• Anthes, R. A., 1983: Regional models of the atmosphere in middle latitudes. Mon. Wea. Rev.,111, 1306–1335.

  • Bélair, S., D.-L. Zhang, and J. Mailhot, 1994: Numerical prediction of the 10–11 June 1985 squall line with the Canadian Regional Finite-Element Model. Wea. Forecasting,9, 157–172.

  • Benoit, R., J. Côté, and J. Mailhot, 1989: Inclusion of a TKE boundary layer parameterization in the Canadian regional finite-element model. Mon. Wea. Rev.,117, 1726–1750.

  • ——, M. Desgagne, P. Pellerin, Y. Chartier, and S. Desjardins, 1997:The Canadian MC2: A semi-Lagrangian, semi-implicit wide-band atmospheric model suited for fine-scale process studies and simulation. Mon. Wea. Rev.,125, 2382–2415.

  • Cardone, V. J., R. E. Jensen, D. T. Resio, V. R. Swail, and A. T. Cox, 1996: Evaluation of contemporary ocean wave models in rare extreme events: “Halloween storm” of October 1991 and the“storm of the century” of March 1993. J. Atmos. Oceanic Technol.,13, 198–230.

  • Chouinard, C., J. Mailhot, H. Mitchell, A. Staniforth, and R. Hogue, 1994: The Canadian regional data assimilation system: Operational and research applications. Mon. Wea. Rev.,122, 1306–1325.

  • Delage, Y., 1988: The position of the lowest levels in the boundary layer of atmospheric circulation models. Atmos.–Ocean,26, 329–340.

  • Doyle, J. D., and T. T. Warner, 1993: The impact of the sea surface resolution on mesoscale coastal processes during GALE IOP 2. Mon. Wea. Rev.,121, 313–333.

  • Fritsch, J. M., and C. R. Chappell, 1980: Numerical prediction of convectively-driven mesoscale pressure systems. Part I: Convective parameterization. J. Atmos. Sci.,37, 1722–1733.

  • Glendening, J. W., 1994: Dependance of boundary-layer structure near ice-edge coastal front upon geostrophic wind direction. J. Geophys. Res.,99 (D), 5569–5581.

  • ——, and J. D. Doyle, 1995: Mesoscale response to a meandering surface temperature interface. J. Atmos. Sci.,52, 505–518.

  • Holton, J. R., 1992: An Introduction to Dynamic Meteorology. 3d ed. International Geophysics Series, Vol. 48, Academic Press, 523 pp.

  • Hsu, S. A., 1988: Coastal Meteorology. Academic Press, 260 pp.

  • Huo, S., D.-L. Zhang, J. Gyakum, and A. Staniforth, 1995: A diagnostic analysis of the superstorm of March 1993. Mon. Wea. Rev.,123, 1740–1761.

  • Kuo, H. L., 1974: Further studies on the parameterization of the influence of cumulus convection on large-scale flow. J. Atmos. Sci.,31, 1232–1240.

  • Low-Nam, S., and Y.-H. Kuo, 1991: The mesoscale structure of sea surface temperature and its impact on rapid cyclogenesis. Preprints, First Int. Symp. on Winter Storms, New Orleans, LA, Amer. Meteor. Soc., 426–430.

  • Mailhot, J., 1994: The Regional Finite-Element (RFE) model scientific description. Part 2: Physics. [Available from RPN, 2121 Trans-Canada, Dorval, PQ H9P 1J3, Canada.].

  • ——, and C. Chouinard, 1989: Numerical forecasts of explosive winter storms: Sensitivity experiments with a meso-α scale model. Mon. Wea. Rev.,117, 1311–1343.

  • ——, and Coauthors, 1995: Change to the Canadian regional forecast system: Description and evaluation of the 50-km version. Atmos.–Ocean,33, 55–80.

  • ——, R. Sarrazin, B. Bilodeau, N. Brunet, and G. Pellerin, 1997: Development of the 35-km version of the Canadian regional forecast system. Atmos.–Ocean,35, 1–28.

  • Nuss, W. A., and C. Lilly, 1989: Boundary layer convergence along the Gulf Stream during ERICA IOP2. Preprints, First Int. Symp. on Winter Storms, New Orleans, LA, Amer. Meteor. Soc., 423–425.

  • Perkey, D. J., W. M. Lapenta, F. R. Robertson, and C. W. Kreitzberg, 1991: The role of the sea surface temperature distribution on numerically simulated cyclogenesis during ERICA IOP-4. Preprints, First Int. Symp on Winter Storms, New Orleans, LA, Amer. Meteor. Soc., 418–422.

  • Shuman, F. G., 1957: Numerical methods in weather prediction: II. Smoothing and filtering. Mon. Wea. Rev.,85, 357–361.

  • Skey, S. G. P., K. Berger-North, and V. R. Swail, 1995: Detailed measurement of winds and waves in high sea states from a moored NOMAD weather buoy. Proc. 4th Int. Workshop on Wave Hindcasting and Forecasting, Banff, AB, Canada, Environment Canada, 213–223.

  • Smith, S. D., 1981: Factors for adjustment of wind speed over water to a 10 metre height. Rep. Series/BI-R-81-3/March 81, Bedford Institute of Oceanography, Bedford, NS, Canada, 29 pp. [Available from Bedford Institute of Oceanography, P.O. Box 10026, Darmouth, NS B2Y 4A2, Canada.].

  • Stommel, H., 1965: The Gulf Stream: A Physical and Dynamical Description. 2d ed. University of California Press, 248 pp.

  • Strong, A. E., and R. J. De Rycke, 1973: Ocean current monitoring employing a new satellite sensing technique. Science,182, 482–484.

  • Sweet, W., R. Fett, J. Kerling, and P. LaViolette, 1981: Air-sea interaction effects in the lower troposphere across the north wall of the Gulf Stream. Mon. Wea. Rev.,109, 1042–1052.

  • Tanguay, M., A. Robert, and R. Laprise, 1990: A semi-implicit semi-Lagrangian fully compressible regional forecast model. Mon. Wea. Rev.,118, 1970–1980.

  • Thomas, B., 1995: Use of an interactive graphical analysis system to hindcast the Storm of the Century, March 12–15, 1993. Proc. Fourth Int. Workshop on Wave Hindcasting and Forecasting, Banff, AB, Canada, Environment Canada, Downsview, 93–108.

  • Warner, T. T., M. N. Lakhtakia, J. D. Doyle, and R. A. Pearson, 1990:Marine atmospheric boundary layer circulations forced by Gulf Stream sea surface temperature gradients. Mon. Wea. Rev.,118, 309–323.

  • View in gallery
    Fig. 1.

    Synoptic surface situation from CMC analysis at 0000 UTC 14 Mar to 0000 UTC 15 Mar 1993. Solid lines represent pressure at sea level (4 hPa). Dashed lines represent 1000-hPa temperatures (5°C). Wind barbs are knots. Hours represented are (a) 0000 UTC 14 Mar, b) 1200 UTC 14 Mar, and (c) 0000 UTC 15 Mar.

  • View in gallery
    Fig. 2.

    Grids for different resolutions used during the cascade process. Shaded areas correspond to the nesting buffer zone of each grid. In (a), cascade grids for 50, 25, and 10 km. In (b), cascade grids for 10, 5, and 2 km on the detailed SST (see Fig. 3a).

  • View in gallery
    Fig. 3.

    Sea surface temperature (2°C) field at (a) 25 km from digitized METOC SST analysis (S0) valid from 0000 UTC 12 Mar to 2100 UTC 15 Mar 1993. In (b), S0 filtered 2500 times (S2)

  • View in gallery
    Fig. 4.

    Difference (2°C, gray scale) between the detailed SST and the smoothest filtered SST (S0 − S2) superimposed over S0 (2°C, black lines).

  • View in gallery
    Fig. 5.

    Location of the buoys in relation to the detailed 25-km SST field.

  • View in gallery
    Fig. 6.

    Wind speed at 10 m (5 kt or 2.5 m s−1, labeled white lines) from run 2, superimposed over the detailed SST field (2°C, gray scale). Hours represented are (a) 1200 UTC 14 Mar 1993, corresponding to a 21-h simulation time, and (b) 0000 UTC 15 Mar 1993, corresponding to a 33-h simulation time.

  • View in gallery
    Fig. 7.

    Bulk Richardson number (gray scale) in the first 100 m from run 2 superimposed over the SST field (2°C, black). Hours represented are the same as in Fig. 6. Gray scale goes from light (unstable) to dark (stable).

  • View in gallery
    Fig. 8.

    Wind field at 10 m (2 kt or 1 m s−1, gray scale and black lines) from run 4 superimposed over the detailed SST field (2°C, labeled white). Hours represented are (a) 2000 UTC 14 Mar, (b) 2200 UTC 14 Mar, and (c) 0000 UTC 15 Mar 1993 being, respectively, 5-, 7-, and 9-h simulation times. The W refers to a maximum created by the warm eddy and C1 and C2 refer to two minimums in the wind speed, created by the cold front and a tongue of cooler water. The location of the East Scotian Slope buoy (44137) is indicated by the encircled cross. Refer to Fig. 2b for the location of the area.

  • View in gallery
    Fig. 9.

    Wind fields at 10 m (5 kt or 2.5 m s−1) from run 3 (S0, gray scale), run 3a (S1, dashed black), and run 3b (S2, thin solid). Hours represented are the same as in Fig. 6 but, respectively, correspond to (a) a 12-h simulation time and (b) a 24-h simulation time. Labeled lines are those corresponding to the same isotachs for runs 3a and 3b. Cold front position is from run 3.

  • View in gallery
    Fig. 10.

    Gridpoint data from runs 1, 2, and 3, representing East Scotian Slope (44137) buoy with the buoy observations: In (a) 10-m wind direction, in (b) 10-m wind speed, in (c) air surface temperature and sea surface temperature, and in (d) sea level pressure. For the definition of lines, see legend in graph.

  • View in gallery
    Fig. 11.

    Same as Fig. 10b but in (a) from run 3, 4, and 5 (S0) and in (b) from runs 3b, 4b, and 5b (S2). For the definition of lines, see legend in graph. Four-hour shifted series are represented by lower series in (a). Note that 10 k (or 5 m s−1) have been subtracted to split the series.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 71 21 2
PDF Downloads 10 6 0

The Influence of Mesoscale Features of the Sea Surface Temperature Distribution on Marine Boundary Layer Winds off the Scotian Shelf during the Superstorm of March 1993

Serge DesjardinsMaritimes Weather Centre, Atmospheric Environment Service, Bedford, Nova Scotia, Canada

Search for other papers by Serge Desjardins in
Current site
Google Scholar
PubMed
Close
,
Robert BenoitRecherche en Prévision Numérique, Atmospheric Environment Service, Dorval, Quebec, Canada

Search for other papers by Robert Benoit in
Current site
Google Scholar
PubMed
Close
, and
Val SwailClimate Research Branch, Atmospheric Environment Service, Downsview, Ontario, Canada

Search for other papers by Val Swail in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

This paper studies the mesoscale wind field during the blizzard of March 1993 off the east coast of North America and examines the influence of the sea surface temperature distribution on surface winds. Can the Gulf Stream and its meanders, by its strong influence on the marine boundary layer, generate mesoscale features in the wind field? Numerical simulations of the storm are carried out using the MC2, a fully elastic nonhydrostatic model. Simulations are conducted at different resolutions (50, 25, 10, 5, and 2 km) with both detailed and smoothed SST fields, so as to examine the influence of these parameters on the marine boundary layer winds. Results from these numerical simulations are compared with surface observations from buoys. The study reveals some mesoscale features in the wind field caused by the Gulf Stream’s meanders and the warm eddies of the SST field. In a stable boundary layer, the meanders shaped a 50–55-kt (26–28 m s−1) band of winds in a general 40–45-kt (21–23 m s−1) wind field. Behind the cold front, local enhancements of 10-kt (5 m s−1) winds were found over the warm water eddies in the unstable boundary layer.

Corresponding author address: Serge Desjardins, Environment Canada, Atlantic Region, 15th floor, Queen Centre, 45 Alderney Drive, Darmouth, NS B2Y 2N6, Canada.

Email: Serge.Desjardins@ec.gc.ca

Abstract

This paper studies the mesoscale wind field during the blizzard of March 1993 off the east coast of North America and examines the influence of the sea surface temperature distribution on surface winds. Can the Gulf Stream and its meanders, by its strong influence on the marine boundary layer, generate mesoscale features in the wind field? Numerical simulations of the storm are carried out using the MC2, a fully elastic nonhydrostatic model. Simulations are conducted at different resolutions (50, 25, 10, 5, and 2 km) with both detailed and smoothed SST fields, so as to examine the influence of these parameters on the marine boundary layer winds. Results from these numerical simulations are compared with surface observations from buoys. The study reveals some mesoscale features in the wind field caused by the Gulf Stream’s meanders and the warm eddies of the SST field. In a stable boundary layer, the meanders shaped a 50–55-kt (26–28 m s−1) band of winds in a general 40–45-kt (21–23 m s−1) wind field. Behind the cold front, local enhancements of 10-kt (5 m s−1) winds were found over the warm water eddies in the unstable boundary layer.

Corresponding author address: Serge Desjardins, Environment Canada, Atlantic Region, 15th floor, Queen Centre, 45 Alderney Drive, Darmouth, NS B2Y 2N6, Canada.

Email: Serge.Desjardins@ec.gc.ca

1. Introduction

The Gulf Stream (Stommel 1965) is like a warm river in the ocean evolving very slowly when compared to the timescale of meteorological synoptic systems. It originates from the Straits of Florida. It follows a narrow northeastward track up to Cape Hatteras, where under the influence of the westerlies and, more importantly, under the Coriolis force, it turns more eastward and becomes diffuse as it continues to the coast of northwestern Europe. On the Scotian Shelf, particularly in winter and spring, it meanders, sometimes with large amplitude, forming large loops that pinch off Gulf Stream water “rings.” Gulf Stream warm core eddies, generated on the poleward side or the north wall side of the Gulf Stream, drift northward and westward, and last for months or sometimes are reabsorbed by the Gulf Stream.

According to Sweet et al. (1981), significant weather changes have often been noted by observers crossing the north wall of the Gulf Stream. Oceanographically, a commonly observed phenomenon is the sudden change in sea state from near-calm seas over the cold slope water to rougher seas over the Gulf Stream. This could be attributed to the difference in wind stress on the surface water as a result of differences in the relative motion of the Gulf Stream and the slope water under a given wind regime (Strong and De Rycke 1973). However, air–sea temperature differences, contributing to greater instability over the Gulf Stream, also act to favor a turbulent transfer of stronger winds to the sea surface over that area, producing rougher seas in comparison to the slope water area. Sweet et al. (1981) also noticed an abrupt increase in the wind speed from 10 to 20 kt (5–10 m s−1) over cold water to 25–35 kt (13–18 m s−1) when crossing the north wall of the Gulf Stream. Hsu (1988) explains this as a sea-breeze-like wind that blows from the cool water side to the warm water side of the Gulf Stream (i.e., a direct thermal circulation).

A number of studies have been carried out on the impact of the mesoscale structure of the sea surface temperature (SST), especially on the influence of the Gulf Stream on East Coast cyclogenesis (Mailhot and Chouinard 1989; Low-Nam and Kuo 1991; Perkey et al. 1991). They showed that the interaction of the SST with the atmosphere through the marine atmospheric boundary layer (MABL) plays a major role in the early stage of cyclogenesis. Other researchers have studied the 3D mesoscale circulations generated by the SST distribution, in the MABL, especially in the vicinity of the Gulf Stream (Sweet et al. 1981; Nuss and Lilly 1989;Warner et al. 1990; Doyle and Warner 1993). Most of these studies were done at high resolutions on very small domains where the Gulf Stream did not meander and where the main feature was the large SST gradient. However, Glendening and Doyle (1995) made a theoretical study of the impact of the scale of the SST meanders on the boundary layer. They found that with relatively large-scale meanders without geostrophic forcing, a meander signature will appear in the boundary layer structure. For relatively small-scale meanders, the meander signature is lost to alongfront blending of the geostrophic adjustment processes, orienting the gradients perpendicular to the mean SST front. Finally, with strong large-scale geostrophic forcing, their study demonstrates that the size of the meanders has to be increased to reach the large-scale limit.

The present study was conducted to investigate the combined impact of several features in the SST field, such as the Gulf Stream’s meanders and the warm and cold eddies of the cold slope water sea, on the synoptic marine boundary layer. Therefore, one needs a larger study area to gauge the impact on the whole synoptic system, that is, more than local adaptation. Could these mesoscale features in the SST field create local enhancements or decreases in the surface wind field that could eventually influence, at the same scale, the generation of ocean waves?

The study is essentially focused on the case of 13–15 March 1993 known as the “blizzard” or “superstorm” of March 1993 or sometimes called the “Storm of the Century” (Huo et al. 1995; Cardone et al. 1996). It was carried out with the help of the Mesoscale Compressible and Community (MC2) model developed at Environment Canada (Recherche en Prévision Numérique or RPN), which by its versatility allowed various high-resolution simulations of the blizzard of March 1993. By using a detailed SST analysis as a control field, the influence of the SST distribution on the surface wind field was tested in sensitivity experiments where some SST features were removed through filtering technique.

The present paper is structured as follows: section 2 describes the blizzard of March 1993 and its wind field, while section 3 contains information on the numerical model used, followed by the methodology. Section 4 includes test results made to ensure that the Canadian Meteorological Centre (CMC) analysis and the 50-km MC2 were valid representations of the real situation before proceeding to higher-resolution simulations. Results of various higher-resolution simulations showing the influence of the Gulf Stream and the SST distribution on surface winds are presented in section 5. Finally, section 6 makes the link between the numerical results and the observations. A summary of the conclusions is contained in section 7.

2. Synoptic situation

a. The storm’s pressure pattern

From 13 to 15 March 1993, the east coast of North America was hit by one of the deepest extratropical low pressure systems ever affecting this part of the world. The blizzard known as the Storm of the Century, deepened explosively in the Gulf of Mexico and over the southeastern United States. A picture of the synoptic situation from the CMC analyses is given in Fig. 1. For a more complete description of the storm, the reader is referred to Cardone et al. (1996), Thomas (1995), and Huo et al. (1995). At 0000 UTC 13 March 1993, a strong cyclonic low pressure system at 996 hPa was located in the Gulf of Mexico. It tracked northeastward and continued to intensify rapidly over the following 24 h. At 0000 UTC 14 March (Fig. 1a) the storm center reached the state of Delaware, where the pressure at the low center bottomed out at 963 hPa. It continued to track northeastward along the East Coast and slowly filled to reach Anticosti Island by 0000 UTC 15 March 1993 (Fig. 1c) with a central pressure value of 970 hPa (968 hPa observed).

An intense hyperbaroclinic zone was one of the important energy sources of the storm. Its southern edge was delimited by a surface warm front extending eastward from the low and a surface cold front extending south–southwestward from a wave. Because of the well-developed isobaric circularity of the storm, no significant surface trough was present, resulting in a rather gradual change in wind direction. The authors would like to point out that this latter fact created great difficulty in accurately locating the surface cold front. However, the presence of the surface cold front was well marked by a significant fall in temperature and a rise in pressure.

b. The storm’s wind field

The surface pressure gradient around the storm was very intense throughout its life cycle. According to the CMC analysis, corresponding surface geostrophic winds of 75–125 kt or 39–64 m s−1 surrounded the center at 0000 UTC 14 March 1993—the deepest stage of the storm. The strongest geostrophic winds were found in the northwest sector of the low, in a narrow band north of the warm front, and in the warm sector ahead of the cold front. North of the warm front, easterlies 45–65 kt or 23–33 m s−1 (50%–60% of the surface pressure gradient, or SPG) were reported by numerous coastal stations. Ahead of the surface cold front, a low-level jet of south-southeasterlies at 50–65 kt or 26–33 m s−1 (65% of the SPG) was present. Behind the surface cold front, although the pressure gradient slackened slightly, southwesterlies at 40–60 kt or 21–31 m s−1 (50%–70% of the SPG) remained.

Continuous cold air advection behind the surface cold front destabilized the marine boundary layer and allowed downward momentum transfer, creating geostrophic to supergeostrophic wind conditions. At 0000 UTC 15 March 1993, southwesterly winds at 40–50 kt or 21–26 m s−1 (100%–130% of the SPG) were blowing over the southern waters of Nova Scotia. For a more complete description of the storm’s wind field, the reader is referred to Thomas (1995) and Cardone et al. (1996).

3. Methodology

a. MC2 model

All numerical simulations were done with the MC2 model, which is based on the discretization of the full-elastic nonhydrostatic equations of Tanguay et al. (1990). The model solves the full set of Euler equations on a limited-area Cartesian domain of the polar projection with time-dependent nesting of the lateral boundary conditions supplied by a large-scale model or an analysis. The MC2 model uses numerical algorithms of semi-Lagrangian advection and semi-implicit time differencing. It has so far proven to be a versatile modeling tool that allows excellent simulations over a wide spectrum of scales. For further information about the MC2 model, the reader is referred to Benoit et al. (1997). All simulations were done using the RPN full physics package (Mailhot et al. 1995; Mailhot et al. 1997). Here we give some brief notes on the boundary layer scheme. The planetary boundary layer is based on a prognostic equation for turbulent kinetic energy (Benoit et al. 1989) together with a stratified surface layer for the turbulent surface fluxes; the sea surface roughness is based on a Charnock formula. Shallow convection is simulated with a method described by Mailhot (1994) and Mailhot et al. (1995), and is treated as a special case of the turbulent planetary boundary layer to include the saturated case in the absence of precipitation.

Vertical transfers due to turbulent air motion are parameterized in the form of vertical diffusion (Benoit et al. 1989). The effect is strongly dependent on the vertical diffusion coefficient, which is locally evaluated every time step. A surface layer diagnostic routine gives a diagnostic (i.e., not derived from a time-dependent differential equation) value of the state of the surface layer variables (only u, υ, T, and humidity) at any reasonable height, based on the current surface stress and turbulent fluxes determined from the rest of the boundary layer scheme, and on some simple assumption on the low-level profiles (Delage 1988).

b. Numerical simulations

The MC2 model was run on an HP 9000 series 755 workstation with 160-MB RAM. The top of the model was set at 25 km. The first two thermodynamic levels were set at 0 and 20 m to fix the first momentum level at 10 m. Winds from the model were extracted at this level. Different numbers of levels (15, 20, 30) were tested. The results showed no significant differences at the surface. Therefore, for an optimization question of space and time of the integration on the machine used, a set of 20 computational levels was eventually chosen, with half of the levels below 2.5 km, including 6 in the first kilometer. Although this is barely enough to resolve the details of the boundary layer mixing process, especially in the stable case, this should at least give a correct simulation of the surface layer winds, which is the focus of the present study. Simulations at different horizontal resolutions—50, 25, 10, 5, and 2 km—were carried out. The MC2 50-km run was generated by using boundary conditions from the CMC 50-km analysis (Chouinard et al. 1994; Mailhot et al. 1995). Higher-resolution runs were obtained by cascade, using the previous run at lower resolution as a “nesting file” to supply the initial and the lateral boundary conditions. Horizontal diffusion was 60 000 m2 s−1 at 50 km then decreased in proportion to Δx. Figure 2 shows the grids employed during the cascade process, which is summarized in Table 1. The resolution is given in the first column. The second column is the starting date of the simulations. In the third column, the time step is found, duration of the integration is given in the fourth column. The fifth column is the nesting frequency. Information on the time offset relative to the starting date is found in the sixth column and, finally, the grid size, for each simulation, is given in the last one.

Two deep convection schemes were used at the different resolutions. A version (Mailhot and Chouinard 1989) of the Kuo scheme (Kuo 1974) parametrized the deep convection at 50 km and was replaced by a modified version (Bélair et al. 1994) of the Fritsch–Chappell scheme (Fritsch and Chappell 1980) for the 25-, 10-, and 5-km simulations. At 2 km, a simple condensation scheme without any subgrid-scale condensation scheme (Mailhot and Chouinard 1989) was employed.

c. Sea surface temperature fields

In the control experiment, the detailed SST field (labeled hereinafter S0) was obtained from the SST analysis from the Canadian Forces Meteorological and Oceanographical Centre Halifax (METOC), valid from 0000 UTC 12 March to 2100 UTC 15 March 1993. This SST analysis was digitized (by hand), on a 25-km grid (Fig. 3a). The scale of eddies was on the order of 100 km while the approximate wavelength of the Gulf Stream meanders was around 500 km. For the corresponding grids at 50, 10, 5, and 2 km, this SST field was simply interpolated on those grids. Therefore, no further detail in the SST field has been added in the higher-resolutions runs.

A Shuman filter (Shuman 1957), for which the response is
i1520-0493-126-11-2793-e1
was applied to the digitized SST field on the control grid (therefore Δx = 25 km) to test the influence of the SST distribution. The number of iterations is given by m while n refers to the wavelength number. Table 2 summarizes the various SST fields used for each run. The first column gives the resolution of the experiment. The next three columns summarize and label all simulations carried out with the different SST fields. The first filtering (250 iterations, cutoff wavelength = 9Δx) eliminated only the shortwave variations in the SST field in order to preserve the main meanders and eddies (hereinafter S1). The S1 SST field (not shown) used in runs 1–3a was similar to the SST from the CMC analysis, except the latter did not have the western cold eddy. The second filtering (2500 iterations, cutoff wavelength = 18Δx) produced an SST field without eddies nor Gulf Stream keeping only the SST gradient between the Sargasso Sea and the Scotian Shelf waters (hereinafter S2, Fig. 3b). Note, however, that the filtering slightly cooled the southern waters and warmed the northern ones as can be seen from Fig. 3.

A subtraction of the S2 SST from the S0 SST field (S0 − S2) superimposed over the S0 SST is shown in Fig. 4. It revealed local differences over cold and warm eddies. A sinusoidal band of warmer water (by 3°–6°C) for the S0 SST followed the meanders of the Gulf Stream, while local maxima of 5°C for the S0 SST indicated the location of the warmer eddies north of the Gulf Stream. Over cold eddies, the S0 SST was 3°C cooler when compared to the S2 SST. Similar results were found when the S0 SST and the S1 SST fields were compared (S0 − S1, not shown here). The differences were 3°C or less and were less localized.

4. Validation

As a preface to this section, it should be noted that for this section and the next ones, all model fields shown and used exclude the lateral buffer zone employed for MC2’s nesting procedure. This nesting, and also the cascade procedure that we employ, is very common and well documented in the review paper of Anthes (1983). Obviously, the errors associated with the coupling of the model to the external flow are not strictly confined to that buffer zone. The higher truncation errors present in the coarser modeling of the external flow are advected toward the interior. Also, wave reflections and other inconsistencies can sometimes occur in that zone and diffuse to a certain extent toward the inner region. The nesting scheme generally succeeds in keeping this noise within or close to the buffer zone. In practice, however, for the case examined here and for the low-level wind field, it appears that the advected truncation errors do not affect severely the results as can be seen for instance from the time series of Fig. 10 (section 6), obtained from the 50-, 25-, and 10-km simulations.

a. Numerical gridpoint data versus buoy observations

The lateral boundary conditions for the MC2 model were supplied by the CMC analysis. Different fields, such as wind, sea level pressure, and temperature, from the analysis were compared with the handmade synoptic fields and buoy observations. Canadian buoys report winds as 10-min vector averages and maximum wind speeds are extracted from the highest running 8-s maximums (Skey et al. 1995). All numerical outputs comparisons with buoy observations were done using a cubic interpolation scheme to produce a numerical value at the buoy location.

First, visual comparisons (not shown here) indicated that the CMC analysis was a good numerical representation of the real situation. Second, comparisons between buoy observations and gridpoint data representing the buoy locations in the CMC analysis (resolution of 50 km) and in the 50-km MC2 simulation were done for eight buoys. Comparisons of times series between data from numerical analysis, prognostics, and observations gave very similar agreement for each buoy. Table 3 lists the buoys used. The last column gives the highest resolution at which a given buoy was employed for comparisons. Figure 5 shows the location of real buoys employed in relationship to the detailed 25-km SST field. Figure 10 (see 50-km curve) shows the case for Canadian buoy 44137.

b. CMC analysis versus 50-km resolution prognostics

In general, from an isobaric pattern perspective, the evolution of the storm was well forecasted by all numerical models. In this research, the nonhydrostatic MC2 50-km output from run 1 (see Table 2) was compared with the operational 50-km Regional Finite Element model (RFE) (Mailhot et al. 1995) output as well as with the CMC analysis. Both models and the analysis agreed closely on the position and center value of the low and the main isobaric features such as troughs. Isobaric patterns, as well as surface winds generated by both numerical models, were very comparable to those derived from the analysis and showed no major differences. Differences in wind speed were less than 10 kt (5 m s−1). The 1000-hPa wind field from the analysis had stronger winds in the southerly low-level jet, while the MC2 developed stronger winds behind the storm. However, the difference was only about 5 kt (2.5 m s−1).

From a temporal perspective, the low pressure system from both (MC2 and RFE) models slowed down when compared to the observations. This had a direct effect on the timing of the wind shift behind the low pressure system and certainly introduced a bias in the nesting of the finer-mesh file right at the beginning. As will be seen later (in section 6), this difference of about 4–6 h between observations and model outputs was present in all simulations.

c. Higher-resolution runs

No major synoptic differences were noticed between lower- and higher-resolution simulations. Differences in the isobaric patterns were observed in the vicinity of the frontal trough causing local differences in the wind direction. These were mainly due to the slight difference in the propagation speed at different resolutions of the surface trough. Run 2 using S0 produced a wind speed field having more mesoscale details when compared to the smoother wind speed field generated by run 1 using S0 interpolated on a 50-km grid. Higher-resolution runs (3, 4, 5) added more details in the wind speed field but differences in wind speed values were never more than 3 kt (1.5 m s−1). Finally, runs 4 and 5 revealed an interesting phenomenon in the wind speed field that will be discussed later (in section 5c).

5. Results from the simulations

a. Stability factor

The sea surface temperature modifies the marine boundary layer above it, by turbulent vertical exchanges of mass, momentum, moisture, and heat, particularly in the region of the Gulf Stream. Important air–sea interaction processes such as ocean wave induced surface stresses, heat, moisture, and momentum fluxes are directly or indirectly interconnected with the stability of the marine boundary layer. In the present paper, the influence of the Gulf Stream and the sea surface temperature distribution on surface wind was assessed by the bulk Richardson number calculated in the first 100 m by using the following relation:
i1520-0493-126-11-2793-e2
Note, in the case of |Rib| > 1 the value was set to unity with the proper sign. Here, Rib > 0.25 refers to statically stable conditions strong enough to damp mechanical turbulence production, while Rib ⩽ 0.25 refers to conditions where mechanical production is intense enough to sustain turbulence in a statically stable boundary layer. Finally, Rib ⩽ 0 denotes a statically unstable layer where convection is the generator of turbulence.

Synoptically, the Gulf Stream and its meanders, as well as the local warm eddies, often act as stationary mesoscale low-level unstable or slightly stable regions. In the unstable case, they can be seen as local shallow convective mixed layers where stronger vertical exchanges of momentum will help to enhance the surface winds. In the weak stable case, very weak vertical shear in the wind is needed to sustain turbulence, which also favors stronger vertical exchanges of momentum. Over colder water areas, the stabilizing effect of the water temperature on the near-surface air temperature mainly reduces the vertical momentum fluxes resulting in lighter surface winds. With adequate model resolution, one should see the influence on the surface wind of these stationary mesoscale features in the SST distribution.

Figure 6 shows two maps of the 10-m wind (labeled white lines) generated by the nonhydrostatic MC2 25-km run (run 2) superimposed over the detailed SST (gray scale). Figure 6a represents the situation ahead of the cold front. Warm air between 10° and 20°C was pushed by the southerlies over cooler seawaters. The magnitude of the wind field was shaped by the meanders of the Gulf Stream. Over the period, in a general south-southwesterly flow at 35–45 kt (18–23 m s−1), maximum wind speeds of 54–58 kt (28–30 m s−1) were found locally over the Gulf Stream’s meanders and the warm eddies. No real decrease in wind speed was noticed over the cold eddies, but the fact that these cold eddies were surrounded by warm meanders created a local minimum in the wind speed.

Figure 6b represents the situation behind the cold front, where a cold air mass at a temperature near 0°C was passing over warmer waters. Over the period, in a general westerly circulation of 35–45 kt (18–23 m s−1), maximum wind speeds of about 50–55 kt (26–28 m s−1) were found over warm eddies. However, the magnitude of the wind field was no longer shaped by the meanders of the Gulf Stream. The passage of the front changed the stability of the marine boundary layer.

Figure 7 shows the bulk Richardson number (gray scale) calculated in the first 100 m for the same time frame as in Fig. 6. The detailed SST (black lines) is superimposed over it. In Fig. 7a, ahead of the front, the bulk Richardson number indicated dynamically stable conditions in the lower levels of the atmosphere over the coldest waters. Over warmer waters, conditions were dynamically unstable with possibly more turbulent conditions over the warm eddies and over the meanders of the Gulf Stream, where statically unstable conditions prevailed. Over those areas, stronger vertical momentum transfer occurred, enhancing the wind speed locally. Over cold eddies, more stable conditions prevailed and diminished the vertical momentum transfer. Cold air advection could already be noticed on the left edge of the figure, where the lighter color indicates unstable conditions. In Fig. 7b, behind the cold front, cold air advection had uniformly destabilized the low levels of the atmosphere over the region. However, maximum destabilization (lowest values of bulk Richardson number) occurred over the meanders and the warm eddies coinciding with the maxima in wind speed (see Fig. 6b). The cold air destabilized the low-level layers of the atmosphere up to 500 m, allowing downward momentum transfer throughout the region, masking the influence of the Gulf Stream and the SST on the wind field. Most of the influence occurred over the Gulf Stream and the warm eddies because the coupled air mass and sea surface temperatures, taken together, favor weak stable to statically unstable conditions over these regions where a weak vertical wind shear was enough to sustain turbulence.

b. Characteristic length scale of the meanders

The following discussion is based on the studies of Glendening (1994) and Glendening and Doyle (1995) on different length scales involved in the interaction of synoptic forcing and air circulation due to SST meanders. The dependance of the size of a meander on the boundary layer response can be estimated by two main internal atmospheric scales: a mesoscale deformation radius and an alongfront advection length scale. For the sake of simplicity let us suppose that the SST distribution near the Gulf Stream is a mean linear SST front with superimposed meanders of varying size.

Strong SST gradients such as those encountered in the vicinity of the Gulf Stream generate local geostrophic adjustments (similar to sea breeze or ice-edge-coast thermally driven circulations) between their warmer and colder side boundary layers (BLs). The fundamental horizontal length scale governing the geostrophic adjustment caused by this differential temperature is called the mesoscale deformation radius R, which gives the horizontal extent influenced by self-induced circulation. It can be estimated, from Glendening and Doyle (1995), by
i1520-0493-126-11-2793-e3
where Dw is the BL depth over the warmer surface and N is the Brunt–Väisälä frequency given by Holton (1992).
The presence of large-scale forcing will also modify the BL response since thermally driven flow has dependence upon geostrophic wind. When the boundary layer friction component of the geostrophic wind optimally counterbalances the thermal flow, vertically averaged BL winds parallel the SST front and consequently sustain the strong gradients of different parameters such as temperature, vertical velocity, stability, and wind speed components. Such an optimal condition is found when the geostrophic forcing is pratically parallel to the SST front. The advective length scale Λ can be estimated by the product of the characteristic velocity along the front and the characteristic time over which this advection is significant (f−1). For simplicity, one uses the geostrophic component parallel to the front instead of the frictionally modified BL velocity; then from Glendening and Doyle (1995)
i1520-0493-126-11-2793-e4

With a meandering Gulf Stream, one needs to define the scale of the meander represented by L, which equals both the amplitude and the quarter-wavelength of the meander. The parameter indicating the dependence on alongfront variations, relating the scale of the meander to the scale of geostrophic adjustment, is therefore L/R. Here, L/Λ shows the importance of large-scale advection for along-SST front variations. In our case, with an estimated meander length scale of 100 km at 40°N, a warm-side BL depth Dw ∼ 800 m with a stability of N ∼ 2.0 × 10−2 s−1, we obtain R ∼ 50 km, and therefore, L/R ∼ 2. With an estimated southerly geostrophic wind speed of 70 kt (35 m s−1) mostly parallel to the mean SST front, we get Λ ∼ 370 km and, therefore, L/Λ ∼ 0.25. According to Glendening and Doyle (1995), for a meander of size L > R, without geostrophic forcing, the boundary layer should contain the signature of the meanders since there is no possiblility of alongfront blending, due to effective lateral interaction of the geostrophic adjustment. In the present case, R < L ≪ Λ indicates that the geostrophic advection is very dominant in the process and, referring to Glendening and Doyle (1995), it could be described as a complex intermediate-scale case where strong alongfront blending is sufficient to erase that meander signature. Why, then is the signature of the Gulf Stream, as revealed in the previous section, so clear in the present case?

In Glendening and Doyle (1995), strong sensible heat fluxes were present. In the present case, weak fluxes were present since there was a persistent southerly flow bringing warm temperatures almost in equilibrium with the sea surface temperature. This created the unstable to slightly statically stable conditions in the boundary layer. In such a condition, no direct thermal ciculation (like a sea breeze) was needed to redistribute a thermal imbalance, since the Gulf Stream did not act as a local heat source. The boundary layer instead responded by generating a weak convection in the lowest levels that produced stronger transfers of momentum, which resulted in an enhancement of the surface winds. Moreover, in response to these local maxima in the wind field, due to the persistence of the flow, a kinematic compensation was found. Surface dipoles of divergence and convergence, corresponding very well to similar dipoles of subsidence and ascendance in the vertical movement in the first few kilometers above the surface, were found in the vicinity of the Gulf Stream. However, this kinematic compensation was found to have no effect on the mass field.

In this present case study, the effect of the detailed SST on the surface wind field is therefore mainly explained by the weak stability of the shallow boundary layer, which allows stronger turbulent momentum transfers over the the Gulf Stream meanders and the warm eddies.

c. The influence of the eddies on the wind field

Results to this point have shown that the Gulf Stream influenced the wind speed. In particular, the meanders of the Gulf Stream seem to increase the surface wind speed locally by about 10 kt (5 m s−1). The eddies also influence, to a lesser extent and on a smaller scale, the marine boundary layer winds. Figure 8 shows the passage of the cold front over the warm eddy where the East Scotian Slope (44137) buoy was located (see Figs. 2 and 5). The figure is composed of the 10-m wind speed field (gray scale and black lines) from run 4 (see Table 2) superimposed over the detailed SST (labeled white lines). At 2000 UTC 14 March 1993 (Fig. 8a), the cold front was at the entry of the warm eddy. A synoptic slackening of the pressure gradient behind the front produced an elongated band of minimum values in the wind speed (hereinafter C1). A tongue of cooler water extending east of the warm eddy created more stable conditions, which generated weaker winds and formed another elongated band of minimum values in the wind speed (hereinafter C2). Finally, between C1 and C2, the warm eddy favored a maximum in the wind speed (hereinafter W) principally because of the less stable conditions prevailing over it. If one characterizes C1 as a frontal passage “tracer,” the sequence of images and a 6-h animation of the event create the illusion that the front never crossed the warm eddy but went around it by splitting in two. The northern part of C1 strengthened the existing C2 creating the effect that the synoptic slackening of the pressure gradient (observed behind the front) was now ahead of the front. The southern part of C1 slid south of the warm eddy. During the sequence, W gradually shrunk between C1 and C2 but never disappeared until the passage of the front. In reality, the cold front did pass over this part of the warm eddy. More turbulent conditions over it sustained higher wind speed values of about 10 kt (5 m s−1). In Fig. 8c, the front was located east of the warm eddy. The elongated band of minimum value in the wind speed (C1) created by the synoptic slackening of the pressure gradient re-formed behind the front. In the northeastern side of the region, the tongue of cooler water maintained weaker wind speed values ahead of the front. Although this is a single event, the last set of images shows that eddies also influence, on a smaller scale, the surface wind field.

d. Sensitivity tests

Now, using the filtered SST fields, one can see the influence of the real SST field on the generation of surface wind. As pointed out briefly at the end of section 3c, the filtering slightly cooled the southern waters and warmed up those located over the Scotian Shelf. Based on previous results, this cooling over the southern waters will diminish the maxima in wind speed that would be found over warm eddies and will slightly decrease the unstable westerlies behind the cold front. On the other hand, the warming of the northern waters will decrease the stability that would be found with normal SST and will slightly augment the unstable westerlies behind the cold front.

Figure 9 shows a superposition of the 10-m wind field resulting from runs 3, 3a, and 3b (see Table 2). Labeled lines are those corresponding to the same isotachs for runs 3a and 3b. The reader is strongly encouraged to refer to Fig. 6 for the location of the SST features causing these mesoscale variations in the wind field. Ahead of the front (see Fig. 9a), the strongest and most detailed wind field was generated by the S0 (gray scale) SST. The S1 (dashed lines) and S2 (thin lines) SST fields generated smoother wind fields. As can be seen, minor differences between the wind fields generated by various runs were noticed. The main differences were found along the meanders of the Gulf Stream. In a general southerly circulation at 40–45 kt (21–23 m s−1), a band of 50–55 kt (26–28 m s−1) was generated over the meanders by the S0 (run 3) SST field. This latter field slightly decreased the wind speed over the cold eddies when compared to those generated by the S1 and S2 SST fields. The simulation done with S1 (run 3a, which was similar to the RFE operational run) captured the large mesoscale features of the SST distribution such as the warm and cold eddies but never got the Gulf Stream’s meanders at the bottom of the figure. The simulation with S2 even lacked the maximum 52 kt (27 m s−1) on the right edge of Fig 9a generated by the warm eddy (see Fig. 6a), thus depicting only the synoptic surface wind field not affected by the mesoscale SST distribution. Behind the front (see Fig. 9b), winds generated by the various runs were again very similar overall. The simulation done with S0 and S1 caught the wind speed maxima generated by the cold air flowing over the northest Gulf Stream meander and the warm eddy where 44137 is located (see Fig. 6b). Simulation done with S2 SST (run 3b), produced lighter winds behind the front and again depicted the synoptic wind field as if the SST distribution had no mesoscale feature.

Since S2 SST generated the largest differences in the wind field when compared with the ones generated with S0, the remainder of the comparsions were done with the smoothest (S2) SST field. A subtraction of the wind speed at 10 m {UVS0UVS2} (not shown here) showed that the meanders of the Gulf Stream (S0) generated a ribbon of winds 8–11 kt (4–6 m s−1) stronger over it while warm core eddies locally increased the wind speed by 5–10 kt (2.5–5 m s−1) when compared with the simulation with no eddies (S2).

Although, all conclusions from this section were essentially based on model output, it revealed the strong influence of the detailed (S0) SST distribution, and particularly its mesoscale features, on the synoptic MABL winds. The strongest influence was found ahead of the front where local destabilization of the MABL was not hidden in the general unstable conditions occurring behind the front. Ten-knot (5 m s−1) stronger winds were found over the Gulf Stream’s meanders and over the warm eddies ahead of the front. In higher-resolution simulations, the local influence of the eddies was studied. It was shown that they modified the wind speeds only very locally.

6. Gridpoint data versus buoy data

a. Lower resolutions (50–10 km)

Up to this point, the results and conclusions with respect to the influence of the Gulf Stream on the surface wind field were based essentially on model output. In this section, the link is made between the model output at different resolutions and the buoy observations. Since the higher-resolution simulations were done over the waters south of the Atlantic provinces, we will principally concentrate our attention on the Canadians buoys (44137, 44138, 44139, 44141), more specifically on buoy 44137, which was chosen as the focal point in the cascade process (Fig. 2b). Buoy 44137 was, in general, representative of the situation at the other buoys (not shown here) and was the only one allowing comparisons between observations and simulations at all resolutions (see Table 2).

Figure 10 is one of eight sets of graphs each representing a buoy used in this study. Each set is composed of four graphs of time series of wind direction, wind speed, temperature, and pressure at sea level, respectively, of 1-h interval for observations and 3-h interval for simulations. Wind model output data were taken at 10 m while the temperature data were from the surface. Following Smith (1981), 5-m wind speeds from buoys were converted to a height of 10 m. This conversion added little to real buoy wind speed observations. No correction was made for pressure, temperature, or wind direction.

From an examination of the four graphs in Fig. 10, one can conclude that the 50-, 25-, and 10-km MC2 simulations closely matched the real situation. Numerical and observed directions of wind (see Fig. 10a) were consistent at all buoy sites. The air temperature from all simulations dropped less rapidly than those observed (see Fig. 10c), implying a less rapid destabilization of the marine boundary layer, and might have generated weaker winds in the numerical simulations. The numerical and the observed sea level pressure closely agreed except at the approach and after the passage of the cold front. Observed sea level pressures were on average 5 hPa higher (lower) than the simulated ones at the Canadian (American) buoys. Wind speed values from numerical outputs were, in general, more comparable to the observed maximum (highest running 8-s maximum over 10-min interval) at the Canadians buoys than the mean (10-min vector averages) winds. At the American buoys (41002, 44004, 44005, 44014), wind speed values from numerical outputs adopted, in general, the mean value instead of the maximum wind speed observed. Finally, as mentioned earlier (see section 4b), a delay of about 6 h between observations and numerical simulations in the minimum value of the pressure (see Fig. 10d) was also observed at each buoy site. This delay was reflected as well in the wind speed but was only about 4 h as can be seen in Fig. 10b.

From Fig. 10 [East Scotian Slope (buoy 44137)], the passage of the front at the buoy was indicated by a weak change in the wind direction (Fig. 10a), a minimum in the time series of the wind speed (Fig. 10b), a rapid falling of the temperature, and a rapid rising of the pressure. Most of the difference between observations and numerical simulations occurred near and after the passage of the cold front. Figure 10a shows that resolution did not have a strong influence on the wind direction. Numerical wind directions adopted a more southerly component than those observed. Figure 10c reveals that the marine boundary layer was always statically unstable at buoy 44137 except during 0600–1500 UTC 14 March 1993 when conditions were stable in reality but nearly neutral in the numerical simulations. Behind the front, the air temperature dropped rapidly, decreasing the stability of the marine boundary layer and allowing the observed maximum wind speed to remain high for a short period and record its largest value around 0000 UTC 15 March 1993 (see Fig. 10b). Finally, from Fig. 10d, it can be seen that the resolution did not have any impact on the pressure field and as expected did not diminish the delay of about 6 h between the observed and numerically simulated pressure.

b. Higher resolutions (10–2 km)

Higher-resolution runs, up to 2 km, were done to investigate if mesoscale structures would appear in the wind field in the vicinity of the front, and to see if the SST would still have a strong impact on the marine boundary layer, despite the fact that the SST analysis was digitized at 25 km.

Higher-resolution simulations were concentrated on the area of the East Scotian Slope (44137) buoy and output data were stored each hour. The reader can refer to Fig. 8 and concentrate on the southern part of the warm eddy where the buoy (encircled cross) is located to have an overview of the situation. It has been noted that the minimum of the wind speed caused by the slackening of the pressure gradient behind the front never crossed the southern part of the warm eddy. Because of that, buoy 44137 never recorded an abrupt decrease of the wind speed but rather a gradual one. This last fact combined with the location of the buoy in the axis of the tongue of cooler water also complicated the signature of the frontal passage at this buoy.

Figure 11 can be seen as an enlargement of Fig. 10b of the strong wind event period (0000 UTC 14 March–0600 UTC 15 March 1993). Time series of higher-resolution (10, 5, and 2 km) simulations are presented. Figure 11a contains data from runs 3, 4, and 5 done with the detailed SST (S0), while data in Fig. 11b come from runs 3b, 4b, and 5b where S2 was used throughout the entire cascade process. As previously noted, wind speed values from numerical simulations reflected more closely the observed maximum values. However, at buoy 44137, in the vicinity of the observed passage of the cold front (around 1700 UTC in Fig. 11a), they adopted a value closer to the observed mean wind speeds. The front itself preceded the minimum in the wind speed (around 1900 UTC) caused by the slackening of the pressure gradient behind it. It again reached the maximum observed wind speed values at the end of the series. Around 1300 UTC, there was a minimum in the observed wind speed that no high-resolution runs seem to reproduce. This minimum in the wind speed observed by the authors could be the result of the slackening of the synoptic pressure gradient, ahead of the cold front. The subsequent increase in the wind speed would be the low-level jet accompanying the cold front.

The “numerical” passage of the front occurred around 2100 UTC, creating a delay (similar to the one noticed in the pressure series) of 4 h between events from numerical simulations and the observations. By comparing an animation of the S0 SST run (run 5) at 2 km to Fig. 8 and shifting back the numerical series by 4 h (represented by the lower curves in Fig. 11a), it appears that the relative minimum in the observed wind speed, around 1500 UTC was generated by the tongue of cooler water (C2 in Fig. 8) penetrating the southern part of the warm eddy. As the front approached, it pushed slightly eastward the squeezed maximum in wind speed (W in Fig. 8) over the buoy area, explaining the rise in the wind speed. The passage of the front created a very narrow axis of wind speed minimum within the larger wind speed maximum persisting over the warm eddy. This second minimum around 1900 UTC for the observation and 2300 UTC for the 10-, 5-, and 2-km simulations series marked the frontal passage. In fact, the 2-km simulation (dotted square) best reflected the rapid 5-kt (2.5 m s−1) fall noticed in the observed time series at the passage of the front. This is evidence of a distinct response of the 2-km run to the higher-resolution SST.

After the passage of the front over the buoy, unstable and strong synoptic southwesterlies became established over the eddy and brought back stronger winds for a limited period of time. Although, buoy 44137 was located in a warm eddy, the minimum in the wind speed created by the tongue of cooler water just east of the warm eddy, and penetrating its southern part, could have generated more stable conditions at the buoy location. This could explain why numerical wind speed values were more squarely between maximum and mean wind speed values rather than closer to the maximum values noticed at the other Canadian buoys. This hypothesis appears more plausible when one considers the wind speed at buoy 44141 (not shown here), which was located right in the middle of a warm eddy; wind speed values were in the range of maximum observed values.

To summarize, because of the buoy location in the axis of the tongue of cooler water, the influence of the warm eddy was diminished and this caused weaker winds at the buoy site. Therefore, one can make the hypothesis, by referring to Fig. 8 and previous ones, that this caused a drop of 5 kt (2.5 m s−1) in the wind speed at the frontal passage instead of at least 10 kt (5 m s−1) if buoy 44137 would have been located outside of the warm eddy.

The time series shown in Fig. 11b strengthens the above hypothesis. First, the simulation done with the S2 SST generated weaker winds than those using the S0 SST. The main minimum in the wind speed occurred around 0000 UTC 15 March 1993 and marked the end of an important fall in the wind speed. Very high-resolution (2 and 5 km) runs generated a more important drop in the wind speed (about 10 kt or 5 m s−1 less) than the 10-km one (about 5 kt or 2.5 m s−1). This latter point reinforces the hypothesis that the presence of the warm eddy in runs done with S0 SST decreased the rapid reduction of the wind speed at the frontal passage and suggests the importance of the detailed SST field at very high resolutions in order to maintain a wind speed value closer to those observed.

7. Summary and conclusions

A study was made of the influence of the SST distribution on the generation of winds, particularly in the vicinity of the Gulf Stream’s meanders, using the blizzard of March 1993 (13–15 March 1993) as a laboratory. All model outputs were generated by the MC2 nonhydrostatic model nested at its lateral boundaries by the CMC analysis or by previous lower-resolution MC2 output. Despite a delay of about 6 h in the pressure field evolution between observations and numerical outputs (a delay, which certainly had an impact on the other fields), the 50-km MC2 outputs were considered as a valid and continuous representation of the atmosphere. Higher resolutions showed no major differences in the pressure and wind fields but increasingly more details in the surface wind speed field were generated at higher resolutions.

A study of the 25-km MC2 simulation at two different times, representing the situation ahead of and behind the front, revealed the influence of the Gulf Stream and its meanders on surface wind speeds. It was demonstrated that the stronger influence was ahead of the front, where the wind speed pattern was shaped by the meanders of the Gulf Stream. A study of the stability of the marine boundary layers demonstrated that statically or dynamically unstable conditions in the lower layers of the atmosphere prevailed in a shallow layer next to the surface, ahead of the front over warm eddies and meanders. This increased the vertical momentum transfer that locally enhanced the wind speed. The Gulf Stream’s meanders created a mesoscale wind field pattern because they locally destabilized the low level of the marine boundary layer when synoptic stable conditions prevailed. Momentum transfer was cut by stable conditions persisting over cold eddies. Behind the front, cold air above warmer seawaters destabilized the marine boundary layer throughout the region. This generalized destabilization masked the effect of the warm eddies and the influence of the Gulf Stream’s meanders on the surface winds. Winds were more uniform in the marine boundary layer and hid the effect of the mesoscale pattern in the SST field. The Gulf Stream no longer shaped the wind field pattern. However, the warmest spot of the meanders and the warm eddies, especially those located north of the Gulf Stream, locally enhanced the instability, allowing an increase in wind speed.

The study of the passage of the cold front at very high resolution, over one of the warm eddies, revealed the influence of the SST distribution on the marine boundary layer winds at very small scales. The low-level instability prevailing over that warm eddy substantially altered the well-defined band of weaker winds following the passage of the front. This created the illusion that the front jumped over the eddy instead of crossing it. The phenomenon was emphasized by the presence of a tongue of cooler water, which created a quasi-stationary minimum in the wind speed east of this warm eddy.

The influence of the resolution of the SST distribution was tested by making simulations with smoother SST fields. A comparison between winds generated by an SST without meanders and eddies and a real SST field revealed that the warm eddies together with the meanders of the Gulf Stream enhanced the wind speeds, by about 10 kt (5 m s−1) in a general flow of 40–45 kt (21–23 m s−1) ahead of the front. Weaker winds [by about 5 kt (2.5 m s−1)] prevailed over cold eddies. Behind the front, the warm eddies increased the instability of the marine boundary layer, locally enhancing the winds by about 10 kt (5 m s−1) in a general flow of 40–45 kt (21–23 m s−1).

A link was made between model outputs at various resolutions and buoy (44137) observations. Time series of different parameters such as wind direction, wind speed, temperature, and pressure, when compared with those observed, demonstrated that model outputs at all resolutions corresponded closely to the observations. The time series for higher-resolution runs explained the behavior of the observed wind speed in the time frame of the frontal passage and also revealed the importance of the SST field details for maintaining relatively strong winds at the passage of the front. The location of the buoy on the axis of the cooler water tongue penetrating the southern part of the warm eddy and the persisting instability prevailing over the warm eddy possibly prevented a drop of at least 10 kt (5 m s−1) at the passage of the front that would have been noticed if the buoy had been located outside of the warm eddy. Higher-resolution simulation with smoother SST fields reinforced this plausible hypothesis.

Although the conclusions from this study were based on only one case, the blizzard of March 1993, and it simply used the static stability to explain the influence of mesoscale SST features on the boundary layer wind field, the study can be seen as another case showing the response of the MABL to mesoscale SST distribution studied by many authors in the last decades. Furthermore, this study revealed that the signature of the Gulf Stream’s meanders in the MABL winds was persistent and stationary. This stationary mesoscale pattern along the meanders of the Gulf Stream ahead of a cold front and the stationary maxima of wind over warm eddies behind the front, will likely have an impact on the generation of ocean waves. As a preliminary conclusion of his present work to develop and to test a Canadian coupled atmosphere–ocean wave model system, the first author can observe the influence of the Gulf Stream’s meanders to a lesser degree on the wave model field for this same storm. Further studies will be done on the wave aspect, and one can anticipate that these mesoscale features in the wind field will generate a similar pattern in the ocean wave field that can be detected with a fine mesh of wave observations.

Acknowledgments

This research was funded by the federal Panel on Energy Research and Development and the Atmospheric Environment Service. The authors would like to thank Jim Abraham (MWC, Bedford, Nova Scotia) for his support and his confidence. The first author would like to thank Bridget Thomas (MWC) and Alan MacAfee (MWC) for his welcome computer support. Without the support and the help of Ralph Bigio (METOC, Halifax) the detailed SST would not exist. Many thanks to the MC2 community support team (Michel Desgagné and Pierre Pellerin, RPN, Dorval) and Yves Chartier (RPN) for their help and their patience. Finally, a big thank you to Jocelyn Mailhot (RPN) for his review and constructive comments on the manuscript.

REFERENCES

  • Anthes, R. A., 1983: Regional models of the atmosphere in middle latitudes. Mon. Wea. Rev.,111, 1306–1335.

  • Bélair, S., D.-L. Zhang, and J. Mailhot, 1994: Numerical prediction of the 10–11 June 1985 squall line with the Canadian Regional Finite-Element Model. Wea. Forecasting,9, 157–172.

  • Benoit, R., J. Côté, and J. Mailhot, 1989: Inclusion of a TKE boundary layer parameterization in the Canadian regional finite-element model. Mon. Wea. Rev.,117, 1726–1750.

  • ——, M. Desgagne, P. Pellerin, Y. Chartier, and S. Desjardins, 1997:The Canadian MC2: A semi-Lagrangian, semi-implicit wide-band atmospheric model suited for fine-scale process studies and simulation. Mon. Wea. Rev.,125, 2382–2415.

  • Cardone, V. J., R. E. Jensen, D. T. Resio, V. R. Swail, and A. T. Cox, 1996: Evaluation of contemporary ocean wave models in rare extreme events: “Halloween storm” of October 1991 and the“storm of the century” of March 1993. J. Atmos. Oceanic Technol.,13, 198–230.

  • Chouinard, C., J. Mailhot, H. Mitchell, A. Staniforth, and R. Hogue, 1994: The Canadian regional data assimilation system: Operational and research applications. Mon. Wea. Rev.,122, 1306–1325.

  • Delage, Y., 1988: The position of the lowest levels in the boundary layer of atmospheric circulation models. Atmos.–Ocean,26, 329–340.

  • Doyle, J. D., and T. T. Warner, 1993: The impact of the sea surface resolution on mesoscale coastal processes during GALE IOP 2. Mon. Wea. Rev.,121, 313–333.

  • Fritsch, J. M., and C. R. Chappell, 1980: Numerical prediction of convectively-driven mesoscale pressure systems. Part I: Convective parameterization. J. Atmos. Sci.,37, 1722–1733.

  • Glendening, J. W., 1994: Dependance of boundary-layer structure near ice-edge coastal front upon geostrophic wind direction. J. Geophys. Res.,99 (D), 5569–5581.

  • ——, and J. D. Doyle, 1995: Mesoscale response to a meandering surface temperature interface. J. Atmos. Sci.,52, 505–518.

  • Holton, J. R., 1992: An Introduction to Dynamic Meteorology. 3d ed. International Geophysics Series, Vol. 48, Academic Press, 523 pp.

  • Hsu, S. A., 1988: Coastal Meteorology. Academic Press, 260 pp.

  • Huo, S., D.-L. Zhang, J. Gyakum, and A. Staniforth, 1995: A diagnostic analysis of the superstorm of March 1993. Mon. Wea. Rev.,123, 1740–1761.

  • Kuo, H. L., 1974: Further studies on the parameterization of the influence of cumulus convection on large-scale flow. J. Atmos. Sci.,31, 1232–1240.

  • Low-Nam, S., and Y.-H. Kuo, 1991: The mesoscale structure of sea surface temperature and its impact on rapid cyclogenesis. Preprints, First Int. Symp. on Winter Storms, New Orleans, LA, Amer. Meteor. Soc., 426–430.

  • Mailhot, J., 1994: The Regional Finite-Element (RFE) model scientific description. Part 2: Physics. [Available from RPN, 2121 Trans-Canada, Dorval, PQ H9P 1J3, Canada.].

  • ——, and C. Chouinard, 1989: Numerical forecasts of explosive winter storms: Sensitivity experiments with a meso-α scale model. Mon. Wea. Rev.,117, 1311–1343.

  • ——, and Coauthors, 1995: Change to the Canadian regional forecast system: Description and evaluation of the 50-km version. Atmos.–Ocean,33, 55–80.

  • ——, R. Sarrazin, B. Bilodeau, N. Brunet, and G. Pellerin, 1997: Development of the 35-km version of the Canadian regional forecast system. Atmos.–Ocean,35, 1–28.

  • Nuss, W. A., and C. Lilly, 1989: Boundary layer convergence along the Gulf Stream during ERICA IOP2. Preprints, First Int. Symp. on Winter Storms, New Orleans, LA, Amer. Meteor. Soc., 423–425.

  • Perkey, D. J., W. M. Lapenta, F. R. Robertson, and C. W. Kreitzberg, 1991: The role of the sea surface temperature distribution on numerically simulated cyclogenesis during ERICA IOP-4. Preprints, First Int. Symp on Winter Storms, New Orleans, LA, Amer. Meteor. Soc., 418–422.

  • Shuman, F. G., 1957: Numerical methods in weather prediction: II. Smoothing and filtering. Mon. Wea. Rev.,85, 357–361.

  • Skey, S. G. P., K. Berger-North, and V. R. Swail, 1995: Detailed measurement of winds and waves in high sea states from a moored NOMAD weather buoy. Proc. 4th Int. Workshop on Wave Hindcasting and Forecasting, Banff, AB, Canada, Environment Canada, 213–223.

  • Smith, S. D., 1981: Factors for adjustment of wind speed over water to a 10 metre height. Rep. Series/BI-R-81-3/March 81, Bedford Institute of Oceanography, Bedford, NS, Canada, 29 pp. [Available from Bedford Institute of Oceanography, P.O. Box 10026, Darmouth, NS B2Y 4A2, Canada.].

  • Stommel, H., 1965: The Gulf Stream: A Physical and Dynamical Description. 2d ed. University of California Press, 248 pp.

  • Strong, A. E., and R. J. De Rycke, 1973: Ocean current monitoring employing a new satellite sensing technique. Science,182, 482–484.

  • Sweet, W., R. Fett, J. Kerling, and P. LaViolette, 1981: Air-sea interaction effects in the lower troposphere across the north wall of the Gulf Stream. Mon. Wea. Rev.,109, 1042–1052.

  • Tanguay, M., A. Robert, and R. Laprise, 1990: A semi-implicit semi-Lagrangian fully compressible regional forecast model. Mon. Wea. Rev.,118, 1970–1980.

  • Thomas, B., 1995: Use of an interactive graphical analysis system to hindcast the Storm of the Century, March 12–15, 1993. Proc. Fourth Int. Workshop on Wave Hindcasting and Forecasting, Banff, AB, Canada, Environment Canada, Downsview, 93–108.

  • Warner, T. T., M. N. Lakhtakia, J. D. Doyle, and R. A. Pearson, 1990:Marine atmospheric boundary layer circulations forced by Gulf Stream sea surface temperature gradients. Mon. Wea. Rev.,118, 309–323.

Fig. 1.
Fig. 1.

Synoptic surface situation from CMC analysis at 0000 UTC 14 Mar to 0000 UTC 15 Mar 1993. Solid lines represent pressure at sea level (4 hPa). Dashed lines represent 1000-hPa temperatures (5°C). Wind barbs are knots. Hours represented are (a) 0000 UTC 14 Mar, b) 1200 UTC 14 Mar, and (c) 0000 UTC 15 Mar.

Citation: Monthly Weather Review 126, 11; 10.1175/1520-0493(1998)126<2793:TIOMFO>2.0.CO;2

Fig. 2.
Fig. 2.

Grids for different resolutions used during the cascade process. Shaded areas correspond to the nesting buffer zone of each grid. In (a), cascade grids for 50, 25, and 10 km. In (b), cascade grids for 10, 5, and 2 km on the detailed SST (see Fig. 3a).

Citation: Monthly Weather Review 126, 11; 10.1175/1520-0493(1998)126<2793:TIOMFO>2.0.CO;2

Fig. 3.
Fig. 3.

Sea surface temperature (2°C) field at (a) 25 km from digitized METOC SST analysis (S0) valid from 0000 UTC 12 Mar to 2100 UTC 15 Mar 1993. In (b), S0 filtered 2500 times (S2)

Citation: Monthly Weather Review 126, 11; 10.1175/1520-0493(1998)126<2793:TIOMFO>2.0.CO;2

Fig. 4.
Fig. 4.

Difference (2°C, gray scale) between the detailed SST and the smoothest filtered SST (S0 − S2) superimposed over S0 (2°C, black lines).

Citation: Monthly Weather Review 126, 11; 10.1175/1520-0493(1998)126<2793:TIOMFO>2.0.CO;2

Fig. 5.
Fig. 5.

Location of the buoys in relation to the detailed 25-km SST field.

Citation: Monthly Weather Review 126, 11; 10.1175/1520-0493(1998)126<2793:TIOMFO>2.0.CO;2

Fig. 6.
Fig. 6.

Wind speed at 10 m (5 kt or 2.5 m s−1, labeled white lines) from run 2, superimposed over the detailed SST field (2°C, gray scale). Hours represented are (a) 1200 UTC 14 Mar 1993, corresponding to a 21-h simulation time, and (b) 0000 UTC 15 Mar 1993, corresponding to a 33-h simulation time.

Citation: Monthly Weather Review 126, 11; 10.1175/1520-0493(1998)126<2793:TIOMFO>2.0.CO;2

Fig. 7.
Fig. 7.

Bulk Richardson number (gray scale) in the first 100 m from run 2 superimposed over the SST field (2°C, black). Hours represented are the same as in Fig. 6. Gray scale goes from light (unstable) to dark (stable).

Citation: Monthly Weather Review 126, 11; 10.1175/1520-0493(1998)126<2793:TIOMFO>2.0.CO;2

Fig. 8.
Fig. 8.

Wind field at 10 m (2 kt or 1 m s−1, gray scale and black lines) from run 4 superimposed over the detailed SST field (2°C, labeled white). Hours represented are (a) 2000 UTC 14 Mar, (b) 2200 UTC 14 Mar, and (c) 0000 UTC 15 Mar 1993 being, respectively, 5-, 7-, and 9-h simulation times. The W refers to a maximum created by the warm eddy and C1 and C2 refer to two minimums in the wind speed, created by the cold front and a tongue of cooler water. The location of the East Scotian Slope buoy (44137) is indicated by the encircled cross. Refer to Fig. 2b for the location of the area.

Citation: Monthly Weather Review 126, 11; 10.1175/1520-0493(1998)126<2793:TIOMFO>2.0.CO;2

Fig. 9.
Fig. 9.

Wind fields at 10 m (5 kt or 2.5 m s−1) from run 3 (S0, gray scale), run 3a (S1, dashed black), and run 3b (S2, thin solid). Hours represented are the same as in Fig. 6 but, respectively, correspond to (a) a 12-h simulation time and (b) a 24-h simulation time. Labeled lines are those corresponding to the same isotachs for runs 3a and 3b. Cold front position is from run 3.

Citation: Monthly Weather Review 126, 11; 10.1175/1520-0493(1998)126<2793:TIOMFO>2.0.CO;2

Fig. 10.
Fig. 10.

Gridpoint data from runs 1, 2, and 3, representing East Scotian Slope (44137) buoy with the buoy observations: In (a) 10-m wind direction, in (b) 10-m wind speed, in (c) air surface temperature and sea surface temperature, and in (d) sea level pressure. For the definition of lines, see legend in graph.

Citation: Monthly Weather Review 126, 11; 10.1175/1520-0493(1998)126<2793:TIOMFO>2.0.CO;2

Fig. 11.
Fig. 11.

Same as Fig. 10b but in (a) from run 3, 4, and 5 (S0) and in (b) from runs 3b, 4b, and 5b (S2). For the definition of lines, see legend in graph. Four-hour shifted series are represented by lower series in (a). Note that 10 k (or 5 m s−1) have been subtracted to split the series.

Citation: Monthly Weather Review 126, 11; 10.1175/1520-0493(1998)126<2793:TIOMFO>2.0.CO;2

Table 1.

Specification of the cascade process. Time offset is relative to 50-km simulation and duration column is the length of the integration. Here, f is the time interval for interpolating the lateral boundary fields.

Table 1.
Table 2.

Summary of the experiments performed as a function of Δx and SST selected.

Table 2.
Table 3.

Specifications of the buoys used. The last column indicates the highest resolution at which a given buoy was used.

Table 3.
Save