The Utility of Additional Soundings for Forecasting Lake-Effect Snow in the Great Lakes Region

Christopher P. J. Scott Department of Atmospheric, Oceanic and Space Sciences, University of Michigan, Ann Arbor, Michigan

Search for other papers by Christopher P. J. Scott in
Current site
Google Scholar
PubMed
Close
and
Peter J. Sousounis Department of Atmospheric, Oceanic and Space Sciences, University of Michigan, Ann Arbor, Michigan

Search for other papers by Peter J. Sousounis in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

The impact of initializing a mesoscale model with additional sounding data over the Great Lakes region is investigated. As part of the Lake-Induced Convection Experiment (Lake-ICE) field study during the winter of 1997/98, six supplementary Cross-chain Loran Atmospheric Sounding System (CLASS) units and three Integrated Sounding System (ISS) units were used in addition to those from the standard synoptic upper-air network. The three ISS units were in the vicinity of Lake Michigan, and the six CLASS units were in the data-sparse region of central and northeastern Ontario and western Quebec.

The Pennsylvania State University–National Center for Atmospheric Research fifth-generation Mesoscale Model running on a doubly nested grid is used to simulate the lake-effect snow event of 4–5 December 1997. This model output from a 30-km horizontal resolution grid shows that the six CLASS soundings capture a warm layer below 850 hPa that appears to be the result of diabatic heating from the Great Lakes. This leads to an improved simulation of the surface pressure fields over the course of the simulation. A nested 10-km horizontal resolution grid shows that the initialization data from the CLASS sites seemed to have a greater influence on the propagation of a mesoalpha-scale trough that caused significant snowfall to the lee of Lake Michigan than data from the ISS sites. The inclusion of the CLASS sounding data changes the track of the precipitation maximum by approximately 25 km and agrees better with reflectivity data from the Weather Surveillance Radar-1988 Doppler. Implications for forecasters in the Great Lakes region are discussed.

Corresponding author address: Dr. Peter J. Sousounis, Dept. of Atmospheric, Oceanic and Space Science, University of Michigan, Ann Arbor, MI 48109-2143. Email: sousou@umich.edu

Abstract

The impact of initializing a mesoscale model with additional sounding data over the Great Lakes region is investigated. As part of the Lake-Induced Convection Experiment (Lake-ICE) field study during the winter of 1997/98, six supplementary Cross-chain Loran Atmospheric Sounding System (CLASS) units and three Integrated Sounding System (ISS) units were used in addition to those from the standard synoptic upper-air network. The three ISS units were in the vicinity of Lake Michigan, and the six CLASS units were in the data-sparse region of central and northeastern Ontario and western Quebec.

The Pennsylvania State University–National Center for Atmospheric Research fifth-generation Mesoscale Model running on a doubly nested grid is used to simulate the lake-effect snow event of 4–5 December 1997. This model output from a 30-km horizontal resolution grid shows that the six CLASS soundings capture a warm layer below 850 hPa that appears to be the result of diabatic heating from the Great Lakes. This leads to an improved simulation of the surface pressure fields over the course of the simulation. A nested 10-km horizontal resolution grid shows that the initialization data from the CLASS sites seemed to have a greater influence on the propagation of a mesoalpha-scale trough that caused significant snowfall to the lee of Lake Michigan than data from the ISS sites. The inclusion of the CLASS sounding data changes the track of the precipitation maximum by approximately 25 km and agrees better with reflectivity data from the Weather Surveillance Radar-1988 Doppler. Implications for forecasters in the Great Lakes region are discussed.

Corresponding author address: Dr. Peter J. Sousounis, Dept. of Atmospheric, Oceanic and Space Science, University of Michigan, Ann Arbor, MI 48109-2143. Email: sousou@umich.edu

1. Introduction

Modern operational weather forecasting of lake-effect snow in the Great Lakes region relies heavily on the output from numerical weather prediction models (Sousounis et al. 1999). Ballentine et al. (1998) documented the skill of a mesoscale model in forecasting the significant lake-effect storm of 4–5 January 1995 in the vicinity of Lake Ontario. This event exemplifies the localized and intense nature of lake-effect snow as 72-h snowfall totals ranged from 15 cm to over 100 cm within the same county. On an annual basis, communities to the lee of Lake Michigan receive as much as 50% of their snowfall from lake-effect precipitation (Kelly 1986). Clearly, there is a need for accurate prediction of this phenomenon.

Despite increases in the complexity of NWP models and computing power, numerical simulations of the atmosphere continue to produce imperfect forecasts. Sousounis et al. (1999) discuss the human–machine blend that is required to successfully predict winter precipitation in the Great Lakes and also stress the imperfections associated with mesoscale model output. Along with model formulation and computing power, one of the primary factors limiting the improvement of these forecasts is the quality and space–time distribution of atmospheric measurements (Emanuel et al. 1997). Certainly, the sensitivity of numerical models to initial conditions has been documented by many studies (e.g., Warner et al. 1989; Rabier et al. 1996; Pu et al. 1998; Gustafsson et al. 1998). Based on experiments at the European Centre for Medium-Range Weather Forecasts, Rabier et al. (1996) state that large forecast errors arise more often from errors in the initial conditions, than from errors in the model physics. Furthermore, Douglas and Stensrud (1996) summarize that small changes in initial conditions can produce significant changes in the model forecast, suggesting “the need for a greater density of mesoscale upper-air observations for model initialization.”

In an attempt to mitigate initial condition errors, operational forecast centers around the world have implemented sophisticated variational data assimilation schemes. For example, Laroche et al. (1999) detail the forecast impact of the Canadian Meteorological Centre's recently implemented three-dimensional variational analysis. Studies by Stauffer and Seaman (1990) and Zheng et al. (1995) have shown that four-dimensional data assimilation (FDDA) can ameliorate the mesoscale spinup problem by generating realistic mesoscale circulations not resolved by the initial data. However, these same studies highlight the fact that FDDA is inherently dependent on initial conditions within the assimilation period and will produce erroneous forecasts in the absence of adequate initial data.

As testament to the importance of this issue, numerous studies have investigated the impact of incorporating additional data into a model initialization. Tuleya and Lord (1997) illustrated that dropwinsonde measurements in the vicinity of hurricanes improved the model forecast track of these systems. Studies by Pouponneau et al. (1999) and Heming (1990) document the positive impact of additional upper-air observations over the data-sparse Atlantic on cyclone track and intensity forecasts. On the mesoscale, Kuo and Guo (1989) demonstrated that a hypothetical network of profilers improved a mesoscale model simulation of precipitation by better defining the divergence field within the boundary layer. Additionally, the model-forecasted initiation and propagation of mesoscale convective systems have been shown to be highly sensitive to mesobeta-scale features that are not resolved by the conventional observing network (Stensrud and Fritsch 1994; Zheng et al. 1995). Of particular note, Gallus and Bresch (1997) incorporated a bogus sounding into the model initial conditions to improve the simulation of a wintertime vortex over the midwestern United States. The inclusion of one additional sounding at initialization significantly improved model-forecasted precipitation, sea level pressure, and boundary layer winds, and the authors suggest that “the addition of actual upper-level observations on the mesoscale can significantly improve short-range mesoscale forecasting.”

The current study assesses the impact of additional rawinsonde observations within the Great Lakes region on a mesoscale model simulation of a lake-effect snow event that occurred on 5 December 1997. As part of the Lake-Induced Convection Experiment (Lake-ICE; Kristovich et al. 2000), six supplementary Cross-chain Loran Atmospheric Sounding System (CLASS) units and three Integrated Sounding System (ISS) units gathered upper-air sounding data. The CLASS sites were located in the data-sparse region of central and northeastern Ontario and western Quebec to capture mesoalpha-scale circulations induced by the aggregate-heating effects of the Great Lakes (Sousounis 1997). Meanwhile, the ISS sites were positioned in the vicinity of Lake Michigan to monitor the mesoalpha-/mesobeta-scale evolution of the boundary layer upwind and downwind of the lake. Collectively, these additional soundings essentially doubled the horizontal resolution of the operational upper-air sounding network (Fig. 1) and thus represent an unprecedented high-resolution dataset for upper-air measurements in the Great Lakes region.

2. Case overview

A lake-effect snow event occurred on 5 December 1997 during the first days of the Lake-ICE field study. Brief heavy snowfall and accumulations of 5–15 cm across southwest Lower Michigan characterized the event.

a. Synoptic and mesoscale conditions for 4–6 December 1997

At 0000 UTC 4 December 1997 (hereafter, times referred to in the format DD/HHMM; e.g., 04/0000), a 1004-hPa low pressure center was located over northern Lower Michigan (Fig. 2a). During the previous 12-h period, a surface trough had migrated eastward from Wisconsin to form the closed surface low in Fig. 2a. Subsequently, Fig. 2b indicates that the surface low deepened by 4 hPa over the next 12-h period as it moved in the vicinity of Lake Huron by 04/1200. Sousounis and Fritsch (1994) demonstrated numerically how heat and moisture fluxes from the Great Lakes typically influence the development of weak lows in the region during winter. Weiss and Sousounis (1999) found that such hydrostatically induced, mesoalpha-scale collective lake disturbances (COLDs) occur approximately 25–40 days per year across the Great Lakes region. Of this total, more than one-third are characterized by a synoptic-scale warm advection scenario with surface temperatures lower than lake temperatures, causing a positive heat and moisture flux from the Great Lakes into the lower troposphere. This suggests that approximately 10 of the 25–40 days associated with COLD development in a given winter are characterized by positive fluxes from the lakes concomitant with synoptic-scale warm advection. The location of a surface trough to the west of the Great Lakes, and its subsequent development into a 1000-hPa surface low by 04/1200, fit this description.

Figure 2 shows the evolution of surface conditions over the Great Lakes for the 36-h period beginning at 04/0000. Surface temperatures of −2° to 4°C were marginally lower than lake temperatures, which averaged between 3° and 7°C, and cyclonic flow around surface low pressure resulted in weak warm advection over the central and eastern Great Lakes prior to 04/1200. Figures 3 and 4 depict the 850- and 500-hPa conditions over the same period. The surface low moves very little over the course of the first 24 h, suggesting that diabatic heating from the lakes plays a role in its maintenance. By 05/1200, a broader area of low pressure begins to develop over southern Quebec and New England in response to the eastward progression of the 500-hPa pattern. Despite the eastward movement of the 500-hPa closed low, and abatement of significant quasigeostrophic forcing, synoptic-scale surface troughing persists over the Great Lakes basin through 06/0000. On the lower mesoalpha-scale, smaller troughs can be seen rotating around the parent area of low pressure. One such trough indicated in Fig. 2 develops over the upper peninsula of Michigan at 04/1200 in response to aggregate heating and moistening of the lower troposphere by the western Great Lakes, particularly Lake Superior. Over the next 24 h, the trough is advected within the broader cyclonic flow and passes through southwest Lower Michigan by 05/1200. This trough was responsible for significant snowfall along the eastern shoreline of Lake Michigan, and was associated with the subsequent development of a lake-effect mesoscale vortex as detailed in Laird et al. (2001).

Overall, a favorable environment for lake-effect snow was in place over Lake Michigan during this period as a northwesterly fetch was coupled with minimal directional wind shear between the surface and 700 hPa (Niziol 1987). More importantly, cold advection at 850 hPa created the necessary thermal instability for lake-effect precipitation. With Lake Michigan water temperatures (TLake) of 4°–7°C and 850-hPa temperatures (T850) ranging from −10° to −8°C by 05/0000, the Holroyd (1971) criterion for an unstable lapse rate (TLakeT850 > 13°C) was exceeded. Data from observing stations throughout southwest Lower Michigan show that an average of 5–15 cm of snow accumulated to the lee of Lake Michigan on 5 December (Fig. 5).

b. Forecast validation

Meteorologists at the Lake-ICE operations center issued daily weather discussions for the project scientists based on a suite of operational and research forecast models (Sousounis et al. 1999). The discussion from 4 December indicates that the lake-effect event was well forecast, and that model guidance was useful in the preparation of the forecast: “Lake-effect convection is expected to continue through the day Friday [5 Dec] over the southern two-thirds of Lake Michigan. Eta and MM5 model soundings for points over central Lake Michigan show that there will be adequate instability below 700 hPa throughout the day to support convective activity… . MM5 PBL winds show a trough moving south from northern Lower Michigan tonight to a BEH to MBS line by 12Z Friday [5 Dec] in response to the passage of the upper low. This will result in a possible re-orientation of the snow bands from WSW-ENE to NW-SE behind the trough; except for one caveat, the potential for a shoreline parallel band north of MKG.”

The discussion even mentions the southward propagation of the trough shown in Fig. 2, indicating some model skill on the mesoalpha and mesobeta scale. Although this was a well-forecasted event on the multicounty level, further improvements to lake-effect forecasts will require the explicit prediction of lake-effect structures and the subsequent snowfall at the (single) county level. This implies having mesoscale model output of at least 10-km horizontal resolution with a temporal resolution approaching 1 h. Ballentine et al. (1998) demonstrated a mesoscale model's ability to simulate an actual lake-effect event at this resolution, but also showed the limitations of the model's ability to predict the intensity and location of precipitation at the county level. It is the current authors' hypothesis that imperfect model initial conditions may be partially responsible for forecast errors associated with lake-effect precipitation, and the case of 5 December 1997 affords an opportunity to further investigate this connection.

3. Mesoscale model simulations

A study was performed to assess the impact of additional sounding data on the model simulation of the 5 December 1997 case.

a. Model parameterizations

The Pennsylvania State University–National Center for Atmospheric Research (PSU–NCAR) nonhydrostatic fifth-generation Mesoscale Model (MM5) was used to simulate the lake-effect event. Specifics of the model formulation are detailed by Grell et al. (1995). The model was run in a doubly nested fashion with a coarse grid domain of 90-km horizontal resolution covering most of North America, a nested interactive 30-km grid covering the Great Lakes region, and a doubly nested 10-km grid covering Lake Michigan (Fig. 6). All three domains contained 35 vertical levels, with a greater density of layers in the lowest 200 hPa of the troposphere to better resolve the convective boundary layer and the vertical motions associated with lake-effect precipitation. Model options that were used included the high-resolution planetary boundary layer scheme of Blackadar (Zhang and Anthes 1982) and the Goddard microphysics package (Tao et al. 1989; Tao and Simpson 1993) for grid-scale precipitation. On the 90-km grid, convective precipitation was parameterized using the Anthes–Kuo scheme (Anthes 1977). For the two fine grids, the Kain–Fritsch convective scheme (Kain and Fritsch 1993) was implemented, with the minimum cloud height for precipitation reduced to 1 km to account for the shallow nature of the lake-effect convection.

b. Model initial conditions

The initial and boundary conditions for the MM5 simulation were provided by the National Centers for Environmental Prediction (NCEP) Eta Model run from 1200 UTC 4 December 1997. At this time, the Eta was being run at a horizontal resolution of 48 km, with 38 levels in the vertical (Mittelstadt 1998). The motivation in using the Eta initial conditions, as opposed to a static NCEP analysis, was to take advantage of the data assimilation system in the Eta Model. Prior to the introduction of a three-dimensional variational data assimilation scheme (3DVAR) in 1998, the Eta employed an optimum interpolation scheme (OI) that assimilated data over a 12-h period prior to each run, assimilating new data every 3 h (Mittelstadt 1998). This assimilation strategy sought to improve the model initial condition at time t = 0 by allowing mesoscale circulations to spinup during a preforecast period. While Laroche et al. (1999) have documented the utility of the more sophisticated 3DVAR, the initial conditions from the Eta using the OI method were the best available for this case. The MM5 was initialized using the Eta analysis and integrated for a period of 48 h beginning 04/1200 December. This control run (CNTRL) provided a baseline forecast against which comparisons could be made.

In order to determine the impact of the additional sounding data from the CLASS and ISS sites, the bogusing procedure within the RAWINS module of MM5 was used (Grell et al. 1995). Gallus and Bresch (1997) used this bogusing procedure to demonstrate the positive impact of an additional sounding on a mesoscale forecast. The primary limitation of this procedure is that the bogused soundings are considered to be “perfect” data, and no error checks are performed. In order to ensure accuracy and consistency, objective quality control by the NCAR Atmospheric Technology Division was performed on the six CLASS soundings and three ISS soundings valid at 1200 UTC 4 December. Further subjective quality control was performed to remove suspect data and ensure dynamical balance at initialization. Table 1 shows the rawinsonde data that were retained for the simulations. The most commonly removed data were boundary layer winds due to navigation-lock problems with the Loran system. Additionally, sensor wetting precluded the use of some dewpoint temperature data. These quality controlled sounding data were then processed by RAWINS and objectively analyzed using a multiquadric interpolation technique (Nuss and Titley 1994) to produce the initial conditions for the model integration. Model simulations were then performed that included both the CLASS and ISS sounding data at initialization (CLASS+ISS), only the CLASS data (CLASSONLY), and only the ISS data (ISSONLY). Comparing these simulations to the baseline CNTRL simulation provided a method of assessing the impact of the additional sounding data on the model forecast.

4. Results

Model output from the 30-km grid indicates that the additional soundings had a significant impact on the mesoalpha-scale environment. Meanwhile, the 10-km grid shows that these large-scale perturbations caused differences in the model solution that were significant on the mesogamma scale.

a. The 30-km grid initialization

The most significant impact of the additional soundings on the model initialization was to effect an area of positive temperature perturbations within the surface to 850 hPa layer in the region of the CLASS soundings. Figure 7a shows the magnitude of the initial 900-hPa temperature perturbations across the Great Lakes region. The plot shows a slight negative perturbation in the vicinity of the ISS sites, while a much larger and stronger positive perturbation exists over Ontario and Quebec where the CLASS sites were located. The cross section of Fig. 7b indicates that the CLASS+ISS simulation initialized temperatures 1°–2.5°C warmer than did the CNTRL simulation below 850 hPa over a large horizontal distance. This warm perturbation is a consequence of the additional rawinsonde observations incorporated into the model initial conditions.

Figure 8 compares the actual sounding taken at Timmins, Ontario, Canada (YTS), to the initialized model profiles from the CLASS+ISS and the CNTRL simulations. The rawinsonde temperature profile is warmer than the CNTRL profile throughout the entire depth of the troposphere. The CLASS+ISS profile does not align perfectly with the rawinsonde profile, but is warmer than the CNTRL, particularly below 850 hPa. Differences in the CLASS+ISS and actual profiles are the result of the interpolation performed when incorporating the sounding data into the model initial condition. While the actual sounding represents one point in horizontal space at any given level, the interpolation scheme uses neighboring data points (i.e., the other CLASS soundings) to create the initial condition profile at YTS.

As illustrated in Fig. 7, the cumulative impact of the CLASS soundings resulted in a thermal perturbation with the envelope of the +1.0°C temperature perturbation contour extending over 800 km in horizontal distance across Ontario. This consideration of scale is important in the context of geostrophic adjustment and the Rossby radius of deformation (Houze 1993), defined as
i1520-0434-16-4-448-e1
for circular flow, where g is the gravitational acceleration, h is the depth of the disturbance, and ζ and f are the relative and planetary vorticities, respectively. As defined here, the Rossby radius of deformation represents a theoretical watershed separating dynamically large (where the momentum field adjusts to the mass field) and small (where the mass field adjusts to the momentum field) systems. At horizontal scales greater than LR, a system is considered dynamically large whereas dynamically small systems exhibit length scales less than LR. Using, h = 1500 m, ζ = 1.2 × 10−4 s−1, and f = 1.1 × 10−4 s−1 for the thermal perturbation over Ontario gives a value for LR of 527 km. Considering that the physical horizontal length scale of the system exceeds 800 km suggests that the Ontario perturbation can be classified as dynamically large. Even without knowledge of the subsequent simulation, a forecaster looking at the scale of this perturbation might suspect that it would have a significant impact on the evolution of the flow over the Great Lakes basin as the momentum field would adjust to the mass field, altering the large-scale flow. It should be noted that some differences were present in the initial momentum fields, particularly near the ISS sites; however, these perturbations were not significant beyond the first 3 h of the simulation.

Notably, the scale of the perturbation over Ontario is similar to that of the lake aggregate and suggests that the lakes may have played a role in its generation. It is hypothesized that this perturbation is the result of sensible and latent heating of the air mass over the Great Lakes (in particular Lakes Huron, Erie, and Ontario) during the 24 h prior to model initialization. The paucity of routine upper-air observations in the area to the north and east of the Great Lakes meant that the CNTRL simulation was unable to resolve this warmer layer. In fact, the positive temperature perturbation could not be resolved because its scale is approximately equal to the operational sounding spacing in this region.

b. The 30-km grid simulation

The impact of the positive temperature perturbation in the region of the CLASS soundings is manifest in the evolution of model simulated sea level pressure depicted in Fig. 9. The positive temperature perturbation in the surface to 850 hPa layer hydrostatically reduced the sea level pressure in the CLASS+ISS simulation. The shaded region in Fig. 9 denotes the envelope of the −1.0 hPa sea level pressure perturbation contour. This feature exhibits a slow drift to the southwest during the first 24 h of the simulation. The southwesterly movement of the negative sea level pressure perturbation is the result of its location within the broad cyclonic flow over the Great Lakes basin. From a Lagrangian standpoint, advections also play a determining role in the evolution of the perturbation. Over a similar period as that shown in Fig. 9, the shaded regions in Fig. 10 denote the evolution of the envelope of 950-hPa negative height perturbations of magnitude larger than −6.0 m. The northeast side of this feature appears to grow between 04/1800 and 05/0000 in response to warm advection by the (CLASS+ISS)–CNTRL 950-hPa perturbation winds. While this alone may not appear to be a critical forecast implication, the maintenance of the perturbation via warm advection may have helped to preserve the dynamically large scale of the system and sustain the magnitude of the feature through the first 24 h of the simulation. It should be noted that the evolution of this 950-hPa height perturbation was nearly identical in the CLASS+ISS and CLASSONLY simulations, while the ISSONLY simulation did not produce such a feature. Certainly, this implies that the positive temperature perturbation initialized by the CLASS soundings over Ontario and Quebec was responsible for the negative perturbations in the 950-hPa height field during the simulation.

Even without the aid of perturbation analysis, the impact of the CLASS soundings can be seen in the sea level pressure field of the CLASS+ISS simulation. While the overall pattern is quite similar to that in the CNTRL simulation, the 1002-hPa isobar extends farther west to encompass northern Lake Huron during the first 18 h of the CLASS+ISS simulation. Comparing the CLASS+ISS and CNTRL simulated sea level pressure fields at 05/0000 (Fig. 9b) to the corresponding surface analysis (Fig. 2c) qualitatively shows that the CLASS+ISS simulation is superior in its depiction of the 1002-hPa isobar over Lake Huron and location of the surface low. Additionally, a comparison of the CLASS+ISS simulated sea level pressure at 05/1200 (Fig. 9d) to the analysis (Fig. 2d) reveals that the location of the 1004-hPa isobar over northern Lake Michigan agrees well with the surface analysis, whereas the CNTRL simulation places this feature 200 km to the east. Overall, the CLASS+ISS simulation was more accurate than the CNTRL run in its depiction of a deeper and broader area of surface low pressure over the Great Lakes.

While the improved sea level pressure solution is a significant result, an operational perspective would frame the impact of the additional soundings in the context of tangible weather. In this spirit, the simulation of the surface trough that propagated down the length of Lake Michigan during the first 24 h of the simulation is an important consideration. Significant snowfall over southwest Lower Michigan accompanied the passage of this trough. Figure 9 demonstrates that both the CLASS+ISS and CNTRL simulations captured the passage of this feature. Notably, the timing and location of the trough are very similar in each simulation on the 30-km grid. Additionally, the model simulations of the trough agree quite well with the analyzed location as shown in Fig. 2. Evaluating the sea level pressure field at this resolution, it would appear that the differences between the two simulations in reference to this trough are cosmetic. However, the boxed area in Fig. 10c reveals significant (CLASS+ISS)–CNTRL perturbation winds at the 950-hPa level coincident with the passage of the trough at 05/0600. This westerly perturbation flow is a response to the negative height perturbations associated with the warm layer initialized by the CLASS soundings, and had a significant impact on the 10-km grid model simulations over Lake Michigan.

c. The 10-km grid simulation

A nested 10-km horizontal resolution grid was placed over Lake Michigan (Fig. 6) to investigate the impact of the additional soundings at the county level. It should be noted that this grid spacing is not sufficient to resolve all lake-effect structures. In an observational study over Lake Michigan, Kristovich (1993) documented the spacing of wind-parallel roll structures associated with lake-effect snow to be between 3 and 10 km. A 10-km grid would not resolve these individual rolls. However, Hjelmfelt (1990) demonstrated that an 8-km grid was sufficient to resolve the mesobeta-scale features associated with different morphological types of lake-effect snow. For the current study, a 10-km grid was sufficient to resolve the heavy snowfall associated with the trough.

The following discussion involving Figs. 11–14 utilizes radar data in an attempt to evaluate the spatial agreement between the 10-km grid simulations and reality. As validated with radar data, all model simulations were approximately 2h too fast with the mesoalpha-scale trough passage through southwest Lower Michigan. The cause of this timing error is unclear, but it is worthwhile to note that Ballentine et al. (1998) also observed a timing error where MM5 was premature with a trough passage. This timing error was common to the four model simulations, and an interrogation of hourly model output showed that no significant differences in the latitudinal progress of the trough existed among the different simulations. As a result of the timing error, spatial comparisons between radar data and model output in Figs. 11–14 show a consistent difference of 1h, 53 min between the two.

Figure 11a shows the 0.5° elevation scan from the Grand Rapids, Michigan (KGRR), Weather Surveillance Radar-1988 Doppler (WSR-88D) depicting equivalent radar reflectivity valid at 05/0733. An east–west-oriented band of higher reflectivity associated with the southward moving trough is evident. The highest reflectivity values are located along the Lake Michigan shoreline in Muskegon County (see Fig. 5 for county locations). It is possible that higher reflectivity values may have existed over Lake Michigan, as the radar would not have been able to sample below 1.5 km at this radial distance of approximately 100 km. However, vertical cross sections through the reflectivity field confirmed that the highest echo tops within the volume scan were associated with the high reflectivity in Muskegon County, suggesting that this precipitation may have been the most intense within the domain. Additionally, wind-parallel rolls can be seen to the south of the major precipitation axis just offshore of Ottawa and Kent Counties.

Figure 11b shows the 1-h quantitative precipitation forecast from the CLASS+ISS and CNTRL simulations valid at 05/0600. As stated previously, all model simulations were early with the passage of the trough. Nevertheless, Fig. 11b indicates that significant differences between the CNTRL and CLASS+ISS simulations exist on the spatial scale. The structure and magnitude of the precipitation fields are similar, but the CLASS+ISS precipitation maximum is shifted east by approximately 30 km (three grid points) and agrees better with the location of the maximum reflectivity as indicated by the KGRR WSR-88D in Fig. 11a. This eastward shift of the precipitation feature in the CLASS+ISS simulation is likely the result of the westerly perturbation flow documented in Fig. 10c. It should be noted that the CLASSONLY and CLASS+ISS simulations were nearly identical in their placement of this feature, while the ISSONLY and CNTRL simulations were also nearly identical to one another, reinforcing the notion that the CLASS soundings, and the initialized positive temperature perturbations, had the greatest impact on the flow within the Lake Michigan convective boundary layer for this particular case.

Despite the increased westerly flow in the CLASS+ISS simulation (Fig. 11c), the precipitation maxima are nearly identical; in fact, the CNTRL simulation maximum is slightly greater. This can be explained by the relative locations of the features. The CNTRL maximum is entirely over the lake where there is a greater diabatic heat source, resulting in a more vortical wind field structure, while the CLASS+ISS maximum occurs in an area of enhanced convergence due to the perturbation westerly flow. Of particular forecasting importance, the location of the CLASS+ISS maximum relative to the highest radar reflectivity values suggests that it gives a more accurate representation of the precipitation field than does the CNTRL simulation.

As the trough propagated to the SSE, a mesoscale vortex developed and propagated onshore in Allegan County, Michigan, around 05/1000. Laird et al. (2001) investigated this feature using a synthetic dual-Doppler analysis. Inspection of the simulated low-level wind field from the various model runs showed that none of them were able to simulate the closed cyclonic circulation associated with the mesoscale vortex that had a scale length of approximately 30 km. The exact reason for the model's inability to accurately simulate the wind field is unclear, but it may be the result of an underestimation of the diabatic heating over the lake, or insufficient resolution. Nevertheless, Fig. 12 shows that the model generated a precipitation feature of similar scale to the radar depiction of the vortex. Figure 12a shows the spiral bands of enhanced reflectivity converging to a weak reflectivity region over southwestern Allegan County. The 1-h precipitation maximum from the CLASS+ISS simulation (Fig. 12b) is almost exactly collocated with this weak reflectivity core. Considering the model's inability to resolve the subsidence associated with this subgrid-scale feature, the CLASS+ISS simulation is very accurate with its depiction of the precipitation field. The CNTRL maximum is centered offshore, and the areal precipitation pattern does not agree as well with the reflectivity display. This shift of nearly 30 km is significant at the county level and would have implications for zone forecasts in the region. The subsequent panels in Fig. 12 demonstrate that the shift in the precipitation feature is almost entirely the result of the CLASS soundings. The ISSONLY simulation is comparable to the CNTRL, while the CLASSONLY simulation is nearly identical to the CLASS+ISS.

Figure 13a shows the 0.5° radial velocity product from KGRR valid at 05/1153. The mesoscale vortex has continued to propagate to the southeast of its position in Fig. 12. The radial velocity display shows a couplet of weak inbound and strong outbound echoes associated with the cyclonic circulation of the vortex. Figure 13b compares the 950-hPa absolute vorticity fields from the CLASS+ISS and CNTRL simulations. Similar to the 1-h precipitation forecast fields, the CLASS+ISS 950-hPa absolute vorticity maximum is displaced east of the CNTRL, and appears to be closer to the center of the circulation as indicated by the radial velocity.

This 950-hPa vorticity feature was produced in all simulations in conjunction with an area of strong 950-hPa convergence and enhanced precipitation, which were all similarly positioned. This relatively coherent feature accompanied the passage of the trough as it moved to the south and east, and was coincident with the precipitation maxima of Figs. 11 and 12 that were better simulated by the model runs that included the CLASS soundings. Figure 14 shows a plot of the CLASS+ISS and CNTRL tracks of the 950-hPa vorticity maximum against the track of the mesoscale vortex as determined by successive 0.5° radial velocity images from KGRR. Despite the model's inability to simulate the closed circulation, there is significant agreement between the onshore track of the vortex and the CLASS+ISS simulated 950-hPa vorticity feature.

It is possible that the track of the actual mesoscale vortex was affected by this vorticity feature and its associated low-level convergence, which resulted in the strong correlation. Radar data suggested that the genesis region of the closed circulation associated with the mesoscale vortex appeared to be offshore from Ottawa County and the concave coastline. Forbes and Merritt (1984) noted that seven out of eight Lake Michigan mesoscale vortices that formed over a 5-yr period originated over the southern third of the lake where the Laplacian of the heating rate and convergence is maximized due to the shoreline configuration. While the model missed the development of the mesoscale vortex, its subsequent propagation correlated well with the 950-hPa vorticity feature of the CLASS+ISS and CLASSONLY simulations. The eastward shift caused by the CLASS soundings resulted in simulated precipitation fields that were remarkably similar to those in radar reflectivity data.

5. Discussion

The current study suggests that the initial conditions for a mesoscale model may not be resolving the entire impact of the lake aggregate during a warm advection scenario. Even with the benefit of the Eta Model's assimilation scheme, the model initial condition did not capture the warmer surface to 850 hPa layer over Ontario and Quebec that the CLASS soundings demonstrated. The impact of this warm perturbation was evident in the eastward shift of a mesoalpha-scale trough associated with lake-effect snow over Lake Michigan, 18–21 h after the model initialization (Figs. 11–14).

The climatological frequency with which such an event may occur is important from a forecasting standpoint. As demonstrated in the case overview, this event can be classified as a warm advection COLD that occurs approximately 10 times during the year. Recognizing that the model initialization over Ontario and Quebec may not capture the effects of the lake aggregate under a warm advection scenario, a forecaster may be able to make the following generalizations about the model forecast. 1) Increased warm advection over northeastern Ontario and western Quebec as a result of the geostrophic adjustment to the positive temperature perturbation (Fig. 10) could result in increased quasigeostrophic forcing and enhanced precipitation in that region. 2) Increased cold advection over the western Great Lakes in response to the positive temperature perturbation may result in decreased static stability and a more favorable thermodynamic environment for lake-effect snow. 3) Subtle changes in the gradient flow as the result of this temperature perturbation may cause shifts in lake-effect precipitation features that may be significant at the county level. This final point is the most important, but it would be a gross generalization to state that every lake-effect feature over Lake Michigan would be shifted east as witnessed in this case. Not surprisingly, Fig. 10 reveals that the 950-hPa perturbation winds are not always in geostrophic balance, and at times appear to have a significant isallobaric component. Even with the understanding that a lake-aggregate-induced low-level warm perturbation may exist, the best adjustment a forecaster can make is to look at high-resolution model output with caution and acknowledge the inherent limitations of county-level quantitative precipitation forecast output given the current observational network.

A plausible solution to improve county-level lake-effect snow prediction would be to implement routine rawinsonde observations in the data-sparse regions of northeastern Ontario and western Quebec where the impact of the Great Lakes aggregate is manifest. This study has shown that a network of CLASS soundings over this area significantly improved the simulation of precipitation in a lake-effect environment over Lake Michigan. Despite the promise of remote sensing systems, studies by Douglas and Stensrud (1996) and Kelly (1997) suggest that rawinsondes continue to be the backbone of the observational system that provides critical data for model initializations. In fact, Douglas and Stensrud (1996) argue that the resolution of the current rawinsonde network should be increased since these observations currently provide the most accurate depiction of the vertical structure of the atmosphere that is needed for model initial conditions. They also suggest that it would be economically viable to implement additional sites in data-sparse regions of North America including Canada.

A final point of discussion focuses on the minimal forecast impact of the ISS sites in the vicinity of Lake Michigan. While this represents only one case, it is possible that routine additional soundings in the region may not have a significant impact on improving model-simulated lake-effect events. Warner et al. (1989) suggest that in areas of intense local forcing, errors in the model initial condition may not be as critical to the forecast evolution. They state that the resistance formulations used to parameterize surface fluxes of heat, momentum, and moisture can cause the model solution at low levels to adjust toward a more-correct value if the initial conditions are in error. Additionally, differential fluxes can cause the model to develop realistic mesoscale circulations that are not resolved by the initial data. Certainly, the lake-effect environment is locally forced to a great extent, as thermal instability is a necessary ingredient in the generation of lake-effect snow. Of course, the operational meteorologist could benefit from additional soundings within the lake-effect environment by having access to a current high-resolution profile of the convective boundary layer. However, in the context of numerical modeling, it appears that additional soundings may be most beneficial in data-sparse areas downwind of the Great Lakes basin. As an area of future work, an optimization study would be useful in quantifying the number of additional soundings necessary to generate a significant improvement in the model simulation.

6. Conclusions

This study has shown that additional upper-air sounding data acquired over the data-sparse regions of northeastern Ontario and western Quebec had a significant impact on a mesoscale model forecast over Lake Michigan. Numerical simulations showed that the passage of a mesoalpha-scale trough and the attendant lake-effect snow over Lake Michigan were more accurately represented when these sounding data were incorporated into the model initial conditions. This suggests that the synoptic-scale flow is very important to the propagation of mesoscale features that develop within the convective boundary layer and that synoptic considerations cannot be neglected when modeling smaller-scale circulations that occur in a lake-effect environment. Additionally, the impact of the CLASS soundings suggests that the model initial condition in the CNTRL simulation was not able to resolve the warm low-level layer over northeastern Ontario and western Quebec that was likely induced by the aggregate effect of the eastern Great Lakes. Considering that this synoptic pattern occurs approximately 10 times per year, forecasters need to be aware of the paucity of upper-air observations in this region, and the corresponding limitations to the accuracy of mesoscale model output.

Furthering the recommendations of Douglas and Stensrud (1996), this modeling study has illustrated that increasing the density of the current operational rawinsonde network in data-sparse regions of North America can have a positive impact on short-term weather prediction, and that better initialization of synoptic features in Canada can have an impact on mesoscale forecasts in the United States. It appears that additional soundings over Ontario and Quebec may improve “upwind” forecasts by resolving the downwind effects of the Great Lakes aggregate. Considering that the high resolution of current operational models has made quantitative precipitation forecast output a useful tool for forecasting lake-effect precipitation at the county level, this improvement in the model initial condition represents the next step in lake-effect forecasting.

Acknowledgments

The authors wish to thank Greg Mann and Dick Wagenmaker of the Detroit/Pontiac National Weather Service Forecast Office for their suggestions, insight, and resources. Additional thanks to Tim Hunter of the Great Lakes Environmental Research Laboratory for snowfall data, Wei Wang of NCAR for modeling support, and Dave Stauffer and Jim Bresch for helpful correspondence. This research was supported in part by NSF Grant ATM-9502009.

REFERENCES

  • Anthes, R. A., 1977: A cumulus parameterization scheme utilizing a one-dimensional cloud model. Mon. Wea. Rev, 105 , 270286.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ballentine, R. J., Stamm A. J. , Chermack E. E. , Byrd G. P. , and Schleede D. , 1998: Mesoscale model simulation of the 4–5 January 1995 lake-effect snowstorm. Wea. Forecasting, 13 , 893920.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Douglas, M. W., and Stensrud D. J. , 1996: Upgrading the North American upper-air observing network: What are the possibilities? Bull. Amer. Meteor. Soc, 77 , 907924.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Emanuel, K., and Coauthors. 1997: Observations in aid of weather prediction for North America: Report of Prospectus Development Team Seven. Bull. Amer. Meteor. Soc, 78 , 28592868.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Forbes, G. S., and Merritt J. H. , 1984: Mesoscale vortices over the Great Lakes in wintertime. Mon. Wea. Rev, 112 , 377381.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gallus Jr., W. A., and Bresch J. F. , 1997: An intense small-scale wintertime vortex in the midwest United States. Mon. Wea. Rev, 125 , 27872807.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Grell, G. A., Dudhia J. , and Stauffer D. A. , 1995: A description of the fifth generation Penn State/NCAR Mesoscale Model (MM5). NCAR Tech. Note NCAR/TN-398+IA, 107 pp.

    • Search Google Scholar
    • Export Citation
  • Gustafsson, N., Kallen E. , and Thorsteinsson S. , 1998: Sensitivity of forecast errors to initial and lateral boundary conditions. Tellus, 50A , 167185.

    • Search Google Scholar
    • Export Citation
  • Heming, J. T., 1990: The impact of surface and radiosonde observations from two Atlantic ships on a numerical weather prediction model forecast for the storm of 25 January 1990. Meteor. Mag, 119 , 249259.

    • Search Google Scholar
    • Export Citation
  • Hjelmfelt, M. R., 1990: Numerical study of the influence of environmental conditions on lake–effect snowstorms over Lake Michigan. Mon. Wea. Rev, 118 , 138150.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Holroyd, III., E. W., . 1971: Lake-effect cloud bands as seen from satellites. J. Atmos. Sci, 28 , 11651170.

  • Houze, Jr., R. A., . 1993: Cloud Dynamics. Academic Press, 573 pp.

  • Kain, J. S., and Fritsch J. M. , 1993: The role of the convective “trigger function” in numerical prediction of mesoscale convective systems. Meteor. Atmos. Phys, 49 , 93106.

    • Search Google Scholar
    • Export Citation
  • Kelly, G., 1997: Influence of observations on the operational ECMWF system. WMO Bull, 46 , 336342.

  • Kelly, R. D., 1986: Mesoscale frequencies and seasonal snowfalls for different types of Lake Michigan snow storms. J. Climate Appl. Meteor, 25 , 308312.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kristovich, D. A. R., 1993: Mean circulations of boundary-layer rolls in lake-effect snow storms. Bound.-Layer Meteor, 63 , 293315.

  • Kristovich, D. A. R., and Coauthors. . 2000: The Lake-Induced Convection Experiment and the Snowband Dynamics Project. Bull. Amer. Meteor. Soc, 81 , 519542.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kuo, Y-H., and Guo Y-R. , 1989: Dynamic initialization using observations from a hypothetical network of profilers. Mon. Wea. Rev, 117 , 19751998.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Laird, N. F., Miller L. J. , and Kristovich D. A. R. , 2001: Synthetic dual-Doppler analysis of a winter mesoscale vortex. Mon. Wea. Rev, 129 , 312331.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Laroche, S., Gauthier P. , St-James J. , and Morneau J. , 1999: Implementation of a 3d variational data assimilation system at the Canadian Meteorological Centre. Part II: The regional analysis. Atmos.–Ocean, 37 , 281307.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Mittelstadt, J., 1998: The eta-32 model. Western Region Tech. Attachment 98-03, NWS.

  • Niziol, T. A., 1987: Operational forecasting of lake effect snowfall in western and central New York. Wea. Forecasting, 2 , 310321.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Nuss, W. A., and Titley D. W. , 1994: Use of multiquadric interpolation for meteorological objective analysis. Mon. Wea. Rev, 122 , 16111631.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Pouponneau, B., Ayrault F. , Bergot T. , and Joly A. , 1999: The impact of aircraft data on an Atlantic cyclone analyzed in terms of sensitivities and trajectories. Wea. Forecasting, 14 , 6783.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Pu, Z-X., Lord S. J. , and Kalnay E. , 1998: Forecast sensitivity with dropwindsonde data and targeted observations. Tellus, 50A , 391410.

    • Search Google Scholar
    • Export Citation
  • Rabier, F., Klinker E. , Courtier P. , and Hollingsworth A. , 1996: Sensitivity of forecast errors to initial conditions. Quart. J. Roy. Meteor. Soc, 122 , 121150.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Sousounis, P. J., 1997: Lake-aggregate mesoscale disturbances. Part III: Description of a mesoscale aggregate vortex. Mon. Wea. Rev, 125 , 11111134.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Sousounis, P. J., and Fritsch J. M. , 1994: Lake-aggregate mesoscale disturbances. Part II: A case study of the effects on regional and synoptic-scale weather systems. Bull. Amer. Meteor. Soc, 75 , 17931811.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Sousounis, P. J., Mann G. E. , Young G. S. , Wagenmaker R. B. , Hoggatt B. D. , and Badini W. J. , 1999: Forecasting during the Lake-ICE/SNOWBANDS field experiments. Wea. Forecasting, 14 , 955975.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Stauffer, D. R., and Seaman N. L. , 1990: Use of four-dimensional data assimilation in a limited-area mesoscale model. Part I: Experiments with synoptic-scale data. Mon. Wea. Rev, 118 , 12501277.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Stensrud, D. J., and Fritsch J. M. , 1994: Mesoscale convective systems in weakly forced large-scale environments. Part II: Generation of a mesoscale initial condition. Mon. Wea. Rev, 122 , 20682083.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Tao, W-K., and Simpson J. , 1993: Goddard cumulus ensemble model. Part I: model description. Terr. Atmos. Oceanic Sci, 4 , 3572.

  • Sousounis, P. J., and McCumber M. , 1989: An ice-water adjustment scheme. Mon. Wea. Rev, 117 , 231235.

  • Tuleya, R. E., and Lord S. J. , 1997: The impact of dropwindsonde data on GFDL hurricane model forecasts using global analyses. Wea. Forecasting, 12 , 307323.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Warner, T. T., Key L. E. , and Lario A. M. , 1989: Sensitivity of mesoscale-model forecast skill to some initial-data characteristics, data density, data position, analysis procedure and measurement error. Mon. Wea. Rev, 117 , 12811310.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Weiss, C. C., and Sousounis P. J. , 1999: A climatology of collective lake disturbances. Mon. Wea. Rev, 127 , 565574.

  • Zhang, D. L., and Anthes R. A. , 1982: A high resolution model of the planetary layer—Sensitivity tests and comparison with SESAME-79 data. J. Appl. Meteor, 21 , 15941609.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Zheng, Y., Xu Q. , and Stensrud D. J. , 1995: A numerical simulation of the 7 May 1985 mesoscale convective system. Mon. Wea. Rev, 123 , 1781179.

    • Crossref
    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

Locations of upper-air rawinsonde stations during Lake-ICE

Citation: Weather and Forecasting 16, 4; 10.1175/1520-0434(2001)016<0448:TUOASF>2.0.CO;2

Fig. 2.
Fig. 2.

Great Lakes surface analysis for 4–5 Dec 1997. Sea level pressure (hPa), solid; temperature (°C), dashed; wind barb, 10 m s−1; half barb, 5 m s−1; (a) 0000 UTC 4 Dec, (b) 1200 UTC 4 Dec, (c) 0000 UTC 5 Dec, and (d) 1200 UTC 5 Dec. Heavy dashed line indicates location of the particular mesoalpha-scale trough of interest

Citation: Weather and Forecasting 16, 4; 10.1175/1520-0434(2001)016<0448:TUOASF>2.0.CO;2

Fig. 3.
Fig. 3.

The 850-hPa analysis for 4–5 Dec 1997. Height (dam), solid; temperature (°C), dashed; wind barb, 10 m s−1; half barb, 5 m s−1: (a) 0000 UTC 4 Dec, (b) 1200 UTC 4 Dec, (c) 0000 UTC 5 Dec, and (d) 1200 UTC 5 Dec

Citation: Weather and Forecasting 16, 4; 10.1175/1520-0434(2001)016<0448:TUOASF>2.0.CO;2

Fig. 4.
Fig. 4.

The 500-hPa analysis for 4–5 Dec 1997. Height (dam), solid; temperature (°C), dashed; wind flag, 50 m s−1; full barb, 10 m s−1; half barb, 5 m s−1: (a) 0000 UTC 4 Dec, (b) 1200 UTC 4 Dec, (c) 0000 UTC 5 Dec, and (d) 1200 UTC 5 Dec

Citation: Weather and Forecasting 16, 4; 10.1175/1520-0434(2001)016<0448:TUOASF>2.0.CO;2

Fig. 5.
Fig. 5.

Map of observed snowfall totals over southwest Lower Michigan for 5 Dec 1997

Citation: Weather and Forecasting 16, 4; 10.1175/1520-0434(2001)016<0448:TUOASF>2.0.CO;2

Fig. 6.
Fig. 6.

Map depicting model domain and two nested grids. The largest domain has 90-km horizontal grid spacing; next largest, 30-km grid spacing; and smallest, 10-km grid spacing

Citation: Weather and Forecasting 16, 4; 10.1175/1520-0434(2001)016<0448:TUOASF>2.0.CO;2

Fig. 7.
Fig. 7.

(CLASS+ISS)–CNTRL simulation temperature perturbations (0.5°C contour interval) valid at model initialization, 1200 UTC 4 Dec, at (a) 900 hPa. Dots denote CLASS and ISS sites. Thick line denotes location of vertical cross section in (b)

Citation: Weather and Forecasting 16, 4; 10.1175/1520-0434(2001)016<0448:TUOASF>2.0.CO;2

Fig. 8.
Fig. 8.

Timmins, Ontario, Canada (YTS), sounding and model initialization temperature profiles valid 1200 UTC 4 Dec

Citation: Weather and Forecasting 16, 4; 10.1175/1520-0434(2001)016<0448:TUOASF>2.0.CO;2

Fig. 9.
Fig. 9.

Mean sea level pressure (2-hPa contour interval) for CLASS+ISS (solid) and CNTRL (dashed) simulations. The (CLASS+ISS)–CNTRL sea level pressure perturbations (dotted, 0.5 hPa contour interval), where the shaded region denotes sea level pressure perturbations of larger magnitude than −1.0 hPa at (a) 1800 UTC 4 Dec, (b) 0000 UTC 5 Dec, (c) 0600 UTC 5 Dec, and (d) 1200 UTC 5 Dec. Heavy dashed line indicates location of the particular mesoalpha-scale trough of interest

Citation: Weather and Forecasting 16, 4; 10.1175/1520-0434(2001)016<0448:TUOASF>2.0.CO;2

Fig. 10.
Fig. 10.

The (CLASS+ISS)-CNTRL 950-hPa perturbation winds (reference vector shown), perturbation heights of larger magnitude than −6 m (shaded region), and CNTRL simulation 950-hPa temperatures (dashed, contour interval 1°C) at (a) 1800 UTC 4 Dec, (b) 0000 UTC 5 Dec, (c) 0600 UTC 5 Dec, and (d) 1200 UTC 5 Dec. Boxed area in (c) denotes approximate area shown in Fig. 11

Citation: Weather and Forecasting 16, 4; 10.1175/1520-0434(2001)016<0448:TUOASF>2.0.CO;2

Fig. 11.
Fig. 11.

(a) Radar reflectivity from Grand Rapids, MI (KGRR), 0.5° elevation scan valid 0753 UTC 5 Dec. (b) Model-simulated 1-h quantitative precipitation forecast (0.25-mm contour interval) valid at 0600 UTC 5 Dec: CLASS+ISS, solid; CNTRL, dashed. (c) Model-simulated 950-hPa winds valid at 0600 UTC 5 Dec: full barb, 5 m s−1, and half barb, 2.5 m s−1; CLASS+ISS, bold black; CNTRL, gray

Citation: Weather and Forecasting 16, 4; 10.1175/1520-0434(2001)016<0448:TUOASF>2.0.CO;2

Fig. 12.
Fig. 12.

(a) Radar reflectivity from KGRR 0.5° elevation scan valid 1053 UTC 5 Dec. (b)–(d) Model-simulated 1-h quantitative precipitation forecast (0.25-mm contour interval) valid 0900 UTC 5 Dec: (b) CLASS+ISS (solid), CNTRL (dashed); (c) CLASS+ISS (solid), CLASSONLY (dash–dot); (d) CNTRL (dashed), ISSONLY (dotted)

Citation: Weather and Forecasting 16, 4; 10.1175/1520-0434(2001)016<0448:TUOASF>2.0.CO;2

Fig. 13.
Fig. 13.

(a) Radar radial velocity from KGRR 0.5° elevation scan valid 1153 UTC 5 Dec. Circle indicates the radial velocity couplet denoted by inbound (negative) and outbound (positive) radial velocity values (m s−1) associated with the lake-effect mesoscale vortex. (b) Model-simulated 950-hPa absolute vorticity values (scaled by 10−5 s−1, 10−4 s−1 contour interval) valid at 1000 UTC 5 Dec; CLASS+ISS, solid; CNTRL, dashed

Citation: Weather and Forecasting 16, 4; 10.1175/1520-0434(2001)016<0448:TUOASF>2.0.CO;2

Fig. 14.
Fig. 14.

Track of 950-hPa vorticity maximum as simulated by CLASS+ISS and CNTRL for the time period shown and track of mesoscale vortex as indicated by Grand Rapids WSR-88D (0.5° radial velocity) for the time period 0953–1253 UTC. Symbols denote hourly positions

Citation: Weather and Forecasting 16, 4; 10.1175/1520-0434(2001)016<0448:TUOASF>2.0.CO;2

Table 1.

Lake-ICE rawinsonde data used in model initial condition valid 1200 UTC 4 Dec. [ol0][pp07-t01]

Table 1.
Save
  • Anthes, R. A., 1977: A cumulus parameterization scheme utilizing a one-dimensional cloud model. Mon. Wea. Rev, 105 , 270286.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ballentine, R. J., Stamm A. J. , Chermack E. E. , Byrd G. P. , and Schleede D. , 1998: Mesoscale model simulation of the 4–5 January 1995 lake-effect snowstorm. Wea. Forecasting, 13 , 893920.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Douglas, M. W., and Stensrud D. J. , 1996: Upgrading the North American upper-air observing network: What are the possibilities? Bull. Amer. Meteor. Soc, 77 , 907924.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Emanuel, K., and Coauthors. 1997: Observations in aid of weather prediction for North America: Report of Prospectus Development Team Seven. Bull. Amer. Meteor. Soc, 78 , 28592868.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Forbes, G. S., and Merritt J. H. , 1984: Mesoscale vortices over the Great Lakes in wintertime. Mon. Wea. Rev, 112 , 377381.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gallus Jr., W. A., and Bresch J. F. , 1997: An intense small-scale wintertime vortex in the midwest United States. Mon. Wea. Rev, 125 , 27872807.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Grell, G. A., Dudhia J. , and Stauffer D. A. , 1995: A description of the fifth generation Penn State/NCAR Mesoscale Model (MM5). NCAR Tech. Note NCAR/TN-398+IA, 107 pp.

    • Search Google Scholar
    • Export Citation
  • Gustafsson, N., Kallen E. , and Thorsteinsson S. , 1998: Sensitivity of forecast errors to initial and lateral boundary conditions. Tellus, 50A , 167185.

    • Search Google Scholar
    • Export Citation
  • Heming, J. T., 1990: The impact of surface and radiosonde observations from two Atlantic ships on a numerical weather prediction model forecast for the storm of 25 January 1990. Meteor. Mag, 119 , 249259.

    • Search Google Scholar
    • Export Citation
  • Hjelmfelt, M. R., 1990: Numerical study of the influence of environmental conditions on lake–effect snowstorms over Lake Michigan. Mon. Wea. Rev, 118 , 138150.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Holroyd, III., E. W., . 1971: Lake-effect cloud bands as seen from satellites. J. Atmos. Sci, 28 , 11651170.

  • Houze, Jr., R. A., . 1993: Cloud Dynamics. Academic Press, 573 pp.

  • Kain, J. S., and Fritsch J. M. , 1993: The role of the convective “trigger function” in numerical prediction of mesoscale convective systems. Meteor. Atmos. Phys, 49 , 93106.

    • Search Google Scholar
    • Export Citation
  • Kelly, G., 1997: Influence of observations on the operational ECMWF system. WMO Bull, 46 , 336342.

  • Kelly, R. D., 1986: Mesoscale frequencies and seasonal snowfalls for different types of Lake Michigan snow storms. J. Climate Appl. Meteor, 25 , 308312.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kristovich, D. A. R., 1993: Mean circulations of boundary-layer rolls in lake-effect snow storms. Bound.-Layer Meteor, 63 , 293315.

  • Kristovich, D. A. R., and Coauthors. . 2000: The Lake-Induced Convection Experiment and the Snowband Dynamics Project. Bull. Amer. Meteor. Soc, 81 , 519542.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kuo, Y-H., and Guo Y-R. , 1989: Dynamic initialization using observations from a hypothetical network of profilers. Mon. Wea. Rev, 117 , 19751998.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Laird, N. F., Miller L. J. , and Kristovich D. A. R. , 2001: Synthetic dual-Doppler analysis of a winter mesoscale vortex. Mon. Wea. Rev, 129 , 312331.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Laroche, S., Gauthier P. , St-James J. , and Morneau J. , 1999: Implementation of a 3d variational data assimilation system at the Canadian Meteorological Centre. Part II: The regional analysis. Atmos.–Ocean, 37 , 281307.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Mittelstadt, J., 1998: The eta-32 model. Western Region Tech. Attachment 98-03, NWS.

  • Niziol, T. A., 1987: Operational forecasting of lake effect snowfall in western and central New York. Wea. Forecasting, 2 , 310321.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Nuss, W. A., and Titley D. W. , 1994: Use of multiquadric interpolation for meteorological objective analysis. Mon. Wea. Rev, 122 , 16111631.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Pouponneau, B., Ayrault F. , Bergot T. , and Joly A. , 1999: The impact of aircraft data on an Atlantic cyclone analyzed in terms of sensitivities and trajectories. Wea. Forecasting, 14 , 6783.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Pu, Z-X., Lord S. J. , and Kalnay E. , 1998: Forecast sensitivity with dropwindsonde data and targeted observations. Tellus, 50A , 391410.

    • Search Google Scholar
    • Export Citation
  • Rabier, F., Klinker E. , Courtier P. , and Hollingsworth A. , 1996: Sensitivity of forecast errors to initial conditions. Quart. J. Roy. Meteor. Soc, 122 , 121150.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Sousounis, P. J., 1997: Lake-aggregate mesoscale disturbances. Part III: Description of a mesoscale aggregate vortex. Mon. Wea. Rev, 125 , 11111134.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Sousounis, P. J., and Fritsch J. M. , 1994: Lake-aggregate mesoscale disturbances. Part II: A case study of the effects on regional and synoptic-scale weather systems. Bull. Amer. Meteor. Soc, 75 , 17931811.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Sousounis, P. J., Mann G. E. , Young G. S. , Wagenmaker R. B. , Hoggatt B. D. , and Badini W. J. , 1999: Forecasting during the Lake-ICE/SNOWBANDS field experiments. Wea. Forecasting, 14 , 955975.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Stauffer, D. R., and Seaman N. L. , 1990: Use of four-dimensional data assimilation in a limited-area mesoscale model. Part I: Experiments with synoptic-scale data. Mon. Wea. Rev, 118 , 12501277.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Stensrud, D. J., and Fritsch J. M. , 1994: Mesoscale convective systems in weakly forced large-scale environments. Part II: Generation of a mesoscale initial condition. Mon. Wea. Rev, 122 , 20682083.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Tao, W-K., and Simpson J. , 1993: Goddard cumulus ensemble model. Part I: model description. Terr. Atmos. Oceanic Sci, 4 , 3572.

  • Sousounis, P. J., and McCumber M. , 1989: An ice-water adjustment scheme. Mon. Wea. Rev, 117 , 231235.

  • Tuleya, R. E., and Lord S. J. , 1997: The impact of dropwindsonde data on GFDL hurricane model forecasts using global analyses. Wea. Forecasting, 12 , 307323.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Warner, T. T., Key L. E. , and Lario A. M. , 1989: Sensitivity of mesoscale-model forecast skill to some initial-data characteristics, data density, data position, analysis procedure and measurement error. Mon. Wea. Rev, 117 , 12811310.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Weiss, C. C., and Sousounis P. J. , 1999: A climatology of collective lake disturbances. Mon. Wea. Rev, 127 , 565574.

  • Zhang, D. L., and Anthes R. A. , 1982: A high resolution model of the planetary layer—Sensitivity tests and comparison with SESAME-79 data. J. Appl. Meteor, 21 , 15941609.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Zheng, Y., Xu Q. , and Stensrud D. J. , 1995: A numerical simulation of the 7 May 1985 mesoscale convective system. Mon. Wea. Rev, 123 , 1781179.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Fig. 1.

    Locations of upper-air rawinsonde stations during Lake-ICE

  • Fig. 2.

    Great Lakes surface analysis for 4–5 Dec 1997. Sea level pressure (hPa), solid; temperature (°C), dashed; wind barb, 10 m s−1; half barb, 5 m s−1; (a) 0000 UTC 4 Dec, (b) 1200 UTC 4 Dec, (c) 0000 UTC 5 Dec, and (d) 1200 UTC 5 Dec. Heavy dashed line indicates location of the particular mesoalpha-scale trough of interest

  • Fig. 3.

    The 850-hPa analysis for 4–5 Dec 1997. Height (dam), solid; temperature (°C), dashed; wind barb, 10 m s−1; half barb, 5 m s−1: (a) 0000 UTC 4 Dec, (b) 1200 UTC 4 Dec, (c) 0000 UTC 5 Dec, and (d) 1200 UTC 5 Dec

  • Fig. 4.

    The 500-hPa analysis for 4–5 Dec 1997. Height (dam), solid; temperature (°C), dashed; wind flag, 50 m s−1; full barb, 10 m s−1; half barb, 5 m s−1: (a) 0000 UTC 4 Dec, (b) 1200 UTC 4 Dec, (c) 0000 UTC 5 Dec, and (d) 1200 UTC 5 Dec

  • Fig. 5.

    Map of observed snowfall totals over southwest Lower Michigan for 5 Dec 1997

  • Fig. 6.

    Map depicting model domain and two nested grids. The largest domain has 90-km horizontal grid spacing; next largest, 30-km grid spacing; and smallest, 10-km grid spacing

  • Fig. 7.

    (CLASS+ISS)–CNTRL simulation temperature perturbations (0.5°C contour interval) valid at model initialization, 1200 UTC 4 Dec, at (a) 900 hPa. Dots denote CLASS and ISS sites. Thick line denotes location of vertical cross section in (b)

  • Fig. 8.

    Timmins, Ontario, Canada (YTS), sounding and model initialization temperature profiles valid 1200 UTC 4 Dec

  • Fig. 9.

    Mean sea level pressure (2-hPa contour interval) for CLASS+ISS (solid) and CNTRL (dashed) simulations. The (CLASS+ISS)–CNTRL sea level pressure perturbations (dotted, 0.5 hPa contour interval), where the shaded region denotes sea level pressure perturbations of larger magnitude than −1.0 hPa at (a) 1800 UTC 4 Dec, (b) 0000 UTC 5 Dec, (c) 0600 UTC 5 Dec, and (d) 1200 UTC 5 Dec. Heavy dashed line indicates location of the particular mesoalpha-scale trough of interest

  • Fig. 10.

    The (CLASS+ISS)-CNTRL 950-hPa perturbation winds (reference vector shown), perturbation heights of larger magnitude than −6 m (shaded region), and CNTRL simulation 950-hPa temperatures (dashed, contour interval 1°C) at (a) 1800 UTC 4 Dec, (b) 0000 UTC 5 Dec, (c) 0600 UTC 5 Dec, and (d) 1200 UTC 5 Dec. Boxed area in (c) denotes approximate area shown in Fig. 11

  • Fig. 11.

    (a) Radar reflectivity from Grand Rapids, MI (KGRR), 0.5° elevation scan valid 0753 UTC 5 Dec. (b) Model-simulated 1-h quantitative precipitation forecast (0.25-mm contour interval) valid at 0600 UTC 5 Dec: CLASS+ISS, solid; CNTRL, dashed. (c) Model-simulated 950-hPa winds valid at 0600 UTC 5 Dec: full barb, 5 m s−1, and half barb, 2.5 m s−1; CLASS+ISS, bold black; CNTRL, gray

  • Fig. 12.

    (a) Radar reflectivity from KGRR 0.5° elevation scan valid 1053 UTC 5 Dec. (b)–(d) Model-simulated 1-h quantitative precipitation forecast (0.25-mm contour interval) valid 0900 UTC 5 Dec: (b) CLASS+ISS (solid), CNTRL (dashed); (c) CLASS+ISS (solid), CLASSONLY (dash–dot); (d) CNTRL (dashed), ISSONLY (dotted)

  • Fig. 13.

    (a) Radar radial velocity from KGRR 0.5° elevation scan valid 1153 UTC 5 Dec. Circle indicates the radial velocity couplet denoted by inbound (negative) and outbound (positive) radial velocity values (m s−1) associated with the lake-effect mesoscale vortex. (b) Model-simulated 950-hPa absolute vorticity values (scaled by 10−5 s−1, 10−4 s−1 contour interval) valid at 1000 UTC 5 Dec; CLASS+ISS, solid; CNTRL, dashed

  • Fig. 14.

    Track of 950-hPa vorticity maximum as simulated by CLASS+ISS and CNTRL for the time period shown and track of mesoscale vortex as indicated by Grand Rapids WSR-88D (0.5° radial velocity) for the time period 0953–1253 UTC. Symbols denote hourly positions

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 227 34 5
PDF Downloads 92 31 6