Characteristics of lake-effect snowstorms associated with the Great Salt Lake are described. Using WSR-88D radar imagery, 16 well-defined and 18 marginal lake-effect events were identified from September 1994 through May 1998 (excluding June–August), with the former used for more detailed analysis. Precipitation during the well-defined events was frequently characterized by the irregular development of radar echoes over and downstream of the Great Salt Lake. The most commonly observed precipitation structures were solitary wind-parallel bands that developed along or near the major axis of the GSL and broad-area precipitation shields with embedded convective elements that formed near the southern shoreline.
Regional-scale composite analyses and rawinsonde-derived statistics showed that the lake-effect events occurred in post frontal westerly to northerly 700-hPa flow following the passage of an upper-level trough and associated low-level cold front. The lake-effect environment was characterized by limited steering layer (800–600 hPa) directional shear (generally 60° or less), moist- to dry-adiabatic low-level lapse rates, and small convective available potential energy (CAPE), although the CAPE may be locally greater over the Great Salt Lake. In all events, the lake–700-hPa temperature difference exceeded 16°C, which roughly corresponds to a dry-adiabatic lapse rate. The lake–land temperature difference was always positive and usually exceeded 6°C, indicating significant potential for the development of land-breeze circulations and associated low-level convergence over the lake. Radar-derived statistics suggest that lake enhancement is strongest during periods of northwesterly to northerly flow and large lake–land temperature differences. These characteristics are compared with those associated with lake-effect snowstorms of the Great Lakes and implications for operational forecasting are discussed.
Accurate prediction of snowstorms that develop over or to the lee of the Great Salt Lake (GSL) is one of the major forecast challenges facing northern Utah meteorologists. Such lake-effect snowstorms occur several times each year, impacting transportation, commerce, and public safety along the densely populated urban corridor that is located south and east of the GSL and includes the cities of Ogden, Salt Lake, and Provo. Although typical lake-effect events produce snowfall accumulations of several cm, more intense and long-lived events can generate heavy accumulations. One event that occurred from 17 to 18 October 1984 produced up to 69 cm of snow, resulting in one million dollars of property damage (Carpenter 1993).
Understanding these lake-effect snowstorms has been limited by several factors including a lack of detailed radar and surface observations. Until 1994, when a National Weather Service (NWS) Weather Surveillance Radar–1988 Doppler (WSR-88D) was installed, local forecasters relied entirely on hand-drawn radar analyses from a local Federal Aviation Administration radar. In addition, conventional surface observations did not provide sufficient resolution to resolve local circulations, such as local convergence zones and lake or land breezes. This lack of observations, coupled with the complex topographic relief of the region, greatly limited the ability to examine the mesoscale structure and dynamics of lake-effect snowstorms.
Present understanding of lake-effect snowstorms of the GSL derives from Carpenter (1993), who identified events based on visual observations and spotter reports of both heavy snowfall and convection. The study used surface and 700-hPa observations from the Salt Lake City International Airport (SLC)1 to show that the lake-effect events were associated with post-cold-frontal northwesterly flow at 700 hPa, a lake–700-hPa temperature difference of at least 17°C, which approximately corresponds to a dry-adiabatic lapse rate, and an absence of capping stable layers or inversions below 650–700 hPa. Because of the limited observational data that was available during the period of study, Carpenter (1993) could not examine the mesoscale structure of the lake-effect events.
Lake-effect snowstorms of the GSL likely share some characteristics with those of the Great Lakes region of the eastern United States. In the Great Lakes region, lake-effect snowstorms generally require a lake–850-hPa temperature difference of 13°C, which represents a dry-adiabatic lapse rate (Rothrock 1969; Holroyd 1971;Niziol 1987). The low-level flow usually needs to be oriented along the major axis of a lake for intense, isolated snowbands to develop, whereas winds oriented along the minor axis of a lake usually produce multiple wind-parallel snowbands that are less intense (Niziol 1987; Niziol et al. 1995). Strong snowbands also appear to be favored during periods of unidirectional low-level flow, whereas stratiform cloudiness and light flurries are produced during periods of strong directional shear. Finally, capping stable layers or inversions limit cloud depth and precipitation rate (Kristovich and Laird 1998).
A variety of lake-effect precipitation patterns are produced by boundary-layer and mesoscale circulations associated with localized heating over lake surfaces. Braham and Kelly (1982), Braham (1983), Hjelmfelt (1990), and Niziol et al. (1995) describe several lake-effect precipitation patterns including the following:
Midlake bands that develop during periods of strong lake–land temperature contrasts when the large-scale flow is parallel to the major axis of a lake. Localized heating results in land-breeze-induced low-level convergence and the formation of an elongated band of clouds and precipitation that can extend over the lee shore (Peace and Sykes 1966; Passarelli and Braham 1981; Braham 1983; Hjelmfelt 1990; Niziol et al. 1995).
Mesoscale cyclonic vortices that can develop during periods of weak synoptic-scale flow and strong heat fluxes over the lake surface (Forbes and Merritt 1984; Pease et al. 1988; Hjelmfelt 1990; Laird 1999).
The purpose of this paper is to utilize observations from a recently installed NWS WSR-88D radar, along with other meteorological datasets, to advance understanding of lake-effect snowstorms of the GSL. The remainder of the paper is organized as follows. First an overview of the topography of northern Utah and the unique characteristics of the GSL are described in section 2. Then, section 3 describes the synoptic-scale evolution and environmental characteristics of lake-effect events of the GSL using regional composite analyses and sounding-derived statistics. This section also uses statistics derived from radar imagery to examine how factors such as flow direction, lake–land temperature gradient, and thermodynamic instability influence the position and intensity of lake-effect precipitation. Major results are summarized in section 4, which also includes recommendations for future research and improved forecasting.
2. The GSL and surrounding topography of northern Utah
The GSL is the largest body of water in the continental United States west of the Great Lakes and currently covers an area of approximately 4500 km2 (Fig. 1).2 The lake is approximately 120 km long and 45 km wide, with the major axis oriented in a northwest to southeast direction. Although relatively large in areal extent, the average (maximum) depth of the GSL is currently only 4.8 m (10.5 m). Because the GSL is a terminal lake with no outlets, its depth, areal coverage, and salinity vary significantly. Proxy data suggest that over the past 10 000 years the lake has never been completely dry and has exhibited a maximum elevation of 1285–1292 m. Since the mid 1800s, the average elevation has been near 1280 m, with a peak elevation of 1284 m achieved in 1873, 1986, and 1987. With these changes in elevation, areal coverage has varied from 2500 to 5900 km2.
Mean bimonthly lake-surface temperatures obtained from a United States Geological Survey (USGS) observing site at the Saltair Boat Harbor (SBH) show little lag relative to climatological mean air temperatures at nearby SLC (Fig. 2; see Fig. 1 for locations). The climatological maximum (minimum) lake-surface temperature is observed around 1 August (1 February), similar to the timing of the climatological maximum (minimum) mean air temperature. During the spring, mean lake and air temperatures are similar, while during the fall the lake tends to be slightly warmer than the ambient air temperature. It has been suggested that, on shorter timescales, the lake-surface temperature can deviate significantly from climatology and may correlate well with the preceding week’s mean air temperature at SLC (Carpenter 1993). During periods of cold weather, subfreezing lake temperatures can be observed since, due to its high salinity, the lake freezes only near freshwater inlets. Because the GSL never completely freezes and can warm rapidly due to its shallow depth, lake-effect precipitation can be observed from early fall through late spring.
The saline solution of the GSL is called brine and includes sodium chloride and several other salt and mineral constituents. Salinity varies greatly between the northern and southern halves of the lake, which are separated by an earthen railroad causeway that limits mixing (Fig. 1). The northern half, known as Gunnison Bay, receives little freshwater inflow and typically has a salinity near saturation (27%). To the south, Gilbert Bay has several freshwater inlets and a lower salinity, which is currently near 12% but has varied between 6% and 15% since 1970.
This brine is characterized by reduced saturation vapor pressure and associated latent heat flux compared to a body of freshwater. Because of the complexities of ionic dissociation and interaction in aqueous solutions, laboratory measurements are needed to precisely determine the magnitude of this reduction. For very high salinity, similar to that currently found in Gunnison Bay, a 20%–40% saturation vapor pressure reduction was found by Dickson et al. (1965) (Fig. 3). The reduction at lower salinity has not been experimentally determined, although Fig. 3 presents estimates using two approaches. The first uses Raoult’s Law and assumes 100% ionic dissociation of Salt Lake brine constituents and no attractions between solute ions and solvent molecules. The second estimate uses Eq. 7 of Low (1969) with mean sodium-chloride activity coefficients obtained from Harned and Owen (1958). This approach implicitly assumes that the brine is composed entirely of sodium chloride. These estimates suggest that the current reduction in saturation vapor pressure over Gilbert Bay, compared to a body of freshwater, is approximately 10%.
Understanding of lake-effect snowstorms of the GSL is further complicated by the surrounding complex topography (Fig. 1). To the east and southeast are the Wasatch Mountains, which rise abruptly 1000–2000 m to elevations of 2500–3500 m. Two other high-elevation mountain ranges, the Oquirrhs and Stansburys, are located south of the lake. Lowland regions include the Salt Lake and Tooele valleys, which contain the cities of Salt Lake (SLC) and Tooele (TOO). A large fraction of the region’s population lives along the Wasatch Front urban corridor that extends from Ogden (OGD), through SLC, to Provo (PVU). The Great Salt Lake Desert, with an elevation of about 1280 m, lies west of the GSL, while the Raft River Range rises to the northwest. Thus, flow from the northwest, which is associated with lake-effect storms (Carpenter 1993), must traverse substantial topography before moving over the GSL.
Lake-effect events were identified using all available lowest-elevation (0.5°) reflectivity images from the Promontory Point WSR-88D (KMTX; see Fig. 1 for location) during the period from September 1994 to May 1998, excluding the summer months of June–August. Radar images were generated using the Generalized Meteorological Analysis Package (GEMPAK; Koch et al. 1983) from data that was received and archived at the University of Utah in NEXRAD Information Dissemination Service (NIDS) format (Baer 1991). This data was generally available every 6 (10) min when the radar operated in precipitation (clear-air) mode and featured a spatial resolution of 1° × 1 km. The NIDS product suite provides information on 16 categories of reflectivity values (Baer 1991; Crum et al. 1993), resulting in an approximate data resolution of 5 dBZ. Events were classified subjectively as lake effect if reflectivity structures developed over and to the lee of the lake that did not appear to be associated with a larger-scale feature, or if precipitation associated with a larger-scale feature or local orographic lifting was clearly augmented or enhanced by the development of locally high radar reflectivities over and to the lee of the lake.
Although this classification scheme may appear straightforward, the complexities of local reflectivity patterns often made the identification of lake-effect events difficult. Apparent lake-effect precipitation bands frequently occur within, or coincident with, large-scale or orographically induced precipitation features (e.g., Slemmer 1998). Furthermore, reflectivity features may develop over the lake due to factors unrelated to the direct influence of the lake. As a result, potential lake-effect events were carefully examined, resulting in the identification of 16 well-defined events that featured precipitation structures that were clearly initiated by the lake (Table 1). In these events, lake-effect precipitation did not necessarily occur in isolation but may have occurred coincident with precipitation associated with other processes, such as orographic lifting. An additional 18 events (not shown) were classified as marginal because precipitation structures that formed over the lake were either short lived or may have developed due to other factors.
Lake-effect precipitation was often characterized by the irregular development of radar echoes over and downstream of the GSL. The most commonly observed precipitation structures were solitary wind-parallel bands that developed along or near the major axis of the GSL and may share some characteristics with midlake bands that have been observed over Lake Michigan (e.g., Braham and Kelly 1982; Hjelmfelt 1990). Figure 4a presents an example of such a band, which developed on 27 November 1995 and produced localized snow accumulations of 25 cm in the western Salt Lake Valley. Another event, which occurred from 17–18 January 1996, featured a midlake band and a region of orographic precipitation that extended northward along the Wasatch Mountains (Fig. 4b). It is possible that these midlake bands are associated with thermally driven low-level convergence as found for similar features over Lake Michigan (Hjelmfelt 1990). Other precipitation features included broad-area precipitation shields with embedded convective elements that were observed during some events along the southern lake shore. Such a feature formed on 30 March 1998 following passage of a large-scale precipitation shield from the north (Fig. 4c). There were no cases that featured multiple wind-parallel precipitation bands associated with periodic horizontal-roll convection as is found in some Great Lakes events (e.g., Kelly 1982). The lack of such bands may be due to the relatively small cross-flow axis of the lake (45 km) and the presence of substantial topography, which limits the development of periodic horizontal-roll circulations.
a. Event characteristics
The identified events represent an average of 4 well-defined and 4.5 marginal lake-effect events per year (September–May), although these may be underestimates due to gaps in the available radar data. The number of events observed in any given year ranged from 3 (2 well-defined, 1 marginal) in 1994–95 to 16 (6 well-defined, 10 marginal) in 1996–97. Lake-effect events were most common from October through February (Fig. 5) with the average ranging from 1 to 1.5 per calendar month. Events were less frequent during the months of March, April, and May. Only a single marginal event was observed in September. The maximum number of events (well-defined and marginal) observed in any given month was three, which occurred in December 1996 and February 1997.
The maximum lowland snowfall accumulations for well-defined events, as determined from NWS spotter reports and snowfall surveys by Slemmer (1998), are listed in Table 1. Isolating the snowfall produced exclusively by lake-effect processes was generally not possible due to the complexities of many events and the fact that these reports are storm-total accumulations. Several events exceed 20 cm, including the multiday snowstorm of 25–27 February 1998 that produced 129 cm in Bountiful, Utah, and widespread accumulations of more than 50 cm. This event is one of the largest lowland snowstorms in Utah history, with much of the precipitation orographically and lake enhanced during a period of cold, northwesterly flow (Slemmer 1998). An event that illustrates the amount of snow that can be produced exclusively by lake effect occurred on 27 November 1995 when a solitary snowband (see Fig. 4a) produced 25 cm in the west Salt Lake Valley. In extreme events, solitary snowbands have produced accumulations of over 40 cm.
A tendency for lake-effect precipitation to begin during the evening or overnight hours is illustrated by Fig. 6, which shows that 13 of the 16 well-defined events were initiated between 0000 and 1200 UTC (1700 and 0500 LST). There was also a tendency for lake-effect precipitation to be less common during the late afternoon hours [2100–0000 UTC (1400–1700 LST); not shown]. These characteristics may be related to diurnal modulation of the lake-land temperature gradient and associated land-breeze circulations, as will be discussed later in this paper. Slightly more than half (9) of the well-defined events persisted for 13–24 h, with one event (25–27 February 1998) occurring over a 58-h period (Fig. 7).
b. Synoptic-scale aspects
To elucidate the synoptic setting in which these lake-effect snowstorms occur, a composite analysis of the 16 well-defined lake-effect events was generated from operational three-dimensional gridded analyses produced by the National Centers for Environmental Prediction (NCEP) Rapid Update Cycle (RUC; Benjamin et al. 1991, 1994), which were archived at the University of Utah during the study period. These analyses feature 60-km horizontal, 25-hPa vertical, and 3-h temporal resolutions. Although composite analyses are presented every 12 h (Figs. 8–12), the compositing approach used the 3-h resolution of the RUC to minimize the temporal shifting required to construct averaged fields at each composite time.
The composite evolution illustrates that these lake-effect events typically occur during relatively cold northwesterly flow following the passage of an upper-level trough and associated surface-based cold front. One day prior to the onset of lake-effect precipitation (hour −24), an upper-level trough is located over the west coast of the United States, while a surface trough and associated wind shift extend southwestward from Idaho into southern Nevada (Figs. 8a,c). Baroclinity and cold advection located upstream of the surface trough at 700 hPa suggests that this feature marks the position of a surface cold front (Fig. 8b). Significant low-level moisture is located in the postfrontal environment over the northwest United States, where the 700-hPa relative humidity is generally greater than 70%. Southwesterly flow is evident over northern Utah, with weak warm advection implied by a slight veering of the winds with height (Fig. 8d). The near-surface temperature (850 hPa) is 8°C, implying that precipitation reaching the valley floor would likely be rain (Fig. 8d).
Twelve hours later (hour −12), the upper-level trough has moved over Nevada and the surface trough and associated wind shift have pushed into Utah (Figs. 9a,c). Concurrently, 700-hPa baroclinity and cold advection have moved over extreme northwest Utah (Fig. 9b). At SLC, the composite sounding shows west-southwesterly flow continues to extend throughout the troposphere (Fig. 9d). Low-level temperatures have cooled slightly compared to 12 h earlier, and the sounding has moistened.
By the onset time of lake-effect precipitation (hour 0), the cold front has passed through northern Utah as the upper-level trough continued to progress eastward (Figs. 10a,c). Winds at both the surface and 700 hPa have shifted to northwesterly (Figs. 10a,b), an orientation that is favorable for lake-effect and orographic precipitation over the Salt Lake Valley and western slopes of the Wasatch Mountains (Dunn 1983; Carpenter 1993). In fact, locally high 700-hPa relative humidity values are found over the Wasatch Mountains due in part to orographic ascent (Fig. 10b). The composite sounding shows that considerable cooling has occurred in the lower troposphere with 850-hPa temperatures near freezing and 700-hPa temperatures near −11°C (Fig. 10d). The lower atmosphere is conditionally unstable with mean 850–700-hPa and 850–500-hPa lapse rates of 7.6 and 7.3°C km−1, respectively. Although a hypothetical surface parcel lifted using the parcel method has little convective available potential energy (CAPE), the lower troposphere is approximately neutral to convective motions and exhibits no capping inversions or stable layers. The wind profile is characterized by 5 m s−1 near-surface winds that back approximately 45° to 15 m s−1 westerly winds at 500 hPa.
Composite analyses over the next 12 h showed continued eastward progression of the upper-level trough and associated surface front (Fig. 11). Lower-tropospheric temperatures over northern Utah decreased as winds throughout the troposphere veered to northwesterly (Fig. 11d). At SLC, 700-hPa temperatures have fallen to −13°C (Fig. 11b), approximately 20.6°C colder than the composite lake-surface temperature of 7.6°C.3 Below 500 hPa, the composite sounding remained conditionally unstable (Fig. 11d).
Lake-effect precipitation for most of the events in the composite had ceased by hour 24 (Fig. 12). At this time, the upper-level trough was located east of the Rockies and ridging was building over the western United States with the sea level pressure anticyclone centered over Utah. Surface winds were now westerly and veered with height to northwesterly at 500 hPa. With an upper-level ridge developing over the region, temperatures above 700 hPa were beginning to rise, limiting the vertical extent of any convection that may be initiated at low levels.
An important question concerns the representativeness of the composite evolution relative to the individual cases. Although the composite sample size is too small for meaningful estimates of statistical confidence, it does allow for careful examination of individual events. Of the 16 well-defined events, 11 (69%) were very similar to the composite and featured the passage of an upper-level trough and attendant frontal system originating from the west or northwest. In these cases, lake-effect precipitation occurred in relatively cold, postfrontal, westerly to northerly 700-hPa flow. The remaining cases featured the intrusion of low-level cold air with lake-effect precipitation falling in postfrontal westerly to northerly flow, but the evolution of the upper-level trough was different than that of the composite. Some cases featured amplifying upper-level troughs that became closed and stationary over the western United States whereas other events were associated with the movement of closed upper-level lows over Utah.
c. Sounding statistics
Conventional (0000 UTC and 1200 UTC) rawinsonde observations from SLC illustrate the atmospheric conditions that were evident southeast (and generally downwind) of the GSL during the 16 well-defined events. Upstream flow characteristics could not be examined because of a lack of representative surface and upper-air observations north and west of the GSL. This is due to the existence of significant topography (i.e., the Raft River Range) between the GSL and the closest observing sites in the Snake River Valley of Idaho.
1) Thermodynamic structure
Thermodynamic characteristics from the 29 rawinsonde observations taken during the 16 well-defined lake-effect events are summarized in the histograms presented in Figs. 13 and 14. At 700 hPa, temperatures ranged from −6.8°C to −18.2°C, with the majority of cases occurring at temperatures between −10°C and −16°C (Fig. 13a). Corresponding lake temperatures varied from 2°C to 18°C (Fig. 13b) and the lake–700-hPa temperature difference in all events exceeded 16°C, which roughly corresponds to a dry-adiabatic lapse rate (Fig. 13c). This lake–700-hPa-temperature-difference threshold agrees well with the 17°C threshold identified by Carpenter (1993), and with the experience of operational forecasters in the Great Lakes region who consider a lake–850-hPa temperature difference exceeding that of the dry-adiabatic lapse rate to be the minimum required for lake-effect snow (Rothrock 1969; Holroyd 1971; Niziol 1987). The observed surface–700-hPa lapse rates in all but two soundings averaged between moist and dry adiabatic (Fig. 13d).
Observed and lake-modified lifted index and CAPE histograms are presented in Figs. 14a–d. Since no direct observations of atmospheric conditions were available over the lake, the lake-modified statistics are presented to illustrate the potential for sounding modification due to sensible and latent heating over the lake. The surface-parcel characteristics used for the lake-modified calculations were based on an average of the SLC and lake temperatures and dewpoints, with the lake dewpoint set equal to the lake temperature.
Observed 500-hPa lifted indices were negative in only 9 of the 29 rawinsondes, and in many cases were strongly positive (Fig. 14a). Modification of the surface parcel resulted in lower lifted indices, but several soundings still exhibited negative parcel buoyancy at 500 hPa. Thus, many lake-effect events do not exhibit deep instability and the traditional 500-hPa lifted index can be a misleading forecast diagnostic.
Neutral or small positive parcel buoyancy was more commonly found at lower levels. Figure 14b shows that the minimum lifted index, defined as the lowest value of the difference between the environmental and parcel temperatures (Te − Tp), was below zero in a majority of the soundings and was never higher than 0.7 (Fig. 14b). The minimum lifted index pressure, which represents the level of the minimum lifted index, was located below 500 hPa in all but two events and was most frequently found between 750 and 550 hPa [∼1–3.5 km above ground level (AGL), Fig. 14c]. The observed CAPE was less than 50 J kg−1 in more than half the cases and exceeded 200 J kg−1 in only three soundings (Fig. 14d). Thus, a limited amount of positive parcel buoyancy was observed in a majority of the soundings. The SLC rawinsonde site is, however, located downstream of the GSL and convective overturning may have already led to near-neutral conditions at this location. Histograms using modified surface parcel temperatures and dewpoints suggest that additional parcel buoyancy may have been generated over the lake with lake-modified minimum lifted indices and CAPE values ranging from −2.8 to −12.6 and 211 to 2198 J kg−1, respectively (Figs. 14b,d). In most cases, the level of minimum lifted index and maximum parcel buoyancy for lake-modified parcels remains below 500 hPa (Fig. 14c). These findings are consistent with the fact that lake-effect convection is generally confined to the boundary layer (e.g., Chang and Braham 1991).
Carpenter (1993) suggested that lake-effect precipitation would be unlikely if the base of a capping inversion or stable layer was below 650 to 700 hPa (the GSL surface is located near 850 hPa). Figure 14e shows that the lowest observed capping inversion or stable-layer base during the 16 well-defined events was located at 700 hPa. Approximately half (14) of the 29 rawinsonde observations exhibited capping inversions or stable layers between 700 and 500 hPa, while the remaining soundings featured no significant capping inversions or stable layers below 500 hPa. One sounding did not feature a pronounced stable layer, but instead featured a deep layer of moderate stability above 700 hPa and was classified as undefined.
A histogram of lake–land temperature difference, calculated using estimated lake temperatures for each event and hourly surface observations from SLC, is presented in Fig. 14f. The lake–land temperature difference was 2°C or greater during all events and was frequently more than 6°C. Due to the large lake–air temperature contrast, such conditions would produce strong localized sensible and latent heat fluxes and would favor the development of land breezes and low-level convergence over the GSL, which may act to trigger and organize lake-effect precipitation. Studies over the Great Lakes have found that the lake–land temperature difference is a key variable in the development of lake-effect snowstorms (e.g., Hjelmfelt 1990).
In summary, the downwind thermodynamic environment during lake-effect snow events is characterized by moist- to dry-adiabatic low-level lapse rates, lake–700-hPa temperature differences that correspond to or exceed the dry-adiabatic lapse rate, small amounts of CAPE, and the absence of significant capping inversions or stable layers below 700 hPa. In some cases, near-neutral buoyancy can extend into the middle troposphere. Modified soundings suggest that CAPE and parcel buoyancy may, however, be greater over the GSL due to localized sensible and latent heat fluxes. Lake-effect events are also frequently characterized by large lake–land temperature differences, which may be associated with the development of land-breeze-induced low-level convergence over the GSL. Testing this hypothesis may be possible in the near future using new surface observing stations that will be placed around the GSL as part of the Utah Mesonet (Stiff 1997).
2) Wind and moisture characteristics
Local meteorologists consider the steering level for GSL lake-effect precipitation to be 700 hPa, which is approximately 1.5 km AGL. The histogram displayed in Fig. 15a shows that the 700-hPa wind direction during the 16 well-defined events was most frequently between 285° and 345°. Although none of the soundings featured 700-hPa winds from 345° to 360°, lake-effect precipitation did occur when the flow was from this direction, as verified by RUC 700-hPa wind analyses for periods intermediate to upper-air observing times (not shown).
Forecasters in the Great Lakes region have speculated that strong directional shear (>60°) in the steering layer is detrimental to the development of lake-effect precipitation (e.g., Niziol 1987; Niziol et al. 1995). Figure 15b is a histogram of the observed steering-layer (800–600 hPa) directional shear during GSL lake-effect events (the GSL surface is located near 850 hPa). All but one observation featured directional shear of less than 60°, and this observation (81°) was characterized by a light wind of 2.5 m s−1 at 800 hPa. It should be noted, however, that in some events the surface–600-hPa directional shear exceeds 60°. This is because southeasterly surface flow is commonly observed at SLC due to local nocturnal drainage flows (not shown), and may be enhanced during periods of strong lake–land temperature differences by an offshore land breeze circulation.
Speed shear in the steering layer was most frequently between −4 and 8 m s−1 [negative values arise during periods where the 800-hPa wind speed is larger than that at 600 hPa (Fig. 15c)]. Nevertheless, lake-effect precipitation was observed during periods of relatively strong speed shear and even strong reverse shear where the winds were weakening with height.
Given the reduction in saturation vapor pressure (Fig. 3) and relatively small size of the GSL, it is possible that the amount of upstream moisture is an important variable in lake-effect events. Unfortunately, there are no representative surface or upper-air observing sites located northwest of the GSL. The histogram presented in Fig. 15d illustrates the mean surface–700-hPa relative humidity downstream of the GSL. All events featured mean surface–700-hPa relative humidities of 54% or greater, and lake-effect precipitation was more commonly observed with larger relative humidity values.
d. Radar-derived precipitation climatology
To better understand the distribution and intensity of precipitation during lake-effect events, NIDS-formatted lowest-elevation radar scans were used to determine the mean reflectivity, median reflectivity, and the percentage of time that reflectivities equaled or exceeded several threshold values (e.g., 5, 10, or 20 dBZ) during the 16 well-defined lake-effect events identified in Table 1. This analysis approach was originally developed by Slemmer (1998) and utilized a total of 3048 radar scans. Statistics were also generated for subsets of the lake-effect events to examine the impact of large-scale atmospheric conditions (e.g., 700-hPa wind direction, lake–land temperature difference, time of day) on precipitation distribution and intensity. Analyses of the percentage of time that reflectivities equaled or exceeded 10 dBZ (hereafter the 10 dBZ frequency of occurrence or 10 dBZ FO) are presented since this approximately represents the threshold for accumulating snow. Statistics based on reflectivity were used instead of algorithm-derived accumulated precipitation because the latter is limited by the altitude of the KMTX radar, which is located 850 m above the surrounding lowlands, and uncertainties that arise from estimating snowfall rate from reflectivity data (e.g., Doviak and Zrnić 1993).
The 10-dBZ FO analysis for all events is presented in Fig. 16a. Beam blockage is evident in this and subsequent figures over and down radial from the higher peaks of the Wasatch, Oquirrh, and Stansbury Mountains. Additionally, a terrain-blocked radial extends southeastward from KMTX over the GSL and Salt Lake Valley due to high topography on Promontory Point (identified by a dashed line). Overall, the 10 dBZ FO pattern appears to represent a superposition of lake-effect and local orographic precipitation with the region of maximum FO located southeast of the GSL where 10-dBZ reflectivities were observed 15%–35% of the time. Significant orographic enhancement is implied by the large FO values that are found along the western slopes of the Wasatch and Oquirrh Mountains.
Figures 16b–d present the 10-dBZ FO analysis as a function of RUC-analyzed 700-hPa wind direction. These statistics were generated from subsets of radar scans that were based on the RUC analyzed 700-hPa wind direction at SLC. Only scans taken within ±1 h of the RUC analysis time were considered, with FOs calculated relative to the total number of scans in each subset.4 Periods with 270°–300° 700-hPa winds (173 radar scans) appeared to be less active, with 10-dBZ FO values significantly lower than during all events (cf. Figs. 16a,b). The influence of the lake for this flow direction appears to be relatively weak since the overall pattern is characterized by orographic precipitation along the western slopes of the Wasatch and Oquirrh Mountains and limited lake enhancement in the Salt Lake Valley. Some influence of the lake is, nevertheless, suggested by the fact that 10-dBZ FO values are somewhat higher along the Wasatch Mountains directly east of the GSL. Ten-dBZ reflectivities are more common when the 700-hPa flow is from 300° to 330° (1053 scans), with lake enhancement of precipitation suggested by the high FO values located downstream (southeast) of the GSL (Fig. 16c). As the flow becomes more northerly (i.e., 330°–360°; 630 scans), the region of maximum FO shifts to south of the GSL to include the eastern Tooele Valley and western Salt Lake Valley (Fig. 16d).
The lake–land temperature difference has been shown to play an important role in the development of lake-effect precipitation in the Great Lakes region (e.g., Hjelmfelt 1990). Such temperature contrasts lead to the development of land breezes and low-level convergence which play a role in lake-effect snowstorms over the Great Lakes (e.g., Peace and Sykes 1966; Passarelli and Braham 1981; Ballentine 1982; Hjelmfelt and Braham 1983; Braham 1983; Hjelmfelt 1990). To examine the importance of the lake–land temperature contrast, Fig. 17 presents 10-dBZ FO analyses as a function of the lake–land temperature difference, which was calculated from estimated lake temperatures and hourly SLC surface observations.
During periods of relatively weak lake–land temperature contrast (<4°C; 73 scans) the 10-dBZ FO pattern appeared to be primarily orographic with the largest FO values located along the western slopes of the Wasatch Mountains (Fig. 17a). Lake-induced precipitation features appeared to occur relatively infrequently. Periods of moderate lake–land temperature differences (937 scans) were more active, particularly over and southeast of the GSL where a large area of >25% 10-dBZ FO was evident (Fig. 17b). Embedded in this region of high FO was a line of >45% FO along the Wasatch Mountains. When the lake–land temperature difference was largest (>8°C; 866 scans) radar returns were focused to the southeast of the GSL and appeared to be dominated by lake-effect precipitation (Fig. 17c).
Because the lake–land temperature difference may be modulated diurnally by radiative processes, it might be anticipated that lake-effect precipitation exhibits a diurnal signal. Such a signal is also suggested by the tendency for lake-effect events to be initiated after sunset and be less frequent in the late afternoon (e.g., Fig. 6). An analysis of 10-dBZ FO as a function of time of day did exhibit weak diurnal variability with more frequent radar echoes observed from 0900 to 1200 UTC (0200–0500 LST; 465 scans) than from 2100 to 0000 UTC (1400–1700 LST; 265 scans; Figs. 18a,b).
The final characteristic examined was the lake–700-hPa temperature difference, which for the purposes of the radar climatology was calculated using estimated lake temperatures and RUC-analyzed 700-hPa temperatures. Lake–700-hPa temperature differences of <19°C (738 scans) showed a region of frequent 10-dBZ returns located south and east of the GSL with a northward extension of high FO along the western slopes of the Wasatch Mountains (Fig. 19a). Curiously, the frequency of 10-dBZ returns was lower for 19°–21°C temperature differences (600 scans), although the FO maximum did appear to become more confined to the southeast of the lake (Fig. 19b). Since one might expect that over a large sample size the frequency of precipitation to the lee of the lake would increase with decreasing stability, this observation may be the result of the relatively small 16 event sample size. During the most unstable periods, with lake–700-hPa temperature differences exceeding 21°C (642 scans), 10 dBZ reflectivities were concentrated southeast of the lake with a >45% maximum located in the western Salt Lake Valley (Fig. 19c).
4. Summary and conclusions
This paper has examined the characteristics of lake-effect snowstorms associated with the Great Salt Lake. A total of 16 well-defined and 18 marginal lake-effect events were identified from KMTX radar imagery during the period from September 1994 to May 1998 (excluding June–August). Lake-effect precipitation during well-defined events was frequently characterized by the irregular development of radar echoes over and downstream of the lake. The most commonly observed precipitation structures were solitary wind-parallel bands that sometimes aligned along the major axis of the lake as a midlake band, and broad-area precipitation shields with embedded convective elements that would occasionally develop along the south shore of the GSL. Events were most common from October through February when an average of 1–1.5 events per month (well-defined and marginal) occurred. Although the lake-effect contribution to snowfall during these events could not be isolated, several events featured lowland storm totals in excess of 20 cm, including the February 1998 event, which produced 129 cm in Bountiful, Utah.
Regional-scale composites using NCEP RUC analyses from the 16 well-defined events showed that lake-effect snowstorms are associated with the passage of an upper-level trough and associated surface-based cold front. Prior to the development of lake-effect precipitation, these features approach northern Utah from the west-northwest. Lake-effect snow begins following passage of the low-level cold front when the 700-hPa flow has shifted to northwesterly, a direction that is parallel to the major axis of the GSL and is favorable for lake-effect and orographic precipitation. This postfrontal environment is characterized by conditionally unstable low-level lapse rates with composite 700-hPa temperatures over SLC near −11°C. Over the next 12 h, 700-hPa temperatures fall to −13°C, resulting in further destabilization. The composite lake–700-hPa temperature difference at this time reaches 20.6°C, which exceeds a dry-adiabatic lapse rate. By 24 h after the onset time, the upper-level trough has moved east of the Rockies, upper-level ridging is building over the western United States, surface winds have backed to westerly, and the troposphere is stabilizing. Inspection of individual cases showed that 69% were similar in evolution to the composite and featured the passage of an upper-level trough and attendant frontal system from the west or northwest. The remaining cases also featured the intrusion of cold air at low levels and the development of westerly to northerly 700-hPa flow, but upper-level trough evolutions that differed from the composite. Such evolutions included amplifying upper-level troughs that became closed and stationary over the western United States or closed upper-level lows that moved from the Pacific over northern Utah.
Rawinsonde statistics from SLC, which lies downstream of the GSL, were used to illustrate the environmental conditions in which lake-effect precipitation occurs. Thermodynamically, the soundings were characterized by conditionally unstable low-level lapse rates, low CAPE values, and an absence of significant capping inversions or stable layers below 700 hPa. Modified soundings suggested that greater CAPE and convective instability may have been present over the GSL due to localized sensible and latent heat fluxes. The 700-hPa (steering-level) flow during the 16 well-defined events ranged from westerly to northerly with northwesterly flow most common. Steering-layer (800–600-hPa) directional shear was less than 60° in all but one event, which was characterized by light steering-layer winds. The lake–land temperature difference was positive during all events and was frequently larger than 6°C, suggesting significant potential for the development of land-breeze circulations and associated low-level convergence over the GSL.
The precipitation distribution during the 16 well-defined lake-effect events was examined by calculating the frequency of occurrence of 10 dBZ or greater reflectivities in the lowest-elevation (0.5°) radar scans. Cumulative statistics for all events showed 10-dBZ returns to be most common southeast of the GSL and along the western slopes of the Wasatch and Oquirrh Mountains. Ten-dBZ returns during periods of 270°–300° 700-hPa flow were most common to the lee (east) of the GSL and along the western (windward) slopes of the Wasatch Mountains. Periods with 700-hPa flow from 300° to 330° and 330° to 360° were characterized by more frequent 10-dBZ returns to the southeast and south of the lake, respectively. Additionally, orographic precipitation enhancement by the Wasatch and Oquirrh Mountains was observed during 700-hPa flow from 300° to 330° and 330° to 360°, respectively.
A strong focusing of precipitation to the lee of the lake was also evident during periods of large lake–land temperature differences. Periods with relatively small lake–land temperature differences (<4°C) appeared to feature predominantly orographic precipitation with relatively infrequent lake-effect precipitation. Periods with large lake–land temperature differences (>8°C) appeared to be strongly lake enhanced with 10-dBZ reflectivities most common to the south and southeast of the GSL.
These results suggest that lake-effect snowstorms of the GSL share many similarities with those of the Great Lakes (see Niziol et al. 1995 for a review). In both regions, the environment in which lake-effect storms develop is characterized by moist- to dry-adiabatic low-level (within ∼150 hPa of the surface) lapse rates, lake–700-hPa temperature differences (lake–850 hPa over the Great Lakes) that correspond to or exceed a dry-adiabatic lapse rate, and the absence of capping inversions or stable layers within 150 hPa of the surface. Additionally, there tends to be small (<60°) steering-layer directional shear. Finally, the lake–land temperature difference has been shown to be important in Great Lakes lake-effect snowstorms (e.g., Hjelmfelt 1990), and also appears to contribute to the development of lake-effect precipitation over the GSL. It is also possible that land breeze–induced convergence leads to the development of GSL midlake bands (e.g., Fig. 4a) in a manner similar to those observed over Lake Michigan (e.g., Passarelli and Braham 1981; Hjelmfelt 1990).
As a result of these common characteristics, many of the variables that should be considered when forecasting lake-effect snowstorms of the GSL are similar to those used by forecasters in the Great Lakes region (e.g., Niziol al. 1995). These characteristics include
a lake–700-hPa temperature difference of at least 16°C,
an absence of capping inversions or stable layers within 150 hPa of the surface (or below 700 hPa in the case of GSL lake effect),
small (<60°) steering-layer (800–600 hPa) directional shear except when the flow is weak, and
the existence of low-level convergence over the GSL.
Two other characteristics that may be important are the large-scale vertical motion, which may act to enhance (in the case of ascent) or suppress (in the case of subsidence) lake-effect precipitation (Niziol et al. 1995), and the upstream low-level moisture. The latter may be significant due to the relatively small size and hypersaline composition of the Great Salt Lake.
Two important caveats should be considered when assessing the potential for GSL lake-effect snowfall. The first is the potential for orographic precipitation, which is favored over the Wasatch Mountains and adjoining lowland region (including the Salt Lake Valley) during periods of northwesterly flow (Dunn 1983). Thus, significant snowfall may occur during periods that are not favorable for lake-effect precipitation if conditions are right for orographic precipitation. The second is the impact of numerical-model forecast errors on lake-effect precipitation forecasts. Average 12-h (36-h) root-mean-squared errors produced by present-day forecast models over the western United States are roughly 1.4°–1.7° (1.7°–2.2°)C for 700-hPa temperature, 18%–25% (22%–28%) for 700-hPa relative humidity, and 5.4–6.2 (6.4–7.7) m s−1 for 700-hPa vector wind, with the latter produced by errors in wind magnitude and direction (White 1997; Cook 1998; White et al. 1999). Such errors are significant given the sensitivity of lake-effect precipitation to environmental lapse rates, moisture availability, and wind direction, and will limit the predictability of lake-effect events. Even with improved higher resolution models, it is unclear if, for these relatively small spatial scales, initial-condition uncertainty will limit forecast improvements beyond 24 h (e.g., Lorenz 1969), or if the fixed surface forcing of the GSL and surrounding topography will result in enhanced mesoscale predictability at longer lead times (e.g., Paegle et al. 1990).
Finally, significant potential for advancing the understanding and short-range prediction of GSL lake-effect snowstorms will develop over the coming years as observations from new surface stations over and around the Great Salt Lake are collected as part of the Utah Mesonet (Stiff 1997). Such observations should allow for the diagnosis of local mesoscale circulations and their role in generating lake-effect snowbands. Improved observations will also allow for the validation of mesoscale-model simulations that can be used to further examine the dynamics of lake-effect events.
This work was conducted with support provided from National Science Foundation Grant ATM-9634191 and NOAA Grant OGP-526404. Special thanks to Eric Grimit, who assisted in the identification of lake-effect events, and to Larry Dunn, John Horel, Jan Paegle, Tom Potter, David Schultz, Mike Splitt, and Jonathan Slemmer for their contributions, advice, and scientific support. Comments from two anonymous reviewers greatly improved the manuscript.
Corresponding author address: Dr. W. James Steenburgh, Meteorology Dept., University of Utah, 135 S. 1460 E., Room 819, Salt Lake City, UT 84112.
Due to the elevation of SLC (1288 m), surface and 700-hPa observations are used instead of surface and 850-hPa observations as is commonly done in studies of lake-effect snowstorms over the Great Lakes.
Information on lake hydrology and chemical content was obtained from the United States Geological Survey, Salt Lake City, Utah, via their world wide web page (http://wwwdutslc.wr.usgs.gov/greatsaltlake/saltlake.html) and from Gwynn (1980).
Estimated lake temperatures for each event were determined from observations compiled by the USGS at intervals ranging from 2 to 4 weeks.
The calculation of FO relative to the number of scans in each subset is used throughout this section.