Introduction

Lake-effect (LE) bands often produce precipitation, typically snow, during the cold season (October-March) that can result in damage to infrastructure or injuries and fatalities due to accidents (e.g., Kunkel et al. 2002). Cold-season LE environments over the Great Lakes (GL) typically occur as polar air extends equatorward over relatively warm lake water (Fig. 1) and involve interactions of synoptic and mesoscale conditions with localized lake surface moisture and heat fluxes. Within the boundary layer, the lake acts to destabilize and moisten the overlying polar air mass, initiating LE cloud and snowband development (Niziol et al. 1995;Laird et al. 2017). The mesoscale conditions favorable for LE bands over individual GL have been studied extensively because of the substantial impact heavy snowfall from LE storms can have on localized areas, typically downstream of a lake. An important factor that has been identified for enhancement LE snowfall over an individual lake is the atmospheric environment upstream of the lake, especially when a coherent LE band from an upstream lake extends across an intervening land area, and continues development over the downstream lake (e.g., Mann et al. 2002;Villani et al. 2017). However, only a small number of studies have examined the occurrence and interaction of LE bands over two or more lakes during lake-aggregate disturbances (e.g., Sousounis and Mann 2000) and lake-to-lake (L2L) LE bands (e.g., Kristovich et al. 2018;Lang et al. 2018).

Further understanding of events with LE bands occurring over multiple lakes simultaneously, that may include L2L bands, allows for the identification and improved forecasting of locations most likely to be impacted by LE snowfall over larger regions than simply those isolated to the downstream area of an individual lake. Investigating the occurrence of L2L bands and the synoptic-scale patterns that are associated with them is especially important as significant enhancement of LE snowfall in the vicinity of a downstream lake has been shown to occur during these events when compared to single-lake LE bands (Lang et al. 2018). Sousounis and Mann (2000) showed that LE occurring over multiple lakes within a GL lake-aggregate circulation could enhance LE precipitation in some areas and diminish LE precipitation in other areas. Additionally, given that numerous areas can be impacted when all five lakes in the GL region have LE bands occurring on the same day, the identified synoptic-scale patterns could be used to aid seasonal forecasting of LE snowfall throughout the GL region and to examine past long-term variability of LE occurrence, topics that only a few studies have explored (e.g., Notaro et al. 2015;Suriano and Leathers 2017a,b).

Several previous studies have used the occurrences of specific synoptic patterns as an indirect method to examine the variability of LE by implementing approaches similar to the temporal synoptic index (TSI; Kalkstein and Corrigan 1986) rather than identifying observed LE bands. The TSI applies principal component analyses to measurements from a single surface weather station which are then clustered into groups of days with similar atmospheric conditions. The grouped days are then used to derive composites of synoptic patterns from an archived gridded dataset. This approach has been used in examining the variability of LE conditions in the vicinity of one or two of the GL (e.g., Ellis and Leathers 1996;Suriano and Leathers 2017a,b;Suriano 2019;Suriano and Wortman 2021;Ellis and Suriano 2022). Recently, Ellis and Suriano (2022) found that using TSI in tandem with the indirect spatial synoptic classification (SSC) approach (Kalkstein et al. 1996), which examines climatological surface station data to identify days with LE modification of an air mass, resulted in identifying fewer events with more confidence than either approach individually. Another indirect method has merged an approach similar to TSI with information about LE snowfall events reported in the National Oceanic and Atmospheric Administration (NOAA) Storm Event database to examine the composite synoptic patterns that supported LE conditions near Lakes Erie and Ontario (Wiley and Mercer 2020), as well as near Lakes Superior and Michigan (Wiley and Mercer 2021).

There are spatial limitations and uncertainty with these indirect methods as the occurrence of LE bands must be inferred from synoptic patterns developed using the TSI method, which is based on data from a single surface weather station and includes the subjective determination of atmospheric conditions favorable for the formation of LE bands (Ellis and Leathers 1996). Most importantly, past studies that used the TSI approach make no determination whether the mesoscale system (e.g., LE bands) actually occurred using observations during the identified time period (e.g., Ellis and Leathers 1996;Suriano and Leathers 2017a,b;Ellis and Suriano 2022). Other approaches have also utilized indirect methods to identify LE occurrences using networks of surface climatological stations to differentiate modification of an air mass that would be expected during a LE event (e.g., Ellis et al. 2021) or used a mixture of indirect and direct approaches to identify storm type and LE events with information from networks of surface snowfall observations, reanalysis datasets, daily weather maps, weather satellites, and radars (e.g., Hartnett 2021).

Several studies have described the most favorable synopticscale patterns for LE over a portion of the GL region (e.g., Wiggin 1950;Niziol 1987;Ellis and Leathers 1996;Liu and Moore 2004;Suriano and Leathers 2017a,b). For example, Niziol (1987) summarized the basic synoptic-scale pattern for prolonged LE over the eastern GL as including 1) a stationary or slow-moving low at 500 hPa in the vicinity of James Bay, Canada; 2) Arctic air moving over the GL associated with a surface cyclone that tracks eastward between the GL and James Bay, Canada; and 3) a southward extending surface trough across the GL allowing for favorable fetch over the eastern GL. Additionally, this pattern can often be influenced by a collective contribution from the surface heat and moisture fluxes provided by all five lakes in the GL, resulting in an enhancement of a surface cyclone or intensification of a trough over the region (e.g., Petterssen and Calabrese 1959). Studies have not typically examined collective lake disturbances (also referred to as lake-aggregate mesoscale disturbances) using observations because of their expansive horizontal and vertical structure (i.e., approximately 500-1000 km wide and 2-4 km deep). Rather, investigations have primarily used model simulations that compare atmospheric conditions when the GL are present, to alternate simulations when the GL are not included (e.g., Sousounis 1997). Most studies of lake-aggregate mesoscale disturbances have used either a theoretical modeling approach (e.g., Sousounis and Shirer 1992) or examined model simulations of an individual event (e.g., Sousounis et al. 2001;Mann et al. 2002). The only climatological study that examined lake-aggregate disturbances was performed by Weiss and Sousounis (1999) using a 10-yr archive of operational limited-area fine mesh model initializations and forecasts. From this study, it was found that on average 33 days each winter experienced a lake-aggregate disturbance most often associated with a defined surface trough over the GL region, with a negative sea level pressure perturbation of 3-4 hPa.

While there are difficulties with using large observational databases, such as those associated with weather satellites and radars, identifying past LE events directly based on observed coherent characteristics of their mesoscale circulation is a logical and necessary first step to systematically determining synoptic patterns associated with LE across the GL region. Extensive, multiyear spatiotemporal databases of LE events have been created using satellite imagery (e.g., Kristovich and Steve 1995;Rodriguez et al. 2007;Laird et al. 2017) and Doppler radar observations (e.g., Laird et al. 2009a,b;Alcott et al. 2012;Veals and Steenburgh 2015). The use of satellite imagery, for example, allows for extensive spatial coverage and identification of the mesoscale cloud structures. While creating a comprehensive archive of LE band occurrences using these measurement systems has challenges, their use provides the most direct method of confidently and effectively developing multiyear databases of LE bands that exhibit well-known structure in cloud and precipitation patterns resulting from mesoscale circulations within the boundary layer. One such challenge with using satellite imagery is the constraint to daytime visible imagery as the thermal properties of LE clouds are difficult to distinguish with infrared imagery because of their relatively shallow nature (e.g., Kristovich and Steve 1995). Additionally, using satellite imagery to identify the presence of lower-tropospheric mesoscale weather systems (e.g., LE bands) when widespread synoptic clouds are present over a region is often not possible. While radar observations allow for greater temporal coverage and focus on precipitation, they are limited spatially by location and range of individual radars, as well as the distribution of radars within an established network (e.g., Brown et al. 2007).

The current study utilizes a direct approach to identify the occurrences of observed LE cloud bands to establish a strong foundation for determining the synoptic patterns, as well as examining the climatological characteristics, of days with LE occurring on all five lakes within the GL region [hereafter referred to as five-lake days (5LDs)] for a time period spanning 25 cold seasons. This study focuses on 5LDs because of the expansive nature and impacts of the mesoscale systems throughout the entire GL region, the pronounced occurrence of L2L bands on 5LDs, and the likelihood that 5LDs have linkages to GL lake-aggregate mesoscale disturbances. Section 2 of the article describes the observed LE database and the synoptic pattern composite approach. Section 3 presents the climatological characteristics and the synoptic environments found for days having LE bands on all five GL. Last, a discussion relating current results to previous studies that have examined LE synoptic environments, L2L bands, and lakeaggregate mesoscale disturbances in the GL region is provided in section 4.

Data and methods

a. Database of LE occurrences

Geostationary Operational Environmental Satellite (GOES) visible imagery was used to expand an existing daily LE database created by Laird et al. (2017) to 25 cold seasons (October-March) spanning from 1997/98 through 2021/22. Analyses used GOES Imager data obtained from the NOAA Comprehensive Large Array-data Stewardship System (CLASS) website (www. avl.class.noaa.gov) for GOES-8 through GOES-16 satellites with a temporal resolution of approximately 15 min. Visible imagery was chosen during this time period since the spatial resolution of 1 km remained consistent across GOES satellites, which allowed for clear identification of mesoscale cloud patterns and LE cloud band structure connected to specific boundary layer circulations and dynamics. The analysis of the number and types of LE days for the temporal coverage from different GOES satellites within the LE database revealed no statistically significant changes between transitions in satellites (i.e., cold season of 2001/02, 2006/07, and 2016/17). Niziol et al. (1995) and Laird et al. (2017) are examples of studies that describe several mesoscale LE morphologies recorded over lakes within the GL region. These include windparallel bands (WPB; widespread LE that commonly forms when the wind has a short overlake fetch and has a cloud field comprised of numerous horizontal convective rolls), shoreparallel bands (SPB; often one or two dominant cloud bands that form when winds are parallel to the long axis of the lake or a land breeze initiates an overlake convergence zone), mesoscale vortices (MSV; isolated mesoscale cyclonic circulation of clouds in an overlake region or lake basin), unclear LE organization (UNCL; mesoscale cloud features geolocated in an overlake region, but not clearly WPB, SPB, or MSV), and L2L bands (comprised of a dominant mesoscale cloud band extending from one lake across an intervening land area to another lake).

Animations of satellite imagery and individual images on each cold-season day were inspected for the presence of LE clouds and whether LE cloud patterns contained WPB, SPB, MSV, UNCL, and L2L, as well as simultaneous or sequential organizations. The methodology used in this study to identify LE clouds was identical to that discussed in Laird et al. (2017) with criteria that included (i) the cloud pattern needed to be visible for at least 1 h, (ii) the clouds must have originated over the lake, and (iii) the cloud patterns needed to exhibit mesoscale features consistent with LE bands forced by boundary layer circulations linked to a stationary surface heat and moisture source (i.e., lake). If more than one LE morphology was present over an individual lake or if the LE morphology on an individual lake changed over the course of a day, it was documented. For example, a 5LD occurred on 7 January 2004 when Lake Superior had WPB and synoptic clouds, Lake Michigan had WPB, Lake Huron had WPB and SPB, Lake Erie had WPB, and Lake Ontario had SPB (Fig. 2a). A second example of a 5LD that had L2L cloud bands occurred on 10 February 2001 when Lake Superior had WPB with L2L to northern Lake Michigan, Lake Michigan had WPB, Lake Huron had WPB and SPB with L2L to Lake Ontario, Lake Erie had UNCL, and Lake Ontario had SPB (Fig. 2b). All LE morphologies on each lake were aggregated by day, resulting in a daily LE database for the GL region across 25 cold seasons. Although local atmospheric temperature, wind, and moisture measurements were not examined in the vicinity of each LE morphology identified, the mesoscale structure and evolution of the cloud patterns determined in this study from satellite imagery provided confidence that local atmospheric conditions, in addition to the related synoptic environments, led to LE band formation.

The restriction of using only daytime hours from visible satellite imagery may have resulted in an underestimate of the total number of LE days across the 25 cold seasons examined in this study if LE occurred during overnight time periods on any day. Few comprehensive databases of LE band occurrences in the GL region across multiple cold seasons exist, and only one study has examined event start time, end time, and duration. Veals and Steenburgh (2015) used weather radar to identify periods of LE bands over Lake Ontario during 13 cold seasons and found that the mean (median) duration of LE events was 19.5 (13.2) hours with no statistical tendency for winter LE events to initiate or occur more frequently at a preferred hour. Although this information is limited to Lake Ontario, their findings suggest that any underestimate of the total number of LE occurrences from using visible satellite imagery may be relatively small, since LE events, on average, will be present during some daytime hours.

Additionally, the presence of synoptic clouds was recorded over specific lakes within the GL region. When widespread synoptic clouds were present over an entire lake area, mesoscale LE cloud bands that may have been present in the lower troposphere underneath the synoptic clouds were not able to be identified. The frequency of this type of occurrence could not be quantified, although its magnitude is suspected to be relatively small given that LE bands typically occur after a cyclone with widespread synoptic cloud cover has passed through the GL region and a polar or Arctic air mass with minimal synoptic clouds has moved into the region (e.g., Niziol 1987). This is evident for both 5LD examples presented in Fig. 2. Last, the NOAA CLASS archive had a relatively small number of days without GOES imagery available; on average 4.5 days each cold season had missing visible imagery which equates to approximately 2.5% of the days across the 25 cold seasons.

b. Composites of synoptic environments

To examine the synoptic characteristics of 5LDs, archived meteorological data associated with each LE day were acquired from the hourly ERA5 global reanalysis (hereafter ERA5) which has a 0.258 3 0.258 spatial resolution (Hersbach et al. 2018a(Hersbach et al. ,b, 2020)). ERA5 has been used in previous studies to identify synoptic-scale environments associated with precipitation events, such as extreme winter precipitation regimes over the eastern United States (Low et al. 2022) and extreme precipitation in the GL region (Paxton et al. 2021). To develop composites for environments associated with 5LDs, the mean and standard deviation of meteorological fields were computed using the 1800 UTC analysis on 5LDs within the 25 cold seasons for each grid point located across the GL region (defined here as 938-738W, 388-58.58N). The 1800 UTC analysis time was chosen to represent a mid-daylight hour time within the cold season consistent with the methodology of using daytime visible GOES satellite imagery to identify LE cloud bands each day. Geopotential height (hereafter height), temperature, specific humidity, and winds at both 500 and 850 hPa, as well as the 10-m winds, 2-m temperature, and sea level pressure (SLP), were examined. Additionally, standardized anomalies were created for all parameters by subtracting the long-term mean (average across all days within the 1997/98-2021/22 cold seasons) from the mean across 5LDs which was then divided by the long-term standard deviation (Grumm and Hart 2001).

Results

a. LE climatology of five-lake days

The 25 cold-season (i.e., 150-month) LE database allowed for an extensive spatiotemporal examination of LE cloud band occurrences across the entire GL region, specifically providing details of the seasonality, the lake(s) that had LE on a particular day, and the LE morphology. LE occurred on 2870 days (64%) during the 25 cold seasons and was observed over one, two, three, four, or five lakes on 772 (26.9%), 614 (21.4%), 569 (19.8%), 423 (14.7%), and 492 (17.1%) days, respectively. Numerous past studies have examined LE bands that have occurred over a specific GL or over individual lakes within part of the GL region, such as the Lakes Ontario and Erie region (e.g., Dewey 1977;Niziol 1987;Wiley and Mercer 2020;Ellis et al. 2021) or Lakes Superior and Michigan region (e.g., Kristovich and Spinar 2005; Wiley and Mercer 2021).

The unique focus of the current study is on 5LDs, which occurred when atmospheric conditions were favorable over the entire GL region which allowed LE bands to develop on each lake at the same time. Nearly a third (31.3%) of all LE bands that were observed across 25 cold seasons occurred during 5LDs even though 5LDs consisted of just 17.1% (492) of LE days identified. Additionally, L2L was present during 56.3% of all 5LDs and L2L on 5LDs comprised 43.5% of all L2L occurrences during the 25 cold seasons.

The number of 5LDs ranged from seven occurrences in a cold season (2014/15 and 2020/21) to 36 occurrences (2005/06) with a mean of 19.7 (SD 5 8.5) days per cold season (Fig. 3). 5LDs occurred with a mean of 26.1 (SD 5 6.2) days per cold season before the 2009/10 season and then decreased to a mean of 13.8 (SD 5 5.5) days across the subsequent cold seasons, signifying a reduction in the frequency of favorable conditions for 5LDs in recent years. There was an absence or substantial reduction in the number of 5LD occurrences in October, November, and March beginning in the 2009/10 cold season (Fig. 3), which confined 5LDs toward the months of December, January, and February in the more recent cold seasons. In general, the largest number of 5LDs occurred in December, January, and February (collectively 69.5% of 5LDs), with the most 5LD in January (29.5%) and the fewest in October (8.9%; Fig. 4). This is consistent with the monthly variation of total LE days and with the monthly occurrence of L2L days, both having the largest percentage of occurrence during January (Fig. 4).

WPB occurred most often on 5LDs either as the only LE morphology (56.7%) over a lake or with other LE morphologies, such as WPB and SPB on 15.2% of 5LDs (Fig. 5). Examples of WPB as the only LE morphology on a lake are shown in Figs. 2a and 2b for Lakes Superior and Michigan. Although Laird et al. (2017) found that WPB occurred most frequently over Lakes Superior and Michigan during LE days, the occurrence of WPB as the only LE morphology on any individual lake during 5LDs was observed with the same percentage over each lake in the GL region. SPB occurred less frequently as the only LE morphology over an individual lake on 5LDs with 9.1% of LE bands identified. The occurrence of SPB as the only LE morphology on an individual lake during 5LDs was observed most often over Lake Ontario and Lake Erie with 42.1% and 19.7%, respectively, of the 223 occurrences of SPB as the only LE morphology on a lake. Percent of Cold-Season Days Non-LE LE L2L 5LD 14 12 10 8 6 4 2 0 OCT NOV DEC JAN FEB MAR Month FIG. 4. Percentage of LE days, non-LE days, L2L days, and 5LDs each month of all days across the 25 cold seasons.

On average, a 5LD with L2L (278 days) had one or two L2L bands in the GL region. The largest number of L2L occurrences on 5LDs happened between Lakes Huron and Erie, Lakes Superior and Michigan, and Lakes Huron and Ontario (Fig. 6a). Similar to findings from Laird et al. (2017), the aforementioned L2L locations also had the largest number of occurrences of L2L across all LE days, not just on 5LDs. Most L2L locations experienced a maximum in occurrences on 5LDs during either December or January (Fig. 6b), the two most active months for L2L on both 5LDs and all LE days. The annual mean number of L2L occurrences on 5LDs during the months of December and January was 8.4 (SD 5 6.5) and 9.4 (SD 5 4.4), respectively (Fig. 6b).

An analysis of LE occurrences on the day before and day after each LE day during the 25 cold seasons was completed to understand the duration of LE events over the GL (i.e., with events defined as consecutive days with LE bands observed over at least one of the GL). This is an important consideration for relating the timing and significance of impacts to LE snowfall, especially for 5LDs which have an influence on several areas in the GL, since polar and arctic air masses progress through the region with a range of speeds and pathways (e.g., Shadbolt et al. 2006). The atmospheric conditions supporting observed LE over one or more lakes were found to continue from 2 to 10 consecutive days for the majority (62.5%) of LE events (Fig. 7). 5LDs were often found to occur consecutively with 30.7% of 5LDs followed by another 5LD. On average across 25 cold seasons, LE was present on one or more lakes for a mean of 4.0 consecutive days in the GL region. A single isolated LE day (no LE day occurring before or after) occurred for 29.8% of LE days and only 6.9% of LE days occurred as part of events with greater than 10 consecutive LE days. This suggests that the majority of cold air masses that initiate LE in the GL region are widespread and typically overtake the entire region rather than being restricted to a portion of the GL region. LE events with the largest number of consecutive LE days in the GL region occurred during January and February with mean (median) values of 6.0 (4) days and 5.7 (4) days, respectively (Fig. 7). As the number of consecutive LE days increase, both the number of 5LDs and consecutive 5LD increase. The longest event within the 25 cold seasons when LE was ongoing in the GL region spanned a 32-day time period in January and February 2007 during which an extreme coldair outbreak occurred in the eastern United States (Smith and Sheridan 2018). During this event, 16 days of the 32-day time period were 5LDs with eight consecutive 5LD having occurred from 3 to 10 February 2007.

S-H S-M M-E M-H H-E H-O E-O

b. Synoptic-scale patterns and environments of fivelake days

Surface, 850-and 500-hPa composite analyses were examined for the day before, day of, and day after a 5LD using the 1800 UTC analysis time on each day in order to understand the temporal evolution and characteristic patterns of the synoptic environment.

1) SURFACE COMPOSITES

On the day before a 5LD, the mean SLP pattern featured an area of lower pressure over the eastern GL with a trough extending westward across the GL and an area of higher pressure across the western GL (Fig. 8a). Mean 2-m temperatures were lower than 08C over the entire GL region with cold-air advection ongoing over Lakes Superior and Michigan (Fig. 8b). On the day of a 5LD, a composite 1028-hPa high pressure system and large positive SLP anomaly were centered southwest of the GL region with an area of lower pressure outside of the GL region to the north and east (Fig. 8c). Relatively slow mean 10-m wind speeds and weak cold-air advection existed throughout the GL region with anomalously cold surface temperatures ranging from 248 to 2108C (Fig. 8d). On the day after a 5LD, weaker mean surface high pressure was positioned south of the region (Fig. 8e) and mean 2-m temperature remained lower than 08C with weak west-southwesterly and westerly winds across the region (Fig. 8f). With cold air in place and varying winds over the GL regions in the synoptic patterns on the day before, day of, and day after a 5LD, allowed conditions in the lower troposphere to remain favorable for LE over multiple lakes, consistent with the result that LE often occurs across multiple sequential days in the GL region (Fig. 7).

2) 850-HPA COMPOSITES On the day before a 5LD, composite 850-hPa heights revealed a positively tilted trough with an axis positioned over the western GL region with a large negative height anomaly (Fig. 9a). The mean wind direction at 850 hPa was northnorthwesterly with cold-air advection over Lakes Superior, Michigan, and Huron and westerly winds over Lakes Erie and Ontario (Fig. 9c). Mean 850-hPa height contours showed ridging to the west of the GL which transitioned to weak troughing to the east on a 5LD (Fig. 9d). This pattern led to anomalously low 850-hPa heights over the eastern lakes and anomalously high 850-hPa heights over the western lakes (Fig. 9d). Mean and anomaly specific humidities and temperatures at 850 hPa showed that there was a drier, colder air mass present on 5LDs across the entire GL region (Figs. 9e,f) compared both to long-term mean and conditions on the day before. On a 5LD, mean wind directions at 850 hPa were predominantly northwesterly with cold-air advection spread across the entire GL region (Fig. 9f). The presence of an air mass with these attributes would provide an environment with greater upward surface heat and moisture fluxes over lake surfaces, reduction of boundary layer stability within the modified continental air mass, and generally stronger LE convection. Relatively strong northwesterly 850-hPa winds with a cold, dry air mass in place across the GL region would likely provide an environment for WPB over Lakes Superior, Huron, and Michigan while setting up either WPB or SPB over Lakes Erie and Ontario. The direction of the 850-hPa winds would also provide favorable air parcel trajectories across the GL region supportive of L2L connections, especially for the most prominent L2L pathways from Lake Superior to northern Lake Michigan, Lake Huron to eastern Lake Erie, and Lake Huron to Lake Ontario. On the day after a 5LD, mean 850-hPa winds transitioned to westerly as the continental polar air mass exited the GL region resulting in warm-air advection over each of the GL (Fig. 9i).

3) 500-HPA COMPOSITES On the day before a 5LD, composite 500-hPa heights showed anomalously low heights within a broad trough and colder than average temperatures with an area of cold advection over the western GL (Figs. 10a,c). On 5LDs, the trough deepened to near 21.0 standardized anomalies while positioned over and just east of Lake Ontario (Fig. 10d). This pattern has strong northwesterly 500-hPa winds and cold advection spreading a drier and colder-than-average air mass throughout the entire GL region (Figs. 10e,f). This also demonstrates that a deep polar or Arctic air mass is present in these situations with robust cold advection occurring within the lower troposphere below 500 hPa. While the composite analysis showed a broad trough in place, LE occurring in a portion of the GL region has been found to be responsive with changes in snowfall intensity and LE snowband positioning when short-wave troughs propagate through this type of broad long-wave trough pattern (e.g., Metz et al. 2019). On the day following a 5LD, the winds slowed slightly and became more westerly with warmer temperatures having moved into the GL region (Fig. 10i).

Discussion and conclusions

The current study identified LE on 2870 days (64% of all cold-season days) with 492 5LDs during 25 cold seasons. Greater than half of all cold-season LE days in the GL region (i.e., 51.6%) had LE over at least three lakes and 31.3% of all LE bands identified occurred during 5LDs. This signifies that favorable atmospheric conditions for LE were typically widespread across much of the GL region even though most past studies of LE have focused on the mesoscale and local aspects of snowstorms associated with a single lake. The occurrence of LE over all or a majority of lakes during a day is likely linked to the expansive dimensions and characteristics of the winter air masses that move over the GL region (e.g., Brunnschweiler 1952;Hankes and Walsh 2011;Hartig et al. 2023). Days with multiple lakes having LE bands, often combined with the presence of L2L and consecutive days with LE, create challenging forecasts of location, duration, and amount of snowfall.

Mesoscale factors that greatly influence LE, such as gradient wind speed and direction, vertical wind shear, temperature advection, and atmospheric stability, are directly linked to the specific synoptic pattern (e.g., Niziol 1987). Previous studies have identified synoptic patterns conducive for the development of LE in the vicinity of one or two specific lakes (e.g., Lakes Erie and Ontario). Given the limits of the data and methods used by these previous studies to identify LE occurrences over the entire GL region, it is unclear whether LE was simultaneously occurring in additional areas of the GL region that were not a focus of their studies (i.e., on other lakes). A clear distinction of the current study is that the daily occurrence of LE bands on each lake in all areas of the GL region across the 25 cold seasons was examined when creating the extensive LE database. Therefore, the current study was distinctive in the ability to confidently examine days with LE bands occurring over all five lakes (i.e., 5LDs). No previous study, to the authors' knowledge, has examined the synoptic pattern associated with 5LDs.

On 5LDs, the lakes were warming and moistening the overlying air acting to collectively contribute to a lake-aggregate disturbance across the GL region (e.g., Petterssen and Calabrese 1959). Given this, the results of the current study were compared to the results of Weiss and Sousounis (1999) that presented the only climatological study that has been conducted examining the frequency, magnitude, and synoptic patterns associated with lakeaggregate disturbances in the GL region. Weiss and Sousounis (1999) defined the existence of a lake-aggregate disturbance as the presence of a synoptic-scale negative pressure perturbation rather than the observance of LE bands over the GL region. In the current study, SLP on the day before a 5LD (Fig. 8a) had a similar pattern to the mean SLP composite for all lake-aggregate cases presented by Weiss and Sousounis (1999) where a pronounced pressure trough extended westward across the GL region (Fig. 11b). On 5LDs, the SLP pattern (Fig. 11a) was similar to the mean SLP composite for weak (less than 3-hPa negative anomaly) lake-aggregate disturbances found by Weiss and Sousounis (1999) (Fig. 11d) with a high pressure system centered southwest of the GL region and weak cold-air advection throughout the GL region (Figs. 8c,d). The mean composite synoptic pattern for 5LDs also had a low pressure system to the north and east of the GL region which differed from the synoptic pattern found by Weiss and Sousounis (1999) for strong (greater than 9-hPa negative anomaly) lake-aggregate disturbances (Fig. 11f) that had a low pressure system and strong cold advection within the GL region. An inspection of SLP patterns on the 492 individual 5LDs found that a closed low pressure system existed over the GL region (with varying magnitudes) on 49 5LDs (10%).

Several past studies using the TSI approach have identified synoptic SLP patterns that would be favorable for LE in a portion of the GL region. For Lakes Ontario and Erie, 1) Ellis and Leathers (1996) determined five synoptic LE patterns; 2) Suriano and Leathers (2017a) found seven synoptic LE patterns that were subsequently used in studies by Suriano and Leathers (2017b), Suriano (2019), and Suriano and Wortman (2021); 3) Ellis and Suriano (2022) periods; and 4) Wiley and Mercer (2020) found three synoptic LE patterns. Additionally, Wiley and Mercer (2021) determined four synoptic LE patterns for Lake Michigan and three synoptic LE patterns for Lake Superior. Although each study identified a variety of synoptic patterns linked to LE conditions in the vicinity of one or two lakes, at least one synoptic pattern from each study had a strong similarity to the mean SLP composite for 5LDs. These included W-S type (Fig. 11e; Ellis and Leathers 1996), WNW-1 (Fig. 11c; Suriano and Leathers 2017a,b; Suriano 2019; Suriano and Wortman 2021), early season and midseason (Figs. 11g,h; Ellis and Suriano 2022), Cluster 2 (Fig. 11i; Wiley and Mercer 2020), and Cluster 1 (Fig. 11j; Wiley and Mercer 2021). While these studies investigated LE in the vicinity of a local area of the GL (i.e., Lake Ontario, Lake Erie, Lake Superior, and Lake Michigan), similarities between the mean SLP composite for 5LDs in the current research and SLP patterns in previous research suggest that LE conditions may have been more widespread throughout the GL region beyond the smaller area investigated by each previous study.

The primary conclusions from the current research are as follows:

• Nearly a third of all LE bands that were observed across 25 cold seasons (i.e., 150 months) occurred during 5LDs although 5LDs consisted of just 17.1% of LE days identified. • 5LDs occurred with a mean of 26.2 (SD 5 6.2) days per cold season until 2008/09 and then decreased to a mean of 13.8 (SD 5 5.5) days across the subsequent cold seasons, signifying a reduction in the frequency of favorable conditions for 5LDs in recent years. • A majority of 5LDs (56.5%) were observed to have at least one or two L2L connections making up 43.5% of all L2L occurrences identified across the 25 cold seasons. This demonstrates the complexity and frequency of mesoscale and boundary layer interactions occurring across the GL region as L2L bands extend from an upstream lake, across an intervening land area, and continue development over a downstream lake. L2L bands can also result in an enhancement of LE snowstorms which could impact multiple, localized regions of the GL region with heavy snowfall.

• On average, across 25 cold seasons, LE bands were present on one or more lakes for a mean of 4.0 consecutive days in the GL region. 5LDs were often found to occur consecutively with 30.7% of 5LDs followed by another 5LD. As the number of consecutive LE days increase, both the number of 5LDs and consecutive 5LD increase. This demonstrates that LE synoptic patterns may progress slowly through the GL region and/or have subtle variations on the mean composite pattern that continue to support ongoing LE bands over several lakes with varying LE morphology.

• The mean SLP composite pattern for the day of a 5LD was found to have strong similarity to one of the composite SLP patterns found in several previous studies of LE on individual lakes in the GL region (e.g., Ellis and Leathers 1996). While previous studies have identified SLP patterns favorable for LE over one or two lakes, the current results demonstrate that several additional areas of the GL often experience LE bands when a localized region has active LE bands occurring. Therefore, several of the SLP patterns previously identified for LE occurrences on individual lakes provide favorable LE conditions in geographic areas outside the focus of the previous studies.

• The mean SLP composite pattern for the day before and day of a 5LD exhibited a strong similarity to the composite SLP pattern for lake-aggregate disturbances in the GL region found by Weiss and Sousounis (1999). This result would be expected given that each lake was observed to have LE bands present during a 5LD and therefore have conditions of upward directed heat and moisture fluxes in place, contributing to a collective reduction in SLP across the GL region.

Future work will include an analysis of the variability of mesoscale environments in the vicinity of each GL on 5LDs and multiday LE events that include a 5LD, especially since 30.7% of 5LDs were followed by another 5LD. With LE events typically occurring with a duration of 3-5 days, these multiday 5LD events have the potential for large societal impacts due to the longevity of LE snowfall throughout the GL region and therefore require further study. Additionally, the notable decrease in the number of 5LDs per cold season from the first to the second portion of the 25 cold-season time period may provide some added insights toward temporal trends of LE activity in the GL region that previous studies have examined (e.g., Burnett et al. 2003;Bard and Kristovich 2012;Hartnett et al. 2014;Suriano and Leathers 2017b), especially considering the absence, or substantial reduction, in the number of 5LDs in the months of October, November, and March since the 2009/10 cold season (Fig. 3). Last, we hope to use the 25 cold-season database of observed LE cloud bands to conduct a collaborative study examining the ability of the TSI approach to identify LE synoptic-scale patterns favorable for LE over individual lakes or collections of lakes. This is needed since past studies that used the TSI approach have made no determination of whether LE bands actually occurred using observations during the identified time periods. FIG. 11. Surface SLP composites of (a) 5LD in the current study, (b) all lake-aggregate disturbances in Weiss and Sousounis (1999), (c) west-southwest-type 1 (WSW-1) for eastern GL in Suriano and Leathers (2017a), (d) weak lake-aggregate disturbances in Weiss and Sousounis (1999), (e) W-S type for eastern GL in Ellis and Leathers (1996), (f) strong lake-aggregate disturbances in Weiss and Sousounis (1999), (g) early season for eastern GL in Ellis and Suriano (2022), (h) midseason for eastern GL in Ellis and Suriano (2022), (i) Cluster 2 for eastern GL in Wiley and Mercer (2020), and (j) Cluster 1 for western GL in Wiley and Mercer (2021) with SLP color shading and 500-hPa heights solid contours. Inset boxes in (b)-(j) correspond to the approximate region shown in (a). website (www.avl.class.noaa.gov). The 25 cold-season LE database is available through the Hobart and William Smith Colleges Geoscience Department upon request. Additionally, the meteorological data from the ERA5 climate reanalysis used for composites in this study are publicly available from the Copernicus Climate Data Store at https://cds.climate.copernicus.eu/ cdsapp#!/home.