High Frequency Radar Observing Systems in SEACOOS

Download this document as:

Adobe PDF • Open Document • Microsoft Word

Lynn K. Shay¹

Dana Savidge²

Richard Styles³

Harvey Seim⁴

Robert H. Weisberg⁵

Sponsored by:

¹ Rosenstiel School of Marine and Atmospheric Science, University of Miami

²Skidaway Institute of Oceanography. University of Georgia

³Department of Marine Science, University of South Carolina

⁴Department of Marine Sciences, University of North Carolina

⁵Department of Marine Science, University of South Florida

1. Introduction

As part of the Office of Naval Research-sponsored SouthEast Atlantic Coastal Ocean Observing System (SEACOOS), High Frequency (HF) Radars were deployed in four differing venues over the five years of the program (Figure 1). These HF radar systems used either direction-finding (Coastal Ocean Dynamics Application Radar SeaSonde : CODAR) or beam-forming techniques (Wellen Radar : WERA) to acquire radial currents from the Bragg peaks in the Doppler spectra. By mapping the radial current structure from at least two stations, the vector surface velocity fields were mapped in near real-time along the continental shelves of North Carolina, South Carolina and Georgia, Southeast Florida Coast and the West Florida Shelf. One of the programmatic goals focused on testing the latest technologies to acquire data from both long-range (lower-resolution) and medium range (high-resolution) HF radars using both systems. The experimental program sought to exploit other measurement capabilities such as surface waves (both significant wave heights and directional wave spectra) as well as surface wind direction.

One of the concepts introduced in this program was the development of HF radar testbeds where sensors and instruments could be tested. For example, during the summer of 2003, a dual-station WERA system was deployed along the West Florida Shelf overlooking acoustic Doppler current profilers (ADCP) moorings deployed within the University of South Florida Coastal Ocean Modeling and Prediction System (COMPS). These cross-shelf arrays provided an opportunity to assess WERA-derived surface currents over these moorings where the uppermost bin was located at ~4-m depth. In 2005, a "mini-waves" experiment was conducted where tri-axial surface wave instruments (courtesy of National Data Buoy Center and Georgia Institute of Technology) were deployed on two moorings over a two-month period in assessing WERA-derived wave measurements within the Florida Straits. The results indicated fairly good agreement between the buoy and WERA-derived significant wave heights and directional wave spectra using algorithms developed by Wyatt et al. (2003). Another important aspect of the SEACOOS HF radar undertaking was a link to the data management activity. The interaction permitted the near real-time aggregation and visualization of the current observations from the HF radar, in-situ ADCPs and drifters in the SEACOOS footprint and demonstrates the feasibility of sharing the observations with the community of interested users.

The objective of this paper is to provide a perspective on these differing radar systems based on our collective SEACOOS experiences; and, more importantly lessons learned from these deployments in differing venues where the dynamic range of the currents is large as shown in Figure 1. The manuscript describes the basic radar premise and experimental designs in Section 2 including results of the acceptance test along the WFS. In Section 3, time lines and a subset of interesting observations are described. In section 4 an overview of surface wave and wind measurements from HF radar techniques including directional waves is presented. Section 5 is a summary of lessons learned followed by concluding remarks in Section 6. The paper contains an updated bibliography of publications resulting from SEACOOS.

2. Experimental Design

For both DF and BF techniques, the approach utilizes backscatter from surface waves of one-half the radar wavelength (i.e. Bragg wave) to form a Doppler Spectrum (Crombie 1955; Stewart and Joy 1974). First-order returns in this spectrum are associated with the frequency shift off of the Bragg frequency that are proportional to the radial surface current of the dominant peak in the spectrum for either receding or advancing waves toward the radar site (Figure 2). To map the two-dimensional surface current vector, each experimental domain requires at least two radar sites. The ratio of the first-order Bragg peaks of the advancing and receding waves is proportional to the wind direction, and with two or more stations the ambiguity in resolving the wind direction of the wind is removed (Citation) Surface wave signatures, however, are derived from the second-order returns for both the significant wave heights and directional wave spectrum (WERA only) following Wyatt et al. (2003). The domain over which surface waves can be mapped is a function of the noise floor that increases with distance offshore. Waves can be mapped over 55 to 65% of the radar domain from the first-order returns.

2.1 Radar Characteristics

Radar characteristics for both DF and BF techniques are summarized in Tables 1 and 2. The key aspects for SEA-COOS are spatial and temporal resolution, range and data processing requirements. Determining how often the surface currents are needed is a function of the oceanic regime and the HF coastal radar application that differs for each regional association. Horizontal resolution depends on bandwidth and must be part of the licensing request from the Federal Communication Center. A principle data processing issue is enabling a real-time capability in observing the coastal surface currents and waves. Basic characteristics of the CODAR Seasonde systems are listed for the normal-range (medium-resolution) to long-range (low-resolution) systems in Table~1. There is the tradeoff between transmission frequency and bandwidth which determine range and resolution, respectively. The Seasonde system uses a whip transmit (Tx) antennae and one crossed-loop/monopole receiver (Rx) antennae. A least-squares fitting procedure is use to determine the azimuthal directions of the currents based on the MUSIC algorithm (Lipa and Barrick 1983; DePalo and Terrill 2007). An important consideration for CODAR systems is the measurement of the receiver beam pattern. Especially for deployment sites in developed areas the measured beam pattern can depart significantly from the ideal beam pattern and failure to use the correct beam pattern in processing can significantly impact the accuracy of the current estimates. At lower transmitting frequencies in the 5 to 10 MHz range for longer range transmissions (Bragg wavelengths of 30 and 15 m, respectively), available bandwidth tends to be limited by the FCC compared to Very High Frequencies (VHF: >50 MHz) where bandwidth is available to sample high spatial resolution (< kilometer). The region of the shelf where standard processing can produce reliable velocity estimates is limited to water depths where the surface waves producing the Bragg scattering can be considered deep water waves. This constraint limits the inshore coverage region on shallow shelves and is of concern for the longer range systems used. All CODAR systems are omni-directional and use a pulsed swept frequency continuous wave with pulse widths of 100-200 ï­s in normal mode compared to 1000-2000 ï­s in low-resolution mode. For the long-range systems there is typically an hourly output that is an updated, three-hour running average .

The WERA system transmits a frequency modulated continuous wave (FMCW) chirps that avoids the 3 km blind range in front of the radar which is an issue with pulsed systems (Gurgel et al., 1999; Essen et al., 2000). The temporal resolution of WERA is a function of the chirp characteristics (0.26 s) and can be as a few minutes up to hourly intervals (Table 2). For transmission frequencies of 8 and 16 MHz, Bragg wavelengths are 18.7 and 9.35 m, respectively. The four-element transmitter is arranged to encompass about a 120^o swath. WERA has the flexibility to be configured into a DF array (such as CODAR) where 4 antennae may be set up in a square where the distance between element is proportional to the Bragg wavelength. A linear array is set up consisting of 4n elements (n =2,3,4...) using BF techniques. This gives the WERA system considerable flexibility for deployments as noted by Gurgel et al. (1999). Generally, the more antennae, the better the resolution of the current direction (Teague et al., 2001) and for accurate surface wave directional spectra 16-element arrays are needed (Wyatt et al., 2003). While higher spatial resolution requires bandwidth (~200 KHz), temporal sampling can be as low a few minutes as the system is FMCW as compared to a pulsed radar such as the Ocean Surface Current Radar (OSCR) system (Shay et al. 1995). Cable calibrations should be checked periodically to monitor any variations in signal amplitudes and phases. One of the new features that have been recently installed is the BFI approach (K.-W. Gurgel, 2006, personal communication). This approach allows a specific frequency band to be scanned to adjust signal transmission to a frequency where there is minimal noise within the band.

2.2 WFS Acceptance Test

A 16 MHz dual-station WERA system was deployed along the WFS starting 23 Aug and ending 25 Sept 2003 to sample surface circulation over moored ADCPs deployed by USF (Shay et al. 2007, Liu et al. 2007). This short-term deployment was conducted to confirm the functionality of the WERA system which was the first deployed on the US mainland. A 33-d nearly continuous time series of radial and vector surface currents were acquired at 30-min intervals. Coastal surface currents were mapped over an approximate 40 km x 80 km domain (Figure 3) with sites located in Venice Beach, FL and along Coquina Beach, FL, equating to a baseline distance of 45 km (i.e. ~half the radar range). Each site consisted of a four-element Tx and sixteen-element Rx arrays. A total of 1628 snapshots of the vector surface currents were acquired with only 70 samples missing from the vector time series. Comparisons to subsurface measurements from ADCP profiles revealed RMS Differences of 1 to 5 cm s^-1 for both radial and current components. Regression analyses for the vector current components (u-positive east, and v-positive north) indicated slopes close to unity with small biases between surface and subsurface measurements at 4-m depth at EC4 (20-m isobaths) and NA2 (25-m isobaths) moorings from the COMPS array.

Radial current accuracy estimates, based in part on SNR, are an important feature to build into the HF radar network (Fig. 3). An approach is to estimate radial current accuracy proposed by the University of Hamburg, that has been generalized to both DF and BF techniques based on Doppler spectra. The SNR is used as a weighting function and the backscattered power in these estimates. The difference is that the samples of radial currents are taken directly from the first-order peaks in the Doppler Spectrum in BF compared to the boxes from the spectral lines for a given incidence angle sorted into bins in DF mode. Radial current accuracy estimates ranged from 2 to 7 cm s^-1. A similar approach can be implemented for direction finding algorithms where"¦

2.3 Southeast Florida Shelf

After completion of the acceptance test in summer 2003, WERA was deployed at two sites (Key Biscayne, Key Largo) in May 04 and became a real time, web-based product in Sept 04 (Figure 4). Each site consists of a 16-element Rx array and a 4-element Tx array (16 MHz) configured in a square. The HF radar products are surface current speed and direction, and significant wave heights. Sampling intervals have ranged from 10 to 20 minutes where velocities are accurate to less than 6 cm s^-1. Since June 2004, WERA measurements have stretched from the shallow near-shore waters to approximately 100 km offshore (~50%) of the time, and penetrate ~0.7 m. With the exception of a few interruptions (i.e. Nov, Dec 05 after Hurricane Wilma), these radar sites have been working nearly continuously for more than three years. Measurements acquired hurricane Jeanne indicated an energetic coastal ocean response to surface winds of up to 25 m s^-1 (not shown). While Jeanne made landfall in West Palm Beach, an eastward current response emanated from the Biscayne Bay on the south side of Jeanne. As the hurricane made landfall, winds increased to more than 24 m s^-1, forcing surface currents of 1 m s^-1 out of Biscayne Bay. During strong events, the far-field radar response may deteriorate due to the wind seas swamping the backscattered signals (Shay et al. 1998). Given fetch-limited conditions (offshore wind) and that the measurements were acquired along the south side of Jeanne's wind field, the observations suggest WERA can remain operational for moderate winds up to 25 m s^-1 winds if the power grid remains on.

From September 2004 to June 2005, a bottom-mounted, upward-looking ADCP operating at 300 kHz was deployed at 86-m depth. This instrument was deployed at a location inside the WERA array at approximately 0.5 km from the nearest cell. As shown (Table 3), monthly RMS differences were calculated as well as slopes and biases from a regression analysis over the 9-month record. RMS differences ranged from 0.1 to 0.25 m s^-1 between the surface and 14-m depth. First-order statistics reveal a highly variable domain dominated by the Florida Current (FC) where the northward surface velocity in the v-component were more than four times larger than in the u-component of current. These weaker subsurface currents in the east-west direction emphasize dramatic differences throughout the period. However, of interest are periods of flow reversal between surface and subsurface currents that decouple (or decorrelate) surface from subsurface currents as observed in Oct 2004. By contrast, the v-component has a dynamic range of 2.2 to -0.7 m s^-1 and the surface and subsurface currents are highly correlated.

2.4 Georgia-South Carolina Shelf

Two shore-based WERA systems have been operational since April 2006 over the Georgia Carolina Bight with sites on St. Catherine and Prichard Islands in Georgia and South Carolina, respectively. Based on the design of Gurgel et al. (1999), this long-range WERA system operates at a frequency of 8.3 MHz with a daytime range of 220 km, reaching across the broad shelf and over the shoreward flank of the Gulf Stream (Figure 5). At the present time, this WERA system includes 12-element Rx arrays, allowing determination of range and azimuth for current estimation, and permit use of the second-order returns for mapping significant wave heights. Percent data return for vector velocities over the entire record illustrates the areal coverage achieved with these two installations. A strong diurnal cycle is seen with lower areal coverage occurring between the hours of 8 and 11:30PM EST when background noise is at its peak, decreasing the SNR. During peak SNR hours (11AM to 2:30PM EST), the areal coverage is quite good, and at the shelf edge, the 70-100% coverage zones extend along the shelf approximately 140 km. Thus, the engineering design criteria for this radar installation for sensing the Gulf Stream have been achieved in the highly dynamic regime.

2.5 West Florida Shelf

Three long-range CODAR Seasondes were deployed at Redington shores, Venice, and Naples, FL in Fall 2003 to Spring 2004. The Venice station is in collaboration with Mote Marine Laboratory and Rutgers University close to a nearby U.S. Coast Guard Station. As shown in Figure 6, a recent interval when all three radars were working and the winds were sufficient to support a well developed wave field. Observations like this have been the exception rather than the norm over the WFS as results have been mixed over the three-year deployment period. For example, a myriad of problems have ranged from lightening strikes to interference by Homeland Security communications on nearby frequencies in the 5 Mhz range. When the systems have been working well, a measure of their collective performance based on radial coverages from three sites are shown in Figure 7 for Redington Shores, Venice, and Naples. When compared to radials from to in situ observations by ADCPs in the COMPS array at 2-m depth, the RMS differences have been 8 cm s^-1 between surface and near-surface radials. This equates to about 4 cm s^-1 m^-1 compared to those from the phased array deployment in summer 2003 of 1.5 to 2 cm s^-1 m^-1 and more recently on the East Florida Shelf.

2.6 North Carolina Shelf

Two SeaSonde Long Range CODAR systems were deployed at the USACE Field Research Facility (FRF) in Duck, NC in May 2003 and at the USCG station in Buxton, NC, just north of Cape Hatteras in August 2003. . These sites on the Outer Banks were chosen because of readily available power and communications infrastructure. The systems have experienced a variety of problems but have remained in operation since deployment (Figure 8). Coverage has been highly variable but under optimal conditions vector currents are measured more than 175 km from shore. The operating frequency of the system, near 5 MHz, is in a range of active frequencies and suffers from large day/night variations in noise, presumably associated with vertical movement of the ionosphere (Teague et al...), and from intermittent noise sources that are likely of more local origin. More typical daytime and nighttime coverages are shown in Figure 9 for a time period when other noise sources were minimal. The 15 m isobath off the Outer Banks is relatively close to shore and therefore does not too significantly limit the inshore coverage of the radar system .

Over the past three years, limited evaluations have been conducted between surface and subsurface currents in this regime. An ADCP mooring has been maintained in 30 m of water in the footprint of the system since 2004. Comparison of the standard processed CODAR velocity estimates with ADCP currents from 3-m below the surface that have been boxcar-filtered in a manner to mimic the CODAR processing, suggest RMS differences of 15 to 25 cm s^-1 in both components of the flow. Regression of the signals against each other indicates little bias but indicates limited correlation between the two signals (Figure 10). The V-component visually follows a 1:1 relationship for the majority of the dataset but large outliers in the CODAR data skew the regression statistics. The same is not true for the U component and suggests an underestimate of velocities by the CODAR in the cross-shelf direction. As discussed in the next section there are identified problems with the standard processed current estimates; once a revised procedure is finalized a more thorough validation will be undertaken.

From an operational perspective, the main challenges faced in maintaining a near real-time data feed have been physical damage, power and communications issues. The systems suffered several lightning strikes in the first year of operation that highlighted the lack of lightning protection in the CODAR system design. This oversight wasn't too surprising given the lack of systems deployed in the SE and has since been addressed. The other big physical issue has been erosion. At Buxton the installation site was renowned for high erosion rates (the system was deployed where the Hatteras lighthouse used to be - not a wise choice in hindsight) and the antennas have been moved regularly because of the changing shoreline. The moves have required repeated beam pattern mapping efforts that are particularly challenging at this location because it involves running Hatteras Shoals (Graveyard of the Atlantic). Duck has been much more stable in the long term but both systems have been hit by hurricane storm surges that did damage to the systems. We have considered relocating the Buxton system because of all these problems but the National Park Service permitting requirements make moving the system an involved process.

Power and communications have also been challenging to establish in a sustained fashion. Power at Duck is quite secure, but power at Buxton has been shut off several times during mandatory evacuations. A backup generator was installed in 2006 prior to the storm season to address these power outages but it has yet to be pressed into dedicated service. Communications have been troubling. The initial use of phone lines proved quite unreliable. Switching to cable modems has improved communications but leave the systems at the mercy of the cable provider. Service interruptions are not uncommon, as are configuration changes that go unannounced. The other main challenge faced at Duck is that of shared use of the modem with the USACE. Security issues have complicated the configuration and forced abandonment of the FRF internet service, however the FRF accommodated the HF radar system by purchasing a separate cable line to be shared by the CODAR installation and visiting scientists at the FRF. Sharing this service with FRF has led to additional hardware challenges but after a couple of years these are now manageable. Locating the radar at the FRF, which is a test bed for other ocean observing instruments including other radars has also proven to be a challenge due to local interference (specifically from microwave radar energy) arising from the other ongoing experiments.

3. Observed Surface Current Variability:

3.1 Southeast Florida Shelf

As shown in Figure 11, comparison of surface and subsurface velocity time series suggest fairly good correlation between the WERA and ADCP measurements. Surface winds at 10-m from NOAA Coastal Marine Automated Network (CMAN) Fowey Rocks were used to estimate the surface friction velocity (u_*) that ranged between 0 to 0.4 m s^-1 (the higher value observed during a strong atmospheric front). Weaker east-west currents are primarily associated with eddy-like activity along the western flank of the Florida Current. This contrasts with the strong northward current of the Florida Current where the amplitude was a maximum of ~2 m s^-1. The complex correlation coefficients ranged from 0.8 to 0.9 with small phases over the 14-m separation. In this highly energetic regime, driven by variability in the Florida Current, typical RMS differences of 1 to 2 cm s^-1 m^-1 indicate significant near-surface current shears. A potential cause is exemplified by a sub-mesoscale surface current feature that occurred during 17 to 22 January 2005 (Year-Day 17-22 in Fig. 6) within the radar domain shown in Figure 7. During this period an atmospheric front moved over the radar domain and excited a small-scale vortex along the western flank of the Florida Current where the current response indicated currents of about 50 cm s^-1 with an elongated vortex along the 200-m contour and across-shelf scale of about 20 km. Subsequently, the southward flow increased as the vortex center moved north along the 200-m contour as wind decreased to about 8.5 m s^-1. By 2100 GMT 20 Jan, the vortex was reduced in relative size as the wind subsided as the atmospheric front moved further offshore. The vortex moved out of the domain and the Florida Current moved back towards the coast with larger northward surface velocities, suggesting an energetic surface current response (Shay et al. 2007).

3.2 Georgia-South Carolina Shelf

Similar frontal eddies are seen off the GA coast, an example of which is shown in a surface current snapshot and a zoomed in image (Figure 11). The grid spacing is 3 km, which is correct at the mid-shelf, but increases to 6 km at the shelf edge. Along the transect (Fig. 4), the mean speed is calculated from April to August 2006 records to show the offshore distance where total data return falls below 25%. The mean curve is bounded by +/- the average accuracy of the measurement, based on the scatter in the spectra from which the velocity estimates are derived (the Rx arrays each have 12 antennae). GS structure is evident: speed averages about 0.2 m s^-1 over the shelf, rapidly increases across the shoreward, cyclonic flank of the GS beginning at 120 km, reaches a maximum at the GS jet axis at about 155 km, and falls off more slowly across the seaward, anticyclonic flank to the east of the axis. Overall data return is good along this transect - over 40% out to the GS axis.

3.3 North Carolina Shelf

From a data quality perspective it is now understood that the environment off the Outer Banks is particularly challenging for a DF radar. Two main issues are: 1) the large dynamical range of velocities present at a given distance from the antenna; and 2) the relatively high noise levels present at the 5 MHz frequencies at which the system operates. These two factors combine to produce radar cross-spectra that have very broad and structured Bragg scattering regions. The Bragg peak at significant range often barely rises above the noise floor, at least over part of its bandwidth (Figure unc_spectra). Algorithms to identify the Bragg region for subsequent processing to determine current direction and magnitude fail to capture the full Bragg peak and resulting total vector fields can be wildly wrong over some of the coverage area (Figure 12).

Experimentation with the parameters of the algorithm that identify the first-order Bragg peak indicate that improved current fields can be achieved with appropriately modified parameters in the MUSIC algorithm (DePalo and Terrill 2007). The other important change that appears to have a significant impact on data quality is raising the minimum number of valid solutions used in each average. Experimentation with the Bragg peak selection algorithm indicates a single set of revised parameters are not sufficient to reprocess data from different time periods. Evaluation of the best way to reprocess the dataset is ongoing. The vendor has supported the installations on the Outer Banks by allowing more user control over the processing algorithms so that users can experiment with differing settings that may improve performance. Their willingness in this regard has permitted the progress in understanding of the issues faced on the Outer Banks seen to date. As of July 2007, the OBX system has expanded coverage to the north through a partnering with NASA Wallops which has several installations in VA. Early indications are that the additional coverage has a positive influence on the quality of the vector solutions in the area where the problems have been most noticeable and may largely alleviate the data quality issues in the northern half of the domain. If confirmed this finding suggests more complete coverage has the added benefit of improving data quality. (Not sure how bigger coverage adds improved data quality?)

4. Other HF Radar Capabilities

4.1 Significant Wave Height

Estimation of the wave field at discrete gridpoints is an advantage of the BF-approach over the DF approach (Figure 13) in estimating significant wave height and wave direction as part of the standard distribution, using the empirical estimation technique of Essen et al. (1999) and Gurgel et al. (2006). The estimation of directional wave spectra at each point is also possible based on the inversion techniques of Wyatt et al. (2003). Preliminary assessment of the standard distribution wave product has begun suggesting that inclusion of a bathymetric correction (possible in existing software) may be necessary for the lower frequency systems (i.e. < 10 MHz). Note the wedges of nearly uniform wave heights near the shore-based HF-radar installations in the uncorrected wave field example. The significant wave heights can be determined over about 60% of the radar footprint. By contrast, DF radar uses the Doppler spectrum measured at a fixed range close to the radar site to estimate significant wave heights (Wyatt, 2005). This spectrum integrates backscatter over all angles from the radar site and its inversion requires knowledge of the beam pattern. DF approach either doesn't have sufficient SNR to extend the wave measurement beyond a range close to the radar site or current variability makes it much more difficult to separate 1^st and 2^nd order parts of the spectrum at the longer ranges that cover a wider area. Note that this separation will be especially difficult in high seas.

4.2 Directional Wave Spectra

Directional wave spectra have been estimated using the Wyatt et al. (2003) techniques on both the South Atlantic Bight and Southeast Florida Shelf WERA data using a few day samples. Results suggest that the directional wave estimates are sensitive to both the length of the phased array (12 Rx versus 16 Rx) and the length of the sample. Tests with 2048 samples (20-minute samples) revealed less noise and more steady estimates than those with just 1024 samples (10 minutes). Directional wave spectra from the WERA measurements are shown in Figure 14 for the approach and passage of Jeanne over the WERA grid near to the Fowey Rocks measurement site. These wave spectra are possible from the second-order Doppler spectral returns by inverting an integral technique. The wind seas respond to the strong wind stresses containing most of the wind-driven energy. By contrast, there was little indication of a strong low-frequency wave (swell) component moving with the storm since the islands presumably filtered out this faster moving wave component. These strong wind-driven current events are being looked at more closely to assess the performance of WERA surface current mapping and the accurate determination of the forced surface wave directional spectrum from the mini-waves experiment conducted in Spring 05.

4.3 Interoperability

An important issue for the U.S. National Network is combining radial data from BF and DF approaches to form surface current vectors. Since the frequency of the transmitted signal and the corresponding Bragg wavelength set the integrating depth of the remotely sensed measurement (Stewart and Joy 1974), combining radials from two systems require careful development of the technique to map them onto a common grid. For any vector retrievals, the Geometric Distortion of Precision (GDOP), which is a function of the angles of intersection of the radial current measurements, limits the quality of the retrieved radar data (Chapman et al. 1997). This optimal angle lies between 30 to 150^o as shown in many peer-reviewed, published studies. Although this can be relaxed somewhat given the radar configuration, the proof is in how well vectors are resolved in the near and far-fields of these radars. A second aspect is the horizontal resolution of the radars and how a 3 or 6 km resolution maps onto a 1 to 3 km resolution. In this framework, initial comparisons between a 25 MHz CODAR and a 16 MHz WERA data set in the southern part of the domain from an 8-day concurrent time series are shown in (Figure 15). During the period of 16 to 23 April 2005, there was also a moored ADCP as discussed above. Generally, the directions of the 8-day averaged vector currents were in good agreement over both domains with maximum surface currents of about 1.4 m s^-1. However, the variance and standard deviations of the surface velocity signals varied significantly over the domains. The standard deviations for the u-component were a maximum of 28 cm s^-1 along the inshore edge of the Florida Current compared to those from CODAR of about 15 cm s^-1. In the v-component, standard deviations distributions reversed with a standard deviation of almost 80 cm s^-1 for CODAR compared to a maximum of 28 cm s^-1 from WERA.

5. Lessons Learned:

SEA-COOS experiences for HF radars were positive as it allowed us to assess system performances of both systems under differing venues with large differences in the dynamic ranges and horizontal scales. Collectively, a near real-time surface velocity measurement system was developed where data were placed data on the SEA-COOS WEB. For the WERA technology (Gurgel et al. 1999), this was a significant step forward and provided a viable alternative of BF capability to the HF radar community and eventually will be posted on the U.S. National Network. Strong collaborative ties were established within the southeast with USF, UNC, USC, SKIO and GT regarding HF radar technologies and in situ measurements. That said, splitting one pair of installations between two Institutions (i.e., SkIO and USC), so that neither institution has the minimum two stations necessary to assure useful vector data is a strategic error. Command and control, troubleshooting, data-flow, and data analysis are all unduly complicated, and require inordinate levels of coordination between multiple institutions at multiple levels of management in order to make progress. Furthermore, based on our collective experiences, the HF radars used here were not necessarily turn key and all systems need to be maintained with spares and routine maintenance to insure the highest quality of data return possible. In the southeast as well as in the Gulf of Mexico, we must consider hurricane season and potential landfall scenarios (Marks and Shay 1998). Not only will hurricanes disrupt data acquisition procedures, but potential inflict severe damage to the radar sites as in hurricane Wilma (2005).

The two radar groups running using BF techniques (WERA) were in general pleased with the wealth of data provided by this system, including the possibility of near real-time directional wave capabilities. These measurements are not only important to the modeling programs, but are needed to interpret radar-derived surface velocity fields and directional waves in strongly sheared ocean regimes (i.e. Florida Current). In collaboration with our European colleagues, more significant inroads must be made in this area of radar-derived directional waves as it is an exciting area of scientific and research inquiry that has operational potential. This remote sensing capability is a plus in regimes such as the Gulf Stream and Florida Current where surface buoys are difficult to deploy and maintain. Notwithstanding, there were drawbacks with BF: 1) Cabling necessary to support the independent Rx antennae makes the system difficult to relocate quickly. The criticism on the number of Rx antennae along the beach, deemed a drawback by the radar community, has not been an issue for our installations. 2) Processing and post-processing software is in need of improved documentation, but it is open source to the user groups, which we consider a significant advantage. 3) Support of the system is forthcoming from the vendor, but is logistically difficult to acquire, given the time zone offset between the U.S. East Coast and Germany, and some communication difficulties. This issue has been minimized since the vendor now has a North American partner in Canada. And, 4) there is a need to determine the time integration to acquire good directional wave estimates where the installations each have 16-element Rx arrays.

The two radar groups using DF techniques (CODAR) experienced a number of difficulties as well: 1) The 4-5 MHz band is noisy and at times is used in Homeland Security operations (UNC had to stop transmitting at one of its permitted frequencies at the request of the FCC). 2) measurements of surface currents off the Outer Banks can be problematic due to the combined effect of increased noise levels and broad Bragg peaks leads to a decrease in useful data owing to vectors that were not oriented correctly. 3) Significant wave height is valid over the domain and not individual cells or bins as in BF mode. Since only one antennae system is used, the DF algorithms do not provide the directional wave capability. And, 4) There seems to be this relatively large parameter space in the MUSIC algorithm (DePalo and Terrill 2007) that potentially needs more exploratory studies to exploit both the strengths and weaknesses of the system for the U. S. National Network. Error statistics are important for this purpose on not only radial, but vector currents as well.

Acknowledgments: LKS and Team WERA (Thomas Cook, Jorge Martinez-Pedraja, Peter Vertes, Brian Haus and Brad Parks) were supported by ONR through the SEA-COOS program (N00014-02-1-0972) administered by the University of North Carolina Chapel Hill. We would like to thank the following people who have helped us with locations for the radars: Dr. Renate Skinner, Mr. Jim Duquesnel, and Mr. Eric Kiefer from Florida DEP for our site in North Key Largo Botanical Reserve. Mr. Kevin Kirwan and Mr. Ernest Lynk from Miami-Dade County Parks and Recreation for the site in Crandon Park on Key Biscayne. We also thank K.-W. Gurgel and Lucy Wyatt for their valuable assistance in the WERA software and directional wave spectral estimates.

6. References

Crombie, D.D., 1955: Doppler spectrum of sea echo at 13.56 Mc.s^-1. Nature, vol. 175, 681-682.

Chapman, R., L. K. Shay, H. C. Graber, J. B. Edson, A. Karachintsev, C. L. Trump, and D. B. Ross, 1997: On the accuracy of HF radar surface current measurements: Intercomparisons with ship-based sensors." J. Geophys. Res., 102, 18,737- 18,748.

De Paolo, T., and E. Terrill, 2007: Skill assessment of resolving ocean surface current structure using compact anettena style HF radar and the MUSIC direction finding algorithm. J. Atmos. Oceanogr. Tech., 24, 1277-1300.

Emery, B. M., L. Washburn and J. Harlan, 2004: Evaluating radial current measurements from CODAR high-frequency radar with moored current measurements. J. Atmos. Oceanogr. Tech., 21, 1259-1271.

Essen, H.-H., K.-W. Gurgel, and T. Schlick, 1999: Measurement of ocean waveheight and direction by means of HF-radar: an empirical approach. German Hydrograph. J., 51, 369-383.

Essen, H.-H., K.-W. Gurgel, and T. Schlick, 2000: On the accuracy of current measurements by means of HF radar. IEEE J. Oceanic. Engin., vol. 25, no. 4, 2000, pp. 472-480.

Fairall, C., E. F. Bradley, D. P. Rogers, J. B. Edson, and G. S. Young, 1996: Bulk parameterization of air-sea fluxes for the Tropical Ocean Global Atmosphere Coupled Ocean-Atmosphere Response Experiment. J. Geophys. Res., vol. 101, no. C2, 3747-3764.

Gurgel, K.-W., G. Antonischki, H.-H. Essen, and T. Schlick, 1999; Wellen Radar (WERA): a new ground wave HF radar for remote sensing. Coastal Engineering, vol. 37, 219-234.

Gurgel, K.-W., H.-H. Essen and T. Schlick, 2006: An empirical method to derive ocean waves from second-order Bargg scattering: prospects and limitations. IEEE J. of Oceanic. Engin., 31(4), 804-811.

Haus, B. K., 2007: Surface current effects on fetch limited growth of wave energy. J. Geophys. Res., 112, DOI:10.1029/2006JC003924, 1-15pp.

Helzel, T., D. Savidge, R. Styles, T. McKisseck, H. Bull, M. Kniephoff, S. Rehder, First Results from Long Range WERA Sites, Proceedings, Radiowave Oceanography Workshop, Hamburg, Germany, May 15-18, 2006.

Lipa, B. J., and D. E. Barrick, 1983: Least-Squares methods for the extraction of surface currents from CODAR crossed loop data : Application at ARSLOE. IEEE J. Ocean Engineering, 8, 226-253.

Liu, Y., R. H. Weisberg, and L. K. Shay, 2007: Characteristic current patterns on the west Florida shelf from an adjoint, self organizing map analysis of HF radar and ADCP data. J. Atmos. Oceanogr. Tech., 24, 702-712.

Peters, H., L. K. Shay, A. J. Mariano, and T. M. Cook, 2002: Current observations over a narrow shelf with large ambient vorticity. J. Geophys. Res., vol. 107, no. C8, 2-1,15, 2001JC000813.

Shay, L. K., H. C. Graber, D. B. Ross, and R. D. Chapman, 1995: Mesoscale ocean surface current structure detected by HF radar.J. Atmos. and Ocean. Tech., 12, 881-900.

Shay, L. K., T. M. Cook, H. Peters, A. J. Mariano, P. An, A. Soloviev, R. Weisberg, and M. Luther, 2002: Very high frequency radar mapping of surface currents. IEEE J. of Oceanic. Engin., 27, 155-169.

Shay, L. K., J. Martinez-Pedraja, T. M. Cook, B. K. Haus, and R. H. Weisberg, 2007: High frequency radar surface current mapping using WERA. J. Atmos. Oceanogr. Tech. 24, 484-503.

Stewart, R. H. and J. W. Joy, 1974: HF radio measurements of surface currents. Deep-Sea Res., 21, 1039--1049, 1974.

Teague, C. A., J. F. Vesecky, and Z. R. Hallock, 2001: A comparison of multifrequency HF radar and ADCP measurements of near-surface currents during COPE-3. IEEE J. of Oceanic. Engin., 26, 399-405.

Wyatt, L. R., J. J. Green, K.-W. Gurgel, J. C. Nieto Borge, K. Reichert, K. Hessner, H. Gunther, W. Rosenthal, and M. Reistad 2003: Validation and intercomparisons of wave measurements and models during EuroROSE experiments. Coastal Engineering, 48: 1-28.

Wyatt, L., 2005: HF radar for coastal monitoring-a comparison of methods and measurements. Oceans05, Brest, France, 20-23 June 2005, 5 pp.

Table 1: Codar Seasonde specifications and capabilities in long-range (~6Mhz) and medium-range (~25 Mhz) systems.

Table 2: Capabilities of the WERA system in Beam Forming (BF) using a phased array for the 8 and 16 MHz. The system can be configured in Direction Finding (DF) mode where the array is arranged in a square. For waves 16-elements are needed to resolve the directional part of the signals.

Table 3: Summary of comparisons such as RMS differences (cm s^-1) between WERA and a moored ADCP on the shelf from Sept 04 to June 05. Regression statistics are also included such as slope and bias (cm s^-1). The last column is for 9 months of the ADCP deployment.

Table 4: Summary of an eight-day time series comparison between the 14-m bin from an ADCP and the WERA (NKL) and the CODAR (JHPN) that includes the complex correlation coefficient (Υ), phase (φ), and RMS differences between the surface and subsurface (14 m) currents for the east-west current component (u_o-u₁₄) and the north-south (v_o-v₁₄). The measured beam pattern is used for these comparisons for the JHPN data.

Figure 1: HF radar deployments and approximate vector coverage in the SEA-COOS domain relative to bottom terrain.

Figure 2: Doppler power spectrum from a 16 MHz WERA operated in the Florida Straits where υ_b represents the Bragg frequency (~0.408 Hz: red line)) and Δυ is the frequency shift off the Bragg frequency that represents a radial current. The gray area represents the second-order returns where wave data exists.

Figure 3: Percentages of good data (colors : left panel) and accuracy estimates HF-radar derived radial currents (cm s^-1: right panel) from Coquina and Venice Beach based on a weighted signal to noise ratio from the Doppler Spectra averaged over the 33-days of measurements over the WFS. Triangles (black) represent moored COMPS ADCPs and the diamond is a CMAN station.

Figure 4: Time line of operational WERA measurements beginning in May 2004 till August 2007 for Crandon Park (upper), North Key Largo (middle) radial currents and the resultant vector time series (lower) in the Southeast Florida Shelf. Black bars depict the down time for each radar site resulting in no vector.

Figure 5: Percentage of HF-radar data return for day (left) and night (right) between 4/06 installation and 2/07.Coverage area from two sites. Dotted line is the 70% return contour. The offshore transect is shown as the solid white line in the left panel.

Figure 6: An example of a velocity vector field sampled by the CODAR HF-radar array on the WFS with stations at Redington shores, Venice, and Naples, FL.

Figure 7. Radial performance for each of the three sites based on nearly three years of data. The color coding provides the percentage of time for which radials returned data.

Figure 8: Timeline of operational CODAR measurements on the Outer Banks, NC beginning Sept 2003 for two sites at Duck, NC and near Cape Hatteras, NC (Haty). The availability of vector currents is shown as OUBA (Outer Banks).

Figure 9: Scatter plots of CODAR and ADCP currents from 1 April to 25 August 2004 for U and V velocity components and histograms of current differences between surface and 3-m depth. There is essentially no bias but correlations are low. The V component visually follows a 1:1 relationship for the majority of the dataset but large outliers in the CODAR data skew the regression. The same is not true for the U component and suggests an underestimate of velocities by the CODAR in the cross-shelf direction.

Figure 10: Time series of winds (40-h low pass filtered) and surface friction velocity (m s^-1) from Fowey Rocks CMAN station in the upper two panels. The current comparisons of U and V (cm s^-1) with the surface velocity (blue) and the 14-m current (dash-dot:black). The lower panel represents the daily correlation coefficient (Υ) and phase (radians) relative to the surface values. The gray areas represent times of eddy passage in Dec 04 through Feb 05.

Figure 11: Evolution of a sub-mesoscale vortex through the EFS radar domain on 20 and 21 Jan 05 associated with atmospheric frontal passage with predominately southward winds (large yellow arrow) of 12 m s^-1 based on Fowey Rocks CMAN station (star). The magnitude of the current is shown in each panel.

Figure 12: Snapshots of surface current over the entire domain (left panel) and a zoom in at the shelf edge (right panel) from 10:15 EST 7 August 2006 where a frontal eddy is located at ~ 31.5^oN and 79.8^oW. Black dots on shelf are towers with planned or present instrumentation.

Figure 13: Mean current speed along the transect in Figure 1, bounded by accuracy, from spectral scatter in 12 Rx antennae returns. Percent data return along the transect is also shown, for the entire record and for daytime returns between 1:00 and 2:30 PM local time. x-axis is distance from a point onshore midway between the two installations, and not distance from the installations themselves, so does not accurately represent the ranges of the individual installations, which routinely exceed 220 km.

Figure 14: An example of unvalidated wave estimates from two HF-radar installations (black dots on St. Catherine's Island GA and Pritchard's Island SC). The color field represents significant wave height (green is less than 1m, red is more than 2m, yellow is between 1 and 2 m). The blue arrows show wave direction, estimated where coverage from both stations overlaps. Data are collected at half-hourly intervals, at ~3km horizontal resolution. Averaging in space or time may reduce noise and improve estimates. Grey's Reef and tower locations are shown. R8 is near the shelf edge at the 44-m isobath.

Figure 15: Spectra - (placeholder until a decent graphic is generated) - spectra from Duck (top) and Hatteras (bottom) installation as a function of frequency and range for the three components of the receive antenna, the two loops and monopole (from top to bottom on each plot). Note the increase is breadth of the Bragg region at about 100 km range at the Duck site and about 25 km range at the Hatteras site. Light white lines mark the region identified as the Bragg peak for subsequent processing and note that especially for Hatteras the outline fails to capture the full Bragg region, at 100 km range for the positive peak, and for all ranges on the larger negative peak.

Figure 16: Vector current map for the same time period as shown in Figure 15 spectra. Note the anomalous southeastward current in the northern part of the coverage, a fairly common occurrence in the standard processed data from the Outer Banks. The existence of this feature is related to the failure to fully capture the Bragg peak in the processing.

Figure 17: Polar plots of the directional wave spectral energy measurements during Hurricane Jeanne passage on 24 Sept (left) and 25 Sept (right) relative to the direction of the surface winds observed at Fowey Rocks. WERA cell 595 was located close to Fowey Rocks CMAN station. Notice the agreement in the direction of winds between WERA and the Fowey Rocks CMAN station, which is yet another application of phased array radar technology (Processed wave data courtesy of Seaview Remote Sensing LTD).

Figure 18: WERA grid with percentages of good vector data and radial coverage over an 11-month record relative to the vector and radial grids (insets) from the CODAR measurements from 02 UTC 16 April to 2300 UTC 23 April 2005. Radial comparisons have been made between the WERA Key Largo site and the CODAR JHPN site located a few kilometers south. The ADCP position is shown that overlapped this period of the CODAR deployment.

Figure 19: Eight-day averaged comparison between a) BF and b) DF radial current from the NKL and JHPN sites, and c) BF and d) DF (cm s^-1) vectors from concurrent records in April 2005. Black vectors from panel d are superposed on the BF averaged vector map in panel d. The area of interference derived from the DF measurements is shown.

Bibliography (*Peer Reviewed):

Cook, T.M., L.K. Shay, B.K. Haus and J. Martinez, 2005: Backscatter in the Stream: Initial results from East Florida Shelf WERANET. 5^th Radio Oceanography Workshop, Costanoa, CA, 3-6 May 2005.

*Drennan, W. M. and L. K. Shay, 2006: On the variability of the fluxes of momentum and sensibleheat. Bound. Layer Meteor., 119(1), 81-107.

*Hallock, Z.R., P. Pistek, J.L. Miller, L.K. Shay and H.T. Perkins, 2003: A description of tides near the Chesapeake Bay entrance using in situ data with an adjoint model. J. Geophys. Res., 108(C8), 20-1-9 (DOI 10.1029/20015C001128)

*Haus, B. K., 2007: Surface current effects on fetch limited growth of wave energy. J. Geophys. Res., 112, DOI:10.1029/2006JC003924, 1-15pp.

*Haus, B.K., R. Ramos, H. C. Graber, L. K. Shay, and Z. Hallock , 2006: Remote observation of the shoaling of surface waves propagating into an estuarine outflow. IEEE Journal of Ocean. Eng. 31(4), 1-15.

*Haus, B. K., H.C. Graber, L.K. Shay and T.M. Cook, 2003: Alongshelf variability of a coastal buoyancy current during the relaxation of downwelling favorable winds. J. of Coast. Res., 19 (2), 409-420.

Haus, B. K., L. K. Shay, J. Martinez-Pedraja and T. M. Cook, 2006: Surface Currents and Waves generated by Tropical Cyclones as observed by HF Radars. Presented at IEEE, Geosciences and Remote Sensing Symposium, Denver, CO.

Haus, B.K., R. Ramos, T.M. Cook, L.K. Shay, P.A. Work, G. Voulgaris, L. R. Wyatt, and J. Martinez-Pedraja, 2006: HF Radar wave measurement over a high current shear region in the Florida Straits. Ocean Sciences Meeting. Honolulu, Hawaii, February 20-24, 2006.

Haus, B. K., L. K. Shay, R. J. Ramos, J. Martinez-Pedraja, T. M. Cook, P. A. Work, and G. Voulgaris, 2006: Current-shear modification of the fetch limited growth of wave energy. 6^th Radiowave Oceanography Workshop, Hamburg, Germany

*Lekien, F., C. Coulliette, A.J. Mariano, E.H. Ryan, L.K. Shay, G. Haller and J. Madsen, 2005: Pollution release tied to invariant manifolds: A case study along the coast of Florida. Elsevier Science Physica D, 210, 1-20.

*Liu, Y., R. H. Weisberg, and L. K. Shay, 2007: Characteristic current patterns on the west Florida shelf from an adjoint, self organizing map analysis of HF radar and ADCP data. J. Atmos. Oceanogr. Tech., 24, 702-712.

Martinez-Pedraja, J., L.K. Shay, T.M. Cook, B.K. Haus, 2004: Very high frequency radar measurements of surface currents along the inshore boundary of the Florida Current during NRL2001: Rosenstiel School of Marine and Atmospheric Science, University of Miami, RSMAS 2004-03, 32 pp.

*Olascoaga, I. I. Rypina, M. . G. Brown, F. J. Beron-Vera, H. Kocak, L. E. Brand, G. Halliwell, and L.K. Shay, 2006: Persistent transport barrier on the West Florida Shelf, Geophys. Res. Lett., 33, L22603, DOI:10.1029/2006GL027800.

Parks, A.B., L.K. Shay, W.E. Johns, T.M. Cook, B.K. Haus, and J. Martinez-Pedraja, 2006: Observing eddy activity using Wellen Radars and a moored acoustic Doppler current profiler in the Florida Straits. Ocean Sciences Meeting. Honolulu, Hawaii February 20-24, 2006.

*Peters, H., L.K. Shay, A.J. Mariano and T.M. Cook. 2002: Current variability on a narrow shelf with large ambient vorticity. J. Geophys. Res., 107(C8), 2-1-15, 2001JC000813.

*Seim, H., B. Bacon, C. Barans, M. Fletcher, K. Gates, R. Jahnke, E. Kearns, R. Lea, M. Luther, C. Mooers, J. Nelson, D. Porter, L.K. Shay, M. Spranger, J. Thigpen, R. Weisberg, F. Werner. 2003: SEA-COOS - A Model for a Multi-State, Multi-Institutional Regional Observation System, MTS Journal, 37(3), 92-101.

*Shay, L. K., J. Martinez-Pedraja, T. M. Cook, B. K. Haus, and R. H. Weisberg, 2007: High frequency radar surface current mapping using WERA. J. Atmos. Oceanogr. Tech. 24(3), 484-503.

Shay, L. K., A. B. Parks, B. K. Haus, J. Martinez-Pedraja, W. E. Johns, K-W Gurgel, and T. M. Cook, 2007: Resolving coastal ocean eddy activity in surface velocity signatures from Wellen Radars and Acoustic Doppler Current Profiler. Oceans 07, 18-21 June 2007, Aberdeen, Scotland, 6pp.

Shay, L.K., T.M. Cook, B.K. Haus, J. Martinez-Pedraja, H. Peters, W.E. Johns, G. Klaus-Werner, 2006: . Real-time surface velocity mapping using Wellen Radars in the Florida Straits. Ocean Sciences Meeting. Honolulu, Hawaii February 20-24, 2006.

Shay, L.K., T. M. Cook, B.K. Haus, J. Martinez and R. Weisberg, 2005: High frequency radar mapping of surface currents using WERA. 5^th Radio Oceanography Workshop, Costanoa, CA, 3-6 May 2005.

Shay, L.K., T. M. Cook, B.K. Haus, K.-W. Gurgel, H. Peters, J. Martinez, T. Helzel, W.E. Johns, 2005: Real-time surface velocity mapping using Wellen radars in the Florida Straits. The Oceanography Society (TOS)-Int'l Ocean Research Conference, Paris, France, June 3-11, 2005, (abstract/poster)

*Shay, L.K., T.M. Cook, H. Peters, A.J. Mariano, P.E. An, A. Soloviev, R. Weisberg, and M. Luther, 2002:Very high frequency radar mapping of surface currents. IEEE J. of Ocean Engineer., 27, 155-169.

*Shay, L.K., T.M. Cook, and P.E. An. 2003: Submesoscale coastal ocean flows detected by very high frequency radar and autonomous underwater vehicles. J. Atmos. Oceanogr. Tech., 20,1583-1600.

Shay, L.K., T.M. Cook, B.K. Haus, H.C. Graber, J. Martinez, and Z. Hallock, 2003: Observed surface currents during the Chesapeake Bay Outflow Plume Experiments. In 1^st International Radar Oceanography Workshop. Timberline, OR, ONR, 137-145.

Shay, L.K., T.M. Cook, and P.E. An, 2003: Submesoscale flows detected by very high frequency radar and autonomous underwater vehicles. In 1^st International Radar Oceanography Workshop. Timberline, OR, ONR, 129-136.

*Venezia, W. et. al., SFOMC: 2003: A successful Navy and academic partnership providing sustained ocean observation capabilities in Florida Straits. Marine Tech. Soc. Journal, 37(3), 81-91.