J. A. Harlan,1 T. M. Georges,2 and D. C. Biggs3
Abstract. On June 14, 1995, the U. S. Navy's Relocatable Over-the-Horizon Radar (ROTHR) west of Corpus Christi, TX, mapped the radial component of ocean surface current with 15-km resolution over a 230,000-km2 area in the Gulf of Mexico. Concurrently, an oceanographic research vessel measured near-surface currents within part of the area illuminated by the radar, providing an opportunity to compare radar-derived surface currents with in-situ sea truth. The R/V Gyre, operated by Texas A&M University, twice traversed the Gulf of Mexico Loop Current while measuring current vectors with an acoustic Doppler current profiler (ADCP). We compared radar-derived currents with currents measured in the uppermost ADCP bin (centered at 10-m depth). If only radar data exceeding a quality threshold are considered, the rms difference in the radial currents measured by the two techniques is 27 cm s-1. This difference most likely reflects the different sampling employed by these instruments, as well as unremoved ionospheric biases in the radar measurements.
_________________________
1, 2. NOAA Environmental Technology Laboratory, Boulder, CO 80303
3. Department of Oceanography, Texas A&M University, College Station, TX 77843
1. Introduction
Ground-based over-the-horizon (OTH) radars offer a new technology for mapping ocean surface circulation with mesoscale resolution over synoptic-scale areas. Two such radars have been recently deployed by the U.S. Navy in Virginia and Texas for military surveillance. The radars' nominal coverage area includes the entire Caribbean Sea and the southern Gulf of Mexico. The Navy has recently permitted us to conduct tests of these radars' abilities to map ocean surface currents over small portions of their ocean footprint. Current patterns have been observed that qualitatively and quantitatively resemble features observed by other surface- and space-based means [Georges and Harlan, 1995a, 1995b; Georges et al., 1996, 1998a]. However, direct comparisons with established sensors are difficult because there are few current-meter moorings that make routine near-surface current measurements in the open ocean, and none are presently deployed within the radars' field of view. Furthermore, an OTH radar averages its current measurements over approximately 100-km2 of the ocean surface, complicating comparisons with point sensors. Thus, a need exists to reconcile radar-derived surface-current measurements with independent sea truth.
Here, we describe OTH radar measurements made on 14 June 1995, specifically for the purpose of comparison with simultaneous acoustic-Doppler current profiler (ADCP) measurements. The ADCP measurements were made over a three-day period by the R/V Gyre during two transects of the Gulf of Mexico Loop Current. The OTH radar in Texas illuminated an ocean area in the Gulf of Mexico that included the cruise track. We briefly describe how OTH radars measure ocean surface currents, the specific measurements of this test, and sources of error in the measurements. Then we describe the ADCP installation on the R/V Gyre, the details of its measurements during this cruise, and their interpretation. Finally, we describe how the measurements made by these two disparate sensors are reconciled, and we assess the degree of agreement.
2. OTH Radar Measurements
High-frequency (HF or 3-30 MHz) current-mapping radars have been around since the 1970s and are now available commercially. Based on concepts developed by Barrick et al. [1977] at NOAA, shore-based radars illuminate the sea surface using line-of-sight and ground-wave propagation and measure the spectrum of echoes backscattered from Bragg-resonant ocean waves, typically 6 m long. In the absence of currents, the sea-echo spectrum consists, to first order, of two sharp lines with equal positive and negative Doppler shifts corresponding to the phase velocity of radially propagating (incoming and outgoing) ocean waves of one-half the radar wavelength. In the presence of underlying currents, the apparent phase velocity, as measured by a stationary observer, is the sum of the phase velocity in still water and the component of near-surface current in the direction of wave travel. One HF radar can therefore infer radial surface current from the amount of displacement of the "Bragg lines" of the echo spectrum from symmetry about zero Doppler shift [Stewart and Joy, 1974].
To measure surface-current vectors, two radars must interrogate a common ocean area and combine their radial current measurements. Employing the usual radar techniques for resolving echoes in range and azimuth, a pair of such shore-based radars can effectively map surface currents, for example in bays and estuaries, to ranges of about 50 km, with a resolution of a few kilometers. The effective current-averaging depth of HF radar measurements is typically about 1 m. At least two kinds of HF current-mapping radar systems are now commercially available [Shay et al., 1995; Lipa and Barrick, 1995].
An over-the-horizon radar can use exactly the same principle for measuring ocean currents, except it uses reflections from the ionosphere to extend the usable range to more than 2,000 km. The price paid for this greatly extended coverage is in the very large antenna arrays required for useful angular resolution, and in the additional precautions required to avoid and remove ionospheric distortions [Georges et al., 1998b]. Previous attempts to measure ocean surface currents in this way are described by Maresca and Carlson [1980] and Trizna [1982].
2.1 Radar Configuration
The OTH radar measurements were made with the U.S. Navy Relocatable Over-the-Horizon Radar (ROTHR) system west of Corpus Christi, TX. The ROTHR achieves its nominal 0.5 angular resolution with a 2.58-km linear phased receiving array consisting of 372 twin-monopole elements. Range resolution is achieved by performing a DFT on a 25-kHz FM-CW waveform. A radar resolution cell on the ocean surface is therefore about 6 km in range by about 15 km in azimuth, for the frequency and range used.
For this test, a 230,000-km2 area in the Gulf of Mexico was illuminated that imaged that part of the R/V Gyre cruise track that lies east of 89W and south of 27N. (Fig. 1). Its range from the radar extended from 1,034 km to 1,685 km. Data were collected intermittently between 1243Z and 1625Z on 14 June 1995, but only radar data from the first and last hours were used, based on quality considerations. A frequency of about 10.6 MHz was selected using real-time sweep-frequency diagnostic soundings over the oblique radar path. These soundings also permit correction of the radar range to account for the sloping ionospheric path. A 24.6-s Doppler frequency transform was used, giving a nominal radial-current resolution of about 0.42 m s-1. Using parabolic interpolation to locate the spectral peaks more accurately, we can achieve a current resolution exceeding 0.1 m s-1. Greater Doppler resolution is possible using longer coherent integration, but the ultimate limit is imposed by ionospheric shifting and smearing, described next.
2.2 Sources of Error and Data Quality Control
The accuracy of current estimates using ground-wave HF radars depends on the accuracy with which the Doppler shift of the Bragg lines in the sea-echo spectrum can be estimated. This depends on the radar's coherent integration time (Doppler resolution) and on physical sources of spectral broadening, such as nonlinear wave-wave interactions on the sea surface and current variations within a radar cell. The manufacturers of the Ocean Surface Current Radar (OSCR) claim a 4 cm s-1 speed accuracy and a 5 direction accuracy for their instrument.
Additional error in OTH radar current measurements is caused by distortions suffered by the radar wave after two reflections from the ionosphere (one on the outgoing path and the other on the return path). Ionospheric motions and irregularities continuously shift the phase of the received sea echo by changing the length of the two-way radar path through it. These phase fluctuations are normally indistinguishable from, and are often of the same order of magnitude as, the Doppler shifts (rate of change of phase) caused by radial ocean currents, that is, a few tenths of a hertz [Georges and Harlan, 1995b]. If radial currents are derived from ionospherically shifted sea echoes, current measurements become biased. Ionospheric bias is caused both by the diurnal formation and dissipation of the ionospheric layers themselves (whose climatology is more or less known) and by shorter-term dynamical features, such as wavelike traveling ionospheric disturbances. In addition, spectral broadening, caused by ionospheric multipath, prevents accurate measurements of Bragg-line Doppler shifts. Severe broadening can prevent any current measurement at all.
For current measurements, we used only sea echoes that passed an objective quality test based on sharpness of the sea-echo spectrum. We have found that sharpness is an accurate predictor of the usability of OTH radar sea echoes for oceanographic measurements. We monitor spectral sharpness to avoid, rather than correct, these distortions, both in the selection of a radar frequency and in data editing.
2.3 Current Map
Fig. 1 maps in color the radial component of radar-derived surface current in the southeastern Gulf of Mexico. Currents flowing toward the radar (that is, toward the WNW) appear in red and orange, while currents flowing away from the radar (toward the ESE) appear as green and blue. The core of the northwestward-flowing Loop Current is visible in dark red near the southwest corner of the field of view. The maximum radar-derived current speed there is about 1.6 m s-1. The maximum radial speed (eastward) at the apex of the Loop Current (the dark blue region near the top of the field of view) is about 1.5 m s-1. Radar data quality is good over most of the illuminated area, except for its northeastern sector, where the mesoscale patchiness corresponds to less reliable data. Elsewhere, temporal consistency of the map over nearly 4 h [Georges et al., 1998b], along with the generally high data quality, lend confidence to its main features. The location of the Loop Current is consistent with the TOPEX/Poseidon sea-surface topography data for 12-22 June 1995.
Estimating and removing ionospheric biases in the radar-derived currents was a problem for this data set, because there are no land masses near the region of interest to provide zero-Doppler references. Near the end of the 4-h period, land echoes were obtained from Florida and Cuba that indicated an ionospheric bias of about +9 cm s-1. During the first hour, land echoes obtained from Cuba and the Yucatan Peninsula, using an adjacent radar dwell, indicated a bias of about +45 cm s-1. These biases were subtracted from the current measurements in hour 4 and hour 1, respectively. Data from hours 1 and 4 were then combined, averaging the two where both were high quality and selecting the higher quality data elsewhere.
3. ADCP Transect
On 11 June 1995, the R/V Gyre, operated by Texas A&M University, set out on a one-week cruise to measure currents in Loop Current Eddy 'Z' in the western Gulf of Mexico, and the Loop Current proper, guided by the latest TOPEX/Poseidon images of sea-surface topography. The cruise departed Galveston on a course toward Key West, turned around at 1700Z, 15 June, at 24.5N, 84W, and headed back toward Galveston by a more southerly route (Fig. 1). The vessel carried an RD Instruments four-beam 153-kHz ADCP with a 30 concave transducer that is mounted through the hull at a depth of 4 m. The ADCP measures the E-W and N-S components of the current relative to the vessel, while recording the ship's position as determined by differential Global Positioning System (GPS) measurements. Mounting and calibration of the ADCP aboard the R/V Gyre has been described by Murphy et al. [1992]; the principles of ADCP operation are further described by RD Instruments [1989]. The system on the R/V Gyre was configured to collect data in 5-min ensembles, in 4-m depth bins, with the shallowest set at 8-12 m. We used only the data from this shallowest bin (centered at 10-m depth) for comparison with the OTH radar data.
3.1 Sources of Error and Data Quality Control
The ADCP manufacturer claims an rms accuracy of 4 cm s-1 for the E-W and N-S components of currents measured. This is typically degraded, however, by sensitivity and alignment errors. As Joyce [1989] has described, sensitivity errors may arise because the orientation of the acoustic beams is not correct due to factors such as nonzero trim to the transducer and ship, small errors in the beam geometry, or overall system bias. Alignment errors are caused by misalignment between the reference frames of the ADCP and the ship gyrocompass, which is connected to the ADCP to provide heading information. Joyce notes that the two types of errors arise from independent sources and produce errors that are approximately orthogonal. The misalignment introduces an error in the velocity component perpendicular to the ship that is linearly related to ship speed, while the sensitivity error occurs in the ship-parallel component, again in linear proportion to ship speed.
Because the R/V Gyre's ADCP transducer heads are fixed-mounted to the hull, its misalignment angle is typically less than 1.5 degrees. Experience from four cruises of R/V Gyre carried out in support of the Texas-Louisiana Shelf Circulation and Transport Processes Study (LATEX) has shown that the sensitivity error was typically 1 to 4 percent [F.J. Kelly, communication]. The greatest source of remaining error is the ship's gyrocompass because it is an electro-mechanical servo system with inertia and, therefore, lag. ADCP ensembles collected while the ship is turning frequently exhibit vectors in some depth bins that are inconsistent in direction and/or magnitude with vectors from surrounding ensembles at a given depth and those immediately above and below. Therefore, ADCP data collected when the vessel heading changed significantly have been ignored. Moreover, because the ADCP current components are calculated as differences from ship's velocity, the measurements are quite sensitive to small changes in ship speed. Since the R/V Gyre's ADCP has no pitch-and-roll sensors that could be used to provide data to compensate for these modes of motion, data collected when the ship slowed to less than 5 kt or when it stopped (e.g., to make CTD casts) were also ignored.
4. Reconciling ADCP and OTH Radar Measurements
Fig. 1 shows the track of the ADCP transects superimposed on a map of the radial surface current measured by the Texas ROTHR on 14 June. For comparison with the radar data, only the ADCP data collected between 0200Z, 14 June and 2300Z, 16 June were used. Two major differences between these disparate measurements must be dealt with. First, one radar measures only the radial component of the surface current. Therefore, to compare the radar and ADCP current measurements, the ADCP current vectors are first projected onto the radar-radial direction; that is, only the radial current components are compared.
Second, the ADCP currents are essentially a sequence of point measurements made at a mean depth of 10 m over almost three days, whereas the radar extracts current estimates virtually simultaneously from a large number of 100-km2 cells on the ocean surface. To partly deal with the different spatial sampling, we interpolated the ROTHR current measurements to a 1-km grid for comparison with the ADCP measurements. Ignoring the time difference of the two measurements (up to 58 h) assumes that mesoscale currents in the open ocean do not change significantly in that interval. To the extent that this may not be the case, it is useful to keep in mind that the radar measurements were made between 1243Z and 1625Z on 14 June. This interval is shown at the bottom of Fig. 2. Ignoring the depth difference assumes that surface currents do not significantly differ from those at 8-12 m in this part of the ocean.
Figure 2 summarizes the result of the ROTHR-ADCP comparison. The horizontal axis is ADCP measurement time, where data gaps result from editing unreliable samples, using the quality-control procedure described above. A time mark near the middle shows when the ship turned around and headed back through the Loop Current. It is useful to compare Fig. 2 with Fig. 1 to see which parts of the Loop Current are sampled by the sequence of ADCP measurements. The solid line plots the component of the ADCP current velocity in the radar-radial direction. The dots plot the ROTHR-derived radial current component in the nearest 1-km (interpolated) cell. Solid dots and open circles distinguish radar data falling respectively above and below a quality (spectral sharpness) threshold that we normally use to edit the data. If only the radar data falling above the quality threshold are considered, the rms difference between the two measurements is 27 cm s-1, and the difference in their means (bias) is 16 cm s-1. However, the details of Fig. 2 show certain systematic differences that are buried in these overall statistics.
First, the behavior of the ROTHR quality index shows a clear dependence on range from the radar, the higher quality data being obtained from the mid-ranges of the radar dwell region. Data quality was poorer at the farthest extent of the cruise track from the radar (near the turnaround point at the center of the plot) and at the nearest points. This behavior is a result of the particular radar frequency chosen for this test and is not a general indication of where good and bad radar current measurements are typically obtained. More reliable current measurements over larger areas would use longer radar dwell times and radar frequencies optimized for each dwell region.
Second, a consistent 25 cm s-1 positive offset of the radar currents with respect to the ADCP currents is apparent on the outbound transect but not on the return leg. If this offset were due solely to an unremoved ionospheric bias in the radar measurements, it should affect both legs equally, so the difference may indicate a change in the currents during the time between the outward and returning legs of the cruise.
Finally, radar measurements average over a nominal 6 × 15 km cell, whereas the ADCP provides essentially point measurements. This may account for the relatively smoother radar-derived current profile through the narrow portion of the Loop Current at 85W.
5. Conclusion
It is clear that both techniques saw the same Loop Current structures and that the ROTHR maps these features much more efficiently and with much greater two-dimensional coverage. However, the systematic differences between the two measurements are greater than the expected ADCP error and can be only partly attributed to the different space and time sampling of the two instruments. The largest disagreements between the two measurements occur where the radar data quality index is relatively low, suggesting that the index works as intended.
Ionospheric biases and other distortions are and will continue to be the main source of error in OTH radar current measurements. Where nearby land echoes or beacons are available (as is often the case in the Caribbean), biases due to nearly uniform ionospheric motions over large areas can be estimated and removed. Elsewhere, accurate current measurements will require repeated interrogations and low-pass filtering, perhaps over many days [Georges et al., 1998b].
A single radar maps only one component of the current vector, the component along a radar radial, so two radars with overlapping coverage are required to map current vectors. The two ROTHR facilities in Texas and Virginia now provide overlapping coverage of the southern Gulf of Mexico, the Caribbean Sea, and part of the tropical Atlantic Ocean. Dual-radar current-mapping tests using this configuration have already been reported elsewhere [Georges, et al., 1996], including one test of mapping surface currents inside a hurricane [Harlan and Georges, 1997] and another showing the spatio-temporal structure of the Florida Current [Georges et al., 1998a]
The extent to which OTH defense radars can be used routinely for mapping ocean currents will depend to a large degree on the availability of existing radars, the siting chosen for future ones, and forming partnerships with other users of these multi-mission sensors. The brief test described here illuminated only about 1% of the total ocean area covered by the ROTHR system. Current mapping now requires coherent integration for at least 24 s and incoherent (power) averaging over 10 such dwells (compared with about 2 s for tracking aircraft), so demands on radar resources could be substantial, if large areas were to be covered. Routine sharing of operational radars, such as the ROTHR, would therefore require more efficient spectral estimators that could achieve comparable current accuracy with much shorter radar dwells.
Acknowledgments
We thank the Office of the Secretary of Defense and the U.S. Navy for granting access to their OTH radar systems at no cost to NOAA, and R. A. Lematta in particular, for acquiring the Texas ROTHR data for this test. Ship time for the R/V Gyre was funded by Texas A&M University, as part of a two-ship cooperative program with the Direccion General de Oceanografia Naval (DGON) of the Mexican Navy, whose purpose was to survey the geostrophic circulation of the Gulf of Mexico. J. H. Wormuth was chief scientist aboard the Gyre, and C. W. Pollock performed the quality control on the ADCP data.
References
Barrick, D. E., M. W. Evans, and B. Weber, Ocean surface currents mapped by radar, Science, vol. 198, pp. 138-144, 1977.
Georges, T. M. and J. A. Harlan, Gulf Stream anomaly seen by defense radar, Eos, Trans. Amer. Geophys. Union, vol. 76, no. 15, p. 146, 1995a.
Georges, T. M. and J. A. Harlan, Mapping surface currents near the Gulf Stream using the Air Force Over-the-horizon radar, Proc. IEEE Fifth Working Conf. on Current Measurements, St. Petersburg, FL, IEEE Press, Piscataway, NJ, pp. 115-120, 1995b.
Georges, T. M., J. A. Harlan, R. R. Leben, and R. A. Lematta, A Test of Ocean Surface-Current Mapping with Over-the-Horizon Radar, IEEE Trans. Geosci. Remote Sens., vol. 36, no. 1, pp. 101-110, 1998b.
Georges, T. M., J. A. Harlan, and R. A. Lematta, Large-scale mapping of ocean surface currents with dual over-the-horizon radars, Nature, vol. 379, pp. 434-436, 1996.
Georges, T. M., J. A. Harlan, T. N. Lee, and R. R. Leben, Observations of the Florida Current with two over-the-horizon radars, Radio Sci., this issue, 1998a.
Harlan, J. A. and T. M. Georges, Observations of hurricane Hortense with two over-the-horizon radars, Geophys. Res. Lett., vol. 24, no. 24, pp. 3241-3244, 1997.
Joyce T. M., In-situ 'calibration' of shipboard ADCPs, J. Atmos. Oceanic Technol. vol. 6, pp. 169-172, 1989.
Lipa, B. J. and D. E. Barrick, Networking multiple HF radar systems with common coverage overlap, Proc. IEEE Fifth Working Conf. on Current Measurements, St. Petersburg, FL, IEEE Press, Piscataway, NJ, pp.127-130, 1995.
Maresca, J. W., Jr. and C. T. Carlson, Comment on 'Longshore currents on the fringe of hurricane Anita' by Ned P. Smith, J. Geophys. Res., vol. 85, no. C3, pp. 1640-1641, 1980.
Murphy, D. J., D. C. Biggs, and M. L. Cooke, Mounting and calibrating an acoustic Doppler current profiler, Mar. Tech. Soc. Jour. vol. 26(3), pp. 334-38, 1992.
RD Instruments, Acoustic Doppler current profilers, Principles of operation: A practical primer, RD Instruments, San Diego, January 1989.
Shay, L. K., H. C. Graber, D. B Ross, and R. D. Chapman, Mesoscale ocean surface current structure detected by high-frequency radar, J. Atmos. Oceanic Technol. vol. 12, pp. 881-900, 1995.
Stewart, R. H. and J. W. Joy, HF radio measurement of surface currents, Deep Sea Res., vol. 21, pp. 1039-1049, 1974.
Trizna, D. B., Mapping ocean currents using over-the-horizon radar, Int. J. Remote Sens., vol. 3, no. 3, pp. 295-309, 1982.
FIGURE CAPTIONS
Figure 1. Map of the radial component of surface current in the Gulf of Mexico, made with the U.S. Navy Relocatable Over-the-Horizon Radar (ROTHR) near Corpus Christi, TX, on 14 June 1995. Currents flowing toward the radar (that is, toward the WNW) appear in red and orange, while currents flowing away from the radar appear as green and blue. The map shows the structure of the Loop Current and a counter-clockwise eddy on its west edge. Land echoes from Florida, Cuba, and the Yucatan are used to estimate and remove ionospheric biases. Also shown is the track of the R/V Gyre, which performed acoustic-Doppler current profiling on 14-16 June.
Figure 2. Comparison of the radial current measured by the Texas ROTHR (dots) with the same component of the current measured by the ADCP on board the R/V Gyre (dashed line). The left half of the plot shows the outward transect of the Gulf of Mexico Loop Current, while the right half shows the return transect. Solid dots and open circles distinguish radar data falling above and below a quality threshold we use to edit such data. The quality (spectral sharpness) index is plotted along the bottom, with a horizontal line indicating the threshold. If only the radar data falling above the quality threshold are considered, the rms difference between the two measurements is 27 cm s-1, and the difference in their means (bias) is 16 cm s-1.
