Observations of the Western Caribbean Sea Circulation: A Multi-Sensor Approach
Jack A. Harlan
Abstract -The surface circulation of the western Caribbean Sea has been examined using the U.S. Navy Relocatable Over-the-Horizon Radar (ROTHR), blended TOPEX/POSEIDON and ERS-2 altimetry and surface drifters. The ROTHR data are surface current velocities from Bragg scattering of typically 10 m wavelength ocean surface waves. The altimetry-derived data are geostrophic velocities. The two remote sensors were well correlated and reveal the same meso- and basin scale features. The drifters, drogued at 15 m, and the HF radar-derived currents illustrate a number of similar features, at both scales. Both the ROTHR and drifter measurements reveal the importance of shallow bathymetric features in directing the flow. Altimetry-derived geostrophic currents and HF radar-derived currents are well correlated when the altimetry data has been symmetrically weighted in time, centered around the comparison date. When an asymmetric weighting method was used, the two datasets were uncorrelated. We believe that this suggests that the asymmetric technique may require some adjustment.
1. Introduction
The Yucatan Channel (YC) provides the sole entry from the Caribbean Sea into the Gulf of Mexico. As such, the circulation in the Gulf of Mexico is determined by this influx and in turn the Florida Current-Gulf Stream (FC-GS) system is also governed by this influx. Presently unknown, however, is the extent to which changes in the flow in the YC affect the FC-GS system. Likewise, the general circulation of the western Caribbean Sea (WCS) has not been well described either quantitatively or qualitatively. Up until recently, the estimates of surface circulation made in the WCS have been by analysis of drifting buoy tracks (Kinder, 1983; Kinder et al., 1985; Molinari et al., 1981) and by analysis of ship hydrographic measurements (Gordon, 1967; Roemmich, 1981). Both these measurement techniques have helped describe the general WCS circulation but have been limited in space and time resolution.
The Gulf Stream is preceded by the Florida Current which flows between Cuba and Florida (Florida Straits) then turns northward at the east coast of Florida. The Florida Current is in turn preceded by the Loop Current system which becomes intensified in the western Caribbean Sea just south of the Yucatan Channel that separates Cuba and Mexico. The Loop Current within the Gulf of Mexico and the Florida Current have been studied extensively. It has already been observed that the Loop Current can undergo large changes over the course of months: extending far into the northern Gulf of Mexico; pinching off large warm eddies; receding to almost no "loop" so that it simply turns sharply eastward around Cuba and into the Florida Straits.
By understanding the dynamics of the Loop Current before it enters the Gulf of Mexico, it may be possible to improve the modelling and forecasting of the Loop Current-Florida Current-Gulf Stream system and consequently assist in improved weather forecasting and fisheries management.
The intent of this paper is to present some observations of the Loop Current as it approaches the Yucatan Channel from the south. The observations are the result of a multi-sensor approach of in situ and remotely sensed data including drifting buoys, satellite altimetry, satellite radiometry and high frequency radar. We will attempt to analyze these data in order to provide a more coherent view of the surface circulation in the western Caribbean Sea. However, the data sets often do not overlap in time or in space so that some of our comparisons must remain qualitative until the various sensors can be deployed simultaneously sometime in the future.
2. Data Description
2.1 High Frequency (HF) Over-the-Horizon Radar
Since 1994, the NOAA Environmental Technology Laboratory has been acquiring data from the U.S. Navy Relocatable Over-the-Horizon Radar (ROTHR) (Georges et al., 1998a; Georges et al., 1998b; Harlan and Georges, 1997; Harlan et al., 1998). These data are in the form of Doppler frequency spectra which are then processed and analyzed to obtain radial velocities of ocean surface currents. Up until 1995, ETL had a program in which data from the U.S. Air Force OTH-B radar system were obtained on a semi-regular basis. The signal processing algorithms for surface current extraction from the spectra were originally developed with OTH-B data (Georges and Harlan, 1995a; Georges and Harlan, 1995b).
Over-the-horizon skywave radar systems use ionospheric reflection in order to extend the range of high frequency (HF) wave propagation. The HF electromagnetic band extends from 3 to 30 MHz i.e., radio wavelengths of 10 to 100 m. The ROTHR can selectably operate at frequencies from 5 to 28 MHz. At these frequencies, ranges of 3500 km are possible with skywave propagation although the systems are limited to an azimuthal window of about 80 degrees (± 40 degrees from boresight of the receiving antenna). This is a result of the antenna type: a linear phased array which will incur large sidelobes in the antenna pattern if it is steered more than about 40 degrees from boresight. Sidelobes add uncertainty to the determination of the direction from which the backscattered radiation arrives. The ROTHR receiving array is extremely long, approximately 2.58 km which enables it to obtain a nominal beamwidth of 0.5 degrees. This narrow beamwidth results in typical azimuthal resolution of 10-15 km, varying with distance from the radar. The range resolution is approximately 6 km, resulting from FM-CW modulation and a 128-point range transform.
The method for extraction of ocean surface current velocity from Doppler spectra of HF radars has been known for decades (Barrick et al, 1977; Barrick, 1972). Bragg scattering from ocean waves results in two spectral peaks, one positive and one negative. In the absence of currents, the two peaks of equal positive and negative Doppler shifts correspond to the phase velocity of ocean waves, whose wavelength is one-half the radar wavelength (or 5 to 50 m for the ROTHR), traveling toward and away from the radar. When a surface current is advecting the waves, the Doppler shift corresponds to the sum of the phase velocity and the surface current's component in the direction of the radar. By measuring the difference of the measured shift with that of the theoretical case of zero current, one can obtain the radial current velocity. This current velocity represents approximately the upper meter of the ocean (Stewart and Joy, 1974).
While the ionospheric propagation used by OTH skywave radars has the advantage of increased range coverage, the ionosphere itself fluctuates, which in turn causes Doppler shifts on the backscattered radiation. These ionospherically-induced Doppler shifts add to the shift caused by the ocean current velocity, thus causing a bias. When available, land echoes within the radar footprint are used to remove the ionospheric-induced shift. A much more difficult problem, however, is ionospheric disturbances that cause the radar radiation to be reflected off multiple ionospheric sublayers before returning to the receiving antenna. This "multi-path" distortion of the Doppler spectra creates more than just the two Bragg-resonance peaks thus effectively precluding accurate velocity measurement. While there is no method for removing this distortion (other than discarding the spectra), the ionospheric conditions that cause the distortion can be monitored in real-time. This ionospheric information can guide the radar operator to select the proper radar frequency so as to minimize multi-path effects. As explained in Georges et al., 1998, an objective spectral quality algorithm has been developed that categorizes data as having been collected during stable or unstable ionospheric conditions.
When ionospheric conditions are stable, it has been shown that the ROTHR-derived radial velocities have an rms difference of about 0.27 m s -1 when compared with shipboard acoustic Doppler current profiler-derived velocities at 10 m depth. This difference is quite large but the two datasets have significant spatial and temporal measuring differences (Harlan et al., 1998). Noted therein is a 0.25 m s-1
ROTHR-ADCP bias during the first half of the ship cruise but absent in the return trip along a similar track. This alone could account for much of the bias. Because of the sporadic nature of ROTHR data collection, no other ship data has been collected simultaneously with ROTHR data to date.
The archived data for this project were collected during 1995, 1996 and 1998 at many different radar frequencies. Generally, 10 to 15 spectra are collected and averaged before the radial velocities are computed. These spectra generally span about 20 minutes to an hour, depending on other activities by the radar system. There are two ROTHR systems (one in Texas and one in Virginia) that can simultaneously produce radial Doppler frequency spectra. When both systems are available and good ionospheric conditions exist, the radial velocities computed from each site can be combined to produce a velocity vector map. Figure 1 shows the nominal coverage area of the ROTHR systems.
For the western Caribbean, only radial velocities from the Texas site were available until 1998. During 1998, vector velocities were occasionally available. Data for Figure 2 were collected during about a 3 hour time span on June 14, 1995. Orange and red colors indicate flow toward the Texas ROTHR (NWward) and greens and blues indicate SEward flow away from the site. This figure reveals a number of surface flow features including: 1) the broad strong northwest-northward flow seen in the western half of the region from 15.5 N latitude to the YC; 2) southward flow at the western tip of Cuba; 3) an apparent counterclockwise eddy centered at 20.5 N, 81.5 W whose SEward flow is indicated by the blue area (20 N, 82 W) and NWward flow by the orange area (21 N, 81.2 W); 4) a small area of near zero velocity embedded in the strong core at 19.2 N, 84.2 W; 5) a bifurcation of the main NWward flow at 16 N, 80 W that continues until about 18 N, 83 W and 6) an area of weak, slightly SEward flow centered near 16N, 82 W just off the Nicaraguan coast.
Examination of the bathymetry of the WCS (Figure 3) indicates that several of these flow features are associated with bottom topographic features. The largest feature is the Nicaraguan Rise (see flow feature 6 above) which is an extremely broad shallow bank with depths generally less than 30 m with numerous above surface outcroppings. Separated from the Nicaraguan Rise by a narrow (20 km) but deep (500 m) channel, is Rosalind Bank. The bifurcation (flow feature 5) appears to coincide with flow on either side of Rosalind Bank. The main NWward flow (feature 1) seems to follow the deepest part of the WCS by steering to the west of the Cayman Ridge which extends east-west across the middle of the region. However, the abrupt feature at 19 N, 84 W causes an area of near zero velocity (flow feature 4) within the main flow. This bathymetric feature actually extends to within 13 m of the surface according to nautical charts (Defense Mapping Agency, 1993).
Figures 4-7 represent the 1998 data used for comparison.
2.2 TOPEX/POSEIDON and ERS-2 Altimetry
The TOPEX/POSEIDON (T/P) satellite was launched in August,1992 and was later put into a 10-day exact repeat orbit. This was the first satellite that was specifically designed for studying global ocean circulation. There are two altimeters on board (one dual-frequency: 5.3 and 13.6 GHz, one single frequency: 13.65 GHz) which share the same antenna. The reason for the dual-frequency altimeter is to minimize the error from ionospheric electron content by combining data from the two frequencies (Fu et al., 1994). ERS-2 was launched in 1995 as a follow-on replacement for ERS-1 which was operational from 1991. ERS-2 carries a 13.8 GHz altimeter with a 35-day exact repeat orbit. These multi-year, ongoing altimetry projects have provided global coverage for sea surface height measurement. The precision of the T/P altimeters is 2 cm while that of the ERS-2 is 7 cm. The accuracy of the T/P altimetry measurements are on the order of 4 to 8 cm as compared with tide gauge measurements (Mitchum, 1994). Figure 8 illustrates the ground tracks for the combined ERS and T/P satellites.
For this project, a blended T/P and ERS-2 altimetry-derived SSH data with an included a model mean height were used to compute geostrophic velocities for the western Caribbean Sea (WCS). These data were provided by Dr. R. Leben. The velocities are gridded to a 0.25 degree latitude-longitude grid and then can be compared with the ROTHR-derived current velocities. Often, the ROTHR data consists only of radial velocities. In these cases, the altimetry-derived current velocities can be projected into the ROTHR radial direction and then compared with the ROTHR radial velocities.
It should be noted that the blended altimetry data has a spatial resolution of about 25 km whereas the ROTHR spatial resolution is 6 km in range by 10-15 km in azimuth. In the time domain, the ROTHR data is collected in approximately one hour whereas the altimetry data is the result of a weighted average of about 35 days of ERS-1 &2 satellite data and 20 days of T/P altimetry centered at the time stamp of the data. For the 1998 blended altimetry, the data is not center-weighted at the time stamp, but is weighted such that only past times are used, each day having less weight than the day of the time stamp.
From these altimetry-derived sea surface heights, one can compute geostrophic velocities. We used a level of no motion of 1000 m depth. The data, prior to velocity computation, were gridded using an objective analysis method (Hendricks et al., 1996).
The altimetry-derived velocities were obtained for those times when ROTHR data was available. Two example dates, 14 Jun 1995 and 18 September 1998, are given in Figure 9 and 10, respectively. The comparison for the combined 31 May 1995 and 14 June 1995 ROTHR radial velocities with the altimetry-derived velocities (Figure 9) indicate a correlation of 0.78 and a mean and variance of the velocity difference of 0.1 m s and 0.04, respectively (Figure 11) for this set of approximately 600 points. These comparisons were made by a nearest-neighbor-in-space method.
For the 1998 comparisons, several days of comparisons of ROTHR-altimeter data have been combined
in Figure 12. The correlation is only 0.08 which is statistically insignificant from zero correlation for this dataset of about 570 points. The comparison method was as above.
The discrepancy between the 1995 and 1998 ROTHR-altimeter comparisons is discussed in section 3.
2.3 Satellite-Tracked Drifting Buoys
For the Year of the Ocean (YOTO) Project, NOAA began deploying drifting buoys in the Caribbean Sea in March, 1998. These buoys are drogued at 15 m depth and use the "holey sock" drogue method. Therefore, they do not represent exactly the same layer of ocean as that measured by the ROTHR which is measuring the current in approximately the upper meter. Generally, this difference in measuring depths is insignificant since the surface circualation usually extends to at least 15 m. The YOTO buoys have been deployed throughout the Caribbean Sea in an effort to effectively sample a number of different surface circulation features (http://drifters.doe.gov). The drifting buoy data is a record of buoy position reported every three days. These times and positions are then used to compute a Lagrangian view of the surface current velocity. For this study, five YOTO drifters were analyzed (Figure 13). Each drifter has its own line color. The scalar speed between reporting points is indicated by a colored box at the reporting point. Intuitively, it is likely that speeds computed in this manner are biased high, since the drifters probably do not flow in a straight line between points. With that caveat in mind, the general features of the drifter tracks reveal: the strong westward flow which turns north through the Yucatan Channel, the Loop Current and the presence of eddies south of Cuba.
2.4 AVHRR Infrared Imagery
The NOAA 12 and 14 satellites carry the Advanced Very High Resolution Radiometer (AVHRR) with two infrared channels (11 and 12 um wavelength) which can be used to measure sea surface temperature (SST). Although this temperature is representative of only the upper 0.1 mm of the ocean surface, it is often used as an indicator of the water temperature throughout the mixed-layer depth. At infrared wavelengths, clouds will obscure the ocean from the satellite because of complete attenuation of the radiation from the sea surface. For the WCS, this cloud obscuration is an important factor in eliminating many time periods of AVHRR data, since convection is very strong in the tropics. Also, the "clearing" atmospheric frontal systems that travel southward off the North American continent rarely reach as far south as the WCS region. Perhaps an even larger problem is the effect of solar radiation during most months of the year. This solar insolation is strong enough to warm the upper skin of the ocean everywhere in the region to temperatures much warmer than the underlying surface layer, thus eliminating the spatial heterogeneity that is necessary for detection of oceanic surface circulation features. From the analysis of approximately five years of NOAA VHRR data from the YC and the Gulf of Mexico, Vukovich et al., 1979 noted that the months of May through October seldom provide the needed thermal contrast.
Given these problems in obtaining useful AVHRR images of the WCS, we have only been able to acquire a few images that offer some qualitative corroboration of the other sensors' measurements, although scores of images have been viewed. These images were obtained from the NOAA Coastwatch Internet site which requires a user account. They use the NOAA NL-SST algorithm for SST computation and are geo-referenced, having a spatial resolution of nominally 1 km at nadir. Figure 14 shows one of the clearest images we were able to obtain: Feb 08 98.
3. Discussion
Much of the knowledge of the WCS circulation has been the result of hydrographic measurements and their analysis (Gordon,1967; Roemmich,1981). Additionally, there have been studies done with satellite-tracked drifting buoys (Molinari, 1981; Kinder,1985). However, neither of these methods gives a synoptic view of the circulation. In order to adequately characterize the mesoscale aspects of surface circulation, it is desirable to have measurements whose spatial resolution is on the order of 50 km or less and whose temporal resolution is on the order of 10 days. From Figure 13, it can be estimated that drifting buoys require approximately 3 to 4 weeks to travel from the southern edge of our study area (14 N latitude) to the Yucatan Channel. Spatially, however, their resolution is only limited by the frequency of their reported position. Ship cruises are usually limited to a few transects through the region. The data for the WCS used by Roemmich, for example, consisted of just two north-south transects (separated by nearly five degrees of longitude) and a single transect across the Yucatan Channel. Obviously, the value of in situ shipboard measurements is that they are not limited to the ocean surface and provide data about the water column that is impossible with present remote sensors.
A. ROTHR and Altimetry
The two remote sensing devices used in this study, HF radar and altimeters, were analyzed with respect to each other by performing a cross-correlation. This was essentially an effort to determine if the two sensors are measuring roughly the same velocities, even though it is clear that the ROTHR measures a surface current which contains tides and wind-drift currents, and the altimeter-derived velocities are computed assuming geostrophy.
As mentioned, very good correlation was found from the 1995 data sets, while no correlation existed for the 1998 comparisons. Given that good correlation occurred for the 1995 data, it is unlikely that the spatial-temporal differences of the two sensors can account for the poor correlation of the 1998 datasets. However, it should be noted from Figures 2, 4 - 7 of ROTHR data and from Figure 9 and 10 of altimeter velocities that the areas of cross-comparison in 1998 had more small-scale features than the large regions of the 1995 comparisons which may account for some of the poor correlation. A more likely candidate for the discrepancy may be the two different methods of time-weighting on the altimetry data that went into the velocity computations for the two data sets. It has been found that altimeter-derived velocities compared better with 1-m drogued drifters when the 1995 time-weighting method was used as compared with the 1998 method (R. Leben, 1998). It is likely that the ROTHR data and the drifters drogued at 1 m are measuring the same layer of the ocean and that these two types of data should compare favorably, although no such comparison has yet been made.
From Figure 12, it is also apparent that the altimeter-derived velocities are confined to a band of ±0.5 m s-1 whereas the ROTHR velocities span -1.0 to +1.5 m s-1. This could be interpreted as due to excessive smoothing of the altimeter data.
B. YOTO Drifting Buoys
The drifters confirm several features of the surface circulation that are indicated by the ROTHR and altimetry: 1) the tendency of most of the flow to travel westward and then turn northward through the Yucatan Channel; 2) the avoidance of the Rosalind Bank (the large black square in Figure 13) leading to a bifurcation as noted by the ROTHR June 1995 data. The altimeter data is generally not used in such shallow areas due to the presence of tides which require a much higher time sampling frequency than exists for the altimeter; 3) the presence of eddies near the middle of the Yucatan Channel; 4) slower, small eddies along the shelf south of Cuba; 5) the southward flow near the western tip of Cuba; 6) the tendency for the bulk of the flow to stay to the south of the Cayman Ridge whose westernmost point is the small black square in Figure 13.
Since the drifters are a Lagrangian measurement and the ROTHR and altimeter measurements are Eulerian, intercomparison of the datasets is problematic. This is exacerbated by the availability at this time of only 3-day reporting frequency.
C. AVHRR
As mentioned in Section 2, no images were found during January-March, 1998 that provided clear views of the WCS. The clearest image acquired does show the Loop Current but little information can be obtained from the area south of the Yucatan Channel because of lack of thermal contrast.
4. Conclusions
The surface circulation of the western Caribbean Sea has been examined using HF skywave radar, TOPEX/POSEIDON and ERS-2 altimetry and surface drifters. The drifters, drogued at 15 m, and the HF radar-derived currents illustrate a number of similar features, at meso- and basin scales. Both types of measurements reveal the importance of shallow bathymetric features in directing the flow. Altimetry-derived geostrophic currents and HF radar-derived currents are well correlated when the altimetry data has been symmetrically weighted in time, centered around the comparison date. When an asymmetric weighting method was used, the two datasets were uncorrelated. We believe that this suggests that the asymmetric technique may require some adjustment since this technique also exhibits poor agreement with 1-m drogued drifters. We will attempt to obtain 1-m drogued drifter data that is coincident with ROTHR data, in an effort to shed more light on the inter-comparison of the three data sets: ROTHR, altimetry and 1-m drogued drifters since ROTHR-derived velocities have never been compared with 1-m drogued surface drifters. Future work also includes quantitative comparison of 15-m drogued drifter velocities with HF radar-derived velocities and with altimeter-derived velocities. Additionally, more HF-derived data is desired in order to obtain further evidence of a bifurcation of the main flow in the western Caribbean Sea near the Nicaragua Rise and to reveal more information about the possible trapped eddy approximately 20 km in diameter in the lee of the Cayman Ridge.
References
Defense Mapping Agency, Map no. WOAGN400.
Fu, L.L., E.J. Christensen, C.A. Yamarone, M. Lefebvre, Y. Menard, M. Dorrer, P. Escudier, TOPEX/POSEIDON mission overview, J. Geophys. Res., 99 (C12), 24369-24381, 1994.
Georges,T. M. and J. A. Harlan, Gulf Stream anomaly seen by defense radar, Eos Trans. AGU, 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 and R. A. Lematta, Large-scale mapping of ocean-surface currents with dual over-the-horizon radars, Nature, 379, 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 Science, 33 (4), 1227-1239, 1998a.
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., 36(1), 101-110, 1998b.
Gordon,A.L., Circulation of the Caribbean Sea, J. Geophys. Res., 72, 6207-6223, 1967.
Harlan, J.A. and T. M. Georges, Observations of Hurricane Hortense with two over-the-horizon radars, Geophys. Res. Lett., 24(24), 3241-3244, 1997.
Harlan, J.A., T. M. Georges and D. C. Biggs, Comparison of over-the-horizon radar surface-current
measurements in the Gulf of Mexico with simultaneous sea truth, Radio Science , 33 (4), 1241-1247, 1998.
Kinder, T. H., Shallow currents in the Caribbean Sea and Gulf of Mexico as observed with satellite-tracked drifters, Bull. Mar. Sci., 33(2), 239-246, 1983.
Kinder, T.H., G.W. Heburn and A.W. Green, Some aspects of the Caribbean circulation, Mar. Geol., 68, 25-52, 1985.
Leben, R.R., unpublished data, 1998.
Mitchum, G.T., Comparison of TOPEX sea surface heights and tide gauge sea levels, J. Geophys. Res., 99 (C12), 24541-24553, 1994.
Molinari, R.L., M. Spillane, I. Brooks, D. Atwood, and C. Duckett, Surface Currents in the Caribbean Sea as deduced from Lagrangian drifters, J. Geophys. Res., 86, 6537-6542, 1981.
Roemmich, D., Circulation of the Caribbean Sea: a well-resolved inverse problem, J. Geophys. Res., 86, 7993-8005, 1981.
Vukovich, F.M., B.W. Crissman, M. Bushnell, and W.J. King, Some aspects of the oceanography of the Gulf of Mexico using satellite and in situ data, J. Geophys. Res., 84, 7749-7768, 1979.
Acknowledgements
The author thanks Dr. R. Leben for supplying the blended altimetry data and for numerous helpful discussions.