The shrinking Great Salt Lake contributes to record high dust-on-snow deposition in the Wasatch Mountains during the 2022 snowmelt season

Seasonal snowmelt from the Wasatch Mountains of northern Utah, USA, is a primary control on water availability for the metropolitan Wasatch Front, surrounding agricultural valleys, and the Great Salt Lake (GSL). Prolonged drought, increased evaporation due to warming temperatures, and sustained agricultural and domestic water consumption have caused GSL water levels to reach record low stands in 2021 and 2022, resulting in increased exposure of dry lakebed sediment. When dust emitted from the GSL dry lakebed is deposited on the adjacent Wasatch snowpack, the snow is darkened, and snowmelt is accelerated. Regular observations of dust-on-snow (DOS) began in the Wasatch Mountains in 2009, and the 2022 season was notable for both having the most dust deposition events and the highest snowpack dust concentrations. To understand if record high DOS concentrations were linked to record low GSL levels, dust source regions for each dust event were identified through a backward trajectory model analysis combined with aerosol measurements and field observations. Backward trajectories indicated that the exposed lakebed of the GSL contributed 23% of total dust deposition and had the highest dust emissions per surface area. The other potential primary contributors were the GSL Desert (45%) and the Sevier + Tule dry lakebeds (17%), both with lower per-area emissions. The impact on snowmelt, quantified by mass and energy balance modeling in the presence and absence of snow darkening by dust, was over 2 weeks (17 d) earlier. The impact of dust on snowmelt could have been more dramatic if the spring had been drier, but frequent snowfall buried dust layers, delaying dust-accelerated snowmelt later into the melt season.


Introduction
Utah's Wasatch Mountains, located at the eastern edge of the Great Basin, USA, hold a seasonal snowpack which serves as a natural reservoir, annually containing the majority of the region's water supply (Bardsley et al 2013). Snowmelt-driven runoff supports the livelihoods of downstream agricultural valleys and the urban Wasatch Front, home to 80% of Utah's population. Water yield in the region correlates to the peak annual snow water equivalent (SWE) of the mountain snowpack (e.g. Brooks et al 2021), and infrastructure has been designed around capturing seasonal runoff in man-made reservoirs for distribution to downstream water users throughout the year (Sturm et al 2017). For Salt Lake City, 50%-60% of the municipal water supply comes from four streams that drain out of the Central Wasatch, but there is relatively low reservoir capacity (0.026 km 3 ; Bardsley et al 2013), rendering controls on snowmelt timing and magnitude critical. In addition to its role as a natural reservoir, the mountain snowpack is crucial to the Utah economy. The Wasatch is well known for The Greatest Snow on Earth ® , which contributes $1.1 million per centimeter of snow to Utah's snowsports industry, totaling over $1.5 billion year −1 (Leaver 2021).
Snowmelt from Wasatch also determines the annual recharge of the Great Salt Lake (GSL). From north to south, the mountainous Bear, Weber, and Jordan watersheds internally drain into the GSL. The GSL has no minimum lake level, and extensive and consistent water use (primarily in the form of agricultural diversions) since the mid-1800s, coupled with warming and a succession of below-average snow years, have resulted in all-time low GSL levels (Null and Wurtsbaugh 2020;United States Geological Survey (USGS) 2023). The GSL itself plays a critical role for the Utah economy, supporting the $1.48 billion year −1 mineral extraction, lake recreation, and brine-shrimp industries (Null and Wurtsbaugh 2020). Additionally, the GSL is a vital hub for migratory bird species, and recent declines have increased lake salinity and threatened biological collapse of the GSL's ecosystem (Barnes and Wurtsbaugh 2015). Furthermore, large expanses of exposed lakebed area are susceptible to dust emission (Perry et al 2019) and contain elevated levels of heavy metals (Adams et al 2015), highlighting health concerns for 2.6 million people directly downwind. Finally, dust deposition onto the snowpack of the neighboring Wasatch Mountains influences snowmelt rates, which in turn directly alters first order controls on watershed storages and fluxes, potentially impacting consumptive water use and lake levels (Skiles et al 2018b).
Across Utah, observations have shown decreasing snow cover and snow depth in recent decades, largely attributed to an increasing fraction of wintertime precipitation falling as rain (Gillies et al 2012). In high elevation regions this effect is less pronounced, and wintertime precipitation remains snow dominant. Despite this resilience, such regions are susceptible to deposition of light absorbing particles (LAPs), including mineral dust and black carbon (BC) since widespread modern settlement of the Western US in the 19th century (Neff et al 2008, Dozier 2011, Steenburgh et al 2012. The occurrence of LAPs lowers snow albedo, causing more solar radiation to be absorbed into the snowpack (e.g. Skiles et al 2018a). Absorbed solar radiation is the primary driver of snowmelt, and dust-on-snow (DOS) events are known to cause earlier and more rapid melt. The impacts of DOS on snowmelt have been studied extensively across the upper Colorado River Basin, where the radiative forcing of the snowpack has accelerated snowmelt by 3-5 weeks, reduced total water yield, and has been correlated to springtime runoff rates and snowmelt forecasting errors (Painter et al 2010, Bryant et al 2013. In addition to the exposed dry lakebed of the GSL, the Wasatch Mountains are also downwind of barren lands and seasonally dry lakebeds of the eastern Great Basin (known in Utah as the West Desert, figure 1).
Barren land surfaces across the West Desert are naturally resistant to aeolian erosion unless disturbed. However, decreases in seasonal lakebed volumes have slowed the growth of wind-resistant saline crusts, which along with human-related disruption, have allowed for increased dust emissions Nicoll 2014, Perry et al 2019). The passage of strong synoptic cold fronts in the spring months result in winds that exceed the threshold frictional velocity for particle transport (Gillette et al 1980, Steenburgh et al 2012, resulting in regular seasonal dust emissions and DOS events in the Wasatch Mountains. In addition to dust, it is likely that there is LAP deposition on snow from aerosols, including BC, emitted from populations and industry along the adjacent Wasatch Front, although no effort has yet been undertaken to partition LAP types and their impacts. Previous DOS work in the Wasatch Mountains focused on specific dust events; this is the first comprehensive 'source-to-sink' study of DOS events across an entire snow season. This was motivated by the 2021-2022 snow cover season (October-June), which had record-high dust concentrations and frequent spring dust events, in addition to concerns about increasing dust emissions from the dry lakebed of the GSL following consecutive years of record low lake levels. A combination of regular field observations, laboratory analyses for dust and BC, per-event backward-trajectory atmospheric modeling, and snow energy and mass balance modeling were used to study DOS events and subsequent melting. The objectives were to assess; (1) the location of dust sources and relative contribution to dust loading, (2) accelerated snowmelt due to snow darkening, and (3) the potential of frequent late season snowfall to mitigate dust impacts.

Atwater Study Plot
Observations for this study were collected at the Atwater Study Plot (ASP) in the town of Alta in Little Cottonwood Canyon (LCC), Utah (40.591206 • N, 111.637685 • W). The LCC watershed drains steep high-elevation alpine terrain in the central Wasatch Mountains and is the greatest contributor to Salt Lake City's water supply. The snow study site was originally established for avalanche studies by the Forest Service in 1939. The site, now operated by the Utah Department of Transportation for snow safety observations, has snow energy balance instrumentation funded and maintained by the Snow Hydrology Research to Operations Laboratory (Snow HydRO Lab) at the University of Utah (Skiles et al 2022). The study plot lies within a generally flat opening in a mixed aspen and conifer forest on an otherwise S-SE facing slope at approximately 3667 m, directly adjacent to Alta Ski Area (figure 1).

Figure 1.
Map of known dust source regions across Utah's West Desert. Land cover dataset was obtained from USGS, and Lake Bonneville levels were acquired through the Utah geospatial resource center. Source regions were documented in Hahnenberger and Nicoll (2014) and Steenburgh et al (2012). * Note the land cover dataset displayed does not reflect the current GSL levels.
Station instrumentation is installed on a fixed platform and measured variables include air temperature, relative humidity, wind speed, and wind direction. Radiation balance is monitored with a four-component radiometer for incoming and outgoing broadband solar and longwave radiation (4.5-40 µm), supplemented with a downlooking snow surface temperature sensor, and a pair of uplooking and downlooking visible solar radiation pyranometers. The near infrared solar radiation (0.70-3.0 µm) is calculated from the difference between the visible (0.29-0.70 µm) and broadband solar values (0.29-3.0 µm). Snow depth is measured from an ultrasonic sensor mounted on a stake to the west side of the instrumentation platform, and SWE is quantified from regular snow pit observations of snow depth and density collected on the east side of the platform. Snow samples are also collected in each pit for analysis of LAP content (described in more detail in section 3.1). Seasonally, during snowmelt, the particle flux and size of aerosols in transport are monitored by a portable aerosol spectrometer maintained by the University of Utah's Department of Atmospheric Science (Skiles et al 2018b).

Snow measurements
Routine field measurements were conducted during the 2021-2022 snow season at ASP. Field collection occurred weekly to biweekly during the snowmelt season when dust events were most frequent (figure 2(a)). A full-depth snow pit was dug during each field session. Observations included snow stratigraphy and profiles of density and temperature. Temperature was recorded every 10 cm. Density, measured continuously using a 1000 cm 3 density cutter, was used to calculate SWE. Across-site snow depth was differentially mapped at cm-scale resolution during each field session with structure-frommotion (SfM) photogrammetry, using imagery from a consumer grade unmanned aerial vehicle (UAV). The UAV mapped snow depths were combined with snow pit densities to estimate spatial variations in SWE. See supplement A for SfM acquisition and processing methods.

Dust sampling and analyses
Dust was sampled within respective snow layers when present. Dust layers were identified by visual inspection within the snow pit face. Sampling followed the methodology described in . In summary, a 25 × 25 × 3 cm volume of snow was sampled that encompassed the dust layer of interest. Later in the melt season, multiple dust layers combined into a broad layer of dust (∼10-20 cm) near the top of the snow surface, across which dust was continuously sampled every 3 cm. All samples were bagged, placed in a cooler, and taken to a cold storage facility. Samples were analyzed for BC content using laser light incandescence and filtered for total LAP mass, which was dominated by dust. A more detailed description of this methodology is described in supplement B.

Energy and mass balance modeling
To assess the impacts of dust radiative forcing on snowmelt rates, we used the physically based 1-d snowmelt model, Snobal, to simulate snowpack mass and energy balance (Marks 1988, Marks et al 1999. The model simulates hourly snow accumulation and melt using meteorological, radiation, and precipitation inputs, and has been used extensively to estimate the impacts of DOS-events (e.g. Painter et al 2007. The snowpack is represented as two-layers, one active 25 cm fixed-thickness upper layer, and the second representing the remainder of the snowpack. Snowmelt occurs when the cold content of the layer is depleted and the liquid water holding capacity has been exceeded. Model runs were initiated on 1st April, with snow depth, density, and temperature estimated from snow pit measurements on 28th March and 6th April. The forcing variables, measured in situ at ASP, included net solar radiation, incoming longwave radiation, air temperature, vapor pressure, and wind speed. Precipitation was acquired from Alta Guard, an adjacent (∼80 m southwest) snow safety study plot (Skiles et al 2022). Snow-fraction, snow temperature, and new snow density were calculated based on air temperature. The reference model run represented observed (with dust) conditions. Solar radiation inputs were modified for the different scenarios described below. The difference in snow depletion date, relative to the observed run, was used to quantify impacts on snowmelt timing.
(1) Clean snow scenario: To simulate higher reflectance from clean snow, the additional solar radiation absorbed due to the presence of dust, known as dust radiative forcing (RF) was subtracted from the observed net solar radiation input. Dust RF was quantified following Painter et al (2007) and Skiles et al (2012). See supplement C for more detail.
(2) No new snow scenario: To understand how much new snow (and the ensuing reset to higher albedo) delayed the impact of dust radiative forcing, no new precipitation was input in the model run following peak SWE (2nd May). Additionally, the visible albedo of the snow surface was set to the value it had when dust was first exposed, allowing for a continuous decay of surface albedo after 2nd May.  (2017) and Skiles et al (2018b). The HRRR-STILT footprints were then convolved with a dust erodibility potential product gridded on a 0.05 × .05 • mesh that covers the Western U.S. The dust erodibility potential product estimates the potential for dust to be emitting for any given grid cell. This dust emission model uses soil and land use data to estimate the friction velocity needed to loft dust into the atmosphere. This friction velocity is often referred to as the friction velocity threshold and is a function of the land use and soil type. Gridded friction velocities from the HRRR model are then used to determine whether winds were strong enough to exceed the friction velocity threshold. The GSL water levels were adjusted within the dust erodibility potential product to reflect lake level for the spring of 2022 (lake level = 1277 m asl). Footprints from our backward trajectory model were convolved with the dust erodibility potential product to estimate upwind areas with the highest potential of contributing dust to ASP. See supplement D for additional descriptions of model inputs.

Season summary
The 2021-2022 snow season started with aboveaverage precipitation in October and December. In early 2022 snowfall was infrequent due to atmospheric high pressure until late February when snowfall increased again. Frequent frontal passages and precipitation were observed throughout the months of April and May. Most of these cold fronts were preceded by DOS events driven by sustained high wind speeds across western and northern Utah (as in Steenburgh et al 2012 andNicoll 2012). Late in the season (May), DOS events were sometimes associated with large local convective systems. Dust was most frequently deposited days to hours before snowfall or through wet deposition during snowfall (figures 2(a) and 3). Peak SWE (∼77 cm) occurred at ASP on 2nd May and was 75%-84% of average relative to in situ SWE at SNOTEL sites of similar elevation in the central Wasatch (National Resources Conservation Service 2023). Melt was rapid when combined dust layers became exposed at the surface, which happened periodically between snowfall events in May, and the ASP snowpack was depleted by the first week of June.

Dust and BC concentrations
A record of 16 dust events were observed in the 2022 snowmelt season ( figure 2(b)), 14 of which deposited a combined dust concentration (collected on 20th May) of 1.95 mg g −1 , which is by far the highest concentration ever recorded in LCC since records began in 2009 ( figure 2(b)). Based on cross referencing with particle count data and regional webcams at particulate matter sensors, most visible layers within the snowpack (A-F, table 1) include dust deposition by multiple events, as opposed to dust from a single event. The majority (75%) of the total dust (1.46 mg g −1 ) was deposited by events between 15th March and 20th May. The most notable dust layers in the snowpack were layer B (0.469 mg g −1 , 28th March and 5th April) and layer F (0.444 mg g −1 , 6th-8th May), which together represent almost half of the sampled dust concentrations (table 1). Per-event dust concentrations were estimated based on sampled layer concentrations and the erodibility potential obtained from each dust event simulated by the STILT model (see section 3.4) (table 2). This assumes that all dust deposition at ASP directly corresponded to a simulated dust event and does not account for smaller events that may have gone unobserved. From this, it can be inferred that the three most notable events at Atwater were: (1) 4-5th April (27.8% of combined concentrations), (2) 2nd − 3rd May (21.3%), and (3) 21st April (10.6%, figure 3). The BC concentrations in snow were slightly elevated above those documented in other locations throughout the intermountain western United States (Doherty et al 2014, Skiles and. However, total LAP concentrations were still strongly dominated by dust. The ratio of BC concentrations to total LAP concentrations was generally <0.002. Like dust concentrations, the highest BC concentrations were observed in layer B (57.8 ppb) and F (31.1 ppb) (table 1). Unlike the dust samples, the 20th May BC concentration (98.7 ppb) was significantly lower than the layer-summed BC concentration (167 ppb), suggesting that BC deposited within the snow may have been scavenged by melt water, as observed in Doherty et al (2013), which would limit the impact on melt. The BC concentrations reported here, though, may be biased by the sampling strategy that targeted dust events; BC is emitted and deposited onto the snowpack during the entire winter, and indeed, the highest BC concentration sampled was in the first dust layer. Still, it is useful to understand the partitioning of BC relative to dust during melt, when radiative forcing is most relevant, and results indicate that dust radiative forcing dominates total LAP forcing by 88%-90% (supplement C, figure C1).

Dust source regions
Dust events were most frequently associated with frontal passages, which are often preceded by strong S-SW winds (Steenburgh et al 2012). In the context of dust transport to ASP, these winds transport dust mainly from the Sevier and Tule lakebeds as well as the GSLD . In contrast, strong winds from the northwest transport dust from the GSL lakebed and the northern GSLD to the Salt Lake Valley , and to ASP and the central Wasatch Mountains (Skiles et al 2018b). Particle size distributions in Skiles et al (2018b) and field observations suggest that these regional contributions dominate dust loading on the snowpack.
Our dust modeling framework suggests that 85% of dust deposited at ASP from 15th March to 20th May originated from established dust source regions within Utah's West Desert, outlined in figures 1 and 4. Nearly half (45.2%) of all dust potentially originated  from the GSLD. However, relative to the total area, the dust contribution was relatively low (<.3% for each ∼25 km 2 pixel, figure 4). The GSL lakebed was likely the second largest dust contributor overall (22.5%). Notably, GSL dust came from a relatively small area compared to the GSLD, and therefore was the most efficient source region (contributing 1%-2% per ∼25 km 2 of lakebed area) (figure 4). Part of this is driven by the GSL's proximity to ASP relative to other dust source regions. The combined contribution of the Sevier and Tule Dry lakes was 17.3%, with the Sevier lakebed contributing more strongly to the dust emissions ( figure 4). The remaining 15% of dust originates largely from western Nevada's Carson Sink (5.7%), as well as numerous other smaller contributors (figure 4). It is worth noting that the GSL erodibility fraction was assumed to be 9% within our dust erodibility potential product, which  could be a conservative estimate (Perry et al 2019).
Other dust emission sources, such as the GSLD, had an erodibility fraction of 25%. A higher erodibility fraction (>9%) for the GSL would dramatically shift the distribution in dust source contributions to favor the GSL. Dust source regions exhibited spatial and temporal variability (table 2). The GSL lakebed was likely the primary dust source for the 31st March and 1st-2nd May events, as well as significant proportions of dust for the 15th-16th March and 4th May events. Dust layers with the highest total concentrations (B, F; table 2) were likely dominated largely by dust from the GSLD, which was the largest dust contributor for 4th-5th April, 2-3rd May, 6th May, 7th May, and 8th May events. Dust from the Sevier and Tule dry lakebeds had the highest contributions on 21st April (figure 3), as well as 26th-29th April, when strong winds were predominately out of the southwest.

Radiative forcing and snowmelt timing
Results from snowmelt modeling on observed 'dustpresent' conditions differed by no more than 15 cm from observed SWE, and fell within the distribution of study plot SWE, except for the final observation date (20th May). Generally, the model underestimated SWE when snow was accumulating (April) and then overestimated SWE during melt, due to slightly slower than observed melt rates (figure 5). When dust was exposed in early May, daily peak RF ranged from 100-400 W m −2 and reached a maximum of 502 W m −2 just prior to snow depletion, which is approximately half of total incoming solar radiation (figure 6). When dust RF was removed from observed net solar radiation, to represent clean 'dustfree' snow, the snow season was extended by over 2 weeks (17 d) (scenario 1, section 3.3, figure 5). The accelerated snowmelt shifts both melt timing and magnitude; peak SWE occurred ∼10 d later in the clean snow scenario and was ∼10 cm higher, and clean snow SWE melted at a rate of 1.4 cm d −1 from during ablation (peak SWE to depletion) while dirty snow melted at a rate of 2 cm d −1 .
The 'no new snow' model runs, designed to estimate extended snow duration due to snowfall covering initial dust exposure, indicate that in the absence of snowfall after peak SWE, snow depletion would have occurred 5.5 d earlier then observed ( figure 5).
Analyzing the model outputs indicates that this was due primarily to the albedo reset following the snowfall, lowering absorption of solar radiation, rather than the precipitation magnitude. This implies that the frequency of late season snowstorms is a key variable in predicting the magnitude of accelerated snowmelt from dust radiative forcing.

Key findings
This study was the first to investigate the processes and impacts of DOS across an entire snowcover season in the Wasatch Mountains. Frequent strong winds typically associated with the passage of cold fronts contributed to 14 dust events recorded at ASP between 15th March and 20th May.
Multi-year drought status across Utah's West Desert region (University of Nebraska-Lincoln 2023, notably June 2020-June 2022) may have changed land cover characteristics in dust source regions, for example less perennial vegetation cover (Munson et al 2011), which coupled with frequent springtime frontal passages and low winter precipitation potentially contributed to greater dust emissions. Atmospheric modeling revealed that dust sources were constrained predominately within the West Desert of Utah, with largest contributions from GSLD (45%), the GSL dry lakebed (23%) and the Sevier and Tule dry lakebeds (17%) (figure 4). Large contributions from the GSLD are consistent with isotopic signatures from dust in previous years (Carling et al 2020), and backward trajectory modeling of a dust event in 2017  (Skiles et al 2018b). This consistently high modeled contribution is not unexpected given the GSLD's large surface-area, and its position in the path of both preand post-frontal wind trajectories (figures 1 and 4).
The total end of season dust concentration was 1.95 mg g −1 , with individual events likely accounting for 1%-28% of total loading (table 1). The BC concentrations roughly mimicked dust concentrations but were a minor fraction of the total LAP loading (table 1). Dust layers became frequently exposed starting in early May (figure 5), after which melting of the dust-loaded snowpack was delayed by ∼6 d due to frequent late season snowfall. Once fully exposed, combined dust layers resulted in a maximum radiative forcing of 502 W m −2 , which accelerated snowmelt and shifted the snow depletion date by ∼17 d relative to clean snow conditions, a threetimes greater decline in the length of the snow cover season than previously recorded (Skiles et al 2018b) (figure 5).

Implications of a shrinking GSL
The GSL lakebed is the closest dust contributor to the Wasatch Mountains (figure 1). Our study estimated that nearly a quarter of the sampled dust originated from the GSL lakebed. These results are alarming amid recent field evidence suggesting at least 9% of the GSL lakebed is susceptible to dust emission (Perry et al 2019), implying that even small increases in erodible lakebed area can result in record-high snowpack dust concentrations. Protections for the GSL have been the focus of recently passed state legislation; Utah House Bill 33 allows water users to lease their water rights to maintain lake levels and Utah House Bill 410 allocates funding to maintain the GSL's ecosystem and economy. Policymakers have declined, though, to take up legislation that would set a minimum lake level. If declines in GSL levels continue, accelerated snowmelt may be similar or greater in magnitude, than the 2 weeks we found here. This shift could alter the snowmelt regime to mimic the San Juan Mountains of Colorado more closely, where dust concentrations exhibit high interannual variability, but can exceed 4 mg g −1 (e.g. Painter et al 2012, Skiles et al 2015, which accelerates melt by months rather than weeks . Such changes will contribute to a declining snow cover season, notably in high elevation areas. Fewer days of snow cover and rapid melt will impact the Utah snowsports economy, alter local watershed storages and fluxes, and add complexities to water management across Utah's snow-dependent downstream communities and the GSL.

Data availability statements
The data that support the findings of this study are openly available at the following URL/DOI: 10.5281/ zenodo.7796477. The in situ instrumentation data from Atwater Study Plot can be viewed and downloaded from MesoWest using Site ID ATH20. The Snobal model is open source and can be accessed on GitHub (https://github.com/USDA-ARS-NWRC/ pysnobal). ongoing development of the Snobal model. We aknowledge two anonymous reviewers, whose feedback improved the manuscript. This work was funded by the National Science Foundation Earth Science Division through the Critical Zone Collaborate Network (Award #2012091).