Synthetic aperture radar (SAR) has proven to be a useful tool for observing a wide variety of oceanographic and atmospheric phenomena. This is because capillary waves whose amplitudes are modulated in space and time by oceanic and atmospheric processes are efficient scatterers at SAR wavelengths. In this paper, a SAR image of Lake Michigan taken during the Lake-Induced Convection Experiment is analyzed. The image shows three broad parallel bands identifiable as the components of a shallow, Great Lake–induced thermal circulation:two bands associated with opposing land-breeze circulations, and a middle band containing the signature of boundary layer convection. A cross-frontal cut shows that the width of the two land-breeze fronts varies in a manner consistent with previously reported observations of land and sea breezes superimposed on synoptic flows. The SAR image analysis in conjunction with a mesoscale analysis of a Great Lake–scale convection pattern substantially increases the available knowledge of that pattern. Specifically, the SAR image provides information concerning the precise placement of the surface land-breeze fronts not available from other means. Finally, the SAR analysis shows that the western land-breeze brightness patterns are affected by the shallow terrain along the western shore of Lake Michigan. The latter point therefore suggests that SAR can provide valuable information about the link between variations in surface roughness and/or land use patterns and the horizontal structure of the surface wind stress over coastal regions.
Synthetic aperture radar (SAR) has been shown to be an effective tool for studying a vast array of oceanographic and atmospheric phenomena. This is because the spatial variation of the radar-backscatter field depicted on SAR imagery is directly related to the spatial variation of ocean surface roughness elements that have scales comparable to the wavelength of radar signal transmitted by the SAR. For the SAR aboard the RADARSAT satellite, this wavelength is 5.6 cm. Thus the spatial patterns of centimeter-scale capillary waves on the surface of the water are primarily responsible for the spatial patterns of the radar backscatter depicted in SAR images.
Centimeter-scale capillary waves can be generated and modulated by both atmospheric and oceanographic processes. The signatures of oceanographic features imaged by SAR have been studied extensively (e.g., Alpers et al. 1981; Beal et al. 1981; Vesecky and Stewart 1982;Gasporovic et al. 1988; Thompson and Jensen 1993; Nilsson and Tildesley 1995). In recent years, research has shifted to the study of the many atmospheric processes observable by SAR (reviewed in Mourad 1999). Examples of atmospheric phenomena that appear in SAR imagery include atmospheric gravity waves (Vachon et al. 1994), synoptic-scale lows and synoptic-scale fronts (Kalmykov et al. 1985; Velichko et al. 1989; Johannessen et al. 1991; Vachon et al. 1998), heavy rain cells (Atlas 1994; Atlas and Black 1994; Atlas et al. 1995; Iguchi et al. 1995), boundary layer rolls (Gerling 1986; Müller et al. 1999; Alpers and Brümmer 1994; Mourad 1996; Mourad and Walter 1996; Korsbakken et al. 1998; Lehner et al. 1998), convection (Mitnik 1992;Sikora et al. 1995; Mourad and Walter 1996; Sikora et al. 1997), and katabatic winds (Winstead and Young 2000; Alpers et al. 1998). All of these studies have taken advantage of the relatively high resolution offered by synthetic aperture radar to view the footprints of atmospheric phenomena not easily observed by other means.
In this paper, data from the Lake-Induced Convection Experiment (Lake-ICE) are used to aid the interpretation of a RADARSAT SAR image from 1208 UTC 14 January 1998. The results show that this SAR image contains the signature of a land breeze and associated land-breeze front from both the eastern and the western shores of Lake Michigan; these fronts almost meet over the lake and are separated by the signature of atmospheric boundary layer convection. This structure is created by a lake-induced thermally driven circulation embedded within weak synoptic easterly flow. These results are not just of academic interest; we argue that a forecast of this mesoscale structure could be improved with the addition of a SAR analysis like the one presented here. Also of interest is the role that coastal surface roughness variations play in the modulation of the near-surface wind field as it is manifested in the SAR image. Specifically, there is evidence that the surface footprint of the land breeze is modulated in some areas by variations in the surface roughness properties caused by terrain fluctuations and variations in land use along the Lake Michigan coastline.
2. Data sources
The meteorological data used in this study were obtained during Lake-ICE, performed during December 1997 and January 1998. A complete description of Lake-ICE is given by Kristovich et al. (2000). Relevant to this study, the Lake-ICE dataset offers the advantage of well-organized supplemental observations in the vicinity of Lake Michigan at the time of our SAR image. These observations provide the high spatial resolution necessary for detailed mesoscale analyses of the Great Lakes region.
Hourly surface observations along the shores of all the Great Lakes were obtained from the Great Lakes Environmental Research Laboratory (GLERL). These data were originally archived during the ongoing National Oceanic and Atmospheric Administration Coast Watch/Great Lakes Node program (Leshkovich et al. 1993). Other surface data were obtained through the CODIAC data management system at the Joint Office for Scientific Support. They included observations reported by ships and buoys, reports from Coastal Marine Automated Network stations, observations at U.S. Coast Guard stations, the Surface Airways Observations dataset, and other marine reports covering the entire nearshore and offshore regions of the Great Lakes. Supplemental data included observations from Automated Surface Observing System sites in the Great Lakes region. These additional data were necessary because GLERL did not archive data taken far from the shores of the Great Lakes.
In addition to the routinely available surface data described above, the Lake-ICE field project was ongoing. This field program consisted of a variety of facilities whose purpose was to collect in situ data over Lake Michigan and surrounding areas. Examples of these facilities include the National Center for Atmospheric Research Electra and University of Wyoming King Air aircraft, Lidar from the University of Wisconsin and supplemental soundings from two sites: Sheboygan, Wisconsin, and Montague, Michigan. Unfortunately, the only additional data available near the time of the SAR image were the two National Center for Atmospheric Research Integrated Sounding System (ISS) supplemental soundings and operational imagery from the weather service radar (Weather Surveillance Radar-1988 Doppler; WSR-88D) network. In addition to standard rawinsonde data, the supplemental sounding sites also provided wind profiler data from the Remote Acoustic Sounding System (RASS). The first of these soundings, at 1200 UTC 14 January, 1998, was launched from Sheboygan (ISS2) while the second, at 1205 UTC on the same day, was launched from Montague (ISS1). The sounding sites are labeled in Fig. 1a. Of interest to this study, RASS data were available from both sites for several hours surrounding 1200 UTC. The WSR-88D base reflectivity data presented here were obtained from the Milwaukee, Wisconsin (labeled MKX in Fig. 1a), radar site. All base reflectivity plots were from the 0.5° elevation angle scan. Velocity–azimuth display (VAD) wind profiles were obtained from the Grand Rapids, Michigan (labeled GRR in Fig. 1a), radar site.
a. Synoptic overview
At 1200 UTC 14 January 1998, the Great Lakes were under the influence of a retreating area of arctic high pressure centered over southern Canada just to the east of Lake Superior. Temperatures within this air mass were significantly lower than surface temperatures of the Great Lakes, indicating the general conditions necessary for cold-season thermal circulations over the lakes (Petterson and Calabrese 1959; Lenschow 1973; Weiss and Sousounis 1999). The predominant synoptic-scale airflow at 1200 UTC was light and easterly over the area surrounding Lake Michigan, the focal lake of this study. However, as shown later, a westerly flow existed along the immediate western shore of Lake Michigan and an enhanced easterly flow existed over the southern half of the eastern shore of Lake Michigan. This mesoscale airflow pattern is consistent with opposing land-breeze circulations as reported by Passarelli and Braham (1981) and Hjelmfelt and Braham (1983). These circulations and their manifestation in a SAR image are the subject of this study.
b. Mesoscale features
The focus here is on the immediate vicinity of Lake Michigan in order to document the lake-induced thermal circulation that was present at 1200 UTC. The daily weather summary from the Lake-ICE operations staff (obtained from the Joint Office for Scientific Support) at 2223 UTC 13 January 1998 had predicted the development of a land breeze. The subsequent summary at 2317 14 January 1998 indicated that a shallow land breeze had occurred on both sides of the lake but was not deep enough to generate a shore-parallel snowband. Previous observations indicate that when the boundary layer is shallow and the static stability within the air mass is high, shore-parallel snowbands do not always develop (Hjemfelt and Braham 1983; Passarelli and Braham 1981).
Figure 1 shows (a) the location of supplemental Lake-ICE data and WSR-88D sites, (b) sea level pressure and significant weather features, and (c) temperature analyses of the Great Lakes region at 1200 UTC. The pressure field was subjectively analyzed from the archived surface data described in section 2 using a pressure contour of 1 hPa. The analysis followed the conventions of Young and Fritsch (1989) and Fritsch and Vislocky (1996). For all locations along the shores of Lake Michigan (e.g., Milwaukee) we used the observations at the coast guard stations rather than inland airports, because the coast guard stations are located directly along the lake shore. The pressure of 1027.0 hPa reported by a ship of opportunity at 45.0°N, 86.5°W was not included in the analysis. This value was deemed to be unreasonably low given the surrounding pressures and the relatively light winds reported along the northern shores of Lake Michigan. A pressure analysis including this ship observation (not shown) was qualitatively similar to that shown in Fig. 1b; the only significant difference was that it indicated a closed mesoscale low over Lake Michigan, rather than the trough discussed below.
The mesoscale analysis of Fig. 1b shows a clearly defined trough over Lake Michigan. This type of thermally induced trough is well documented in the literature (see, e.g., Hjemfelt and Braham 1983; Weiss and Sousounis 1999; Nicosia et al. 1999). This particular trough persisted for several hours; similar analyses at 1100 and 1300 UTC (not shown) indicated that it was present at those times as well. The placement and symbology of the land-breeze fronts in Fig. 1b—although not their existence—was based on the analysis of the SAR image (placement) and WSR-88D radar reflectivities (symbology) from the Milwaukee radar site. This is discussed below.
The temperature analysis shown in Fig. 1c also indicates the mesoscale conditions necessary to support a thermally induced lake circulation. The temperature field was subjectively analyzed using a 2°C contour interval. Warm and cold anomalies are labeled. The temperature analysis shows that all of the Great Lakes were associated with warm anomalies and that there were cold anomalies over the adjacent land surfaces. This is typical when the lakes are warm relative to the air (Passarelli and Braham 1981; Hjelmfelt and Braham 1983; Nicosia et al. 1999; Weiss and Sousounis 1999). The warm anomaly over Lake Michigan was well pronounced at 1200 UTC. Admittedly, the strength of this anomaly is based strongly on the single ship observation on Lake Michigan; however, the warm anomaly is consistent with the warm anomalies over the other Great Lakes, and with the results of Simpson (1987) and Stull (1988) and is supported by the relatively warm conditions reported at observing sites along the immediate Lake Michigan shoreline.
In order to study the vertical extent of the land-breeze circulation on both shores of Lake Michigan, RASS wind profiler data from Sheboygan and Montague (Fig. 2) were examined. In addition, VAD wind profiles from Grand Rapids, Michigan (Fig. 3), and Milwaukee (not shown) were studied. At both sounding sites, the surface winds as entered by the operator were offshore, indicating that both sounding sites were influenced by the land-breeze circulation.
The RASS data from Sheboygan (Fig. 2a) indicated uniformly easterly flow from the lowest level (95 m) to a level just below 1 km at 1200 UTC. This pattern could either reflect the land-breeze return circulation or, more likely, the ambient synoptic flow over the western Great Lakes at 1200 UTC. The fact that all coastal stations along the western shore of Lake Michigan were reporting offshore flow at 1200 UTC suggests that a land breeze was established at that time. However, the RASS data from Sheboygan suggest that the offshore component of the land breeze was confined to the lowest 100 m.
The RASS data from Montague (Fig. 2b) indicated that the surface ridge axis had just passed by the site between 0800 and 1200 UTC (as indicated by the light and variable near-surface winds during that time). At 1200 UTC, easterly flow was becoming established at the Montague sounding site. This easterly flow on the eastern shore of the lake is consistent with the offshore component of a land-breeze circulation. However, easterly flow there is also consistent with the general synoptic flow pattern at 1200 UTC. The fact that there was no evidence of a westerly return circulation in the lowest 500 m at Montague suggests that these offshore winds were of synoptic origin. However, as discussed below, the VAD wind profiles from Grand Rapids did provide evidence that a return circulation existed on the eastern shore of Lake Michigan.
In addition to the PASS wind profiler data described above, VAD wind profiles from Grand Rapids (GRR) and Milwaukee (MKX) were examined. The lowest reported wind observations from the MKX VAD winds were at 2000 ft (∼600 m) and agreed with the RASS data from Sheboygan. Therefore, they are not shown here. The GRR VAD wind profile was also consistent with the nearby Montague sounding; however, some important differences exist. Figure 3 shows the VAD winds for the time period from 1150 to 1330 UTC for the GRR radar site. Note that the lowest wind information at this site is at 1000 ft (∼300 m). Further, note that there is some evidence of weak onshore flow at the lowest level from 1240 to 1320 UTC. This may be the VAD wind signature of the uppermost part of the return circulation associated with the eastern land breeze. The RASS data from Sheboygan coupled with the GRR VAD wind profiles indicate that the proposed eastern land breeze evident in the surface data is confined to the lowest 300 m of the boundary layer while the proposed western land breeze is confined to the lowest 100 m of the boundary layer.
In order to study the time evolution of the land-breeze circulation, a series of base-level reflectivity plots from the WSR-88D radar at Milwaukee were examined. The times studied were from 1000 to 1400 UTC. Figure 4 shows radar reflectivity from the 0.5° elevation angle scan every 30 min starting at 1133 UTC and ending at 1302 UTC. Within this series of images, a distinct feature in the ground clutter around Milwaukee develops and expands offshore. Of particular interest is the development of a band of 8+ dBZ reflectivity that develops between 1203 (Fig. 4b) and 1233 UTC (Fig. 4c). This feature is also present at 1302 UTC. This structure is similar to those seen in previous radar studies of land- and sea-breeze circulations (Meyer 1971). Furthermore, the location is generally consistent with the location of the sharp backscatter front evident in the SAR image. We therefore hypothesize that this feature is the WSR-88D signature of the land breeze and land-breeze front. Because this structure is developing and propagating offshore, the appropriate symbology is that of a developing land-breeze front (Young and Fritsch 1989) for the western reflectivity structure.
4. SAR image analysis
In this section, we show how the two land-breeze circulations and the possible convection between them are evident in a RADARSAT SAR image (Fig. 5) taken at 1208 UTC 14 January 1998. This image justifies the exact placement of the frontal symbology shown in Fig. 1b.
Figure 5a shows the image as originally collected; Fig. 5b is the same as Fig. 5a but with additional information. For geographical reference, the major cities along the western shore of Lake Michigan, the location of ISS2, and the location (point A) of the ship observations reported in section 3 are indicated. At each of these locations, surface wind observations are also indicated in Fig. 5b.
a. Western land breeze
The primary feature of interest in Fig. 5 is the bright area extending eastward from the western shore of Lake Michigan. The eastern edge of this bright area extends northward from southern Milwaukee to the peninsula between Sturgeon Bay and Lake Michigan. Given that the wind direction was offshore at all of the coastal stations along the immediate western shore of Lake Michigan, the bright area is hypothesized to be the surface footprint of the land breeze as imaged by the SAR. Supportive of this hypothesis is the fact that when a sea or land breeze is embedded in opposing synoptic flow, the leading edge of the flow is characterized by a sharp wind shift and strong temperature gradient (Estoque 1962; Passarelli and Braham 1981; Meyer 1971; Chiba 1993; Arritt 1993). In addition, Meyer (1971) reported ground-based radar observations that indicated a sharp reflectivity front at the leading edge of a land-breeze circulation off the east coast of the United States. Therefore the fact that the boundary at the eastern edge of the first SAR feature is sharp (quantified in section 4c) makes it a likely candidate for the leading edge of the land breeze: essentially, the land-breeze front. Further support for this interpretation is the fact that the wind observed at 1200 UTC by the ship (point A) was from 40° at 3.5 m s−1. This observation is consistent with the fact that the ship was located east of the sharp boundary, where an easterly component to the winds would be expected. The choice of developing land-breeze front symbology was driven by the MKX reflectivity analysis indicating that the land breeze was strengthening at 1200 UTC.
Another important observation to be made from the brightness pattern visible in the SAR image is the apparent influence of horizontal variations in surface roughness characteristics along the western shore of the lake on the radar backscatter offshore of these variations. Figure 6 shows a close-up of the Sturgeon Bay Ship Canal and the brightness patterns visible offshore. The fan-shaped brightness signature at the mouth of the Sturgeon Bay Ship Canal indicates that there is a connection between variations in the surface properties of the coast and the brightness signatures over the lake. Similar signatures exist at other points along the lake. Terrain-influenced signatures have previously been observed in SAR imagery (e.g., Pan and Smith 1999; Alpers et al. 1998; Winstead and Young 2000).
b. Eastern land-breeze signature
The eastern part of the SAR image is characterized by relatively uniform brightness with a detectable but somewhat diffuse western boundary. This structure is similar to that near the western shore. Unfortunately, the lack of in situ observations over Lake Michigan and the limited number of observations along the eastern shore of Lake Michigan preclude full knowledge of this feature. Thus any explanation of its origins is only speculative. One possible explanation is that the features responsible for this SAR image signature is the land breeze from the eastern shore of the lake superimposed on the easterly synoptic flow. Such an explanation is plausible for the following reasons. First, the ambient flow is easterly. Thus, any land breeze off the eastern shore would be flowing in the same direction as the ambient flow and would be displaced farther west. This type of behavior is well documented in the literature for sea-breeze circulations (Atkinson 1981; Pielke 1984), and the principle should also apply to land-breeze circulations. Second, the diffuse nature of this boundary (quantified below) is consistent with what would be expected for a land breeze propagating with the ambient flow, based on previous studies that found that land- and sea-breeze fronts become diffuse when the ambient flow is in the same direction as the propagation of the fronts (Meyer 1971; Atkinson 1981; Passarelli and Braham 1981; Hjelmfelt and Braham 1983; Arritt 1993). Finally, as discussed above, the VAD wind profile from Grand Rapids did indicate a return circulation consistent with the existence of a land-breeze circulation originating from the eastern shore of Lake Michigan near the time of the image.
c. Analysis of frontal zone width from SAR
In order to quantify the widths of the fronts described in sections 4a and 4b, a cross-lake cut was taken through a portion of the SAR image. (Its location is indicated by the long, horizontal rectangle in Fig. 5b.) This cut was averaged over 50 pixel rows (5 km) to remove small-scale fluctuations and was taken along a section of the coast where both land-breeze fronts were approximately aligned in a north–south orientation. The purpose of this analysis was to detect and quantify the width of each of the two land-breeze frontal zones as defined by a change in the magnitude of the image brightness. Figure 7 shows the results of this analysis. The two land-breeze regions, the two land-breeze fronts, and the central lake-modified air mass are labeled. The two land-breeze fronts are detectable as significant changes in the magnitude of the grayscale values over relatively short horizontal distances. Specifically, the image brightness decreases from 121.3 to 80.7 over a distance of 6 pixels (600 m) for the western front and increases from 89.2 to 127.4 over a distance of 29 pixels (2900 m) for the eastern front. Thus the eastern boundary is significantly wider than the western boundary. This fact is consistent with previous observations of land- or sea-breeze circulations and their interaction with the ambient synoptic flow (Meyer 1971; Atkinson 1981; Chiba 1993). Also, the width of the western front is within the range of widths previously reported for land-breeze frontal zones. However, the eastern front is wider than the widest fronts reported in the literature for this kind of land breeze (Atkinson 1981; Chiba 1993). Chiba (1993), in particular, reported a range of frontal widths from 130 to 1120 m with a mean value of 560 m. Similar analyses (not shown) at other positions along the front show similar results.
d. Central convective region
Finally, between both proposed land-breeze signatures in the SAR image is a narrow region characterized by the mottled pattern reported by various authors (Sikora 1995, 1997; Zechetto et al. 1998) as being associated with a highly convective boundary layer. Published work shows that this mottled pattern is the surface signature of the updraft and downdraft patterns associated with boundary layer convective eddies under conditions of weak mean wind. The presence of boundary layer convection is not surprising given the air–sea temperature difference of −9°C reported by the ship at point A (Fig. 5b).
The existence of this intermediate zone is consistent with what would be expected in the region between two opposing-shore land breezes. In their study, Passarelli and Braham (1981) reported on aircraft observations taken during a traverse of Lake Michigan that transected land-breeze circulations from both shores. The aircraft data indicated that the land-breeze fronts from either side of the lake had not yet met. Between the fronts, they found a region with little or no potential temperature structure (horizontal potential temperature gradient) where the air was in approximate thermal equilibrium with the lake. In our case, there is no information on the horizontal potential temperature structure over the lake; however, the ship-observed temperature of −5°C at 1200 UTC is significantly higher than those along either shore of the lake. This temperature observation, coupled with the SAR signature of boundary layer convection, suggests that the ship was embedded within a region of air that was significantly influenced by Lake Michigan.
A lake-induced thermal circulation over Lake Michigan was analyzed using conventional meteorological data supplemented by additional meteorological observations taken over the Great Lakes region during Lake-ICE. It was found that a pronounced warm anomaly and associated pressure trough existed over Lake Michigan at 1200 UTC 14 January 1998. Associated with this structure was a land-breeze circulation that was evident on both sides of Lake Michigan. The RASS wind profiles from Sheboygan and Montague coupled with VAD wind profiles from Grand Rapids suggested that the land-breeze circulations on both sides of the lake were relatively shallow with the western circulation confined to the lowest 100 m of the boundary layer and the eastern circulation confined to the lowest 300 m of the boundary layer.
As a supplement to the surface observations, a RADARSAT SAR image taken at the same time was analyzed. The SAR image provided valuable information about the precise location of the land-breeze fronts associated with the eastern and the western shores of the lake and the convective region between them. This information was used to augment the mesoscale analysis in Fig. 1b, which showed that land breezes should exist but with an unknown reach away from the shore. Specifically, land-breeze symbology (Fig. 5b) was added to the SAR image to pinpoint the eastern and western extent of the offshore flow. From an operational forecasting standpoint, SAR imagery may be useful for predicting the evolution of land breeze (induced snowbands) over the Great Lakes.
An analysis of the grayscale values for a cross-lake cut of the SAR image showed that the western land-breeze front was a factor of 2 narrower than the eastern land-breeze front. This was explained by the dynamics of sea- and land-breeze circulations embedded within ambient synoptic flow.
Finally, it was shown that upstream terrain and/or surface land use variations affected the surface footprint of the land-breeze flow on the western side of the lake. The fact that SAR has sufficient resolution to resolve these variations suggests that SAR may be an important remote sensing asset for observing the subtle structure in the nearshore surface wind field.
The authors gratefully acknowledge Harry Stern and Yanling Yu of the Applied Physics Laboratory, University of Washington, for their help with processing the SAR image and the AVHRR image shown in this paper. The authors are grateful to Paul Zibton for putting together the figures and Agnes Sieger for her help with the text. In addition, the authors thank the Joint Office for Scientific Support (JOSS) for providing the data in an easily accessible format and the Great Lakes Environmental Laboratory, specifically the NOAA Coast Watch/Great Lakes Node program, the National Climatic Data Center, and the Satellite Active Archive, for providing the original data archived by JOSS. Finally, the authors are grateful to David Kristovich of the Illinois State Water Survey for conceptualizing, organizing, and running Lake-ICE. This study was funded by NSF Grant ATM-9707730.
Corresponding author address: Nathaniel S. Winstead, The Johns Hopkins University Applied Physics Lab, 11100 Johns Hopkins Road, Laurel, MD 20723.