1. Introduction
For more than 30 years, high-frequency radars (HFRs) have been used to map ocean surface currents over wide areas for several practical applications including support for commercial navigation and coastal surveillance (Olascoaga et al. 2006), tracking oil spills (Hodgins 1994) and other toxic materials (Lekien et al. 2005; Coulliette et al. 2007), search and rescue operations (Ullman et al. 2006), and biological research through aiding studying larval dispersal and predicting the population dynamics of commercially exploited species (Bjorkstedt and Roughgarden 1997; Graber and Limouzy-Paris 1997).
High-frequency (HF) remote sensing is based on the scattering of electromagnetic (EM) waves from the rough sea surface, which can be described by the theory of wave–wave interaction (Bragg scattering; Crombie 1955). HFRs transmit EM waves in the HF band (3–30 MHz), which propagate along the sea surface, and measure the Doppler shift of the backscattered EM waves caused by ocean surface waves and currents. First-order backscattering resonance occurs when the wavelength of the surface waves is one-half of the transmitted EM wavelength (Fig. 1). Surface gravity waves responsible for the resonant Bragg backscattering of the transmitted EM waves will be called Bragg waves hereinafter. Two or more shore-based HFRs allow mapping ocean surface currents over large coastal areas with a horizontal resolution on the order of 1 km.
A schematic showing how HFRs measure surface currents. The HFR emits EM waves that travel in coupled mode along the sea surface via the induction of electrical currents in the conductive seawater. This current induction penetrates only the first few centimeters below the surface (δ). In the presence of surface gravity waves, the EM waves can be backscattered by Bragg waves, i.e., waves having a wavelength λB equal to half of the transmitted wavelength λEM: Since gravity waves propagate at the phase velocity Cp relative to the surface waters, and the latter flow at a velocity Cw relative to the earth, the backscattered signals received by the antennas have a slightly different frequency, due to the Doppler effect. The known wave phase velocity can be subtracted from the total velocity Cp + Cw estimated from the measured Doppler shift to obtain the surface radial component of the surface current Cw. The effective depth of measurement d depends on how surface waves are advected by the vertically sheared near-surface currents.
Citation: Journal of Atmospheric and Oceanic Technology 33, 3; 10.1175/JTECH-D-15-0143.1
The capability of HFRs to measure currents is affected by the presence of sea ice (Fig. 2), which limits the wind fetch for open adjacent waters, hence limiting the development of Bragg waves for a given wind speed and duration (WMO 1998). Furthermore, sea ice quickly dampens Bragg waves, which have short periods of 2–6 s (Squire 2007; Dumont et al. 2011). Despite this, mapping surface currents in partially sea ice–covered waters using HFRs has already been done in the past. A 25-MHz Coastal Ocean Dynamics Applications Radar (CODAR) was deployed in Prudhoe Bay, Alaska, during ice breakup in July 1984 to investigate the possibility of using HFRs to observe ice and water velocities as well as monitoring the ice cover (Lipa et al. 1986). No independent measurement of ice drift or current was collected to validate the HFRs measurements, but they were highly correlated with nearby measurements of wind, encouraging their potential for measuring surface currents or ice drift in winter. Flocco et al. (2003) demonstrated the feasibility of measuring surface currents in the polynya of Terra Nova Bay in the Ross Sea, with shore-based HF radars at 27 MHz. Potter and Weingartner (2010) investigated the performance of shore-based CODARs at 25 and 13 MHz for partially ice-covered waters in the Beaufort Sea. They found that generally ocean currents cannot be measured in a grid cell containing ice, except for very thin ice or small and isolated ice floes. When a band of ice floes was present between 20 and 30 km offshore with open water near the coast and offshore of the band, radio waves were able to propagate above the band of ice and currents could be measured offshore of the band. However, depending on the speed and direction of the wind, ice-free waters between the coast and the band could limit the fetch for the development of Bragg waves, thus preventing measurement of currents in this area. These interesting results have not yet been published in the refereed literature, and the authors noted that they were unable to investigate the complete parameter space that allows HF radars to measure ocean currents within partially ice-covered waters.
Current maps obtained from a CODAR HFR at Sainte-Flavie (STF) (Lower St-Lawrence Estuary, Quebec, Canada) in the (right) presence and (left) absence of sea ice. Ice concentration data come from the Canadian operational ice–ocean forecasting system.
Citation: Journal of Atmospheric and Oceanic Technology 33, 3; 10.1175/JTECH-D-15-0143.1
With sea ice coverage and duration decreasing in many polar regions, deployments of HFRs to monitor partially ice-covered waters may increase in the future. It is therefore needed to quantify the impact of sea ice on HFR current measurements.
The aim of this work is to assess experimentally the performance of two types of coastal HF radars, CODAR and Wellen Radar (WERA), to map surface currents in partially ice-covered waters using different frequencies as a function of ice concentration and wind conditions. The performance of HFR is defined here as the ratio between daily mean ranges achieved by HFR antennas in the presence of sea ice, which will be called Ri, and the estimated daily mean ranges in the absence of sea ice in the same wind conditions, called Rm. Performance is then defined as Γ = Ri/Rm. Using daily rather than hourly ranges allows for reducing the effect of ambient electromagnetic noise, which has a characteristic daily cycle.
Performance therefore depends on the radar cross section of water surface within radar cell σ, the propagation factor F, and the noise factor N, which depend on the frequency, salinity, sea ice, sea state, and EM noise (Crombie 1955; Gurgel et al. 1999b; Barrick and Long 2006; Potter and Weingartner 2010).
This paper is organized as follows: the study area, radar system details, and oceanographic and meteorological data sources are described in section 2, which provides also data processing steps applied on HFR and ice data. The results are presented in section 3. The final section summarizes the results and provides a discussion and the conclusions of the study.
2. Data and methodology
a. The study area
The study area is located in the lower St. Lawrence estuary (LSLE; Fig. 3), Canada, characterized by a nonuniform and variable thin ice cover (typically 0.1–0.7 m thick; Saucier et al. 2003) from January to March. Data acquisition was made during the 2012/13 winter, despite the fact this winter represented the sixth lowest ice volume since 1969 (Galbraith et al. 2014).
(bottom-right insert) Study area with the location of HFR and IML-4 buoy. I.B. denotes Bic Islands. P.M. is for Manicouagan Peninsula.
Citation: Journal of Atmospheric and Oceanic Technology 33, 3; 10.1175/JTECH-D-15-0143.1
The LSLE has a width ranging from 20 to 50 km from the mouth of the Saguenay Fjord to Pointe-des-Monts, Quebec, Canada. It is characterized by an estuarine circulation with the presence of at least two water layers having different physical characteristics: a deep (>150 m) salty layer coming from the northwestern Atlantic with a salinity of 33–35 psu and a temperature of 3°–5°C, and a fresher surface layer affected by the watershed runoff with a salinity of 20–31 psu and a temperature of −1.9° to 14°C (Ingram and El-Sabh 1990). The wave height and period gradually decrease from downstream to upstream of the LSLE due to the decreasing wind fetch. Waves generally observed in the LSLE are produced locally by winds blowing predominantly from the west, while storm waves are generated by northeasterly winds that blow over much longer fetches. The tide is predominantly semidiurnal and its amplitude increases from downstream, where it is about 4.2 m at Pointe-des-Monts (Drapeau 1992), to upstream.
The LSLE is a well-suited natural laboratory to conduct the study, since it is usually partially ice covered from December to March and is easily accessible for HFR installation and maintenance. Furthermore, most environmental parameters affecting HFR performance are observed or forecasted at high spatial and temporal resolutions. In addition, there is a surface oceanographic buoy moored in the radars’ field of view {IML-4 buoy operated by the L’Institut Maurice-Lamontagne [Maurice Lamontagne Institute (IML)]; Fig. 3} during ice-free conditions (May–October) that measures waves and winds.
Barrick and Long (2006) showed that EM waves are affected by sea surface salinity only over a depth given by their Eq. (3). A numerical calculation for our HFRs gives a depth of 15 cm for a salinity of 25 psu (≈surface salinity of the LSLE). Since we have no salinity measurements as close to the surface and the LSLE is strongly stratified in salinity, especially near the surface, we cannot quantify the effect of salinity on the radars coverage.
b. HFRs
One CODAR was deployed in November 2012 on the south shore of the LSLE with a frequency of 13.5 MHz at Sainte-Flavie, Quebec (STF; 48.61°N, 68.23°W). On the north shore, one WERA with Northern Radar Inc. antennas was deployed at Pointe-aux-Outardes, Quebec (PAO; 49.04°N, 68.46°W), with 16.15-MHz frequency (Fig. 4). Measurements were taken over 10-min periods. The transmitted chirp bandwidth for all sites is 100 kHz, yielding a range resolution of 1.5 km. The characteristics of each HFR are summarized in Table 1.
Map showing the study area and HFR locations with their corresponding polar grids. GENER-predicted waves positions are also indicated.
Citation: Journal of Atmospheric and Oceanic Technology 33, 3; 10.1175/JTECH-D-15-0143.1
HFR characteristics.
c. Oceanographic and meteorological data
Oceanographic and meteorological data used in this study come from numerical models and a surface met–ocean buoy moored near the center of the LSLE. Hourly ice concentrations were obtained from the Canadian operational ice–ocean forecasting system (Smith et al. 2012) through the St. Lawrence Global Observatory (OGSL) web portal, while winds come from the Canadian Global Environmental Multiscale Model (GEM) weather forecasting system (Côté et al. 1998). To check the model skill, we compared its prediction with satellite imagery obtained from NASA Worldview website showing the spatial distribution of sea ice in the study area for a randomly selected image (Fig. 5). Although the image reveals a complex distribution of sea ice at scales not resolved by the model, the general pattern of distribution is relatively well predicted, as the system assimilates RADARSAT ice analyses produced by the Canadian Ice Service.
Comparison between forecasted ice concentration and a MODIS image obtained from NASA Worldview for approximately the same moment of the day.
Citation: Journal of Atmospheric and Oceanic Technology 33, 3; 10.1175/JTECH-D-15-0143.1
A parametric wave model called GENER was forced by GEM winds to estimate significant wave heights and peak periods in the LSLE. GENER was developed to predict wind waves at one deep-water location using 2D wind fields and the computed effective fetches for 16 directional sectors around the point of interest (Desjardins and Ouellet 1984; Ruest et al. 2013). The IML-4 buoy (Fig. 4) measured waves and winds during summer from May to October 2013.
d. Data processing
Since CODARs and phased-array WERAs use different techniques for estimating radial currents (direction finding and beam forming, respectively), HFR ranges were determined using the α of receive antennas. Term α has been calculated for each receive antenna for the WERA and for the three components of the receive antenna for the CODAR HFR. Then, for a critical value of α (αc = 6 dB), the corresponding range achieved by each antenna (rmax) is determined and daily averaged to obtain Ri. Figure 6 shows an example of rmax determination from one raw data file of the PAO WERA for receive antenna 10. Similar results are obtained for the other receive antennas. For simplicity, for the rest, we show results only for the monopole element of the CODAR receive antenna and for antenna 10 of the WERA HFR.
Example showing the critical value αc of the signal-to-noise ratio and the corresponding rmax for the receive antenna 10 of the PAO WERA.
Citation: Journal of Atmospheric and Oceanic Technology 33, 3; 10.1175/JTECH-D-15-0143.1
3. Results
a. HFR range versus ice concentration
Hourly ice concentration forecasts were daily and spatially averaged over the domain shown in Fig. 4 (between 48.3°–49.15°N and 67.7°–69.25°W). Figure 7 shows Ri versus daily sea ice concentration.
HFR daily mean ranges vs ice concentration for CODAR and WERA.
Citation: Journal of Atmospheric and Oceanic Technology 33, 3; 10.1175/JTECH-D-15-0143.1
The daily mean range for CODAR and WERA decreases with increasing sea ice concentration. However, since the HFRs operate at different frequencies, other environmental parameters such as wind must be taken into account before their sensitivities to sea ice concentration can be compared.
b. Effect of wind and waves on HFR performance
1) Bragg energy density from measured wave spectra
The Bragg energy density is the gravity wave energy density at the Bragg frequency of the HFR. It has, in principle, a direct effect on HFR measurements by affecting the power of the backscatter radar signal [affecting the radar cross section σ in Eq. (1)].
Raw data of vertical accelerations recorded at the IML-4 buoy at 4 Hz every 10 min have been used to calculate wave spectra (it is assumed that the accelerations of the buoy are only associated with the movement of the waves). The wave spectra were daily averaged over the period from May to October 2013, and then linearly interpolated at the HFR Bragg frequencies (Table 1) to obtain Bragg energy densities. Figure 8 shows HFR daily mean ranges versus daily Bragg energy densities. For both radars, the range decreases when the Bragg energy density approaches 0 m2 Hz−1. For the CODAR the range saturates at 50 km for large Bragg energy densities, whereas it reaches 70–80 km at intermediate Bragg energy values and decreases slightly for larger values for the WERA. Therefore, it is necessary to take these relationships into account before investigating the effect of sea ice. First, HFR daily mean ranges in ice-free conditions can be estimated from Bragg energy densities. To do so, HFR daily mean ranges in ice-free conditions have been empirically fitted to Bragg energy densities by averaging ranges within Bragg energy density bins (the red curve in Fig. 8). Then these averaged daily mean ranges have been interpolated to the Bragg energy densities in winter in order to obtain the estimated daily mean ranges Rm.
Daily mean ranges vs Bragg energy density for CODAR and WERA antennas.
Citation: Journal of Atmospheric and Oceanic Technology 33, 3; 10.1175/JTECH-D-15-0143.1
2) Two-parameter Bretschneider’s wave spectrum
Since wave observations are not available in winter, we investigated whether a theoretical wave spectrum built with significant wave height and peak periods estimates obtained from a parametric fetch model could be used to predict Bragg energy density in winter.
Daily-averaged measured spectra were normalized by the maximum energy density for periods shorter than 6 s (to avoid swells), and frequencies were normalized by the corresponding peak frequency. These normalized spectra were averaged during the whole observational period (May–October 2013). The resulting average measured spectrum is very similar to the normalized Bretschneider’s spectrum (Fig. 9), except at low frequencies.
Comparison between the averaged measured spectrum (blue) and the theoretical spectrum of Bretschneider (red). The standard deviation of the measured spectrum is shown in light gray.
Citation: Journal of Atmospheric and Oceanic Technology 33, 3; 10.1175/JTECH-D-15-0143.1
To check whether the Bretschneider’s spectrum could be used in winter to estimate Bragg energy densities, we computed Hs and ωm from daily-averaged measured spectra in summer and used Eq. (3) to estimate Bragg energy densities. Figure 10 shows the Bragg energy densities from IML-4 spectra versus the Bragg energy densities inferred from the Bretschneider’s spectrum built with Hs and ωm measured by the same buoy. The high correlation coefficients (R2 ≥ 84%) obtained confirm that the Bretschneider spectrum can be used to estimate Bragg energy densities.
Observed Bragg energy densities from IML-4 buoy vs estimated Bragg energy densities from the theoretical two-parameter wave Bretschneider’s spectrum, with R2, for Bragg frequencies corresponding to each HFR.
Citation: Journal of Atmospheric and Oceanic Technology 33, 3; 10.1175/JTECH-D-15-0143.1
c. GENER predictions
To separate the effects of wind and sea ice on HFR coverage, we need predictions of waves in the absence of sea ice during winter 2013. Since operational wave forecasting systems parameterize the effect of sea ice on waves, we used a simple fetch model called GENER (Desjardins and Ouellet 1984; Ruest et al. 2013, 2016).
Wind speed predicted by GEM of Environment and Climate Change Canada has been used to force GENER during winter 2013 at two positions, each in the corresponding polar grid of the HFR coverage: (48.75°N, 68.31°W) for STF and (48.97°N, 68.42°W) for PAO (stars in Fig. 4). Waves predicted by GENER are very similar at the two positions (not shown), confirming that using a single position for each HFR is sufficient to represent wave conditions over the entire observational area.
To validate GENER predictions, we compare daily-averaged predicted (from the GENER model) and measured (from the IML-4 buoy) significant wave heights (<Hs>_model vs <Hs>_obs) and peak periods (<Tp>_model vs <Tp>_obs) (Fig. 11). Wave spectra measured every 10 min at the buoy were daily averaged and <Hs>_obs and <Tp>_obs were estimated using the first moment and the peak frequency of the daily-averaged spectra, respectively. Since GENER predicts Hs and Tp every hour, for consistency with the averaging procedure of the buoy measurements, hourly Bretschneider spectra were computed using the predicted Hs and Tp. Then, the daily-averaged <Hs>_model and <Tp>_model were estimated from daily-averaged spectra.
Comparison between predicted (blue) and measured (red) (top) significant wave heights and (bottom) peak periods, with HFRs Bragg periods shown with dotted lines (STF: red, PAO: blue).
Citation: Journal of Atmospheric and Oceanic Technology 33, 3; 10.1175/JTECH-D-15-0143.1
Observed vs predicted Bragg energy densities for CODAR and WERA. Fitted linear relationships are shown in red and the 1:1 lines are shown in green.
Citation: Journal of Atmospheric and Oceanic Technology 33, 3; 10.1175/JTECH-D-15-0143.1
d. Normalized ranges versus ice concentration
For both HFR types, CODAR (STF) and WERA (PAO), Ri obtained during winter 2013 were normalized by Rm. Figure 13 shows the performance Γ = Ri/Rm versus the daily-averaged concentration of sea ice spatially averaged over the entire observational area shown in Fig. 4. Although the normalization of ranges does not affect dramatically the results (cf. Figs. 7 and 13), it allows for comparing the different HFRs together. The performance Γ sometimes exceeds 1 because the fit used to estimate the expected ice-free coverage is a fit to data with a strong scatter (Fig. 8).
CODAR and WERA HFR normalized daily mean ranges (Γ) vs ice concentration. Least squares fitted linear relationships are shown by black lines.
Citation: Journal of Atmospheric and Oceanic Technology 33, 3; 10.1175/JTECH-D-15-0143.1
4. Discussion and summary
The fact that CODAR and WERA were operating at different frequencies, hence required different Bragg wavelengths (Table 1), has been taken into account by computing the wave energy at the Bragg frequencies of each radar. We found that both WERA and CODAR antennas were sensitive to the Bragg wave energy density with slightly different relationships (Fig. 8). However, sensitivity to sea ice concentration was similar for both HFR types (Fig. 13). Indeed, backscattering of the signal transmitted by the HF radar is due to the short ocean waves, which are rapidly damped by sea ice, suppressing the constructively interfering return signal. Moreover, the presence of sea ice limits the wind fetch over adjacent open waters, therefore limiting the development of Bragg waves for a given wind speed and duration (WMO 1998). Normalized ranges are sometimes small for CODAR and WERA at low ice concentration and this may be due to the presence of frazil, or landfast ice along the coast in front of the instrument, which was not monitored during this study. Indeed, during winter, ice melts and reforms alternately according to air and sea surface temperature and wind conditions. Before sea ice consolidates and emerges from the water surface as gray-white brash or floes, ice crystals form in the water, which are mixed down to a certain depth (a few tens of centimeters typically) depending on the wave activity. This is called frazil ice and this type of ice does not appear explicitly in ice analyses or in models and is invisible for satellite remote visible or radar sensors. However, frazil can affect the HFR ranges by attenuating ocean waves, decreasing the wave-generation rate by the wind, and modifying the surface salinity. The ice concentration provided by the Canadian operational ice–ocean forecasting system does not represent frazil and neither can we measure it.
The relatively large distribution of the scatters may also be due to uncertainties in our normalization method (large scatter around the relationship between ice-free radar coverage and Bragg energy density, see Fig. 8; uncertainties in GENER Bragg energy density predictions, see Fig. 12), to ambient radio noise variability, and to sea surface salinity variability (Crombie 1955; Potter and Weingartner 2010; Gurgel et al. 1999a; Barrick and Long 2006). Although the noise level was very variable, there is no clear relationship with the HFR range (Fig. 14). Another important factor that can affect the HFR performance is the spatial distribution of sea ice. For example, for approximately the same predicted spatially averaged ice concentration (≈0.2), the spatial distribution of sea ice can vary significantly (Fig. 15), and the corresponding HFR average ranges also differ significantly from 23 January 2013 (≈38.6 km for STF and ≈36 km for PAO) to 31 January 2013 (≈11 km for STF and ≈12.5 km for PAO). However, taking this into account is beyond the scope of the present study, which focused on obtaining a simple relationship relating HFR ranges to sea ice concentration for future site planning purposes.
Daily mean range of the WERA radar vs the ambient radio noise detected by antenna 10.
Citation: Journal of Atmospheric and Oceanic Technology 33, 3; 10.1175/JTECH-D-15-0143.1
Ice spatial distribution for similar spatially averaged ice concentration over the study area (≈0.2). The average range for the corresponding HFRs is about 38.6 km for STF and 36 km for PAO on 23 Jan 2013 and about 11 km for STF and 12.5 km for PAO on 31 Jan 2013.
Citation: Journal of Atmospheric and Oceanic Technology 33, 3; 10.1175/JTECH-D-15-0143.1
Despite the various limitations of our study (predicted rather than observed sea ice concentration, simple parametric wave model to predict expected ice-free Bragg energy densities in winter), it is the first study to quantify experimentally the relationship between HFR ranges and sea ice concentration. Empirical relationships between HFR ranges and environmental parameters (wind and sea ice) will allow for predicting ranges that could be achieved by HFRs installed in other seasonally ice-covered areas.
Acknowledgments
We thank Céline Quentin for the helpful suggestions that improved an earlier version of this work. We acknowledge the University Mission of Tunisia in Montreal, the FRQNT, the Marine Environmental, Observation, Prediction and Response Network (MEOPAR), Canada Economic Development for Quebec Regions, and Quebec-Ocean for their financial support. We thank Urs Neumeier for producing wave predictions using the GENER model, Simon Senneville and Simon St-Onge Drouin for providing the GEM winds, Fisheries and Oceans Canada for the IML-4 buoy data, and James Caveen for his technical support.
REFERENCES
Barrick, D. E., and Long R. , 2006: How salinity affects radar performance: A case study of 24 MHz HF radar performance in San Francisco Bay during the 2006 New Year storm floor. CODAR White Paper, 7 pp.
Bjorkstedt, B. P., and Roughgarden J. , 1997: Larval transport and coastal upwelling: An application of HF radar in ecological research. Oceanography, 10 (2), 64–67, doi:10.5670/oceanog.1997.25.
Bretschneider, C. L., 1959: Wave variability and wave spectra for wind-generated gravity waves. U.S. Army Corps of Engineers Tech. Memo. No. 118, 192 pp.
Côté, J., Gravel S. , Méthot A. , Patoine A. , Roch M. , and Staniforth A. , 1998: The operational CMC–MRB Global Environmental Multiscale (GEM) model. Part I: Design considerations and formulation. Mon. Wea. Rev., 126, 1373–1395, doi:10.1175/1520-0493(1998)126<1373:TOCMGE>2.0.CO;2.
Coulliette, C., Lekien F. , Paduan J. D. , Haller G. , and Marsden J. E. , 2007: Optimal pollution mitigation in Monterey Bay based on coastal radar data and nonlinear dynamics. Environ. Sci. Technol., 41, 6562–6572, doi:10.1021/es0630691.
Crombie, D. D., 1955: Doppler spectrum of sea echo at 13.56 Mc./s. Nature, 175, 681–682, doi:10.1038/175681a0.
Desjardins, L., and Ouellet Y. , 1984: Modèles numériques utilisés pour la conception des ouvrages maritimes. Comptes rendus du Colloque sur la simulation numérique appliquée au domaine de la ressource hydrique: Dans le cadre du 52e congrès annuel de l’ACFAS, Université Laval, 187–224.
Drapeau, G., 1992: Dynamique sédimentaire des littoraux de l’estuaire du Saint-Laurent. Geogr. Phys. Quat., 46, 233–242, doi:10.7202/032907ar.
Dumont, D., Kohout A. , and Bertino L. , 2011: A wave‐based model for the marginal ice zone including a floe breaking parameterization. J. Geophys. Res., 116, C04001, doi:10.1029/2010JC006682.
Flocco, D., Falco P. , Wadhams P. , and Spezie G. , 2003: Surface current measurements in Terra Nova Bay by HF radar. Antarct. Sci., 15, 55–62, doi:10.1017/S0954102003001147.
Galbraith, P. S., and Coauthors, 2014: Physical oceanographic conditions in the Gulf of St. Lawrence in 2013. Fisheries and Oceans Canada Doc. 2014/062, vi + 84 pp.
Graber, H. C., and Limouzy-Paris C. B. , 1997: Transport patterns of tropical reef fish larvae by spin-off eddies in the Straits of Florida. Oceanography, 10 (2), 68–71, doi:10.5670/oceanog.1997.26.
Gurgel, K.-W., Antonischki G. , Essen H.-H. , and Schlick T. , 1999a: Wellen Radar (WERA): A new ground-wave HF radar for ocean remote sensing. Coastal Eng., 37, 219–234, doi:10.1016/S0378-3839(99)00027-7.
Gurgel, K.-W., Essen H.-H. , and Kingsley S. P. , 1999b: HF radars: Physical limitations and recent developments. Coastal Eng., 37, 201–218, doi:10.1016/S0378-3839(99)00026-5.
Headrick, J. M., and Skolnik M. I. , 1974: Over-the-Horizon radar in the HF band. Proc. IEEE, 62, 664–673, doi:10.1109/PROC.1974.9506.
Hodgins, D. O., 1994: Remote sensing of ocean surface currents with the SeaSonde HF radar. Spill Sci. Technol. Bull., 1, 109–129, doi:10.1016/1353-2561(94)90020-5.
Ingram, R. G., and El-Sabh M. I. , 1990: Fronts and mesoscale features in the St Lawrence estuary. Oceanography of a Large-Scale Estuarine System, M. I. El-Sabh and N. Silverberg, Eds., Coastal and Estuarine Studies, Vol. 39, Amer. Geophys. Union, 71–93, doi:10.1002/9781118663783.ch4.
Lekien, F., Coulliette C. , Mariano A. J. , Ryan E. H. , Shay L. K. , Haller G. , and Marsden J. , 2005: Pollution release tied to invariant manifolds: A case study for the coast of Florida. Physica D, 210, 1–20, doi:10.1016/j.physd.2005.06.023.
Lipa, B. J., Crissman R. D. , and Barrik D. E. , 1986: HF radar observations of Arctic pack-ice breakup. IEEE. J. Oceanic Eng., 11, 270–275, doi:10.1109/JOE.1986.1145160.
Ochi, M. K., 1998: Ocean Waves: The Stochastic Approach. Cambridge Ocean Technology Series, Vol. 6, Cambridge University Press, 332 pp.
Olascoaga, M. J., Rypina I. I. , Brown M. G. , Beron-Vera F. J. , Koak H. , Brand L. E. , Halliwell G. R. , and Shay L. K. , 2006: Persistent transport barrier on the West Florida Shelf. Geophys. Res. Lett., 33, L22603, doi:10.1029/2006GL027800.
Potter, R. A., and Weingartner T. J. , 2010: Surface circulation radar mapping in Alaskan coastal waters: Beaufort Sea and Cook Inlet. University of Alaska Fairbanks Rep. OCS Study MMS 2009-049, 144 pp.
Ruest, B., Neumeier U. , Dumont D. , and Lambert A. , 2013: Wave climate evaluation in the Gulf of St. Lawrence with a parametric wave model. Extended Abstracts of Coastal Dynamics 2013: Coastal Dynamics Research Emphasizing Practical Applications, P. Bonneton and T. Garlan, Eds., SHOM, 1363–1374. [Available online at http://www.coastaldynamics2013.fr/pdf_files/132_Ruest_Benoit.pdf.]
Ruest, B., Neumeier U. , Dumont D. , Bismuth E. , Senneville S. , and Caveen J. , 2016: Recent wave climate and expected future changes in the seasonally ice-infested waters of the Gulf of St. Lawrence, Canada. Climate Dyn., 46, 449–466, doi:10.1007/s00382-015-2592-3.
Saucier, F. J., Roy F. , Gilbert D. , Pellerin P. , and Ritchie H. , 2003: Modeling the formation and circulation processes of water masses and sea ice in the Gulf of St. Lawrence, Canada. J. Geophys. Res., 108, 3269, doi:10.1029/2000JC000686.
Smith, G. C., Roy F. , and Brasnett B. , 2012: Evaluation of an operational ice–ocean analysis and forecasting system for the Gulf of St Lawrence. Quart. J. Roy. Meteor. Soc., 139, 419–433, doi:10.1002/qj.1982.
Squire, V. A., 2007: Review of ocean waves and sea-ice revisited. Cold Reg. Sci. Technol., 49, 110–133, doi:10.1016/j.coldregions.2007.04.007.
Ullman, D. S., O’Donnell J. , Kohut J. , Fake T. , and Allen A. , 2006: Trajectory prediction using HF radar surface currents: Monte Carlo simulations of prediction uncertainties. J. Geophys. Res., 111, C12005, doi:10.1029/2006JC003715.
WMO, 1998: Guide to wave analysis and forecasting. 2nd ed. WMO-702, 159 pp.