Abstract

Although smaller lakes are known to produce lake-effect precipitation, their influence on the precipitation climatology of lake-effect regions remains poorly documented. This study examines the contribution of lake-effect periods (LEPs) to the 1998–2009 cool-season (16 September–15 May) hydroclimate in the region surrounding the Great Salt Lake, a meso-β-scale hypersaline lake in northern Utah. LEPs are identified subjectively from radar imagery, with precipitation (snow water equivalent) quantified through the disaggregation of daily (i.e., 24 h) Cooperative Observer Program (COOP) and Snowpack Telemetry (SNOTEL) observations using radar-derived precipitation estimates. An evaluation at valley and mountain stations with reliable hourly precipitation gauge observations demonstrates that the disaggregation method works well for estimating precipitation during LEPs. During the study period, LEPs account for up to 8.4% of the total cool-season precipitation in the Great Salt Lake basin, with the largest contribution to the south and east of the Great Salt Lake. The mean monthly distribution of LEP precipitation is bimodal, with a primary maximum from October to November and a secondary maximum from March to April. LEP precipitation is highly variable between cool seasons and is strongly influenced by a small number of intense events. For example, at a lowland (mountain) station in the lake-effect-precipitation belt southeast of the Great Salt Lake, just 12 (13) events produce 50% of the LEP precipitation. Although these results suggest that LEPs contribute modestly to the hydroclimate of the Great Salt Lake basin, infrequent but intense events have a profound impact during some cool seasons.

1. Introduction

Lake-effect precipitation is a potentially important component of the water cycle near large inland bodies of water, including the Great Salt Lake (GSL) of northern Utah. As a terminal lake within a closed hydrologic basin, the GSL serves as a collector and integrator of climate variability and change (Lall and Mann 1995; Lall et al. 1996; Mohammed and Tarboton 2011). Imbalances between lake inflows, which are dominated by surface-water runoff (66%) and direct precipitation on the lake (31%), and outflows, which consist entirely of evaporation, cause changes in lake level, area, and composition (e.g., salinity) that affect mineral industries, shoreline and aquatic ecosystems, natural resource management, and transportation (Arnow 1980; Gwynn 1980; Mohammed and Tarboton 2011; see also information from the U.S. Geological Survey Utah Water Science Center available online at http://ut.water.usgs.gov/greatsaltlake/).

Lake-effect precipitation (primarily snow) occurs over northern Utah several times per year (Steenburgh et al. 2000; Alcott et al. 2012) and contributes to the GSL water budget through direct precipitation on the lake and the buildup of a mountain snowpack that drives much of the surface-water runoff within the GSL basin, which serves as the primary water resource for 400 000 people in Salt Lake City (Salt Lake City Department of Public Utilities 1999). Lake-effect precipitation also provides a path for water recycling (Eltahir and Bras 1996) because evaporation from the lake contributes to a portion of the water mass that falls as precipitation (Onton and Steenburgh 2001) and eventually returns as surface-water runoff.

Lake-effect precipitation can contribute to substantial snow accumulations over northern Utah. Carpenter (1993) describes a lake-effect snowstorm that produced up to 69 cm of snow in the Salt Lake Valley and 102 cm in the adjacent Wasatch Mountains (see Fig. 1 for geographic locations). In an analysis of the so-called 22–27 November 2001 Hundred-Inch Storm, Steenburgh (2003) attributed 1.45 cm of the snow water equivalent (SWE) that fell at Salt Lake City International Airport (KSLC) and 5.54 cm of the SWE that fell at the Alta–Collins (CLN) observing station in the Wasatch Mountains to two lake-effect periods. Beyond potential impacts on water resources, lake-effect snowstorms help to fuel Utah’s $1.2 billion yr−1 ski industry and reputation for the “Greatest Snow on Earth” (Steenburgh and Alcott 2008; Gorrell 2011).

Fig. 1.

GSL basin, subbasins, and topography (following inset scale).

Fig. 1.

GSL basin, subbasins, and topography (following inset scale).

Previous studies illustrate the environmental conditions, seasonality, and spatial distribution of the GSL effect, which occurs during cold-air outbreaks when localized sensible heating and latent heating over the relatively warm water lead to the development of precipitating moist convection (Carpenter 1993; Steenburgh et al. 2000; Steenburgh and Onton 2001; Alcott et al. 2012). Alcott et al. (2012) identify an average of 13 GSL-effect periods per cool season (16 September–15 May, with the year defined by the ending calendar year), with autumn and spring peaks in frequency separated by a midwinter minimum. Radar reflectivities indicate that lake-effect precipitation is greatest to the south and east of the GSL and is most common in the overnight and early-morning hours (Steenburgh et al. 2000; Steenburgh and Onton 2001; Onton and Steenburgh 2001).

No previous study quantifies how much precipitation is produced seasonally during lake-effect periods (LEPs) around the GSL or, to our knowledge, similar meso-β-scale (20–200 km; Orlanski 1975) bodies of water in other regions. Such estimates have been made, however, for larger bodies of water such as the Laurentian Great Lakes by using a variety of approaches (e.g., Changnon 1968; Eichenlaub 1970; Braham and Dungey 1984; Scott and Huff 1996, 1997). For example, Changnon (1968) compares snowfall amounts near the climatological upwind (western) and downwind (eastern) shorelines of Lake Michigan, Eichenlaub (1970) examines how much snow is produced during periods when the synoptic conditions are favorable for lake effect, and Braham and Dungey (1984) and Scott and Huff (1996, 1997) calculate the enhancement relative to an estimate of non-lake-effect precipitation obtained by interpolating precipitation amounts from outside the lake-effect snowbelts. Scott and Huff (1996, 1997) estimate that lake effect more than doubles the mean wintertime snowfall east of Lake Superior and yields increases of 90% southeast of Lake Huron, 35% east of Lake Michigan, and 40% east of Lakes Erie and Ontario.

Complex topography strongly influences precipitation around the GSL and precludes the application of relatively simple approaches like those described above. The GSL is oriented from northwest to southeast, with the Wasatch Mountains to the east and the Oquirrh and Stansbury Mountains to the south (Fig. 1). The GSL has an average surface area of ~4400 km2, making it much smaller than Lake Ontario (~19 000 km2), which is the smallest of the Laurentian Great Lakes. The hydrologic basin of the GSL spans four states (Utah, Wyoming, Nevada, and Idaho) and encompasses a total area of 89 000 km2. Because of the small contribution of groundwater from the west desert, however, the basin has an effective area of 55 000 km2 (Lall and Mann 1995; Great Salt Lake Information System 2011). The lower-elevation basins and valleys receive 10–65 cm of precipitation (SWE) annually, whereas much larger amounts (100–130+ cm) fall in the mountains.

In this research, we develop and apply a technique to estimate the contribution of precipitation (SWE) produced during LEPs to the 1998–2009 cool-season hydroclimate of the Great Salt Lake basin. The technique uses high-frequency radar-derived precipitation estimates to produce an hourly precipitation dataset from gauge-based daily (24 h) observations, which can then be used to partition the observed precipitation into lake-effect and non-lake-effect periods. Although our results suggest that LEPs contribute modestly to the cool-season hydroclimate of the GSL basin, infrequent but intense events have a profound influence during some cool seasons.

2. Data and methods

LEP SWE estimates were produced for the 1998–2009 cool seasons. These cool seasons include 128 LEPs (Table 1), which were identified by Alcott et al. (2012) on the basis of visual inspection of the “KMTX” Weather Surveillance Radar-1988 Doppler (WSR-88D) radar-reflectivity imagery (see Fig. 1 for KMTX location).1 The identification method used by Alcott et al. (2012) follows that of Laird et al. (2009) and defines LEPs as periods of at least 1 h during which precipitation features 1) were coherent and quasi-stationary with a distinct connection to the lake, 2) were shallow and distinguishable from large, transitory synoptic features, and 3) exhibited increasing depth and/or intensity in the downwind direction.

Table 1.

LEP onset and ending times and evaluation stations. The latter identify the subset of LEPs used in section 3 for the evaluation of the disaggregation technique at KSLC (105 total LEPs) and CLN (78 total LEPs). Coverage is not complete for each station because of missing or unavailable radar or hourly precipitation data. Blank indicates that missing or unavailable radar (as indicated by footnote) or hourly precipitation data precluded evaluation of the disaggregation technique at both sites.

LEP onset and ending times and evaluation stations. The latter identify the subset of LEPs used in section 3 for the evaluation of the disaggregation technique at KSLC (105 total LEPs) and CLN (78 total LEPs). Coverage is not complete for each station because of missing or unavailable radar or hourly precipitation data. Blank indicates that missing or unavailable radar (as indicated by footnote) or hourly precipitation data precluded evaluation of the disaggregation technique at both sites.
LEP onset and ending times and evaluation stations. The latter identify the subset of LEPs used in section 3 for the evaluation of the disaggregation technique at KSLC (105 total LEPs) and CLN (78 total LEPs). Coverage is not complete for each station because of missing or unavailable radar or hourly precipitation data. Blank indicates that missing or unavailable radar (as indicated by footnote) or hourly precipitation data precluded evaluation of the disaggregation technique at both sites.

Generation of LEP SWE estimates follows the approach of Wüest et al. (2010) who disaggregated daily precipitation gauge observations using radar-derived SWE estimates to produce an hourly SWE dataset for Switzerland. Radar-derived SWE estimates have high temporal resolution (typically every 6–10 min) but suffer from low absolute accuracy, especially in complex terrain (e.g., Westrick et al. 1999; Rasmussen et al. 2001). Daily precipitation gauge observations provide greater absolute accuracy but lack the temporal resolution needed to isolate the SWE produced during LEPs, which are typically less than 24 h in duration and frequently cross the boundaries of observing periods. Therefore, we use the temporally resolved radar-derived SWE estimates to disaggregate the daily precipitation gauge observations into hourly intervals, a procedure that preserves the daily SWE totals and enables the separation of accumulated SWE into lake-effect and non-lake-effect periods. The method involves four steps: 1) construction of a serially complete precipitation gauge dataset, 2) calculation of hourly radar-derived SWE estimates, 3) disaggregation of daily precipitation gauge observations into hourly intervals, and 4) partitioning of SWE amounts into lake-effect and non-lake-effect periods.

a. Construction of a serially complete precipitation gauge dataset

The serially complete precipitation gauge dataset uses daily SWE observations from the National Weather Service (NWS) Cooperative Observer Program (COOP) and Natural Resources Conservation Service (NRCS) Snowpack Telemetry (SNOTEL) stations (Fig. 2a). COOP and SNOTEL data were obtained from the Utah Climate Center at Utah State University and the NRCS website, respectively. Most COOP stations are in valley locations, with citizen weather volunteers providing manual SWE measurements at 0.01-in. (0.25 mm) resolution from an 8-in. diameter (20.3 cm) precipitation gauge (NWS 1989; Daly et al. 2007). Most of the precipitation gauges are unshielded, with two known exceptions in the study area: KSLC and the Ogden Pioneer Power House (NWS 1989; S. Summy, NWS, 2011, personal communication). In addition, if the precipitation gauge measurement at KSLC is questionable, a manual measurement is taken and the observation record is adjusted accordingly (S. Summy, NWS, 2011, personal communication). Since the time of the observation is not consistent throughout the COOP network (varying from 1400 to 0700 UTC for the stations used in this study), the times were obtained from the National Climatic Data Center (NCDC) to enable more accurate disaggregation of the daily SWE amounts.

Fig. 2.

Surface stations used in the study (open circles indicate COOP, × symbols mark SNOTEL, the asterisk is KSLC, and the filled diamond indicates CLN). Blue indicates stations that are located to the southeast of GSL that were used in Fig. 9, and red indicates stations that were eliminated from the analysis because of full blockage of the KMTX 0.5° radar scan. (a) Terrain background (as in Fig. 1). (b) Radar beam blockage (following inset scale).

Fig. 2.

Surface stations used in the study (open circles indicate COOP, × symbols mark SNOTEL, the asterisk is KSLC, and the filled diamond indicates CLN). Blue indicates stations that are located to the southeast of GSL that were used in Fig. 9, and red indicates stations that were eliminated from the analysis because of full blockage of the KMTX 0.5° radar scan. (a) Terrain background (as in Fig. 1). (b) Radar beam blockage (following inset scale).

SNOTEL stations are located primarily in the mountains and provide automated hourly and daily accumulated SWE measurements from a large storage precipitation gauge at 0.10-in. (2.54 mm) resolution (Hart et al. 2004; NRCS 2011). The precipitation gauge is approximately 12 in. (30.5 cm) in diameter, has an Alter shield around the orifice to reduce wind effects on catchment, and contains antifreeze to melt frozen precipitation and oil to prevent evaporation (Serreze et al. 1999; Wallis et al. 2007; R. Julander, NRCS, 2010, personal communication). This design makes the SNOTEL precipitation gauge more accurate for frozen precipitation than a conventional tipping-bucket gauge, but foreign objects falling into the gauge and thermal expansion and contraction of the aluminum cylinder can produce false precipitation fluctuations (Kuligowski 1997; R. Julander, NRCS, 2010, personal communication). The thermal expansion and contraction combined with low (2.54 mm) data resolution limit the accuracy of the hourly accumulations. Therefore, we opt to disaggregate the daily data.

Precipitation gauges provide a direct measurement at a discrete point but are susceptible to systematic and random errors (Kuligowski 1997; Sieck et al. 2007; Vasiloff et al. 2007). Systematic errors include undercatch during high winds, which likely averages ~10%–15% for SNOTEL gauges (Rasmussen et al. 2011) and higher percentages for unshielded COOP gauges. Such errors are not accounted for in our statistics but are discussed where relevant in the paper. SNOTEL observations are otherwise generally reliable, but mechanical issues and infrequent maintenance due to their remote location can result in errors. COOP observations are subject to observer bias, which may include underreporting of light events and overreporting of events divisible by 0.05 and 0.1 in. (Daly et al. 2007).

NRCS performs several levels of quality control on the daily SNOTEL observations, including manual inspection by a hydrologist, before archiving them on their website (R. Julander, NRCS, 2010, personal communication; NRCS 2011). Although this does not eliminate all sources of error, we have assumed that the data are reliable, and we perform no additional quality control. To help to limit some of the problems with COOP observations, we use only COOP stations that report nearly continuously during the study period. Here, nearly continuously means missing less than 290 (~10%) of all possible daily SWE observations and less than 13 (~10%) of all possible daily SWE observations during LEPs. These criteria limit our analysis to the most frequently reporting COOP stations, which we hope reduces issues related to observer bias and identifies the most reliable stations. We also examined all daily SWE amounts of ≥50 mm and eliminated five that were clearly erroneous as deduced from manual checks of surrounding observations.

For COOP stations that meet the criteria above, we used the normal-ratio method (Paulhus and Kohler 1952; Young 1992; Eischeid et al. 2000) to estimate SWE on days with missing or erroneous data. First, we calculate the climatological linear correlation coefficient between the observed daily SWE at the missing data station and surrounding COOP stations. This is done by month to account for seasonality in the spatial distribution of SWE. Then, we compute a weight Wi for each surrounding station i from

 
formula

where ri is the correlation coefficient between station i and the missing data station and ni is the number of days used to calculate the correlation coefficient [Eischeid et al. 2000, their Eq. (1); Young 1992]. We then calculate the SWE estimate M from

 
formula

where N is the number of surrounding stations, si is the SWE observation at station i, and wi is the relative weight at surrounding station i, as given by

 
formula

In calculating M, we use only the four most highly correlated surrounding stations (i.e., N = 4) since Eischeid et al. (2000) found that the inclusion of more than four stations does not significantly improve the estimate and, in some cases, may actually degrade it. Further, because of the wide variation in COOP station observing times, only surrounding stations that report within 3 h of the missing-data station are used in the SWE estimate.

The SNOTEL records are more complete than the COOP records, with no more than 9 days of missing SWE observations at any station. Since SNOTEL stations report accumulated SWE, we assume that no SWE fell on these missing days if there was no change in accumulated SWE between the preceding and following days. Otherwise we use an estimate that is based on the application of the normal-ratio method described above to surrounding SNOTEL stations.

b. Calculation of hourly radar-derived SWE estimates

Estimating precipitation rate from radar reflectivity typically involves the use of an empirically derived power-law (i.e., Z–R) relationship of the form

 
formula

where Z is the radar-reflectivity factor (mm6 m−3), R is the rainfall rate (mm h−1), and a and b are constants (Doviak and Zrnic 1993; Rinehart 2004). The optimal Z–R relationship varies with storm type, precipitation type, and location (Rinehart 2004; Doviak and Zrnic 1993; Rasmussen et al. 2003). For estimating precipitation rates during snow, previous studies have derived an analogous Z–S relationship (where S is the SWE rate), with a ranging from 40 to 3300 and b ranging from 0.88 to 2.2 (Gunn and Marshall 1958; Ohtake and Henmi 1970; Sekon and Srivastava 1970; Carlson and Marshall 1972; Puhakka 1975; Fujiyoshi et al. 1990; Rasmussen et al. 2003; Warning Decision Training Branch 2011). For KMTX, Vasiloff (2001) recommends Z = 75S2, which provides a nearly one-to-one linear fit between storm-total radar estimates and precipitation gauge observations in the GSL basin, although considerable variability exists in the quality of fit from station to station. The NWS Warning Decision Training Branch currently recommends Z = 40S2 for snowstorms in the Intermountain West. We have used the Vasiloff (2001) relationship, although both relationships yield the same results since the disaggregation process is sensitive only to the exponent b and not to the coefficient a. For convenience, we apply this relationship to all LEPs, although a small fraction likely produced rain in the lower elevations.

Reflectivity values come from the lowest elevation scan (~0.5°) of the KMTX radar. For each scan, the maximum radar reflectivity in a nine-pixel stencil centered on each COOP and SNOTEL station is identified and converted to a SWE rate. The nine-pixel stencil helps to minimize the effects of wind displacement of the snow from the elevated radar scan to the ground-level observing station (Doviak and Zrnic 1993). SWE rates are converted into SWE amounts over each interval, defined as the time between radar scans, by taking the average SWE rate of the surrounding radar scans and multiplying it by the interval. Hourly SWE amounts are based on the integration of the SWE amounts calculated for all intervals during a given hour. This process is completed for all hours during all observing days that include an LEP.

To minimize storage space and processing time, we primarily use radar data archived in level-III [also known as Next Generation Weather Radar Information Dissemination Service (NIDS)] format (Baer 1991), which has a reflectivity resolution of 5 dBZ and a spatial resolution of 1° × 1 km. The level-III data are typically available every 6–10 min, but there are reporting gaps. SWE amounts during radar outages of 3 h or less are estimated in the same manner as above, by taking the average SWE rate of the scans surrounding the outage and multiplying it by the length of the outage. For lake-effect periods with radar outages that are greater than 3 h (9.4% of all lake-effect periods; Table 1), hourly SWE amounts are estimated using level-II data, which have the same spatial resolution (1° × 1 km) as level III but a much higher data resolution (0.5 dBZ; Crum et al. 1993). If level-II radar data are also missing (5.5% of all lake-effect periods; Table 1), hourly SWE amounts are estimated by interpolating 3-h SWE rates from the North American Regional Reanalysis (Mesinger et al. 2006).

c. Disaggregation of daily precipitation gauge observations

Using the hourly radar-derived SWE estimates, the disaggregated hourly precipitation gauge SWE Gt is calculated from

 
formula

where Et is the radar-derived hourly SWE estimate, Gd is the daily precipitation gauge observation, and t is the hour [Wüest et al. 2010, their Eq. (1)]. This disaggregation is performed for all days that include an LEP.

The disaggregation method significantly reduces quantitative biases produced by radar-derived SWE estimates (e.g., Doviak and Zrnic 1993; Vasiloff 2001; Rasmussen et al. 2003) but does not completely eliminate them (Wüest et al. 2010), as discussed in section 3. Further, in areas with complete radar beam blockage, the disaggregation will completely smooth the daily precipitation gauge total over all 24 h (i.e., the same SWE amount will be recorded for each hour). For this reason, 14 stations lying in areas with full radar beam blockage, as determined by following the method of Wood et al. (2003), were eliminated (Fig. 2b), leaving 55 for analysis.

d. Partitioning of SWE amounts

The disaggregated hourly SWE amounts are then partitioned into LEPs and non-LEPs on the basis of the onset and ending times of each LEP (Table 1). Since it is not possible to separate the precipitation produced by synoptic, orographic, and lake-effect processes and there are some LEPs during which these processes operate in concert (Steenburgh et al. 2000; Alcott et al. 2012), the LEP SWE may overestimate the total lake-effect precipitation, as discussed where appropriate later in this paper.

3. Evaluation of disaggregation method

To evaluate the accuracy of the disaggregation method, we compare disaggregated estimates of hourly and total SWE during LEPs with reliable hourly precipitation gauge observations at two surface stations: KSLC and CLN. The disaggregated estimates derive from daily (0000–0000 UTC) SWE accumulations summed from the hourly observations. This enables a direct evaluation that is not possible at stations with only daily data.

KSLC is a manually augmented Automated Surface Observing System station operated by the Salt Lake City NWS Forecast Office at an elevation of 1288 m in the Salt Lake Valley (Fig. 2a). During the first part of the study period, the station was equipped with a heated tipping bucket, which was replaced with an all-weather precipitation accumulation gauge (AWPAG) in July of 2004 (Groisman et al. 1999; Greeney et al. 2007; NWS 2011). The heated tipping bucket measures hourly SWE at a resolution of 0.01 in. (0.25 mm) and works by melting frozen precipitation before it enters the tipping apparatus (Groisman et al. 1999). The AWPAG weighs accumulated SWE at a resolution of 0.01 in. (0.25 mm) and includes a wind shield to reduce undercatch (Greeney et al. 2007; Tokay et al. 2010). KSLC data were obtained from NCDC.

CLN is a midmountain (2945 m) station in the Wasatch Mountains southeast of KSLC (Fig. 2a). Hourly SWE is measured at a resolution of 0.01 in. (0.25 mm) using a shielded 8-in. weighing gauge that contains antifreeze and a circulating device to minimize snow buildup on gauge walls. The snow safety staff at Alta Ski Area collected and provided the CLN data.

The hourly SWE observations from KSLC and CLN were not subjected to quality control. We concentrate the evaluation on a subset of LEPs that coincide with days with complete hourly data coverage at each station (Table 1). The coverage at KSLC is largely complete (105 out of 128 possible LEPs), but at CLN the LEPs available for evaluation begin in December of 1998 and are confined mainly to November–April, which are the months when the ski area is in operation (78 LEPs).2

We evaluate the accuracy of the disaggregation technique using scatterplots, frequency distributions, and three metrics—the mean bias error, mean absolute error, and total percent error:

 
formula
 
formula
 
formula

where E is the disaggregated estimate, O is the observed value, N is the number of data pairs, and i is the hour.

Figure 3a presents a scatterplot of hourly disaggregated versus observed SWE at KSLC during LEPs. There are 1199 hourly estimates, with a correlation of 0.75. The hourly estimates have some errors that appear to be quasi-randomly distributed, but with a tendency for underestimation for hours with high observed SWE. Figure 4a, however, shows that the total disaggregated SWE at KSLC for the LEPs has a much higher correlation (0.92) with observed amounts. Although this partly reflects the quasi-randomness of the hourly errors, longer time periods are inherently more likely to yield a better fit. For example, the disaggregation technique would yield a perfect estimate for a 24-h LEP that coincides with the period of the daily precipitation gauge data. Consistent with these results, the mean absolute and bias errors for disaggregated LEP SWE at KSLC are 0.46 and −0.02 mm, respectively. The small bias error is reflected in the bias-error frequency distribution, which is quasi normal with limited skew (Fig. 5a).

Fig. 3.

Hourly disaggregated vs observed SWE (mm) during LEPs at (a) KSLC and (b) CLN. The solid and dashed lines indicate one-to-one and regression lines, respectively.

Fig. 3.

Hourly disaggregated vs observed SWE (mm) during LEPs at (a) KSLC and (b) CLN. The solid and dashed lines indicate one-to-one and regression lines, respectively.

Fig. 4.

Disaggregated vs observed LEP SWE (mm) at (a) KSLC and (b) CLN. The solid and dashed lines indicate one-to-one and regression lines, respectively.

Fig. 4.

Disaggregated vs observed LEP SWE (mm) at (a) KSLC and (b) CLN. The solid and dashed lines indicate one-to-one and regression lines, respectively.

Fig. 5.

Frequency distribution of disaggregated LEP SWE bias error (mm) at (a) KSLC and (b) CLN. The bins are based on ranges of two-thirds. CLN has two outlying bias errors of −6.56 and −9.21 mm (not shown). Inset abbreviations include mean bias error (MBE), mean absolute error (MAE), and total percent error (TPE).

Fig. 5.

Frequency distribution of disaggregated LEP SWE bias error (mm) at (a) KSLC and (b) CLN. The bins are based on ranges of two-thirds. CLN has two outlying bias errors of −6.56 and −9.21 mm (not shown). Inset abbreviations include mean bias error (MBE), mean absolute error (MAE), and total percent error (TPE).

The scatterplot of hourly disaggregated versus observed SWE at CLN during LEPs exhibits similar scatter but has a slightly larger correlation (0.76) than is seen for KSLC (Fig. 3b). Like for KSLC, there is some tendency to underestimate high SWE hours, but, when integrated for the LEPs, the agreement between the total disaggregated and observed SWE is very good (0.99 correlation; Fig. 4b). The mean absolute and bias errors for disaggregated LEP SWE are 1.04 and −0.77 mm, respectively. The bias-error frequency distribution is quasi normal, but with a larger negative skew than at KSLC (Fig. 5b), illustrating that the disaggregation method tends to underestimate LEP SWE at CLN.

Hourly disaggregation errors stem from several sources. The first is representativeness error arising from differences between the volume (1° × 1 km) and point measurements made by the radar and precipitation gauge, respectively. In some instances, the radar-derived SWE might not be representative of the SWE falling at a point beneath the radar volume (e.g., Kitchen and Blackall 1992; Habib et al. 2004).

The second is the use of a single ZS relationship, which cannot represent actual precipitation rates in all storms given the wide variety of hydrometeors and precipitation processes (Doviak and Zrnic 1993; Rasmussen et al. 2001). Also contributing to ZS errors are issues related to the overshooting of shallow storms, evaporation and sublimation below the lowest-elevation radar scans, incomplete beam filling, and bright banding (Doviak and Zrnic 1993; Vasiloff 2001; Rasmussen et al. 2003; Wüest et al. 2010). For example, if the radar overshoots a storm for a portion of the day, the disaggregation will underestimate the precipitation rate during that period and overestimate it during the remainder of the day (Wüest et al. 2010). Likewise, if the precipitation seen by the radar evaporates or sublimates before reaching the gauge, the disaggregation will overestimate the precipitation rate during that period. Both of these types of errors are common at valley stations like KSLC because the average altitude of the center of the KMTX 0.5° radar scan is ~1500 m above the valley floor (Wood et al. 2003). When the radar beam intersects a melting layer it causes high-reflectivity returns, resulting in false precipitation intensity peaks in the disaggregation (Doviak and Zrnic 1993; Wüest et al. 2010). Incomplete beam filling can result in an underestimation of precipitation in the disaggregation as the radar is only partially sampling the storm.

These quantitative errors appear to largely cancel when integrating over long periods of time, resulting in a small total percent error of −1.5% and −11.7% at KSLC and CLN, respectively. Thus, we conclude that the method works reliably for estimating climatological LEP SWE totals.

4. Results

a. Cool-season mean and variability

To provide spatial context for the results for individual stations, Fig. 6 shows the frequency (%) of radar reflectivities ≥ 10 dBZ (an approximate threshold for accumulating snow; Steenburgh et al. 2000) during the 128 LEPs. If one neglects the large localized maxima produced by vehicle traffic along Interstate Highways 15, 80, and 215 (Slemmer 1998), the frequencies are greatest to the south and east of the GSL, indicating that these areas should receive more LEP SWE.

Fig. 6.

Frequency of occurrence of radar reflectivity that is ≥10 dBZ (%, following inset scale) during LEPs.

Fig. 6.

Frequency of occurrence of radar reflectivity that is ≥10 dBZ (%, following inset scale) during LEPs.

Outside of this region, there are also maxima over the Stansbury Mountains and the Wasatch Mountains southeast of Utah Lake (see Fig. 1 for locations). The maximum over the Stansbury Mountains largely reflects orographic precipitation that occurs during LEPs since lake effect occurs infrequently to the southwest of the Great Salt Lake (see section 4c). As discussed by Alcott et al. (2012), precipitation produced by non-lake-effect processes, including orographic precipitation, occurs in concert with lake effect 38% of the time. The maximum over the Wasatch Mountains southeast of Utah Lake likely also results from orographic precipitation. The terrain in this area has a southwest–northeast orientation, which is orthogonal to flow from the northwest, the most common wind direction during LEPs (Alcott et al. 2012). It could also be related to lake effect produced by Utah Lake, which has been observed by local forecasters but, to our knowledge, remains undocumented in the peer-reviewed literature. These examples illustrate that the LEP SWE statistics presented here include some precipitation produced by non-lake-effect processes and, at some locations, lake effect produced by Utah Lake.

Consistent with the frequency of radar reflectivities of ≥10 dBZ, the mean cool-season LEP SWE is greatest at stations to the south and east of the GSL, with the largest amounts at the Snowbird (SBDU1; 60.4 mm)3 and Dry Fork (DRFU1; 60.1 mm) SNOTEL stations in the Wasatch and Oquirrh Mountains, respectively (Fig. 7). High LEP SWE is also found at upper-elevation stations on the northern slope of the Uinta Mountains, in the Stansbury Mountains, and in the Wasatch Mountains southeast of Utah Lake. High LEP SWE in the latter two mountain regions likely reflects concomitant orographic precipitation or lake effect generated by Utah Lake. In the case of the station on the northern slope of the Uinta Mountains, however, an analogous maximum is not evident in the frequency of radar reflectivities of ≥10 dBZ because of partial beam blockage (cf. Figs. 6 and 7). Given the lack of radar coverage, it is unclear if the high LEP SWE reflects orographic precipitation during LEPs, disaggregation errors, or observational errors. Given that this site is well east of the GSL, the high LEP SWE is likely not representative of a large lake-effect contribution.

Fig. 7.

Mean cool-season LEP SWE (in millimeters, following the inset scale) at stations in the GSL basin. Maximum is 60.4 mm.

Fig. 7.

Mean cool-season LEP SWE (in millimeters, following the inset scale) at stations in the GSL basin. Maximum is 60.4 mm.

Despite being located in the region with the highest frequency of radar reflectivities of ≥10 dBZ, some COOP stations in the Salt Lake and Tooele Valleys south and east of the GSL observe much less LEP SWE than do SNOTEL stations in the adjacent mountains. This discrepancy partly reflects orographic effects. The fraction of mean cool-season SWE produced during LEPs (hereinafter the LEP fraction) helps to adjust for the inherent SWE gradient between valley and mountain stations and provides a more spatially coherent map of the mean contribution of LEPs to cool-season SWE (Fig. 8). LEP fractions above 5% occur at many stations south and east of the GSL, including those in the Oquirrh Mountains, Salt Lake Valley, and Wasatch Mountains east of the Salt Lake Valley, as well as two stations southeast of the Utah Lake. Although the greatest LEP SWE occurs at SBDU1 in the Wasatch Mountains east of the Salt Lake Valley, the highest LEP fractions are found in the Oquirrh Mountains and the Wasatch Mountains southeast of Utah Lake (cf. Figs. 7 and 8). This occurs because SBDU1 receives more precipitation during non-lake-effect periods given the diversity of storms and flow directions that affect that portion of the Wasatch Mountains (Dunn 1983). In contrast, the Oquirrh Mountains and Wasatch Mountains southeast of Utah Lake are climatologically drier, and therefore LEP SWE contributes to a greater fraction of the cool-season precipitation.

Fig. 8.

Mean cool-season LEP fraction (in percent, following the inset scale) at stations in the GSL basin. Maximum is 8.4%.

Fig. 8.

Mean cool-season LEP fraction (in percent, following the inset scale) at stations in the GSL basin. Maximum is 8.4%.

It appears, however, that measurement bias is an additional contributor to the low LEP SWE at some valley COOP stations. Figure 9 shows the frequency distribution of LEP fraction for COOP and SNOTEL stations south and east of the GSL (see Fig. 2 for station locations). LEP fractions average 3.4% at COOP stations and 5.0% at SNOTEL stations, and there are three COOP stations with an LEP fraction of <1.5%, which is inconsistent with the uniformly high frequency of radar reflectivities of ≥10 dBZ in this region (Fig. 6). Furthermore, the KSLC COOP station, which features a shielded precipitation gauge and manually augmented observations, has the highest lake-effect fraction (5.8%) of the COOP stations. Since LEP precipitation falls primarily as snow, the low LEP fraction at many COOP stations likely reflects undercatch of snowfall by unshielded precipitation gauges.

Fig. 9.

Frequency distribution of LEP SWE fraction at COOP (black) and SNOTEL (gray) stations southeast of the GSL (indicated by blue in Fig. 2).

Fig. 9.

Frequency distribution of LEP SWE fraction at COOP (black) and SNOTEL (gray) stations southeast of the GSL (indicated by blue in Fig. 2).

LEP SWE varies greatly between cool seasons, as illustrated by two stations in the LEP SWE maximum southeast of the GSL, KSLC, and SBDU1. KSLC, which is located at 1288 m in the Salt Lake Valley, has a mean cool-season LEP SWE of 16 mm, with a range of 3.9–36.6 mm (Fig. 10a). The corresponding LEP fraction mean is 5.8%, with a range of 2.9%–14.5%. At SBDU1, which is located at 2938 m in the Wasatch Mountains, the LEP SWE mean is much larger, 60.4 mm, with a range of 13.6–127.4 mm (Fig. 10b). The larger value reflects orographic precipitation enhancement at SBDU1 during LEPs. The LEP fraction mean and range at SBDU1, however, are 5.1% and 1.4%–11.6%, respectively, which are comparable to but slightly smaller than those observed at KSLC.

Fig. 10.

Cool-season LEP SWE (mm; solid black line) and fraction (%; dashed gray line) at (a) KSLC and (b) SBDU1. Black asterisk and gray triangle indicate the LEP SWE and fraction, respectively, after removal of LEPs during the Hundred-Inch Storm.

Fig. 10.

Cool-season LEP SWE (mm; solid black line) and fraction (%; dashed gray line) at (a) KSLC and (b) SBDU1. Black asterisk and gray triangle indicate the LEP SWE and fraction, respectively, after removal of LEPs during the Hundred-Inch Storm.

Both KSLC and SBDU1 exhibit a prominent peak in LEP SWE and fraction during the 2002 cool season when two intense LEPs occurred during the 22–27 November 2001 Hundred-Inch Storm (Steenburgh 2003).4 The first LEP produced 9.7 and 27.3 mm at KSLC and SBDU1, respectively, and the second LEP produced 20.8 and 80.1 mm, the latter being the largest LEP SWE observed at each site during the study period (Figs. 11a,b). Removing the SWE produced by these two LEPs reduces the 2002 LEP fraction to 2.7% and 1.9% at KSLC and SBDU1, respectively (gray triangles in Figs. 10a and 10b). Cumulative distribution functions of LEP SWE at KSLC and SBDU1 suggest that lake-effect events as large as the second LEP during the Hundred-Inch Storm are extremely infrequent (<1% probability of occurrence during the 12 cool-season study period; Fig. 12). Nevertheless, infrequent but intense LEPs have a profound impact on the overall climatology. A mere 12 (25) LEPs at KSLC and 13 (32) LEPs at SBDU1 account for 50% (75%) of the total SWE produced during the 128 LEPs included in the 12 cool-season climatology (Figs. 13a,b).

Fig. 11.

Frequency distribution of LEP SWE (mm) at (a) KSLC and (b) SBDU1. Gray dashed line shows the 75th percentile. Vertical dotted lines marked with “break” indicate a break in the numbering on the x axis. Largest event in (a) and (b) is the second LEP during the 22–27 Nov 2001 Hundred-Inch Storm.

Fig. 11.

Frequency distribution of LEP SWE (mm) at (a) KSLC and (b) SBDU1. Gray dashed line shows the 75th percentile. Vertical dotted lines marked with “break” indicate a break in the numbering on the x axis. Largest event in (a) and (b) is the second LEP during the 22–27 Nov 2001 Hundred-Inch Storm.

Fig. 12.

Cumulative distribution function of LEP SWE at KSLC (black line) and SBDU1 (gray line).

Fig. 12.

Cumulative distribution function of LEP SWE at KSLC (black line) and SBDU1 (gray line).

Fig. 13.

Total accumulated LEP SWE (mm) vs individual LEP SWE (mm) at (a) KSLC and (b) SBDU1. Upper and lower horizontal gray dashed lines indicate 50% and 25% of total accumulated LEP SWE, respectively.

Fig. 13.

Total accumulated LEP SWE (mm) vs individual LEP SWE (mm) at (a) KSLC and (b) SBDU1. Upper and lower horizontal gray dashed lines indicate 50% and 25% of total accumulated LEP SWE, respectively.

The importance of infrequent but intense LEPs, which may occur in isolation or possibly in sequence during episodes with highly favorable lake and atmospheric conditions (e.g., the Hundred-Inch Storm), is further demonstrated by the maximum monthly LEP SWE during the study period (Fig. 14a). The maximum monthly LEP SWE exceeds 54 mm at several stations in the Oquirrh and Wasatch Mountains south and east of the Great Salt Lake and 27 mm at lower-elevation stations in the northern Salt Lake Valley. These amounts are comparable to or larger than the mean cool-season LEP SWE at most stations (Fig. 14b), indicating that episodes with one or more intense LEPs play a dominant role in the lake-effect hydroclimate of the region.

Fig. 14.

(a) Maximum monthly LEP SWE (in millimeters, following the inset scale) at stations in the GSL basin during the study period. Maximum is 119 mm. (b) Ratio of largest monthly LEP SWE to the mean cool-season LEP SWE (%, following inset scale) at stations in the GSL basin. Maximum is 665%.

Fig. 14.

(a) Maximum monthly LEP SWE (in millimeters, following the inset scale) at stations in the GSL basin during the study period. Maximum is 119 mm. (b) Ratio of largest monthly LEP SWE to the mean cool-season LEP SWE (%, following inset scale) at stations in the GSL basin. Maximum is 665%.

b. Monthly mean and variability

At most stations, the mean monthly LEP SWE exhibits a bimodal distribution (Figs. 15a–i) with a primary peak in the autumn (October–November), a secondary peak in the late winter or early spring (March–April), and an intermediate winter minimum (January–February). This bimodal distribution resembles that of the monthly frequency of LEPs (Table 1; Alcott et al. 2012).

Fig. 15.

Monthly mean LEP SWE (in millimeters, following the inset scale) at stations in the GSL basin for (a) 16–30 September, (b) October, (c) November, (d) December, (e) January, (f) February, (g) March, (h) April, and (i) 1–15 May. Maximum at any station is 22.1 mm in November.

Fig. 15.

Monthly mean LEP SWE (in millimeters, following the inset scale) at stations in the GSL basin for (a) 16–30 September, (b) October, (c) November, (d) December, (e) January, (f) February, (g) March, (h) April, and (i) 1–15 May. Maximum at any station is 22.1 mm in November.

At KSLC, maxima in LEP SWE occur in November and April (5.0 and 2.3 mm, respectively) and are separated by a January minimum (0.4 mm; Fig. 16a). SBDU1 exhibits a less pronounced bimodal distribution with LEP SWE more heavily skewed toward the autumn months (Fig. 16b). The autumn maximum occurs in November (22.1 mm), and a less prominent secondary maximum occurs in March (5.4 mm). The winter minimum occurs in February (3.0 mm). The higher mean monthly LEP SWE at SBDU1 relative to KSLC reflects orographic enhancement. At both locations, the November maximum is amplified by the two LEPs during the Hundred-Inch Storm. Removing the SWE produced during these two LEPs reduces the mean November LEP SWE to 2.4 and 13.1 mm at KSLC and SBDU1, respectively (black asterisks in Figs. 16a,b).

Fig. 16.

Mean monthly LEP SWE (mm) at (a) KSLC and (b) SBDU1. Black asterisk indicates the SWE after the removal of LEPs during the Hundred-Inch Storm described by Steenburgh (2003).

Fig. 16.

Mean monthly LEP SWE (mm) at (a) KSLC and (b) SBDU1. Black asterisk indicates the SWE after the removal of LEPs during the Hundred-Inch Storm described by Steenburgh (2003).

c. Environmental conditions

We examined the environmental conditions during LEPs by accumulating the hourly LEP SWE into 12-h windows centered on KSLC upper-air sounding times (0000 and 1200 UTC). Because of missing sounding data, this analysis incorporates 92% of the LEP hours included in the 12-cool-season climatology discussed above.

The 700-hPa wind direction is typically used to anticipate the location of lake-effect storms and affects the fetch across the GSL (Carpenter 1993; Steenburgh et al. 2000; Alcott et al. 2012). Steenburgh et al. (2000) showed that the frequency of radar reflectivities of ≥10 dBZ during LEPs is greatest when the 700-hPa wind direction is 300°–360° (see their Fig. 16) but did not examine the SWE. Figure 17 presents the fraction of LEP SWE produced for various 700-hPa wind directions D. At most stations, the fraction of LEP SWE is greatest for 300° < D ≤ 330°, followed by 270° < D ≤ 300° and 330° < D ≤ 360°, respectively (Figs. 17b–d). For comparison, Alcott et al. (2012) found that the majority (~70%) of LEPs occur when the 700-hPa wind direction is 300°–360°. The importance of fetch, which is maximized at 325°, is underscored by the substantially lower fraction of LEP SWE for 240° < D ≤ 270°, 360° (north) < D ≤ 30°, and 30° < D ≤ 60° (Figs. 17a,e,f), although a lower frequency of favorable synoptic conditions during such flow regimes may also contribute.

Fig. 17.

Fraction of LEP SWE (percent, following the inset scale) for 700-hPa wind directions D of (a) 240° < D ≤ 270°, (b) 270° < D ≤ 300°, (c) 300° < D ≤ 330°, (d) 330° < D ≤ 360°, (e) 360° (north) < D ≤ 30°, and (f) 30° < D ≤ 60°. Maximum at any station is 56% at 300° < D ≤ 330°.

Fig. 17.

Fraction of LEP SWE (percent, following the inset scale) for 700-hPa wind directions D of (a) 240° < D ≤ 270°, (b) 270° < D ≤ 300°, (c) 300° < D ≤ 330°, (d) 330° < D ≤ 360°, (e) 360° (north) < D ≤ 30°, and (f) 30° < D ≤ 60°. Maximum at any station is 56% at 300° < D ≤ 330°.

A lake−700-hPa temperature difference ΔT of at least 16°C has historically been used as a necessary but not sufficient condition for the GSL effect (Steenburgh et al. 2000). Alcott et al. (2012) recently found, however, that LEPs during winter frequently feature lower ΔT values (as low as 12.4°C), whereas during the spring ΔT values are higher. They developed a seasonally varying relationship for the minimum ΔT required for lake effect, ΔTmin, that is based on a best-fit curve applied to the minimum ΔT observed during LEPs in each month:

 
formula

where d is the number of days since 15 September. They then defined ΔTexcess as the difference between the observed ΔT during an LEP and ΔTmin (i.e., ΔTexcess = ΔT − ΔTmin). A ΔTexcess that is ≥0 indicates that the seasonally varying threshold is met or exceeded.

Figure 18 shows that at most stations the fraction of LEP SWE is largest for 2° ≤ ΔTexcess < 4°C, followed by 4° ≤ ΔTexcess < 6°C. The LEP SWE fraction is somewhat lower for 0° ≤ ΔTexcess < 2°C and ΔTexcess ≥ 6°C. Lower LEP SWE for large ΔTexcess may seem counterintuitive but likely reflects a combination of factors. First, the climatological frequency of cold-air intrusions that yield large ΔTexcess values is low. Second, the development of moist convection requires not only instability, but also moisture and a mechanism to lift parcels to their level of free convection (Doswell 1987; Johns and Doswell 1992). For moisture, Alcott et al. (2012) found that the mean 850–700-hPa relative humidity RH850–700 provided the strongest discrimination between soundings with and without GSL-effect precipitation.

Fig. 18.

Fraction of LEP SWE (percent, following the inset scale) for (a) 0° ≤ ΔTexcess < 2°C (149 total LEP hours), (b) 2° ≤ ΔTexcess < 4°C (371 h), (c) 4° ≤ ΔTexcess < 6°C (362 h), and (d) ΔTexcess ≥ 6°C (350 h). Maximum at any station is 83% for 2 ≤ ΔTexcess < 4°C.

Fig. 18.

Fraction of LEP SWE (percent, following the inset scale) for (a) 0° ≤ ΔTexcess < 2°C (149 total LEP hours), (b) 2° ≤ ΔTexcess < 4°C (371 h), (c) 4° ≤ ΔTexcess < 6°C (362 h), and (d) ΔTexcess ≥ 6°C (350 h). Maximum at any station is 83% for 2 ≤ ΔTexcess < 4°C.

Figure 19 presents the fraction of LEP SWE at KSLC and SBDU1 as a function of ΔTexcess and RH850–700 (based on intervals of 1°C and 5%, respectively). At KSLC, the majority of LEP SWE occurs when 0 ≤ ΔTexcess ≤ 8°C and RH850–700 ≥ 70% (Fig. 19a). Similar results are found at SBDU1, although some LEP SWE is produced at lower (60%–70%) RH850–700 values (Fig. 19b). At KSLC and SBDU1, the absolute maximum at ΔTexcess ~3.5°C and RH850–700 ≈ 95% largely reflects the environmental conditions during one 12-h period with high precipitation rates during the Hundred-Inch Storm. Secondary maxima at other times reflect other intense LEPs. LEP SWE occurring when ΔTexcess < 0°C is a result of two LEPs that developed after the 700-hPa temperatures fell dramatically following the sounding time. In these cases, the actual ΔTexcess during lake-effect precipitation was likely larger and also greater than 0°C.

Fig. 19.

Fraction of LEP SWE at (a) KSLC and (b) SBDU1 during the 12-yr cool-season study period as a function of ΔTexcess and RH850–700.

Fig. 19.

Fraction of LEP SWE at (a) KSLC and (b) SBDU1 during the 12-yr cool-season study period as a function of ΔTexcess and RH850–700.

d. Perspectives on LEP trends and variability

The area of the GSL varies dramatically on interannual and interdecadal time scales (Lall and Mann 1995; Lall et al. 1996; Mohammed and Tarboton 2011) and could influence the frequency and magnitude of LEPs. In the historical record (1847–2011), the maximum, average, and minimum GSL area are 8550, 4400, and 2460 km2, respectively. During the study period, the mean cool-season area of the GSL featured an overall decline from 4500 to 3100 km2. Nevertheless, there is little correlation between standardized anomalies (i.e., departures from the study period mean expressed as the number of the standard deviations; Grumm and Hart 2001) of mean cool-season GSL area and cool-season LEP SWE at KSLC (R = 0.23) and SBDU1 (R = 0.36) (Figs. 20a,b). Thus, year-to-year variations in LEP SWE are poorly explained by variations in GSL area during the 12-cool-season study period. This suggests that knowledge of the area of the GSL has limited influence on the LEP SWE during the forthcoming cool season, although larger lake-area variations might have a more significant influence.

Fig. 20.

Standardized anomalies of cool-season LEP SWE (gray bars), mean cool-season GSL area (solid lines), and 500-hPa trough days (dash–dot lines) at (a) KSLC and (b) SBDU1.

Fig. 20.

Standardized anomalies of cool-season LEP SWE (gray bars), mean cool-season GSL area (solid lines), and 500-hPa trough days (dash–dot lines) at (a) KSLC and (b) SBDU1.

Alcott et al. (2012) identified a stronger relationship between the frequency of cool-season LEPs and 500-hPa trough days, defined as a day on which the 500-hPa relative vorticity exceeds 2 × 10−5 s−1. There is negligible correlation between the amount of cool-season LEP SWE at KSLC (R = 0.01) and SBDU1 (R = 0.11) and the frequency of 500-hPa trough days (Figs. 20a,b), however. Instead, as noted previously, cool-season LEP SWE is dominated by a small number of trough passages during which the environmental conditions enable the development of major lake-effect storms. Although it is recognized that both instability (i.e., ΔTexcess ≥ 0) and moisture (i.e., RH850–700 > ~60%) are needed for lake-effect storms, the factors contributing to intense events remain unclear. This lack of knowledge, combined with the infrequent nature of intense events, serves as a barrier to better understanding trends and variations in LEP SWE.

5. Summary and conclusions

We have evaluated and applied a method to estimate the amount of precipitation (snow water equivalent) produced during lake-effect periods in the region surrounding the Great Salt Lake. The method follows Wüest et al. (2010) and uses high-temporal-resolution radar-derived SWE estimates to disaggregate daily precipitation gauge observations to hourly time resolution. By combining these two datasets, we preserve the daily precipitation gauge totals obtained from SNOTEL and COOP stations and enable the separation of SWE into lake-effect and non-lake-effect periods.

The method was applied over the 1998–2009 cool seasons (16 September–15 May, with the year defined by the ending calendar year), which encompasses 128 LEPs, and was evaluated at valley (Salt Lake City International Airport) and mountain (Alta–Collins) stations that are located in the lake-effect precipitation belt southeast of the GSL. At both stations the method works well for estimating LEP SWE. Scatter exists in the hourly SWE estimates, but the errors are quasi random, and estimates for longer periods inherently yield a better fit. The sources of these errors include differences between the volume and point measurements made by radar and precipitation gauges, respectively, the use of a single ZS relationship for all LEPs, and radar sampling issues.

Analysis of the disaggregated COOP and SNOTEL data shows that the mean cool-season LEP SWE is greatest to the south and east of the GSL, including in the Oquirrh Mountains, Salt Lake Valley, and adjoining Wasatch Mountains. Upper-elevation SNOTEL stations generally receive more LEP SWE than valley COOP stations, reflecting both orographic precipitation enhancement and a probable undercatch of snow at COOP stations that use unshielded precipitation gauges. Relatively high LEP SWE at stations in other areas, including the Stansbury Mountains and Wasatch Mountains southeast of Utah Lake, is likely due primarily to precipitation phenomena that sometimes occur in concert with the GSL effect, including orographic precipitation and lake effect generated by Utah Lake. Therefore, the LEP SWE likely overestimates the precipitation produced solely by the GSL effect.

The fraction of cool-season SWE produced during LEPs (i.e., the LEP fraction) is also greatest to the south and east of the GSL. Although the largest LEP SWE occurs at the Snowbird SNOTEL (SBDU1) in the Wasatch Mountains east of the Salt Lake Valley, the largest LEP fractions are found in the Oquirrh Mountains and in the Wasatch Mountains southeast of Utah Lake, which are climatologically drier during non-lake-effect periods. Throughout the GSL basin, the LEP fraction is small, with a maximum of 8.4% in the Oquirrh Mountains. Although previous studies do not enable a direct comparison, such LEP fractions are likely much lower than are found downstream of larger bodies of water, such as the Laurentian Great Lakes. For comparison, Scott and Huff (1996, 1997) estimate that lake effect doubles the mean winter snowfall east of Lake Superior and increases snowfall east of Lake Ontario, the smallest of the Great Lakes, by 40%.

At most stations the mean monthly LEP SWE exhibits a bimodal distribution with a primary peak in autumn (October–November) and a secondary peak in late winter and spring (March–April), which closely resembles the monthly frequency of LEPs described by Alcott et al. (2012). The secondary late-winter and spring maximum contrasts with the Laurentian Great Lakes, which can become partially ice covered during winter and warm slowly in the spring (e.g., Niziol et al. 1995). In contrast, the shallow (mean depth ~3 m), hypersaline GSL never freezes and thus warms rapidly during the spring.

LEP SWE and fraction are highly variable among cool seasons and are strongly influenced by infrequent but intense LEPs. At KSLC and SBDU1, 50% of the total LEP SWE during the study period was produced by just 12 and 13 LEPs, respectively. The fraction of LEP SWE is greatest when the 700-hPa wind direction D is 300° < D ≤ 330°, which roughly corresponds to the maximum fetch (325°). Multiple variables are needed to help to identify the environmental conditions that produce the most LEP SWE, including a seasonally varying lake −700-hPa temperature threshold ΔTexcess and mean 850–700-hPa relative humidity RH850–700. Most of the LEP SWE at KSLC and SBDU1 falls when 0 ≤ ΔTexcess ≤ 8°C and RH850–700 ≥ 70%, although some LEP SWE falls at SBDU1 at lower RH850–700 values. Diagnosis of the factors leading to intense events, which may be related to poorly resolved (or understood) mesoscale phenomena such as thermally driven flows (e.g., Steenburgh and Onton 2001; Onton and Steenburgh 2001), remains an important topic for future research.

These results illustrate that LEPs contribute modestly to precipitation in the GSL basin, although infrequent but intense LEPs can have a profound influence in some years. Improved quantification of the contribution of the GSL to the hydroclimate of northern Utah may require efforts to distinguish between lake-effect and non-lake-effect precipitation during LEPs. Alternatively, regional climate simulations could be used to better understand the influence of the GSL. Such simulations would need to be run at high resolution (2 km or less) and adequately simulate lake-surface temperature, which plays a critical role in lake-effect storms and varies rapidly on time scales of a few days.

Acknowledgments

We thank Neil Laird and the summer-2007 undergraduate students Benjamin Albright and Jessica Popp at Hobart and William Smith Colleges who performed the initial LEP identification. Comments and suggestions from John Horel, Andy Wood, Court Strong, and two anonymous reviewers greatly aided the research effort and preparation of this paper. We gratefully acknowledge the provision of datasets, software, and/or computer time and services by NCDC, NCEP, NCAR, Unidata, the Center for Ocean–Land–Atmosphere Studies, and the University of Utah Center for High Performance Computing. This paper is based upon work supported by the National Science Foundation under Grant AGS-0938611. Any opinions, findings, and conclusions or recommendations expressed in this paper are those of the authors and do not necessarily reflect the views of the National Science Foundation.

REFERENCES

REFERENCES
Alcott
,
T. I.
,
W. J.
Steenburgh
, and
N. F.
Laird
,
2012
:
Great Salt Lake–effect precipitation: Observed frequency, characteristics, and associated environmental factors
.
Wea. Forecasting
,
27
,
954
971
.
Arnow
,
T.
,
1980
: Water budget and water-surface fluctuations of Great Salt Lake. Great Salt Lake: A Scientific, Historical and Economic Overview, J. W. Gwynn, Ed., Utah Geological and Mineral Survey, 255–264.
Baer
,
V. E.
,
1991
:
The transition from the present radar dissemination system to the NEXRAD Information Dissemination Service (NIDS)
.
Bull. Amer. Meteor. Soc.
,
72
,
29
33
.
Braham
,
R. R.
, and
M. J.
Dungey
,
1984
:
Quantitative estimates of the effect of Lake Michigan on snowfall
.
J. Climate Appl. Meteor.
,
23
,
940
949
.
Carlson
,
R. E.
, and
J. S.
Marshall
,
1972
:
Measurement of snowfall by radar
.
J. Appl. Meteor.
,
11
,
494
500
.
Carpenter
,
D. M.
,
1993
:
The lake effect of the Great Salt Lake: Overview and forecast problems
.
Wea. Forecasting
,
8
,
181
193
.
Changnon
,
S. A.
, Jr.
,
1968
: Precipitation climatology of Lake Michigan basin. Illinois State Water Survey Bulletin 52, Urbana, IL, 46 pp.
Crum
,
T. D.
,
R. L.
Alberty
, and
D. W.
Burgess
,
1993
:
Recording, archiving, and using WSR-88D data
.
Bull. Amer. Meteor. Soc.
,
74
,
645
653
.
Daly
,
C.
,
W. P.
Gibson
,
G. H.
Taylor
,
M. K.
Doggett
, and
J. I.
Smith
,
2007
:
Observer bias in daily precipitation measurements at United States Cooperative Network stations
.
Bull. Amer. Meteor. Soc.
,
88
,
899
912
.
Doswell
,
C. A.
, III
,
1987
:
The distinction between large-scale and mesoscale contribution to severe convection: A case study example
.
Wea. Forecasting
,
2
,
3
16
.
Doviak
,
R. J.
, and
D. S.
Zrnic
,
1993
: Doppler Radar and Weather Observations. Academic Press, 562 pp.
Dunn
,
L. B.
,
1983
: Quantitative and spacial distribution of winter precipitation along Utah’s Wasatch Front. NOAA Tech. Memo. NWS WR-181, 72 pp. [Available from NOAA/NWS Western Region Headquarters, 125 S. State St., Rm. 1311, Salt Lake City, UT 84138-1102.]
Eichenlaub
,
V. L.
,
1970
:
Lake effect snowfall to the lee of the Great Lakes: Its role in Michigan
.
Bull. Amer. Meteor. Soc.
,
51
,
403
412
.
Eischeid
,
J. K.
,
P. A.
Pasteris
,
H. F.
Diaz
,
M. S.
Plantico
, and
N. J.
Lott
,
2000
:
Creating a serially complete, national daily time series of temperature and precipitation for the western United States
.
J. Appl. Meteor.
,
39
,
1580
1591
.
Eltahir
,
E. A. B.
, and
R. L.
Bras
,
1996
:
Precipitation recycling
.
Rev. Geophys.
,
34
,
367
379
.
Fujiyoshi
,
Y.
,
T.
Endoh
,
T.
Yamada
,
K.
Tsuboki
,
Y.
Tachibana
, and
G.
Wakahama
,
1990
:
Determination of a Z–R relationship for snowfall using a radar and high sensitivity snow gauges
.
J. Appl. Meteor.
,
29
,
147
152
.
Gorrell
,
M.
,
2011
: Blue skies shine on Utah’s ski season openings. Salt Lake Tribune, 10 November, 1st ed. [Available online at http://www.sltrib.com/sltrib/outdoors/52882948-117/ski-season-utah-brighton.html.csp.]
Great Salt Lake Information System
, cited
2011
: Great Salt Lake basin watershed description. [Available online at http://www.greatsaltlakeinfo.org/Background/Description.]
Greeney
,
C. M.
,
J. V.
Fiore
,
J. M.
Dover
, and
M. L.
Salyards
,
2007
: Winter test of production all-weather precipitation accumulation gauge for ASOS 2005–2006. Preprints, 16th Conf. on Applied Climatology, San Antonio, TX, Amer. Meteor. Soc., JP1.11. [Available online at https://ams.confex.com/ams/pdfpapers/116371.pdf.]
Groisman
,
P. Ya.
,
E. L.
Peck
, and
R. G.
Quayle
,
1999
:
Intercomparison of recording and standard nonrecording U.S. gauges
.
J. Atmos. Oceanic Technol.
,
16
,
602
609
.
Grumm
,
R. H.
, and
R.
Hart
,
2001
:
Standardized anomalies applied to significant cold season weather events: Preliminary findings
.
Wea. Forecasting
,
16
,
736
754
.
Gunn
,
K. L. S.
, and
J. S.
Marshall
,
1958
:
The distribution with size of aggregate snowflakes
.
J. Meteor.
,
15
,
452
461
.
Gwynn
,
J. W.
, Ed.,
1980
: Great Salt Lake: A Scientific, Historical and Economic Overview. Utah Geological and Mineral Survey, 480 pp.
Habib
,
E.
,
G. J.
Ciach
, and
W. F.
Krajewski
,
2004
:
A method for filtering out raingauge representativeness errors from the verification distributions of radar and raingauge rainfall
.
Adv. Water Resour.
,
27
,
967
980
.
Hart
,
K. A.
,
W. J.
Steenburgh
,
D. J.
Onton
, and
A. J.
Siffert
,
2004
:
An evaluation of mesoscale-model-based model output statistics (MOS) during the 2002 Olympic and Paralympic Winter Games
.
Wea. Forecasting
,
19
,
200
218
.
Johns
,
R. H.
, and
C. A.
Doswell
III
,
1992
:
Severe local storms forecasting
.
Wea. Forecasting
,
7
,
588
612
.
Kitchen
,
M.
, and
R. M.
Blackall
,
1992
:
Representativeness errors in comparisons between radar and gauge measurements of rainfall
.
J. Hydrol.
,
134
,
13
33
.
Kuligowski
,
R. J.
,
1997
: An overview of National Weather Service quantitative precipitation estimates. NOAA/TDL Office Note 97-4, 27 pp.
Laird
,
N. F.
,
J.
Desrochers
, and
M.
Payer
,
2009
:
Climatology of lake-effect precipitation events over Lake Champlain
.
J. Appl. Meteor. Climatol.
,
48
,
232
250
.
Lall
,
U.
, and
M.
Mann
,
1995
:
The Great Salt Lake: A barometer of interannual climatic variability
.
Water Resour. Res.
,
31
,
2503
2515
.
Lall
,
U.
,
T.
Sangoyomi
, and
H. D. I.
Abarbanel
,
1996
:
Nonlinear dynamics of the Great Salt Lake: Nonparametric forecasting
.
Water Resour. Res.
,
32
,
975
985
.
Mesinger
,
F.
, and
Coauthors
,
2006
:
North American Regional Reanalysis
.
Bull. Amer. Meteor. Soc.
,
87
,
343
360
.
Mohammed
,
I. N.
, and
D. G.
Tarboton
,
2011
:
On the interaction between bathymetry and climate in the system dynamics and preferred levels of the Great Salt Lake
.
Water Resour. Res.
,
47
,
W02525
,
doi:10.1029/2010WR009561
.
Natural Resources Conservation Service
, cited
2011
: SNOTEL and snow survey and water supply forecasting. [Available online at http://www.wcc.nrcs.usda.gov/snotel/SNOTEL-brochure.pdf.]
Niziol
,
T. A.
,
W. R.
Snyder
, and
J. S.
Waldstreicher
,
1995
:
Winter weather forecasting throughout the eastern United States. Part IV: Lake effect snow
.
Wea. Forecasting
,
10
,
61
77
.
NWS
,
1989
: Cooperative station observations. National Weather Service Observing Handbook 2, 83 pp.
NWS
, cited
2011
: AWPAG Installation Dates for NWS and FAA owned ASOS sites. National Weather Service, 7 pp. [Available online at http://www.nws.noaa.gov/asos/pdfs/AWPAG_stat.pdf.]
Ohtake
,
T.
, and
T.
Henmi
,
1970
: Radar reflectivity of aggregated snowflakes. Proc. 14th Radar Meteorology Conf., Tucson, AZ, Amer. Meteor. Soc., 209–210.
Onton
,
D. J.
, and
W. J.
Steenburgh
,
2001
:
Diagnostic and sensitivity studies of the 7 December 1998 Great Salt Lake–effect snowstorm
.
Mon. Wea. Rev.
,
129
,
1318
1338
.
Orlanski
,
I.
,
1975
:
A rational subdivision of scales for atmospheric processes
.
Bull. Amer. Meteor. Soc.
,
56
,
527
530
.
Paulhus
,
J. L. H.
, and
M. A.
Kohler
,
1952
:
Interpolation of missing precipitation records
.
Mon. Wea. Rev.
,
80
,
129
133
.
Puhakka
,
T.
,
1975
: On the dependence of ZR relation on the temperature in snowfall. Preprints, 16th Radar Meteorology Conf., Houston, TX, Amer. Meteor. Soc., 504–507.
Rasmussen
,
R.
, and
Coauthors
,
2001
:
Weather Support to Deicing Decision Making (WSDDM): A winter weather nowcasting system
.
Bull. Amer. Meteor. Soc.
,
82
,
579
595
.
Rasmussen
,
R.
,
M.
Dixon
,
S.
Vasiloff
,
F.
Hage
,
S.
Knight
,
J.
Vivekanandan
, and
M.
Xu
,
2003
:
Snow nowcasting using a real-time correlation of radar reflectivity with snow gauge accumulation
.
J. Appl. Meteor.
,
42
,
20
36
.
Rasmussen
,
R.
,
J.
Hallett
,
R.
Purcell
,
S. D.
Landolt
, and
J.
Cole
,
2011
:
The Hotplate precipitation gauge
.
J. Atmos. Oceanic Technol.
,
28
,
148
164
.
Rinehart
,
R. E.
,
2004
: Radar for Meteorologists. 4th ed. Rinehart Publications, 428 pp.
Salt Lake City Department of Public Utilities
, cited
1999
: Salt Lake City watershed management plan. Salt Lake City Department of Public Utilities, 128 pp. [Available online at http://www.slcclassic.com/utilities/PDF%20Files/slcwatershedmgtplan.pdf.]
Scott
,
R. W.
, and
F. A.
Huff
,
1996
:
Impacts of the Great Lakes on regional climatic conditions
.
J. Great Lakes Res.
,
22
,
845
863
.
Scott
,
R. W.
, and
F. A.
Huff
,
1997
: Lake effects on climatic conditions in the Great Lakes Basin. Midwestern Climate Center Research Rep. 97–01, 73 pp.
Sekon
,
R. S.
, and
R. C.
Srivastava
,
1970
:
Snow size spectra and radar reflectivity
.
J. Atmos. Sci.
,
27
,
299
307
.
Serreze
,
M. C.
,
M. P.
Clark
, and
R. L.
Armstrong
,
1999
:
Characteristics of the western United States snowpack from Snowpack Telemetry (SNOTEL) data
.
Water Resour. Res.
,
35
,
2145
2160
.
Sieck
,
L. C.
,
S. J.
Burges
, and
M.
Steiner
,
2007
:
Challenges in obtaining reliable measurements of point rainfall
.
Water Resour. Res.
,
43
,
W01420
,
doi:10.1029/2005WR004519
.
Slemmer
,
J. W.
,
1998
: Characteristics of winter snowstorms near Salt Lake City as deduced from surface and radar observations. M.S. thesis, Department of Meteorology, University of Utah, 138 pp.
Steenburgh
,
W. J.
,
2003
:
One hundred inches in one hundred hours: Evolution of a Wasatch Mountain winter storm cycle
.
Wea. Forecasting
,
18
,
1018
1036
.
Steenburgh
,
W. J.
, and
D. J.
Onton
,
2001
:
Multiscale analysis of the 7 December 1998 Great Salt Lake–effect snowstorm
.
Mon. Wea. Rev.
,
129
,
1296
1317
.
Steenburgh
,
W. J.
, and
T. I.
Alcott
,
2008
:
Secrets of the “Greatest Snow on Earth.”
Bull. Amer. Meteor. Soc.
,
89
,
1285
1293
.
Steenburgh
,
W. J.
,
S. F.
Halvorson
, and
D. J.
Onton
,
2000
:
Climatology of lake-effect snowstorms of the Great Salt Lake
.
Mon. Wea. Rev.
,
128
,
709
727
.
Tokay
,
A.
,
P. G.
Bashor
, and
V. L.
McDowell
,
2010
:
Comparison of rain gauge measurements in the mid-Atlantic region
.
J. Hydrometeor.
,
11
,
553
565
.
Vasiloff
,
S.
,
2001
: WSR-88D performance in northern Utah during the winter of 1998–1999. Part I: Adjustments to precipitation estimates. NOAA/Western Regional Tech. Attachment 01-03, 5 pp.
Vasiloff
,
S.
, and
Coauthors
,
2007
:
Improving QPE and very short term QPF
.
Bull. Amer. Meteor. Soc.
,
88
,
1899
1911
.
Wallis
,
J. R.
,
M. G.
Schaefer
,
B. L.
Barker
, and
G. H.
Taylor
,
2007
:
Regional precipitation-frequency analysis and spatial mapping for 24-hour and 2-hour durations for Washington State
.
Hydrol. Earth Syst. Sci.
,
11
,
415
442
.
Warning Decision Training Branch
, cited
2011
: WSR-88D winter weather precipitation estimation. [Available online at http://www.wdtb.noaa.gov/courses/winterawoc/IC7/lesson2/part1/player.html.]
Westrick
,
K. J.
,
C. F.
Mass
, and
B. A.
Colle
,
1999
:
The limitations of the WSR-88D radar network for quantitative precipitation measurement over the coastal western United States
.
Bull. Amer. Meteor. Soc.
,
80
,
2289
2298
.
Wood
,
V. T.
,
R. A.
Brown
, and
S. V.
Vasiloff
,
2003
:
Improved detection using negative elevation angles for mountaintop WSR-88Ds. Part II: Simulations of the three radars covering Utah
.
Wea. Forecasting
,
18
,
393
403
.
Wüest
,
M.
,
C.
Frei
,
A.
Altenhoff
,
M.
Hagen
,
M.
Litschi
, and
C.
Schar
,
2010
:
A gridded hourly precipitation dataset for Switzerland using rain-gauge analysis and radar-based disaggregation
.
Int. J. Climatol.
,
30
,
1764
1775
.
Young
,
K. C.
,
1992
:
A three-way model for interpolating for monthly precipitation values
.
Mon. Wea. Rev.
,
120
,
2561
2569
.

Footnotes

*

Current affiliation: NOAA/National Weather Service Forecast Office, Cleveland, Ohio.

+

Current affiliation: NOAA/National Weather Service Western Region Headquarters, Salt Lake City, Utah.

1

This is one fewer cool season than the number investigated by Alcott et al. (2012), who extended their climatological dataset to include 2010 during the course of this project.

2

Because the LEP coverage is incomplete, the hourly SWE observations are only used only for the evaluation of the disaggregation technique. The climatology presented in section 4 is based on disaggregated daily data from KSLC and does not include CLN because of its more limited observing period.

3

SBDU1 is ~2 km west of CLN. As discussed in section 3, CLN is not used for the climatology because of its limited period of record.

4

Each of the two LEPs during the Hundred-Inch Storm includes an intense midlake band period and portions of the preceding and following postfrontal periods identified by Steenburgh (2003).