DCT-based gap-fill in post-processing. Don't try to gap-fill at query time.

Common collections it pairs with

• HRRR — atmospheric winds for drift modeling
• Glider data (MARACOOS / CENCOOS / etc.) — Eulerian-Lagrangian comparison
• COOPS — moored ADCP at PORTS stations for ground-truth subsurface currents
• MARINECADASTRE_AIS — vessel drift validation

Worked example

"Map HF radar surface currents off the California coast right now."

1. bbox = [-124.0, 32.0, -117.0, 39.0]   (CA coastal bounds)
2. search_datasets(
     collections=["IOOS_HFRADAR"],
     bbox=bbox,
     datetime="<6-hours-ago>/<now>",
     q="6km",
     limit=10
   )
3. Pull the most recent grid; download NetCDF; plot u/v as quivers

Pitfalls

• Tidal vs sub-tidal currents. HF radar measures total surface velocity — a tidal cycle dominates short-window queries. For "mean flow" questions, average over ≥48h to remove tides.
• Grid edges are noisy. First and last 5–10 km of each grid are often biased.
• Direction conventions vary. Some products use oceanographic ("flowing toward") direction; others meteorological ("coming from"). Always check metadata.