1. Introduction

When cold polar air flows across the relatively warm Great Lakes in winter, strong sensible and latent heat fluxes are generated. These fluxes create a favorable environment for the development of storms, mainly by destabilizing and moistening the lower atmosphere. The Great Lakes can influence the weather on a variety of scales. Forcing from individual lakes generates individual lake-scale lake-effect storms (LESs) that affect the local weather downwind of each lake. The combined forcing from all the Great Lakes can influence the regional weather by inducing or altering synoptic-scale disturbances.

Studies regarding aggregate effects of the Great Lakes have been few. Cox (1917) observed the frequent occurrence of synoptic-scale lows in the Great Lakes region during winter. He suggested that the cooperative heating from the Great lakes “attracted” the synoptic-scale cyclones to the Great Lakes region and intensified them. Petterssen and Calabrese (1959) noted that the collective impacts of the Great Lakes aggregate may generate a 6–7 hPa sea level pressure (SLP) perturbation that is manifested either as a trough or a closed low depending on the speed of the prevailing flow. Danard and Rao (1972) and Boudra (1981) performed numerical simulations with the Great Lakes (WL) and without the Great Lakes (NL) on strong winter cyclones. By comparing WL and NL simulations, these investigators found that the Great Lakes reduced surface pressure by 5 hPa, increased 850-hPa temperature by 5°C, and increased the large-scale precipitation by 3 mm after a 36-h simulation in one case. However, the path of the cyclone was not changed much by the Great Lakes. Sousounis and Fritsch (1994) conducted higher-resolution but otherwise similar WL and NL numerical simulations on a weak synoptic-scale low in late autumn. They found that the Great Lakes not only increased the surface pressure deepening rate but also significantly changed the path of the cyclone. Additionally, they identified a meso-α-scale (200–2000 km; Orlanski 1975) perturbation circulation by subtracting the WL wind fields from the NL ones.

Sousounis (1997) defined a mesoscale aggregate vortex (MAV) after studying the kinematic structure of the aforementioned meso-α-scale perturbation circulation. He defined a MAV as a perturbation vortex that is inertially stable, meso-α-scale, and has a warm core that is ∼500–1000-km wide and 2–4 km deep and that develops from aggregate heating and moistening by the Great Lakes. The MAV usually exhibits cyclonic circulation on an aggregate scale below 800 hPa and anticyclonic circulation from 800 to 600 hPa. An aggregate-scale circulation was defined as a closed circulation with size similar to that of the lake aggregate and with intensity stronger than those at the individual lake scale. The MAV is identifiable on standard surface weather charts as a low with one to two closed SLP contours at 4-hPa interval and a meso-α-scale thermal ridge over the lake region.

Sousounis (1998) investigated the mechanisms during MAV development by attributing the 1000-hPa height tendency to different physical processes, including horizontal temperature advection, adiabatic heating or cooling, and diabatic heating. He concluded that the strong sensible heating that resulted from colder northwesterly winds at an early stage (e.g., preconditioning) and the strong warm advection that resulted from southwesterly winds at a later stage were important for MAV development. These two mechanisms, strong sensible heating and warm advection, contributed to a favorable environment for development of the mesoscale disturbance that later moved into the region by increasing low-level temperatures, reducing the low-level static stability, and consequently enhancing height falls.

Many aspects of the development of lake-aggregate effects and MAVs are still not well understood. For example, it is not known if the Great Lakes aggregate can produce a synoptic-scale cyclone or a MAV when a preexisting disturbance does not already exist. If a synoptic-scale disturbance or MAV can not develop simply from thermal forcing by the Great Lakes, then it would mean that the Great Lakes can only enhance development of a preexisting synoptic-scale or meso-α-scale disturbance, respectively. Furthermore, because there have been only a few case studies regarding the impacts of the lake aggregate on synoptic-scale systems, conclusions drawn from these studies may or may not apply to other aggregate-scale lake-effect storms' events.

The objectives for this study are 1) to systematically intercompare how different flow regimes interact with a group of relatively warm lakes, 2) to examine quantitatively the contributions of the lake aggregate on various fields, such as SLP, temperature, and precipitation, and 3) to identify physical processes that are important for the intensification of the lake-aggregate–enhanced cyclones.

2. Methodology

A series of six numerical experiments (cf. Table 1) was performed using the fifth-generation Pennsylvania State University–National Center for Atmospheric Research (PSU–NCAR) Mesoscale Model (MM5). Three different idealized flow patterns were used as initial conditions: 1) zonal flow, 2) solitary trough, and 3) continuous sinusoidal waves. With-lake (WL) and no-lake (NL) simulations were performed for each flow pattern. The purpose of the zonal flow experiment was to examine whether a synoptic-scale disturbance would develop from heating and moistening by a group of meso-β-scale lakes when there are no preexisting disturbances. The solitary trough experiment was performed to see whether stronger synoptic-scale development occurs when an isolated vorticity maximum is present in the initial conditions. The purpose of the continuous sinusoidal waves experiment, which serves as a relatively more realistic example of weather conditions, was to study the importance of preconditioning from a ridge that precedes a trough and possibly from the interaction between the ridge and the lakes. The corresponding NL experiments were performed to understand how the atmosphere evolves without the lakes. The initial conditions will be described in more detail in the next section.

Two domains were used for the simulations. The outer (coarse) mesh (CGM) had 50 × 37 horizontal grid points and 24 vertical levels with 90-km horizontal grid spacing. The inner (fine) mesh (FGM) had 61 × 61 horizontal grid points and 24 vertical levels with 30-km horizontal grid spacing. The coarse mesh was used mainly to ensure that the preexisting disturbance of choice was fed properly into the fine mesh through its upwind boundary. The coarse mesh was centered at 43.5°N 87.9°W. However, the geographical locations of the domains were not important, except for determining values of the Coriolis parameter at each grid point, because idealized terrain and land use were specified. Flat terrain and almost-uniform land use were adopted for all the simulations. The terrain height was set to 1 m at each grid point to eliminate topographic effects and to facilitate the study of dynamical and thermodynamical forcing. Because the forest land was found to be the most representative category in the Great Lakes region in the real terrain simulation, the land use was assumed to be forest land for each grid point except where there were lakes.

In the WL simulations, four identical circular lakes with diameters of 268 km and separation distances of 360 km were placed in the upwind half of the fine mesh domain. The diameter was specified to make the total area of the four idealized lakes equal to that of all the Great Lakes. A schematic of the domains and the lake-aggregate configuration is shown in Fig. 1. The lake temperatures were specified to be 2°C for the “upper” lakes and 8°C for the “lower” lakes, which are typical lake temperatures for the upper (e.g., Lakes Superior, Michigan, and Huron) and lower Great Lakes (Lakes Erie and Ontario) in November (Eichenlaub 1979). In the NL simulations, the lakes were replaced with surrounding land use values. The ground temperatures for the land grid points were specified by extrapolating from the temperatures at the lowest two model layers. All the simulations were run to 60 h.

In the current study, the more realistic mixed-phase explicit moisture scheme was chosen to parameterize the hydrological cycle. The shallow convection option was used in all the simulations to help transport moisture fluxes generated by the lakes from boundary layer to free atmosphere. The Kuo scheme was chosen as the cumulus parameterization scheme because it was an established cumulus scheme appropriate for 30-km resolution at the time. Most cumulus schemes performed reasonably well in the winter situation when the atmosphere is stable. As indicated in Wang and Seaman's study (1997), the precipitation-prediction skills of the Kuo scheme were almost as good as those of the Kain–Fritsch scheme for the three cold-season cases they examined. Because sensible and latent heating from the lakes are fundamentally important processes for the development of cyclones passing over the lakes in winter, the Blackadar PBL scheme was used. More details of the MM5 modeling system are available in Anthes et al. (1987) and Grell et al. (1994).

The specified idealized initial conditions (ICs) and boundary conditions (BCs) allowed simple scenarios to be created, which included only one or two physical processes, so that the sensitivity of the atmosphere to each process can be studied more easily. The idealized ICs and BCs were defined to mimic atmospheric conditions found typically in winter over the Great Lakes. The generation of idealized initial and boundary conditions included the following eight steps (Details of this idealized initialization technique are described in Chuang and Sousounis 2000):

• step 1—specify idealized terrain height, land use, and model domain characteristics,

• step 2—specify temperature and humidity profile at a chosen grid point,

• step 3—generate the height fields using an analytic formulation,

• step 4—specify the velocity fields using geostrophic relationship,

• step 5—specify the relative humidity fields,

• step 6—construct the virtual temperature fields using hydrostatic relationship,

• step 7—generate the surface pressure fields, and

• step 8—specify the idealized ground and lake temperature fields.

Note that the relative humidity at every grid point is specified to be the same as that at the sounding location on each isobaric level. In addition to performing the simulations, the extended Zwack–Okossi (ZO) equation (Vasilj and Smith 1997) was applied to analyze the effects of vorticity advection, temperature advection, diabatic heating, and adiabatic heating on surface pressure tendency. This technique was used because it maintains identification of the forcing as a set of distinct forcing processes while at the same time accounting for those processes at essentially all atmospheric levels by vertically integrating through the troposphere and lower stratosphere. The simplified form (Vasilj and Smith 1997) of the extended ZO equation—after neglecting terms that are generally small, such as ageostrophic vorticity tendency, vertical vorticity advection, tilting effect, and divergence effect—may be expressed as follows:

View Expanded

where ξ_gl is the geostrophic vorticity at 950 hPa, Pd = (p_l − p_t)⁻¹, p_t is a sufficiently high upper level, p_l is the near-surface pressure level, ξ_a is the absolute vorticity, F is the frictional force, Q̇ is the diabatic heating, S = −(T/θ)(∂θ/∂p), and all other variables have their usual meaning. The processes represented by each of the terms on the right-hand side (rhs) of (1) are vorticity advection; frictional forcing; and thermal forcing that includes temperature advection, diabatic heating, and adiabatic heating. After all the terms on the rhs of (1) are computed, the surface pressure tendency attributable to each process can be obtained by solving the equation:

View Expanded

Equations (1) and (2) were applied to the model output from the series of experiments to determine which physical processes are responsible for development of the cyclones in both WL and NL simulations. The only diabatic heating term that was computed explicitly in this study was the sensible heating Q̇_s. The contribution from all the other diabatic heating sources Q̇_l, which include solar and terrestrial radiation, stable latent heat release, convective latent heat release, and evaporative cooling, were computed as a lump-residual from the temperature tendency equation:

View Expanded

The temperature tendency at τ h was computed using model output at τ − 2 h and τ + 2 h. The horizontal temperature advection, adiabatic heating, and sensible heating terms at τ h were obtained as temporal means of τ − 2 h, τ h, and τ + 2 h using the trapezoidal rule. The sensible heating Q̇_s was calculated by assuming that surface sensible heat fluxes f decrease linearly to zero at 850 hPa:

View Expanded

where ρ, c_p, and z have been defined previously.

The forcing terms were integrated from 200 hPa (p_t) to 950 hPa (p_l) at a 50-hPa interval using the trapezoidal rule. The frictional force plus the terms that were neglected in (1) were not computed explicitly but instead were treated as residuals. Equation (2) was solved using the relaxation technique for each of the forcing terms.

3. Numerical simulations

Initial conditions and boundary conditions were generated at 12-h intervals using the technique described in Chuang and Sousounis (2000). The CGM initial sea level pressure and 1000–500-hPa thickness fields for the zonal flow case [experiment 1 (expt 1)] are shown in Fig. 2 in Chuang and Sousounis (2000). The solitary-trough (expt 2) and continuous sinusoidal waves (expt 3) cases had the same wavelength and wave amplitude except that the wave amplitude in the solitary-trough case was specified to be nonzero in the vicinity of the trough and zero everywhere else. The initial SLP and 1000–500-hPa thickness fields for expts 2 and 3 are shown every 12 h from 0 to 24 h in Fig. 1 of this paper and in Fig. 4 in Chuang and Sousounis (2000), respectively. The surface low in the solitary-trough scenario and the featured surface low (indicated by dashed line) in the continuous waves scenario were specified to enter the western boundary of the FGM and reach optimal vertical tilt with their corresponding 500-hPa troughs at 36 h.

Figure 2 shows the simulated SLP and 12-h precipitation, 1000-hPa temperature and wind fields, and 700-hPa height and wind fields for expt 1 at 48 h (left columns) and 60 h (right columns). At 48 h, a weak surface trough was moving out of the northeastern boundary of the FGM. This surface trough deepened by 2 hPa from 36 h (not shown) to 48 h due to warm advection. The 12-h precipitation amounts and patterns remained very similar from 36 h (not shown) to 60 h. The 1000-hPa wind fields showed zonally elongated two-lake-scale cyclonic circulations over the northern lakes both at 48 and 60 h. The 700-hPa height fields do not appear to have been disturbed much by the lake aggregate throughout the entire simulation. However, the lake-induced impacts may not be apparent until the WL simulation results are compared to the NL ones in the next section.

The simulation results from expt 1 indicate that, when there is no preexisting synoptic-scale disturbance, the forcing from the lake aggregate alone only generated a very weak aggregate-scale surface trough at a later time. Additionally, this weak trough was not developed enough to generate any aggregate-scale precipitation during the 60-h simulation period. Most of the precipitation was located downwind from or over each lake and was the result of individual lake-scale lake-effect storms.

Figure 3 shows simulation results for expt 2 (solitary-trough scenario) at 48 and 60 h. At 48 h, the presence of the synoptic-scale trough in the vicinity of the lakes resulted in heavier and more widespread individual lake-scale precipitation across the southern lakes and more intensified individual lake-scale lows compared to expt 1 at the same time. These features developed because warm advection ahead of the synoptic-scale trough contributed to surface pressure deepening and stronger large-scale ascent, which then enhanced the precipitation. By 60 h, the synoptic-scale trough had intensified slightly. The 1000-hPa horizontal temperature gradient over the lakes became larger and formed a warm front¹ over the northern lakes at this time. There was a weak cold front (e.g., weakly enhanced horizontal temperature gradient) south of the lakes. The combination of dynamical forcing from a developing front and thermal forcing from the lakes brought precipitation to a much larger area compared to expt 1. The 700-hPa wind and height fields showed very little impacts from the lakes throughout the entire 60-h period. Positive vorticity advection (PVA) at 700 hPa over the surface trough at 48 h provided a mechanism for enhancing low-level ascent and surface pressure deepening.

The simulation results from expt 2 show that, with a weak preexisting solitary trough, a weak aggregate-scale surface low with one closed isobar at 2-hPa interval did develop at 60 h after passing the lakes. The signatures of the individual lake-scale impacts were still strong in both the SLP and 1000-hPa wind fields. The existence of the preexisting trough enhanced the individual lake-scale precipitation at 48 h and resulted in large-scale precipitation by 60 h.

Figure 4 shows simulation results for expt 3 (continuous sinusoidal wave scenario) at 48 and 60 h. In this scenario, the lake aggregate was under the influence of a synoptic-scale trough at 0 h, then a synoptic-scale ridge at 24 h, and then the featured synoptic-scale trough at 48 h. By 48 h, the featured trough had developed into a closed low over the lake region with a minimum pressure of ∼997 hPa and four closed isobars at 2-hPa interval, reflecting a SLP drop of about 6 hPa over the last 12 h. The synoptic forcing dominated over the lake forcing because there was very little signal at the individual lake scale in the mass and momentum fields. It is not too surprising that the forcing of the lakes was not as strong at this time due to the fact that the air over the lakes was relatively warm from the warm advection that had been occurring during the past 12 h. The large-scale precipitation had a 12-h maximum of over 9 mm southwest of the surface low center. A surface warm front was evident from the strong horizontal temperature gradient along the northern lakes. A surface cold front was also evident farther to the south and west. At 60 h, the surface warm front weakened after moving out of the lake-aggregate region. The cold advection over the lakes behind the cold front enhanced the surface fluxes from the lakes and maintained the individual lake effects. The precipitation at 60 h was mainly located along the fronts. Similar to expts 1 and 2, the 700-hPa height and wind fields for expt 3 do not show obvious impacts of the lakes.

The simulation results from expt 3 indicate that a relatively strong synoptic-scale cyclone developed when continuous sinusoidal waves were used as ICs. Stronger surface fronts were generated in expt 3 than in expt 2. The continuous sinusoidal waves scenario also allowed stronger upper-level PVA, which supported surface development. Additionally, as will be shown in section 5, the preceding ridge and its interactions with the lakes (preconditioning) in expt 3 allowed the featured trough to become more developed in terms of scale and intensity.

4. Comparison of WL and NL simulations

Figure 5 shows the SLP and 12-h precipitation fields, 1000-hPa temperature and wind fields, and 700-hPa height and wind fields for the NL zonal flow simulation (expt 4) at 48 and 60 h. As expected, the NL simulation for the zonal flow scenario was relatively uneventful because there was no dynamical or lake-aggregate forcing. Figure 6 shows that, compared to 36 h (not shown), the WL–NL perturbation surface low downwind from the lakes intensified and moved farther east at 48 h. By 60 h, the eastward-moving perturbation low had left the FGM domain. At 48 h, a 700-hPa perturbation low was developing northwest of the surface low that was moving out of the FGM. This upper-level perturbation low likely developed corresponding to the development at the surface. The WL–NL differences for the zonal flow scenario suggest that weak perturbation vortices with characteristics similar to those of MAVs were generated, even though there was no dynamical forcing from a preexisting disturbance.

Figure 7 shows simulation results for expt 5 (NL solitary-trough scenario) at 48 and 60 h. The simulation results show that a weak synoptic-scale low with one closed isobar formed at 48 h over the southwestern lake. By 60 h, the low had deepened by 2 hPa and was propagating eastward past the southeastern lake. A comparison between Figs. 3 and 7 indicates that the existence of the lakes did not change the path of the surface low but did strengthen the 1000-hPa temperature gradient and change precipitation patterns associated with the warm front. Compared with the zonal flow scenario, the solitary-trough scenario generated a more intense surface perturbation low over the northeastern lake at 60 h (cf. Fig. 8). The 700-hPa trough in the NL simulation was as weak as in the WL simulation. The 700-hPa height perturbations were very similar to those for the zonal flow scenario at 48 and 60 h. A MAV, that was stronger than the one in the zonal flow scenario, was also generated in the solitary-trough scenario.

Comparison of WL and NL SLP fields for the continuous sinusoidal wave scenario (e.g., Figs. 4 and 9) shows that a weaker and smaller-scale synoptic-scale cyclone still developed without the lakes when continuous sinusoidal waves were used as ICs but with a weaker warm front. The reason why a stronger warm front was generated in expt 3 than in expt 6 was because aggregate heating in the WL simulation provided diabatic heating to only the warm side of the warm front, which increased the temperature gradient and strengthened the warm front. The differential diabatic heating that strengthened the WL warm front also enhanced the northwestern part of its cold front as shown by comparing the WL and NL 1000-hPa temperature fields. However, a comparison of the WL and NL SLP and 1000-hPa wind fields show that the northern part of the NL cold front was stronger because of its sharper wind shift. The more gradual wind shift in the WL simulation developed because flow over the southern lakes was deflected to the right due to a decrease in surface roughness. From a perturbation perspective of the continuous sinusoidal waves scenario (cf. Fig. 10), the lake aggregate generated a MAV centered over the eastern lakes and enhanced precipitation along the existing synoptic-scale warm front (cf. Figs. 4 and 9). The reduced precipitation along the northern part of the cold front was likely related to the more gradual wind shift in the WL scenario, which reduced the convergence and ascent along the front.

5. Diagnosing surface pressure tendency using the extended ZO equation

Comparing the surface pressure tendencies from expts 2 and 3 will help explain why a much more intensified synoptic-scale cyclone developed when using continuous sinusoidal waves as ICs as opposed to a solitary trough. Comparing the surface pressure tendencies for expts 3 and 6 will help explain which physical processes the lake aggregate altered to generate a stronger synoptic-scale cyclone.

6. Summary and conclusions

A series of idealized numerical simulations was conducted to examine the impacts of an idealized array of meso-β-scale circular lakes on synoptic features in cold air situations. Three with-lake scenarios were examined: expt 1—zonal flow; expt 2—solitary trough; and expt 3—continuous sinusoidal waves. Corresponding no-lake simulations were also performed (expts 4–6). The fact that only individual lake-scale lows were generated in expt 1 suggests that the lake aggregate alone was not enough to force the development of a synoptic-scale cyclone when no dynamical forcing was present. Although both expts 2 and 3 generated synoptic-scale closed lows (e.g., synoptic-scale cyclones), the cyclone in expt 2 was much weaker, smaller in terms of scale, and generated much less precipitation compared to that in expt 3. The results of expt 2 suggest further that the solitary-trough scenario, with a combination of upper-level support and lake-aggregate forcing, still does not intensify much. However, compared to expt 1, because of enhanced convergence, warm advection, and ascent owing to the solitary trough, more precipitation was produced. The results of expt 3 indicated that a strong synoptic-scale cyclone developed when a preexisting trough was preceded by a synoptic-scale ridge. The preconditioning from the preceding ridge was concluded to be very important for further development of the synoptic-scale cyclone in the vicinity of the lake aggregate. This conclusion is also supported by the ZO analysis results in section 5a.

The simulation results of expts 5 and 6 show that, compared with their corresponding WL simulations (expts 2 and 3), weaker synoptic-scale closed lows still developed without the lakes. The existence of weak synoptic-scale closed lows in expts 5 and 6 suggests that, for these two flow regimes, the lake aggregate did not induce the formation of the synoptic-scale low but rather enhanced further development. While a comparison between the WL and NL simulations for the solitary-trough scenario showed that the lake aggregate induced the warm front, the corresponding comparison for the continuous sinusoidal waves scenario indicated that the lake aggregate only enhanced the warm front. In either scenario, the lake aggregate induced or enhanced the warm fronts via differential diabatic heating. Sensible heating from the lake aggregate heated the warm sides of the warm fronts and hence increased the temperature gradients across the fronts. The lake aggregate also enhanced the northwestern part of the cold fronts in expt 3. However, surface frictional decreases over the lakes also deflected the flow to the right and contributed to smaller wind shifts along the northern part of the cold front.

The evolution of the MAV for all three scenarios had some common characteristics. When the surface perturbation lows first developed, they were usually accompanied by 700-hPa perturbation highs to their east. After the surface perturbation lows moved away from the lake region, besides the aforementioned 700-hPa perturbation highs, 700-hPa perturbation lows also formed to the west of the surface perturbation lows as the depths of the heat plumes increased with time (Sousounis 1997). A MAV was generated in each of the three different flow scenarios. The most intense MAV occurred in the continuous sinusoidal waves scenario when the dynamical forcing was strongest. The fact that the MAV has different intensity in different background flow scenarios suggests that there was nonlinear interaction between the lakes and dynamical forcing. Otherwise, the superposition of the MAV from the zonal flow scenario and the simulation results in expt 6 would be equal to those in expt 3. This nonlinear interaction fed back to the synoptic systems and resulted in stronger MAVs as the synoptic forcing increased from zonal flow to solitary trough and then to continuous sinusoidal waves.

A diagnosis of surface pressure deepening using ZO analysis showed that warm advection was the main cause for surface pressure deepening for expts 2 and 3 at both 24 and 36 h. Moreover, the preconditioning from the western half of the ridge ahead of the featured trough in expt 3 contributed to a longer fetch of warm advection and greater surface pressure deepening. It was also found that the presence of northwesterly flow in the eastern part of the ridge, ahead of the featured trough, brought colder air into the lake region and enhanced the surface heat fluxes over the lakes. The lake-aggregate–warmed air provided lower static stability and an environment more conductive for surface development.

The ZO analysis for expts 3 and 6 suggests that the lake aggregate not only contributed directly to the surface pressure deepening through sensible heating, but it also contributed indirectly to surface pressure deepening by enhancing warm advection, which subsequently increased surface pressure deepening. The direct contributions of the lake aggregate to surface pressure deepening were confined to the areas over the lakes and had smaller magnitude than the indirect contributions located near the warm front. Similar to what happened in expts 2 and 3, warm advection was also found to be a primary driving mechanism for surface pressure deepening in expt 6. The relatively smaller contributions of vorticity advection to surface pressure deepening in this study were due to the specified ICs. Thus, surface pressure deepening attributable to positive vorticity advection could be enlarged by properly changing ICs such as increasing the amplitude of the disturbance or decreasing the phase lag in the vertical direction.

The results of idealized numerical simulations and ZO analysis in this study lead to the following conclusions. First, the preconditioning from the preceding ridge is very important for the further development of synoptic-scale cyclones, especially for the lake-aggregate–enhanced synoptic-scale cyclones. Second, the lake aggregate not only contributes to surface pressure deepening locally through latent and sensible heating, but it also results in large-scale pressure falls by enhancing or generating the warm front.