• Ashley, W. S., , Mote T. L. , , Dixon P. G. , , Trotter S. L. , , Powell E. J. , , Durkee J. D. , , and Grundstein A. J. , 2003: Distribution of mesoscale convective complex rainfall in the United States. Mon. Wea. Rev., 131, 30033017.

    • Search Google Scholar
    • Export Citation
  • Augustine, J. A., , Woodley W. L. , , Scott R. W. , , and Changnon S. A. , 1994: Using geosynchronous satellite imagery to estimate summer-season rainfall over the Great Lakes. J. Great Lakes Res., 20, 683700.

    • Search Google Scholar
    • Export Citation
  • Bryan, G. H., , Knievel J. C. , , and Parker M. D. , 2006: A multimodel assessment of RKW theory’s relevance to squall-line characteristics. Mon. Wea. Rev., 134, 27722792.

    • Search Google Scholar
    • Export Citation
  • Fowle, M. A., , and Roebber P. J. , 2003: Short-range (0–48 h) numerical prediction of convective occurrence, mode, and location. Wea. Forecasting, 18, 782794.

    • Search Google Scholar
    • Export Citation
  • Gallus, W. A., Jr., , Snook N. A. , , and Johnson E. V. , 2008: Spring and summer severe weather reports over the Midwest as a function of convective mode: A preliminary study. Wea. Forecasting, 23, 101113.

    • Search Google Scholar
    • Export Citation
  • Graham, R., , Bentley M. , , Sparks J. , , Dukesherer B. , , and Evans J. , 2004: Lower Michigan MCS climatology: Trends, pattern types, and marine layer impacts. Preprints, 22nd Conf. on Severe Local Storms, Hyannis, MA, Amer. Meteor. Soc., 7B.6. [Available online at http://ams.confex.com/ams/pdfpapers/81343.pdf.]

  • Laing, A. G., , and Fritsch J. M. , 1997: The global population of mesoscale convective complexes. Quart. J. Roy. Meteor. Soc., 123, 389405.

    • Search Google Scholar
    • Export Citation
  • Lericos, T. P., , Fuelberg H. E. , , Weisman M. L. , , and Watson A. I. , 2007: Numerical simulations of the effects of coastlines on the evolution of strong, long-lived squall lines. Mon. Wea. Rev., 135, 17101731.

    • Search Google Scholar
    • Export Citation
  • Parker, M. D., 2008: Response of simulated squall lines to low-level cooling. J. Atmos. Sci., 65, 13231341.

  • Rotunno, R., , Klemp J. B. , , and Weisman M. L. , 1988: A theory for strong, long-lived squall lines. J. Atmos. Sci., 45, 463485.

  • Scott, R. W., , and Huff F. A. , 1996: Impacts of the Great Lakes on regional climate conditions. J. Great Lakes Res., 22, 845863.

  • Weisman, M. L., , and Rotunno R. , 2004: “A theory for strong, long-lived squall lines” revisited. J. Atmos. Sci., 61, 361382.

  • Weisman, M. L., , Klemp J. B. , , and Rotunno R. , 1988: Structure and evolution of numerically simulated squall lines. J. Atmos. Sci., 45, 19902013.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Locations of observation sites used in this study. Surface observation sites over land are denoted by black diamonds, precipitation observation sites denoted by numbered circles, buoys denoted by black ovals, radar locations denoted by , and raob sounding sites by . Radar ranges for KCLE and Detroit, MI (KDTX), are denoted by a dashed circle (50 km) and a solid circle (100 km), respectively. Dashed line denotes 42.25°N, the northern boundary of the study. Names and locations of precipitation observation sites are located in appendix A.

  • View in gallery

    Base reflectivity images from the Cleveland WSR-88D radar showing examples of (a) isolated, (b) cluster, (c) complex, and (d) linear storm modes used in this study.

  • View in gallery

    Number of occurrences (111 total) of each storm mode by month.

  • View in gallery

    Same as Fig. 3 except by time of day (using TMOW). Time of day shown in local standard time (LST = UTC − 4/5 h).

  • View in gallery

    Flowcharts for (a) cluster and isolated, and (b) linear and complex storms. Numbers given in parentheses are the number of observed storm occurrences for each given parameter combination. The numbers below each parameter give the percentage of events that exhibited, after 60 min over the water, no change or moderate strengthening/major weakening. Percentages are not shown for conditions that had less than five storm occurrences.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 6 6 2
PDF Downloads 7 7 0

Influence of the Lake Erie Overlake Boundary Layer on Deep Convective Storm Evolution

View More View Less
  • 1 Department of Atmospheric Sciences, and Center for Atmospheric Science, Illinois State Water Survey, Prairie Research Institute, University of Illinois at Urbana–Champaign, Urbana, Illinois
  • 2 Center for Atmospheric Science, Illinois State Water Survey, Prairie Research Institute, University of Illinois at Urbana–Champaign, Urbana, Illinois
  • 3 Hobart and William Smith Colleges, Geneva, New York
  • 4 NOAA/National Weather Service Forecast Office, Cleveland, Ohio
  • 5 NOAA/National Weather Service Forecast Office, Phoenix, Arizona
© Get Permissions
Full access

Abstract

The influence that the overlake boundary layer has on storm intensity and structure is not well understood. To improve scientists’ understanding of the evolution of storms crossing Lake Erie, 111 events during 2001–09 were examined using observations from Weather Surveillance Radar-1988 Doppler (WSR-88D), surface, buoy, and rawinsonde sites. It was found that on average, all storm modes tended to weaken over the lake; however, considerable variability in changes of storm intensity existed, with some storms exhibiting steady-state or increasing intensity in specific environments. Noteworthy changes in the storm maximum reflectivity generally occurred within 60 min after storms crossed the upwind shoreline. Isolated and cluster storm modes exhibited much greater weakening than those storms organized into lines or convective complexes. The atmospheric parameters having the greatest influence on storm intensity over Lake Erie varied by mode. Isolated and cluster storms generally weakened more rapidly with increasingly cold overlake surface air temperatures. Linear and complex systems, on the other hand, tended to exhibit constant or increasing maximum reflectivity with cooler overlake surface air temperatures. It is suggested that strongly stable conditions near the lake surface limit the amount of boundary layer air ingested into storms in these cases.

Current affiliation: Systems Research Group, Inc., NOAA/Hydrometeorological Prediction Center, Camp Springs, Maryland.

Corresponding author address: David Kristovich, ISWS, PRI, University of Illinois at Urbana–Champaign, 2204 Griffith Dr., Champaign, IL 61820. E-mail: dkristo@illinois.edu

Abstract

The influence that the overlake boundary layer has on storm intensity and structure is not well understood. To improve scientists’ understanding of the evolution of storms crossing Lake Erie, 111 events during 2001–09 were examined using observations from Weather Surveillance Radar-1988 Doppler (WSR-88D), surface, buoy, and rawinsonde sites. It was found that on average, all storm modes tended to weaken over the lake; however, considerable variability in changes of storm intensity existed, with some storms exhibiting steady-state or increasing intensity in specific environments. Noteworthy changes in the storm maximum reflectivity generally occurred within 60 min after storms crossed the upwind shoreline. Isolated and cluster storm modes exhibited much greater weakening than those storms organized into lines or convective complexes. The atmospheric parameters having the greatest influence on storm intensity over Lake Erie varied by mode. Isolated and cluster storms generally weakened more rapidly with increasingly cold overlake surface air temperatures. Linear and complex systems, on the other hand, tended to exhibit constant or increasing maximum reflectivity with cooler overlake surface air temperatures. It is suggested that strongly stable conditions near the lake surface limit the amount of boundary layer air ingested into storms in these cases.

Current affiliation: Systems Research Group, Inc., NOAA/Hydrometeorological Prediction Center, Camp Springs, Maryland.

Corresponding author address: David Kristovich, ISWS, PRI, University of Illinois at Urbana–Champaign, 2204 Griffith Dr., Champaign, IL 61820. E-mail: dkristo@illinois.edu

1. Introduction and background

It is well known that on climatic time scales the Great Lakes suppress clouds and precipitation near the coastlines during the warm season, particularly in regions downwind of the lakes (e.g., Augustine et al. 1994; Scott and Huff 1996). However, operational weather forecasts for these regions must account for the rate of change of storm intensity as well as situations where the storm evolution differs from climatology. Few published studies have examined how the lakes and their associated overlake boundary layers (OLBLs)1 affect preexisting deep convective storms, particularly during the warm season where the water surface is predominately cooler than the surrounding land.

To develop better forecasting techniques for areas in western lower Michigan, Graham et al. (2004) constructed a 5-yr climatology of the behavior of severe mesoscale convective systems (MCSs) crossing Lake Michigan. They found that MCSs responded variably to passing over the water surface, depending on time of year and structure. For example, MCSs were more likely to weaken or dissipate in the summer months (June–August, JJA) than in the spring or fall. In addition, the stability of the OLBL appeared to influence storm evolution over the lake. The study showed that the convection tended to maintain intensity when the OLBL was strongly stable (lake surface ≥2.5°C colder than the overlake air temperature), despite the reduction in surface-based instability. However, their study did not examine the rate at which storm intensity changes occurred.

Other studies have used numerical modeling techniques to examine the effect of surface-based stable layers on convective systems. Parker (2008) used an idealized model to simulate squall-line interactions with such layers. In his study, a surface-based isothermal cool layer, approximately 1 km deep, was introduced to a mature squall line. He found that storm evolution is associated with the relative buoyancies of the cool layer and the cold pool produced by the convective system. When the cold pool is denser, lifting along the leading edge of the cold pool can overcome a significant amount of surface-based convective inhibition (CIN) if convective available potential energy (CAPE) is present. This, in turn, allows updrafts to maintain their intensity despite interaction with convectively unfavorable near-surface conditions. The results of Parker (2008) also showed that in instances where the surface cool layer ahead of the storm is cooler than the cold pool air, storms can become elevated atop the surface layer, with continuous updrafts being driven by bore propagation.

The effect of changing low-level conditions near shorelines on mature squall lines was examined in an idealized modeling study by Lericos et al. (2007). Their study focused on squall lines moving eastward from the Gulf of Mexico onto the northern Florida coast. In cases of moderate to strong ambient low-level vertical wind shear, land breezes were found to reduce the magnitude of ambient shear. This, in turn, caused the updraft of the squall line to tilt in an upshear direction, resulting in weakening of the squall line.

While the Graham et al. (2004), Parker (2008), and Lericos et al. (2007) studies provide important insights into the evolution of convection interacting with varying near-surface conditions in the vicinity of coastlines, they consider only organized convective systems (e.g., MCSs), leaving the potential effect of a large lake on other convective modes unknown. In addition, the rate of change in the storm system, critical in an area of widely varying lake sizes, was not examined. The current study uses a climatological approach incorporating observations commonly available to forecasters (e.g., radar, hourly surface observations, rawinsonde observations) to examine how large lakes influence the intensity of the wide range of convective modes.

2. Data and methodology

Statistical analyses were performed by season, time of day, and storm mode to identify common storm responses and the potential causes. Weather Surveillance Radar-1988 Doppler (WSR-88D) data from Cleveland, Ohio (KCLE) (Fig. 1), were used for this study. The area of study chosen was south of 42.25°N over Lake Erie; this was selected to include storms within a range of ~100 km of the radar when crossing the lake. This is an area where convective storms frequently pass over Lake Erie (Laing and Fritsch 1997) and are routinely monitored by the Cleveland National Weather Service Forecast Office.

Fig. 1.
Fig. 1.

Locations of observation sites used in this study. Surface observation sites over land are denoted by black diamonds, precipitation observation sites denoted by numbered circles, buoys denoted by black ovals, radar locations denoted by , and raob sounding sites by . Radar ranges for KCLE and Detroit, MI (KDTX), are denoted by a dashed circle (50 km) and a solid circle (100 km), respectively. Dashed line denotes 42.25°N, the northern boundary of the study. Names and locations of precipitation observation sites are located in appendix A.

Citation: Weather and Forecasting 27, 5; 10.1175/WAF-D-11-00076.1

a. Identification of events

Appropriate events (preexisting convective storms that advected from land to over Lake Erie) were identified by a three-step process: 1) surface observations were used to identify dates when significant rain occurred, 2) composite radar reflectivity imagery was used to remove events where the observed precipitation was clearly not associated with convective storms moving over Lake Erie, and 3) more detailed analyses of WSR-88D Level II data were conducted to identify which of the remaining events were appropriate for this study.

Precipitation observations at sites around Lake Erie (Fig. 1, appendix A) were used to identify potentially appropriate events. The azimuth range of 120°–340° with respect to the location of KCLE was chosen to capture the most likely path of storms crossing the lake while staying inside the study area, despite the lack of suitable precipitation observation stations on the southern and eastern shores of Lake Erie. Days when snow occurred or surface air temperatures were <0°C were excluded. Over the period 2001–09, a total of 504 dates were identified with ≥1.27 cm (0.5 in.) of precipitation and above freezing temperatures.

National composite radar reflectivity imagery obtained from the National Center for Atmospheric Research (NCAR) was examined for each of the 504 dates to identify those with convective storms developing upwind and moving over Lake Erie. While the national composite lacks the desired spatial and temporal resolutions, it allows for identification of precipitation systems that are clearly not convective. Any date that had a maximum base reflectivity <35 dBZ on the national composite was eliminated. These criteria reduced the list of dates with possible convective storms moving over the Lake Erie study region to 280. It should be noted that an individual date may have included multiple convective storm events.

More detailed analysis of Level II radar data from KCLE was conducted for each of the remaining 280 potential events using GR2Analyst (Gibson Ridge, version 1.60). These data were obtained from the National Climatic Data Center (NCDC). In order for the event to qualify for this study, the base reflectivity observed at the 0.5° elevation level with values >45 dBZ, a common threshold for convective rainfall, needed to be observed during at least three consecutive radar scans (≈10–15 min) prior to moving over the lake. If a reflectivity value of ≥45 dBZ was not observed at least 15 min before the storm crossed the shoreline, the event was not included in the study. This was done to ensure storms included in the study were not experiencing large changes in intensity as they approached the lake; therefore, the evolution of intensity observed over the water was likely the result of OLBL interaction. Note that this criterion would remove cases where the reflectivity increased to >45 dBZ over the lake; however, this was an infrequent occurrence and is not thought to change the findings of this study.

In some instances, areas of >45-dBZ reflectivity were embedded in areas of widespread rainfall associated with synoptic disturbances. Since the presence and organization of convective storms is not always obvious in such situations, these events were also eliminated from the database. Finally, events when the convective storm moved north of the 42.25°N boundary were removed. This resulted in a total of 111 events (on 101 days) used for analysis of storm evolution over Lake Erie.

b. Determination of convective mode

Each storm was classified by mode to investigate its role in convective evolution over the lake. During events when the storm mode underwent a change over the lake, the storm was classified by the mode at the time it moved over the upwind shoreline. Previous studies (e.g., Fowle and Roebber 2003; Gallus et al. 2008) tended to use three dominate convective modes (linear, isolated, multicellular); a fourth mode (complex) was added for this study due to the propensity for nonlinear mesoscale convective complex (MCC) development in the study area, as well as to allow for a more direct comparison with the findings from Graham et al. (2004).

Examples of the four convective storm modes used in this study are shown in Fig. 2 and were defined as follows:

  • Isolated (I)—individual convective storms, including supercells, with a continuous area of reflectivity >45 dBZ that was <500 km2 (Fowle and Roebber 2003); storms must have been separated from other reflectivity maxima by greater than two diameters of the area containing reflectivity values of >45 dBZ (Fig. 2a);
  • Cluster (CL)—areas of unorganized convection, with several (three or more) reflectivity maxima located within a distance equivalent of two diameters of the 45-dBZ reflectivity area of each individual storm (Fig. 2b); areas of >45-dBZ reflectivity were generally small (<40 km2) for individual storms and were separated by reflectivities <35 dBZ;
  • Complex (C)—nonlinear, organized storm structure having an area of >500 km2 with continuous reflectivity values of >45 dBZ (Fig. 2c); and
  • Linear (L)—an area with reflectivity values >45 dBZ organized in a curvilinear manner (Fig. 2d); storms were considered linear if they were organized in a line <50 km wide, exhibited a length–width ratio of at least 3:1 (Fowle and Roebber 2003), and areas of reflectivity >45 dBZ were separated by less than two of their diameters.
Fig. 2.
Fig. 2.

Base reflectivity images from the Cleveland WSR-88D radar showing examples of (a) isolated, (b) cluster, (c) complex, and (d) linear storm modes used in this study.

Citation: Weather and Forecasting 27, 5; 10.1175/WAF-D-11-00076.1

c. Analysis of storm evolution

Radar reflectivity changes were used as an indication of storm intensity variations. The maximum base reflectivity, recorded to the nearest whole decibel, at the 0.5° elevation angle was determined from KCLE Level II data for each volume scan (approximate time interval of 5 min). Maximum reflectivity was recorded from 30 min before (−30 min) the time the storm moved over the water (TMOW) to at least 60 min following the TMOW (+60 min), or until the storm arrived at the downwind shoreline. In events when the storm had not developed at −30 min, the maximum reflectivity was determined from the volume scan at the time when the maximum reflectivity exceeded 35 dBZ prior to TMOW.

The storm evolution over Lake Erie was determined by comparing the maximum reflectivity at TMOW with both the maximum reflectivity at 30 (+30 min) and at 60 min (+60 min) over the lake. For simplification of discussion, five categories were used to describe the evolution of an individual storm:

  • major strengthening, with ≥ +8 dBZ maximum reflectivity change,
  • moderate strengthening, with +3 to +7 dBZ maximum reflectivity change,
  • no change, with 2 to +2 dBZ maximum reflectivity change,
  • moderate weakening, with −3 to −7 dBZ maximum reflectivity change, and
  • major weakening, with ≤ −8 dBZ maximum reflectivity change.

d. Environmental and storm parameters

To examine the influence of environmental conditions on storm evolution over Lake Erie, several parameters were developed to represent the atmospheric conditions that storms encountered. These parameters and their associated definitions are given in appendix B. The data used to calculate these parameters for each storm are described below.

Observations of temperature, wind speed, and wind direction collected at surface sites over land near the upwind (downwind) shore of Lake Erie were used to represent the upwind (downwind) conditions encountered by each storm (Fig. 1). The upwind and downwind observation sites were relative to individual storm motion as determined from radar data. The majority (about 75%) of the storms examined had mean motions in the range from southwest (225°) through northwest (315°C). Observation sites used for each storm were those deemed closest to the storm path. Only observations taken before the storm appeared to influence conditions at each site were used for this purpose. Overlake conditions were derived from air and water temperature data collected at National Oceanic Atmospheric Administration (NOAA) buoy 45005 (Fig. 1) for each storm when the buoy was operational (generally April–November). All buoy data were acquired from the National Data Buoy Center. Storms for which buoy data were not available were excluded from any analyses that included overlake conditions. Velocity azimuth display (VAD) wind analyses provided in the KCLE WSR-88D Level III dataset were used to represent surface to 3-km and surface to 6-km wind velocities.

3. Results

a. Temporal distribution of storms

The largest frequency (76%) of storms moving over Lake Erie occurred in the months of May–August, with the maximum in July (Fig. 3). Isolated, complex, and linear storms showed similar monthly trends while cluster storms exhibited a peak in May. As expected, storms were more frequent during the afternoon and evening hours (Fig. 4) with differences in timing between modes. Cluster and isolated storms had a relative maximum during the late afternoon and evening [1400–2200 local standard time (LST)], indicative of the importance of surface buoyancy to their development. Linear storms showed a broad maximum during the late morning to late evening (1000–0200 LST) and minimum near sunrise (0600–1000 LST). The relative maximum for complex storms occurred during the overnight (2200–0200 LST) and morning (0600–1000 LST) hours, similar to the pattern of initial convective development in the late evening and maximum extent in the overnight hours observed by Ashley et al. (2003). This is consistent with the findings of Graham et al. (2004) and Parker (2008) that MCS events can maintain their evolution over shallow stable layers.

Fig. 3.
Fig. 3.

Number of occurrences (111 total) of each storm mode by month.

Citation: Weather and Forecasting 27, 5; 10.1175/WAF-D-11-00076.1

Fig. 4.
Fig. 4.

Same as Fig. 3 except by time of day (using TMOW). Time of day shown in local standard time (LST = UTC − 4/5 h).

Citation: Weather and Forecasting 27, 5; 10.1175/WAF-D-11-00076.1

b. Evolution of storms over Lake Erie

The evolution of storm intensity, estimated by changes in the maximum base reflectivity, as a function of time over Lake Erie and storm mode are shown in Table 1. Regardless of storm mode, increases of ≥3 dBZ are rare (5% of events) and no storms exhibited major strengthening (≥8 dBZ).

Table 1.

Percentage of events of each storm mode (rows) that experienced associated changes in maximum reflectivity at (top) 30 and (bottom) 60 min over water. The bold italic values denote the highest percentage for each mode. The “onshore” column denotes percentage of events that moved over the downwind shoreline. The numbers in parentheses below the convective mode represent the number of events in each category. Far-right column shows the average maximum reflectivity change of each storm mode over water.

Table 1.

One of the most notable features is the difference in storm evolution at +30 min after TMOW and +60 min after TMOW (Table 1). At +30 min, the majority of storms experienced no change in intensity (−2/+2 dBZ), with smaller percentages of all storms experiencing moderate weakening (−3/−7 dBZ) and significant weakening (≤−8 dBZ). This was consistent for all storm modes. On average, the maximum base reflectivity only changed by −1 to −2 dBZ from the time the storm crossed the upwind shore. In contrast, excluding storms that moved onshore, by +60 min only about a third of all storms experienced no change in reflectivity and 28% of the storms experienced major weakening. Average reflectivity changes ranged from −2 to −9 dBZ at +60 min from TMOW.

Differences in storm evolution over Lake Erie were noted between storm modes at +60 min (Table 1). Linear and complex storms tended to weaken less than isolated and cluster storms. For example, average reflectivity changes were only −2 to −3 dBZ for linear and complex storms, while for isolated and cluster storms the average reflectivity changes were about −7 to −9 dBZ. Additionally, only 10%–15% of linear and complex storms experienced major weakening at +60 min, compared to 40% of isolated and cluster storms.

c. Relationships with environmental parameters

Graham et al. (2004), Lericos et al. (2007), and Parker (2008) found that changes in environmental conditions associated with a land–water boundary can play important roles in subsequent storm evolution. While some factors affecting the role of the boundary slowly change over long time scales (e.g., surface friction changes over land due to plant growth), forecasting applications focus on shorter time-scale changes due to variations in atmospheric conditions. This study focuses on how short-term changes in atmospheric conditions influence storm evolution over Lake Erie.

To investigate which parameters or combinations of parameters are most closely related to storm evolution, stepwise linear multivariable regression (SLMR) analysis was conducted. SLMR selects only the most statistically significant parameters for inclusion in the linear regression model, which makes it useful for identifying parameters that exhibit the largest influence on storm evolution. The regression model takes the form
eq1
where ΔdBZ is the change in reflectivity at +60 min, C is a constant, Vi are the predictor parameters, and βi are the coefficients of Vi. Only parameters that have a statistical significance level (α) of α ≤ 0.10 are retained. In events where no parameters met the statistical significance criteria, the parameter with the highest statistical significance is shown.

The SLMR used +60-min dBZ change as the predictand and five predictor parameters: 1) the near-surface air temperature over the lake (LAT); 2) OLBL stability, estimated by the difference between the water temperature and the overlake near-surface air temperature (LT-LAT, negative values imply stable conditions); 3) spatial variations in near-surface conditions, estimated by the difference between the overlake near-surface air temperature and the upwind (relative to storm motion) overland near-surface temperature (LAT-UWT); 4) larger-scale spatial variations in near-surface conditions not associated with the lake, estimated by the difference between the downwind and upwind near-surface equivalent potential temperature (θe); and 5) 3-km VAD wind velocities, which may also give some indication of the ambient low-level vertical wind shear. Changes in surface temperature, low-level stability, and low-level wind speed were included, as they are factors generally cited as influencing convective structure and evolution.

Table 2 shows which atmospheric parameters had the most influence on storm evolution determined by the SLMR analysis: overlake air temperature (LAT) for cluster and complex storms, horizontal temperature differences (LAT-UWT) for isolated storms, and 3-km wind speed for linear storms. The results suggest that near-surface temperature changes associated with the OLBL are the most influential parameters for all storm modes, with the exception of linear. However, the differing signs of the β1 coefficients suggest the effect of lake-induced temperature differences depends on storm mode. The positive β1 found for cluster and isolated modes shows a direct relationship between reduced overlake air temperature and storm intensity decrease. However, the negative β1 value for complex storms shows that cooler overlake temperatures are associated with increasing storm intensity.

Table 2.

Summary of the results from SLMR using atmospheric parameters to predict storm intensity changes over Lake Erie (at 60 min over the water), for each storm mode. Columns 1–3 (italicized) display information about the linear regression model developed with the selected parameters. Columns 4 and 5 display information about the chosen parameters inside the model. Here, N is the number of storms included in the regression, R2 represents the coefficient of determination (goodness of fit) of the regression model, and significance displays the statistical significance level (α) of the model.

Table 2.

It should be noted that very few events exhibited environmental or storm evolution changes far removed from all of the other events. Such events can have large influences on the relationships determined by SMLR. If statistical outliers (defined as greater than three standard deviations) are removed, the same parameters were chosen by SMLR. However, the significance of the regression model was greatly improved for the isolated (α = 0.250) mode.

Another method of illustrating the effect of multiple parameters on storm evolution over the lake is flowchart analysis. Flowcharts (Fig. 5) were created to demonstrate how maximum reflectivity changed, at 60 min over the water, with regard to combinations of two parameters: overland–lake temperature difference (LAT-UWT+/−) and overlake air temperature (LAT greater than or less than 22°C). The value of 22°C for the overlake air temperature was chosen to approximately denote transitions between cool and warm seasons. For this analysis, cluster and isolated storms were combined into an unorganized mode and linear and complex storms were combined as organized modes due to their similar evolutionary patterns while traversing the lake (Table 1).

Fig. 5.
Fig. 5.

Flowcharts for (a) cluster and isolated, and (b) linear and complex storms. Numbers given in parentheses are the number of observed storm occurrences for each given parameter combination. The numbers below each parameter give the percentage of events that exhibited, after 60 min over the water, no change or moderate strengthening/major weakening. Percentages are not shown for conditions that had less than five storm occurrences.

Citation: Weather and Forecasting 27, 5; 10.1175/WAF-D-11-00076.1

Each row of the flowchart represents a way of categorizing the storms: 1) convective mode, 2) overland–lake temperature difference, and 3) overlake air temperature. For example, the first two rows in Fig. 5a illustrate that cluster and isolated storms were divided into 7 events with conditions over the lake warmer than those over the upwind shore area (LAT-UWT > 0) and 36 events occurred when the overlake air temperature was cooler than at the upwind shore (LAT-UWT < 0).

Each box in the flowchart gives information on the parameter used to categorize the events and the number of events (above the horizontal line) and convective trends (below the horizontal line). Convective trends are summarized by two categories: no change/moderate strengthening or major weakening (refer to section 2 for evolution definitions).

Results shown in Fig. 5 suggest that cluster/isolated storms tend to be sensitive to overlake changes in near-surface air temperature, and a greater percentage weakened as the air temperature decreased from land to lake (LAT-UWT <0). This is the opposite of the results for linear/complex storms, which showed less of a decrease in intensity when the overlake environment was cooler.

4. Discussion and conclusions

Results of a climatological study examining the evolution of convective storms crossing over Lake Erie suggest that storm evolution is strongly related to storm mode, amount of time the convection has spent over the water, low-level wind speed, and near-surface temperatures over and near the lake. Most observed storms weakened after crossing the upwind shore of Lake Erie; storms rarely exhibited an increase in intensity of >3 dBZ over the lake. While minor weakening of the convective storms was observed at 30 min after crossing the upwind shore, most substantial decreases in intensity were not observed until convection had spent 60 min over the water. This may reflect a common time scale for atmospheric circulations associated with the convection to incorporate surface-influenced air parcels and influence precipitation rates.

An important finding is that storm mode plays a critical role on the amount of weakening that storms experience while over Lake Erie. Cluster and isolated storms tended to decrease in intensity much more rapidly than linear and complex storms. This decrease in intensity was found to be most strongly correlated with the overlake near-surface air temperature (LAT) and its relation to the overland air temperature (LAT-UWT). Overlake air temperature and horizontal land–lake temperature differences were diagnosed as the most influential parameters by stepwise linear multivariate regression (SLMR). It is thought that these relationships reflect the dependence of cluster and isolated convection on local near-surface stability, in agreement with their peak occurrence during times of greatest surface heating: late afternoon and evening. As a result, the majority of these storms were observed to undergo substantial weakening while over the lake, particularly when the overlake air temperature was cooler than the upwind overland air temperature (LAT-UWT <0).

Linear and complex storms were shown to be less sensitive to local lake-induced atmospheric changes and more influenced by the 3-km wind speed. This finding appears to be consistent with previous studies (Weisman et al. 1988; Rotunno et al. 1988; Weisman and Rotunno 2004; Bryan et al. 2006) that suggest dependence between linear storm intensity and low-level vertical wind shear.

While the relationship between surface temperatures and intensity changes of the storms tended to be weaker than 3-km wind speed for linear and complex storm modes, surface temperatures did exhibit an interesting influence. The rate of decrease of reflectivity within linear and complex storms was inversely correlated with land–lake air temperature differences. SLMR found this relationship to be nonstatistically significant. However, this relationship is supported by the flowchart analyses, which showed that organized (linear and complex) storms maintained intensity more often in events where the OLBL was cooler (relative to land). These results are in agreement with the earlier observational results of Graham et al. (2004) for Lake Michigan and the idealized modeling results of Parker (2008).

It is important to note that the data used in this study do not capture potentially important features in the synoptic environment associated with the storms. Features such as elevated or midlevel instability, or the location of the convection relative to the original convective forcing, could not always be determined. Indeed, preliminary examination of storm environments suggested that complex-mode storms were generally found to form north of surface warm fronts or during the local overnight/morning hours, suggesting that a majority of them are elevated. This would likely result in minimal interactions with the OLBL. This is a topic of that should be explored in more detail in future studies. It would also be anticipated that the thermodynamic structure within the OLBL, as well as its depth, which could not be determined in this study, play significant roles in storm evolution over Lake Erie.

The influences of the OLBL on convective storms are anticipated to be different depending on the size of the water body. For the Great Lakes it is possible that in events where the OLBL is highly stable, unstable air from nearby land areas may move over the OLBL and help maintain the convective storm. This seems likely for organized convective storms, where bore propagation atop a highly stable OLBL would work to lift unstable air atop the OLBL. For an oceanic coastal region, no downwind shore is available to maintain convective buoyancy. Further observational and numerical modeling studies of convective changes over lakes of various sizes and convective types are needed. To fully understand the mechanisms influencing convective storm evolution over a water body, further investigation of the relative importance of the sources and sinks of buoyant energy would be highly beneficial.

Finally, smaller-scale variations in surface conditions can play important roles in storm evolution. For example, lake surface temperatures can vary rapidly with time and space over a lake. Likewise, spatial variations in atmospheric conditions, such as stability and wind characteristics, could not be investigated with currently available observational datasets. Future studies should seek to quantify the influences of such surface and atmospheric conditions on storm evolution over the Great Lakes.

Acknowledgments

This research was supported by the National Science Foundation Mesoscale and Dynamic Meteorology Program (Grant ATM 07-11033) and the Cooperative Meteorology, Education, and Training (COMET) program (Grant COM UCAR S09-71437). We appreciate the comments and suggestions from Michael Timlin and James Angel from the University of Illinois at Urbana–Champaign, the anonymous reviewers, and the editor. This manuscript represents the opinions of the authors and does not necessarily reflect the views of the funding agencies or the authors’ affiliations.

APPENDIX A

Observation Stations

Table A1.

The station location, station ID, and type of data collected for each of the observation stations used in this study. Stations where “precip” data were used are considered to be “upwind” of Cleveland for this analysis.

Table A1.

APPENDIX B

Study Parameters

Table B1.

Definitions and abbreviations of each of the parameters used for analyses in this study.

Table B1.

REFERENCES

  • Ashley, W. S., , Mote T. L. , , Dixon P. G. , , Trotter S. L. , , Powell E. J. , , Durkee J. D. , , and Grundstein A. J. , 2003: Distribution of mesoscale convective complex rainfall in the United States. Mon. Wea. Rev., 131, 30033017.

    • Search Google Scholar
    • Export Citation
  • Augustine, J. A., , Woodley W. L. , , Scott R. W. , , and Changnon S. A. , 1994: Using geosynchronous satellite imagery to estimate summer-season rainfall over the Great Lakes. J. Great Lakes Res., 20, 683700.

    • Search Google Scholar
    • Export Citation
  • Bryan, G. H., , Knievel J. C. , , and Parker M. D. , 2006: A multimodel assessment of RKW theory’s relevance to squall-line characteristics. Mon. Wea. Rev., 134, 27722792.

    • Search Google Scholar
    • Export Citation
  • Fowle, M. A., , and Roebber P. J. , 2003: Short-range (0–48 h) numerical prediction of convective occurrence, mode, and location. Wea. Forecasting, 18, 782794.

    • Search Google Scholar
    • Export Citation
  • Gallus, W. A., Jr., , Snook N. A. , , and Johnson E. V. , 2008: Spring and summer severe weather reports over the Midwest as a function of convective mode: A preliminary study. Wea. Forecasting, 23, 101113.

    • Search Google Scholar
    • Export Citation
  • Graham, R., , Bentley M. , , Sparks J. , , Dukesherer B. , , and Evans J. , 2004: Lower Michigan MCS climatology: Trends, pattern types, and marine layer impacts. Preprints, 22nd Conf. on Severe Local Storms, Hyannis, MA, Amer. Meteor. Soc., 7B.6. [Available online at http://ams.confex.com/ams/pdfpapers/81343.pdf.]

  • Laing, A. G., , and Fritsch J. M. , 1997: The global population of mesoscale convective complexes. Quart. J. Roy. Meteor. Soc., 123, 389405.

    • Search Google Scholar
    • Export Citation
  • Lericos, T. P., , Fuelberg H. E. , , Weisman M. L. , , and Watson A. I. , 2007: Numerical simulations of the effects of coastlines on the evolution of strong, long-lived squall lines. Mon. Wea. Rev., 135, 17101731.

    • Search Google Scholar
    • Export Citation
  • Parker, M. D., 2008: Response of simulated squall lines to low-level cooling. J. Atmos. Sci., 65, 13231341.

  • Rotunno, R., , Klemp J. B. , , and Weisman M. L. , 1988: A theory for strong, long-lived squall lines. J. Atmos. Sci., 45, 463485.

  • Scott, R. W., , and Huff F. A. , 1996: Impacts of the Great Lakes on regional climate conditions. J. Great Lakes Res., 22, 845863.

  • Weisman, M. L., , and Rotunno R. , 2004: “A theory for strong, long-lived squall lines” revisited. J. Atmos. Sci., 61, 361382.

  • Weisman, M. L., , Klemp J. B. , , and Rotunno R. , 1988: Structure and evolution of numerically simulated squall lines. J. Atmos. Sci., 45, 19902013.

    • Search Google Scholar
    • Export Citation
1

Overlake boundary layers are often identified in the scientific literature and forecast discussions as marine internal boundary layers. OLBL is used in this article to emphasize the more limited spatial extent of lakes relative to oceans.

Save