Abstract

The development of three summertime mesoscale cyclones (MCs) over the northern Amundsen and Bellingshausen Seas from 10 to 12 January 1995 (during FROST SOP-3) is studied by means of AVHRR data, ERS and SSM/I retrievals, and mesoscale numerical model data. The most pronounced MC is investigated in detail. It had a diameter of about 800 km, a lifetime of more than 24 h, and reached the intensity of a polar low. The developments take place far away from the sea-ice front or topography. The MCs are detected as cyclonic cloud signatures in the AVHRR imagery, and SSM/I retrievals show a distinct mesoscale signal in the fields of cloud liquid water, wind speed, and integrated water vapor (IWV). The frontal structure of the most intense MC is depicted by high IWV gradients and a large near-surface wind shear. The collocation of ERS- and SSM/I-derived wind speeds shows good agreement (bias, 1.1 m s−1; std dev, 1.2 m s−1). ERS-derived wind vectors give no insight into the structure of the MCs, because of missing direct overpasses over the MCs by the narrow ERS scatterometer swaths, but they are used to validate numerical simulations. The numerical simulations using the mesoscale Norwegian Limited Area Model (NORLAM) show two of the MCs as short-wave baroclinic developments triggered by an upper-level trough, while a less significant third MC is not simulated by the model. In contrast to the satellite retrievals, the simulations give insight into the three-dimensional structure of the MCs. Model results are validated using satellite retrievals and some few available in situ observations. This validation study shows the good quality of the numerical simulations for the IWV and the near-surface wind speed from SSM/I as well as for the near-surface wind vector from ERS over the simulation time of 36 h. The differences between ERS and NORLAM wind vectors are 1.1 ± 2.5 m s−1 (mean bias ± std dev) and −3 ± 25° for wind speed and direction, respectively. The validation using SSM/I retrievals yields a mean bias of 0.3 m s−1 (std dev, 2.9 m s−1) for the wind speed, and of −2.5 kg m−2 (std dev, 2.9 kg m−2) for the IWV.

1. Introduction

Summertime mesoscale cyclones over Antarctic oceans have been the subject of several investigations in recent years. In this paper, the term mesoscale cyclone (MC) will be used for cyclonic vortices poleward of the main polar front up to a horizontal scale of 2000 km and the polar low being a subgroup of intense maritime MCs with scales smaller than about 1000 km and a near-surface wind speed exceeding 15 m s−1. Since Antarctic MCs often occur in data-sparse regions, satellite data and simulations using mesoscale numerical models are important tools for the investigations of these phenomena. Antarctic MCs have been observed for different regions of the Antarctic by means of satellite imagery, geophysical parameters retrieved from satellite data, and—in the case that an MC was close to a research station or automatic weather stations—by direct observational data. Detailed observational studies have been carried out for the areas of the Ross Sea (e.g., Carrasco and Bromwich 1993) and also for the area of the Weddell Sea (e.g., Heinemann 1996a; Turner et al. 1993).

Multisensor satellite studies using quantitative retrievals of geophysical parameters from different satellite sensors and high-resolution digital satellite images have been the basis for recent MC studies in the Antarctic. Important satellite sensors used for retrievals of geophysical parameters are the Special Sensor Microwave/Imager (SSM/I) on board the Defense Meteorological Satellite Program (DMSP) satellites, the European Remote Sensing Satellite Scatterometer (ERS-SCAT), and the TIROS-N Operational Vertical Sounder (TOVS) on board the National Oceanic and Atmospheric Administration (NOAA) satellites. TOVS data have been used for retrievals of layer-mean atmospheric temperatures and humidities in polar regions (e.g., Heinemann et al. 1995; Claud et al. 1992) but have a relatively coarse spatial resolution. Passive microwave SSM/I data (Hollinger et al. 1987) allow the determination of near-surface wind speed (WS) and integrated atmospheric water parameters over ocean areas, but low-level structures are often smoothed out by averaging over the whole atmospheric column. In contrast to SSM/I, near-surface wind vectors can be obtained from ERS-SCAT data over ocean areas. The main restriction of ERS-SCAT is the narrowness of the scatterometer swath (about 500 km only), which leads to large gaps between subsequent satellite overpasses and therefore to an insufficient areal coverage. But, in some cases, the circulation structure of Antarctic MCs can be well documented using ERS-SCAT data (e.g., McMurdie et al. 1997; Lieder and Heinemann 1996).

Mesoscale limited-area models with a horizontal resolution comparable to SSM/I or ERS retrievals have been used successfully for numerical simulation of MCs in many cases of Arctic polar lows (e.g., Nordeng 1990;Nordeng and Rasmussen 1992), but only a few numerical case studies exist for the Antarctic. Simulations of summertime coastal MC developments are shown by Engels and Heinemann (1996) and a simulation of a wintertime MC event is presented in Heinemann (1998).

The current paper investigates a group of three summertime MCs occurring from 11 to 12 January 1995 over the Antarctic region of the Southern Pacific, that is, over the northern Amundsen and Bellingshausen Seas. January 1995 is one of the special observing periods (SOPs) of the Antarctic First Regional Observing Study of the Troposphere (FROST; Turner et al. 1996). The goals of FROST are to study the meteorology of the Antarctic and to gain insight into the quality of operational analyses and forecasts and the value of the available satellite data in this region. For three SOPs comprehensive datasets containing in situ and satellite observations, model fields, and manual analyses (surface and upper air) exist for the area south of 50°S (some datasets up to 40°S). Details of the FROST analyses and the assessment of operational analyses using FROST data can be found in papers in the present special issue (e.g., Turner et al. 1999; Hutchinson et al. 1999; Bromwich et al. 1999). Unfortunately, there are no FROST manual analyses yet for the period of interest in January 1995.

The FROST dataset offers an excellent opportunity for studies of Antarctic meteorology combining multisensor satellite data and model simulations. In the present study we want to stress the value of this dataset for investigation of Antarctic mesoscale cyclones. Since in situ measurements are very rare over the southern oceans, satellite data represent almost the only observational data for the MCs investigated (only four drifting buoys are available for the complete ocean region of interest and neighboring coastal areas). Satellite-based studies of Antarctic MCs (e.g., McMurdie et al. 1997) can give some insight into their structure, but fail in explaining the three-dimensional structure and development mechanisms. On the other hand, model-based studies (which are also very rare for the Antarctic) lack a validation by observational data in general.

The goal of the present study is to use multisensor satellite data of the FROST dataset in conjunction with mesoscale model simulations for the investigation of Antarctic MCs. Satellite imagery and retrievals are used for synoptic and subsynoptic description of the MCs, but also serve to validate the model simulations by comparing near-surface wind and integrated water vapor from the model fields with satellite-retrieved fields.

The data for the satellite-based study as well as the mesoscale numerical model are introduced in section 2. A synoptic description by infrared Advanced Very High Resolution Radiometer (AVHRR) imagery and numerical model simulations is given in section 3. The next section includes the mesoscale analysis using SSM/I and ERS-SCAT retrievals. Numerical model simulations with 50- and 25-km horizontal resolution allow a model-based mesoscale analysis and comparisons with the satellite retrievals (section 5). Conclusions are given in section 6.

2. Satellite data, retrieval algorithms, and numerical model

a. Satellite data

For the case study the complete FROST database for this region of the Antarctic was used. Of the available satellite data only TOVS measurements were not used directly, but they are assimilated operationally into the European Centre for Medium-Range Weather Forecasts (ECMWF) model fields, which represent the initial fields for the mesoscale model (see below).

The imagery in this study is based on the radiance measurements in the infrared channel 4 (10.5–11.5 μm) of the AVHRR instrument on board the polar orbiting satellites NOAA-9 and NOAA-12. AVHRR data from NOAA-9 was obtained as High Resolution Picture Transmission (HRPT) data with a nadir resolution of 1.1 km from the British Antarctic Survey (BAS), operating an HRPT receiver at Rothera Station, Antarctic Peninsula. The areal coverage of the receiver has a typical radius of 2500 km, so that the development of the investigated MCs (see below) is only partly covered by HRPT data. Global area coverage (GAC) data with a reduced resolution of 4 km (nadir) was acquired from the Satellite Active Archive (SAA) of NOAA/National Environmental Satellite, Data and Information Service. GAC data for the region of interest is available as two sets of overpasses daily, because solely NOAA-12 data was archived at SAA for the period of January 1995. All GAC and HRPT data used in the present study are summarized in Table 1.

Table 1.

AVHRR data available for the region of interest (GAC; NOAA-12; HRPT; NOAA-9; radiometric resolution, 10 bit). Italic times indicate the orbits presented in Fig. 2, bold times indicate orbits (partly) covering the MC M3.

AVHRR data available for the region of interest (GAC; NOAA-12; HRPT; NOAA-9; radiometric resolution, 10 bit). Italic times indicate the orbits presented in Fig. 2, bold times indicate orbits (partly) covering the MC M3.
AVHRR data available for the region of interest (GAC; NOAA-12; HRPT; NOAA-9; radiometric resolution, 10 bit). Italic times indicate the orbits presented in Fig. 2, bold times indicate orbits (partly) covering the MC M3.

The scatterometer mode of the Active Microwave Instrument on board the polar orbiting ERS-1 satellite allows us to determine near-surface wind vectors over ice-free ocean from the microwave radiation backscattered by the wind-induced surface waves. The ERS-SCAT data presented in the following sections are the ERS-1 offline scatterometer wind products, which were generated at the Institut Français de Recherche pour l’Exploitation de la Mer (IFREMER) using the Centre ERS d’Archivage et de Traitement (CERSAT) algorithms (version 2.2). The accuracies are 10% of the wind speed value (minimum 2 m s−1) and 20° for wind direction. The spatial resolution is about 50 km with a pixel spacing of 25 km and a 500-km swath width. The sun-synchronous orbit of ERS-1 with a period of approximately 100 min results in a distance of 25° longitude between two subsequent orbits, which causes gaps of 900 km between two swaths at 60°S, equivalent to an areal coverage of 35% for this latitude. ERS-SCAT overpasses during the period of interest are listed in Table 2.

Table 2.

Time and position of the ERS-1 orbits available for the region of interest.

Time and position of the ERS-1 orbits available for the region of interest.
Time and position of the ERS-1 orbits available for the region of interest.

The multifrequency, dual-polarization measurements of the SSM/I sensor allow the retrieval of several atmospheric parameters over ice-free ocean with a resolution of 30–70 km, depending on the channels used for the retrieval. For the present study sea ice extent (being an important information for further retrievals) was determined using the algorithm of Cavalieri et al. (1995). The near-surface wind speed was retrieved with an accuracy of better than 2 m s−1 using the algorithm of Goodberlet and Swift (1992). For areas with moderate influence of cloud liquid water (CLW) and rain, the accuracy of the retrieval degrades to 5 m s−1, and pixels with high cloud and rain contamination are discarded completely. The algorithm of Petty (1994) was used to determine the integrated water vapor (IWV) excluding areas with high probability for intensive rain. Comparisons to radiosonde data showed rms differences of 2.6 kg m−2 (Petty 1993). The algorithm of Alishouse et al. (1990) was applied for CLW retrieval. As for IWV the results are rejected for pixels flagged as highly rain contaminated. Data from the SSM/I sensors on the DMSP satellites F-10 and F-11 were obtained from the Marshall Space Flight Center Distributed Active Archive Center (MSFC DAAC) as NESDIS level 1b and NESDIS temperature data record (TDR) data (Table 3). Because the SSM/I swath width is 1400 km and the sun-synchronous orbit has a duration of 100 min, one sensor covers twice daily the complete area poleward of 60° latitude.

Table 3.

SSM/I orbits available for the region of interest. Italic times indicate the orbits presented in Figs. 5–7, bold times indicate orbits (partly) covering the MC M3.

SSM/I orbits available for the region of interest. Italic times indicate the orbits presented in Figs. 5–7, bold times indicate orbits (partly) covering the MC M3.
SSM/I orbits available for the region of interest. Italic times indicate the orbits presented in Figs. 5–7, bold times indicate orbits (partly) covering the MC M3.

b. Description of the numerical mesoscale model

In this study, the Norwegian Limited Area Model (NORLAM) (version 9) developed at the Norwegian Meteorological Institute (Grønås et al. 1987) is used. NORLAM is a full-physics hydrostatic primitive equation model, which is run with a resolution of 121 × 97 horizontal grid points and 30 vertical sigma levels. Forecast runs are carried out in a nested mode, a first run is made for a 50-km grid (LAM50) with lateral boundaries provided by 6-hourly ECMWF analyses (with a resolution of 1.125° × 1.125° for this study). The model domain with 25-km horizontal grid distance (LAM25) is nested in the integration area of LAM50, taking forecasts of LAM50 as initial fields (6-h forecast) and time-dependent boundary conditions during the simulation. Topography data are taken from a dataset with a resolution of 1/12° × 1/12° (NOAA 1995). The SST data from ECMWF analyses is used, but areas with unrealistic low SST are set to a lower limit of −1°C. The sea ice coverage is taken from daily maps produced by the Pelicon project (Heygster et al. 1996).

The sea ice front and the ECMWF analysis of mean sea level pressure for 0000 UTC 11 January 1995 are shown in Fig. 1a for a part of the model domain of LAM50; the model topography and geopotential height at 500 hPa for 0000 UTC are given in Fig. 1b. Solid triangles indicate stations performing radiosonde ascents, solid squares indicate buoys and automatic weather stations in the model domain. The data of these stations have been collected in the FROST database during the period of interest, but it is not clear if these data were used in the operational ECMWF data assimilation. A solid dot marks the center position of the most pronounced MC (M3) at 0400 UTC 11 January (obtained from satellite imagery). This MC developed over the open ocean at about 57°S, 120°W, that is, far away from the sea ice front in an area where no radiosonde stations can be found. A synoptic-scale cyclone lay over the partly ice-covered Bellingshausen Sea (a more detailed synoptic description is given in section 3).

Fig. 1.

(a) Subsection of the LAM50 domain with MSLP (solid, isolines every 4 hPa) and IWV (dashed, isolines every 3 kg m−2) from the ECMWF analysis at 0000 UTC 11 Jan 1995, the land areas (dark grayshading, including the ice shelf areas) and the sea ice front (thick dashed line, sea ice–covered areas shaded light gray). Triangles mark the radiosonde stations: Punta Arenas in South America, Mount Pleasant Airport on the Falkland Islands, Belgrano and Marambio at the Antarctic Peninsula, and Halley at the coast of the eastern Weddell Sea. The center of M3 at 0400 UTC 11 Jan obtained from satellite images is marked by the full circle, L marks the center of a synoptic-scale cyclone. (b) Same subsection as in (a) but with the geopotential height at 500 hPa (solid, isolines every 60 gpm) from the ECMWF analysis at 0000 UTC 11 Jan 1995, the Antarctic topography (contour intervals of 500 m, shaded for heights exceeding 500 m), the coastline, and the sea ice front. The solid box marks the area of the LAM25. In addition to M3, L, and the radiosonde stations, the synoptic stations are indicated by small triangles and small squares indicate positions of automatic weather stations and drifting buoys (for the period 10–13 Jan 1995).

Fig. 1.

(a) Subsection of the LAM50 domain with MSLP (solid, isolines every 4 hPa) and IWV (dashed, isolines every 3 kg m−2) from the ECMWF analysis at 0000 UTC 11 Jan 1995, the land areas (dark grayshading, including the ice shelf areas) and the sea ice front (thick dashed line, sea ice–covered areas shaded light gray). Triangles mark the radiosonde stations: Punta Arenas in South America, Mount Pleasant Airport on the Falkland Islands, Belgrano and Marambio at the Antarctic Peninsula, and Halley at the coast of the eastern Weddell Sea. The center of M3 at 0400 UTC 11 Jan obtained from satellite images is marked by the full circle, L marks the center of a synoptic-scale cyclone. (b) Same subsection as in (a) but with the geopotential height at 500 hPa (solid, isolines every 60 gpm) from the ECMWF analysis at 0000 UTC 11 Jan 1995, the Antarctic topography (contour intervals of 500 m, shaded for heights exceeding 500 m), the coastline, and the sea ice front. The solid box marks the area of the LAM25. In addition to M3, L, and the radiosonde stations, the synoptic stations are indicated by small triangles and small squares indicate positions of automatic weather stations and drifting buoys (for the period 10–13 Jan 1995).

3. Satellite images and synoptic-scale analysis

The focus of this paper is a family of three MCs developing from 11 to 12 January 1995 over the northern Bellingshausen and Amundsen Seas. For an interpretation based on satellite images, all NOAA overpasses with digital AVHRR (GAC) data archived by NOAA/NESDIS and AVHRR (HRPT) data archived by BAS are available. Table 1 shows the orbits with AVHRR data covering the area of interest. In general, a combination of two or three subsequent calibrated digital GAC images or a combination of nearly simultaneous HRPT and GAC data will be used in order to achieve a better coverage of this relatively large area.

Figure 2a shows the satellite image (AVHRR infrared channel) for 0400 UTC 11 January 1995, which is a composite of the HRPT overpass at 0403 UTC and the GAC overpass at 0413 UTC. The large-scale synoptic situation was characterized by a synoptic-scale cyclone (L in the following) lying over the Bellingshausen Sea with its center at 66°S, 85°W. The main frontal cloud band of L extended to the east over the southern Bellingshausen Sea, then to the north at about 65°W; that is, the main front (marked by the cloud band) lay approximately parallel to the Antarctic coast of West Antarctica and the Antarctic Peninsula (cf. Fig. 1a). Figure 1a also shows the ECMWF analysis of the mean sea level pressure (MSLP) for a subsection of the LAM50 domain at about the same time. The ECMWF analysis at 0000 UTC 11 January 1995 was taken as the initial field for the LAM50 simulations. The analyzed position of the center of L agrees well with the position seen on the satellite image. The sea ice front in the areas of the Bellingshausen and Amundsen Seas extended approximately along the 70°S latitude circle. The center of L therefore lay at the sea ice edge. The three MCs (M1, M2, and M3 in the following) can be detected as cloud signatures northwest of L between 55°S and 60°S.

Fig. 2.

AVHRR channel 4 images from GAC and HRPT data (cf. Table 1, polar stereographic projection, brightness temperature scale in K) during 11 and 12 Jan 1995. (a) Mosaic (reduced resolution) for 0400 UTC 11 Jan. (b) As in (a) but for 1400 UTC 11 Jan.

Fig. 2.

AVHRR channel 4 images from GAC and HRPT data (cf. Table 1, polar stereographic projection, brightness temperature scale in K) during 11 and 12 Jan 1995. (a) Mosaic (reduced resolution) for 0400 UTC 11 Jan. (b) As in (a) but for 1400 UTC 11 Jan.

The initial stage of the most intense and longest-living MC (M3) can be seen in Fig. 2a at 57°S, 122°W about 2000 km northwest of the center of L in a strong zonal flow. East of M3, two small areas of higher clouds at distances of 700 and 1400 km from M3 mark the positions of M2 and M1. The curved cloud band of M2 (at 59°S, 110°W) indicated a mesoscale circulation at that time.

In addition to MSLP, the IWV of the ECMWF analysis at 0000 UTC 11 January 1995 is displayed in Fig. 1a. As it was found in satellite-based studies (Katsaros et al. 1989), this parameter represents a good indicator for fronts of synoptic-scale cyclones. This is also obvious in the ECMWF analysis for the main front of L, which is marked by a pronounced gradient of IWV. The zones of high IWV gradients in the model analysis are also associated with maxima of low-level temperature gradients (not shown). The geopotential height field at 500 hPa in Fig. 1b shows the well-developed upper-level trough of the synoptic low L, a strong zonal flow northwest of L, and a weak trough upstream of the center of M3 (position taken from satellite image at 0400 UTC).

About 10 h later, at 1400 UTC 11 January 1995, M3 developed a pronounced cloud band and moved eastward to 58°S, 112°W (Fig. 2b), while the circulation associated with M2 (at 59°S, 95°W) seemed to be dissipating. The details of the cloud structures of M3 and M2 can best be seen in the full-resolution HRPT image at 1404 UTC (Fig. 2c). Low-level clouds near the center and midlevel clouds south of the center clearly indicated the circulation associated with M3. The appearance of M3 on the satellite image resembled a short baroclinic wave with shallow convection west and north of the center. Mesoscale cyclone M1 (with its center at 60°S, 83°W) was associated with a cloud band with a south–north orientation at that time (Fig. 2b). The center of the large-scale cyclone L had remained almost stationary at 67°S, 90°W, but its front had moved farther eastward and began to pass over the Antarctic Peninsula. This is also reflected by the 12-h simulation of LAM50 valid at 1200 UTC 11 January (Fig. 3a), which shows the highest IWV gradients just west of the Antarctic Peninsula. The blocking effect of the topography and the beginning formation of a lee cyclone east of the Antarctic Peninsula is also visible in the model simulations of the MSLP field. Along the 60°S latitude circle, two weak troughs had formed at 110° and 80°W, being approximately coincident with the positions of M3 and M1, respectively.

Fig. 3.

Simulated MSLP (solid, isolines every 4 hPa) and IWV (dashed, isolines every 3 kg m−2) after 12- and 24-h simulation time for a subsection of the LAM50 domain. The simulations are valid at (a) 1200 UTC 11 Jan and (b) 0000 UTC 12 Jan. The center of M3 at (a) 1400 UTC 11 Jan and (b) at 0200 UTC 12 Jan obtained from satellite images is marked by the full circle and L indicates the center of a synoptic-scale cyclone.

Fig. 3.

Simulated MSLP (solid, isolines every 4 hPa) and IWV (dashed, isolines every 3 kg m−2) after 12- and 24-h simulation time for a subsection of the LAM50 domain. The simulations are valid at (a) 1200 UTC 11 Jan and (b) 0000 UTC 12 Jan. The center of M3 at (a) 1400 UTC 11 Jan and (b) at 0200 UTC 12 Jan obtained from satellite images is marked by the full circle and L indicates the center of a synoptic-scale cyclone.

The 24-h simulation of LAM50 valid at 0000 UTC 12 January 1995 (Fig. 3b) shows the eastward movement of the two mesoscale troughs. The front of L had passed over the Antarctic Peninsula and a well-developed lee cyclone was now centered at 69°S, 60°W. The corresponding satellite image (Fig. 2d) shows the center of M3 at 57°S, 96°W. The cloud structures of M3 resemble again a short baroclinic wave but also indicate the beginning dissipation of the vortex. This dissipation process can be seen to continue on satellite images later during 12 January (not shown).

The only in situ measurements (MSLP and temperature) over the ocean in the model region were carried out by the two drifting buoys marked with black squares in Fig. 1b (around 55°S, 110°W) and with white squares near M3 in Fig. 2b. The northern buoy (buoy 54 807) lay relatively far away from the center of M3, while the southern buoy (buoy 54 808) was almost passed by the center of M3 during its eastward movement (Fig. 2b and Fig. 2d). The MSLP was measured by the buoys approximately every 2 h (Fig. 4). At both buoys the passage of M3 was measured as a drop in pressure between 0600 and 1500 UTC 11 January, the latter time being 1 h after the time of the satellite image in Fig. 2b. For the southern buoy, which was nearer to the center of M3, the drop was more pronounced (about 6 hPa) and the minimum was 992.8 hPa. The lowest value for the northern buoy was about 10 hPa higher, which reflects the pressure gradient associated with the strong zonal flow. The MSLP simulations at the grid points being nearest to the buoy positions are also displayed in Fig. 4. No pressure drop can be found in the simulations at the northern buoy position and only a 2-hPa drop at the southern buoy, where the minimum is reached at 1500 UTC as well, but with a value 10 hPa higher than the measurement. This difference of 10 hPa between the simulated and measured curves of both buoys continues for the next 20 h of the simulation. It will be shown below that the model simulations overestimate the eastward movement of M3 and the center of the simulated MC M3 is found to be 300 km southeast of the observed one after 18 h of simulation. To enable a comparison considering the displacement of the simulated MC M3, the curve with triangles in Fig. 4 displays the MSLP measurements for a grid point about 300 km southeast of buoy 54 808. The simulated pressure drop is more pronounced (3.6 hPa between 0900 and 1500 UTC 11 January) but is still weaker than the measured one at buoy 54 808. From the comparison between MSLP measurements in the vicinity of M3 and the model simulation it can be concluded that the simulation underestimates the strength of the MSLP anomaly associated with M3, while comparisons to other surface observations yield good agreement (not shown).

Fig. 4.

Buoy measurements of MSLP (thick solid line: buoy 54 807, northern buoy in Fig. 2b; thick dashed line: buoy 54 808, southern buoy in Fig. 2b) and LAM50 simulations at the nearest grid point (thin lines without symbols). The thin line with triangles displays the MSLP simulations at a grid point about 300 km southeast of buoy 54 808. Vertical bars mark the times of the AVHRR data shown in Fig. 2.

Fig. 4.

Buoy measurements of MSLP (thick solid line: buoy 54 807, northern buoy in Fig. 2b; thick dashed line: buoy 54 808, southern buoy in Fig. 2b) and LAM50 simulations at the nearest grid point (thin lines without symbols). The thin line with triangles displays the MSLP simulations at a grid point about 300 km southeast of buoy 54 808. Vertical bars mark the times of the AVHRR data shown in Fig. 2.

4. Satellite-based mesoscale analysis

The spatial resolution of SSM/I retrievals of typically 50 km (see section 2a) is well suited for the analysis of mesoscale phenomena. The good temporal and spatial coverage in polar regions of the two SSM/I sensors in orbit in January 1995 permits following the development of the MCs with four sets of overpasses daily (Table 3). In the following, the analysis will be concentrated on three atmospheric parameters retrieved from SSM/I measurements: the two column-integrated parameters CLW and IWV and, additionally, the near-surface wind speed (WS). Combinations of three subsequent SSM/I overpasses are presented in the figures to get an optimal coverage of the regions of interest.

Figure 5 shows SSM/I-derived CLW and WS for 1000–1300 UTC 11 January 1995 for ice-free ocean areas and covering the areas of the three MCs. The sea ice coverage is marked by the isoline of 50% ice concentration. The areas south of 70°S, where high CLW is detected, are mainly artifacts. If the SSM/I sea ice algorithm detects polynias with less than 15% ice concentration, the atmospheric parameters are retrieved but the results are usually highly contaminated due to pixels erroneously detected ice-free or due to the satellite footprint partly covering the ice edge. The three MCs—M3 (57°S, 113°W), M2 (59°S, 100°W), and M1 (61°S, 85°W)—are clearly marked by bands of high CLW. Mesoscale cyclone M3 had developed a pronounced comma-shaped cloud band with high CLW values. In contrast to AVHRR images, where only the highest cloud layers contribute to the radiance signal, the SSM/I-derived CLW represents the whole atmospheric column. Compared to the infrared imagery (Fig. 2) the baroclinic cloud structure of M3 is much clearer in the CLW field, because the high clouds between the cloud bands northeast of the center in Fig. 2b turn out to be cirrus clouds, which are almost transparent at microwave frequencies.

Fig. 5.

SSM/I-derived cloud liquid water, near-surface wind speed, and integrated water vapor for 1000–1300 UTC 11 Jan 1995 (time of orbits from east to west: 0945, 1126, 1308 UTC). (a) CLW (grayshaded, scale in kg m−2) and WS (contour interval 2 m s−1) over ice-free ocean with coastline and sea ice front (thick isoline of 50% ice concentration). (b) IWV (isolines every 1 kg m−2) over ice-free ocean with coastline and sea ice coverage (grayshaded, light shading: >50% ice concentration).

Fig. 5.

SSM/I-derived cloud liquid water, near-surface wind speed, and integrated water vapor for 1000–1300 UTC 11 Jan 1995 (time of orbits from east to west: 0945, 1126, 1308 UTC). (a) CLW (grayshaded, scale in kg m−2) and WS (contour interval 2 m s−1) over ice-free ocean with coastline and sea ice front (thick isoline of 50% ice concentration). (b) IWV (isolines every 1 kg m−2) over ice-free ocean with coastline and sea ice coverage (grayshaded, light shading: >50% ice concentration).

The WS field (isolines in Fig. 5a) shows the three MCs being associated with zones of WS exceeding 14 m s−1, their maxima extending to the west of the corresponding cloud bands. In contrast, only weak winds were present in large areas south of M3 (values around 6 m s−1). North of the center of M3 an extended zone with high WS up to 20 m s−1 was present, and a very pronounced shear zone with a gradient of 12 m s−1 over 200 km was associated with the southern part of the cloud band. In accordance with the AVHRR image in Fig. 2b, the SSM/I-derived CLW shows the cloud band of M1 as an area of higher CLW content and the associated WS field shows higher values over a larger region compared to 10 h earlier (not shown). The cloud band of M2 is also represented in the CLW field (but with slightly lower values than for 0300 UTC, not shown). The WS field stayed constant in the area of M2.

The IWV field in Fig. 5b clearly shows the signature of the frontal zones of M3, separating dry polar air with values of 7–8 kg m−2 and the midlatitude humid air with values of about 11–12 kg m−2. Both M2 and M1 are also visible as IWV gradients but being weaker than for M3. The main front of L had moved over the sea ice and the Antarctic Peninsula so that the cloud band cannot be seen any longer in the SSM/I-derived fields. The high IWV gradients associated with M3 coincide perfectly with the cloud bands of M3, while for M2 and M1 the IWV field shows only rather broad areas with increased humidity.

At 0115 UTC 12 January 1995 M3 had moved to 57°S, 98°W (Fig. 6a). The central area is marked by high WS exceeding 18 m s−1, moderate CLW values, and a broad local maximum of IWV content (Fig. 6b). On the eastern side of M3 two CLW bands extended from south to north (along 95°W) and from west to east (along 59°S) being coincident with the two frontal zones marked by strong IWV gradients. The high clouds visible in this sector of humid air in the IR image (Fig. 2c) again consisted mainly of cirrus clouds, since low values of CLW dominated this region in Fig. 6a. During the next 12 h M3 dissipated while moving farther eastward toward the Drake Passage (not shown). The cloud structure in CLW and the gradients in IWV were less pronounced while the WS was found to have high values of 16–18 m s−1 over a large area.

Fig. 6.

As in Fig. 5 but for 0000–0300 UTC 12 Jan 1995 (orbits 2336, 0116, 0257 UTC).

Fig. 6.

As in Fig. 5 but for 0000–0300 UTC 12 Jan 1995 (orbits 2336, 0116, 0257 UTC).

The spatial resolution of the wind products of ERS-SCAT is comparable to the SSM/I resolution, but the data are of only limited use for mesoscale analyses because of the poor spatial coverage of the sensor (cf. section 2a). The chance of hitting MCs of typical sizes between 200 and 1000 km with a swath width of 500 km and gaps between subsequent orbits of 900 km (at 60°S) is rather low. For this case study none of the MCs was fully covered by the scatterometer swaths during its lifetime. As an example for a typical coverage Fig. 7 shows the wind vectors for the three ERS-1 overpasses between 0600 and 0900 UTC 11 January 1995 superimposed on the grayshaded SSM/I-derived WS field for 0400–0700 UTC. ERS-SCAT wind vectors with wind speeds less than 2 m s−1 are rejected by the retrievals because of the poor quality of the retrieval for low wind speeds. The SSM/I wind field represents an intermediate time between the situations shown in Fig. 2a and Figs. 2b and 5a. The wind maxima associated with the three MCs at that time can be found between 55° and 60°S at longitudes of 120°, 108°, and 95°W, respectively, which again reflects the zonal movement of these wind patterns with a speed of about 15 m s−1. As in Fig. 5a, the highest wind speeds (exceeding 20 m s−1) can be found near M3 (at 57°S, 120°W). The circulation around the center of L is clearly visible in the measured near-surface wind field. The low wind speeds in the area west of 110°W and south of 62°S are nicely reproduced in both data types. The center of M3 at 57°S, 120°W was not covered by the ERS-SCAT swath, but ERS-SCAT shows the strong northwesterly flow east of M3. For M2 (approximately at 59°S, 108°W) only parts of the associated high wind speed area are covered by the overpass, but no mesoscale circulation was present in the ERS-SCAT wind vectors. This indicates that M2 can be regarded as a less significant vortex with only a weak surface circulation.

Fig. 7.

Near-surface wind vectors determined from ERS-SCAT data for 0600–0900 UTC 11 Jan 1995 (time of orbits from east to west: 0543, 0724, 0904) and SSM/I-derived WS for 0400–0700 UTC (orbits 0350, 0530, 0709 UTC) over ice-free ocean with coastline and sea ice front (thick isoline of 50% ice concentration). Only every fourth wind vector and only vectors with more than 2 m s−1 are plotted. A scaling vector is shown in the lower-right corner. Areas with SSM/I WS of more than 8 m s−1 are grayshaded (scale in m s−1).

Fig. 7.

Near-surface wind vectors determined from ERS-SCAT data for 0600–0900 UTC 11 Jan 1995 (time of orbits from east to west: 0543, 0724, 0904) and SSM/I-derived WS for 0400–0700 UTC (orbits 0350, 0530, 0709 UTC) over ice-free ocean with coastline and sea ice front (thick isoline of 50% ice concentration). Only every fourth wind vector and only vectors with more than 2 m s−1 are plotted. A scaling vector is shown in the lower-right corner. Areas with SSM/I WS of more than 8 m s−1 are grayshaded (scale in m s−1).

5. Numerical model-based analysis and comparisons with satellite data

a. Mesoscale model-based analysis

As discussed in section 3, the LAM50 simulations of the MSLP have shown M3 and M1 as mesoscale troughs moving eastward with the mean zonal flow. Mesoscale cyclone M3 is found to be best developed for the 18-h simulation valid at 1800 UTC 11 January. Figure 8 shows the relative vorticity at 950 hPa at that time for a subsection of the LAM50 domain. Several pronounced cyclonic vorticity maxima are present. The first one at 70°S, 90°W is associated with the synoptic-scale cyclone L (values of −1.6 × 10−4 s−1). The second one with values of −2.8 × 10−4 s−1 can be found east of the Antarctic Peninsula associated with the lee cyclone. Both M3 and M1 can be identified as cyclonic vorticity maxima at 59°S, 101°W (−2.0 × 10−4 s−1) and at 59°S, 76°W (−2.0 × 10−4 s−1), respectively. Mesoscale cyclone M2 has neither been simulated as a surface trough nor as a distinct mesoscale cyclonic vorticity maximum, but lies in a belt of moderate cyclonic vorticity south and southeast of M3. The comparison of the vorticity centers with the MC positions obtained from the satellite images reveals that the zonal translation is too fast in the simulation and that the simulated MC centers are about 300 km southeast of the observed ones. This overestimation of the eastward translation in the simulations could be caused by a too-strong zonal wind in the inital and boundary fields from the ECMWF operational model analyses.

Fig. 8.

Relative vorticity at 950 hPa (positive values dashed, contour interval 0.4 × 10−4 s−1, shaded for values lower than −1.0 × 10−4 s−1) after 18 h simulation time (valid at 1800 UTC 11 Jan) for a subsection of the LAM50 domain. The center of M3 at 1800 UTC 11 Jan obtained from satellite images is marked by the full circle.

Fig. 8.

Relative vorticity at 950 hPa (positive values dashed, contour interval 0.4 × 10−4 s−1, shaded for values lower than −1.0 × 10−4 s−1) after 18 h simulation time (valid at 1800 UTC 11 Jan) for a subsection of the LAM50 domain. The center of M3 at 1800 UTC 11 Jan obtained from satellite images is marked by the full circle.

A more detailed mesoscale analysis for M3 will now be given using LAM25 forecasts nested in the LAM50 domain (see Fig. 1b). LAM25 starts with the 6-h forecast of LAM50 and uses the LAM50 forecasts as boundary conditions during the whole simulation. The LAM25 simulation after 12 h (valid at 1800 UTC 11 January) is shown in Fig. 9. The wind vector field and potential temperature at 850 hPa (Fig. 9a) reflect the mesoscale circulation of M3 (superimposed on the strong zonal wind) with the circulation center at about 60°S, 100°W. The temperature field reveals moderate baroclinicity and a wavelike structure. The result is weak cold/warm advection northwest/northeast of the center; that is, the simulated structure is similar to that of a short baroclinic wave. The intensification of the low-level circulation of M3 is triggered by a short-wave 500-hPa trough (Fig. 9b). The maximum upward vertical velocity at 700 hPa lies downstream of the 500-hPa trough axis and over the low-level circulation center. A belt of pronounced lifting extends to the east and northeast with a maximum of 2.2 Pa s−1. The areas with a cloud coverage of more than 90% at 500 hPa reveal similarities to the midlevel clouds associated with M3 on the satellite image 4 h earlier (Fig. 2b). The same kind of triggering is simulated also for M1, but the upward vertical velocities are much smaller (0.8 Pa s−1, not shown).

Fig. 9.

LAM25 simulations after 12 h simulation time (valid at 1800 UTC 11 Jan): (a) Potential temperature (solid, isolines every 1 K) and wind vectors (every grid point) at 850 hPa. A scaling vector is shown in the lower-right corner. (b) Geopotential height at 500 hPa (solid, isolines every 20 gpm) and upward vertical velocity in pressure coordinates (dashed, isolines every 20 × 10−2 Pa s−1) at 700 hPa for the same area as in (a). Shaded areas indicate simulated cloud coverage at 500 hPa exceeding 90%.

Fig. 9.

LAM25 simulations after 12 h simulation time (valid at 1800 UTC 11 Jan): (a) Potential temperature (solid, isolines every 1 K) and wind vectors (every grid point) at 850 hPa. A scaling vector is shown in the lower-right corner. (b) Geopotential height at 500 hPa (solid, isolines every 20 gpm) and upward vertical velocity in pressure coordinates (dashed, isolines every 20 × 10−2 Pa s−1) at 700 hPa for the same area as in (a). Shaded areas indicate simulated cloud coverage at 500 hPa exceeding 90%.

b. Comparisons of near-surface wind

The model results allow a comparison to ERS and SSM/I retrievals presented in section 4. Although the ERS-SCAT overpasses did not cover the areas of M1 and M3, the simulated near-surface wind field structure in the area of the ERS-SCAT swaths can be compared to the ERS wind vectors. Figure 10 shows the simulated near-surface wind field at 0600 UTC 11 January for the region presented in Fig. 7. The thick lines mark the borders of ERS-SCAT overpasses, regions with wind speeds higher than 8 m s−1 are grayshaded as in Fig. 7. There is a very good agreement in the wind directions and the general structure of the wind speed distribution between model simulation and the satellite data. Some small differences in wind direction can be found west of L (67°S, 73°W), south of M2 (60°S, 110°W), and in the region 67°S, 125°W, while the latter deviation probably is caused by the difference in time of 3 h between the westernmost satellite overpass and the simulation.

Fig. 10.

The 6-h prognoses valid at 0600 UTC 11 Jan of the horizontal wind field at the lowest level of the model (30 m above the surface) for a subsection of the LAM50 domain; the shaded areas indicate the wind speed (lightest shading, 8–12 m s−1; darkest shaded, >20 m s−1). The center of M3 at 0400 UTC 11 Jan obtained from satellite images is marked by the full circle and the thick lines indicate the positions of the ERS-SCAT swaths for 0600–0900 UTC shown in Fig. 7.

Fig. 10.

The 6-h prognoses valid at 0600 UTC 11 Jan of the horizontal wind field at the lowest level of the model (30 m above the surface) for a subsection of the LAM50 domain; the shaded areas indicate the wind speed (lightest shading, 8–12 m s−1; darkest shaded, >20 m s−1). The center of M3 at 0400 UTC 11 Jan obtained from satellite images is marked by the full circle and the thick lines indicate the positions of the ERS-SCAT swaths for 0600–0900 UTC shown in Fig. 7.

The quantitative comparison of the simulated near-surface wind at 0600 UTC and the satellite data closest in time (orbit 0543 UTC) shows that the model underestimates the wind speed by 0.6 m s−1 with a standard deviation of 1.3 m s−1, which is in accordance with the 2 m s−1 accuracy of the CERSAT algorithm. The difference in wind direction is 2.6° with a standard deviation of 25.4°. The collocation of four orbits with time differences less than 1 h to the corresponding model wind fields (Table 4) shows a mean difference for the wind speed (satellite measurement minus simulation) of 1.1 m s−1 with standard deviation 2.5 m s−1, and for the wind direction a bias of −3.2° with a standard deviation of 25.3°.

Table 4.

Difference of wind speed and wind direction between ERS-SCAT measurements and the 10-m wind of the LAM50 simulation (bias = satellite − model) for four ERS-1 satellite orbits. N = number of grid points used; Δt = time difference.

Difference of wind speed and wind direction between ERS-SCAT measurements and the 10-m wind of the LAM50 simulation (bias = satellite − model) for four ERS-1 satellite orbits. N = number of grid points used; Δt = time difference.
Difference of wind speed and wind direction between ERS-SCAT measurements and the 10-m wind of the LAM50 simulation (bias = satellite − model) for four ERS-1 satellite orbits. N = number of grid points used; Δt = time difference.

The comparison of the wind speed of the simulations at the lowest sigma level (at approximately 30 m) of the LAM50 model at 0600 UTC (Fig. 10) and the SSM/I-derived near-surface WS at 19.5 m (Fig. 7) again shows that the wind speed structures of the model and the satellite retrievals agree well. A quantitative comparison between simulated and SSM/I-derived WS was also carried out. In order to reduce the collocation errors, 3-hourly model fields were interpolated in time to the times of the corresponding SSM/I overpasses. The differences were then calculated only for ocean areas far away from the coast and the sea ice edge in order to minimize the influence of SSM/I retrieval errors due to erroneous ice detection. For the situations presented in Fig. 7 and Fig. 10, the underestimation of the wind speed by the model is found to be 0.8 m s−1 with a standard deviation of 3.1 m s−1, which is clearly higher than the typical retrieval accuracy of 2 m s−1. The comparison of all SSM/I-derived wind fields during the complete LAM50 simulation yields a mean bias of 0.3 m s−1 with a standard deviation of 2.9 m s−1 (Table 5).

Table 5.

Difference between SSM/I-derived near-surface WS (reference height 19.5 m) and WS of the LAM50 simulation (lowest level, 30 m). SSM/I pixels were chosen to be far enough from coast and sea ice to avoid retrieval errors. The model data were interpolated in time to the corresponding SSM/I orbits. N = number of grid points used.

Difference between SSM/I-derived near-surface WS (reference height 19.5 m) and WS of the LAM50 simulation (lowest level, 30 m). SSM/I pixels were chosen to be far enough from coast and sea ice to avoid retrieval errors. The model data were interpolated in time to the corresponding SSM/I orbits. N = number of grid points used.
Difference between SSM/I-derived near-surface WS (reference height 19.5 m) and WS of the LAM50 simulation (lowest level, 30 m). SSM/I pixels were chosen to be far enough from coast and sea ice to avoid retrieval errors. The model data were interpolated in time to the corresponding SSM/I orbits. N = number of grid points used.

In contrast to these results, the difference between ERS-SCAT and SSM/I retrievals for 10 to 12 January 1995 (Table 6) lies in the expected error range with a mean bias of 1.1 m s−1 and a standard deviation of 1.2 m s−1. The slightly higher WS of the SSM/I retrieval may be due to different reference heights of the satellite retrievals (SSM/I, 19.5 m; ERS-SCAT, 10 m). The high correlation of both datasets (correlation coefficient R = 0.91) and an approximately constant bias for the complete WS range in the scatterplot (not shown) confirm this explanation for the mean bias.

Table 6.

Difference of wind speed derived from SSM/I data and ERS-SCAT measurements, interpolated on the LAM50 grid. Additionally the results of comparison for subsets of the data are given for rain flag, CLW, and IWV classes (see text). Bias = SSM/I–ERS-SCAT. R = correlation coefficient. Δt = time difference, averages are weighted with number of grid points.

Difference of wind speed derived from SSM/I data and ERS-SCAT measurements, interpolated on the LAM50 grid. Additionally the results of comparison for subsets of the data are given for rain flag, CLW, and IWV classes (see text). Bias = SSM/I–ERS-SCAT. R = correlation coefficient. Δt = time difference, averages are weighted with number of grid points.
Difference of wind speed derived from SSM/I data and ERS-SCAT measurements, interpolated on the LAM50 grid. Additionally the results of comparison for subsets of the data are given for rain flag, CLW, and IWV classes (see text). Bias = SSM/I–ERS-SCAT. R = correlation coefficient. Δt = time difference, averages are weighted with number of grid points.

Since the ERS-SCAT retrievals are obtained at a much lower frequency of the microwave spectrum than the SSM/I sensor, they are less sensitive to cloud and rain contamination. A comparison of both WS retrievals for subsets of the data separated by weather conditions allows a more differentiated analysis of the SSM/I retrieval quality. For this analysis the interpolation of SSM/I wind speeds to the 50-km grid is carried out for satellite pixels selected by SSM/I-retrieved weather conditions in terms of rain flag, CLW, or IWV. The results of the comparison to the ERS-SCAT retrievals given in Table 6 underline the role of the rain flag in the SSM/I WS algorithm. As explained in section 2a, the accuracy of the retrieved WS with rain flag 1 (according to Goodberlet and Swift 1992) is lower than for fair weather conditions (rain flag 0), which can be seen in a moderately higher bias and standard deviation and a lower correlation. But in this case study the error is still much lower than the typical 2–5 m s−1 for flag 1. The situation is very similar when distinguishing between low and moderate CLW content. The influence of IWV on the WS retrieval seems to be negligible regarding the standard deviation but is clearly visible for the correlation and the bias, which changes sign. The difference in bias can be attributed to the differences in the general wind distribution of the two datasets; the mean wind conditions are 8.0 ± 2.2 m s−1 for the humid and 11.6 ± 2.7 for the dry subset. It can be concluded that the retrieval quality of the SSM/I wind speeds used in this study is only slightly influenced by water vapor and cloud water conditions (in the range where the retrieval algorithm is applicable).

While the bias of WS between the model simulation and both satellite sources is relatively small, the standard deviation of the differences is higher than the global retrieval accuracies for both satellite types. This seems to be largely a result of a time lag or spatial displacement between the development in reality and in the simulation. Also the mesoscale structures not adequately reproduced by the simulation contribute to the error. This is shown in the following by a more detailed qualitative comparison of WS and IWV derived from SSM/I and the numerical simulations for a subsection of the LAM25 domain.

Figure 11 shows results after 6 h of simulation of LAM25 (valid at 1200 UTC 11 January) for the mature stage of M3. The comparison with SSM/I WS for 1000–1300 UTC in Fig. 5a and WS from LAM25 simulations (Fig. 11a) shows the same good general agreement in the WS values as LAM50 6 h before: the low wind speed areas (<6 m s−1) in the sector 100°–120°W along 65°S and the high wind speed zones north of the center of L (>14 m s−1), in the area of M1 (>16 m s−1) and for M3 (>16 m s−1). Larger differences between simulation results and satellite observations occur in the position of the wind maxima associated with L and M3. The simulated WS maximum of M3 (at 58°S, 108°W) is positioned between the maxima of the observed M3 (at 57°S, 113°W) and of the observed M2 (at 59°S, 100°W), which was not simulated by the model.

Fig. 11.

(a) Simulated wind speed at the lowest model level (isolines every 2 m s−1) and (b) IWV (isolines every 1 kg m−2) after 6 h simulation time for a subsection of the LAM25 domain. The simulations are valid at 1200 UTC 11 Jan. The center of M3 at 1400 UTC 11 Jan obtained from satellite images is marked by the full circle and L indicates the center of a synoptic-scale cyclone. Land and sea ice areas are shaded dark gray and light gray, respectively.

Fig. 11.

(a) Simulated wind speed at the lowest model level (isolines every 2 m s−1) and (b) IWV (isolines every 1 kg m−2) after 6 h simulation time for a subsection of the LAM25 domain. The simulations are valid at 1200 UTC 11 Jan. The center of M3 at 1400 UTC 11 Jan obtained from satellite images is marked by the full circle and L indicates the center of a synoptic-scale cyclone. Land and sea ice areas are shaded dark gray and light gray, respectively.

c. IWV comparisons

The same findings can also clearly be seen in the comparison of IWV retrieved from SSM/I (Fig. 5b) and of IWV of the LAM25 simulation (Fig. 11b). The IWV values and the structure of the IWV field near M3 are very similar in the model and observations, but with a difference of 300 km in position. The second obvious difference is the position of the main frontal zone of L. In the SSM/I retrieval (time of orbit covering the eastern part of Fig. 5: 0945 UTC) the front already has reached or passed the Antarctic Peninsula while in the simulations valid at 1200 UTC the high IWV gradients of the frontal zone still lie northwest of this barrier. In contrast to the SSM/I retrievals, the model results show a much clearer structure in IWV and WS for M1.

As a general tendency for the whole model area, a slightly drier atmosphere is observed in the SSM/I retrieval than in the simulation. Looking at the comparison of the IWV fields during the complete LAM50 simulation (Table 7) supports this result. The mean bias (SSM/I retrieval minus simulation) is −2.5 kg m−2 with a standard deviation of 2.93 kg m−2. For this calculation the LAM50-simulated IWV fields were again interpolated in time to the corresponding satellite orbits.

Table 7.

Difference between SSM/I-derived IWV and IWV of the LAM50 simulation. The model data were interpolated in time to the corresponding SSM/I overpasses. N = number of grid points used.

Difference between SSM/I-derived IWV and IWV of the LAM50 simulation. The model data were interpolated in time to the corresponding SSM/I overpasses. N = number of grid points used.
Difference between SSM/I-derived IWV and IWV of the LAM50 simulation. The model data were interpolated in time to the corresponding SSM/I overpasses. N = number of grid points used.

The relative good agreement between the satellite-retrieved and the corresponding simulated parameters gives confidence in the quality of the numerical simulations. This is important, since only few observational data enter the ECMWF analyses of the area of interest, which can lead to large errors in the analyses of different weather centers for this area, as was demonstrated for several situations during the FROST SOPs (Turner et al. 1996). The main shortcomings of the simulations with respect to M3 are the underestimation of the pressure anomaly (compared to buoy data) and a faster zonal translation (compared to satellite data).

6. Summary and conclusions

The development of a group of mesoscale cyclones over the northern Amundsen and Bellingshausen Seas from 11 to 12 January 1995 was studied by means of satellite and numerical model data. AVHRR data, ERS, and SSM/I retrievals were used for the observational description of the synoptic and subsynoptic environment associated with the development of the MCs. The most intense MC (M3) had a diameter of about 800 km, a lifetime of more than 24 h, and reached the intensity of a polar low. SSM/I wind speeds showed values up to 20 m s−1 in the vicinity of this MC. The development phase of M3 took place during 11 January, and a distinct signal in the infrared imagery and in SSM/I retrievals of CLW, wind speed, and IWV can be found. The development occcurred over the Amundsen Sea in a strong zonal flow west of a main synoptic-scale cyclone with its center over the Bellingshausen Sea. The overall appearance of M3 on the satellite images and in the fields of satellite-retrieved parameters resembled a short baroclinic wave. In analogy to synoptic-scale cyclones, the fronts of M3 were clearly marked by the SSM/I-derived IWV gradients and were associated with strong wind shear.

ERS-SCAT wind vectors give no insight into the structure of the MCs, because of the narrowness of the ERS-SCAT swaths, but can be used to validate numerical simulations performed using the mesoscale model NORLAM with horizontal resolutions of 50 and 25 km. The comparison between the numerical model simulations and the corresponding satellite-retrieved parameters IWV and near-surface wind (SSM/I and ERS-SCAT) shows a relative good agreement. This finding gives confidence in the quality of the numerical simulations with NORLAM and the quality of the ECMWF analyses, in which the limited area model is nested. A more general assessment of the operational analyses during FROST for the whole Antarctic and for longer timescales is given in Turner et al. (1999).

In accordance with the appearance of the MC M3 in the satellite-based study, the numerical model results show M3 as a short-wave baroclinic development triggered by an upper-level trough. Although the comparisons between model simulations and buoy MSLP measurements reveal an underestimation of the pressure anomaly of M3 by the model, a clear low-level circulation associated with M3 is simulated. The model simulations show M1 also being forced by the upper-level flow, but the forcing is much weaker than for M3. The developments occurred far away from the sea ice front or topographic structures. The synoptic-scale environment is quite different from that associated with summertime MC developments over the Weddell Sea region (Heinemann 1990) but has similarities to wintertime polar lows over the Weddell Sea (Heinemann 1996b) as well as over the Ross and Bellingshausen Seas (Carleton and Fitch 1993).

Fig. 2.

(Continued) White squares indicate positions of drifting buoys (northern buoy: 54 807, southern buoy: 54 808). (c) As in (b) but for the full-resolution HRPT image. (d) As in (a) but for 0200 and 0500 UTC 12 Jan.

Fig. 2.

(Continued) White squares indicate positions of drifting buoys (northern buoy: 54 807, southern buoy: 54 808). (c) As in (b) but for the full-resolution HRPT image. (d) As in (a) but for 0200 and 0500 UTC 12 Jan.

Acknowledgments

This research was supported by the Deutsche Forschungsgemeinschaft under Grant Kr228/16. The authors thank the Norwegian Meteorological Institute (Oslo) for support. AVHRR data was made available by the FROST database (BAS) and by NOAA/NESDIS. Analyses for the simulations were provided by the ECMWF (Reading, United Kingdom). SSM/I data were obtained from the MSFC DAAC (Alabama), and ERS-1 data were made available by IFREMER (Brest, France).

REFERENCES

REFERENCES
Alishouse, J. C., J. B. Snider, E. R. Westwater, C. T. Swift, C. S. Ruf, A. S. Snyder, J. Vongsathorn, and R. R. Ferraro, 1990: Determination of cloud liquid water content using the SSM/I. IEEE Trans. Geosci. Remote Sens., 28, 817–822.
Bromwich, D. H., R. I. Cullather, and R. W. Grumbine, 1999: An assessment of the NCEP operational global spectral model forecasts and analyses for Antarctica during FROST. Wea. Forecasting, 14, 835–850.
Carleton, A. M., and M. Fitch, 1993: Synoptic aspects of Antarctic mesocyclones. J. Geophys. Res., 98, 12 997–13 018.
Carrasco, J. F., and D. H. Bromwich, 1993: Mesoscale cyclogenesis dynamics over the southwestern Ross Sea. J. Geophys. Res., 98, 12 973–12 995.
Cavalieri, D. J., K. M. St. Germain, and C. T. Swift, 1995: Reduction of weather effects in the calculation of sea-ice concentration with the DMSP SSM/I. J. Glaciol., 41, 455–464.
Claud, C., N. A. Scott, and A. Chedin, 1992: Use of TOVS observations for the study of polar and Arctic lows. Int. J. Remote Sens., 13, 129–139.
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 Sys., 4, 181–208.
Goodberlet, M. A., and C. T. Swift, 1992: Improved retrievals from the DMSP wind speed algorithm under adverse weather conditions. IEEE Trans. Geosci. Remote Sens., 30, 1076–1077.
Grønås, S., A. Foss, and M. Lystad, 1987: Numerical simulations of the polar lows in the Norwegian Sea. Tellus, 39A, 334–353.
Heinemann, G., 1990: Meso-scale vortices in the Weddell Sea region. Mon. Wea. Rev., 118, 779–793.
——, 1996a: A wintertime polar low over the eastern Weddell Sea (Antarctica): A study with AVHRR, TOVS, SSM/I and conventional data. Meteor. Atmos. Phys., 58, 83–102.
——, 1996b: On the development of wintertime meso-scale cyclones near the sea ice front in the Arctic and Antarctic. Global Atmos.–Ocean Sys., 4, 89–123.
——, 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.
——, S. Noël, A. Chedin, N. A. Scott, and C. Claud, 1995: Sensitivity studies of TOVS retrievals with 3I and ITPP retrieval algorithms:Application to the resolution of meso-scale phenomena in the Antarctic. Meteor. Atmos. Phys., 55, 87–100.
Heygster, G., and Coauthors, 1996: PELICON—Project for estimation of long-term variability in ice concentration. Final Report for the European Community, Contract EV5V-CT93-0268 (DG 12 DTEE), University of Bremen, 188 pp. [Available from Institut für Fernerkundung, Universität Bremen, Postfach 330 440, D-28 334 Bremen, Germany.]
.
Hollinger, J. P., R. Lo, G. Poe, R. Savage, and J. Pierce, 1987: Special Sensor Microwave/Imager user’s guide. Rep. 579 269, Naval Research Laboratory, 120 pp. [Available from Naval Research Laboratory, 4555 Overlook Avenue S.W., Washington, DC 20375-5000.]
.
Hutchinson, H., and Coauthors, 1999: On the reanalysis of Southern Hemisphere Charts for the FROST project. Wea. Forecasting, 14, 909–919.
Katsaros, K. B., I. Batty, L. A. McMurdie, and G. W. Petty, 1989: Identification of atmospheric fronts over the ocean with microwave measurements of water vapor and rain. Wea. Forecasting, 4, 449–460.
Lieder, M., and G. Heinemann, 1996: Satellite investigation of Antarctic Mesocyclones over the Southern Pacific and the Bellingshausen Sea. Extended Abstracts, Workshop on Theoretical and Observational Studies of Polar Lows, St. Petersburg, Russia, European Geophysical Society Polar Lows Working Group, 55–60. [Available from G. Heinemannn, Meteorologisches Institut der Universität Bonn, Auf dem Hügel 20, D-53 127 Bonn, Germany.]
.
McMurdie, L. A., C. Claud, and S. Atakturk, 1997: Satellite-derived atmospheric characteristics of spiral and comma-shaped Southern Hemisphere mesocyclones. J. Geophys. Res., 102, 13 889–13 905.
NOAA, 1995: TerrainBase Global DTM. CD-ROM, version 1.0. [Available from NESDIS National Geophysical Data Center, Boulder, CO 80303-3328.]
.
Nordeng, T. E., 1990: Model-based diagnostic study of the development and maintenance mechanism of two polar lows. Tellus, 42A, 92–108.
——, and E. A. Rasmussen, 1992: A most beautiful polar low. A case study of a polar low development in the Bear Island region. Tellus, 44A, 81–99.
Petty, G. W., 1993: SSM/I algorithms for the estimation of column water vapor. Proc. Shared Processing Network DMSP Special Sensor Microwave Imager Symp., Naval Postgraduate School, Monterey, CA.
——, 1994: Physical retrievals of over-oceans rain rate from multichannel microwave imagery. Part II: Algorithm implementation. Meteor. Atmos. Phys., 54, 101–121.
Turner, J., T. A. Lachlan-Cope, D. E. Warren, and C. N. Duncan, 1993: A meso-scale vortex over Halley station, Antarctica. Mon. Wea. Rev., 121, 1317–1336.
——, and Coauthors, 1996: The Antarctic First Regional Observing Study of the Troposphere (FROST) project. Bull. Amer. Meteor. Soc., 77, 2007–2032.
——, S. Leonard, G. J. Marshall, M. Pook, L. Cowled, R. Jardine, S. Pendlebury, and N. Adams, 1999: An assessment of operational antarctic analyses based on data from the FROST project. Wea. Forecasting, 14, 817–834.

Footnotes

Corresponding author address: Dr. Günther Heinemann, Meteorologisches Institut, der Universität Bonn, Auf dem Hügel 20, 53121 Bonn, Germany