National Weather Service Forecast Office
White Lake, Michigan
Previous research has identified the following variables as most important in forecasting the location and intensity of lake effect snow:
Most published procedures for forecasting lake effect snow include a section where the vertical velocity at 700-mb is considered (i.e., Dockus 1985, Niziol 1987). While 700-mb may still be the most common level at which vertical velocity is diagnosed, the introduction of new technology to the operational community makes it possible to examine vertical velocity at other levels. Since most lake effect snow occurs in clouds with tops at or below the 700-mb level, it is likely that synoptic-scale vertical motion at levels below 700-mb would have an effect at least comparable in magnitude to the effect of vertical velocity at 700-mb.
The purpose of this study was to develop a record of the characteristics of synoptic-scale model forecasts of vertical velocity during potential lake effect snow events in southwest lower Michigan. (The definition of a potential lake effect snow event is given in Section 2.) Model forecasts of vertical velocity at 700-mb and 850-mb will be recorded for several events. Also, since relative humidity is closely related to vertical velocity, model forecasts of relative humidity at 700-mb and 850-mb will be recorded. Finally, to study the effect of the forecasted vertical velocity on the local static stability, several soundings will be examined and related to the model forecasts. The study will include events where lake effect snow occurred, and events where several factors pointed to the possibility of lake effect snow, yet no significant snows accumulated. In Section 2 of this paper, the method for developing the record is described. Section 3 gives the results of the study. A summary and discussion will be presented in Section 4.
To develop a database for this work, data was collected from 31 distinct 24-hour periods during which several favorable characteristics for lake effect snow was present across southwest lower Michigan. To simplify the study, it was decided to focus on times where the snow was not being strongly enhanced by synoptic-scale lifting mechanisms. Therefore, times were chosen during which no major synoptic-scale surface troughs or cyclones were located near western Michigan. While this constraint confined the study to events where the model forecasts of upward vertical velocity were less than 4 µb/s, it is hypothesized that even small synoptic scale vertical velocities can still play an important role in determining snow amounts by providing either favorable or unfavorable "background" conditions for lake effect snow. In addition, the following characteristics were present at some time during each 24-hour period:
For each period, snowfall data was taken from nine cooperative observation stations that reported 24-hour snowfalls within two hours of 12 UTC (Figure 1). In addition, 24-hour snowfall totals from 12 UTC to 12 UTC at WSO Muskegon (MKG) and WSO Grand Rapids (GRR) were incorporated into the study. Snowfall for each period was categorized as either trace, light, or heavy, based on the following definitions:
|trace:||No more than one station reported a 24-hour snowfall of an inch or more.|
|light:||Two or more stations reported a 24-hour snowfall of an inch or more, with fewer than 2 stations reporting 3 inches or more.|
|heavy:||Two or more stations reported a 24-hour snowfall of 3 inches or more.|
Figure 1. A map indicating the location of the nine cooperative observation sites used for 24 hour snowfalls, plus the location of Muskegon and Grand Rapids.
For each period, model forecasts of vertical velocity, relative humidity, wind direction, wind speed and 850-mb temperature were recorded. For periods that occurred before 11/93, values were taken from alphanumeric displays of NGM gridded forecasts, available at 6 hourly increments and interpolated between grid points to a point near Muskegon. For periods from 11/93 through 4/95, values were inferred by viewing ETA model forecast time-cross Sections generated by PCGRIDDS at 43°N, 86°W (again, near Muskegon). For each event, values were taken from the model run that began less than 12 hours before the beginning of the lake-snow event. From each of these model runs, the values were taken from the 12 through 36 hour forecasts.
Using the guidelines from Section 2, 31 events were identified. The snowfall was classified as "trace" for nine events, "light" for ten events, and "heavy" for 12 events. For the "trace" events, the average 850-mb wind direction for every 12 hourly (12 UTC or 00 UTC) time associated with an 850-mb temperature less than or equal to -12°C was 302°. Likewise, the average wind direction was 310° for the "light" events and 287° for the "heavy" events. The most variation in wind direction occurred with the "heavy" events, when minor short-wave troughs sometimes produced wind shifts from WSW to NW during the event.
Figures 2 through 6 show plots of forecast 850-mb temperature, vertical velocity and relative humidity, delineated by snowfall category. For the "trace" events, values in Figures 2-6 are plotted for every 12 hourly (12 or 00 UTC) time during which the average 1000-850-mb wind was greater than or equal to 290° and < 340°, the 850-mb temperature was less than or equal to -12°C, and the 1000-700-mb directional wind shear was less than or equal to 40°. As a result, 16 x's are plotted representing the nine "trace" events, since several 24-hour events contained two 12 hourly periods that met the criteria stated above.
Since it was impossible to know the exact time during which snow was falling during the 24-hour periods of "light" or "heavy" snow, the conditions that were most likely during the heaviest and most widespread snowfall during those periods had to be estimated. Therefore, the following conventions are used for the plots of the data from the "light" and "heavy" events in Figures 2-6. The lowest 24-hour 850-mb temperature forecast for each event is plotted. The lowest 24-hour value of forecast vertical velocity (weakest subsidence or strongest ascent) is plotted, given that the vertical velocity must have coincided with an 850-mb temperature forecast less than or equal to -12°C. Finally, the highest 24-hour values of forecast relative humidity are plotted, given that the relative humidity must have coincided with an 850-mb temperature forecast less than or equal to -12°C.
The data from Figure 2 indicates that the "trace" events all occurred with 850-mb temperature forecasts greater than or equal to -20°C. Meanwhile, a greater spread in the data is indicated for the "light" and "heavy" groups, with several colder events indicated, especially in the "heavy" group. It should be noted that even though the "heavy" group contained the events with the lowest forecast 850-mb temperatures in the study, it still contained seven events where the lowest predicted 850-mb temperature was greater than or equal to -20°C.
Figure 2. Plots of model forecasts of 850 mb temperatures (°C) for the 28 cases in the study delineated by snowfall category.
Figure 3 shows plotted values of 700-mb vertical velocity forecasts for the 31 times in the study. The data from Figure 3 indicates that mostly subsidence was forecasted during the "trace" and "light" events. Predicted subsidence appears to have been most pronounced in the "trace" cases, with values mostly greater than 1 µb/s indicated, compared to values mostly between 0 µb/s and 1 µb/s in the "light" cases. Meanwhile, mostly ascent was predicted for the "heavy" events. Although none of the "heavy" events occurred with a major synoptic-scale trough or cyclone in the vicinity (recall Section 2), evidently low amplitude short-wave troughs were still forecast to produce periods of weak upward vertical motion across the area in most of these cases.
Figure 3. Plots of model forecasts of 700 mb vertical velocity (µb/s) for the 28 cases in the study delineated by snowfall category.
Figure 4 shows plotted values of 850-mb vertical velocity forecasts for the 31 times in the study. Subsidence was forecast for all of the "trace" cases, with values mostly from 1 µb/s to 2 µb/s. Meanwhile, weak vertical velocity was predicted for the "light" cases, with values mostly between -1 µb/s to +1 µb/s. Finally, weak ascent was forecast for most of the "heavy" cases, with values mainly between 0 µb/s and -1 µb/s.
Figure 4. Plots of model forecasts of 850 mb vertical velocity (µb/s) for the 28 cases in the study delineated by snowfall category.
Figure 5 shows plotted values of 700-mb relative humidity forecasts for the 31 times in the study. Values less than or equal to 30 percent were forecasted for most of the "trace" cases, while values between 30 and 60 percent are indicated for most of the "light" cases. Meanwhile, values greater than or equal to 60 percent were forecasted for the majority of the "heavy" cases, with values of greater than or equal to 80 percent predicted for nearly half of those cases.
Figure 5. Plots of model forecasts of 700 mb relative humidity (%) for the 28 cases in the study delineated by snowfall category.
Figure 6 shows plotted values of 850-mb relative humidity forecasts for the 31 times in the study. For the "trace" cases, a large spread is indicated, with most forecasted values less than or equal to 70 percent. Meanwhile, higher numbers are indicated for the "light" events, with most of the predicted values > 50 percent. Finally, very high numbers are indicated for the "heavy" events, with all of the forecasted values greater than or equal to 80 percent.
Figure 6. Plots of model forecasts of 850 mb relative humidity (%) for the 28 events in this study delineated by snowfall category.
A significance test on the differences between the means of the data from the "trace" vs. "light" events, and the "light" vs. "heavy" events were performed using a non-parametric Kruskal- Wallis multiple comparison technique with a significance level of 0.10 (Gibbons 1976). The differences between all of the means were significant, except the differences between the means of the 850-mb temperatures for the "light" vs. "heavy" cases.
Figure 7 shows soundings from two of the "trace" events, taken at Green Bay, Wisconsin (GRB), on the upstream side of Lake Michigan. These soundings are representative of the temperature and moisture profiles that were most often observed at GRB during the "trace" events in this study.
Figure 7a shows the GRB sounding from 12 UTC 11 December 1994. A shallow low-level moist layer is evident, capped by a low-based subsidence inversion at 900-mb. In this case, the inversion probably inhibited the development of lake effect snow on the downstream side of the lake, despite any low-level moistening and destabilization that may have occurred as the air crossed Lake Michigan. Figure 7b shows the GRB sounding from 12 UTC 4 April 1994. In contrast to Figure 7a, a relatively deep mixed boundary layer is indicated, with a subsidence inversion above 850-mb. In this case, the inversion at GRB was located at a height that is often associated with significant lake effect snow, but any lake effect snow was probably quickly shut off as subsidence resulted in a lowering of the dry air and the inversion as the air passed across Lake Michigan.
Figure 8 shows soundings taken at GRB during two of the "heavy" events. Again, these soundings were chosen because they were somewhat representative of the temperature and moisture profiles most commonly observed during the "heavy" events in the study.
Figure 8a shows the sounding taken at 00 UTC 5 December 1992. In contrast to the soundings in Figure 7, no low-level temperature inversion is indicated. Also, more moisture is indicated, especially in the layer from 850 to 700-mb. In summary, the sounding was much more favorable for convection than were the soundings in Figure 7. Figure 8b shows the sounding taken at 12 UTC 25 December 1993. Compared with the sounding in Figure 8a, a more stable temperature profile is indicated, especially below 850-mb. However, as in Figure 8a, ample moisture is indicated in the layer from 850 to 700-mb. This sounding could have become very similar to the sounding shown in Figure 8a, if low-level destabilization and moistening were to occur as the air traveled across Lake Michigan.
Model forecasts of several variables were recorded for 31 periods during which several factors were favorable for the development of lake effect snow in southwest Lower Michigan. It was found that predictions of synoptic-scale vertical velocity were directly related to snow amounts across the area, with forecasted subsidence associated with light accumulations, and forecasted ascent associated with heavier accumulations. The vertical velocity predictions at 850-mb were found to have a relationship comparable in magnitude to the forecasts of vertical velocity at 700-mb. As a preliminary rule of thumb, it appears that "trace" snowfalls are favored when the subsidence is forecast to exceed 1 µb/s at 850-mb and 2 µb/s at 700-mb. Likewise, "heavy" snowfalls are favored when any ascent is forecast at those levels.
Because of the relatively narrow east-west width of Lake Michigan (50-70 miles), the over-water fetch length is often marginal (80-120 miles) for "pure" lake effect snow in northwesterly low-level flow across southwest Michigan. Another factor that makes it difficult for heavy lake effect snow to develop across southwest Michigan without synoptic-scale upward motion is the lack of raising terrain downstream of the lake. As a result, lake effect snow in southwest Michigan often occurs with multiple wind-parallel bands of weak to moderate intensity. In contrast, the east-west orientation of Lakes Erie and Ontario, the upwind presence of Lake Huron, and the presence of higher terrain, all allow for a more favorable environment for heavier, shoreline-parallel snow bands across the northeast Ohio and western New York snow-belts. As a result, the synoptic-scale vertical motion may frequently play a bigger role in determining snow amounts across southwest Michigan, especially in the northwest flow regimes examined in this study.
Heavy lake-effect snow sometimes develops along the eastern shoreline of Lake Michigan in cases with a very light low-level flow when an elongated north-south snow band associated with some mid-lake convergence zone drifts east to the eastern shore. In these types of cases, heavy snow may accumulate despite synoptic-scale subsidence, since air parcels have particularly long residence times over the lake due to the light flow. Also, heavy snow may fall with strong synoptic-scale subsidence in strong northerly flow, when over-water fetch lengths can approach or exceed 200 miles. These types of cases were not considered in this study, since it was decided that they were significantly different from the majority of cases in the data sample.
Vertical profiles of model forecasts of relative humidity were also examined for the same 31 periods. It was found that the magnitude of the relative humidities at 700-mb and 850-mb were both directly related to snowfall amounts. As a preliminary rule of thumb, "trace" amounts of snow are favored when 850-mb relative humidity is forecast > 50 percent, and 700-mb relative humidity is forecast below 30 percent. Also, "heavy" amounts are favored when 850-mb relative humidities are forecast > 80 percent, and 700-mb relative humidities are forecast above 70 percent. To expand on these findings, Green Bay soundings were examined from days when the snowfall in southwest Michigan was classified as "trace", and days when the snowfall was classified "heavy". It was found that soundings taken on days where the snowfall was classified as "trace" were sometimes characterized by strong low-based temperature inversions, and were usually characterized by dry air in the layer from 850-700-mb. In contrast, soundings taken on days when the snowfall was classified as "heavy" were usually characterized by plentiful moisture in the layer from 850-700- mb.
The results from this study suggest that low-level subsidence inhibits lake effect snow by contributing to drying in the layer from 850 to 700-mb. Lake effect snow is most favored in an environment where low-level moisture (and high theta-e) is plentiful, since that would often be associated with convective instability through a fairly deep layer of the lower troposphere. The results of the study also suggest that subsidence and its associated drying frequently inhibit lake effect snow by forming a low-level subsidence inversion. In summary, it would seem that examining model forecasts of low-level vertical velocity and relative humidity in a lake effect snow environment may be a quick way to roughly assess how favorable the local sounding will be for the development of moist convection.
From these results, it seems that the temperature profile of the lowest 100 mbs at GRB is not necessarily a reliable indicator for lake effect snow accumulations in southwest Michigan (compare Figures 2b and 3b). This is probably because the temperature profile of the lowest 100 mbs of the troposphere can be strongly modified when air travels from the GRB area southeast across Lake Michigan, and weak low-based temperature inversions can sometimes be eliminated.
Figure 9 shows a flow chart that has been used for several years to help with forecasting lake effect snow in Michigan. While the large-scale temporal nature of the data in this study (24-hour snowfalls) makes it impossible to rigorously verify the flow chart, some conclusions can still be drawn. First, adding a check of the vertical velocity at 850-mb might be a useful addition to the "dynamics" check Section. Also, a distinction between subsidence with magnitudes above and below 1 µb/s might be useful. Finally, there is strong evidence for adding a check of 700-mb and 850-mb relative humidities.
Figure 9. A flow chart used for forecasting lake effect or lake-enhanced snow in Michigan. (Note, flow chart is a three page document.)
Burrows, W.R., 1991: Objective Guidance for 0-24-Hour and 24-48-Hour Mesoscale Forecasts of Lake-Effect Snow Using CART. Wea. Forecasting, 6, 357-378.
Dockus, D.A., 1985: Lake effect snow forecasting in the computer age. Natl. Wea. Dig., 10, 5-19.
Gibbons, J.D., 1976: Nonparametric Methods for Quantitative Analysis. Holt, Rinehart and Winston, 463pp.
Holroyd, E.W., III, 1971: Lake effect cloud bands as seen from weather satellites. J. Atmos Sci.,28, 1165-1170.
Niziol, T.A., 1987: Operational forecasting of lake effect snow in western and central New York. Wea. Forecasting, 1, 311-321.
____________, 1982: A record-setting lake effect snowstorm at Buffalo NY. Natl. Wea. Dig., 7(4), 19-24.
Rothrock, H.J., 1969: An aid in forecasting significant lake snows. ESSA Tech. Memo. WBTM CR-30, NOAA/NWS, Kansas City, MO, 18 pp.
Wagenmaker, R.B., 1987: A Case Study of a Significant Lake Enhanced Snow Event in Upper Michigan. NOAA Technical Memorandum NWS CR-88. 10-25.