• Andrade-Canto, F., J. Sheinbaum, and L. Zavala Sansón, 2013: A Lagrangian approach of the Loop Current eddy separation. Nonlinear Processes Geophys., 20, 8596, doi:10.5194/npg-20-85-2013.

    • Search Google Scholar
    • Export Citation
  • Athie, G., J. Candela, J. Ochoa, and J. Sheinbaum, 2012: Impact of Caribbean cyclones on the detachment of Loop Current anticyclones. J. Geophys. Res., 117, C03018, doi:10.1029/2011JC007090.

    • Search Google Scholar
    • Export Citation
  • Beron-Vera, F. J., M. J. Olascoaga, and G. J. Goni, 2008: Oceanic mesoscale eddies as revealed by Lagrangian coherent structures. Geophys. Res. Lett., 35, L12603, doi:10.1029/2008GL033957.

    • Search Google Scholar
    • Export Citation
  • Bretherton, F. P., R. E. Davis, and C. B. Fandry, 1976: A technique for objective analysis and design of oceanographic experiments applied to MODE-73. Deep-Sea Res. Oceanogr. Abstr., 23, 559582, doi:10.1016/0011-7471(76)90001-2.

    • Search Google Scholar
    • Export Citation
  • Brodeau, L., B. Barnier, A.-M. Treguier, T. Penduff, and S. Gulev, 2010: An ERA40-based atmospheric forcing for global ocean circulation models. Ocean Modell., 31, 88104, doi:10.1016/j.ocemod.2009.10.005.

    • Search Google Scholar
    • Export Citation
  • Candela, J., J. Sheinbaum, J. Ochoa, A. Badan, and R. Leben, 2002: The potential vorticity flux through the Yucatan Channel and the Loop Current in the Gulf of Mexico. Geophys. Res. Lett., 29, 2059, doi:10.1029/2002GL015587.

    • Search Google Scholar
    • Export Citation
  • Chang, Y.-L., and L.-Y. Oey, 2012: Why does the Loop Current tend to shed more eddies in summer and winter? Geophys. Res. Lett., 39, L05605, doi:10.1029/2011GL050773.

    • Search Google Scholar
    • Export Citation
  • Chérubin, L. M., Y. Morel, and E. P. Chassignet, 2006: Loop current ring shedding: The formation of cyclones and the effect of topography. J. Phys. Oceanogr., 36, 569591, doi:10.1175/JPO2871.1.

    • Search Google Scholar
    • Export Citation
  • Dai, A., and K. Trenberth, 2002: Estimates of freshwater discharge from continents: Latitudinal and seasonal variations. J. Hydrometeor., 3, 660687, doi:10.1175/1525-7541(2002)003<0660:EOFDFC>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Donohue, K. A., D. R. Watts, P. Hamilton, R. Leben, M. Kennelly, and A. Lugo-Fernández, 2016: Gulf of Mexico Loop Current path variability. Dyn. Atmos. Oceans, doi:10.1016/j.dynatmoce.2015.12.003, in press.

    • Search Google Scholar
    • Export Citation
  • Dubranna, J., P. Brunius, J. M. Lopez, and J. Candela, 2011: Circulation over the continental shelf of the western and southwestern Gulf of Mexico. J. Geophys. Res., 116, C08009, doi:10.1029/2011JC007007.

    • Search Google Scholar
    • Export Citation
  • Dussin, R., B. Barnier, and L. Brodeau, 2014: The making of Drakkar forcing set DFS5. DRAKKAR/MyOcean Rep. 05-10-14, 29 pp. [Available online at https://www.drakkar-ocean.eu/forcing-the-ocean/the-making-of-the-drakkar-forcing-set-dfs5.]

  • Fratantoni, P. S., T. N. Lee, G. P. Podesta, and F. Muller-Karger, 1998: The influence of Loop Current perturbations on the formation and evolution of Tortugas eddies in the southern Straits of Florida. J. Geophys. Res., 103, 24 75924 779, doi:10.1029/98JC02147.

    • Search Google Scholar
    • Export Citation
  • Hallock, Z. R., W. J. Teague, and E. Jarosz, 2009: Subinertial slope-trapped waves in the northeastern Gulf of Mexico. J. Phys. Oceanogr., 39, 14751485, doi:10.1175/2009JPO3925.1.

    • Search Google Scholar
    • Export Citation
  • Hurlburt, H. E., and J. D. Thompson, 1982: The dynamics of the Loop Current and shed eddies in a numerical model of the Gulf of Mexico. Hydrodynamics of Semi-Enclosed Seas, J. C. J. Nihoul, Ed., Elsevier, 243–298.

  • Large, W. L., and S. G. Yeager, 2004: Diurnal to decadal global forcing for ocean and sea-ice models: The data sets and flux climatologies. NCAR Tech. Rep. TN-460+STR, 105 pp.

  • Leben, R. R., 2005: Altimeter-derived loop current metrics. Circulation in the Gulf of Mexico: Observations and Models, Geophys. Monogr., Vol. 161, Amer. Geophys. Union, 181–201.

  • Leclair, M., and G. Madec, 2009: A conservative leapfrog time stepping method. Ocean Modell., 30, 8894, doi:10.1016/j.ocemod.2009.06.006.

    • Search Google Scholar
    • Export Citation
  • Le Hénaff, M., V. H. Kourafalou, Y. Morel, and A. Srinivasan, 2012: Simulating the dynamics and intensification of cyclonic Loop Current frontal eddies in the Gulf of Mexico. J. Geophys. Res., 117, C02034, doi:10.1029/2011JC007279.

    • Search Google Scholar
    • Export Citation
  • Le Hénaff, M., V. H. Kourafalou, R. Dussurget, and R. Lumpkin, 2014: Cyclonic activity in the eastern Gulf of Mexico: Characterization from along-track altimetry and in situ drifter trajectories. Prog. Oceanogr., 120, 120138, doi:10.1016/j.pocean.2013.08.002.

    • Search Google Scholar
    • Export Citation
  • Luther, M. E., and J. M. Bane, 1985: Mixed instabilities in the Gulf Stream over the continental slope. J. Phys. Oceanogr., 15, 323, doi:10.1175/1520-0485(1985)015<0003:MIITGS>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Madec, G., 2016: NEMO ocean engine. Institut Pierre-Simon Laplace Note du Pole de Modélisation 27, 406 pp.

  • Masina, S., S. G. H. Philander, and A. B. G. Bush, 1999: An analysis of tropical instability waves in a numerical model of the Pacific Ocean. 2: Generation and energetics of the waves. J. Geophys. Res., 104, 29 63729 661, doi:10.1029/1999JC900226.

    • Search Google Scholar
    • Export Citation
  • Nencioli, F., C. Dong, T. Dickey, L. Washburn, and J. C. McWilliams, 2010: A vector geometry-based eddy detection algorithm and its application to a high-resolution numerical model product and high-frequency radar surface velocities in the Southern California Bight. J. Atmos. Oceanic Technol., 27, 564579, doi:10.1175/2009JTECHO725.1.

    • Search Google Scholar
    • Export Citation
  • Oey, L.-Y., 2008: Loop Current and deep eddies. J. Phys. Oceanogr., 38, 14261449, doi:10.1175/2007JPO3818.1.

  • Pichevin, T., and D. Nof, 1997: The momentum imbalance paradox. Tellus, 49A, 298319, doi:10.1034/j.1600-0870.1997.t01-1-00009.x.

  • Reffray, G., R. Bourdalle-Badie, and C. Calone, 2015: Modelling turbulent vertical mixing sensitivity using a 1-D version of NEMO. Geosci. Model Dev., 8, 6986, doi:10.5194/gmd-8-69-2015.

    • Search Google Scholar
    • Export Citation
  • Renault, L., M. J. Molemaker, J. C. McWilliams, A. F. Shchepetkin, F. Lemarié, D. Chelton, S. Illig, and A. Hall, 2016: Modulation of wind work by oceanic current interaction with the atmosphere. J. Phys. Oceanogr., 46, 16851704, doi:10.1175/JPO-D-15-0232.1.

    • Search Google Scholar
    • Export Citation
  • Rudnick, D. L., G. Gopalakrishnan, and B. D. Cornuelle, 2015: Cyclonic eddies in the Gulf of Mexico: Observations by underwater gliders and simulations by numerical model. J. Phys. Oceanogr., 45, 313326, doi:10.1175/JPO-D-14-0138.1.

    • Search Google Scholar
    • Export Citation
  • Schmitz, W. J., Jr., 2005: Cyclones and westward propagation in the shedding of anticyclonic rings from the Loop Current. Circulation in the Gulf of Mexico: Observations and Models, Geophys. Monogr., Vol. 161, Amer. Geophys. Union, 241–261.

  • Schultz, D. M., W. E. Bracken, and L. F. Bosart, 1998: Planetary- and synoptic-scale signatures associated with Central American cold surges. Mon. Wea. Rev., 126, 527, doi:10.1175/1520-0493(1998)126<0005:PASSSA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Seo, H., A. J. Miller, and J. R. Norris, 2016: Eddy–wind interaction in the California Current System: Dynamics and impacts. J. Phys. Oceanogr., 46, 439459, doi:10.1175/JPO-D-15-0086.1.

    • Search Google Scholar
    • Export Citation
  • Sheinbaum, J., G. Athie, J. Candela, J. Ochoa, and A. Romero, 2016: Structure and variability of the Yucatan and Loop Currents along the slope and shelf break of the Yucatan Channel and Campeche Bank. Dyn. Atmos. Oceans, doi:10.1016/j.dynatmoce.2016.08.001, in press.

    • Search Google Scholar
    • Export Citation
  • Sturges, W., N. G. Hoffmann, and R. R. Leben, 2010: A trigger mechanism for Loop Current ring separations. J. Phys. Oceanogr., 40, 900913, doi:10.1175/2009JPO4245.1.

    • Search Google Scholar
    • Export Citation
  • Sudre, J., C. Maes, and V. Garçon, 2013: On the global estimates of geostrophic and Ekman surface currents. Limnol. Oceanogr. Fluids Environ., 3, 120, doi:10.1215/21573689-2071927.

    • Search Google Scholar
    • Export Citation
  • Vukovich, F. M., and G. A. Maul, 1985: Cyclonic eddies in the eastern Gulf of Mexico. J. Phys. Oceanogr., 15, 105117, doi:10.1175/1520-0485(1985)015<0105:CEITEG>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Walker, N., R. Leben, S. Anderson, J. Feeney, P. Coholan, and N. Sharma, 2009: Loop Current frontal eddies based on satellite remote-sensing and drifter data. U.S. Dept. of the Interior, Minerals Management Service, Gulf of Mexico OCS Region, OCS Study MMS 2009-023, 88 pp.

  • Zavala-Hidalgo, J., S. L. Morey, and J. J. O’Brien, 2003a: Cyclonic eddies northeast of the Campeche Bank from altimetry data. J. Phys. Oceanogr., 33, 623629, doi:10.1175/1520-0485(2003)033<0623:CENOTC>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Zavala-Hidalgo, J., S. L. Morey, and J. J. O’Brien, 2003b: Seasonal circulation on the western shelf of the Gulf of Mexico using a high-resolution numerical model. J. Geophys. Res., 108, 3389, doi:10.1029/2003JC001879.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Sea surface temperature (°C) over the LC on (a) 4 Feb 2011 and (b) 6 Feb 2011. It illustrates the presence of a cyclonic eddy with a horizontal scale of about 100 km forming on the western front of the LC at the location of the mooring array. Temperatures are from the MUR SST product. (c) Sea level height (m) and corresponding surface geostrophic currents (m s−1) from AVISO on 6 Feb 2011. Contours represent the 100-, 500-, 1000-, and 2000-m isobaths. The PN moorings locations are indicated with black dots, and vectors represent the daily mean current average between 50 and 100 m, obtained from ADCP measurements.

  • View in gallery

    Cross-section horizontal velocity (m s−1) obtained from rotated ADCP observations at the PN array (see Fig. 1), between January and April 2011. Positive values are associated with currents oriented northwestward. The currents are vertically averaged between different depths: (a) 50–100 m, (b) 200–300 m, and (c) 400–500 m. Sea level height (m) observations from AVISO have been interpolated along the PN section and are represented with labeled contours. The bold points at 1 May indicate the latitude of the PN moorings. The vertical dashed lines indicate 4 and 6 Feb, when a frontal eddy is crossing the mooring array (see Fig. 1).

  • View in gallery

    Objective analysis of (left) cross-section velocity (m s−1) and (right) relative vorticity (s−1) through the PN array for (from top to bottom) 3, 4, 5, and 6 Feb. Mooring velocities have been low-pass filtered with a cutoff period of 48 h and interpolated using objective mapping (Bretherton et al. 1976) with horizontal decorrelation scale Ls = 20 km and vertical decorrelation scale Lz = 150 m.

  • View in gallery

    Multitaper variance-conserving power spectra (m−2 s−2) of cross-section velocity at mooring PN2 for three different depth ranges (as in Fig. 2) from (a) observations and (b) model. The spectra in (a) are built with in situ data from January to May 2011 and five windows and in (b) using model data from January to May 2008 and five windows. Dotted lines represent the corresponding 90% confidence limit based on the theoretical spectra of an AR(2) process with variances equal to that of the analyzed signals. The number of degrees of freedom for the 90% level calculation is estimated as 2K − 1, with K as the number of windows.

  • View in gallery

    Model domain and mean sea level height (m) simulated by the model. Contours represent 30-, 500-, and 3000-m isobaths. The red points mark distances of 500 km along the 30-m isobath in order to help the interpretation of the Hovmöller diagram in Fig. 19. The dashed black line along the CB indicates the path used to build the Hovmöller diagram in Fig. 21d.

  • View in gallery

    (a),(c) Mean surface currents (m s−1) and (b),(d) surface eddy kinetic energy (m2 s−2) from (left) observations and (right) the model for the period 2003–12. Observed surface currents are from the GEKCO database (Sudre et al. 2013) and consist of the sum of altimetry-derived geostrophic currents and estimated Ekman currents.

  • View in gallery

    Cross-section horizontal velocity (m s−1) of numerical model along the PN mooring section from January to April 2008. The currents are vertically averaged between different depths: (a) 50–100, (b) 200–300, and (c) 400–500 m. The model sea level height (m) is shown in contours. Dashed lines indicate 6, 10, and 14 Mar 2008 for comparison with model snapshots shown in Fig. 8.

  • View in gallery

    (a)–(c) Model sea level (m), (d)–(f) surface temperature (°C), (g)–(i) relative vorticity at 100 m (s−1), and (j)–(l) FTLE at 100 m (day−1) at 6, 10, and 14 Mar 2008. The black line represents the position of the PN section as used in Fig. 7. The black contours in (a),(b), and (c) show the eddies identified by the tracking method.

  • View in gallery

    (a)Variance of the relative vorticity (s−2) at 100 m. Multitaper variance-conserving power spectra of surface vorticity (s−2), computed at each model grid point and averaged over four different zones along the continental shelf of Yucatan Coast and Campeche Bank for experiments (b) REF and (c) MONTHLY. The four zones are shown in (a). Model data from January 2003 to December 2012 and 29 windows are used for this spectral analysis.

  • View in gallery

    Time–latitude diagram of (a) model SSH along the mean Loop Current position (identified with the 0.1 m isocontour of mean SSH as indicated with a dashed line in Fig. 5), (b) high-pass filtered model SSH anomalies, and (c) low-pass filtered SSH anomalies. A cutoff period of 30 days is used to separate the high-frequency (b) and low-frequency (c) contributions to the SSH signal. Vertical dashed lines indicate the time of LCE detachments (identified by visual inspection of the model outputs).

  • View in gallery

    Linear lagged regressions of model sea level height (m) on a North Campeche LCFE index for (top) model and (bottom) altimetry for the period 2003 to 2012. Lags range from −45 to 0 days. To compute the index, the SSH is spatially averaged in the area from 23.8° to 24.2°N and from 88.4° to 87.2°W, that is, north CB in the area of the PN moorings array. The index is filtered with a Lanczos filter to retain the variability between 30 and 90 days and weighted by the standard deviation. The anomalous patterns are linearly correlated with one standard deviation anomaly of the SSH north of CB and units of the regression coefficients are meters. Data were high-pass filtered (90-day cutoff period) before the computation of the regression coefficients. Regression coefficients less than the 99% confidence level are set to 0. Contours represent the 100- and 500-m isobaths.

  • View in gallery

    Eddy kinetic energy (m2 s−2) computed from surface currents filtered in different bands of periods: 0–15, 15–30, and 30–90 days, in experiments (a)–(c) REF and (d)–(f) MONTHLY. Data from January 2003 to December 2012 are used. Contours represent the 100-, 500-, and 1000-m isobaths.

  • View in gallery

    Tracks and statistics of eddies identified during the period 2003–12 and lasting more than 5 days: (a) trajectories of the eddies (cyclones in blue, anticyclones in red), (b) origin position of the cyclones, (c) final position of the cyclones, (d) eddy lifetime (days) of the cyclones, and (e) diameter of the cyclones (km). (f) The LC path in has been identified as a mean sea level isocontour of 0.1 m and (g) is used to sort the diameter of the cyclones along the path of the LC . The color codes in (f) and (g) correspond. The crosses in (f) indicate locations where merging has been identified (in the sense that an eddy is identified within another one).

  • View in gallery

    Maps of baroclinic conversion of (a),(e) EKE (kg m−1 s−3), (b),(f) barotropic conversion of EKE (kg m−1 s−3), (c),(g) vertical velocity shear (s−2), and (d),(h) horizontal velocity shear (s−2) for (top) simulation REF and (bottom) simulation MONTHLY. In (a),(b),(e), and (f), positive means a transfer of energy toward the turbulent field. Mean velocity shear is computed as the time-mean velocity shear computed from the 30-day low-pass filtered current ULP and VLP. Energy transfer terms and velocity shears are averaged between the surface and 100-m depth and are computed using daily model outputs from 2003 to 2012.

  • View in gallery

    (a) Multitaper variance-conserving power spectra of the zonal and meridional components of the wind stress averaged over the northwestern corner of the Gulf of Mexico. Data were averaged between 24° and 30°N and 98° and 90°W before spectral analysis. The spectra are built with model data from January 2003 to December 2012 and 29 windows. (b) Wavelet of the wind stress averaged over the same region as in (a). The power spectra is normalized by σ2 the variance of the corresponding signal. The thick contour encloses regions of greater than 90% confidence for an AR(2) process. Vertical dashed lines indicate 1 Jan of each year. (c) Longitudinal position of the LC front at 24°N, identified by the position of the maximum gradient of SSH between 89° and 86°W.

  • View in gallery

    (a),(b) Winter (November–March) and (c),(d) summer (May–August) mean eddy kinetic energy (m2 s−2) in experiments (a),(c) REF and (b),(d) MONTHLY computed from surface currents high-pass filtered (15-day cutoff period). Only periods with the LC front at 24°N located between 88° and 87°W (see Fig. 15c) are retained in the average.

  • View in gallery

    Linear lagged regressions of model sea level height (m) and wind stress (N m−2) on a cold surge wind index. Lags range from 0 to 11 days. To compute the index, the southward meridional wind stress is spatially averaged in the northwestern Gulf of Mexico (from 24° to 30°N and from 98° to 90°W), filtered with a high-pass Lanczos filter to retain the variability under 15 days and weighted by the standard deviation. Thus the anomalous patterns are linearly correlated with one standard deviation anomaly of the meridional wind stress in the northwestern Gulf of Mexico (about −0.05 N m−2) and units of the regression coefficients are N m−2 for the wind stress and meters for the SSH. Data were high-pass filtered (30-day cutoff period) before the computation of the regression coefficients. Regression coefficients less than the 90% confidence level are set to 0. Red (gray) vectors indicate the northward (southward) anomaly of the regressed meridional component of the wind stress. Contours represent the 100- and 500-m isobaths. See text for explanation of letter labels.

  • View in gallery

    Lagged cross-correlations based on model fields between (a) SSH at different locations, (b) zonal wind stress and SSH, (c) meridional wind stress and SSH, and (d) SSH and surface relative vorticity at the location of the PN array. Time series were high-pass filtered (cutoff period of 30 days) and normalized before computation of the correlation coefficients. The positions of the time series are indicated with labels in the different figures and correspond to the positions of the tide gauges indicated in Table 2.

  • View in gallery

    (a) Along-coast Hovmöller diagrams of wind stress (positive means that the along-coast component of the wind stress is oriented with the land on the right-hand side) and sea level anomalies (m) in (b) REF experiment and (c) MONTHLY experiment. The Hovmöller diagrams are built using data along a path described by the 30-m isobath from the Florida coast to the Yucatan coast. To compute the alongshore component, the wind stress has been rotated using local orientation of the 30-m isobaths. Daily data from January to April 2008 are used. Anomalies are computed by subtracting the mean values of the sea level height during this period. The x axis indicates the distance from the starting point on the west Florida shelf (km) together with the corresponding longitudes and latitudes. The path together with points separated by 500 km distance is shown in Fig. 5.

  • View in gallery

    a) Position of the 14 tide gauges used to estimate propagation speeds of coastal sea level signals. b) Lagged cross correlation between selected tide gauges a: RK-CD, b: VE-MU, c: PT-PR, and d: GA-PR. c) Lag (days) of the maximum cross-correlation coefficient (as shown in b) against distance along the coast between stations (103 km) computed for each of the tide gauges pairs with separation distance higher than 500 km. A least squares fit gives a mean phase speed estimate of 3.4 m s−1.

  • View in gallery

    Hovmöller diagrams of (a) SSH (m) and contours of zonal gradients of SSH (contours show values of 0.8, 0.9, 1, 1.1, and 1.2 × 10−5), (b) surface speed (m s−1), (c) vorticity (s−1) at 21.5°N, and (d) vorticity along the shelf break of the Campeche Bank (s−1). The path used to build the along-shelf diagram in (d) is indicated in Fig. 5 with a dashed black line. In (d), the x axis represents the distance from Cozumel Channel. Model outputs from REF experiment between January and March 2008 are used, as in Fig. 19b.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 200 192 13
PDF Downloads 197 190 19

Loop Current Frontal Eddies: Formation along the Campeche Bank and Impact of Coastally Trapped Waves

View More View Less
  • 1 LEGOS, Université de Toulouse, IRD, CNRS, CNES, UPS, Toulouse, France
  • | 2 Departamento de Oceanografía Física, CICESE, Ensenada, Baja California, Mexico
  • | 3 CNRS, Université Grenoble Alpes, LGGE (UMR5183), F-38041, Grenoble, France
© Get Permissions
Full access

Abstract

Velocity data from a mooring array deployed northeast of the Campeche Bank (CB) show the presence of subinertial, high-frequency (below 15 days) velocity fluctuations within the core of the northward flowing Loop Current. These fluctuations are associated with the presence of surface-intensified Loop Current frontal eddies (LCFEs), with cyclonic vorticity and diameter < 100 km. These eddies are well reproduced by a high-resolution numerical simulation of the Gulf of Mexico, and the model analysis suggests that they originate along and north of the CB, their main energy source being the mixed baroclinic–barotropic instability of the northward flow along the shelf break. There is no indication that these high-frequency LCFEs contribute to the LC eddy detachment in contrast to the low-frequency LCFEs (periods > 30 days) that have been linked to Caribbean eddies and the LC separation process. Model results show that wind variability associated with winter cold surges are responsible for the emergence of high-frequency LCFEs in a narrow band of periods (6–10 day) in the region of the CB. The dynamical link between the formation of these LCFEs and the wind variability is not direct: (i) the large-scale wind perturbations generate sea level anomalies on the CB as well as first baroclinic mode, coastally trapped waves in the western Gulf of Mexico; (ii) these waves propagate cyclonically along the coast; and (iii) the interaction of these anomalies with the Loop Current triggers cyclonic vorticity perturbations that grow in intensity as they propagate downstream and develop into cyclonic eddies when they flow north of the Yucatan shelf.

Corresponding author address: Julien Jouanno, LEGOS, 14 Ave. Edouard Belin, 31400 Toulouse, France. E-mail: julien.jouanno@ird.fr

Abstract

Velocity data from a mooring array deployed northeast of the Campeche Bank (CB) show the presence of subinertial, high-frequency (below 15 days) velocity fluctuations within the core of the northward flowing Loop Current. These fluctuations are associated with the presence of surface-intensified Loop Current frontal eddies (LCFEs), with cyclonic vorticity and diameter < 100 km. These eddies are well reproduced by a high-resolution numerical simulation of the Gulf of Mexico, and the model analysis suggests that they originate along and north of the CB, their main energy source being the mixed baroclinic–barotropic instability of the northward flow along the shelf break. There is no indication that these high-frequency LCFEs contribute to the LC eddy detachment in contrast to the low-frequency LCFEs (periods > 30 days) that have been linked to Caribbean eddies and the LC separation process. Model results show that wind variability associated with winter cold surges are responsible for the emergence of high-frequency LCFEs in a narrow band of periods (6–10 day) in the region of the CB. The dynamical link between the formation of these LCFEs and the wind variability is not direct: (i) the large-scale wind perturbations generate sea level anomalies on the CB as well as first baroclinic mode, coastally trapped waves in the western Gulf of Mexico; (ii) these waves propagate cyclonically along the coast; and (iii) the interaction of these anomalies with the Loop Current triggers cyclonic vorticity perturbations that grow in intensity as they propagate downstream and develop into cyclonic eddies when they flow north of the Yucatan shelf.

Corresponding author address: Julien Jouanno, LEGOS, 14 Ave. Edouard Belin, 31400 Toulouse, France. E-mail: julien.jouanno@ird.fr

1. Introduction

The Loop Current (LC) is the most characteristic feature of the circulation in the eastern Gulf of Mexico. It takes the form of an anticyclonic flowing ribbon, which joins the Yucatan Current to the Florida Current and episodically sheds large anticyclonic eddies with diameters on the order of 200–400 km; these are called Loop Current eddies (LCEs). The shedding of LCEs is a complicated process since eddies can detach and reattach to the LC several times before they travel to the west of the basin. Different mechanisms have been proposed to explain the LCEs shedding: momentum imbalance (Pichevin and Nof 1997), barotropic–baroclinic instability of the LC (Hurlburt and Thompson 1982), pulses of increased transport through the Florida Straits (Sturges et al. 2010), seasonal modulation by the local wind stress (Chang and Oey 2012), advection of cyclonic vorticity from the Cayman Basin (Athie et al. 2012), or interaction with cyclonic eddies generated locally (Fratantoni et al. 1998; Zavala-Hidalgo et al. 2003a; Chérubin et al. 2006; Le Hénaff et al. 2012).

The presence of cyclonic eddies moving along the outer edge of the LC, known as Loop Current frontal eddies (LCFEs), has been documented all along the LC: northeast of the Campeche Bank (CB), on its northern edge, and along the coast of Florida. Vukovich and Maul (1985) were the first to observe cyclonic eddies on the northern and eastern fractions of the LC. They showed that these cyclones take the form of cold domes with a diameter between 80 and 120 km located on the LC cyclonic shear side. Using altimetry and drifter data, Walker et al. (2009) further describe their characteristics, and Le Hénaff et al. (2014) show that the LCFEs are in solid-body rotation close to their core and confirms that the LCFEs have diameters of 80–120 km. The glider observations of Rudnick et al. (2015) suggest LCFEs are more barotropic than the LCEs. They attribute this behavior to the origin of the LCFEs from the less stratified Gulf Common Water, on the dense side of the LC.

Observations suggest that LCFEs exist over a broad spectrum of temporal variability. Using along-track altimetry, Zavala-Hidalgo et al. (2003a) described the formation and development of cyclonic eddies on the southwestern side of the LC, northeast of the Campeche Bank. They noticed that the cyclones can remain between the CB and 26°N for periods between 1.3 to 9.6 months. They observed eight cyclonic eddies in the region in a period of 7 yr. Analyzing 5 yr of altimetry and in situ data, Athie et al. (2012) show low-frequency cyclonic anomalies (of the order of two to three cyclonic eddies each year) crossing the Yucatan Channel from the Caribbean Sea. Recently, Donohue et al. (2016) used an array of moorings and inverted echo sounders with pressure gauges to show there is a wide spectrum of spatial and temporal variability along the LC front. They find the strongest variability in the 40–100-day low-frequency band (460-km wavelengths) concentrated east of the Mississippi Fan. They also documented high-frequency variability (3–10 days, 240-km wave length) concentrated over the western side of the array (but it should be noted that their array does not cover the eastern slope of the Campeche Bank). Finally, using 3 yr of direct current measurements along the eastern Campeche Bank [including the Plataforma Norte (PN) moorings discussed in this paper], Sheinbaum et al. (2016) also found a large variability of the along-shelf currents at low frequencies (40–100 days or longer periods) but with a substantial contribution at higher frequencies (5–25-day periods), especially during October–March. These observational evidences of high-frequency variability along the LC call for a better understanding of their nature and origin.

It has been surmised that the interaction of LCFEs with the LC play a key role in the shedding of LCEs (Vukovich and Maul 1985). Fratantoni et al. (1998) and Zavala-Hidalgo et al. (2003a) associated the LC “pinching-off” process (Schmitz 2005) with LCFEs cutting, either traversing eastward or westward, the LC. Athie et al. (2012) observed that 16 (76%) of the 21 detachments that occurred during the period 2005–09 are related to the cyclonic perturbations coming from the Caribbean Sea. However, the recent analysis of along-track altimetry and drifters by Le Hénaff et al. (2014) show that LCFEs on the eastern CB are actually preferably found immediately after a LCE detachment or separation. These authors suggest that the cyclonic eddies observed near the CB, which were thought to participate in the LCE detachment process, are only involved in the shedding process after a reattachment of the LCE. The previous studies dedicated to the influence of the LCFEs on the LC detachment did not formerly distinguish between low- and high-frequency eddies, but the relationship between the cyclones and the LC detachments was mostly established by looking at low-frequency LCFEs (e.g., Zavala-Hidalgo et al. 2003a; Athie et al. 2012). Whether or not high-frequency eddies contribute to the LCE separation remains an open issue.

Several mechanisms for the origin of the low-frequency LCFEs have been put forward. Chérubin et al. (2006) provided some evidence in favor of a mixed barotropic–baroclinic instability process as the cause for the formation of LCFEs. They also pointed out the key role of the cyclonic vorticity belt around the LC in the instability process. Oey (2008) identified the northern CB and northeastern Florida slope as enhanced instability regions where cyclones are generated or intensified. While Le Hénaff et al. (2012) pointed out the formation of LCFEs on the western side of the LC along the CB during a LC eddy detachment, they did not provide any explanation on the mechanisms involved in the formation of such LCFEs. The CB, however, does exhibit a large promontory where the shelf cuts westward and intensification of LCFEs near a promontorylike topographic feature were found by Le Hénaff et al. (2012) in the northern Gulf shelf slope southeast of the Mississippi delta. No mechanisms have yet been put forward for the origin of the high-frequency perturbations observed along the eastern Campeche Bank, in large part because of the limitations of the mooring array (Sheinbaum et al. 2016). Remote mechanisms could also play a significant role in the formation of LCFEs. In situ and satellite observations suggest indeed that the cyclonic eddies observed between the CB and the LC may not necessarily originate locally but could result from the northward advection of cyclones or perturbations generated upstream of the Yucatan Channel, in the western Caribbean region, or farther upstream (Candela et al. 2002; Athie et al. 2012).

In summary, numerous questions remain regarding the characteristics, origin, and fate of the LCFEs, in particular the high-frequency eddies near the CB. The main goal of this study is to further describe the high-frequency LCFEs occurring near the CB and to shed light on the environmental factors controlling their formation and characteristics. Our main focus will not be on the large and low-frequency cyclonic anomalies crossing the Yucatan Channel (Athie et al. 2012) or those discussed by Zavala-Hidalgo et al. (2003a), but on the high-frequency (period < 30 days) and small-scale (<100 km) frontal eddies developing around the LC. Numerical simulations will be used to complement observations. While in situ mooring measurements provide an observational truth of the LCFEs characteristics at mooring locations, they are too limited in space to investigate their origin and fate at regional scale. Altimetry does provide a synoptic view of the circulation, but it does not allow a proper investigation of the formation and fate of all LCFEs that may occur along the CB since altimetry tends to filter out the high-frequency fraction of the LCFEs. The paper is organized as follows: Observational evidences of high-frequency LCFEs traveling along the CB are presented and discussed in section 2. Section 3 describes the spatial and temporal characteristics of the LCFEs as resolved in a high-resolution numerical experiment. Section 4 investigates some of the mechanisms controlling their formation. Section 5 provides a discussion and summary of the results.

2. Observations of the LCFEs at the Campeche Bank

High-resolution observations of Sea surface temperature (SST) from the multiscale ultrahigh resolution (MUR) sea surface temperature product (merged product using data from MODIS, AMSR-E, and AVHRR) on 4 February 2011 and 6 February 2011 are shown in Figs. 1a and 1b. At these times, the temperature of the waters advected by the LC from the Caribbean Sea (>26°C) contrasts with the temperature of the surrounding Gulf common waters (<24°C). This contrast allows us to infer the formation of a cyclonic eddy along the western edge of the LC (Figs. 1a,b), with horizontal scale of about 100 km. Several similar examples of these can be found in Walker et al. (2009). There is no clear signature of this eddy in the altimetry data, as shown by the surface geostrophic currents derived from it on 6 February 2011 (Fig. 1c). This is due to the low spatial resolution (¼°) and low temporal resolution (a 7-day time-averaging procedure is used to compute the altimetry gridded product) compared to the characteristics of these eddies as is discussed below.

Fig. 1.
Fig. 1.

Sea surface temperature (°C) over the LC on (a) 4 Feb 2011 and (b) 6 Feb 2011. It illustrates the presence of a cyclonic eddy with a horizontal scale of about 100 km forming on the western front of the LC at the location of the mooring array. Temperatures are from the MUR SST product. (c) Sea level height (m) and corresponding surface geostrophic currents (m s−1) from AVISO on 6 Feb 2011. Contours represent the 100-, 500-, 1000-, and 2000-m isobaths. The PN moorings locations are indicated with black dots, and vectors represent the daily mean current average between 50 and 100 m, obtained from ADCP measurements.

Citation: Journal of Physical Oceanography 46, 11; 10.1175/JPO-D-16-0052.1

An array of four moorings deployed along a section north of the CB (hereinafter referred to as the PN section; see Fig. 1) measured the full water column from April 2010 to May 2011. These moorings were mounted with upward-looking acoustic Doppler current profilers (ADCPs; positions shown in Table 1)1 that allow us to infer the vertical characteristics of the LCFEs identified in the SST data. The velocities measured by the ADCPs at the PN mooring array are averaged in the following ranges of depth: 50–100, 200–300, and 400–500 m. The velocity component perpendicular to the PN section, for the different ranges of depth, and the sea level height from AVISO are shown in Fig. 2 as a function of time from January to April 2011. During this time, as is the common situation, the LC flowed along the CB, with near-surface maximum velocity exceeding 1.2 m s−1 and centered at 24.5°N (Fig. 2a). We remark that the along-section fluctuations of the maximum northwestward velocity varies less at depth (Fig. 2c) compared to the near-surface (Fig. 2a) or intermediate depths (Fig. 2b). The large velocity variations with depth (vertical shear) illustrate the baroclinic character of the Loop Current in this region.

Table 1.

Position and characteristics of the ADCPs deployed north of the CB.

Table 1.
Fig. 2.
Fig. 2.

Cross-section horizontal velocity (m s−1) obtained from rotated ADCP observations at the PN array (see Fig. 1), between January and April 2011. Positive values are associated with currents oriented northwestward. The currents are vertically averaged between different depths: (a) 50–100 m, (b) 200–300 m, and (c) 400–500 m. Sea level height (m) observations from AVISO have been interpolated along the PN section and are represented with labeled contours. The bold points at 1 May indicate the latitude of the PN moorings. The vertical dashed lines indicate 4 and 6 Feb, when a frontal eddy is crossing the mooring array (see Fig. 1).

Citation: Journal of Physical Oceanography 46, 11; 10.1175/JPO-D-16-0052.1

The cross-section currents present high-frequency variations at all depths all along the period of observation. Northwestward acceleration at the core of the LC is generally associated with southeastward acceleration at the shallowest mooring of the section (at 23.8°N in Figs. 2a,b; at 24.2°N in Fig. 2c). This suggests that these anomalies are cyclonic eddies traveling along the western edge of the LC. This is in line with SST observations of the cyclonic pattern crossing the array on 4 February (Fig. 1a), which coincides with increased northward flow at the core of the LC and a southward flow on the westernmost mooring (see vectors in Fig. 1a and the 50–100-m currents at day 4 February in Fig. 2a). In the following section, we will show that this behavior is consistent with the numerical simulations. An objective mapping of the cross-section currents from 3 to 6 February (Fig. 3) further illustrates the characteristics of the cyclonic eddy crossing the PN array in February. The passage of the cyclone through the section on 4 February results in an intensification of the LC (Fig. 3b) and a southwestward flow at moorings PN1 and in the deep part of PN2. This led to increased positive vorticity at mooring PN2 (Fig. 3f). Power spectra of the cross-section velocity component at the PN2 mooring show that the perturbations have a peak frequency in the 8–10-day period band at all depths (Fig. 4a). The position of the LC core as inferred from altimetry SSH is in good agreement with the near-surface ADCP observations (black contours in Fig. 2a), but there is no signature of short-period fluctuations in the altimetry.

Fig. 3.
Fig. 3.

Objective analysis of (left) cross-section velocity (m s−1) and (right) relative vorticity (s−1) through the PN array for (from top to bottom) 3, 4, 5, and 6 Feb. Mooring velocities have been low-pass filtered with a cutoff period of 48 h and interpolated using objective mapping (Bretherton et al. 1976) with horizontal decorrelation scale Ls = 20 km and vertical decorrelation scale Lz = 150 m.

Citation: Journal of Physical Oceanography 46, 11; 10.1175/JPO-D-16-0052.1

Fig. 4.
Fig. 4.

Multitaper variance-conserving power spectra (m−2 s−2) of cross-section velocity at mooring PN2 for three different depth ranges (as in Fig. 2) from (a) observations and (b) model. The spectra in (a) are built with in situ data from January to May 2011 and five windows and in (b) using model data from January to May 2008 and five windows. Dotted lines represent the corresponding 90% confidence limit based on the theoretical spectra of an AR(2) process with variances equal to that of the analyzed signals. The number of degrees of freedom for the 90% level calculation is estimated as 2K − 1, with K as the number of windows.

Citation: Journal of Physical Oceanography 46, 11; 10.1175/JPO-D-16-0052.1

Other in situ measurements are available on the eastern Campeche Bank [e.g., the Plataforma Este (PE) cross-shelf array near 22°–23°N already described in Athie et al. (2012)] but we chose to discuss here only the variability at the PN array in which the frontal eddies are more developed. The similarities and differences between the variability at the PN and PE arrays are thoroughly discussed in Sheinbaum et al. (2016).

3. LCFEs in a high-resolution model of the Gulf of Mexico

a. Numerical setup

A regional model of the Gulf of Mexico has been developed for this study. The numerical code is the oceanic component of the Nucleus for European Modeling of the Ocean program (NEMO3.6; Madec 2016). It solves the three-dimensional primitive equations in curvilinear coordinates discretized on a C grid and fixed vertical levels (z coordinate). The model configuration consists of a grid with 1/36° horizontal resolution (Δx, Δy ~2.8 km) encompassing the Gulf of Mexico and the Cayman Sea (from 14° to 31°N and from 98° to 78°W; see model domain in Fig. 5). There are 75 levels in the vertical (with 12 levels in the upper 20 m and 24 levels in the upper 100 m). Temperature and salinity are advected using a total variance dissipation scheme (TVD) with nearly horizontal diffusion parameterized as a Laplacian isopycnal diffusion, with a coefficient of 45 m2 s−1. Horizontal diffusion of momentum is implicit since a third-order upwind advection scheme (UP3) is employed. The vertical diffusion coefficients are given by a generic length scale (GLS) scheme with a k–ε turbulent closure (Reffray et al. 2015). Bottom friction is quadratic with a bottom drag coefficient of 10−3, and partial-slip boundary conditions are applied at the lateral boundaries. The temporal integration is achieved by a modified leapfrog Asselin filter (Leclair and Madec 2009), with a coefficient of 0.1 and a time step of 150 s. The free surface is solved using a time-splitting technique with the barotropic part of the dynamical equations integrated explicitly with a time step of 2.5 s.

Fig. 5.
Fig. 5.

Model domain and mean sea level height (m) simulated by the model. Contours represent 30-, 500-, and 3000-m isobaths. The red points mark distances of 500 km along the 30-m isobath in order to help the interpretation of the Hovmöller diagram in Fig. 19. The dashed black line along the CB indicates the path used to build the Hovmöller diagram in Fig. 21d.

Citation: Journal of Physical Oceanography 46, 11; 10.1175/JPO-D-16-0052.1

The model is forced at its lateral boundaries with daily outputs from the MERCATOR global ocean reanalysis GLORYS2V3. The open boundary conditions radiate perturbations out of the domain and relax the model variables to 1-day averages of the reanalysis data. Details of the method are given in Madec (2016). At the surface, the atmospheric fluxes of momentum, heat, and freshwater are computed by bulk formulae (Large and Yeager 2004). The model is forced with the Drakkar Forcing Sets version 5.2 product (DFS5.2) (Dussin et al. 2014), which is based on ERA-Interim reanalysis and consists of 3-h fields of wind, atmospheric temperature, and humidity and daily fields of longwave and shortwave radiation and precipitation. DFS5.2 is an update of the product described in Brodeau et al. (2010). The shortwave radiation forcing is modulated online by a theoretical diurnal cycle. A monthly climatological runoff based on the dataset of Dai and Trenberth (2002) is prescribed near the river mouths as a surface freshwater flux with increased vertical mixing in the upper 10 m. The model is run from 2000 to 2012, and daily averages from 2003 to 2012 are used in the present study, which is long enough for our focus on high-frequency and seasonal variability.

In addition to this reference experiment (REF), an experiment forced with monthly wind stress has been carried out. In this second experiment, which will be referred as MONTHLY, the surface wind stress is computed using monthly averages of the DFS5.2 winds used in the REF experiment. To obtain a minimal alteration of the oceanic background state, the air–sea heat and freshwater fluxes (specifically the latent and sensible heat fluxes and the evaporation) in MONTHLY are computed using the high-frequency DFS5.2 winds used in REF as well.

b. Validation of the background circulation

The observed mean surface currents in the Gulf of Mexico inferred from altimetry and Ekman drift [geostrophic and Ekman current observatory (GECKO) product; Sudre et al. 2013] show an intensification of the Yucatan Current along the CB (Fig. 6a) that is well represented in the model (Fig. 6c). Overall, the surface mean extension of the LC and the surface eddy activity within the LC [measured as eddy kinetic energy (EKE) = 1/2(u*2 + υ*2), where velocity anomalies u* and υ* are defined with respect to 5-yr-averaged velocities] are also well reproduced by the model (Fig. 6). We note in both model and observations two distinct maxima of EKE within the LC, one north of the CB and another east of the west Florida shelf.

Fig. 6.
Fig. 6.

(a),(c) Mean surface currents (m s−1) and (b),(d) surface eddy kinetic energy (m2 s−2) from (left) observations and (right) the model for the period 2003–12. Observed surface currents are from the GEKCO database (Sudre et al. 2013) and consist of the sum of altimetry-derived geostrophic currents and estimated Ekman currents.

Citation: Journal of Physical Oceanography 46, 11; 10.1175/JPO-D-16-0052.1

c. Characteristics of the simulated LCFEs

1) High-frequency LCFEs

Time series of model currents across the PN section are shown in Fig. 7 for the period January to April 2008. From February to April, the LC in the model flows along the shelf break of the CB and passed though the PN section as observed in 2011. The cross-section velocities at the different depths (Fig. 7) are in good agreement with observations (Fig. 2). As in the observations, the cross-section velocities (i.e., velocity component normal to the section) at the core of the LC show short-period fluctuations in opposite phase with current fluctuations on the southwestern side of the section (i.e., at the mooring PN1). During this 4-month period, the cross-section velocity spectrum at 24.2°N peaks in the 6–10-day period band (Fig. 4b) as in observations. Snapshots of SST, vorticity, and finite-time Lyapunov exponents (FTLEs)2 at 100 m are shown at 4-day intervals from 6 to 14 March in Fig. 8. They show that these short-period velocity anomalies correspond to cyclonic structures traveling downstream along the outer western edge of the LC. For example, the peak velocity on 6 March near 24.4°N in Fig. 7a, coincides with a depression of the sea level (Fig. 8a), a perturbation of the SST field (Fig. 8d), a local maximum of positive vorticity (Fig. 8g), and an eddylike coherent structure at 100 m (Fig. 8j). On 14 March, the cyclone located at the PN section (24°N) moved to near 26°N and merged with a larger cyclone (see Figs. 8j,k,l), while a new cyclone occupies the PN section (in agreement with a maximum of northward velocity at the PN section; Fig. 7). These snapshots illustrate that LCFEs propagate relatively fast along the LC and that their period of occurrence is short.

Fig. 7.
Fig. 7.

Cross-section horizontal velocity (m s−1) of numerical model along the PN mooring section from January to April 2008. The currents are vertically averaged between different depths: (a) 50–100, (b) 200–300, and (c) 400–500 m. The model sea level height (m) is shown in contours. Dashed lines indicate 6, 10, and 14 Mar 2008 for comparison with model snapshots shown in Fig. 8.

Citation: Journal of Physical Oceanography 46, 11; 10.1175/JPO-D-16-0052.1

Fig. 8.
Fig. 8.

(a)–(c) Model sea level (m), (d)–(f) surface temperature (°C), (g)–(i) relative vorticity at 100 m (s−1), and (j)–(l) FTLE at 100 m (day−1) at 6, 10, and 14 Mar 2008. The black line represents the position of the PN section as used in Fig. 7. The black contours in (a),(b), and (c) show the eddies identified by the tracking method.

Citation: Journal of Physical Oceanography 46, 11; 10.1175/JPO-D-16-0052.1

Variance-conserving power spectra of the relative vorticity at 100 m are averaged over four areas along the shelf break of the CB and Yucatan coast (Fig. 9a). The variance of high-pass filtered (30-day cutoff period) surface vorticity is high along the CB (Fig. 9a). At the three northernmost locations (A, B, and C), the vorticity spectra show peaks in the 8–10-day band (Fig. 9b). This is coherent with spectra of cross-section velocity at the PN2 mooring computed from observations (Fig. 4a) or model (Fig. 4b). These locations contrast with the Yucatan coast (location D), where the variance of vorticity is lower and there are no marked spectral peaks in the 8–10-day band (Fig. 9b). This suggests that the intense high-frequency variability (below 15 days) observed along the CB, can be generated from the entrance of the Yucatan Channel between 21° and 22°N.

Fig. 9.
Fig. 9.

(a)Variance of the relative vorticity (s−2) at 100 m. Multitaper variance-conserving power spectra of surface vorticity (s−2), computed at each model grid point and averaged over four different zones along the continental shelf of Yucatan Coast and Campeche Bank for experiments (b) REF and (c) MONTHLY. The four zones are shown in (a). Model data from January 2003 to December 2012 and 29 windows are used for this spectral analysis.

Citation: Journal of Physical Oceanography 46, 11; 10.1175/JPO-D-16-0052.1

This is confirmed by a time–latitude diagram of the sea surface height (SSH) along the LC from 18° to 25°N (Fig. 10a). The high- and low-frequency contributions to the SSH variability are isolated using high- and low-pass Lanczos filters with a cutoff period of 30 days. Most of the short-period anomalies form north of 21°N (Fig. 10b), and there is no indication that they are advected from the Caribbean Sea. Instead, we can distinguish many propagating features propagating southward from the Yucatan Channel at 21° to 18°N, indicative of coastal-trapped waves or sea level response to southward-propagating wind events. The eddy detachment times (shown with dashed lines in Fig. 10b) do not reveal any relationship between the detachments and the presence of short-period cyclonic anomalies. This suggests they do not have a first-order influence on the eddy separation process.

Fig. 10.
Fig. 10.

Time–latitude diagram of (a) model SSH along the mean Loop Current position (identified with the 0.1 m isocontour of mean SSH as indicated with a dashed line in Fig. 5), (b) high-pass filtered model SSH anomalies, and (c) low-pass filtered SSH anomalies. A cutoff period of 30 days is used to separate the high-frequency (b) and low-frequency (c) contributions to the SSH signal. Vertical dashed lines indicate the time of LCE detachments (identified by visual inspection of the model outputs).

Citation: Journal of Physical Oceanography 46, 11; 10.1175/JPO-D-16-0052.1

2) Low-frequency LCFEs

The low-frequency part of the SSH signal (periods > 30 days) along the western edge of the LC is shown in Fig. 10c. It illustrates the propagation of negative SSH anomalies rising at 19°N along the Yucatan coast, that is, south of the Yucatan Channel. We remark that some cyclonic anomalies north of the Yucatan Channel do not show any straightforward link with the circulation upstream, suggesting they are formed in this region or are advected from the eastern bound of the LC. In contrast to the short-period anomalies discussed above (Fig. 10b), a large fraction of the low-frequency cyclonic anomalies north of the Yucatan Channel is related to the Loop Current detachment events, as suggested by the concomitancy of the separation time (dashed lines) with negative SSH anomalies in Fig. 10c and confirmed by visual inspection of sea level maps (not shown). This is in agreement with the results of Athie et al. (2012) suggesting that Caribbean cyclones could contribute significantly to the Loop Current eddy-shedding process. This also suggests that the cyclones observed in Athie et al. (2012) or Zavala-Hidalgo et al. (2003a) are more tied to low-frequency LCFEs (periods > 30 days) rather than short-period LCFEs.

The link between low-frequency negative SSH anomalies near Campeche and Caribbean eddies is confirmed with a lagged regression analysis (Fig. 11, top) that indicates some correlation between negative SSH anomalies at the northeastern shelf of the CB (lag 0 in Fig. 11) and cyclonic eddies interacting 1.5 months before with the Yucatan coast (near 20°N, lag −45 days in Fig. 10). The regression analysis also indicates that some cyclones observed along the CB may also be related to large cyclones coming from the west Florida shelf that recirculate to the west. A regression analysis performed with 10 yr of altimetry data show very similar patterns (Fig. 11b, bottom), suggesting that the model represents quite well these low-frequency dynamics.

Fig. 11.
Fig. 11.

Linear lagged regressions of model sea level height (m) on a North Campeche LCFE index for (top) model and (bottom) altimetry for the period 2003 to 2012. Lags range from −45 to 0 days. To compute the index, the SSH is spatially averaged in the area from 23.8° to 24.2°N and from 88.4° to 87.2°W, that is, north CB in the area of the PN moorings array. The index is filtered with a Lanczos filter to retain the variability between 30 and 90 days and weighted by the standard deviation. The anomalous patterns are linearly correlated with one standard deviation anomaly of the SSH north of CB and units of the regression coefficients are meters. Data were high-pass filtered (90-day cutoff period) before the computation of the regression coefficients. Regression coefficients less than the 99% confidence level are set to 0. Contours represent the 100- and 500-m isobaths.

Citation: Journal of Physical Oceanography 46, 11; 10.1175/JPO-D-16-0052.1

3) EKE distribution

Taken as a rough estimate of mesoscale activity, the EKE is computed from surface velocity anomalies filtered in different frequency bands (Figs. 12a–c). Within this range of variability, the distribution of energy depends on the frequency band considered. At periods between 0 and 15 days, the highest EKE occurs along and north of the CB (Fig. 12a), while at periods between 15 and 30 days, the EKE is more homogeneous along the LC (Fig. 12b) with a second maximum along the Florida slope. At periods above 30 days, the maximum EKE occurs along the west Florida shelf (Fig. 12c). Such distribution of EKE between the different frequency bands is in agreement with growth and merging of the LCFEs during their travel around the LC. This is confirmed in the following section.

Fig. 12.
Fig. 12.

Eddy kinetic energy (m2 s−2) computed from surface currents filtered in different bands of periods: 0–15, 15–30, and 30–90 days, in experiments (a)–(c) REF and (d)–(f) MONTHLY. Data from January 2003 to December 2012 are used. Contours represent the 100-, 500-, and 1000-m isobaths.

Citation: Journal of Physical Oceanography 46, 11; 10.1175/JPO-D-16-0052.1

d. Eddy tracking

Eddy identification and tracking is now performed in order to complete our understanding of the fate of the LCFEs. The identification of eddies is based on the automated detection algorithm developed by Nencioli et al. (2010). The eddy centers are identified using geometrical constraints on the velocity field (a velocity minimum, tangential velocities increasing linearly proportional to the distance from the center and decreasing after reaching a maximum value, and reversal of the sign of the velocity components cross the center of the eddy). Eddy dimensions are computed from the integration of the streamfunction and searching for the largest closed contours around the eddy centers. The eddy tracks are derived by comparing the detected eddy fields at successive time steps within an area defined by a radius of 60 km. The method has been applied to daily currents averaged between the surface and 100-m depth, and we restrict the analysis to eddies lasting more than 5 days. Although the method provides a lower end estimate of the radius of the eddies [see Fig. 8 here or Fig. 6 in Nencioli et al. (2010)], it guarantees to return the dimension of vortical (closed) structures.

A large amount of trajectories of cyclonic eddies follow the main path of the LC (Fig. 13a) and can be associated with LCFEs. Most of these eddies form on the eastern flank of the CB (Fig. 13b), while the location where they dissipate is sparser (Fig. 13c), suggesting there is not a preferential area where dissipation of the eddies takes place. The typical diameter of the cyclonic eddies that have been identified is between 20 and 60 km (Fig. 13e). The diameter of the LCFEs increases along the main LC path (Figs. 13f,g) from ~40 km along the CB to ~80 km near the west Florida shelf. Merging events have been identified in the entire LC area (cross in Fig. 13f), suggesting that merging may contribute to this growth. Growth and merging of the LCFEs during their travel around the LC has already been observed by Schmitz (2005) and Le Hénaff et al. (2012). This analysis points out that a large faction of the LCFEs originates near the CB.

Fig. 13.
Fig. 13.

Tracks and statistics of eddies identified during the period 2003–12 and lasting more than 5 days: (a) trajectories of the eddies (cyclones in blue, anticyclones in red), (b) origin position of the cyclones, (c) final position of the cyclones, (d) eddy lifetime (days) of the cyclones, and (e) diameter of the cyclones (km). (f) The LC path in has been identified as a mean sea level isocontour of 0.1 m and (g) is used to sort the diameter of the cyclones along the path of the LC . The color codes in (f) and (g) correspond. The crosses in (f) indicate locations where merging has been identified (in the sense that an eddy is identified within another one).

Citation: Journal of Physical Oceanography 46, 11; 10.1175/JPO-D-16-0052.1

4. Formation of the high-frequency LCFEs along the Campeche Bank

In this section, we investigate the energy sources for the high-frequency LCFEs developing near the CB, and the reason why their signature is visible, from an Eulerian perspective, in a narrow band of periods.

a. Instability of the main flow

Oey (2008) has already shown that intense baroclinic conversion occurs below 250-m depth along and north of the CB. The author did not examine the upper-ocean dynamics, but it is expected that surface-intensified eddies along the CB are also produced by instability of the main flow. To verify this hypothesis, we compute the barotropic (BT) and baroclinic (BC) energy conversion terms following Masina et al. (1999):
eq1
where u′ and υ′ are the horizontal components of the perturbation velocity, ULP and VLP are the horizontal components of the low-pass velocity, w′ is the vertical velocity perturbation, ρ′ is the density perturbation, and ρ0 is the mean density. The perturbations are constructed with respect to their low-frequency time series that were built using a 30-day running mean; the low-frequency contribution and the perturbation add up to the full series. So for each daily model output, the low-frequency variables are representative of the time-mean flow over a 30-day period centered at the date of the model output (so the low-pass flow also includes large-scale and low-frequency eddies). This methodology allows us to identify the energy sources for the eddy variability occurring at periods of less than 30 days. The mean baroclinic and barotropic conversion terms averaged from 0 to 100 m are both large along and north of the CB (Figs. 14a,b). The barotropic conversion dominates in this part of the water column. This is in agreement with Chérubin et al. (2006) who found that over the LC (23°–29°N, 91°–83°W) barotropic instability (that arises from lateral velocity shear) is intensified at the surface, while baroclinic instability (that arises from vertical velocity shear) is intensified at depth. Finally, these results suggest that the release of energy from the large-scale flow to the turbulent flow contributes to the formation and growth of the LCFEs in the region. The strong vertical and horizontal shear along the shelf break (Figs. 12c,d) is a key ingredient for the development of these mixed barotropic–baroclinic instabilities.
Fig. 14.
Fig. 14.

Maps of baroclinic conversion of (a),(e) EKE (kg m−1 s−3), (b),(f) barotropic conversion of EKE (kg m−1 s−3), (c),(g) vertical velocity shear (s−2), and (d),(h) horizontal velocity shear (s−2) for (top) simulation REF and (bottom) simulation MONTHLY. In (a),(b),(e), and (f), positive means a transfer of energy toward the turbulent field. Mean velocity shear is computed as the time-mean velocity shear computed from the 30-day low-pass filtered current ULP and VLP. Energy transfer terms and velocity shears are averaged between the surface and 100-m depth and are computed using daily model outputs from 2003 to 2012.

Citation: Journal of Physical Oceanography 46, 11; 10.1175/JPO-D-16-0052.1

b. Impact of the high-frequency wind variability

The concentration of the EKE in the region of the CB in a narrow band of periods (6–10 day) raises the question of which environmental factors control the period of formation of the eddies. This narrow band is also characteristic of the wind variability in the region, and the following results suggest the synoptic-scale wind variability as a key factor.

The power spectra of the zonal and meridional components of the wind stress, computed from wind stress averaged in the northwestern part of the Gulf of Mexico (24°–30°N, 98°–90°W), show one main peak of energy at periods between 6 and 10 days (Fig. 15a), which correspond to peaks already noted in the vorticity power spectra (Fig. 9b). This suggests a link between the wind stress variability and the formation of the LCFEs. To verify this hypothesis, we carried out a simulation referred to as MONTHLY in which the model is forced with monthly average ERA-Interim wind stress, instead of the 3-h and daily fields (see section 3a for further details), neglecting the input of energy by the wind at periods of less than 60 days. We verified that the MONTHLY results in terms of LCFE characteristics are not sensitive to a choice of 1-day or monthly averages for the lateral boundary conditions (not shown). In this simulation, the dynamics near the CB is significantly different compared to REF. In MONTHLY, the power spectra of the surface vorticity do not show marked peaks at periods between 6 and 10 days (Fig. 9c). In terms of EKE, the main differences are a reduction of the energy in the MONTHLY experiment at periods of less than 15 days all along the CB and farther downstream (Figs. 12a,d) and an increase of the EKE at larger periods (Figs. 12e,f). These differences at periods larger than 15 days may be due to the damping effect of the wind work on the eddies, which mainly arise when the surface currents are taken into account in the computation of the wind stress (e.g., Renault et al. 2016; Seo et al. 2016). We expect this effect to be weaker when the model is forced with monthly winds (MONTHLY) instead of 3-h winds (REF), but this would deserve further investigation. The eddy tracking analysis for the MONTHLY experiment indicates a 20% decrease of the number of eddies (124 eddies in 5 yr in REF vs 102 eddies in 10 yr in MONTHLY) formed north of the CB (in the area between 23° and 24.5°N and 89° and 88°W). In conclusion, the comparison between both simulations suggests that 1) the absence of synoptic winds does not eliminate the formation of LCFEs, but 2) the synoptic winds contribute to the emergence of more frequent eddies near the CB. In other words, the high-frequency winds tend to reinforce the instability process along and north of the CB (as confirmed by the larger baroclinic and barotropic energy conversion terms in REF compared to MONTHLY; Figs. 14e,f) but are not necessary for LCFEs to develop.

Fig. 15.
Fig. 15.

(a) Multitaper variance-conserving power spectra of the zonal and meridional components of the wind stress averaged over the northwestern corner of the Gulf of Mexico. Data were averaged between 24° and 30°N and 98° and 90°W before spectral analysis. The spectra are built with model data from January 2003 to December 2012 and 29 windows. (b) Wavelet of the wind stress averaged over the same region as in (a). The power spectra is normalized by σ2 the variance of the corresponding signal. The thick contour encloses regions of greater than 90% confidence for an AR(2) process. Vertical dashed lines indicate 1 Jan of each year. (c) Longitudinal position of the LC front at 24°N, identified by the position of the maximum gradient of SSH between 89° and 86°W.

Citation: Journal of Physical Oceanography 46, 11; 10.1175/JPO-D-16-0052.1

The high-frequency variability of the wind stress has its maximum during winter (Fig. 15b). So, it is expected that the influence of the high-frequency winds on the eddy variability will be maximum during this part of the year. The contribution of the wind forcing to this seasonal variability is verified through REF/MONTHLY comparison of the surface EKE computed from velocities high-pass filtered at periods below 15 days and averaged for winter (November–March) and summer (May–August). The main differences between REF and MONTHLY experiments are observed during winter (Figs. 16a,b), with increased EKE along and north of the CB in REF. During summer, the difference of EKE between both simulations is weaker.

Fig. 16.
Fig. 16.

(a),(b) Winter (November–March) and (c),(d) summer (May–August) mean eddy kinetic energy (m2 s−2) in experiments (a),(c) REF and (b),(d) MONTHLY computed from surface currents high-pass filtered (15-day cutoff period). Only periods with the LC front at 24°N located between 88° and 87°W (see Fig. 15c) are retained in the average.

Citation: Journal of Physical Oceanography 46, 11; 10.1175/JPO-D-16-0052.1

As observed in Figs. 2 and 7, the position of the LC front relative to the eastern CB slope has a strong effect on the presence of high-frequency LCFEs. When the LC separates from the slope, the generation of high-frequency LCFEs is cut off. These separations last approximately one month and generally occur when the LC sheds a ring. The winter and summer composites in Fig. 16 are constructed by removing the periods with the LC front far from its mean position (these periods are identified from Fig. 15c, when the LC front at 24°N is east of or west of 87°W) so that the more frequent production of LCFEs during winter cannot be explained by seasonal changes of the LC position.

A linear regression analysis was performed in order to get further insight into the dynamical link between wind stress and the formation of the LCFEs. Regression coefficients at different lags were computed between a wind stress index, representative of the synoptic perturbations passing over the Gulf of Mexico, and model surface fields (sea surface height). The index is defined in such a way that it isolates atmospheric perturbations with periods shorter than 15 days for the entire period of 2003–12. It is computed as the time-filtered and spatially averaged southward wind stress (the wavelet analysis of this signal is shown in Fig. 15b) between 24° and 30°N and 98° and 90°W. The high-pass time filtering is performed using a Lanczos filter and the index is weighted by its standard deviation. Regression coefficients between the index and model fields are computed at each model grid point for different lags ranging between 0 and 11 days (Fig. 17). The model fields are high-pass filtered with a cutoff period of 30 days before computation of the regression coefficients.

Fig. 17.
Fig. 17.

Linear lagged regressions of model sea level height (m) and wind stress (N m−2) on a cold surge wind index. Lags range from 0 to 11 days. To compute the index, the southward meridional wind stress is spatially averaged in the northwestern Gulf of Mexico (from 24° to 30°N and from 98° to 90°W), filtered with a high-pass Lanczos filter to retain the variability under 15 days and weighted by the standard deviation. Thus the anomalous patterns are linearly correlated with one standard deviation anomaly of the meridional wind stress in the northwestern Gulf of Mexico (about −0.05 N m−2) and units of the regression coefficients are N m−2 for the wind stress and meters for the SSH. Data were high-pass filtered (30-day cutoff period) before the computation of the regression coefficients. Regression coefficients less than the 90% confidence level are set to 0. Red (gray) vectors indicate the northward (southward) anomaly of the regressed meridional component of the wind stress. Contours represent the 100- and 500-m isobaths. See text for explanation of letter labels.

Citation: Journal of Physical Oceanography 46, 11; 10.1175/JPO-D-16-0052.1

By construction, the wind stress field regressed with lag 0 reproduces almost entirely its southward component within the Gulf (Fig. 17). The wind stress pattern in correlation with the index shows southeastward winds toward the Caribbean Sea at lag-1 day and northward wind anomalies in the western Gulf of Mexico by lags 2 to 4. The regressed wind stress field is not significant at subsequent lags. There are several indications that such wind variability captured by the regression analysis corresponds to Central American cold surges (nortes; e.g., Schultz et al. 1998): 1) the events last a few days, 2) meridional wind stress anomalies are dominant (see also power spectra in Fig. 15a), 3) the wind stress anomalies travel from the northwestern to the southeastern Gulf of Mexico with a marked velocity (and temperature) front, and 4) this high-frequency variability occurs predominantly during winter (Fig. 15b).

The regression of the SSH field on the wind index suggests that anomalies of SSH are directly forced between lags 0 and 3 by the intense wind event captured by the regression analysis (Fig. 17). Looking downwind, coastal winds blowing with the land on the right-/left-hand side is favorable for coastal downwelling/upwelling and sea level rising/lowering. This is verified since the along-coast SSH anomalies at lag 0 are consistent with offshore or onshore Ekman drift in response to wind stress anomalies. This close relationship between coastal SSH and local wind stress is also verified by the computation of cross-correlation coefficients between model fields extracted at Galveston, Port Isabel, and Progreso tide gauges stations (the geographical positions of the model extraction match the position of the tide gauges given in Table 2). At these locations, the cross correlation between SSH and either the zonal or the meridional component of the wind stress (depending on the orientation of the coast) is maximum at lag 0 and overpasses 0.5 (Figs. 18b,c).

Table 2.

Tide gauge record locations.

Table 2.
Fig. 18.
Fig. 18.

Lagged cross-correlations based on model fields between (a) SSH at different locations, (b) zonal wind stress and SSH, (c) meridional wind stress and SSH, and (d) SSH and surface relative vorticity at the location of the PN array. Time series were high-pass filtered (cutoff period of 30 days) and normalized before computation of the correlation coefficients. The positions of the time series are indicated with labels in the different figures and correspond to the positions of the tide gauges indicated in Table 2.

Citation: Journal of Physical Oceanography 46, 11; 10.1175/JPO-D-16-0052.1

The regressed SSH field shows at lags 6 to 11 a succession of small-scale positive and negative anomalies north of the CB (Fig. 17), which are the signature of LCFEs. The regression analysis does not capture LCFEs on the northern and eastern side of the LC neither at the lags shown in Fig. 17 nor at higher lags (not shown). This is explained 1) by the reduction of the frequency of the LCFEs when traveling around the LC because of merging and growth but also 2) by displacements of the LC. We note that there is no signature of LCFEs at lag 0. This suggests that LCFEs are not triggered directly by the wind but result from a remote effect of the wind forcing.

The main candidate to explain a link between the synoptic wind events and the presence of LCFEs a few days after is the interaction between coastal-trapped waves over the CB and the LC front. The main argument is that the anomalies of SSH within the LC and along the CB connect backward in time with anomalies of SSH along the coast of the Gulf (Fig. 17). This is also suggested by the correlation between the surface relative vorticity along the CB and the SSH at Progreso and Port Isabel (Fig. 18d). The maximum correlation occurs at lags 2 and 7 days, respectively. We cannot conclude which of the locally forced or remotely forced SSH anomalies over the CB are the most important for LCFE triggering. It is probable that both contribute, as illustrated by the following: The formation of the negative anomaly of SSH (which corresponds to a cyclonic anomalous circulation) evidenced in Fig. 17 (lag 11) north of the CB and referred as D appears to originate from direct wind forcing over the CB. Indeed, this anomaly D can be traced backward in time to the near entrance of the Yucatan Channel at lag 3 (marked as E). Following backward in time, the anomaly E observed at lag 3 is connected with the negative SSH anomaly at lags 1 and 2 on the CB (F), which at first order is generated locally by the wind. On the other hand, the positive anomaly of SSH at lag 11 located within the notch of the CB (G) originates in the western Gulf of Mexico. Indeed, the positive anomaly can be traced backward in time to the near entrance of the Yucatan Channel at lag 6 (H) and in the western Gulf of Mexico at lag 0 (A), where it is forced by a southward synoptic wind event. Note also that some SSH anomalies related to LCFEs appear to be forced or reinforced at the northern tip of the CB (J and D). Some mechanisms that could explain how a coastal signal over the CB can produce sea level and vorticity fluctuations along the LC are discussed in the following section.

c. Origin and propagation of the coastal-trapped waves

Most of the SSH anomalies captured by the regression analysis along the western coast of the Gulf of Mexico propagate anticlockwise toward the CB and Yucatan Channel afterward. The positive anomaly observed at lag 0 (this pattern is referenced as A in Fig. 17) can be traced to the Bay of Campeche at lag 2 (B) and on the CB at lag 5 (C). Such propagation is also observed for a negative SSH anomaly from lags 4 to 10. These anomalies travel ~2000 km from the western coast of the Gulf of Mexico to the CB in about 6 days (Fig. 17). This leads to a propagation speed of about 3.8 m s−1. A finer estimation of this propagation speed can be inferred from a Hovmöller diagram of along-coast model SSH between January and March 2008 (Fig. 19b) that shows short-period variability (<10 days) propagating around the Gulf of Mexico. Between distance 1000 and 4000 km, these anomalies of SSH propagate with a phase speed close to 3.3 m s−1.

Fig. 19.
Fig. 19.

(a) Along-coast Hovmöller diagrams of wind stress (positive means that the along-coast component of the wind stress is oriented with the land on the right-hand side) and sea level anomalies (m) in (b) REF experiment and (c) MONTHLY experiment. The Hovmöller diagrams are built using data along a path described by the 30-m isobath from the Florida coast to the Yucatan coast. To compute the alongshore component, the wind stress has been rotated using local orientation of the 30-m isobaths. Daily data from January to April 2008 are used. Anomalies are computed by subtracting the mean values of the sea level height during this period. The x axis indicates the distance from the starting point on the west Florida shelf (km) together with the corresponding longitudes and latitudes. The path together with points separated by 500 km distance is shown in Fig. 5.

Citation: Journal of Physical Oceanography 46, 11; 10.1175/JPO-D-16-0052.1

The comparison of along-coast SSH Hovmöller diagrams between experiments REF and MONTHLY confirms that the short-period signals are produced by synoptic winds (Fig. 19). Indeed, the coastal SSH signals in MONTHLY (Fig. 19c) have much less amplitude than in REF (Fig. 19b). Moreover, the variability of along-coast SSH at periods less than 10 days is nearly absent in MONTHLY. But interestingly, lower-frequency (>10 days) and lower-amplitude signals are observed to propagate from the south of the Florida Peninsula (~700 km in Fig. 19b). Since these signals are not present in the atmospheric and lateral forcing fields (which force waves of higher frequency that hardily travel up to a distance of 700 km in Fig. 19c), it has to be mainly generated by the intrinsic ocean variability, probably by the oscillation and interaction of the LC with the Florida banks. Eddies impinging the western coasts of the Gulf also induce coastal waves, as suggested, for example, by the positive anomaly of SSH, which develops in the western Gulf of Mexico at the beginning of February 2008 (at distance 4000 km in Fig. 19c).

Note that the propagation of SSH anomalies from the northeast to the western Gulf of Mexico is not clear either in the regression analysis (Fig. 17) or in the along-coast Hovmöller diagram of model SSH, as shown between January and April 2008 in Fig. 19b. In this latter case, most of the patterns between distance 1000 and 3000 km are not propagating (suggesting they originate nearly simultaneously), and as mentioned previously, the SSH anomalies that propagate up to a distance of 6000 km (i.e., to the entrance of the Yucatan Channel) preferably originate near 3000 km (i.e., in the northwestern Gulf of Mexico).

The propagation of SSH anomalies from the western Gulf of Mexico to the Yucatan Channel (Figs. 17, 19b) and their ability to survive the sharp topographic changes when passing from the deep Bay of Campeche to the shallow CB is supported by the analysis of coastal sea level observations from 14 tide gauge stations. The observations were taken from stations located between Galveston in the northwestern Gulf of Mexico and Calica in the Caribbean Sea (Table 2; Fig. 20a). Hourly values of tide gauge data have been used. The mean and trend are removed via a least squares linear fit. Tides and high frequency are removed, subtracting a tidal harmonic analysis and applying a low-pass filter with 1 cpd cutoff frequency. The sea level correlations between Progreso and Port Isabel (label c in Fig. 20b) or between Progreso and Galveston (label d in Fig. 20b) show relative maxima at lags close to 5 and 7 days. The propagation speeds estimated from these lags are 3.7 and 3.9 m s−1, respectively. This confirms that sea level anomalies observed in the northwestern Gulf of Mexico can propagate to the CB. The separation times estimated from all the tide gauges pairs with distance larger than 500 km are shown in Fig. 20c. A least squares fit gives a mean propagation speed estimate of 3.4 m s−1 that is in very good agreement with the propagation speed estimated from model data in Fig. 19 (3.3 m s−1) and reinforces the evidence of coastal-trapped wave propagation and the validity of our numerical simulations.

Fig. 20.
Fig. 20.

a) Position of the 14 tide gauges used to estimate propagation speeds of coastal sea level signals. b) Lagged cross correlation between selected tide gauges a: RK-CD, b: VE-MU, c: PT-PR, and d: GA-PR. c) Lag (days) of the maximum cross-correlation coefficient (as shown in b) against distance along the coast between stations (103 km) computed for each of the tide gauges pairs with separation distance higher than 500 km. A least squares fit gives a mean phase speed estimate of 3.4 m s−1.

Citation: Journal of Physical Oceanography 46, 11; 10.1175/JPO-D-16-0052.1

The tide gauge data show, for some stations pairs, a relative maximum at or very close to zero lag (Fig. 20b). This maximum is unrelated to coastal-trapped waves, as it requires very high propagation speed. Although cold fronts are expected to generate coastal-trapped waves, their passage is quite fast and may generate sea level fluctuations with very small delays over the entire Gulf. These fast passages of atmospheric perturbations are the likely direct source of the relative maxima in correlations at short and null delays. Correlations computed from model data extracted at the location of the equivalent station pairs are given in Fig. 18a. They are generally in good agreement with the correlations obtained from observations, at the exception of the correlation for the Galveston–Progreso and Port Isabel–Progreso station pairs at zero lag. This might be explained by the fact that the simulated sea level does not include the inverse barometer contribution due to changes in atmospheric pressure, which is known to significantly contribute to the sea level variations in the region (e.g., Zavala-Hidalgo et al. 2003b).

d. Coastal-trapped waves and LCFEs formation

The regression analysis in section 4b suggested a link between coastal-trapped waves and LCFEs’ formation. But this raises the question of what mechanisms are at work in the CB to explain how incoming waves drive the formation of the LCFEs. The analysis of model fields at the CB for the period January to March 2008 provides some elements to answer this question.

The Hovmöller diagram of the SSH at 21.5°N shows high-frequency anomalies on the CB, west of 86.5°W (Fig. 21). These anomalies can be identified at a distance of 6000 km from Fig. 19b. Most are related to features propagating from the western Gulf of Mexico, although some with the shortest periods appear to be forced on the CB (between distances 5500 and 6000 km in Figs. 19a,b). The northward flow of the LC, identified by its maximum velocities between 86.5° and 86.3°W in Fig. 21b, shows short-period fluctuations. Peaks of velocity are associated with negative anomalies of SSH at the CB (Fig. 21a), increased zonal gradients of SSH (contours in Fig. 21a), and positive anomalies of surface relative vorticity (Fig. 21c). We interpret this as follows: the negative anomaly of SSH of the incoming coastal wave sharpens the zonal gradient of SSH when impacting the LC at the shelf break. This results in the acceleration of the northward flow (the Loop Current) that produces a pulse of positive vorticity (Fig. 21c) through modulation of the zonal gradients of meridional velocity. The Hovmöller diagram of surface relative vorticity along the shelf break of the CB confirms that the anomalies that propagate along the CB are generated near 21.2°N (Fig. 21d). They travel northward at a speed of about 0.3 m s−1.

Fig. 21.
Fig. 21.

Hovmöller diagrams of (a) SSH (m) and contours of zonal gradients of SSH (contours show values of 0.8, 0.9, 1, 1.1, and 1.2 × 10−5), (b) surface speed (m s−1), (c) vorticity (s−1) at 21.5°N, and (d) vorticity along the shelf break of the Campeche Bank (s−1). The path used to build the along-shelf diagram in (d) is indicated in Fig. 5 with a dashed black line. In (d), the x axis represents the distance from Cozumel Channel. Model outputs from REF experiment between January and March 2008 are used, as in Fig. 19b.

Citation: Journal of Physical Oceanography 46, 11; 10.1175/JPO-D-16-0052.1

The time taken by the vorticity anomalies to travel the 200 km separating the entrance of the Yucatan Channel (21°N) to the northern tip of the CB (23°N) is close to 8 days (Fig. 21d). As seen previously, this time scale is also characteristic of the dominant period for the LCFEs formation in the region [see also Table 1 in Walker et al. (2009)]. The regression analysis suggests that the SSH anomalies at the CB can trigger SSH (vorticity) anomalies at the entrance of the Yucatan Channel but can also trigger or reinforce the anomalies at the northern tip of the CB. It is noteworthy that a large part of the SSH signal reaching the CB does not travel south of the Yucatan Channel as seen in the along-coast Hovmöller diagram (Fig. 19b). This means that a significant fraction of the incoming energy is transferred to the LC by the process described above.

5. Summary and discussion

This work investigates the origin and fate of LCFEs along the Campeche Bank (CB) using in situ and high-resolution numerical simulations of the Gulf of Mexico. Comparison with in situ data shows that the model adequately represents the variability along the CB. So the model is used to further describe the characteristics of the LCFEs and to investigate the mechanisms whereby these cyclonic eddies are formed.

Measurements from January to April 2011 at an array of ADCPs located at the northeastern tip of the CB show intense 8–10-day velocity fluctuations in the LC at different depths between the near surface and 500 m. These fluctuations are of order ±0.3 m s−1 in the core of the LC and tend to be of opposite phase on its cyclonic shear side, suggesting that these perturbations are the signature of cyclonic LCFEs. This is in agreement with the analysis of the high-resolution model outputs. Indeed, the model shows short-period (period < 10 days), surface-intensified cyclonic eddies with horizontal scale of about 100 km traveling along the LC. The cyclonic eddies accelerate locally the core of the LC.

The analysis of the model outputs shows that these short-period LCFEs should be distinguished from the large and low-frequency LCFEs discussed in Athie et al. (2012) or in Zavala-Hidalgo et al. (2003a). In agreement with these previous studies, we found that (i) the low-frequency LCFEs are often related to LC detachments, and (ii) they are partly fed by Caribbean cyclones impinging on the Yucatan coast. In contrast, the short-period LCFEs form mainly at the entrance of the Yucatan Channel, and there is no indication that they contribute to the LC detachments.

Spectral analysis from 10-yr model outputs at different locations along the shelf break of the CB and Yucatan Channel confirms that the intraseasonal anomalies of surface vorticity (which peak at periods between 6 and 10 days) originate along the CB. Indeed, there is no evidence that these vorticity anomalies are advected from regions south of the entrance of the Yucatan Channel (21°N). First, the power spectrum of the vorticity south of Cozumel Island does not show spectral power as strong as at locations along the CB. Second, the mean kinetic energy of the perturbation with periods less than 10 days is low south of the entrance of the Yucatan Channel (Fig. 9b).

The mean kinetic energy of the perturbations with periods less than 15 days is high along the CB, as mentioned before but also in the open ocean between the shelf and 24.5°N (Fig. 10b). In the eastern part of the LC (along the west Florida shelf), most of the intraseasonal kinetic energy is contained at higher periods (30–90 days). This is in line with the spatial growth of the LCFEs from 40 to 80 km and merging as they propagate along the LC, as shown by the eddy-tracking analysis in section 3d. Such spatial distribution of the characteristic periods of the LC perturbations probably explains why the LCFEs from satellite data have been first pointed out in the northern and eastern LC. The high-frequency (periods < 15 days) and small horizontal-scale (<100 km) LCFEs, however, are not captured in gridded altimetry products (Fig. 2 and Walker et al. 2009). Note that cyclonic vorticity anomalies have been observed in this region from along-track altimetry (Zavala-Hidalgo et al. 2003a) or in situ data (Athie et al. 2012). But the variability described by these authors occurs at periods above 2 months and appears to have larger scale compared to the ~100-km radius of the LCFEs inferred in this study. These authors described another regime of variability that has been linked either with the advection of cyclonic vorticity anomalies from the Cayman Basin (Athie et al. 2012) or large-scale LC fluctuations (Zavala-Hidalgo et al. 2003a).

Diagnostics indicate that mixed baroclinic–barotropic instabilities are responsible for the growth of the short-period perturbations along and north of the CB. The presence of intense horizontal and vertical shears in the upper 100-m depth explains why the LC releases energy to the turbulent flow in this region. These results are in agreement with the numerical findings of Oey (2008), who showed that a preferred site for deep cyclogenesis in the Gulf of Mexico is north of CB and at the northeastern side of the shelf (near 23.5°–25°N, 87°–88°W).

One important result of this study is to point out the role played by atmospheric variability (with a dominant contribution of the winter cold surges) in driving the period of formation of the LCFEs along the eastern shelf break of the CB. The results suggest that the dynamic link between the wind variability and LCFEs is provided by sea level anomalies generated on the CB and coastal-trapped waves generated predominantly along the western coast of the Gulf of Mexico.

The propagation speed of the coastal SSH anomalies, estimated in this study (3.3 m s−1), is in agreement with along-coast in situ observations in the western and southwestern shelf of the Gulf of Mexico (Dubranna et al. 2011). These authors also found high energy in the 6–10-day band with signals propagating southward with phase speed between 3.24 and 4.01 m s−1. They associated this variability with first-mode coastally trapped waves. From cross-spectral analysis, Dubranna et al. (2011) showed that these waves are generated locally by the wind blowing along the western shelf of the Gulf. Moreover, they found that alongshore current variations over the western CB (near 92°W) are not correlated with local wind variations, suggesting that the 6–10-day signal on the western CB has a remote origin. From an extensive dataset of tide gauges’ measurements from 1946 to 2012, taken at different locations around the Gulf of Mexico, we provided additional evidence of coastal sea level signals propagating from the western Gulf of Mexico to the entrance of the Yucatan Channel with a mean phase speed of 3.4 m s−1 and surviving the sharp topographic changes when passing from the deep Bay of Campeche to the shallow CB. This is in agreement with our model results that show almost free wave propagation from the western Gulf of Mexico to the CB. Model results also suggest that high-frequency winds during winter favor the development of high-frequency variability along the LC during winter. This will have to be confirmed through analysis of longer time series of in situ data.

Interestingly, the MONTHLY experiment also generates coastally trapped waves propagating in the Gulf of Mexico with periods close to 2 weeks (Fig. 19c). Their amplitude is much lower than the amplitude of the wind-driven and short-period signals seen in REF (Fig. 19b). This signal could be related to the observations of Hallock et al. (2009) in the northeastern Gulf of Mexico. These authors observed coastally trapped waves with periods of about 16 and 21 days, amplitudes of 5–10 cm s−1, and wavelengths of about 500 km. They proposed that these waves could be forced by the interaction between the LC eddies and the topography on the northern coast of the Gulf of Mexico. We did not find evidence that these waves control the formation of LCFEs in MONTHLY experiment. The amplitude of this signal might be too low to exert a significant interaction with the LC. Other parameters such as the wavelength or the period might also play a role in the response of the LC to the incoming waves.

We found that the vorticity anomalies giving rise to the LCFEs with periods between 6 and 10 days mainly originate near 21.2°N on the shelf break at the southernmost part of the CB. Model results suggest that negative SSH anomalies associated with incoming coastal waves on the CB sharpen the zonal gradients of SSH and increase the northward flow when impinging on the LC. This acceleration produces a pulse of positive vorticity that propagates to the north, grows, and takes the form of a cyclonic eddy when detaching the shelf. The phase speed of the vorticity anomalies along the LC in the CB region is estimated to be close to 0.3 m s−1 (Fig. 21d). Interestingly, this value is close to the 0.45 m s−1 phase speed of wavelike meanders with a period of 7–8 days propagating downstream within the Gulf Stream south of Cape Hatteras (Luther and Bane 1985). In the Gulf of Mexico, this phase speed is much lower than the speed of the LC core’s flow near the surface, which generally exceeds 1.2 m s−1. So an open question is what processes determine this phase speed value. This could be related to Rossby waves’ dynamics propagating on the background and the sharp potential vorticity gradient between the LC and surrounding waters of the Gulf of Mexico (e.g., see the potential vorticity field in Fig. 3a of Chérubin et al. 2006). This point, together with further details on the mechanisms of interaction between the incoming waves and the LC, requires further investigations.

In regards to LC eddy shedding, we did not find evidence of any impact of the coastal-trapped waves on the eddy separation process. In both REF and MONTHLY simulations, the number of separation is of one or two per year in agreement with altimetry results (Leben 2005) or state-of-the-art models. This is because the dominant source of energy for the growth of the frontal perturbations involved in the ring liberation process is the baroclinic–barotropic conversion from the large-scale flow rather than the input of energy by the short-period, coastal-trapped waves.

Acknowledgments

We acknowledge the provision of supercomputing facilities by CICESE and GENCI Project GEN7298. The regional configuration was set up in cooperation with the DRAKKAR project (http://www.drakkar-ocean.eu/). Altimetry data were produced by Salto/Duacs and distributed by AVISO, with support from CNES. The GEKCO product used in this study was developed by J. Sudre at LEGOS, France. Sea level data were provided by (i) Servicio Mareográfico Nacional, Universidad Nacional Autónoma de México, Instituto de Geofísica (http://www.mareografico.unam.mx); (ii) the Joint Archive for Sea Level (http://ilikai.soest.hawaii.edu/uhslc/rqdsstc.html); and iii) CICESE. We also acknowledge F. Nencioli for providing the source code of his automatic eddy detection algorithm. Multiscale ultra-high resolution (MUR) SST data were obtained from the NASA EOSDIS Physical Oceanography Distributed Active Archive Center (PO.DAAC) at the Jet Propulsion Laboratory, Pasadena, California. (http://mur.jpl.nasa.gov/). We are grateful to E. Chassignet and two anonymous reviewers for helpful comments on the manuscript.

REFERENCES

  • Andrade-Canto, F., J. Sheinbaum, and L. Zavala Sansón, 2013: A Lagrangian approach of the Loop Current eddy separation. Nonlinear Processes Geophys., 20, 8596, doi:10.5194/npg-20-85-2013.

    • Search Google Scholar
    • Export Citation
  • Athie, G., J. Candela, J. Ochoa, and J. Sheinbaum, 2012: Impact of Caribbean cyclones on the detachment of Loop Current anticyclones. J. Geophys. Res., 117, C03018, doi:10.1029/2011JC007090.

    • Search Google Scholar
    • Export Citation
  • Beron-Vera, F. J., M. J. Olascoaga, and G. J. Goni, 2008: Oceanic mesoscale eddies as revealed by Lagrangian coherent structures. Geophys. Res. Lett., 35, L12603, doi:10.1029/2008GL033957.

    • Search Google Scholar
    • Export Citation
  • Bretherton, F. P., R. E. Davis, and C. B. Fandry, 1976: A technique for objective analysis and design of oceanographic experiments applied to MODE-73. Deep-Sea Res. Oceanogr. Abstr., 23, 559582, doi:10.1016/0011-7471(76)90001-2.

    • Search Google Scholar
    • Export Citation
  • Brodeau, L., B. Barnier, A.-M. Treguier, T. Penduff, and S. Gulev, 2010: An ERA40-based atmospheric forcing for global ocean circulation models. Ocean Modell., 31, 88104, doi:10.1016/j.ocemod.2009.10.005.

    • Search Google Scholar
    • Export Citation
  • Candela, J., J. Sheinbaum, J. Ochoa, A. Badan, and R. Leben, 2002: The potential vorticity flux through the Yucatan Channel and the Loop Current in the Gulf of Mexico. Geophys. Res. Lett., 29, 2059, doi:10.1029/2002GL015587.

    • Search Google Scholar
    • Export Citation
  • Chang, Y.-L., and L.-Y. Oey, 2012: Why does the Loop Current tend to shed more eddies in summer and winter? Geophys. Res. Lett., 39, L05605, doi:10.1029/2011GL050773.

    • Search Google Scholar
    • Export Citation
  • Chérubin, L. M., Y. Morel, and E. P. Chassignet, 2006: Loop current ring shedding: The formation of cyclones and the effect of topography. J. Phys. Oceanogr., 36, 569591, doi:10.1175/JPO2871.1.

    • Search Google Scholar
    • Export Citation
  • Dai, A., and K. Trenberth, 2002: Estimates of freshwater discharge from continents: Latitudinal and seasonal variations. J. Hydrometeor., 3, 660687, doi:10.1175/1525-7541(2002)003<0660:EOFDFC>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Donohue, K. A., D. R. Watts, P. Hamilton, R. Leben, M. Kennelly, and A. Lugo-Fernández, 2016: Gulf of Mexico Loop Current path variability. Dyn. Atmos. Oceans, doi:10.1016/j.dynatmoce.2015.12.003, in press.

    • Search Google Scholar
    • Export Citation
  • Dubranna, J., P. Brunius, J. M. Lopez, and J. Candela, 2011: Circulation over the continental shelf of the western and southwestern Gulf of Mexico. J. Geophys. Res., 116, C08009, doi:10.1029/2011JC007007.

    • Search Google Scholar
    • Export Citation
  • Dussin, R., B. Barnier, and L. Brodeau, 2014: The making of Drakkar forcing set DFS5. DRAKKAR/MyOcean Rep. 05-10-14, 29 pp. [Available online at https://www.drakkar-ocean.eu/forcing-the-ocean/the-making-of-the-drakkar-forcing-set-dfs5.]

  • Fratantoni, P. S., T. N. Lee, G. P. Podesta, and F. Muller-Karger, 1998: The influence of Loop Current perturbations on the formation and evolution of Tortugas eddies in the southern Straits of Florida. J. Geophys. Res., 103, 24 75924 779, doi:10.1029/98JC02147.

    • Search Google Scholar
    • Export Citation
  • Hallock, Z. R., W. J. Teague, and E. Jarosz, 2009: Subinertial slope-trapped waves in the northeastern Gulf of Mexico. J. Phys. Oceanogr., 39, 14751485, doi:10.1175/2009JPO3925.1.

    • Search Google Scholar
    • Export Citation
  • Hurlburt, H. E., and J. D. Thompson, 1982: The dynamics of the Loop Current and shed eddies in a numerical model of the Gulf of Mexico. Hydrodynamics of Semi-Enclosed Seas, J. C. J. Nihoul, Ed., Elsevier, 243–298.

  • Large, W. L., and S. G. Yeager, 2004: Diurnal to decadal global forcing for ocean and sea-ice models: The data sets and flux climatologies. NCAR Tech. Rep. TN-460+STR, 105 pp.

  • Leben, R. R., 2005: Altimeter-derived loop current metrics. Circulation in the Gulf of Mexico: Observations and Models, Geophys. Monogr., Vol. 161, Amer. Geophys. Union, 181–201.

  • Leclair, M., and G. Madec, 2009: A conservative leapfrog time stepping method. Ocean Modell., 30, 8894, doi:10.1016/j.ocemod.2009.06.006.

    • Search Google Scholar
    • Export Citation
  • Le Hénaff, M., V. H. Kourafalou, Y. Morel, and A. Srinivasan, 2012: Simulating the dynamics and intensification of cyclonic Loop Current frontal eddies in the Gulf of Mexico. J. Geophys. Res., 117, C02034, doi:10.1029/2011JC007279.

    • Search Google Scholar
    • Export Citation
  • Le Hénaff, M., V. H. Kourafalou, R. Dussurget, and R. Lumpkin, 2014: Cyclonic activity in the eastern Gulf of Mexico: Characterization from along-track altimetry and in situ drifter trajectories. Prog. Oceanogr., 120, 120138, doi:10.1016/j.pocean.2013.08.002.

    • Search Google Scholar
    • Export Citation
  • Luther, M. E., and J. M. Bane, 1985: Mixed instabilities in the Gulf Stream over the continental slope. J. Phys. Oceanogr., 15, 323, doi:10.1175/1520-0485(1985)015<0003:MIITGS>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Madec, G., 2016: NEMO ocean engine. Institut Pierre-Simon Laplace Note du Pole de Modélisation 27, 406 pp.

  • Masina, S., S. G. H. Philander, and A. B. G. Bush, 1999: An analysis of tropical instability waves in a numerical model of the Pacific Ocean. 2: Generation and energetics of the waves. J. Geophys. Res., 104, 29 63729 661, doi:10.1029/1999JC900226.

    • Search Google Scholar
    • Export Citation
  • Nencioli, F., C. Dong, T. Dickey, L. Washburn, and J. C. McWilliams, 2010: A vector geometry-based eddy detection algorithm and its application to a high-resolution numerical model product and high-frequency radar surface velocities in the Southern California Bight. J. Atmos. Oceanic Technol., 27, 564579, doi:10.1175/2009JTECHO725.1.

    • Search Google Scholar
    • Export Citation
  • Oey, L.-Y., 2008: Loop Current and deep eddies. J. Phys. Oceanogr., 38, 14261449, doi:10.1175/2007JPO3818.1.

  • Pichevin, T., and D. Nof, 1997: The momentum imbalance paradox. Tellus, 49A, 298319, doi:10.1034/j.1600-0870.1997.t01-1-00009.x.

  • Reffray, G., R. Bourdalle-Badie, and C. Calone, 2015: Modelling turbulent vertical mixing sensitivity using a 1-D version of NEMO. Geosci. Model Dev., 8, 6986, doi:10.5194/gmd-8-69-2015.

    • Search Google Scholar
    • Export Citation
  • Renault, L., M. J. Molemaker, J. C. McWilliams, A. F. Shchepetkin, F. Lemarié, D. Chelton, S. Illig, and A. Hall, 2016: Modulation of wind work by oceanic current interaction with the atmosphere. J. Phys. Oceanogr., 46, 16851704, doi:10.1175/JPO-D-15-0232.1.

    • Search Google Scholar
    • Export Citation
  • Rudnick, D. L., G. Gopalakrishnan, and B. D. Cornuelle, 2015: Cyclonic eddies in the Gulf of Mexico: Observations by underwater gliders and simulations by numerical model. J. Phys. Oceanogr., 45, 313326, doi:10.1175/JPO-D-14-0138.1.

    • Search Google Scholar
    • Export Citation
  • Schmitz, W. J., Jr., 2005: Cyclones and westward propagation in the shedding of anticyclonic rings from the Loop Current. Circulation in the Gulf of Mexico: Observations and Models, Geophys. Monogr., Vol. 161, Amer. Geophys. Union, 241–261.

  • Schultz, D. M., W. E. Bracken, and L. F. Bosart, 1998: Planetary- and synoptic-scale signatures associated with Central American cold surges. Mon. Wea. Rev., 126, 527, doi:10.1175/1520-0493(1998)126<0005:PASSSA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Seo, H., A. J. Miller, and J. R. Norris, 2016: Eddy–wind interaction in the California Current System: Dynamics and impacts. J. Phys. Oceanogr., 46, 439459, doi:10.1175/JPO-D-15-0086.1.

    • Search Google Scholar
    • Export Citation
  • Sheinbaum, J., G. Athie, J. Candela, J. Ochoa, and A. Romero, 2016: Structure and variability of the Yucatan and Loop Currents along the slope and shelf break of the Yucatan Channel and Campeche Bank. Dyn. Atmos. Oceans, doi:10.1016/j.dynatmoce.2016.08.001, in press.

    • Search Google Scholar
    • Export Citation
  • Sturges, W., N. G. Hoffmann, and R. R. Leben, 2010: A trigger mechanism for Loop Current ring separations. J. Phys. Oceanogr., 40, 900913, doi:10.1175/2009JPO4245.1.

    • Search Google Scholar
    • Export Citation
  • Sudre, J., C. Maes, and V. Garçon, 2013: On the global estimates of geostrophic and Ekman surface currents. Limnol. Oceanogr. Fluids Environ., 3, 120, doi:10.1215/21573689-2071927.

    • Search Google Scholar
    • Export Citation
  • Vukovich, F. M., and G. A. Maul, 1985: Cyclonic eddies in the eastern Gulf of Mexico. J. Phys. Oceanogr., 15, 105117, doi:10.1175/1520-0485(1985)015<0105:CEITEG>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Walker, N., R. Leben, S. Anderson, J. Feeney, P. Coholan, and N. Sharma, 2009: Loop Current frontal eddies based on satellite remote-sensing and drifter data. U.S. Dept. of the Interior, Minerals Management Service, Gulf of Mexico OCS Region, OCS Study MMS 2009-023, 88 pp.

  • Zavala-Hidalgo, J., S. L. Morey, and J. J. O’Brien, 2003a: Cyclonic eddies northeast of the Campeche Bank from altimetry data. J. Phys. Oceanogr., 33, 623629, doi:10.1175/1520-0485(2003)033<0623:CENOTC>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Zavala-Hidalgo, J., S. L. Morey, and J. J. O’Brien, 2003b: Seasonal circulation on the western shelf of the Gulf of Mexico using a high-resolution numerical model. J. Geophys. Res., 108, 3389, doi:10.1029/2003JC001879.

    • Search Google Scholar
    • Export Citation
1

Detailed positions and characteristics of the instruments are given in Table 1. Inertial and main tidal frequencies have been removed using a low-pass Lanczos filter with cutoff period of 2 days, and daily averages of the filtered data are used in this study.

2

As a diagnostic to identify the presence and position of the cyclonic eddies from model data, we make use of FTLEs. FTLEs provide a measure of the maximum separation rate of initially nearby fluid particles in a finite time and have proven to be useful for identifying mesoscale eddies (e.g., Beron-Vera et al. 2008). A detailed description of the method used to compute FTLEs from high-resolution model data is given in Andrade-Canto et al. (2013). Here, FTLE maps are computed backward in time using 25 days of horizontal velocities at 100-m depth.

Save