HF radar surface currents

A site is a few meters of antenna by the water and a small shelter for the electronics. Most of the network's radars are SeaSondes, made by CODAR Ocean Sensors. A SeaSonde listens on two loop antennas crossed at right angles and a vertical whip, all on one mast. At 12–14 MHz it can transmit from the same mast: a vertical dipole about 7.6 m (25 ft) tall, a metal tube inside a fiberglass mast with the whip above, and the crossed loops sealed in an enclosure partway up. Earlier sites, and some current ones, transmit from a separate antenna. CODAR gives its combined antenna's height as 6.4–7.6 m; high-resolution sites use antennas 5.5–6.5 m tall, and long-range sites a transmit antenna 9.1 m tall with a receive antenna of 6.4 m.

The antenna stands as close to the water as possible, turned so that the loops sit evenly on either side of the line straight out to sea, and clear of metal taller than about a quarter of the radio wavelength (6 m at 13 MHz), which distorts what the loops hear. Radio waves at these frequencies travel along the conducting surface of the sea, beyond the horizon: a site over wet sand reaches farther than one over dry rock, and over fresh water the waves don't travel nearly as far.

The shelter holds a transmitter and a receiver, each a chassis 13 cm high, 49 cm wide and 53 cm deep, and a computer. A site draws about 350 W, and radiates 40 W on average, 80 W at peak. CODAR recommends spacing sites 40–60% of the radar's range apart; the four 13.45 MHz radars on the New Jersey shore are 26–34 km from their neighbors.

A few sites are phased arrays instead, such as the two WERA radars at 13.5 MHz on Florida's Gulf coast: one to four transmit antennas and a line of 12 to 16 receive antennas, which can take up to 100 m of beach and form a narrow beam to determine bearing.

The network's radars operate in five bands. Lower frequencies reach farther but see the sea in coarser cells; higher frequencies see finer cells nearer the shore. Range varies with the sea state, radio interference and the transmit antenna.

Range and range cells are IOOS's figures. The Bragg wave (its length and period) and the depth the current represents (λ/8π) are calculated here at 5, 8.5, 13, 25 and 42 MHz, as the next section explains.

The radar sends its radio waves out along the sea's surface and listens. Nearly all of the echo comes from ocean waves half as long as the radio waves, traveling straight toward or straight away from the radar, whose echoes from crest after crest arrive in step (Bragg scattering; Crombie, 1955). At 13 MHz the radio waves are 23.1 m long and the Bragg waves 11.5 m.

What the radar measures is the Doppler shift of that echo. In deep water, a wave's speed is set by its length: for the Bragg wave it is √(gλ/4π), where λ is the radio wavelength, or 4.24 m/s at 13 MHz, with a period of 2.72 s. In still water, the echo would come back in two peaks, shifted 0.368 Hz up by the waves coming toward the radar and 0.368 Hz down by those going away. A current carries the waves, and its component along the line of sight moves both peaks by a further 2v/λ: 0.087 Hz for each meter per second at 13 MHz. That extra shift is the radial velocity, the current toward or away from the site. The component of the current across the line of sight shifts nothing, so one radar can't see it (Barrick and others, 1977).

Range comes from the time the echo takes to return: a SeaSonde sweeps its frequency, and the wider the sweep, the finer the range cells (3 km for the 50 kHz sweep of the New Jersey radars). A SeaSonde determines an echo's bearing from the ratio of the three signals on its antennas (the two loops and the whip), and places it on a grid of 1° or 5°: a 5° sector is about 0.5 km across near the coast and 10 km across at 100 km. The receiver takes in the three signals; the computer turns them into echo spectra and, each hour, a map of radial velocities on rings of range and spokes of bearing around the site.

The current measured is that of a thin top layer: the Bragg waves feel the water to a fraction of their length, so the radar sees the current of the top layer weighted toward the surface, to an effective depth of about λ/8π (Stewart and Joy, 1974): 0.9 m at 13 MHz, 2.4 m at 5 MHz, 0.5 m at 25 MHz. The network's grids state theirs as the top 0.5 m (1 km grid), 1.0 m (2 km) and 2.4 m (6 km). In the figure's close-up, the cut faces glow according to that weighting, and a dashed line marks the effective depth.

A single radar never measures the current itself, only its component along each line of sight. Where two or more sites see the same water, the network combines their radials into totals every hour, on fixed grids of 6 km and 2 km cells and, in a few bays, 1 km and 500 m. For each cell, it gathers the radials within a search radius (10 km on the 6 km grid, 3 km on the 2 km grid, 1.5 km on the 1 km grid) and finds, by unweighted least squares, the single current whose components along their lines of sight best match them. A total needs radials from at least two sites, and at least three radials in all.

The grids leave out totals with poor geometry (a horizontal dilution of precision above 1.25; see the next section), and those faster than 1 m/s on the West Coast or 3 m/s elsewhere. Each hour's grid fills in over three or four hours as the radars' data arrive. NDBC serves the hourly grids for about 92 days.

Compared with current meters, the radars' currents typically agree to within 5–10 cm/s, across many studies. CODAR gives under 7 cm/s against moored current profilers, and about 2 cm/s for the tidal currents. Part of the difference is not error: the radar averages the top meter or so over a cell kilometers wide for an hour, while a current meter measures at one point, usually meters down. A single radial's precision is estimated at 2–5 cm/s.

Geometry matters just as much. Where two lines of sight cross at a shallow angle, near the line joining two sites (the baseline) or far out to sea, a small error in each radial becomes a large error across them. Totals are trusted where the lines cross at between about 30° and 150°; each total's horizontal dilution of precision (HDOP) measures how much the geometry magnifies the errors.

Coverage changes from hour to hour. High seas and radio interference shorten a radar's range, and a site that stops reporting leaves the cells that only it could see with one radial or none. An hour without a total is a gap, never a zero.

Bluegraph reads the network's 2 km and 6 km grids for the U.S. West, East and Gulf coasts, the Gulf of Alaska, Hawaii, Puerto Rico and the Virgin Islands, and parts of the Great Lakes; the 1 km and 500 m grids aren't used yet. It keeps only totals from two or more sites with an HDOP of 2 or less.

Each station within coverage is assigned one cell: in the finest grid that covers it, the nearest cell with a good total in at least half of the hours in the last three days, within 2.5 nautical miles on the 2 km grid or 4 on the 6 km grid. The daily audit chooses again as radars go down and come back; a station keeps its cell while the cell is good in at least a quarter of those hours. On those terms, 130 stations have a cell. Each hour, Bluegraph reads the cell's last eight hours, since the grids fill in late, so the newest current is a few hours behind the station's other readings.

The readings are the current's speed and its direction, given as the direction the water is going, in degrees true. They appear on the station page (and drift across its 3D sea), in Compare, and in Query's currents table; the Reference says more. They are the current in a cell a mile or two from the station, averaged over the cell and the hour, not at the buoy itself.