JUNE 2001

315

WINSTEAD ET AL.

Synthetic Aperture Radar Observations of the Surface Signatures of Cold-Season Bands over the Great Lakes NATHANIEL S. WINSTEAD,* ROBERT M. SCHAAF,

AND

PIERRE D. MOURAD

Applied Physics Laboratory, University of Washington, Seattle, Washington (Manuscript received 12 April 2000, in final form 28 December 2000) ABSTRACT An important aspect of operational meteorology in and around the Great Lakes region of the United States and Canada in the winter months is the forecasting of lake-effect precipitation. While the synoptic- and mesoscale processes that govern the development of lake-effect precipitation have been well understood for many years, problems observing these bands remain because of the limited boundary layer coverage provided by the Weather Surveillance Radar-1988 Doppler (WSR-88D) network. While traditional visible and infrared satellite imagery helps alleviate these coverage limitations, overcast conditions often negate this advantage. Here, a new method for observing lake-effect bands by using synthetic aperture radar (SAR) to identify and characterize their surface signatures is presented. SAR is a remote sensing tool that images surface roughness. Over water, this roughness is related to the surface wind stress and, hence, surface wind field. Here, three cases are documented where the SAR aboard the Canadian Radar Satellite-1 imaged the footprints of precipitating bands over the Great Lakes: one case with multiple snowbands west of one main band over Lake Superior, and two cases with shore-parallel bands over each of Lakes Ontario and Michigan. These cases are first documented using traditional observing methods: infrared satellite imagery, WSR-88D, and surface observations. Then, each SAR image is interpreted based upon the traditional observations. The ultimate goal is to demonstrate that SAR is capable of detecting the surface signatures associated with Great Lakes precipitation bands that could be of value to forecasters when data from traditional observation platforms are unavailable.

1. Introduction An important aspect of wintertime operational meteorology in and around the Great Lakes region of the United States and Canada is the forecasting of lakeeffect precipitation (Niziol 1987; Reinking et al. 1993; Niziol et al. 1995). Lake-effect precipitation can fall as either rain or snow. It can take the form of cellular convective showers, multiple bands or rolls (e.g., Kelly 1986; Kristovich 1993; Kristovich and Steve 1995), mesoscale vortices (e.g., Forbes and Merritt 1984; Pease et al. 1988; Laird 1999; Minor et al. 2000), and single shore-parallel bands (e.g., Lavoie 1972; Passarelli and Braham 1981; Kristovich and Steve 1995). The specific morphology of a given storm is determined largely by the direction of flow over the lakes, the background stability, the profile of vertical wind shear over the lake, and the specific geometry of the affected lake (Kristovich and Steve 1995). In general, the synoptic- and me* Current affiliation: Applied Physics Laboratory, The Johns Hopkins University, Baltimore, Maryland. Corresponding author address: Pierre D. Mourad, Applied Physics Laboratory, University of Washington, 1013 NE 40th Street, Seattle, WA 98105. E-mail: [email protected]

q 2001 American Meteorological Society

soscale processes that govern the conditions favorable for the development of lake-effect snow have been well understood for many years (e.g., Petterson and Calabrese 1959). Therefore, operational forecasting techniques designed to predict the development, morphology, and location of lake-effect snowbands exist and have continued to improve (Niziol 1987; Niziol et al. 1995). Unfortunately, the scale of these bands is almost always subsynoptic leading to difficulties in observing them without relying heavily on visible/IR satellite imagery, and the Weather Surveillance Radar-1988 Doppler (WSR-88D). While these platforms are critical components of the operational observation network, they are, under certain conditions, insufficient for observing lake-effect precipitation. For example, WSR-88D often overshoots the boundary layer, a problem that is particularly detrimental over certain areas of the Great Lakes. See NRC (1995) for a general discussion of this issue and Nicosia et al. (1999) for a specific example of a case where WSR-88D overshot a lake-effect rain band. Visible and IR satellite imagery fill this gap; however, as reported in Reinking et al. (1993), it is not unusual for lake-effect snowbands to be hidden under a layer of higher clouds. When both WSR-88D and visible and IR imagery are not of use, the operational forecaster/researcher needs other tools to observe ongoing lake-effect circulations.

316

WEATHER AND FORECASTING

VOLUME 16

FIG. 1. AVHRR image (channel 4) of Lake Superior taken at 1134 UTC 9 Jan 1999. The image has been calibrated and converted to brightness temperature using ENVI image processing software. In addition, the image has been projected onto a polar stereographic grid with a 1.1-km spatial resolution. The locations of the dominant band and areas of multiple bands are labeled.

Here, we demonstrate such a tool, one useful for observing the surface footprints of the atmospheric circulations responsible for lake-effect snowbands, namely satellite-borne synthetic aperture radar (SAR). SAR is a radar system that images small-scale horizontal variations in surface roughness. Over water, the SAR beam is most efficiently scattered by centimeter-scale gravity– capillary waves. Because these waves are almost instantaneously generated by the local surface wind stress, spatial variability in the backscatter on SAR imagery can be linked to spatial variability in the surface wind field. Thus, in the absence of contamination from processes intrinsic to the water such as sea ice, temperature fronts, slicks, or currents, horizontal variability in SAR images of lakes and oceans is related to horizontal variability in the surface wind field. The resolution of the horizontal scales of motion resolvable with SAR is limited by both the pixel spacing (small-end limit) and the swath width (large-end limit). For the Canadian Radar Satellite-1 (RADARSAT-1) Wide Swath SAR imagery, the nominal resolution is 100 m and the maximum swath width is 500 km. Thus, atmospheric scales of motion ranging from less than 500 km to subkilometer-scale boundary layer turbulence can be resolved with the SAR instrument. (A brief summary of previous meteorological research efforts using SAR is provided in section 2.) The primary advantages of SAR for the meteorologist include its fine spatial resolution and the fact that the beam is not substantially affected by clouds and snow, although Atlas (1994) showed that rain generated by intense tropical rain cells can locally reduce the surface backscatter beneath these cells. The primary dis-

advantages of SAR include current limitations on spatial and temporal coverage, and timeliness, that is, time between image acquisition, processing, and distribution to the end user. Over the Great Lakes, an additional disadvantage is the existence of ice. These limitations are discussed in detail in section 5. Here, we describe three case studies where the RADARSAT-1 SAR imaged the surface footprints of lakeeffect snowbands of various types. The first case shows that a large-scale (hereafter referred to as a dominant band) band existed over eastern Lake Superior, marking a boundary between a zone of multiple bands to the west and smaller-scale atmospheric boundary layer structure to the east. The second case shows a single, shore-parallel band over central Lake Ontario. The final case shows a shore-parallel band over western Lake Michigan. In each case, data from traditional operational observing methods are first used to independently document the existence and character of the bands. Following this analysis, the SAR image showing the bands is presented and interpreted using what is already known of the case of interest as well as what can be inferred from the SAR image itself, such as local wind direction. From these analyses, we conclude that SAR might be of use as a diagnostic tool to alert the forecaster to the existence of atmospheric circulations that might be responsible for generating lake-effect precipitation when the traditional observational network fails. Specifically, SAR can provide crucial information about what is occurring in the boundary layer over the Great Lakes when it is overcast over those parts of the Great Lakes and where the boundary layer is not sampled by WSR-88D.

JUNE 2001

WINSTEAD ET AL.

317

FIG. 2. WSR-88D four-panel image from the Marquette, MI, radar site from 9 Jan 1999. Base reflectivity from the 0.58 elevation angle was used on all four panels. The times of each image are (a) 1104, (b) 1134, (c) 1203, and (d) 1233 UTC, respectively. The radar reflectivity signature corresponding geographically to the dominant band in the AVHRR image is labeled in each image.

2. Synthetic aperture radar SAR has been shown to be an effective tool for observing many geophysical processes in the upper ocean and lower atmosphere. Oceanographic phenomena observed from SAR include ocean internal waves (Gasparovic et al. 1989), surface currents (Johannessen et al. 1991), and sea surface slicks (Nilsson and Tildesley 1995). SAR has also been used in operational ice forecasting over the Great Lakes (Leshkevich et al. 1999) and over the Bering Sea and Labrador Sea (Vachon et al. 2000). In recent years, a new body of research has been developed focusing on the many atmospheric boundary layer and mesoscale phenomena observable with SAR, reviewed by Mourad (1999) and featured in the January–March 2000 issue of the Johns Hopkins Technical Digest (volume 21, number 1). Some examples of atmospheric phenomena relevant to this study include both cellular and roll convection (Mitnik 1992; Alpers and Bru¨mmer 1994; Sikora et al. 1995; Mourad 1996; Mourad and Walter 1996). These

authors show that marine atmospheric boundary layer convection has a distinct footprint on SAR imagery. Of relevance here, roll vortices show up as a set of parallel, linear features aligned roughly with the mean surface wind. Also, land breezes have been observed in a SAR image of Lake Michigan (Winstead and Mourad 2000). In this case, the SAR image indicated two sharp boundaries that the authors attributed to two opposing land breezes that had not yet converged in the center of the lake. Additional mesoscale meteorological phenomena observed by SAR include orographically induced features such as nocturnal drainage flow exit jets (Winstead and Young 2000), channel flow (Pan and Smith 1999), and upstream surface roughness due to small-scale topography and land use variations (Winstead and Mourad 2000). Other mesoscale meteorological features of interest to the marine forecaster that are observable from SAR include polar lows (Chunchuzov et al. 1999; Sikora et al. 2000a; Vachon et al. 2000) and hurricanes (Friedman and Li 2000; Katsaros et al. 2000).

318

WEATHER AND FORECASTING

VOLUME 16

Monaldo 2000; Pichel and Clemente´-Colon 2000). This project, called the Alaska SAR Demonstration, has demonstrated that reasonable estimates of wind speed are possible in a timely manner (within 5 h of image acquisition). In addition, the high resolution of SAR has led some researchers to extract boundary layer characteristics from SAR wind speed imagery of the convective marine atmospheric boundary layer. Examples of successful retrievals include boundary layer depth (Sikora et al. 1997), Monin–Obukhov length, and buoyancy flux (Sikora et al. 2000b; Young et al. 2000). 3. Data and methods

FIG. 3. Subjective analysis of (a) pressure and significant weather features and (b) temperature over the Lake Superior region at 1200 UTC 9 Jan 1999. The contour intervals for pressure and temperature are 2 mb and 58C, respectively. (a) Troughs and (b) warm and cold anomalies are labeled.

In addition to these examples, parallel work of interest to operational meteorology has been ongoing. If knowledge of the surface wind direction and the surface layer static stability is known a priori or can be accurately predicted using a numerical weather prediction model, for example, calibrated SAR images can be converted to wind speed using techniques similar to those used in scatterometry. Recent examples of algorithms to convert SAR backscatter values to surface wind speed include those of Wakermann et al. (1996), Vachon and Dobson (1996), and Lehner et al. (1998). In addition, recent collaborations between the Johns Hopkins University Applied Physics Laboratory and the National Oceanic and Atmospheric Administration (NOAA) have focused on making these techniques operational over the Gulf of Alaska and the Bering Sea (Thompson and Beal 2000;

In all cases presented here, surface data were obtained from the University of Washington’s Department of Atmospheric Sciences. These data included routinely available surface data from the Great Lakes region. Also, data from the Coastal Marine Automated Network stations in and around the Great Lakes were used when available to provide wind information over the Great Lakes. These data were obtained from the National Data Buoy Center. These surface data in conjunction with the remote sensing data described below provided a framework for the surface analyses presented in section 3. All mesoscale surface maps presented here were subjectively analyzed using contour intervals appropriate for the feature being analyzed. Specifics of each analysis are given with each individual case study. The traditional remote sensing observation platforms included Advanced Very High Resolution Radiometer (AVHRR) satellite imagery and WSR-88D imagery. The specifics of the satellite image times are presented with each case. General characteristics of all AVHRR images are as follows. In each case, channel 4 (centered on 10.82 mm) was used because it was nighttime. These images were calibrated and converted to brightness temperature using the Environment for Visualizing Images (ENVI) image processing software (available from Research Systems Incorporated in Boulder, CO). The images were then projected onto a polar stereographic map with a spatial resolution of 1.1 km. For all three cases, WSR-88D imagery is presented. Base reflectivity plots at an elevation angle of 0.58 are used. For the Lake Superior case, the imagery was obtained from The Pennsylvania State University’s Department of Meteorology. For the remaining two cases, the imagery was obtained from the National Center for Atmospheric Research’s Research Applications Program Web server. The three sites from which data are shown are the Marquette, Michigan (MQT); Buffalo, New York (BUF); and Milwaukee, Wisconsin (MKX); locations. The SAR images used in this study are three RADARSAT-1 ScanSAR Wide B images processed at the Canadian Space Agency (CSA) receiving station at Gatineau, Quebec, and obtained from the Satellite Active Archive. These images are from descending passes and

JUNE 2001

WINSTEAD ET AL.

FIG. 4. RADARSAT-1 ScanSAR Wide B synthetic aperture radar image at 1207 UTC 9 Jan 1999. The original image width was 500 km but the image has been cropped to focus on the Lake Superior region. The pixel spacing of the image is 100 m. The locations of the dominant bands, multiple bands and wind streaks are labeled. SAR image copyright CSA (1999).

FIG. 5. AVHRR image (channel 4) at 1010 UTC 19 Jan 2000. The image has been calibrated and converted to brightness temperature using ENVI processing software. In addition, the image has been projected onto a polar stereographic grid with a 1.1-km spatial resolution. The location of the shore-parallel cloud band is labeled. In addition, the kink (labeled A) corresponds to a kink in the corresponding WSR-88D images (Fig. 6) and SAR backscatter image (Fig. 8).

319

320

WEATHER AND FORECASTING

have a swath width of 500 km. However the images have been cropped to center on the Great Lake of interest. The pixel spacing for these ScanSAR Wide B images is 100 m. No image processing was performed by the authors; however, an incidence angle trend inherent to the surface scattering properties of the SAR beam is removed at the processing facility at Gatineau to aid in the interpretation of image features (discussed in Vachon et al. 2000). 4. Case studies For each case study, we will first analyze the mesoscale meteorological conditions using standard data: cloud imagery, WSR-88D, and surface observations. We will then introduce the SAR image associated with each study and relate the SAR image features with the previously documented mesoscale meteorological phenomena. The result will be a strong correspondence between what the SAR-based analysis shows and what is shown with traditional analysis techniques, thereby corroborating SAR’s ability to capture images of phenomena of operational meteorological interest. The implication is that on days when traditional observational methods are not available but SAR is, SAR might be a useful surrogate remote sensing tool, and when those traditional tools are available SAR may be useful both as corroborating data with unusually fine detail as well as a source of extra information. a. 9 January 1999: Lake Superior Figure 1 shows an AVHRR image of Lake Superior from 1134 UTC. The cloud image shows a mesoscale band (labeled in Fig. 1 and hereafter referred to as the dominant band) separating two distinct cloud pattern orientations. Specifically, to the west of the dominant band, there exist well-defined roll circulations consistent in character with previously observed multiple bands (Kelly 1986; Kristovich 1993; Kristovich and Steve 1995). These bands are aligned in a northwest to southeast direction, parallel to the prevailing wind flow near the lake (see Fig. 3). The predominant roll cloud spacing ranges from 4–5 km near the northern shore of western Lake Superior to 6–12 km near the southern shore of western Lake Superior. East of the dominant band there are fewer linear circulations evident in the cloud pattern. However, there do exist multiple bands (labeled in Fig. 1) that are oriented approximately 908 clockwise from those over western Lake Superior, again consistent with the wind flow over the eastern shore of Lake Superior (see Fig. 3). These cloud signatures have a shorter spacing, ranging between 3 and 4 km. The change in cloud orientation suggests that there is a significant change in the boundary layer wind profile across the dominant band. The dominant band also marks the likely eastern extent of significant lake-effect snowfall. Figure 2 shows

VOLUME 16

a four-panel plot of WSR-88D imagery from Marquette starting at 1104 UTC and ending at 1234 UTC. The radar imagery shows the classic signature of lake-effect snowbands oriented northwest to southeast in the vicinity of southern Lake Superior. The easternmost roll circulation measured by the radar is located where the dominant cloud band is located in Fig. 1. While we cannot be certain that no snow is falling east of this band, because the radar beam is likely above the boundary layer there, the precipitation patterns in Fig. 2 are consistent with the AVHRR cloud observations. The orientation of the cloud patterns, the existence of a dominant band over eastern Lake Superior, and the existence of lake-effect snowbands west of the dominant cloud band suggest that there is mesoscale convergence, and likely a surface trough over the eastern half of Lake Superior. Figure 3 shows mesoscale analyses of surface pressure with significant weather features (panel a) and temperature (panel b) from 1200 UTC 9 January 1999. Both pressure and temperature were subjectively analyzed using 2-mb and 58C intervals, respectively. The temperature analysis (Fig. 3b) indicates a pronounced warm anomaly over Lake Superior. The gradient is particularly strong over the western half of Lake Superior and over the northern shore of eastern Lake Superior. The gradient is less strong over the southeastern half of Lake Superior and the Upper Peninsula of Michigan due to interactions with the boundary layer modifications associated with Lakes Michigan and Huron. The pressure and significant feature analysis (Fig. 3a) indicates that there is convergence and generally lower pressures over Lake Superior. The surface wind and pressure observations mandate the existence of a trough over western Lake Superior. In addition, the pressure and wind observations over the eastern half of the lake support the satellite observations and indicate the existence of a second region of mesoscale convergence there. This is supported by the cloud patterns in Fig. 1 and the radar observations in Fig. 2. Specifically, the orientation of the cloud patterns over western Lake Superior coupled with the existence of the larger-scale cloud band separating the two distinct cloud regimes suggest that the axis of convergence lies along the axis of the dominant cloud band. The precipitation patterns evident in Fig. 2 are qualitatively similar to the cloud patterns in that there is widespread precipitation to the west of the linear reflectivity pattern associated with the dominant cloud band while there are no discernible echoes to the east of this reflectivity band, with the caveat noted earlier. We now turn to the RADARSAT-1 SAR image of Lake Superior taken at 1207 UTC (Fig. 4). In many ways, the SAR image is similar in appearance to the AVHRR image (Fig. 1), except that the values at each pixel are related to surface roughness rather than radiance. Specifically, there exist multiple bands over western Lake Superior oriented in a northwest to southeast manner. Previous studies have shown that these bands are the surface footprints of convective roll vortices (Alpers and

JUNE 2001

WINSTEAD ET AL.

321

FIG. 6. WSR-88D four-panel image from the Buffalo, NY, radar site from 19 Jan 2000. Base reflectivity from the 0.58 elevation angle was used in all four panels. The times of each image are (a) 1107, (b) 1207, (c) 1306, and (d) 1406 UTC, respectively. The reflectivity signature corresponding to the shore-parallel band is labeled. In addition, the kink (labeled A) corresponds to a kink in the corresponding AVHRR image (Fig. 5) and SAR backscatter image (Fig. 8).

Bru¨mmer 1994; Mourad and Walter 1996; Mourad 1996; Mu¨ller et al. 1999). The spacing of these rolls is on the order of 6–9 km near the southern shore of western Lake Superior consistent with the AVHRR observations mentioned above and within the range of previously reported band spacing (see, e.g., Kelly 1986). Over eastern Lake Superior, some wind streaks can be seen oriented in a northeast to southwest direction with a spacing of approximately 1–3 km. Between these two qualitatively different backscatter regimes is a sharply defined boundary that coincides in location to the dominant cloud band in Fig. 1. Because the horizontal variability in backscatter evident in Fig. 1 is in general related to horizontal wind variability, the sharp boundary that is coincident with the cloud bands reveals important information about the spatial variability of the surface wind field over Lake Superior. Moreover, the

trough over eastern Lake Superior can be very precisely located using SAR. b. 19 January 2000: Lake Ontario Figure 5 shows an AVHRR image of the eastern Great Lakes from 1010 UTC 19 January 2000. The image shows a single shore-parallel band over Lake Ontario, oriented along the central axis of the lake. The existence of this band is corroborated by the Buffalo, WSR-88D. Figure 6 shows a four-panel plot of base reflectivity at an elevation angle of 0.58 at 1107 (panel a), 1207 (panel b), 1306 (panel c), and 1406 UTC (panel d). The image shows a linear precipitation band over the center of the lake. The radar signature of the band is consistent with the linear cloud band over Lake Ontario in Fig. 5, both in placement and shape. This band is relatively weak

322

WEATHER AND FORECASTING

VOLUME 16

that is missing is a clearly defined trough over Lake Ontario; however, surface wind observations around Lake Ontario show that offshore flow exists around the entire lake. This suggests that a trough exists over the lake but is unresolved by the synoptic data network. It is therefore reasonable to hypothesize that this is a case where the two opposing land breezes from both shores of the lake meet in the middle creating an axis of surface convergence, upward motion, and ultimately precipitation. The SAR image from the morning of 19 January is consistent with this conclusion. Figure 8 shows a RADARSAT-1 SAR image of Lake Ontario from 1136 UTC. The image indicates a thin, linear feature extending down the central axis of Lake Ontario. This feature is coincident in space and shape with the linear feature visible in both AVHRR and WSR-88D imagery. For example, the band is located approximately 30 km from the southern shore of the lake and has a noticeable kink (labeled point A in Figs. 5, 6b, and 8). From west to east across the SAR image, the image brightness increases, suggesting an increase in surface wind speed over eastern Lake Ontario. Near the southern shore of the lake, the signature of ice exists (labeled). The linear band in central Lake Ontario in both WSR-88D and AVHRR cloud imagery is consistent with previous reported intersecting land-breeze circulations over the Great Lakes (see, e.g., Niziol et al. 1995). We therefore hypothesize that the linear feature evident in the SAR image is located where the two opposing land breezes converge and is therefore the axis of surface convergence. c. 4 February 2000: Lake Michigan

FIG. 7. Subjective analysis of (a) pressure and (b) temperature from the eastern Great Lakes at 1200 UTC 19 Jan 2000. The contour interval for temperature was 58C and the contour interval for pressure was 2 mb. Warm and cold anomalies are labeled in (a).

(;15 dBZ) but persists until 1405 UTC (Fig. 5d) when it becomes indistinguishable from a larger mesoscale area of snow. Figure 7 shows a subjective pressure (2 mb) (panel a) and temperature analysis (58C contour interval) (panel b) of the eastern Great Lakes from 1200 UTC 19 January 2000. The temperature analysis indicates warm anomalies over Lakes Erie, Huron, and Ontario; however, the land–lake temperature gradient is strong around Lake Ontario. On the morning of 19 January, the synoptic-scale pressure gradient and wind flow over the region was weak (Fig. 7a). These conditions are favorable for the development of land breezes over the Lake Ontario region as reported during the Lake Ontario Winter Storms project (Reinking et al. 1993). One element

Figure 9 shows an AVHRR image from 1027 UTC 4 February 2000. The image shows thin, high clouds over southern Lake Michigan with low clouds covering eastern Lake Michigan. There is no clear indication of a cloud band over western Lake Michigan, although the low clouds over the lake suggest that the lake-effect processes are occurring at this time. In order to document the band in this case, we rely on WSR-88D imagery. Figure 10 shows a four-panel plot of level 1 base reflectivity from the Milwaukee radar spanning times from 1104 to 1224 UTC 4 February 2000. This fourpanel plot documents the existence of a precipitation band oriented parallel to the western shore of Lake Michigan. This band is located west of the center axis of Lake Michigan and consists of a quasi-linear band that persists throughout the time period covered. The maximum intensity in this band ranges from 16 to 20 dBZ. Figure 11 shows a subjective analysis of surface pressure with a 2-mb contour interval (panel a) and temperature analysis with a 58C contour interval (panel b) from 1200 UTC 4 February 2000. The temperature analysis indicates weak thermal gradients over Lake Mich-

JUNE 2001

WINSTEAD ET AL.

323

FIG. 8. RADARSAT-1 ScanSAR Wide B image at 1136 UTC 19 Jan 2000. The original image width was 500 km but the image was cropped to focus on the Lake Ontario region. The pixel spacing of the image is 100 m. The shore-parallel band is labeled. The kink in the band (labeled A) represents a feature that is also detectable in the AVHRR image (Fig. 5) and WSR-88D images (Fig. 6). In addition, ice near the southern shore is labeled. Finally, the horizontal bands that appear in the image (such as in region Q) are an image artifact. SAR image copyright CSA (2000).

igan. The surface pressure analysis indicates a moderately strong synoptic-scale pressure gradient over the Lake Michigan region. The surface wind in the vicinity of the lake is from the north-northeast, nearly parallel to the along-lake axis and is blowing at between 10 and 20 kt. Previous researchers have shown that when the flow over a particular Great Lake is shore parallel (see, e.g., Ballantine 1982; Reinking, et al. 1993; Nicosia et al. 1999), conditions are favorable for the development of shore-parallel bands of rain or snow. Two separate physical mechanisms have been identified to explain these bands: opposing land breezes and frictional convergence. Ballantine (1982) presented case studies over Lake Michigan containing cloud and snow bands oriented along the axis of Lake Michigan. He found that these bands were often the result of land-breeze circulations, much like the Lake Ontario band reported in section 3b. However, there is typically a strong temperature gradient evident when land breezes exist. Figure 11b indicates that the thermal gradient over the southern two-thirds of Lake Michigan is weak. A second

possible explanation for the existence of shore-parallel bands was reported by Nicosia et al. (1999). They found that frictional convergence played a role in the development of a lake-enhanced rainband over Erie, Pennsylvania. Similar to here, that case was marked by shoreparallel flow. Both explanations are consistent with the WSR-88D observations from MKX. We now document the existence of the surface signature associated with this band in RADARSAT-1 SAR imagery. Figure 12 shows a RADARSAT-1 SAR image of Lake Michigan taken at 1203 UTC 4 February 2000. The SAR image shows a shore-parallel band whose offshore extent ranges from 20 km over southern Lake Michigan (east of Chicago) to 38 km offshore of Sheboygan, Wisconsin. The band is approximately parallel to the axis of the lake but west of the central axis. This shoreparallel band coincides geographically with the precipitation band shown in Fig. 10. The qualitative shape of the band shown in Fig. 12 and its location relative to shore is similar to that of SAR observations of a land breeze reported by Winstead and Mourad (2000). How-

324

WEATHER AND FORECASTING

FIG. 9. AVHRR image (channel 4) at 1027 UTC 4 Feb 2000. The image has been calibrated and converted to brightness temperature using ENVI processing software. In addition, the image has been projected onto a polar stereographic grid with a 1.1-km spatial resolution.

ever, the synoptic pattern at 1200 UTC is not conducive to land-breeze formation for two reasons. First, the temperature gradient surrounding Lake Michigan is weak (Fig. 11b) with no evidence of a strong lake–land temperature gradient like the one seen over Lake Ontario (Fig. 7b). Second, the synoptic wind flow around the lake is shore parallel and quite strong (Fig. 11a). While land breezes are possible in moderate synoptic flow such as this, the lack of a strong land–lake temperature gradient casts doubt on this process here. What seems more likely is that frictional convergence is playing a role. Other researchers (e.g., Nicosia et al. 1999) have observed that when there is shore-parallel flow over a lake, a wind-parallel band can develop over the lake or just onshore. Given the shore-parallel flow on this day, it seems more likely that this process is occurring. 5. Discussion of limitations While the purpose of this paper is to demonstrate the utility of SAR as an additional remote sensing tool for the marine forecaster, there are some important limitations that must be overcome before SAR can become operational. These include current limitations on spatial and temporal coverage, and contamination from nonatmospherically generated phenomena. Because satellite-borne SAR platforms such as RADARSAT must be

VOLUME 16

mounted on a polar orbiting satellite, one important concern is temporal coverage, particularly at lower latitudes like the Great Lakes (or the Gulf of Mexico or the U.S. southeast coast for hurricane monitoring). RADARSAT ScanSAR Wide B temporal and spatial coverage varies for each lake (based on the size of the lake and its latitude). The most often imaged Great Lakes are Superior and Huron because of their size and location. Portions of these lakes are sometimes imaged twice daily (in the morning on a descending pass and in the evening on an ascending pass). However, the most common repeat cycle for Lake Superior is daily. In order to quantify this, RADARSAT orbit planning software (available from the Alaska SAR Facility in Fairbanks, AK) was used to generate the number of SAR passes that passed over some portion of Lake Superior over the month of January 2001. Over this month, a total of 30 frames that included at least half of Lake Superior occurred, or nearly daily. For the other lakes, coverage is not so good. Lakes Michigan, Erie, and Ontario are imaged approximately once every 2–3 days. Obviously, this coverage frequency is not sufficient for operational meteorology. However, in this decade, a total of four additional SAR satellites are planned (Attema et al. 2000; Winkour 2000). These additional launches, if properly coordinated and if data access to the operational community is provided, should help alleviate the temporal and spatial coverage issues mentioned here (Beal 2000; Holt and Hilland 2000; Winkour 2000). At the very least, these launches should provide an excellent supplemental dataset for use in the operational community. The second logistical limitation to SAR is the time between image acquisition and availability to a forecaster or researcher. In section 2, a demonstration project between NOAA and the Johns Hopkins University Applied Physics Laboratory (APL) designed to test the feasibility of providing near–real time wind speed estimates from SAR for the Gulf of Alaska has proved to be reasonably successful. The average processing time during this demonstration has been approximately 5 h (Monaldo 2000) but the authors point out that much of this 5-h processing time is both the electronic transfer of the raw image files from the Alaska SAR Facility to the Johns Hopkins University APL and retrieving wind direction estimates from a numerical forecast model. If the imagery could be received, processed, and disseminated from the same location, a significant reduction in the time required to produce SAR wind speed imagery could be achieved. Furthermore, continued advances in computer technology should help to alleviate this issue. In addition to the logistical issues above, there are certain issues related to image interpretation that are important to consider. Over the Great Lakes, the primary scientific limitation to using SAR to remotely sense snowbands is surface ice. The presence of lake ice renders SAR useless for atmospheric analysis. Because

JUNE 2001

WINSTEAD ET AL.

325

FIG. 10. WSR-88D four-panel plot from the MKX radar site from 4 Feb 2000. Base reflectivity from the 0.58 elevation angle was used in all panels. The times of each reflectivity image are (a) 1104, (b) 1124, (c) 1204, and (d) 1224 UTC, respectively. The location of the shore-parallel band is labeled.

nearly all of the Great Lakes have some ice cover during the course of the winter, especially Lake Erie, which tends to freeze over completely, the utility of SAR will be maximized in the early winter months before ice cover becomes significant. Fortunately, the processes that lead to lake-effect snow tend to be most efficient without ice. Therefore, this is not a serious limitation, even during that part of the winter when both ice and open water exist, because ice generally has a well-characterized signature in SAR imagery. The other primary scientific limitation is the existence of other nonatmospheric phenomena in SAR imagery. Geophysical phenomena intrinsic to the water such as slicks, internal waves, surface currents, and other processes could interfere with the meteorological interpretation of SAR imagery. Because some of those features can occur on the same scale as atmospheric phenomena, it is possible to misinterpret SAR signatures. Fortunately, atmospheric signatures tend to be repeatable and, as suggested by

Sikora et al. (2000a), pattern recognition techniques could be developed to help alleviate this problem. 6. Conclusions In this paper, we showed three separate case studies where traditional remote sensing assets documented the existence of three different bands over different Great Lakes. The first case study documented a dominant band over Lake Superior. It was hypothesized that this dominant band marked the location of the axis of surface convergence (and hence the location of the surface trough) associated with the dominant band. The second case study documented a shore-parallel band over central Lake Ontario. It was hypothesized that this band was the result of two opposing land breezes converging in the center of the lake. Finally, the third case study documented the existence of a shore-parallel band over west-central Lake Michigan. It was hypothesized that

326

WEATHER AND FORECASTING

VOLUME 16

FIG. 12. RADARSAT-1 ScanSAR Wide B image at 1203 UTC 4 Feb 2000. The original image width was 500 km but the image was cropped to focus on Lake Michigan. The cities of Milwaukee and Chicago are labeled. Finally, the location of the shore-parallel band is indicated. FIG. 11. Subjective analysis of (a) pressure and (b) temperature over the Lake Michigan region at 1200 UTC 4 Feb 2000. The contour intervals are 2 mb for pressure and 58C for temperature.

this band was the result of either land breezes converging over western Lake Michigan or frictional convergence that resulted from the nearly shore-parallel alignment of the synoptic-scale flow over the lake. In all three cases we have shown that SAR is capable of detecting the same atmospheric phenomena observable from traditional satellite and WSR-88D imagery, and in greater detail. This correspondence suggests that SAR can prove to be useful to the operational forecaster when the imaged lake is cloud covered and over those regions of the Great Lakes where WSR-88D overshoots the boundary layer. Finally, as pointed out by one of the reviewers, with precise positioning of a given SAR image of the surface manifestation of a band, coupled with precise position-

ing of a temporally matched WSR-88D or AVHRR image of that band aloft, one could in principle infer important information about the tilt of the band. Moreover, if one pushed this argument to its limit, one could also infer some information about the boundary layer shear environment. Acknowledgments. The authors are grateful to Paul Zibton for his help preparing journal-quality figures. In addition, the authors would like to thank Todd Sikora, Dave Kristovich, and Don Thompson for many valuable discussions regarding the scope and direction of this paper. Also, the reviewers’ comments greatly improved this manuscript. Finally, the authors would like to thank the National Science Foundation (Grant ATM9707730) for supporting this work.

JUNE 2001

WINSTEAD ET AL. REFERENCES

Alpers, W. R., and B. Bru¨mmer, 1994: Atmospheric boundary layer rolls observed by the synthetic aperture radar aboard the ERS1 satellite. J. Geophys. Res., 99, 12 613–12 621. Atlas, D., 1994: Footprints of storms on the sea: A view from spaceborne synthetic aperture radar. J. Geophys. Res., 99, 7961–7969. Attema, E., Y.-L. Desnos, and G. Duchossois, 2000: Synthetic aperture radar in Europe: ERS, Envisat, and beyond. Johns Hopkins APL Tech. Dig., 21, 155–161. Ballentine, R. J., 1982: Numerical simulation of land-breeze-induced snowbands along the western shore of Lake Michigan. Mon. Wea. Rev., 110, 1544–1553. Beal, R. C., 2000: Toward an international Storm Watch using wide swath SAR. Johns Hopkins APL Tech. Dig., 21, 12–20. Chunchuzov, I., P. W. Vachon, and B. Ramsay, 1999: Detection and characterization of mesoscale cyclones in RADARSAT synthetic aperture radar images of the Labrador Sea. Can. J. Remote Sens., 26, 213–230. Forbes, G. S., and J. H. Merritt, 1984: Mesoscale vortices over the Great Lakes in wintertime. Mon. Wea. Rev., 112, 377–381. Friedman, K. S., and X. Li, 2000: Monitoring hurricanes over the ocean with Wide Swath SAR. Johns Hopkins APL Tech. Dig., 21, 80–85. Gasparovic, R. F., D. R. Thompson, and J. R. Apel, 1989: Synthetic aperture radar imaging of ship-generated internal waves. Johns Hopkins APL Tech. Dig., 10, 326–331. Holt, B., and J. Hilland, 2000: Rapid-repeat SAR imaging of the ocean surface: Are daily observations possible? Johns Hopkins APL Tech. Dig., 21, 162–169. Johannessen, J. A., R. A. Shuchman, O. M. Johannessen, K. L. Davidson, and D. R. Lyzenga, 1991: Synthetic aperture radar imaging of upper ocean circulation features and wind fronts. J. Geophys. Res., 96, 10 411–10 422. Katsaros, K. B., P. W. Vachon, P. G. Black, P. P. Dodge, and E. W. Uhlhorn, 2000: Wind fields from SAR: Could they improve our understanding of storm dynamics? Johns Hopkins APL Tech. Dig., 21, 86–93. Kelly, R. D., 1986: Mesoscale frequencies and seasonal snowfalls for different types of Lake Michigan snow storms. J. Climate Appl. Meteor., 25, 308–312. Kristovich, D. A. R., 1993: Mean circulations of boundary layer rolls in lake effect snow storms. Bound.-Layer Meteor., 63, 293–315. Laird, N. F., 1999: Observations of coexisting mesoscale lake-effect vortices over the western Great Lakes. Mon. Wea. Rev., 127, 1137–1141. Lavoie, R. L., 1972: A mesoscale numerical model of lake-effect storms. J. Atmos. Sci., 29, 1025–1040. Lehner, S., J. Horstmann, W. Koch, and W. Rosenthal, 1998: Mesoscale wind measurements using recalibrated ERS SAR images. J. Geophys. Res., 103, 7847–7856. Leshkevich, G. A., S. V. Nghiem, and R. Kwok, 1999: Great Lakes ice cover classification and mapping using satellite synthetic aperture radar (SAR) data. Proc. Fifth Int. Conf. on Remote Sensing for Marine and Coastal Environments, San Diego, CA, ERIM, II-402–II-405. Minor, T., P. J. Sousounis, J. Wallman, and G. Mann, 2000: Hurricane Huron. Bull. Amer. Meteor. Soc., 81, 223–236. Mitnik, L. M., 1992: Mesoscale coherent structures in the surface wind field during cold air outbreaks over the Far Eastern seas from the satellite side looking radar. La Mer, 30, 287–296. Monaldo, F., 2000: The Alaska SAR Demonstration and near-realtime synthetic aperture radar winds. Johns Hopkins APL Tech. Dig., 21, 75–79. Mourad, P. D., 1996: Inferring multiscale structure in atmospheric turbulence using satellite-based SAR imagery. J. Geophys. Res., 101, 18 433–18 449. ——, 1999: Footprints of atmospheric phenomena in synthetic aperature radar images of the ocean surface–A review. Air–Sea Ex-

327

change: Physics, Chemistry, and Dynamics, G. Geernaert, Ed., Kluwer Academic, 269–290. ——, and B. Walter, 1996: Viewing a cold-air outbreak using satellitebased synthetic aperture radar and advanced very high resolution radiometer imagery. J. Geophy. Res., 101, 16 391–16 400. Muller, G., B. Brummer, and W. Alpers, 1999: Roll convection within an Arctic cold-air outbreak: Interpretation of in situ aircraft measurements and spaceborne SAR imagery by a three-dimensional atmospheric model. Mon. Wea. Rev., 127, 363–380. Nicosia, D. J., and Coauthors, 1999: A flash flood from a lake-enhanced rainband. Wea. Forecasting, 14, 271–288. Nilsson, C. S., and P. C. Tildesley, 1995: Imaging of oceanic features by ERS-1 synthetic aperture radar. J. Geophys. Res., 100, 953– 967. Niziol, T. A., 1987: Operational forecasting of lake effect snowfall in western and central New York. Wea. Forecasting, 2, 310– 321. ——, W. R. Snyder, and J. S. Waldstreicher, 1995: Winter weather forecasting throughout the eastern United States. Part IV: Lake effect snow. Wea. Forecasting, 10, 61–77. NRC, 1995: Assessment of WSR-88D coverage and associated weather services. WSR-88D Panel Report, National Weather Service Modernization Committee, National Research Council, 112 pp. Pan, F., and R. B. Smith, 1999: Gap winds and wakes: SAR observations and numerical simulations. J. Atmos. Sci., 56, 905–923. Passarelli, R. E., and R. R. Braham, 1981: The role of the winter land breeze in the formation of Great Lake snow storms. Bull. Amer. Meteor. Soc., 62, 482–492. Pease, S. R., W. Lyons, C. Keen, and M. Hjelmfelt, 1988: Mesoscale spiral vortex embedded within a Lake Michigan snow squall band: High resolution satellite observations and numerical model simulations. Mon. Wea. Rev., 116, 1374–1380. Pettersen, S., and P. A. Calabrese, 1959: On some weather influences due to warming of the air by the Great Lakes in winter. J. Meteor., 16, 646–652. Pichel, W. G., and P. Clemente-Colo´n, 2000: NOAA CoastWatch SAR applications and demonstration. Johns Hopkins APL Tech. Dig., 21, 49–57. Reinking, R. F., and Coauthors,1993: The Lake Ontario Winter Storms (LOWS) project. Bull. Amer. Meteor. Soc., 74, 1828–1849. Sikora, T. D., G. S. Young, R. C. Beal, and J. B. Edson, 1995: On the use of spaceborne synthetic aperture radar imagery of the sea surface in detecting the presence and structure of the convective marine atmospheric boundary layer. Mon. Wea. Rev., 123, 3623–3632. ——, ——, H. N. Shirer, and R. D. Chapman, 1997: Estimating convective atmospheric boundary layer depth from microwave radar imagery of the sea surface. J. Appl. Meteor., 36, 833– 845. ——, K. S. Friedman, W. G. Pichel, and P. Clemente´-Colo´n, 2000a: Synthetic aperture radar as a tool for investigating polar mesoscale cyclones. Wea. Forecasting, 15, 745–758. ——, D. R. Thompson, and J. C. Beidorn, 2000b: Testing the diagnosis of marine atmospheric boundary-layer structure from synthetic aperture radar. Johns Hopkins APL Tech. Dig., 21, 94– 99. Thompson, D. R., and R. C. Beal, 2000: Mapping high-resolution wind fields using synthetic aperture radar. Johns Hopkins APL Tech. Dig., 21, 58–67. Vachon, P. W., and F. W. Dobson, 1996: Validation of wind vector retrieval from ERS-1 SAR images over the ocean. Global Atmos. Ocean Syst., 5, 177–187. ——, and Coauthors, 2000: Canadian progress toward marine and coastal applications of synthetic aperture radar. Johns Hopkins APL Tech. Dig., 21, 33–40. Wackermann, C. C., C. L. Rufenach, R. Schuman, J. A. Johannessen, and K. Davidson, 1996: Wind vector retrieval using ERS-1 synthetic aperture radar. IEEE Trans. Geosci. Remote Sens., 34, 1343–1352.

328

WEATHER AND FORECASTING

Winkour, R. S., 2000: SAR symposium keynote address. Johns Hopkins APL Tech. Dig., 21, 5–11. Winstead, N. S., and P. D. Mourad, 2000: Shallow Great Lake–scale atmospheric thermal circulation imaged by synthetic aperture radar. Mon. Wea. Rev., 128, 3654–3663.

VOLUME 16

——, and G. S. Young, 2000: An analysis of exit-flow drainage jets over the Chesapeake Bay. J. Appl. Meteor., 39, 1269–1281. Young, G. S., T. D. Sikora, and N. S. Winstead, 2000: On inferring marine boundary layer properties from spectral characteristics of satellite-borne SAR imagery. Mon. Wea. Rev., 128, 1506–1520.