Lake-effect snowstorms generally develop within convective boundary layers, which are induced when cold air flows over relatively warm lakes in fall and winter. Mesoscale circulations within the boundary layers largely control which communities near the downwind shores of the lakes receive the most intense snow. The lack of quantitative observations over the lakes during lake-effect storms limits the ability to fully understand and predict these mesoscale circulations. This study provides the first observations of the concurrent spatial and temporal evolution of the thermodynamic and microphysical boundary layer structure and mesoscale convective patterns across Lake Michigan during an intense lake-effect event. Observations analyzed in this study were taken during the Lake-Induced Convection Experiment (Lake-ICE).
Aircraft and sounding observations indicate that the lake-effect snows of 13 January 1998 developed within a convective boundary layer that grew rapidly across Lake Michigan. Boundary layer clouds developed within 15 km and snow developed within 30 km of the upwind (western) shoreline. Near the downwind shore, cloud cover was extensive and snow nearly filled the boundary layer. Extensive sea smoke in the surface layer, with disorganized (or cellular) and linear features, was observed visually across the entire lake. Over portions of northern Lake Michigan, where airborne dual-Doppler radar observations were obtained, the mesoscale circulation structure remained disorganized (random or cellular) across the lake. Given observed shear and stability conditions in this region, this structure is consistent with past theoretical and numerical modeling results. To the south, where surface winds were slightly stronger and lake–air temperature differences were less, wind-parallel bands indicative of rolls were often present.
The horizontal scale of the observed mesoscale convective structures grew across Lake Michigan, in agreement with most previous studies, but less rapidly than the increase of the boundary layer depth. The decreasing ratio of convective horizontal size to boundary layer depth (aspect ratio) is contrary to many recent studies that found a positive correlation between boundary layer depth and aspect ratio.
Lake-effect snowstorms are manifestations of the development and evolution of convective internal boundary layers (CBLs), which occur when cold air flows over the relatively warm surfaces of the Great Lakes and other midlatitude lakes. In recent years, there has been a notable improvement in forecasting the occurrence of lake-effect events. However, mesoscale circulations within lake-effect CBLs, which control the location and intensity of snowfall, are often complex and much more difficult to predict. Improvements in our understanding of mesoscale lake-effect circulations require detailed observations over and near the Great Lakes (e.g., Kristovich et al. 1999; Scott and Sousounis 2001). Observations of the concurrent cross-lake evolution of mesoscale circulations and CBL thermodynamic and microphysical characteristics, however, have not been available.
This study builds on past knowledge of mesoscale boundary layer circulations derived mainly from numerical modeling and nearshore observations. Common mesoscale lake-effect structures over individual lakes can be classified into distinct morphological types (Braham and Kelly 1982; Kelly 1986; Hjelmfelt 1990; Kristovich and Steve 1995; Niziol et al. 1995; Laird et al. 2003). The four most commonly discussed types include 1) widespread coverage (e.g., Kelly 1982, 1984; Kristovich 1993; Steve 1996; Kristovich and Laird 1998), 2) shoreline bands (e.g., Braham 1983; Hjelmfelt and Braham 1983; Schoenberger 1986), 3) midlake bands (e.g., Passarelli and Braham 1981), and 4) mesoscale vortices (e.g., Forbes and Merrit 1984; Pease et al. 1988; Laird 1999).
Kelly (1986) found that widespread coverage of lake-effect clouds and snow was the most common mesoscale structure over Lake Michigan. Kristovich and Steve (1995) later examined satellite data for all of the Great Lakes, and found widespread events were most common over nearly every lake. Widespread lake-effect snow events are most often composed of boundary layer rolls (e.g., Kelly 1982, 1984; Kristovich 1993), cells, or random convection (e.g., Braham 1986), or combinations of these convective structures (e.g., Kristovich et al. 1999; Cooper et al. 2000). Similar linear and cellular features in the lake-effect surface layer have been observed by horizontally scanning lidar near the upwind shore of Lake Michigan (Mayor 2001). However, the relationship between surface layer and linear structures extending throughout the depth of CBLs is not well understood (e.g., Young et al. 2002). Several recent studies have identified the importance of wind speed in the surface layer to the development and structure of boundary layer convection. For lake-effect environments, Kristovich (1993), Kristovich et al. (1999), and Cooper et al. (2000) showed, using observations and numerical modeling, that the near-surface shear magnitude is an important contributor to lake-effect boundary layer convective structure. This relationship also has been shown for other convective environments (e.g., Weckwerth et al. 1997).
Atmospheric measurements collected during the Lake-Induced Convection Experiment (Lake-ICE; Kristovich et al. 2000) provide unique and important simultaneous information on winter mesoscale circulations and smaller-scale turbulent and microphysical characteristics over Lake Michigan. This paper quantitatively describes the spatial and temporal evolution of the CBL and mesoscale circulations observed across Lake Michigan during an intense lake-effect case with widespread convection on 13 January 1998.
2. Data and methodology
a. Data collection on 13 January 1998
An overview of Lake-ICE operations and goals are given in Kristovich et al. (2000). Project facilities used for the present study included three National Center for Atmospheric Research (NCAR) Integrated Sounding System (ISS) sites located in Michigan and Wisconsin, two aircraft (NCAR Electra and the University of Wyoming King Air) deployed over Lake Michigan, and the NCAR Electra Doppler Radar (ELDORA). Figure 1 shows the approximate locations of the Electra and King Air (KA) research flights on 13 January 1998, ISS sites, and operational National Weather Service (NWS) sounding sites. In addition, visible imagery from the National Oceanic and Atmospheric Administration (NOAA) Geostationary Operational Environmental Satellite-8 (GOES-8) and observations from NWS Weather Surveillance Radar-1988 Doppler (WSR-88D) radar sites (Fig. 1) at Green Bay, Wisconsin, and Grand Rapids and Alpena, Michigan, were used to gain further information on convective morphologies over and near Lake Michigan. Lake surface temperatures were retrieved from Advanced Very High Resolution Radiometer (AVHRR) satellite observations of the Great Lakes (Schwab et al. 1999).
Rawinsondes were launched from the three ISS sites every 90 min for a 9 h time period, from 1200 to 2100 UTC on 13 January. Thereafter, rawinsondes were launched every 6 h at ISS2 and ISS1 until 1200 UTC on 15 January. These soundings provided data on the temporal evolution of the CBL and overlying air upwind and downwind of Lake Michigan. Observations from the 0000 and 1200 UTC NWS rawinsondes at Green Bay WI were also used to supplement the ISS soundings.
On 13 January 1998, the Electra flew repeatedly across Lake Michigan nearly parallel to the wind direction within the convective mixed layer (Fig. 1) at an altitude of approximately 550–600 m above the lake surface (ALS). Twelve cross-lake flight segments were conducted between about 1355 and 1710 UTC. The primary purpose of these flight segments was to collect ELDORA and in situ observations across Lake Michigan. For the final hour of the Electra flight over Lake Michigan, a stack of cross-wind, constant-height flight segments were completed close to the eastern shore (not shown). The KA conducted four stacks of level flight segments approximately perpendicular to the Electra flight path (Fig. 1). The flight segments in each KA stack were conducted at multiple altitudes over the same approximate location, extending from about 7 to 30 km north of the Electra flight path. The KA flight stacks consisted of one to three segments within the cloud deck and one segment above the clouds. The stacks closest to the eastern and western shores of Lake Michigan included flight segments in the lower portions of the convective mixed layer, near 150 m ALS. Visual flight restrictions prevented the KA from penetrating below the cloud layer on the two flight stacks over central Lake Michigan due to a lack of breaks in the clouds. The KA and Electra were equipped with instrumentation for measurement of state, dynamic, radiative, and particle variables. These measurements gave valuable information on cross-lake changes in the vertical structure of the CBL and mesoscale convective structures.
The ELDORA provided an unprecedented dataset on the evolution of convective structures across Lake Michigan on 13 January 1998. The ELDORA is composed of two 3-cm-wavelength Doppler radars that scan around the axis of the Electra's fuselage at angles 18° forward and rearward from the perpendicular. Data from these scans can be combined to provide dual-Doppler radar coverage over regions on either side of the Electra. More information on ELDORA and the observation method employed is provided in Hildebrand et al. (1996).
The lake-effect environment in which ELDORA data were collected is not ideal for airborne Doppler radar observations. The convection on this date was shallow, with the boundary layer height increasing from approximately 200 m ALS over the western region of the lake to about 900 m ALS over the eastern region. The primary radar scatterers were snow crystals. The observed effective reflectivity factor (hereafter referred to as reflectivity) was generally less than 10 dBZ. On 13 January 1998, ELDORA was able to observe reflectivity patterns up to 25 km from the Electra (at locations eastward of about midlake, where abundant snow was present), upper-level convergence/divergence patterns near the top of the CBL out to about 10–15 km, and shallow near-surface convergence–divergence patterns out to about 5–10 km from the Electra.
Analyses of ELDORA data were accomplished using methods similar to those described by Hildebrand et al. (1996) and Wakimoto et al. (1996). The NCAR/Mesoscale and Microscale Meteorology (NCAR/MMM) software package SOLO (Oye et al. 1995) was used for editing the radar data and REORDER was used for interpolation of these data to Cartesian coordinates (Oye 1994). Calculation of derived fields (e.g., wind velocities, divergence) was accomplished with the NCAR/MMM Custom Editing and Display of Reduced Information in Cartesian Space program (CEDRIC; Mohr et al. 1986). For this study, Cartesian grid resolutions of 400 m horizontally and 200 m vertically were used and analyses of derived fields were confined to within about 12 km of the aircraft.
It is necessary to carefully evaluate the collected ELDORA data, particularly in lake-effect conditions that were not ideal for radar retrievals. In order to test for self-consistency of the ELDORA observations, the relationship between the convergence and reflectivity fields was examined. Figure 2 shows an example of ELDORA-observed reflectivity and derived convergence fields interpolated to a level near the top of the CBL (0.9 km ALS) within a region near the center of Lake Michigan. The x axis is oriented along the aircraft track, 292°–112°. The reflectivity pattern in this location shows a disorganized (random/cellular) convective structure, with some indication of both along-wind and cross-wind linear patterns. ELDORA wind speeds and directions (6–8 m s−1 from about 320°) were consistent with overlake aircraft in situ observations. Regions of divergence (light shading) indicate areas of mesoscale updraft and were collocated with, or slightly upwind of, regions of higher reflectivity (black lines). These observations demonstrate the quality and consistency between the reflectivity and Doppler velocity measurements and indicate that boundary layer convective circulations can be resolved close to the aircraft in these conditions.
The ELDORA-observed wind and reflectivity fields were consistent with observations from the KA, flown as close as 7 km north of the Electra flight path. Figure 3 shows KA observations of vertical motions taken while the aircraft flew at 550-m height near the location of the ELDORA fields shown in Fig. 2. The heavy line shows 11-s running averages of vertical motions, approximating the ELDORA beamwidth in the radar analysis regions. Peak smoothed vertical motions ranged from about +2.0 to −1.5 m s−1. These values are similar to, but stronger than, the peak vertical motions derived from ELDORA observations—approximately ±1.0 m s−1. The difference may be due, in part, to difficulties in obtaining ELDORA observations of the shallow near-surface divergence fields. Peaks in aircraft-observed vertical motions occurred on average every 30 s, corresponding to a convective updraft spacing of 2–3 km. This agrees well with the spacing between peaks in both vertical motions and reflectivity observed by ELDORA near this location.
Convective structures observed by the ELDORA could not be directly compared with those from WSR-88D sites near Lake Michigan. The Electra flights were located between regions where WSR-88Ds were able to detect the shallow lake-effect convective structures that developed on 13 January 1998. About 70–100 km south of the Electra flight path, the Grand Rapids WSR-88D observed mainly linear convective patterns over land and a combination of linear and random convective patterns over the lake. More discussion of the spatial variation of observed convective patterns is given in section 4.
3. Synoptic conditions and boundary layer growth
A succession of cold fronts and surface pressure troughs moved through the Great Lakes region from 9 to 13 January, followed by surges of cold air and lake-effect snowstorms. Figure 4 shows the NOAA surface weather analysis for 1200 UTC on 13 January 1998. The primary cold front was located over the eastern Great Lakes region, after having crossed Lake Michigan on 12 January. West-northwest winds and low-level cold-air advection were present over the western Great Lakes throughout the day on 13 January.
At 1400 UTC, during Lake-ICE flight operations, surface temperatures were about −18° to −26°C upwind and −10° to −13°C downwind of Lake Michigan. Lake surface temperatures were 2°–5°C, resulting in surface lake–air temperature differences averaging 15°–30°C. Wind speeds of near 8 m s−1 were observed within 200 m of the lake surface by Lake-ICE aircraft.
Surface fluxes were estimated using eddy correlation techniques with KA observations and using bulk techniques (e.g., Garratt 1992) with nearshore air temperatures and lake water temperatures. The KA flight segments at about 120 m ALS near the upwind and downwind shores gave average rates of 120 and 70 W m−2 for sensible and latent heat fluxes, respectively. No significant difference in surface KA-observed fluxes from the upwind to the downwind shore was noted. Since the aircraft flight segments were above the surface layer, fluxes derived from these data are likely underestimates of surface values. Bulk estimates using observations from Muskegon, Michigan, gave surface sensible heat fluxes from 148 to 174 W m−2, latent heat fluxes from 112 to 134 W m−2, Bowen ratios near 1.3, and negative ratios of the CBL depth to the Monin–Obukhov length (−Zi/L) of 110–150. These values are generally consistent with those observed over Lake Michigan during widespread coverage lake-effect events (Kelly 1984; Kristovich 1993; Kristovich and Laird 1998).
The growth and development of the CBL was rapid over the 90–100-km cross-lake fetch between ISS sites (see Fig. 1 for locations). Boundary layer temperatures increased 8°–10°C below 900 hPa and moisture contents approached saturation with respect to ice near the downwind shore. Upwind and downwind rawinsonde observations (Fig. 5) indicate that the well-mixed layer (nearly constant equivalent potential temperature) increased in depth, with the height of the CBL top increasing from 960 hPa (about 200–300 m ALS) at ISS2 (Sheboygan, Wisconsin) to about 910 hPa (900 m ALS) at ISS1 (Muskegon). The CBL remained nearly steady state and exhibited a virtually constant growth rate from the western to eastern shores during a 6-h time period encompassing the aircraft observations. Interestingly, potential temperature profiles changed little from ISS2 to about 90 km inland at ISS3 (Greenville, Michigan). Any cooling from the traverse across snow-covered ground appeared to have been small.
Figure 5b gives profiles of wind speed and direction observed near the downwind shore of Lake Michigan on 13 January at 1500 UTC. Wind speed shear was evident throughout the CBL, particularly below 970 hPa and above about 900 hPa. Little wind direction shear was observed in the CBL, and weak backing of the winds was observed above, indicative of cold air advection. The influences of this wind profile on mesoscale convective structures are discussed in section 4.
A more comprehensive view of the evolution of CBL growth across Lake Michigan on 13 January is examined using a composite vertical cross section of equivalent potential temperature (θe), derived from Lake-ICE aircraft and rawinsonde observations, and NWS observations at Green Bay (Fig. 6). Ice processes in the cloud layer did not significantly alter the profile, and adjustments of θe were therefore not included in this analysis. Despite differences in locations and time of observations, the measurement platforms showed consistent observations within the nearly steady state CBL. Boundary-layer-mean θe increased by approximately 7 K from the upwind to the downwind shore. The height of the lowest inversion rose more rapidly over the upwind half of Lake Michigan than over the downwind half, similar to observations shown in Chang and Braham (1991) and Kristovich et al. (1999).
Figure 7 shows several estimates of CBL-top height increases across Lake Michigan, based on locations of lake-effect cloud tops (estimated as height where cloud cover had decreased to 10% of the aircraft passes), delineation of the height of rapidly decreasing vapor (estimated by mixing ratios of 0.40 g kg−1), and where turbulence rapidly decreased with height (estimated as turbulent motions of 0.50 m2 s−1). These estimates are in close agreement, and indicated the mean increase in height of the lowest inversion across Lake Michigan was approximately 7.0 m km−1. This value was within the range of previously reported growth rates over Lake Michigan in cases with westerly component CBL winds and widespread lake-effect convection (Table 1). The weak correlation between surface heat fluxes and CBL growth rates shown in Table 1 is not surprising, but does not account for other important factors, such as ambient stability (Kristovich and Laird 1998), inversion-layer depths (Byrd et al. 1991), presence of an elevated mixed layer (Agee and Gilbert 1989; Chang and Braham 1991; Schroeder 2002), influences of latent heat release (Cooper et al. 2000), and nearshore topographic influences (Hjelmfelt 1992).
The KA flight stacks allowed for an examination of thermodynamic and microphysical characteristics of the CBL (Fig. 8). Flight-segment-mean equivalent potential temperatures (Fig. 8a) indicate that the KA flew within and above the mixed layer near the upwind and downwind shores. The KA flight stacks in the middle of the lake were within the entrainment zone and stable air above. CBL-mean mixing ratios (Fig. 8b) approximately doubled across the lake as a result of large surface moisture fluxes. Very few clouds were observed at the flight stacks closest to the upwind shore (Fig. 8c). Visually, these few clouds were determined to be the tops of widespread “sea smoke” (for an example, see Fig. 1 in Young et al. 2002). Over the rest of the lake, clouds were observed to increase in coverage up to almost 70% at the aircraft flight levels, similar to observations given by Lenschow (1973), Chang and Braham (1991), and Kristovich (1993). No clouds were observed by the KA at 150 m ALS near the downwind shore; however, sea smoke continued to be visually observed. Snow particles were observed at lowest levels across the entire lake (Fig. 8d). Surprisingly, ice particles were more frequently observed than clouds near the upwind shore. It is possible that the ice particles at this location were previously glaciated sea smoke clouds or snow blown from nearby land areas of Wisconsin. Most of the CBL was filled with snow (over 70%) near the downwind shore, which provided scatterers for mesoscale ELDORA observations.
4. Mesoscale evolution
The synoptic environment gives rise to conditions suitable for lake-effect storms, but mesoscale boundary layer circulations more directly control the organization, location, and evolution of lake-effect snows (e.g., Hjelmfelt 1990; Laird et al. 2003). Figure 9 shows the visible GOES-8 imagery for 1502 UTC 13 January 1998, near the time of aircraft observations. Satellite-observed clouds are visible within about 15 km of the western shore of Lake Michigan. Linear cloud features were visible in the southeastern portions of the lake, as confirmed by observations near the southeastern shore (Miles 2002). Cloud patterns in the northern half of the lake, where Lake-ICE aircraft operated, did not indicate clear mesoscale structures. However, some regions appeared to have cellular cloud structures.
ELDORA observations were utilized to give information on the spatial and temporal evolution of convection across Lake Michigan in the boxed region in Fig. 9. As an example, Fig. 10 provides a plan-view radar reflectivity analysis for about 1600 UTC. A disorganized convective pattern, with a few locations resembling open cellular convection, was observed at all times during the 4-h Electra flight. The observed convective structure is consistent with “random only” convective structures, as expected using criteria such as given by Grossman (1982) for the aircraft-observed lake–air temperature differences (15°–25°C), CBL wind speeds (8 m s−1), and −Zi/L (110–150). Interestingly, the surface flux estimates were within those observed in cases with boundary layer rolls by Kristovich (1993) and lake–air temperature difference and wind speeds reported for roll events by Kelly (1982). Linear cloud features indicative of rolls were often observed in southern regions of the lake (consistent with Miles 2002), where winds were somewhat stronger (and thus the near-surface wind shear was greater; see sounding data from Muskegon in Fig. 5b) and lake–air temperature differences were smaller.
The ELDORA reflectivity analysis shown in Fig. 10 also shows some evidence of linear features oriented at large angles to the CBL wind direction. While these features are nearly obscured by the complex convective field, they appear in portions of nearly all of the ELDORA cross-lake reflectivity fields obtained on 13 January 1998. Spatial autocorrelation analyses [as used by Weckwerth et al. (1997) and Kristovich et al. (1999)] were completed for four adjacent 10 km × 10 km regions along the flight segment to determine the presence of repeated patterns in the reflectivity field. Figure 11 shows an example autocorrelation field, for 1400 UTC. The autocorrelation fields revealed a weak tendency for banded patterns with orientations of approximately 45°–225° to 70°–250°, more than 40° from the mixed-layer wind direction of about 320°. The large difference in angle between the orientation of lines of reflectivity and the mixed-layer wind direction suggests that horizontal convective roll circulations were not responsible for this feature. The example shown in Fig. 11 met the criterion for linearity given by Weckwerth et al. (1997), but only about half of the reflectivity regions met this criterion. Because wind shear directions were roughly perpendicular to the cross-wind bands, convectively induced gravity waves (e.g., Clark et al. 1986; Kuettner et al. 1987) are suspected as the cause of these banded features. However, it is not possible with the current dataset to determine the cause of these bands.
Figure 10 indicates a significant increase of radar reflectivity from west to east. Figure 12 quantifies the average ELDORA-observed reflectivity values on the north and south sides of the Electra flight path at approximately 1400, 1500, and 1600 UTC. The average reflectivities were calculated for four adjacent 10 km × 10 km regions along the Electra path, from −5 to 35 km in Fig. 10. On average, a rapid increase in reflectivity was observed at distances of −5 to about 15 km and a more gradual increase or nearly constant reflectivity to the east (distances over 15 km). This pattern of reflectivity increase agrees with the over lake snowfall rate pattern derived by Braham and Dungey (1995) by combining aircraft precipitation particle observations from a number of widespread snow cases. There was considerable spatial variability in the rate of reflectivity increase along the Electra flight path partially because of variations in mesoscale convective structure at scales close to the 10 km × 10 km sample regions. In addition, substantial temporal variability was observed between flight segments due to the eastward propagation of individual convective elements within the cross-lake reflectivity field.
It might be expected that the dominant mesoscale convective scale would increase as the CBL grew from west to east (Figs. 6 and 7). The mean horizontal scale of convective structures was estimated from ELDORA observations by determining central locations of all high-reflectivity regions. Figure 13 gives the mean distances between local reflectivity peaks as a function of location along the ELDORA path for four time periods. In general, there was a notable, but slow, increase in convective wavelength, on average increasing 12%–21% from west to east.
It is desirable to have additional quantitative information on convective size scales to complement the ELDORA observations. Unfortunately, a direct comparison between these observations and satellite or nearby WSR-88D sites was not possible, and aircraft flights were not taken at midlevel in the CBL at all locations over the lake. However, some information on mesoscale convective scale changes can be gained from KA flight segments above the cloud-topped CBL. Spatial variations in cloud-top temperature, based on observations from the Heimann downward-pointing radiometer, were examined to gain information on spatial variations in mesoscale convective regions. Examination of cloud-top temperatures from four cross-wind KA flight segments from near the western (Fig. 14a) to near the eastern shore (Fig. 14d) shows clear evidence of an increase in convective scale. Regions where the lake surface was visible had temperatures of 0°–3°C, while cloud-top temperatures were 10°–20°C cooler. Individual cloud elements were observed much more frequently over western regions (Figs. 14a and 14b) of the lake than downstream closer to the eastern shore (Figs. 14c and 14d), where larger clouds were observed approximately once per minute on average (about 5-km flight distance). Fourier analyses of these data indicate increasing power in the lower frequencies, suggesting an increase in convective wavelength, but no single peak in wavelength was dominant. Both the ELDORA and KA in situ observations provide consistent qualitative evidence that a discernable increase in convective size scale existed with fetch across Lake Michigan on 13 January 1998.
While observations indicate that the convective wavelength increased 12%–21% across Lake Michigan, the CBL-top height increased over a similar distance by 50% or more (Fig. 6). The specific percentage increases depend on which method is used to estimate CBL-top height. These changes in convective wavelength and CBL depth translate into a decrease of the mean aspect ratios (wavelength/depth) of the mesoscale convection by at least 20%–25% across the lake. This decrease in aspect ratio is contrary to most previous studies, which indicate increased aspect ratios with convective depth. This will be discussed in more detail in section 5.
5. Summary and concluding remarks
This study utilizes observations taken during the Lake-Induced Convection Experiment (Lake-ICE) of an intense lake-effect snowstorm. The first detailed observations of the concurrent spatial and temporal evolution of both the convective boundary layer thermodynamic and microphysical structure and mesoscale convective patterns across Lake Michigan in a lake-effect snowstorm are presented. Aircraft and sounding observations indicate that an inversion-capped CBL grew rapidly across Lake Michigan, with clouds developing within 15 km and snow developing within 30 km of the upwind (western) shoreline. Airborne dual-Doppler radar (ELDORA) observations gave unprecedented information on the evolution of the mesoscale convective structures across Lake Michigan, despite the fact that the convection was quite shallow and the primary radar scatterers were snow. ELDORA observations indicated that the mesoscale circulation structure was dominated by disorganized (cellular/random) convection. Weak bands of high reflectivity were evident at large angles to the mean CBL wind direction, and are unlikely due to roll convection. It is suggested that convection-induced gravity waves near the CBL top may be responsible for these features. Convection in regions to the south of the aircraft observations, where surface winds were slightly stronger and lake–air temperature differences were less, often exhibited features consistent with roll convection. These findings are consistent with theory and past numerical and observational studies. The horizontal scale of the mesoscale convective structures grew across Lake Michigan; however, the rapid growth of the CBL was faster than the horizontal scale growth. This resulted in a decreasing aspect ratio with fetch across the lake.
The observed decrease in aspect ratio with fetch is contrary to the findings of a number of studies of mesoscale structures in convective boundary layers. Many observational and numerical modeling studies have found a positive correlation between aspect ratios of rolls and cellular convection and boundary layer depth. For example, in a study of several cases of roll convection in lake-effect storms, Kelly (1982) observed a tendency for increasing aspect ratio with increasing Zi. In a wide-ranging review, Young et al. (2002) found that aspect ratios of roll convection in cold air outbreaks over the ocean and large lakes tended to increase with boundary layer depth. It should be noted, however, that a few studies found different relationships between aspect ratio and depth. Melfi et al. (1985) found decreasing aspect ratios with distance of cold air flow over the ocean off the east coast of the United States (and therefore increasing Zi). Lohou et al. (1998) and Young et al. (2002) reported no consistent relationship between aspect ratio and boundary layer depth over land areas.
Future observations and numerical modeling work will be required to fully understand the relationships between environmental and surface conditions, convective aspect ratios, and boundary layer depth. Latent heat release in cloud formation (e.g., Sheu et al. 1980; van Delden 1985; Huang and Kallen 1986; Huang 1988; Chlond 1988), entrainment processes in CBL clouds (Sykes et al. 1988), non-linear processes (Rothermel and Agee 1986; Mourad and Brown 1990), differential stability of updrafts and downdrafts (Muira 1986), and interactions with convection-induced gravity waves (e.g., Clark et al. 1986; Hauf and Clark 1989) are among the factors previously found to be related to convective wavelength and aspect ratio. Future studies will need to determine how such factors change in an evolving boundary layer structure with varying surface characteristics, and how they may have resulted in decreasing aspect ratios in this case.
It is interesting to note that where ELDORA was able to determine snowfall fields, the mesoscale structure remained in a disorganized (random/cellular) convective state. Except for increasing in convective size, there was little change in organization over the lake. Close to the upwind shore, the convective structure might be expected to be much more linear [as observed farther south by Mayor (2001)]. This may be due to increased shear in a subsidence region near the upwind shore (as proposed in Kristovich et al. 1999) or due to local variations in rates of surface heating. Near the downwind shore, satellite images suggest that the convective structure rapidly changed from disorganized to roll vortices. Further studies are needed to understand the reason for this change as well as the determination of factors controlling the rate at which the convective structure can adjust to the rapidly changing surface near the shore.
This research was supported by NSF Grants ATM 98-16306 and ATM 98-16203. Comments from Drs. H. Ochs, K. Kunkel, S. Hollinger, and three anonymous reviewers led to significant improvements in this manuscript. The Great Lakes Environmental Research Laboratory CoastWatch program provided lake surface temperatures.
Corresponding author address: Dr. David A. R. Kristovich, Illinois State Water Survey, 2204 Griffith Dr., Champaign, IL 61820-7495. Email: email@example.com