1. Introduction
Climatologies of Antarctic mesocyclones (MCs), which are mainly based on satellite imagery, show that MCs at the meso-α (200–2000 km) and the meso-β scale (20–200 km) occur frequently in the coastal regions of Antarctica as documented, for example, by climatological studies for the Ross Sea (Carrasco and Bromwich 1994) and the Weddell Sea (Heinemann 1990; Carrasco et al. 1997; Turner et al. 1996) regions of Antarctica (Figs. 1 and 2). Although these systems are generally not as intense as the sometimes destructive polar lows of the Northern Hemisphere, they can still complicate logistical operations (e.g., aviation) in these areas and could be important in a climatological sense, for example, by increasing the exchange of momentum and heat between polar and subpolar regions.
For a few individual cases of coastal MCs, detailed investigations using observational data are present in the literature. In the eastern Weddell Sea region (EWSR; see Fig. 1) a coastal polynya is present during the austral summer, and MCs are frequently observed along the coast between the stations Halley and Georg von Neumayer. Available observational data indicate that many of the MCs on the meso-β scale have lifetimes of about 12 h and are associated with wind speeds of about 10 m s−1 [e.g., the coastal MCs studied by Heinemann (1996)]. However, some larger systems (meso-α scale) were found to be more intense. Turner et al. (1993) investigate the case of a vigorous coastal MC, which developed inland of the Brunt Ice Shelf in the Weddell Sea region and caused severe weather with winds of gale force and heavy snowfall. For the Ross Sea area, Carrasco and Bromwich (1993, 1995) investigated series of MCs developing in the vicinity of the Antarctic stations Terra Nova Bay (74.7°S, 164.1°E) and McMurdo (77.9°S, 166.7°E; see Fig. 2 for positions) using observational data and hand-drawn analyses. In these areas, large valleys are present, giving rise to intense katabatic winds, and the water close to the ice slopes is ice free during the austral summer. These facts as well as the observational results of Bromwich (1991), Carrasco and Bromwich (1993, 1995), and Heinemann (1996) suggest that the vortex stretching of the downslope flow is an important mechanism for the development of coastal MCs in the Antarctic.
First attempts to simulate topographically forced MCs in the Antarctic for realistic cases are presented in Engels and Heinemann (1996) for the Weddell Sea region using the Norwegian limited area model (NORLAM), a mesoscale numerical weather forecast model, for the cases described by Heinemann (1996) and Turner et al. (1993). Idealized simulations with NORLAM using idealized synoptic forcings for the same region (Klein and Heinemann 2001) strongly suggest that the coastal polynya and the katabatic flow over the ice slopes are important for the genesis of these MCs. In addition to synoptic support of the katabatic wind system and vortex stretching of the downslope flow, the transport of cold air over the open water of the coastal polynya is found to play an important role in the coastal mesocyclogenesis through the convergence of the sensible and latent heat flux in the boundary layer and subsequent reduction of the static stability in the lower troposphere.
A topographically forced formation of mesoscale circulations for the Ross Sea region is shown by the idealized simulations of Gallée (1995) using a hydrostatic mesoscale model with 10-km grid size and a state of rest as initial conditions. Mesocyclones of about 100-km size associated with a pressure anomaly of about 2 hPa are simulated in the Terra Nova Bay. The specific topographic structure of the Terra Nova Bay is found to be important for the mesocyclogenesis, since it causes a cyclonic shear of two different katabatic airstreams.
Mesocyclones and strong katabatic winds in the Antarctic can be a severe problem for logistic operations and they still present a very difficult task for operational forecasting. In the present study, different MC events are investigated using the mesoscale model NORLAM, in order to elucidate the forcing mechanisms of different types of MCs in both the EWSR and the western Ross Shelf region (WRSR), and in order to identify problems of operational forecasting in the Antarctic.
2. Model setup and design of the studies
The limited area model (LAM) used for the simulations of Antarctic MCs is the former operational model NORLAM (version 9) of the Norwegian Meteorological Institute (DNMI) at Oslo. A detailed description of the model can be found in Grønås and Hellevik (1982), while information on the model's physics and on parameterization schemes is contained in Nordeng (1986). NORLAM has been widely applied to the investigation of MCs and katabatic winds in polar regions (e.g., Grønås et al. 1987; Nordeng 1990; Heinemann 1997, 1998). The model domain consists of 121 × 97 grid points, while a vertical resolution of 30 σ levels is used with about half of the levels located below 850 hPa. A summary of the main model characteristics is listed in Table 1.
Idealized and realistic simulations have been carried out. A nesting mode is used for both types of simulations. For the idealized simulations a first run using a grid spacing of 50 km (LAM50, model domain 6000 km × 5000 km) is performed with analytic fields as initial conditions and constant boundaries. For the simulations of real cases, initial and boundary conditions are provided by European Centre for Medium-Range Weather Forecasts (ECMWF) analyses. A second integration with a 25-km grid (LAM25, model domain 3000 km × 2500 km) is then nested in the LAM50 results starting with the LAM50 0-h initialized analysis. Table 2 contains a list of the datasets used for the simulations. The model domain of the LAM50, which is used for all Weddell Sea simulations, is displayed in Fig. 1. The domain of the LAM25 is indicated by the rectangle and comprises a part of the Antarctic Peninsula and the complete Weddell Sea region. In Fig. 2 the model domains of the LAM50 and LAM25 for the Ross Sea area are shown.
3. Simulations of Weddell Sea MCs
In this section, the results of NORLAM simulations for three cases of mesocyclogenesis in the Weddell Sea region are presented. The first example is the case of a mesoscale vortex on the meso-β scale, which formed over the ice slopes of the EWSR near Halley station. Such vortices are frequent during austral summer with about three systems a week (Heinemann 1990). The second event described in this work is the case of a larger (meso-α scale) MC at Halley. These systems occur less frequently (about one system a week) in the EWSR than the smaller systems, but can be more intense and severe. In addition, an example of a larger (meso-α scale) MC is presented, which formed over the Filchner/Ronne Ice Shelf, where MCs are less often detected than in the EWSR.
a. The Weddell Sea MC case of 17 February 1989
Figure 3 shows the Advanced Very High Resolution Radiometer (AVHRR) visible satellite image for 0727 UTC 17 February 1989. A synoptic-scale cyclone can be seen with its center near 67°S, 10°W, and the eastern part of the Weddell Sea is almost cloud free. The cyclonic cloud pattern of an MC is present near 75°S, 25°W, which is close to Halley station. Figure 4 shows a blowup of that area for the AVHRR infrared image at 0727 UTC 17 February 1989. The MC has a scale of less than 300 km and is associated with a cloud pattern spiraled in. The brightness temperatures indicate very shallow clouds being relatively warm compared to the ice surfaces. At the time of the satellite image (0727 UTC), the center of the MC is located directly at the coast. However, earlier satellite images show that the MC developed over the ice slopes east of Halley during the night of 16–17 February 1989. It then crossed the Brunt Ice Shelf, moved over the open water and dissipated after about half a day (not shown). Observations at Halley station at 0600 UTC show a wind speed of 6.5 m s−1 from 90°, while a drifting buoy north of Halley (at about 73.8°S, 25.1°W; see Fig. 4) measured 3.5 m s−1 from 204° at 0624 UTC, thus indicating the weak cyclonic flow pattern associated with the MC.
The synoptic situation leading to the MC development (1800 UTC 16 February 1989) is displayed in Fig. 5 as the 18-h mean sea level pressure (MSLP) forecast of LAM25. The synoptic-scale cyclone depicted by the satellite image (Fig. 3) is simulated by LAM25. In the coastal areas of the EWSR, easterly synoptic flow from the ice slopes in the direction of the open water is present. As shown in climatological studies (Heinemann 1990), this situation is favorable for MC developments in the EWSR.
In the NORLAM simulation, the MC forms during the first hours of 17 February over the ice slopes (Fig. 6), which is in agreement with satellite observations. Wind speeds of about 10 m s−1 are simulated over the slopes along the eastern coast. However, in contrast to the observations, which showed a westward movement of the MC, no propagation is simulated by NORLAM, and the MC broadens and dissipates over the ice after about 1 day. As a result of the missing propagation of the modeled MC, the agreement with the observations at Halley station is relatively poor. Comparisons to measurements of the drifting buoy north of Halley (see Fig. 4), however, show a relatively good agreement, revealing that the synoptic conditions are relatively well captured by the NORLAM forecast. While the simulated wind speeds are slightly too high (partly due to the difference in height between model and observations), the veering of the wind from easterly to southwesterly directions during the time of the MC development is simulated well for the position of the buoy (not shown). In contrast to the satellite image (Fig. 4) no cloud signature is associated with the MC in the model simulations. Possible explanations for the failure of the model to simulate these shallow clouds are weaknesses in the parameterization of cloud physics, but also too dry initial ECMWF moisture analyses.
Figure 7 shows the results of a time series of the vorticity budget terms at 950 hPa for a box (mean over nine grid points) located in the MC center. As a result of the convergence over the slopes (DIV), the vorticity tendency is negative after 21 h; that is, cyclonic vorticity is being produced. Diffusion and tilting of the vortex compensate this production after 30 h; that is, the simulated MC shown in Fig. 6 has reached its maximum cyclonic vorticity. The horizontal structure of the stretching term at 850 hPa is indicated in Fig. 6. The maximum convergence at this level is present over the ice slopes.
A cross section through the center of the MC after 30 h is shown in Fig. 8. A pronounced katabatic flow is present in the lowest 200 m over the ice slopes. The convergence found at the bottom of the slope (see Fig. 6) is also indicated by the wind vectors in the plane of the cross section. The signal of the MC can be seen in the sign reverse of the wind speed normal to the plane. The simulated MC is very shallow with a height below 1000 m. A belt of large potential vorticity (PV) is present over the slopes (indicated by the shading), which is a result of the horizontal wind shear at the top of the katabatic layer.
The development of the MC on 17 February 1989 appears to be directly connected to the vertical stretching of the synoptically supported downflow over the ice slopes. Even though the further development/movement of the MC is not captured, the results of the NORLAM simulation suggest that the convergence of the katabatic flow in connection with a suitable synoptic background is one essential mechanism for the generation of coastal MCs in the EWSR. The case of 17 February 1989 is one of the rare EWSR cases, where at least a weak signal of the MC can be found in the model results. Simulations for several other cases of meso-β-scale MCs in that area were less successful. This shows that these obviously orographically forced MCs are generally very difficult to capture by mesoscale models. This difficulty could be a result of the oversimplification of the orographical structures in the model or reflect the problem of a too low resolution. However, the NORLAM simulation of the present case confirms that mesoscale models with grid spacings of about 25 km are in principle able to capture such events. A general problem for the simulations could be the sparsity of observational data for the Antarctic. Since the development of these MCs obviously strongly relies on additional synoptic support, a very good quality of the analyses used as forcing for the mesoscale model is a prerequisite for successful MC simulations.
b. A subsynoptic vortex near Halley Station
During the period of 28–30 January 1990, a pronounced meso-α-scale MC (diameter about 1000 km) was present with its center near Halley station (Fig. 9). Harsh weather conditions (heavy snowfall and wind speeds of up to 12 m s−1) were associated with the cyclone. The NORLAM simulations capture the basic development relatively well although underestimating the strong deepening of the low at Halley by about 5–10 hPa. After about noon on 30 January, the cyclone moved northward and dissipated.
In satellite images, the first sign of the MC is visible in the morning of 27 January slightly northeast of Berkner Island of the Filchner/Ronne Ice Shelf (marked in Fig. 10). From there the system moved northeastward in the direction of Halley. After an eventual weakening during that movement phase, a significant intensification took place in the morning of 28 January. A preexisting cloud band associated with a large-scale frontal system was incorporated in the circulation of the MC (Fig. 9). The NORLAM-simulated relative humidity field and wind vectors at 850 hPa of the 18-h LAM25 forecast (Fig. 10) agree well with the cloud structure visible in the satellite image (Fig. 9). At the western side of the MC, cold air is advected from the Filchner/Ronne Ice Shelf over the sea ice of the western Weddell Sea and the coastal polynya.
For the development of the Halley MC, upper-level forcing appears to be important. On 26 January, an upper-level cyclone is present at the relatively high latitude of 80°S (Fig. 11) associated with a trough extending over the southern part of the Weddell Sea. With the northward movement of this upper-level trough, cold air is advected into the region between Berkner Island and Halley. The upper-level support is clearly identifiable by the track of a PV anomaly at 500 hPa (Fig. 11).
While the main development of the MC can be explained by “classical” baroclinic forcing, an orographic forcing appears to be present for the initial stage. The low-level wind field at 0600 UTC 26 January is strongly influenced by the orography (Fig. 12). Katabatic flows around Berkner Island and in the eastern part of the Filchner/Ronne Ice Shelf lead to cyclonic vortices (M1, M2, and M3) over the ice shelf and the sea ice. Considering the initial phase of the Halley MC, the development starts during this time in an area northeast of Berkner Island, and M3 represents the initial stage of the Halley MC. Preexisting low-level baroclinicity in the vicinity of the sea ice edge as well as the convergence of the near-surface flow from the continental ice slopes with the larger-scale flow over and around Berkner Island seem to provide an initial low-level perturbation, which then amplifies as the result of the upper-level forcing. The track of the Halley MC during 26–28 January leading from the vicinity of Berkner Island to Halley station is indicated in Fig. 10.
At a later stage (0000 UTC 29 January), low pressure in the region east of Berkner Island additionally forces the katabatic flow over the ice slopes, leading to several katabatic surges (see Bromwich 1989) and the intensification of M1 and M2 (Fig. 13). Such low-level vortices at the bottom of the ice slopes are also present in the simulation results for other cases (not shown). These vortices are shallow and do not show associated cloud patterns. No sign of M1–M3 in that area is present in the available satellite images, and it cannot be determined, if these systems are real because of missing observational data. Yet, such MCs are consistent with the special orographical characteristics of that particular area and are likely to be relatively stationary phenomena for conditions of a well-developed katabatic wind system. Carrasco and Bromwich (1994) also found that many MCs over the Ross Ice Shelf do not develop a cloud signature.
A satellite image of this later stage of the Halley MC on 29 January is displayed in Fig. 14 with the NORLAM-simulated MSLP field superimposed, showing the overall good agreement of simulation and observation. At that stage, the MC is already relatively large with a diameter of about 1000 km.
The case of this MC event is an example of classical baroclinic MC development. An existing low-level perturbation intensifies in association with the arrival of an upper-level PV anomaly as described by quasigeostrophic theory. For the initial low-level development, convergence of the near-surface flow from the ice slopes east of Berkner Island seems to be important.
c. A subsynoptic vortex over the Filchner/Ronne Ice Shelf
From 11 to 14 February 1990, a subsynoptic vortex with a diameter of about 600–1000 km was present over the Filchner/Ronne Ice Shelf (FRIS). Mesocyclone developments detectable in satellite imagery are rare in that part of Antarctica (Heinemann 1990). Figure 15 shows a satellite image at 1950 UTC 11 February, which is still the early stage of the MC. The MC discussed here formed in the morning of 11 February southwest of Berkner Island over the FRIS and remained quasi-stationary until the evening of 12 February 1990, when it started to move westward. Within the next day, the MC moved northward along the southern part of the Antarctic Peninsula and reached the ice front of the FRIS northwest of Berkner Island sometime on 14 February 1990. After an intermediate reintensification over the open water, the low broadened and dissipated during 15 February 1990.
The 30-h LAM25 forecast valid at 1200 UTC 11 February 1990 (Fig. 16) shows low-level winds (about 30 m above ground) that agree well with the circulation indicated by the cloud structures in the satellite imagery and show a mesoscale vortex (SSC) southwest of Berkner Island. The vortex develops at a low-level trough axis in the vicinity of a large-scale frontal zone. This front results from the flow pattern associated with the large-scale trough, that is, significant warm air advection from the open water north of Berkner Island and cold air advection from the slopes around the Filchner/Ronne Ice Shelf. Strong wind speeds of up to 17 m s−1 are present over the slopes southeast of Berkner Island, leading to an additional cold air transport toward the center of the vortex.
The low-level perturbation southwest of Berkner Island intensifies with the approach of an upper-level PV anomaly associated with a short baroclinic wave. The 500-hPa geopotential heights and temperature fields as well as the PV (shading) are shown in Fig. 17. As in the case of the MC at Halley station discussed in section 3b, the upper-level support is important for the MC development. The long lifetime and the movement of the MC seem to be a result of the northward shifting of a 500-hPa cyclonic PV anomaly from about the South Pole toward the center of the FRIS.
The complex movement of the MC toward the open water during this later stage is partially reproduced in NORLAM simulations started on 12 and 13 February 1990, respectively (not shown). However, since the initial development phase south of Berkner Island is not captured in these simulations, only a subsynoptic trough, instead of a closed circulation, develops, which then reaches the open water. Furthermore, the simulated trough moves too fast in the simulation of 12 February 1990, therefore reaching the open water about 12 h earlier than the observed MC.
Comparing the case of the MC over the FRIS with the case of the MC at Halley discussed in section 3b, the question arises of whether the different frequencies of such systems in these two areas are only a result of the differences in detectability. Considering the simulation result that upper-level forcing is very important for these larger systems, it can be speculated that the different availability of suitable upper-level support could lead to real differences in the frequency of MCs in these two areas. While the FRIS is located in a relatively sheltered position at high latitudes, the Halley region is located closer to the main track of synoptic cyclones. Generally synoptic cyclones rather tend to follow a zonal path around the Antarctic continent than to enter the interior of the ice sheet. A southward movement of synoptic cyclones over the ice sheet implies a vertical shrinking of the vortices and subsequently a reduction of cyclonic vorticity. Thus, the different frequency of the necessary synoptic upper-level forcing could also account for the rareness of intense meso-α MCs over the FRIS compared to developments in the vicinity of Halley station in the EWSR. On the other hand, smaller- (meso-β-) scale MCs may also frequently form over the Filchner/Ronne Ice Shelf due to the topographic forcing, although these MCs cannot be detected by satellite imagery.
4. Simulations of Ross Sea MCs
In this section, an idealized simulation and two case studies of the interaction of katabatic winds and MCs in the areas of Terra Nova Bay and McMurdo are presented (see Fig. 2).
a. Idealized simulation
In order to investigate the forcing of the surface on the boundary layer flow in the WRSR, a model run without synoptic forcing was performed. For the idealized simulation presented in this section, an atmosphere at rest with a single vertical profile of the potential temperature and relative humidity is used as the initial state. This horizontally homogeneous distribution changes during the integration due to atmosphere–surface exchange and other boundary layer processes. The most pronounced effect is the development of katabatic winds over the slopes. In Fig. 18a, the results of the idealized simulation are shown after 24 h of integration using surface and radiation conditions for 18 February 1988. Wind vectors and potential temperature at the lowest NORLAM σ level (about 30 m above ground) are displayed. Already after 24 h, strong katabatic winds can be seen with wind speeds of up to 16 m s−1. The potential temperature field reveals frontal structures along the bottom of the slopes and the transition zones from ice or ice shelf to the open water.
The results after 60 h of integration are displayed in Fig. 18b. The shading of the potential temperature indicates the path of the cold air, which mainly descends through two large valleys, Reeves/David Glacier (RDG) near Terra Nova Bay in the northern part and Mulock Glacier (MG, southwest of McMurdo) in the southern part of the WRSR, respectively. The two tongues of cold air extend from the higher slopes through the two valleys over the open water. The more southern airflow passes south of McMurdo and Ross Island, thereby crossing the northwestern part of the Ross Ice Shelf. Between these two cold airstreams, relatively warm air with values of almost 270 K is present at the lowest NORLAM σ level. The structure of these streams together already yields a cyclonic pattern. This kind of mesocyclogenesis indicates that the specific topographic structure of the WRSR is favorable for mesocyclogenesis. The results of this idealized simulation are in agreement with simulation results of Gallée (1995) for the same area, who uses a similar model setup (atmosphere at rest). While Gallée (1995) assumes ice-free conditions for the whole Ross Sea, the present simulation uses realistic ice conditions valid for 18 February 1988.
b. The Ross Sea MC event of 18 February 1988
On 18 February 1988, an MC was observed in the Ross Sea area between the station Terra Nova Bay and Ross Island. This case was already investigated by Carrasco and Bromwich (1993) using observational data. Figure 19 shows the AVHRR infrared image at 1302 UTC 18 February 1988. The MC M1 between Terra Nova Bay and Ross Island (RI) can easily be identified by its spiral cloud pattern, and its position agrees with the MC of the idealized simulation. The MC developed on the rear side of an eastward moving synoptic cyclone, whose associated pressure gradient on its western side led to a period with significantly intensified katabatic winds in the vicinity of Terra Nova Bay. The cloud band associated with the MC also extends over the northern part of the RIS, but the inland ice in the Terra Nova Bay region is almost cloud free (clearly depicted by the AVHRR imagery in the water vapour channel, not shown).
In Fig. 20 the wind vectors, potential temperature, and relative humidity at 925 hPa of the 18-h NORLAM forecast are displayed (valid at 1800 UTC 18 February 1988). High humidity (likely associated with cloudy areas) is indicated by the shading of areas with values larger than 90%. A clear signal of the MC M1 can be seen in the simulation results, where a spiral cloud pattern and a vortex in the wind field are visible. Since the simulation was started at 0000 UTC on 18 February, the model forecast is valid 5 h later than the satellite image. Yet, the modeled MC is shifted toward the orography by about 50–100 km compared to the real MC. During the further integration, the modeled cyclone moves to the east and merges with a weaker trough south of Ross Island, thereby intensifying, broadening, and finally becoming a larger subsynoptic cyclone, which is in agreement with the observations as discussed in Carrasco and Bromwich (1993). In addition, a synoptic cyclone enters the northern part of the Ross Sea area, partially merging with the MC. An additional MC development takes place during the night from 18 to 19 February 1988 in the southeastern part of the Ross Ice Shelf (not shown).
Figure 21 shows the wind vectors and the potential temperature at the lowest NORLAM σ level (approximately 30 m above ground) after 18 h, where the mesoscale vortex is indicated by M1. A cold airstream is visible at the northern part of M1, while a low-level baroclinic zone with a horizontal gradient of 8 K 100 km−1 is present near the ice edge. A second cold airstream can be seen south of McMurdo, extending over the ice shelf. An analysis of the vertical component of the vorticity budget equation on pressure surfaces shows that the vortex stretching is the dominating process for the production of cyclonic vorticity during the first 18 h of the simulation (not displayed). The vortex stretching of the descending airflow in the valley region near Terra Nova Bay is therefore of great importance for the MC development.
Model runs started earlier than 0000 UTC 18 February give less satisfying results for this MC event. This might indicate that the synoptic forcing necessary for the development of the MC is not contained in the synoptic analyses prior to that date. The effect of the synoptic forcing on the MC development is also essential. The high values of the stretching term are not a result of the stretching of the katabatic drainage flow alone. An approaching synoptic low at 700 hPa provides additional relative vorticity for the generation of the MC by vortex stretching. In Fig. 22, a 15-h forecast of the geopotential height at 700 hPa is shown. Additionally, the potential vorticity at the same level is depicted by the shading of areas with values less than −1 PVU (potential vorticity unit, 1 PVU = 10−6 m2 s−1 kg−1 K). A tongue of potential vorticity extends from the low at 700 hPa into the genesis area of the MC. The flow associated with this system enhances the katabatic downflow in that area and leads to the katabatic surge at Terra Nova Bay. It can therefore be concluded that for the MC case of 18 February 1988, a combination of the katabatic wind and the synoptic forcing leads to the generation of the MC.
1) Sensitivity studies
Sensitivity studies have been performed for this case, in which several individual parameters and processes were changed compared to the simulation described above (control run). In the first sensitivity study, the diabatic heating due to the convergence of the sensible and latent heat fluxes was switched off. This reduces the strength of the katabatic wind over the slopes slightly, since the air is only cooled by the divergence of the net radiation. The main effect, however, is the prevention of diabatic effects at the area of open water close to the coast. The results are very similar to the results of the control run, which indicates that the diabatic heating of the cold air over the open water is not of prime importance for this case. In a second sensitivity test, the model was run with full physics, but the surface parameters were changed. For that run, the whole Ross Sea was covered by ice. This is another way of investigating the importance of the area of open water for the MC case of February 1988. Again, the results were basically the same as in the control run. In a third sensitivity simulation, the latent heat release by cloud physics was switched off in the model. Also the release of latent heat turned out to be of only minor importance. In the last sensitivity study, the orography of the model domain was set to zero at all grid points of the domain; that is, there were no longer any mountains in the model domain in that study. In the run without orography, no MC developed. Therefore, this last study clearly reflects the relevance of the orography of the Ross Sea area for the mesocyclogenesis.
c. The Ross Sea MC event of 7 January 1988
The observational data for this case is documented in Carrasco and Bromwich (1995), and their results will be used for comparison and as reference. A series of MCs developed in the period from 7 to 10 January 1988 in the WRSR, with several MCs merging and forming a larger subsynoptic system.
Figure 23 shows a regional analysis of the sea level pressure for 7 January 1988 taken from Carrasco and Bromwich (1995) and model results of a NORLAM simulation (both valid at 1800 UTC 7 January 1988). Two MCs, L1 and L2, can be seen as well in the regional analysis as in the model results. The surface pressure analysis was performed using automatic weather station (AWS) data. In the AVHRR satellite imagery for 1716 UTC (not shown) no cloud structures associated with the MCs could be detected. Particularly, L1 seems to be a vortex mainly forced by the low-level flow field in the Terra Nova Bay. This indicates again that satellite-based climatologies might miss MCs, if the vortex dynamics does not result in cloud formation (see Carrasco and Bromwich 1994). As in the case of 18 February 1988, the agreement of the simulation and the observations during the initial stage of the MC events is very satisfying.
At later stages, however, the agreement of observation and simulation results is very poor. This appears to be a result of a too strong high pressure system over the Ross Ice Shelf in the ECMWF analyses, which prevents the merging of the simulated MCs and leads to a completely inaccurate forecast of the further development. NORLAM forecasts started on 8 and 9 January 1988, respectively, underestimate the development and show a subsynoptic trough instead of a closed circulation. This could be a result of the missing initial phase during 7 January. The results of these NORLAM simulations underline again that it is important to capture both the initial phase, which is strongly influenced by the orography and the katabatic wind system in the WRSR, and the synoptic forcing at all stages in order to obtain a successful MC forecast.
Figure 24 shows the wind field at pressure level 925 hPa after 24 h (valid at 0000 UTC 8 January 1988). The simulated MCs L1 and L2 can be seen as two pronounced circulations, and a cyclonic cloud pattern associated with L2 is simulated. Although some low-level clouds were present over the RIS according to the satellite imagery, the model seems to overestimate the cloud formation associated with L2. At the lowest NORLAM level, one easily recognizes the familiar cold airstreams through the valleys near Terra Nova Bay and McMurdo (Fig. 25, katabatic flows K1, K2a, and K2b). Besides the low-level baroclinicity, it is again the stretching especially in the valley regions of Reeves/David Glacier, Mulock Glacier, and Byrd Glacier (see Fig. 25), enhanced by a descending upper-level (around 700 hPa) low (not shown), that seems to be important for the development of the two MCs, L1 and L2. The comparably poor results at later stages indicate again that one major problem of the modeling of Antarctic MCs and Antarctic weather forecasting is the synoptic environment provided by the global analyses used as initial fields for mesoscale models. As a result of the observational data sparsity for the Antarctic, the quality of the analyses appears to be in many cases insufficient to satisfyingly reproduce the synoptic conditions during complex MC developments.
1) Comparison of EWSR–WRSR
Considering the regional characteristics, the EWSR and the WRSR clearly show important similarities. In both areas MCs represent a frequent phenomenon, for which coastal polynyas and katabatic winds over the ice slopes appear to be important. The forcing mechanisms of the MCs show several similarities. A strong dependence on the specific topographical structures as well as on the synoptic support are found from the case studies performed. However, the simulations for the EWSR and the WRSR also reveal differences. The katabatic winds in the vicinity of the stations Terra Nova Bay and McMurdo are generally more intense than the katabatic winds in the EWSR. Looking in detail at the regional characteristics of the EWSR and the WRSR, it has to be noted that the coastal slopes are not as steep in the EWSR as in the WRSR, and that the valley structures in the WRSR are more pronounced, which consequently explains the generally stronger katabatic winds in the WRSR. In addition, the slopes of the WRSR coast in the vicinity of Terra Nova Bay extend as far as the coastal polynya, while an ice shelf area is located between the coast and open water in the EWSR. Although surface heat fluxes over the polynya appear to be of minor influence for the summertime MC developments in the WRSR investigated in this study, this could be important for other seasons. These considerations suggest that the WRSR has an even larger potential for mesocyclogenesis than the EWSR.
5. Conclusions
Mesocyclones are a prominent atmospheric phenomenon in the Antarctic. The complex Antarctic orography of the coastal regions can play an important role in mesocyclogenesis. In the EWSR and in the WRSR the orographic structure along the coast leads to a significant channeling of the katabatic boundary layer winds. In connection with additional synoptic forcing, the vertical stretching of the converging flow leads to a production of cyclonic vorticity. Besides this stretching mechanism, katabatic winds can have a second impact on the generation of MCs by transporting cold air into the coastal areas and thereby enhancing the low-level baroclinicity. However, it is important to note that no purely katabatic wind induced MC of the discussed size has ever been observed according to our knowledge. It has to be concluded therefore that the katabatic wind alone is too weak to generate MCs of the observed scales, although it does provide a favorable environment for mesocyclogenesis. Additional synoptic forcing supports the effect of katabatic winds in several ways. Since the vertical extent of the katabatic wind layer is relatively small (only several 100 m), an additional synoptic forcing can help to provide a vertical stretching over a thicker layer. Furthermore, the synoptic forcing can support the katabatic flow and lead to katabatic surges, which extend farther over the open water areas. Without such additional synoptic forcing, the katabatic flow is generally confined to the ice slopes and dissipates relatively soon after crossing the coastline or over the ice shelves.
Important synoptic support is given by the classical baroclinic upper-level forcing, which is associated with upper-level cyclonic PV intrusions. The initial low-level perturbations, however, seem to be strongly connected to the specific topographical features of this region and/or existing low-level baroclinicity.
It is evident from the investigated MC cases that the orography is crucial for the genesis of many MCs in the Antarctic. However, the specific topographic structure of Terra Nova Bay and the associated katabatic winds in the vicinity of the stations Terra Nova Bay and McMurdo seem to be especially favorable for mesocyclogenesis. Intense katabatic airstreams provide a cyclonic shear for the initial MC development, and the slopes of the WRSR coast extend as far as the coastal polynya, while an ice shelf area is located between the coast and open water in the Weddell Sea region. These considerations suggest that the WRSR has a larger potential for mesocyclogenesis than the Weddell Sea region. At the present stage, there is no way to prove this hypothesis. An extensive study using data of many years would be necessary to obtain a general climatology of MCs in the Antarctic. Probably this climatotology should be model based, since satellite-based climatologies will only include MCs associated with cloud formation.
Considering the difficulty of operational forecasting in the Antarctic, it becomes obvious that numerical modeling of MCs still represents a very challenging task. The studies presented in this work show that mesoscale models are in principle able to capture the development of MCs, although some tendency to underestimate the intensity of the developments seems to be present. Comparisons with the relatively successful polar low forecasting for the Northern Hemisphere lead to a question about the reasons for the different levels of success in the forecast attempts. One major difficulty is the description of certain physical processes, such as cloud formation and precipitation, radiative transfer, and turbulence in the stably stratified Antarctic boundary layer in numerical weather prediction models. Since the Antarctic near-surface characteristics turn out to be one important factor for the MC development, poor forecasts could also be a result of an insufficient representation of, for instance, the complex orography of Antarctic in the models or from a too coarse model resolution. But, taking into consideration the importance of the synoptic forcing, it has also to be stated that the quality of the analyses is often insufficient to yield the synoptic forcing with that degree of accuracy as required for a successful MC forecast using a mesoscale limited area model. The latter conclusion clearly underlines the demand to incorporate new observational data (e.g., obtained by remote sensing techniques) in operational Antarctic weather prediction in order to compensate for the general data sparsity in this region.
Acknowledgments
This research was supported by the Deutsche Forschungsgemeinschaft under Grant He 2740/1. The authors thank the Norwegian Meteorological Institute (DNMI) at Oslo, Norway, for providing the NORLAM model. The ECMWF provided the SST data and analyses taken as initial and boundary conditions for the realistic simulations. SSM/I data used for the derivation of the sea ice coverage for January/February 1990 were provided by Remote Sensing Systems (California). The sea ice coverage for February 1989 and for the Ross Sea cases is based on SSM/I-derived sea ice concentrations supplied by the National Snow and Ice Data Center (NSIDC) at Boulder, Colorado. The orography data, GTOPO30, were provided by the U.S. Geological Survey's Earth Resources Observation Systems (EROS) Data Center. AVHRR data were made available by NOAA/NESDIS, Washington D.C.
REFERENCES
Blackadar, A. K., 1979: High resolution models of the planetary boundary layer. Advances in Environment and Scientific Engineering, J. Pfafflin and E. Ziegler, Eds., Gordon and Breach, Vol. 1. 50–85.
Bratseth, A., 1982: A simple and efficient approach to the initialization of weather prediction models. Tellus, 34 , 352–357.
Bromwich, D. H., 1989: Satellite analyses of Antarctic katabatic wind behavior. Bull. Amer. Meteor. Soc., 70 , 738–749.
Bromwich, D. H., 1991: Mesoscale cyclogenesis over the southwestern Ross Sea linked to strong katabatic winds. Mon. Wea. Rev., 119 , 1736–1752.
Carrasco, J. F., and D. H. Bromwich, 1993: Mesoscale cyclogenesis dynamics over the Southwestern Ross Sea, Antarctica. J. Geophys. Res., 98 , 12973–12995.
Carrasco, J. F., and D. H. Bromwich, 1994: Climatological aspects of mesoscale cyclogenesis over the Ross Sea and Ross Ice Shelf regions of Antarctica. Mon. Wea. Rev., 122 , 2405–2425.
Carrasco, J. F., and D. H. Bromwich, 1995: A midtropospheric subsynoptic-scale vortex that developed over the Ross Sea and Ross Ice Shelf of Antarctica. Antarct. Sci., 7 , 199–210.
Carrasco, J. F., D. H. Bromwich, and Z. Liu, 1997: Mesoscale cyclone activity over Antarctica during 1991. Part 2: Near the Antarctic Peninsula. J. Geophys. Res., 102 , 13939–13954.
Engels, R., and G. Heinemann, 1996: Three-dimensional structures of summertime Antarctic meso-scale cyclones. Part II: Numerical simulations with a limited area model. Global Atmos.–Ocean Syst., 4 , 181–208.
Gallée, H., 1995: Simulation of the mesocyclonic activity in the Ross Sea, Antarctica. Mon. Wea. Rev., 123 , 2051–2069.
Grønås, S., and O. E. Hellevik, 1982: A limited area prediction model at the Norwegian Meteorological Institute. Tech. Rep. 61, Norwegian Meteorological Institute, Oslo, Norway, 75 pp.
Grønås, S., A. Foss, and M. Lystad, 1987: Numerical simulations of polar lows in the Norwegian Sea. Tellus, 39A , 334–353.
Heinemann, G., 1990: Mesoscale vortices in the Weddell Sea region. Mon. Wea. Rev., 118 , 779–793.
Heinemann, G., 1996: Three-dimensional structures of summertime Antarctic meso-scale cyclones. Part I: Observational studies with aircraft, satellite and conventional data. Global Atmos.–Ocean System, 4 , 149–180.
Heinemann, G., 1997: Idealized simulations of the Antarctic katabatic wind system with a three-dimensional meso-scale model. J. Geophys. Res., 102 , 13825–13834.
Heinemann, G., 1998: A meso-scale model-based study of the dynamics of a wintertime polar low in the Weddell Sea region of the Antarctic during WWSP86. J. Geophys. Res., 103 , 5983–6000.
Klein, T., and G. Heinemann, 2001: On the forcing mechanisms of mesocyclones in the eastern Weddell Sea region: Process studies using a mesoscale numerical model. Meteor. Z., 10 , 113–122.
Kuo, H. L., 1965: On formation and intensification of tropical cyclones through latent heat release by cumulus convection. J. Atmos. Sci., 22 , 40–63.
Louis, J-F., 1979: A parametric model of vertical eddy fluxes in the atmosphere. Bound.-Layer Meteor., 17 , 187–202.
Nordeng, T. E., 1986: Parameterization of physical processes in a three-dimensional numerical weather prediction model. Tech. Rep. 65, Norwegian Meteorological Institute, Oslo, Norway, 48 pp.
Nordeng, T. E., 1990: Model-based diagnostic study of the development and maintenance mechanism of two polar lows. Tellus, 42A , 92–108.
Reynolds, R. W., and T. M. Smith, 1994: Improved global sea surface temperature analyses using optimum interpolation. J. Climate, 7 , 929–948.
Scientific Committee for Antarctic Research, 1993: Antarctic Digital Database on CD-ROM. Scott Polar Research Institute, Cambridge, United Kingdom, CD-ROM.
Turner, J., T. A. Lachlan-Cope, D. E. Warren, and C. N. Duncan, 1993: A mesoscale vortex over Halley Station, Antarctica. Mon. Wea. Rev., 121 , 1317–1336.
Turner, J., G. Corcoran, S. Cummins, T. Lachlan-Cope, and S. Leonard, 1996: Seasonal variability of meso-cyclone activity in the Bellingshausen/Weddell Sea region of Antarctica. Global Atmos.–Ocean Syst., 5 , 73–97.
USGS, 1997: GTOPO30 Digital Elevation Model. EROS Data Center, U.S. Geological Survey, Sioux Falls, SD. [Available online at http://edcaac.usgs.gov/gtopo30.html.].
Map of the Weddell Sea region of Antarctica comprising the model domain of the LAM50 with orography (solid isolines every 500 m). The domain of the LAM25 is framed by the box labeled LAM25. The region referred to as the “eastern Weddell Sea region” and the Filchner/Ronne Ice Shelf are marked by EWSR and FRIS, respectively. SST (isolines every 2°C, dashed) and the sea ice edge (indicated by the thick dashed line) are shown for 8 Feb 1990. The solid thick line marks the coastline. Antarctic radiosonde stations are indicated by black triangles (Hal, Halley station; GvN, Georg von Neumayer)
Citation: Monthly Weather Review 131, 2; 10.1175/1520-0493(2003)131<0302:SOTFMI>2.0.CO;2
As in Fig. 1 but for the Ross Sea region of Antarctica. The areas referred to in the text as the western Ross Sea region and the Ross Ice Shelf are marked by WRSR and RIS, respectively (SST and sea ice edge for 18 Feb 1988; MCM = McMurdo station, Ter = Terra Nova Bay station)
Citation: Monthly Weather Review 131, 2; 10.1175/1520-0493(2003)131<0302:SOTFMI>2.0.CO;2
AVHRR visible satellite image (NOAA channel 2, albedo in %) for 0727 UTC 17 Feb 1989 for the Weddell Sea region
Citation: Monthly Weather Review 131, 2; 10.1175/1520-0493(2003)131<0302:SOTFMI>2.0.CO;2
AVHRR infrared satellite image (NOAA, channel 4 brightness temperatures, grayscale in K) for 0727 UTC 17 Feb 1989. Squares indicate the location of Halley station (Hal) and a drifting buoy (B). Wind vectors are shown for 0600 UTC at Hal and for 0624 UTC at B (full barb, 10 kt; half barb, 5 kt)
Citation: Monthly Weather Review 131, 2; 10.1175/1520-0493(2003)131<0302:SOTFMI>2.0.CO;2
An 18-h prognosis of the LAM25 valid at 1800 UTC 16 Feb 1989. MSLP (full isolines every 2 hPa), potential temperature at 850 hPa (dashed isolines every 2 K), and relative humidity at 850 hPa (shaded if larger than 90%)
Citation: Monthly Weather Review 131, 2; 10.1175/1520-0493(2003)131<0302:SOTFMI>2.0.CO;2
Wind vectors and potential temperature (dashed isolines every 2 K) at the lowest σ level after +30 h (valid at 0600 UTC 17 Feb 1989). The M indicates the position of a simulated MC near Halley. The square contains an area for which the vorticity budget was analyzed (mean over nine grid points). Here, M1–M2 show the horizontal position of a cross section. Areas with values of the stretching term at 850 hPa ≤ −2.0 × 10−8 s−2 are shaded. The orography is shown by full lines (isolines every 500 m)
Citation: Monthly Weather Review 131, 2; 10.1175/1520-0493(2003)131<0302:SOTFMI>2.0.CO;2
Time series of the budget terms of the vorticity equation at 950 hPa (1200 UTC 16 Feb–1200 UTC 17 Feb 1989), mean values for the box containing the MC M (see Fig. 6): HADV, horizontal advection; VADV, vertical advection; TT, tilting term; DZ, local tendency of ζ; DF, diffusion term; and DIV, divergence term
Citation: Monthly Weather Review 131, 2; 10.1175/1520-0493(2003)131<0302:SOTFMI>2.0.CO;2
Cross section M1–M2 (see Fig. 6) after +30 h (0600 UTC 17 Feb 1989, LAM25): wind vectors tangential to plane; potential temperature (dashed, isolines every 2 K) and wind normal to plane (isolines every 1 m s−1, full lines correspond to winds directed into the plane of the cross section, isolines dashed with short dashes indicate winds directed in the opposite direction). Areas with PV ≤ −1.5 PVU are shaded
Citation: Monthly Weather Review 131, 2; 10.1175/1520-0493(2003)131<0302:SOTFMI>2.0.CO;2
AVHRR satellite image (NOAA, channel 4 brightness temperatures, grayscale in K) for 2245 UTC 28 Jan 1990. Berkner Island (B1), the Filchner/Ronne Ice Shelf (FRIS), and the MC near Halley station are marked
Citation: Monthly Weather Review 131, 2; 10.1175/1520-0493(2003)131<0302:SOTFMI>2.0.CO;2
An 18-h LAM25 forecast (valid at 1800 UTC 28 Jan 1990) of wind vectors (only every fourth vector plotted) and relative humidity (shaded if larger than 90%) at 850 hPa. The orography is shown by full lines (isolines every 500 m). The filled circle and the filled square indicate the approximate positions of the center of the MC at 0000 UTC 26 Jan and 0000 UTC 27 Jan, respectively
Citation: Monthly Weather Review 131, 2; 10.1175/1520-0493(2003)131<0302:SOTFMI>2.0.CO;2
Geopotential height (full isolines every 40 m) and temperature (dashed isolines every 2°C) at 500 hPa of a 24-h LAM25 forecast (valid at 0000 UTC 27 Jan 1990). The dark and light shadings show areas with PV at 500 hPa ≤ −1.5 PVU after 24 and 48 h, respectively
Citation: Monthly Weather Review 131, 2; 10.1175/1520-0493(2003)131<0302:SOTFMI>2.0.CO;2
Wind vectors at the lowest σ level at 0600 UTC 26 Jan 1990 (6-h LAM25 forecast) and orography (full isolines every 500 m) for a part of the model domain. Here, M1, M2, and M3 indicate the positions of simulated MCs
Citation: Monthly Weather Review 131, 2; 10.1175/1520-0493(2003)131<0302:SOTFMI>2.0.CO;2
As in Fig. 12 but for 0000 UTC 29 Jan 1990 (24-h LAM25 forecast) and without M3
Citation: Monthly Weather Review 131, 2; 10.1175/1520-0493(2003)131<0302:SOTFMI>2.0.CO;2
AVHRR satellite image (NOAA, channel 4 brightness temperatures, grayscale in K) for 1846 UTC 29 Jan 1990 with superimposed NORLAM coastline (thick line), ice edge (thick dashed line), and NORLAM-simulated MSLP (18-h forecast, valid at 1800 UTC 29 Jan 1990, isolines every 1 hPa)
Citation: Monthly Weather Review 131, 2; 10.1175/1520-0493(2003)131<0302:SOTFMI>2.0.CO;2
AVHRR infrared satellite image (NOAA, channel 4) at 1950 UTC 11 Feb 1990. Berkner Island (BI), the Filchner/Ronne Ice Shelf (FRIS), and the position of the MC are marked
Citation: Monthly Weather Review 131, 2; 10.1175/1520-0493(2003)131<0302:SOTFMI>2.0.CO;2
Wind vectors (only every second vector plotted) and temperature (isolines every 2°C, temperatures ≥ −8°C shaded) at the lowest σ level at 1200 UTC 11 Feb 1990 (36-h LAM25 forecast) and orography (full isolines every 500 m) for a part of the model domain
Citation: Monthly Weather Review 131, 2; 10.1175/1520-0493(2003)131<0302:SOTFMI>2.0.CO;2
Geopotential height (full isolines every 40 m) and temperature (dashed isolines every 2°C) at 500 hPa of a 6-h LAM50 forecast (valid at 0600 UTC 10 Feb 1990). The dark and light shadings show areas with PV at 500 hPa ≤ −1.5 PVU after 6 and 54 h (valid at 0600 UTC 11 Feb 1990), respectively
Citation: Monthly Weather Review 131, 2; 10.1175/1520-0493(2003)131<0302:SOTFMI>2.0.CO;2
(a) Wind vectors and potential temperature (dashed isolines every 2 K) at the lowest σ level after 24 h of LAM25 integration for the WRSR for an idealized simulation with an atmosphere at rest as initial fields. The part of the model domain shown is identical to the area framed by the smaller box in Fig. 2. A length scale is shown below the lower-right corner. The Ross Ice Shelf (RIS), Ross Island (RI), Reeves/David Glacier (RDG), and Mulock Glacier (MG) are marked. (b) Same as in (a) but after 60 h. Areas with potential temperatures lower than 260 K are shaded in 10-K intervals. Additionally, a filled circle indicates the position of the simulated MC
Citation: Monthly Weather Review 131, 2; 10.1175/1520-0493(2003)131<0302:SOTFMI>2.0.CO;2
AVHRR infrared satellite image (NOAA, channel 4, grayscale in K) for 1300 UTC 18 Feb 1988. Here, M1 indicates a MC. The Ross Ice Shelf (RIS) and Ross Island (RI) are indicated
Citation: Monthly Weather Review 131, 2; 10.1175/1520-0493(2003)131<0302:SOTFMI>2.0.CO;2
Wind vectors, potential temperature (dashed isolines every 2 K), and relative humidity (shaded if larger than 90%) at 925 hPa after 18 h of LAM25 integration (valid at 1800 UTC 18 Feb 1988). Here, M1 indicates the simulated MC near Terra Nova Bay (Ter). The orography is shown by full lines (isolines every 500 m)
Citation: Monthly Weather Review 131, 2; 10.1175/1520-0493(2003)131<0302:SOTFMI>2.0.CO;2
Wind vectors and potential temperature (dashed isolines every 2 K) at the lowest σ level (about 30 m above ground) after 18 h of LAM25 integration (valid at 1800 UTC 18 Feb 1988). Areas with potential temperatures lower than 260 K are shaded in 4-K intervals. Here, M1 indicates a simulated MC, and K indicates a channeled katabatic cold airflow near Ter. The orography is shown by full lines (isolines every 500 m)
Citation: Monthly Weather Review 131, 2; 10.1175/1520-0493(2003)131<0302:SOTFMI>2.0.CO;2
Geopotential height (full isolines every 10 m) and temperature (dashed isolines every 2°C) of a 15-h LAM25 forecast (valid at 1500 UTC 18 Feb 1988) at 700 hPa. The shadings show areas with PV at 700 hPa ≤ −1.0 PVU
Citation: Monthly Weather Review 131, 2; 10.1175/1520-0493(2003)131<0302:SOTFMI>2.0.CO;2
(a) Regional analysis of MSLP using AWS data (taken from Carrasco and Bromwich 1995) for 1800 UTC 7 Jan 1988. (b) LAM25 prognosis of MSLP (isolines every 2 hPa, +18 h) for 1800 UTC 7 Jan 1988. The L1, L2, and the H indicate low and high pressure systems, respectively
Citation: Monthly Weather Review 131, 2; 10.1175/1520-0493(2003)131<0302:SOTFMI>2.0.CO;2
Wind vectors and relative humidity (shaded if larger than 90%) at 925 hPa after 24 h of LAM25 integration (valid at 0000 UTC 8 Jan 1988). The L1 and L2 indicate simulated MCs. The orography is shown by full lines (isolines every 500 m)
Citation: Monthly Weather Review 131, 2; 10.1175/1520-0493(2003)131<0302:SOTFMI>2.0.CO;2
Wind vectors and potential temperature (dashed isolines every 2 K, values lower than 262 K shaded in 4-K intervals) at the lowest σ level after 18 h (1800 UTC 7 Jan 1988, LAM25) and orography. The L1 and L2 indicate simulated MCs, while K1, K2a, and K2b indicate katabatic flow regimes. Reeves/David Glacier (RDG), Mulock Glacier (MG), and Byrd Glacier are marked
Citation: Monthly Weather Review 131, 2; 10.1175/1520-0493(2003)131<0302:SOTFMI>2.0.CO;2
Characteristics of the NORLAM as used for the simulations of mesocyclones
List of the datasets used for the NORLAM simulations
