Introduction

Stormwater (SW) is water resulting from precipitation events and melting snow, running off the urban landscape, collecting in storm sewers, and being released into the environment with little to no treatment. Increasing urbanization in cities has sealed soils, removed vegetation, and changed natural drainage paths resulting in increased quantities and flashiness of SW discharging from these landscapes, in conjunction with significantly decreased SW quality (Yang and Toor 2017; Zgheib et al. 2008). Untreated SW is recognized as a significant transport vector for contaminants to enter receiving water bodies (Barbosa et al. 2012; Blecken et al. 2012; Matos et al. 2015; Qinqin et al. 2015). In cold climates, precipitation falls as snow during the winter, introducing an additional form of runoff that depends on snow management practices. Interest in urban snowmelt (SM) quality started in the late twentieth century in places such as Europe and Canada (Schrimpff et al. 1979; Zinger and Delisle 1988), with increasing research since the 1990s (Mayer et al. 1999; Oberts et al. 2000; Viklander 1996, 1999; White et al. 1995). Interestingly, studies typically examine road runoff (Helmreich et al. 2010; Legret and Pagotto 1999; Taka et al. 2017; Westerlund and Viklander 2006) as opposed to runoff from snow storage facilities which can potentially contain snowpack-derived transformation chemicals (Exall et al. 2011a; Shotyk et al. 2010).

The impact of SW depends on the concentration and mixture of contaminants, which are themselves influenced by climatic and land use conditions within the catchment (Aryal et al. 2010; Matos et al. 2015). In cold climates, urban snow is often managed by pushing snow to the sides of the road, application of road salt and/or sand, and collection of snow for delivery to snow storage sites. Prior to collection and within hours of traffic exposure (Glen and Sansalone 2002), the snow accumulates contaminants such as total suspended solids (TSS) (Sillanpää and Koivusalo 2013), metals (Kuoppamäki et al. 2014), organics (Herngren et al. 2010; Howitt et al. 2014), and total dissolved solids (TDS), notably chloride (Reinosdotter and Viklander 2007; Westerlund and Viklander 2008). Whether they bank the road or occupy storage sites, snowpacks (noted as SP herein) are highly permeable media susceptible to heat and mass fluxes (Albert and Hardy 1995; Calonne et al. 2012; Ebner et al. 2016; Löwe et al. 2011). The unique physicochemical traits of snow as a SW matrix can influence the concentrations of contaminants within the pack (Ebner et al. 2016) and create chemical transformation products (Glen and Sansalone 2002).

The form in which contaminants are found is known to effect potential toxic impacts by affecting bioavailability to aquatic organisms. At lower trophic levels, metals can cause DNA damage, growth and biomass reduction, cellular respiration and enzyme dysfunction, inhibition of photosynthesis, and nervous system damage, among other adverse effects (Jaishankar et al. 2014; Zubala et al. 2017). In humans, arsenic, copper, lead, and zinc are associated with health effects including cancer, bone disease, hypertension, DNA and enzymatic dysfunction, nervous system damage, infertility, and organ failure (Chung et al. 2014; Ma et al. 2016; Zubala et al. 2017). The chloride ion from road salts drives toxicity in Lampsilis freshwater mussels (Prosser et al. 2017), while road salts and dissolved Zn were most closely associated with toxicity endpoints in Ceriodaphnia dubia and Oncorhynchus mykiss (Mayer et al. 2011a). Chloride from winter road maintenance is a significant contributor to dissolved inorganic ligands in solution, thus enhancing the complexation of metals and increasing their dissolved concentration (Mayer et al. 2011a; Valtanen et al. 2014). Elevated concentrations of cadmium, copper, magnesium, nickel, and zinc have been associated with elevated chloride ion concentrations in SM (Lazur et al. 2020; Mayer et al. 2011a; Reinosdotter and Viklander 2007). However, even in the presence of chloride, heavy metal burdens in urban SM were observed mostly (> 90% of total metal concentration) as non-dissolved particulates which remained on the snow storage site (Westerlund and Viklander 2011). Despite potentially reducing the acute aquatic toxicity of SM runoff, these contaminants may instead be transferred to adjacent benthic ecosystems (Karlavičiene et al. 2009; Rentz et al. 2011) or into subsurface and groundwater flows (Mohammed et al. 2019; Pavlovskii et al. 2019; Shotyk et al. 2010). In SW models, contamination is considered to be deposited on the urban landscape in a linear or asymptotic fashion between storm events with some/all contaminants being carried into the SW during the preceding rainfall event (Barbosa et al. 2012; LeBoutillier et al. 2000). Mobile substances such as salts are washed first, while heavier or particle-bound contaminants may require high volumes or intensities to flow into sewers (Mayer et al. 2011b). This phenomenon is referred to as the “first flush,” and though it is used in SW to refer to a single-event phenomenon, the mechanisms of pollutant build-up and wash-off also influence spring melt runoff characteristics. As snow melting occurs on warm days, contaminants are preferentially eluted from the SP (Westerlund et al. 2008; Westerlund and Viklander 2006) causing an SM-driven discharge to receiving water bodies. This occurs in both SPs at storage facilities and those undisturbed within the catchment, with the latter uptaking contaminants as governed by land use classes before entering the sewer similarly to summer SW.

While urban SW quality is well-researched, it is often conducted in climates that receive limited or no snowfall (Galfi et al. 2017; Westerlund and Viklander 2011). Spring SM in cold regions may contribute up to 60% of the annual contaminant loading into receiving environments (Oberts et al. 2000). Despite the potential impacts of this loading, the literature body of urban SM quality, transport, and modeling remains limited (Bartlett et al. 2012; Codling et al. 2020; Galfi et al. 2017; Mayer et al. 2011a; Moghadas et al. 2016; Sillanpää and Koivusalo 2015; Taka et al. 2016; Valeo and Ho 2004; Valtanen et al. 2014; Vijayan et al. 2019; Westerlund and Viklander 2011). Though the toxic potential of winter road runoff to vertebrates is well-discussed (Hintz and Relyea 2017; Hopkins et al. 2013; Meland et al. 2010; Sanzo and Hecnar 2006), few studies examine runoff discharging from urban facilities (Exall et al. 2011a, b) or conduct invertebrate bioassays (Mayer et al. 2011a; Prosser et al. 2017). The spring runoff at these facilities is subject to longer holding conditions than road runoff which is removed expediently. Furthermore, spring runoff draining to SW sewers will follow the same transport path as summer SW and uptake similar characteristics from the landscape. This could confound differences between spring SM and summer SW necessary to determining the impact of SM on its immediately adjacent environments. Previous observations indicate that plowed snow is significantly more burdened than undisturbed SPs, containing up to 50% of areal pollutant mass (Kuoppamäki et al. 2014; Moghadas et al. 2015; Sillanpää and Koivusalo 2013; Viklander 1999). Thus, the goal of this study was to assess the quality of urban snow collected and stored in municipal snow storage facilities in the City of Saskatoon (CoS), Saskatchewan, Canada. The CoS is in a cold climate, semi-arid region of Canada and borders both banks of the South Saskatchewan River (SSR) which is the receiving waterbody for city SW. Previous research into the CoS’s SW runoff quality has been conducted (Al Masum et al. 2021; Codling et al. 2020; McLeod et al. 2006; Challis et al. 2021; Popick et al. 2021); however, no previous study has considered SM runoff quality. The SM quality assessment includes analysis of pH, electrical conductivity (EC), total dissolved solids (TDS), total suspended solids (TSS), chemical oxygen demand (COD), total organic carbon (TOC), dissolved organic carbon (DOC), dissolved metals, chloride, and targeted dissolved polyaromatic hydrocarbons (PAHs). In addition, SM toxicity was assessed using Raphidocelis subcapitata and Vibrio fischeri. Lastly, land use classifications are used to determine potential pollutant loadings into the SSR following similar assessments for rainfall-only SW presented by Al Masum et al. (2021) and Popick et al. (2021).

Methods

Study area

The CoS has a total area of 228 km2 and a population of about 330,000 making it the largest municipality in Saskatchewan (Statistics Canada 2021). Based on climate data collected by the Saskatchewan Research Council’s Climate Reference Station (CRS), the average daily temperature ranges between approximately 18.7 °C in August and − 14.7 °C in January, with an annual average of 3 °C. The CoS receives an average of 355 mm of precipitation per year, with ~ 50 mm of precipitation falling between the months of November and March as snow.

There are four snow storage facilities within the CoS, all of which were included in the present study (Fig. S1). These sampling sites include the impermeable site on the southwest border of the CoS (Valley Rd.), one site north of the CoS along Wanuskewin Rd., one site located in the northeast of the city (Central Ave.), and one located within the University of Saskatchewan grounds (USask). The Valley Rd. site has a paved surface, a settling pond, and a designated outlet where the meltwater enters a series of specially designed barriers before being discharged into the SSR. The other three sites do not regulate SM flow and lie adjacent to vegetated wetlands or swales. Storage facility sites are noted with a blue dot, while red dots note SW outfalls which were sampled as part of a parallel study. See Supporting Information (SI) for snowfall information used in the current studied expressed as snow-water equivalents (SWE) (Table S1).

Sampling

Snow was collected from at least one of the four snow facilities on warm days in which SM was being created (Table S2). Snow directly from the pile (SP) sampled four times in April 2019, and SM puddles were sampled eleven times from March to May 2020. One SM sample was obtained on April 2, 2019, and two SP samples were obtained on March 7, 2020, for SP-SM comparisons within site and event. Plastic scoops were used to collect both snow and SM in pre-cleaned 4-L or 25-L Nalgene containers. Snow from the piles was collected from 8–12 random locations on the surfaces and sides of the SPs to create an aggregate sample. The SM samples were obtained from meltwater pools found at the foot of SPs and also collected from 8–12 locations from on-site puddles to create an aggregate sample of the SM. Samples were sealed, labeled, and transported to the USask Environmental Engineering labs for storage at 4 °C. SP samples were melted in their Nalgene containers at 21 °C in the labs for initial analyses before storage at 4 °C.

Previous studies have collected snow by sampling SPs at different depths (Calonne et al. 2011; McLaren 1982), by coring, by simple grab sampling of the top 20 cm of the SP (McLaren 1982), and by scooping SM runoff samples from puddles (Kuchta and Cessna 2009). However, studies sampling SPs are limited, and there is no standard procedure for sample collection available. Due to the mixing of various roadside snows in removal trucks, their deposition on top of old packs, and potential-freeze thaw events causing partial or preferential elution of contaminants, it was assumed that pollutants would not be uniformly distributed in the pile. Previous research on pollutant loading in snow packs found grid sampling to best minimize variations in mass load estimates (Vijayan et al. 2021), though this team used coring to minimize depth uncertainty, which was not performed in the current study.

A Riegl miniVUX1-UAV lidar with an Applanix APX-20 inertial measurement unit (IMU) mounted on a DJI M600 Pro unoccupied aerial system (UAS) platform was used to scan the Valley Rd. facility. The difference between the rasterized uncovered surface, interpolated from the exposed concrete pad around the SP, and the rasterized snow surface was used to estimate SP volume.

Laboratory analyses

Laboratory-based analysis included physicochemical, biological, and toxicity assessments. Physicochemical analyses of samples included pH, TSS, TDS, EC, COD, TOC/DOC, metals, PAHs, and chloride. The SP density (kg/m3) was calculated by packing snow from the April 2, 2019, and April 13, 2019, snow events into a 1-L plastic container (to obtain a unit volume of snow), weighing the tared and full container, melting the snow at room temperature, measuring the volume of meltwater, and then using the difference in volume to calculate density. The biological analyses comprised the enumeration of fecal coliforms, while two toxicity assays were conducted including Raphidocelis subcapitata algae and Vibrio fischeri bacteria.

Physicochemical analyses

The TSS concentration was measured via vacuum filtration using Whatman™ 934-AH™ glass microfiber filters following Standard Methods 2540 (2018). A HACH sensION 156 digital probe was used to measure pH, TDS, and EC. For DOC, all samples were extracted through a 0.45-µm Teflon filter using a 12-mL Luer-Lock syringe. Approximately 40 mL of filtered sample was placed in a glass vial and analyzed using a Lotix combustion TOC analyzer (Teledyne Tekmar, OH, USA) following the manufacturer-provided method. For COD, samples were added to VWR Mercury-Free High-Range (20–1,500 mg/L) COD digestion vials, and the COD was measured using a HACH DR/4000U Spectrophotometer (HACH USA, CO, USA) set to 625 nm following the HACH COD Method 8000 (HACH 2014). Samples were run in duplicate using either 2 mL of sample or 1 mL of sample and 1 mL of distilled (DI) water. Chloride concentrations were determined by Bureau Veritas (Calgary, Canada) using Auto Colourimetry.

For metal analysis, samples were acidified with 0.02 N nitric acid and vacuum-filtered through a 0.45-µm nitrocellulose filter. A 100-mL sample volume was passed through the filter, and the filtrate was collected in a Nalgene container. Samples were analyzed using inductively coupled plasma mass spectrometry (ICP-MS) at the USask Department of Geological Sciences or the USask Toxicology Centre. The methods at the Department of Geological Sciences included the use of a PerkinElmer 300D ICP-MS, diluting samples 20 × prior to analysis, and using a custom calibration standard (SCP Science) for blanks and standards of 10, 50, and 100 ppb. The certified reference material was NIST-SRM1643f. At the Toxicology Centre, samples were analyzed using an Agilent 8800 ICP-MS QQQ Triple Quadrupole mass spectrometer (Agilent, Santa Clara, USA). OminiTrace Ultra nitric acid (HNO3) (w/w) (Millipore-Sigma, Ontario, Canada) was used for blanks, standards, and sample solutions. High-purity standard stock solutions (1000 mg/L) were purchased from Delta Scientific (Mississauga, Canada). The calibration standard solution, containing 22 multi-elements, was supplied by SCP Science (Quebec, Canada). The standard reference material, natural water 1640a, was supplied by National Institute of Standards and Technology (NIST) (Gaithersburg, USA).

Samples for PAH analyses were pre-filtered (Whatman™ 934-AH™ glass microfiber filters) to remove high TSS concentrations prior to solid-phase extraction (SPE) to eliminate clogging of the filter. After filtration, 2 mL of chloroform was added per 1 L of sample as a preservative, with the samples stored in amber glass bottles at 4 °C prior to extraction. A deuterium-labeled internal standard mix (500 mg/L of acenaphthene-d10, chrysene-d12, and phenanthrene-d10 in acetone) provided by Sigma-Aldrich (Oakville, ON) was added to the sample at a 10 µL/L ratio. Before sample addition, Waters Oasis HLB 500 mg extraction columns were pre-conditioned using 3 mL dimethylene chloride (DCM), 3 mL methanol (LC–MS grade), and 3 mL 18.2 MΩ-cm ultrapure water (EMD Milli-Pore Synergy® system, Etobicoke, ON). Up to 500 mL of each SM sample was vacuum extracted through the column at a rate of 1 drop/second. After extraction, the column was washed with 3 mL of 5% methanol in water and air-dried with suction for 10–30 min. If column elution was not possible immediately following extraction, the columns were stored at − 20 °C. Columns were eluted twice with 5 mL of DCM and once with 5 mL of methanol. The eluate was collected in glass vials, reduced to a volume of 10 mL with a gentle stream of nitrogen gas, and split into two 5-mL portions (one portion used for a parallel research study). The aliquots were reduced to near dryness under nitrogen and reconstituted in 0.5 mL nonane. The reconstituted sample was added to a gas chromatography vial and stored at 4 °C.

Samples were analyzed for PAHs using gas chromatography-mass spectrometry (GC–MS) using A Thermo Scientific Trace 1300 or 1310 gas chromatograph coupled with a Thermo ISQ 7000 single quadrupole or a Thermo QExactive quadrupole-Orbitrap hybrid mass spectrometer, respectively. Helium (99.999% purity) was used as the carrier gas to separate the PAHs on an Agilent DB-5 ms (60 m × 250 μm I.D., film thickness 0.1 μm) fused silica capillary column. Both instruments were operated in full scan mode, and data were analyzed using an isotope dilution workflow, i.e., areas of target compounds were normalized to the areas of recovered deuterium-labeled standards. A seven-point calibration curve, along with extraction and solvent blanks was run with each batch of samples. Limits of detection and quantification are included in Table S3.

Biological and toxicity analyses

The first toxicity assessment was the inhibition of the luminescence of V. fischeri aquatic bacteria. The method is described in EPS RM/1/24 (Environment Canada 1993) using a Microtox M500 machine (Modern Water, DE, USA). Phenol (60 mg/L) was used as the positive control and 20% sucrose as both the diluent (SD) and osmotic adjustment solution (OAS) as recommended for freshwater samples or increased test sensitivity to metals. The SD was also used as the negative control. This method was adapted to measure the luminescence inhibition of dilution series in 96-well microplates using an OptimaSTAR plate analyzer. Prior to the test, 10 mL each of the samples, OAS, and SD were refrigerated for 1 h before being transferred to 25-mL beakers. Plates were run both inoculated and uninoculated to correct for background luminescence. To inoculate the wells, 1 mL of Microtox Reconstitution Solution was mixed with a vial of lyophilized V. fischeri strain NRRL B-11177 (Modern Water, USA); 0.3 mL of this solution was further diluted with 3 mL of SD. This solution was placed in a multichannel pipette reservoir on ice. Placing the plates on a paper towel over ice and using a multi-pipettor, plates were inoculated with 8 µL of the bacteria solution. Luminescence readings were taken immediately, after 5 min, and after 15 min, with plates remaining on ice between measurements. Due to the configuration of the plate, samples were run as seven-step dilutions, while the phenol was a six-step dilution series. Values for both samples and phenol are reported as the EC50 and EC10 relative to the negative control at 15 min.

The second bioassay tested the 72-h chronic toxicity of R. subcapitata green algae following the EPA Method 1003.0 (EPA 2002). Prior to analysis, a dilution series of inoculated algae stock was used to establish a linear relationship between chlorophyll expression and cell count. The chlorophyll in R. subcapitata was found to relate linearly with cell count (R2 = 0.9996), and this was used as a parameter for growth inhibition in that species. Samples were prepared in 1-mL wells of 24-well microplates in a configuration of four replicates of negative control and a five-step dilution series. Wells were inoculated with 100 µL of inoculum (1,000,000 cells/mL) to meet appropriate cell densities (10,000 cells/mL in wells at the start of the test). Using fluorescence as a proxy for algae cell count, growth inhibition was measured as a function of fluorescence at 0 h and at 72 h. Cell counts were observed directly from one random control well per plate at 0 h to ensure proper inoculation. Both algae fluorescence and cell counts were read using the Tecan Spark® multifunction plate reader (Tecan Trading AG, Switzerland). Sample concentration in wells ranged from 100 to 6.25%, both inoculated and uninoculated to account for background fluorescence. Initially, the EC50 at 72 h was calculated; as the EC50 was not observed at the full concentration for all samples, the EC10 was additionally calculated (see “Statistical analysis”).

Land use analysis

The volume of snow transported to municipal storage facilities can be estimated using the volume of snow that is plowed from local traffic surfaces. These snowbanks are removed and hauled to municipal snow storage facilities where they are piled outdoors until they melt during spring and/or summer. The volume of snow plowed to the side of the road is assumed to be equivalent to the volume ultimately transported to snow storage sites, thus providing a quantitative estimate of snow volumes managed at the CoS’s snow storage sites. However, it should be noted that this volume could be partly lost to sublimation or melting over the winter, and an assessment of SP volume immediately prior to melt will most accurately reflect the impact of that melt period. Moreover, facility snow likely contains approximately half of all winter contaminant mass loads (Sillanpää and Koivusalo 2013). Though this snow may not be indicative of non-traffic related land use impacts, it nonetheless represents a significant portion of potential urban stormwater loading into receiving environments. Land use breakdowns for the CoS are included in Table S4 which include each of the local subcatchment characteristics.

As SM discharges from the storage facilities are not currently measured in the CoS, event-mean-concentrations (and corresponding site-mean-concentrations) could not be calculated as more typical for rainfall-based stormwater events (e.g., Popick et al. 2021). Seasonal loading can be estimated, however, using the seasonal mean concentrations measured at each site in relation to the volume of estimated SM discharging from that facility. The runoff volume, in snow-water-equivalents (SWEs), was calculated using SP volume estimation and lab-measured values of late-winter SP density. Currently, Valley Rd. was selected for UAS-LIDAR imaging as it is the main snow facility in the CoS and receives approximately half of the CoS’s plowed snow. UAS-LIDAR imaging of the snow facility was completed to determine the Valley Rd. site snow volume (50% of the CoS), and the remaining volumes were estimated at 25% each for the Wanuskewin Rd. and Central Ave. facilities (Table 3). The USask storage facility is markedly smaller than these other city-wide facilities; thus, the volume of snow was estimated based on GIS-based maps in comparison to the Valley Rd. site. Using the SWEs and site-specific physicochemical and contaminant concentrations, seasonal loadings were estimated for the 2020 SM season (Table 3) as presented in the “Results and discussion”.

Statistical analysis

Statistical analyses were run using GraphPad Prism 9 (GraphPad Software, San Diego, CA). Gaussian distribution of SP and SM datasets were assumed, though these sets were analyzed separately and later compared. Ordinary one-way analysis of variance (ANOVA) was run to examine any site- or event-related variance in runoff characteristics (p ≤ 0.05) along with Tukey’s post hoc test such that all samples would be intercompared. Pearson’s correlations (with two-tailed p value analysis to determine if correlation was due to random sampling) were run to examine any potential relations between pH, TDS, chloride, and EC; chloride and dissolved metals; and the EC10 values of R. subcapitata and V. fischeri against individual SW parameters and metals and PAH species.

To prepare the algae and Microtox data for statistical analysis, the background chlorophyll or luminescence of uninoculated samples was subtracted from that of inoculated samples to remove any baseline. The exponential growth rate between the initial and endpoints was calculated for the algal bioassay. Both algal growth rates and bacterial luminescence values, respectively, were normalized as a percent of the average chlorophyll (R. subcapitata) or luminescence expression (V. fischeri) observed in the negative control. Within GraphPad Prism, four-parameter logistic regression was used to fit dose–response curves to the growth rate and luminescence inhibition data, respectively, to obtain the EC50 and its 95% confidence interval (CI) for each analyzed sample. As the EC50 was not observed for many samples, the EC10 and its 95% CI was also interpolated from the curve. The ECx is defined as the concentration of the bulk SW sample required to reach the endpoint.

Results and discussion

The “Results and discussion” will be considered in three sections including (1) “Physicochemical parameters,” (2) “Toxicology assessments,” and (3) “Land use management.” The physicochemical analyses section will include a discussion of results in three subsections covering the pH, TDS, and EC (“pH, total dissolved solids (TDS), and electrical conductivity (EC)”); DOC, COD, and TSS (“Dissolved organic carbon (DOC), chemical oxygen demand (COD), and total suspended solids (TSS)”); and metals and PAHs (“Metals and polycyclic aromatic hydrocarbons (PAHs)”). The toxicity section presents the two toxicity assays R. subcapitata (“R. subcapitata bioassay”) and V. fischeri (“V. fischeri bioassay”). The final land use section will be used to discuss the impacts of land use on the individual catchment area estimated pollutant loadings into the SSR. Figure 1 presents a map of the CoS and a summary of all physicochemical, metals, and PAH data in the form of box and whisker plots. Further details will be discussed in each of the following subsections. Results by event are included in Table 1 (see Table S5 for results by site; Tables S6-S7 for full SP and SM results).

Fig. 1
figure 1

Map of the City of Saskatoon (CoS) indicating current snow facilities (red). Stormwater (SW) outfalls are indicated with blue dots, which are out of the scope of this study. See Table S2 for timing and distribution of sampling events. Units are contaminant-specific and error bars indicate 95% CIs

Full size image
Table 1 Physicochemical parameters for 2019 snow pile (SP) and 2020 snowmelt (SM). Values are average (standard deviation) for all sites sampled on each occasion. Yearly values are shown in bold. Averages of individual events
Full size table

Physicochemical parameters

pH, total dissolved solids (TDS), and electrical conductivity (EC)

The average pH for SP samples was 9.00 ± 0.54 and 8.07 ± 0.31 for SM samples (Table 1). No variation was observed across either SP or SM by site or event for pH with one exception (p ≤ 0.05): the March 7, 2020, SP samples were significantly different from all 2019 samples. Generally, pH was significantly higher in SP samples than in SM samples (p ≤ 0.05). Curiously, USask SP samples possessed the lowest pH of all SP samples on April 2, 2019 (pH = 8.76), and March 7, 2020 (pH = 7.56) (Table S6), with pH increasing over the 2019 season (which was not observed at other sites). Due to the low SP sampling size, more field seasons are needed to confirm if lower pH and melt trends are characteristic to the site. Compared to other sites sampled on the same dates, Central Ave. SM typically had lower pH values, while Wanuskewin Rd. SM had higher pH values (Table S7). For SP samples, pH had strong inverse correlation with TDS (R =  − 0.938) and EC (R =  − 0.937), though this was driven by TDS content in March 7, 2020, SP samples. This correlation was not observed in SM samples with R =  − 0.243 and R =  − 0.247, respectively.

The EC and TDS are related parameters so will be discussed together herein. The average EC for SP and SM were 124 ± 131 µS/cm and 3,561 ± 4,017 µS/cm, respectively, while the average TDS were 59.4 ± 64.1 mg/L and 1,900 ± 2,290 mg/L, respectively (Table 1). No site variation was observed for EC nor TDS for SP or SM (p ≤ 0.05). Event-wise, with respect to SP and similarly to pH, the March 7, 2020, samples differed significantly from 2019 samples for both TDS and EC (p ≤ 0.05). With respect to SM, variation was observed between March 26, 2020, and April 2, 2019 (TDS only), April 17, 2020, April 23, 2020 (TDS only), and all samples after April 28, 2020 (inclusive) (p ≤ 0.05). No significant differences were observed between any other events. Except the March 7, 2020, SP Valley Rd. and USask samples, EC and TDS results for SP samples were 1–2 orders of magnitude lower than those of SM samples. Most SM samples display elevated EC and TDS relative to summer SW, with the March 26, 2020, Valley Rd. sample having a TDS sevenfold and an EC sixfold greater than those of 2019 SW (collected as part of a parallel study). Though no seasonal trends were observed during the 2019 season due to elution of TDS from the SPs, the TDS in 2020 SM samples remained elevated throughout the sampling season (Table 1), driven by individual site-specific contaminant spikes (Table S7), with a pulse of 1,721 mg/L TDS and 3,350 µS/cm observed at Valley Rd. as late as May 1, 2020.

Measurements of chloride were performed on select samples (Table S8). Generally, chloride comprised 39–67% of TDS concentrations, except the April 28, 2020, USask sample, where it only comprised 11% of TDS (the value returned to a set-typical value on May 5th). Compared to summer SW chloride concentrations (141 ± 56 mg/L), concentrations on April 2, 2019, SM were similar (150 mg/L), but chloride was more abundant in less-melted March 7, 2020, samples. Mirroring TDS and EC concentrations, USask SP contained more slightly more chloride than SM, while at Valley Rd., significantly more chloride was observed in SM. Making this observation within chloride as opposed to TDS is interesting: 62% of the TDS burden in the Valley Rd. SM sample is chloride (46% in SP), while this proportion is only 41% in USask SM (47% in SP). The sites are evidently within different stages of preferential elution early into the sampling season, likely due to the unique impervious surface at the Valley Rd. site. Throughout the 2020 spring melt, the largest three pulses of chloride are observed at this site (2,600–5,800 mg/L), with the next largest three pulses observed at Central Ave. (1,100–1,700 mg/L). As chloride encourages dissolved partitioning of metals, correlations between chloride and metals concentrations were examined. There is some correlation between the presence of chloride and dissolved concentrations of arsenic (R = 0.594), cadmium (R = 0.489), manganese (R = 0.684), nickel (R = 0.658), and strontium (R = 0.482) in tested samples, the implications of which are further discussed in “Toxicology assessments.”

A 2-week melt period was observed prior to sampling in April 2019, which was not observed in March 2020. Additionally, the seasonal sampling events on April 2, 2019, and March 7, 2020, took place on cooler days relative to other events. Both sampling conditions may explain the relative similarity in pH (as temperature governs the chemical composition of the SW matrix), but differences in EC and TDS. As dissolved contaminants in snow are preferentially eluted with meltwater out of SPs (Viklander 1996), it is expected that EC and TDS would be elevated in SM relative to SPs. In cold climates, these parameters are a major concern for toxic impact due to the practice of road salting (Exall et al. 2011b; Mayer et al. 2011a; Murphy et al. 2014). Dissolved salts, especially chloride, affect the partitioning of metals, increasing those in the dissolved phase (Bäckström et al. 2004; Galfi et al. 2017; Mayer et al. 2011b; Reinosdotter and Viklander 2007). A previous study conducted over a period of 2 years found significantly elevated EC, TSS, TOC, copper, and zinc in melt season samples compared to rain season samples (Helmreich et al. 2010). However, the chemodynamics of snow are complex. For example, when sand is used as an anti-slip agent (often in combination with NaCl or MgCl2, CaCl2 salt), SM pH and TSS are elevated. This may counteract the dissolving effect of chloride by providing more particulate surfaces for contaminants to bind (e.g., metals, PAHs). Some of the highest SM pH values in this study occurred in Valley Rd. samples, the impermeable surface of which would retain TDS and TSS in SM runoff.

Dissolved organic carbon (DOC), chemical oxygen demand (COD), and total suspended solids (TSS)

Despite the presence of geese contributing organic waste onsite throughout the melt season (personal observation of H.P.) at the USask and Central Ave. sites, no site-wise significant variation in DOC nor COD was observed for either SP or SM (p ≤ 0.05). The average DOC of SM (6.93 ± 6.25 mg/L) approximately double than that of SP (3.12 ± 2.73 mg/L). Average COD did not follow a similar trend, with SP COD (460 mg/L) slightly higher than SM COD (353 mg/L). There was significant variation in COD obtained from both pile (SD ± 230 mg/L) and melt (SD ± 293 mg/L) samples. Unsurprisingly, there was little correlation between COD and DOC concentrations in SP (R =  − 0.068); conversely, COD and DOC were correlated in SM (R = 0.668). The highest DOC values in SP occurred in USask samples (4.45 ± 1.75 mg/L; Table S5) for all 2019 events. In SM, the highest DOC values occurred in Valley Rd. samples on March 7, 2020 (20.4 mg/L), and March 26, 2020 (27.4 mg/L). The highest COD values were measured in April 13, 2019, pile samples (684 ± 7.4 mg/L) and March 26, 2020, Valley Rd. sample (645 mg/L ± 301 mg/L). For SP, event-wise differences in DOC were observed between March 7, 2020, and April 13, 18, and 24, 2019 (p ≤ 0.05); the sole differences in event-wise COD were observed between April 13, 2019, and all other SP samples (p ≤ 0.05). For SM, one event-specific variation in DOC occurred between March 7, 2020, and May 12, 2020 (p ≤ 0.05); only the March 26, 2020, event otherwise differed from April 2, 2019, April 10, April 17, April 23, April 28, May 5, and May 12, 2020, samples (p ≤ 0.05). Only the March 26, 2020, and May 12, 2020, samples had significantly different COD from one another. The concentrations of both DOC and COD fluctuated between sites and sampling events for both SP and SM with no clear seasonal trend, though no COD values exceeding 300 mg/L were measured in SM samples after April 23, 2020.

The overall range of TSS varied from 31 to 2,982 mg/L in SM and 506 to 3,513 mg/L in SP. Generally, SP samples contained more TSS than SM samples, with an average TSS (1,662 ± 750 mg/L) four times that of SM (400 ± 387 mg/L) (Table 1). It is not unexpected that coarse TSS remain within the SP as opposed to becoming entrained in SM runoff (Westerlund and Viklander 2011). No significant difference related to site nor event was observed for any samples. The highest TSS was observed on April 13, 2019, SP. Elevated TSS values in SM were observed on April 17, 2020 (USask, 1,800 mg/L), and April 28, 2020 (Central Ave., 2,982 mg/L), coinciding with warm days in which higher flow more capable of transporting more, and potentially larger, particles was expected. Interestingly, the SM trend in TSS concentration did not decrease as the spring season proceeded. Though COD and DOC were correlated in SM and not SP, this observation was reversed with respect to TSS, with COD and TSS correlated in SP (R = 0.593) but not SM (R = 0.143).

Larger, less chemically reactive particles remained within the SP, which is consistent with previous findings that large particles are deposited onsite and not transported with meltwater (Viklander 1999). The TSS likely experienced delayed releases from the SP after extended periods of warming melted the surface of the SP and transported those particles through the pack. This would also explain the relationship between TDS and TSS values at this site, where peaks in TSS concentration were observed in the days following TDS peaks (March 26th–March 29th, April 17th–April 20th, April 23rd–April 28th, all 2020 season). Relative to the preceding three sampling dates, elevated TSS was observed at Valley Rd. through April 28, May 1, and May 5, 2020, though the concentration had dropped almost half by May 12, 2020. Similar TSS behavior was observed by Westerlund and Viklander (2011), who noted that over 90% of the metal burden in SPs was in particulate form and followed the same trends as TSS. It is possible the high TSS contained particulate COD, which has been previously observed in snow samples (Westerlund and Viklander 2011). Moreover, particulate COD tends to dissolve in meltwater (Viklander 1996), and some partial dissolution may explain the slightly higher variability in SM (SD ± 101 mg/L) relative to snow (SD ± 79 mg/L).

Metals and polycyclic aromatic hydrocarbons (PAHs)

The average sum of measured metal species was 559 ± 187 µg/L in SP and 725 ± 412 µg/L in SM (Fig. 2 with concentrations of individual elements included in Tables S9-S10). Concentrations of dissolved aluminum, iron, copper, lead, and zinc were 32–68% higher in SP samples, though SP concentrations were comparable to those in summer SW obtained as part of a parallel study (Popick et al. 2021). The largest overall contaminant peaks occurred in Valley Rd. SM on March 26, 2020 (1,455 µg/L), April 17, 2020 (1,988 µg/L), and April 23, 2020 (1,654 µg/L). All three of these contaminant peaks were accompanied by discharges of manganese measuring 575–919 µg/L. Other species-specific pulses in SP included aluminum at 401 µg/L (Valley Rd., April 13, 2019), arsenic at 26.2 µg/L (Wanuskewin Rd., April 18, 2019), copper at 135 µg/L (USask, April 2, 2019), lead at 7.50 µg/L (USask, April 24, 2019), and selenium at 15.8 µg/L (USask, April 2, 2019). In SM, species-specific pulses include mercury at 0.18 µg/L (Central Ave., May 5, 2020) and zinc at 581 µg/L (Valley Rd., March 29, 2020). In SP and SM samples from April 2, 2019, and March 7, 2020, dissolved metal concentrations in SM were similar, but March 7, 2020, SP concentrations were approximately 67% of those observed in April 2, 2019, SP.

Fig. 2
figure 2

Sum of the analyzed suite of dissolved metals for snow piles (SPs; top) and snowmelt (SM; bottom). Note that the individual metal concentrations can be found in the SI Tables S9 and S10

Full size image

While metals are naturally occurring geogenic elements, with their presence in sediments and soils predating industrialization (Owca et al. 2020), their mobilization in the urban environment can be markedly increased by human activities (Soto et al. 2011; Galfi et al. 2017; Sakson et al. 2018). Concentrations of Al, Cu, Pb, and Zn exceeded CCME Guidelines for Aquatic Life in almost all SP samples. Threshold-exceeding pulses were less ubiquitous in SM samples, with the exception of Cr and Cu, which were universally elevated. Comparable concentration ranges of Cu and Zn across SP, SM, and summer stormwater samples (Popick et al. 2021) indicate a probable geogenic source, despite their elevated concentrations relative to CCME guideline thresholds. Additionally, the presence of chloride in SM was not observed to induce dissolved metal concentrations in SM greater than summer stormwater (Popick et al. 2021). Previous research by Westerlund and Viklander (2011) indicates that a major fraction of metals (> 90%) remain in particulate phase in SM. The higher dissolved metals in SP samples may be due to a greater total metal burden and lower TDS concentration compared to SM, encouraging dissolved-phase partitioning. However, this is merely speculation as the dissolved fraction of each metal cannot be determined as total metals burden was not measured.

The average of measured dissolved ΣPAHs were 77 ± 157 µg/L in SP and 0.587 ± 0.962 µg/L in SM (Fig. 3 with individual species in Table S11). Generally, dissolved PAHs in SP were 2 orders of magnitude greater than those detected in SM. With the exception of the Valley Rd. April 24, 2019, sample, at least 20 µg/L ΣPAHs were detected in all SP samples. SM samples typically contained < 2 µg/L ΣPAHs, with the exception of USask April 23, 2020, at 4.82 µg/L and Valley Rd. March 29, 2020, at 1.14 µg/L. The most notable PAH discharge of 565 µg/L occurred in SP samples from Wanuskewin Rd. on April 13, 2019, including notable discharges of pyrene, phenanthrene, and anthracene, while the largest pulses in SM occurred in Valley Rd. March 26, 2020 (1.14 µg/L), and USask April 23, 2020 (4.82 µg/L), samples. Nearly all SP samples exceeded the CCME guidelines for fluorene, benzo[a]pyrene, pyrene, phenanthrene, and anthracene, while 50% of SM samples exceeded guideline thresholds for benzo[a]pyrene, pyrene, and all except one exceeded the threshold for anthracene.

Fig. 3
figure 3

Sum of the analyzed suite of dissolved PAHs measured in snow piles (SPs; top) and snowmelt (SM; bottom). Note the graphs are employed on differently scaled vertical axes and the individual PAH concentrations can be found in the SI Table S11

Full size image

High concentrations of PAHs in SW are often correlated with high concentrations of heavy metals, indicating that roadways are also a major source of PAHs, mainly as vehicle exhaust, diesel soot, or from vehicle and pavement degradation (Aryal et al. 2010; Legret and Pagotto 1999; Moghadas et al. 2015). Considering that the Wanuskewin Rd. April 13 2019, sample contained comparable DOC and COD to other April 13, 2019, samples, its PAHs are likely traffic-derived. Sampling at this site did not disproportionately target contaminated snow relative to other sites or events; it is possible that oil-enriched particles were entrained in the sample and repartitioned into the dissolved phase when melted; however, previous observations of the same suite of PAHs noted 88–95% of species remained particle-bound in SW samples (Mayer et al. 2011b; Vijayan et al. 2019). This would be consistent with observations of preferential elution, in which the heavier and less polar compounds are the last to leave the SP. The April 13, 2019, Wanuskewin Rd. sample may be the result of selecting a localized patch of pollution within the SP; even if this magnitude of pollution only occurred at one location onsite, the effluents may still be highly concentrated even after mixing. A pollution peak may also have been occurring at this site alone on the sampling day. The phenanthrene, pyrene, and anthracene found in this sample are likely compounds derived from coal tar, roadway degradation, and vehicle emissions which contribute the bulk of pollution in roadside snow banks (Kuoppamäki et al. 2014; Moghadas et al. 2015; Vijayan et al. 2019; Viklander 1999).

Toxicology assessments

R. subcapitata bioassay

Significant toxicity was generally not observed in either SP or SM samples, with the calculation of an EC10 necessary to quantify toxic response for many samples (Table 2). It is assumed that chloride might have played a significant role in the osmotic stress experienced by R. subcapitata. While all samples inhibited growth at 100% concentration, few inhibited growth rates more than 50%. Furthermore, all serial dilutions below and including 50% often displayed stimulation instead of inhibition relative to the negative control. Toxicity did not seem related to any site-specific sample characteristics, though toxic responses were observed for all tested April 17, 2020, samples. The most toxic response for which a 95% CI could be calculated was the USask SM sample on this date (EC50 at 43.7% sample concentration; 95% CI 39.5–48.2). This sample most remarkably contains the highest dissolved copper discharge of all SM samples (Table S10). The April 17, 2020, Central Ave. sample generated the second-lowest EC50 threshold (54.3% sample concentration), but no CI could be calculated. This sample is not particularly remarkable compared to inter-event samples (the Valley Rd. site has higher TDS and chloride readings), inter-site samples (the April 23, 2020, Central Ave. sample has similar quality measurements but no detected EC50), or the entire dataset (no contaminant spikes were measured in this sample). However, it does possess the greatest dissolved zinc concentration of any April 17, 2020, sample (119 µg/L). These toxicity results seem to agree with previous observations that idiosyncratic sample mixtures are likely providing additive, synergistic, and antagonistic effects which cannot be attributed to a single species (Fulladosa et al. 2005a, b; Tsiridis et al. 2006).

Table 2 Seventy-two-hour growth inhibition EC50 and EC10 for R. subcapitata and 15-min luminescence inhibition for V. fischeri (Microtox). Results are presented as percent of total sample concentration required to read the endpoint. EC50 and EC10 values were generated using GraphPad Prism as a dose–response curve (see “Methods”) of, in the case of R. subcapitata, the initial (t = 0) fluorescence observed after 72 h and, in the case of V. fischeri, the initial (t = 0) luminescence observed after 15 min. EC10 values were interpolated from relating 10% luminescence inhibition to sample concentration on a standardized curve (see “Methods”)
Full size table

Organic matter and TSS have been linked to acute toxicity as bound PAHs and metals cause mortality at acutely toxic concentrations and can cause adverse, potentially cumulative, effects at chronic concentrations (Barbosa et al. 2012; Glen and Sansalone 2002; Rossi et al. 2013). Concentrations of PAHs exceeding CCME guideline thresholds (see above) did not contribute to observable toxicity in SP over SM samples. This observation is not entirely unexpected, as Bragin et al. (2016) found that PAHs did not contribute significant growth inhibition in R. subcapitata. Despite copper and zinc commonly causing toxicity in aquatic organisms (Babich and Stotzky 1978; Mayer et al. 2011b) alongside notable concentrations in the most toxic samples, no correlation between copper or zinc concentrations and algae toxicity was observed in this study (R = 0.090 and R = 0.263, respectively). As R. subcapitata can acclimate to aquatic concentrations between 0.5 and 100 µg Cu/L (Bossuyt and Janssen 2005), local algae species may accommodate higher geogenic background levels. Brix et al. (2010) identified relatively low zinc toxicity risk for brief, 1-h events and relatively significant toxicity for chronic exposures, which could have toxicity implications for spring melt runoff from galvanized SW pipes. In cold climates, road salts can contribute significantly to toxicity and affect metal toxicity in winter and spring runoff. In a study examining the toxicity of winter highway runoff, undiluted samples containing high road salt observed no survival in D. magna, V. fischeri, Ceriodaphnia dubia, or Oncorhynchus mykiss (Mayer et al. 2011a). Mayer et al. (2011b) note that sodium chloride and calcium chloride are commonly used de-icers in Ontario, while the CoS uses potassium chloride and magnesium chloride for de-icing. Local salt selection should be noted when assessing the toxic impact of SM: calcium and sodium chlorides produce greater toxic responses in O. mykiss than magnesium chloride (Hintz and Relyea 2017), while other species may be predominantly impacted by magnesium chloride (Hopkins et al. 2013).

V. fischeri bioassay

Similar to R. subcapitata, displayed some degree of luminescence inhibition in samples at 100% concentration (Table 2). However, inhibition was not detected in all samples, and stimulation relative to the negative control was often observed in serial dilutions. No remarkable site- or event-related trends in toxicity were observed, though the survival rates in diluted sample wells created difficulties in calculating IC50 with 95% CI for many samples. The April 23, 2020, Valley Rd. sample displayed the highest toxic potential (15-min IC50 = 8.3% sample concentration); this pulse notably had the highest concentration of dissolved manganese (919 µg/L) and the second-highest concentrations of dissolved mercury (0.15 µg/L) and chloride (2,900 mg/L).

Fulladosa et al. (2005a, b) found metals to be toxic to V. fischeri in the order of mercury > silver > copper > zinc, while cobalt, cadmium, chromium(VI), arsenic(V), and arsenic(III) showed no significant toxicity. However, in this study, as with algae, no significant correlations were noted between dissolved metals and toxic response. Moderate correlations were found between toxicity and chromium (R = 0.453), manganese (R = 0.480), nickel (R = 0.479), and strontium (R = 0.354). In testing binary mixtures of metals, Fulladosa et al. (2005a, b) found copper-zinc mixtures to be additive, while Tsiridis et al. (2006) observed a synergistic effect. Interestingly, the latter comments that all copper-zinc IC50 results differed significantly from theoretical predictions; when combined with mixtures of humic acids, the toxicity of the solution decreased relative to the metals-only mixture. Interestingly, most of the aforementioned metal species were found to be significantly correlated with chloride concentrations (see “Physicochemical parameters”). The complexity of sample mixture is potentially reflected in these results. Chloride concentrations have been implicated as the dominant driver in SM toxicity relative to metals and PAHs (Mayer et al. 2011a; Prosser et al. 2017); they appear to have increased the influence of select metals on toxicity, as previously observed (Bäckström et al. 2004; Reinosdotter and Viklander 2007). However, despite the impact which would be expected from, for example, the Valley Rd. March 26, 2020 (5,800 mg/L chloride) or the April 23, 2020, sample (2,900 mg/L chloride, 919 µg/L manganese; 0.15 µg/L mercury), none was observed. That antagonistic mechanisms against chloride or metal toxicity seem apparent in samples which is most easily indicated by comparing R. subcapitata and V. fischeri data: EC50 thresholds for algae often occurred at greater sample concentrations, though algae should be more susceptible to salt toxicity relative to V. fischeri (Cook et al. 2000). Previous studies which found significant salt toxicity sampled winter road runoff in the Ontario (Mayer et al. 2011a; Prosser et al. 2017). Though carbonate-rich soils are local to the Saskatoon area, with the CoS reporting a river water hardness of 191 mg CaCO3/L on the municipal website, Prosser et al. found 3,159 mg/L of chloride impacted 50% mortality in freshwater mussels (Lampsilis siliquoidea) at a water hardness of 237 mg CaCO3/L (Prosser et al. 2017). As local water hardness therefore would explain the lack of toxic chloride impact in this study, it is probable that the aging of SPs and its onsite SM puddles alter their toxic potential relative to fresh road runoff.

Land use management

To estimate the quantity of snow at CoS facilities prior to the melt season, UAS-LIDAR imagery of the Valley Rd. facility was obtained on March 13, 2020, with a snow estimate of 151,910 m3 (see “Methods”). The UAS-LIDAR system possesses a vertical accuracy of ± 0.09 m (Harder et al. 2020) which produces a volumetric uncertainty of ± 3,442 m3. Assuming the Valley Rd. facility receives half of the CoS’s plowed snow (pers. comm.), a city-wide facility volume of approximately 300,000 m3 was obtained. Using samples taken from Valley Rd., a SP density of 520 ± 20 kg/m3 was determined. This density value is typical for snow which has been partially compacted, melted, and refrozen such as managed SPs. This SP density was then used in conjunction with a water density of 1,000 kg/m3 to produce an estimate of 79,000 ± 4,199 m3 in SWE for the Valley Rd. site and 34,500 ± 1,834 m3 at each of CoS other facilities. Similarly, the USask site snow volume was estimated and the SWE determined to be 10,000 ± 532 m3. Assuming these volumes were discharged to the environment through the melt season, the 2020 seasonal loading estimate was calculated using the average measured concentration of TDS, TSS, COD, DOC, copper, chromium, manganese, nickel, lead, selenium, zinc, and ΣPAHs (Table 3). It should be noted that this loading is representative of CoS traffic-related surfaces; however, SM runoff within catchments is likely to follow the same transport paths as SW. However, contaminant concentrations in this non-road snow were not determined in the current study. As 35–85% of SM has been observed to infiltrate on the Canadian Prairies, even over frozen soils (Mohammed et al. 2019), it is not unreasonable to assume roadside SPs (which would discharge directly to storm sewers if not removed) account for a significant portion of urban contamination in snow-derived runoff.

Table 3 Snowmelt loading estimates for 2020 coming from each of four storage facilities included in this study. Note that the snowmelt volumes are based on snow-water-equivalents (SWEs) that were determined using estimates of snow volumes on site and the snowpack density (see “Methods” for further information)
Full size table

Given it is the largest snow storage facility, as expected the Valley Rd. site contributed the highest calculated loadings for all parameters shown in Table 3. This includes exceptionally high values for TDS (169,231 ± 191,920 kg), COD (24,524 ± 14,365 kg), and DOC (611 ± 622 kg). High margins of error are to be expected as the volume of snow received at all facilities is estimated. Vijayan et al. (2021) observed up to 50% variation relative to best-estimate mass loads in samples composed by mixing collected snow in what seems to be a similar manner to that employed in this study. In addition, Vijayan et al. (2021) found single column samples deviated up to 400% from the best-estimate mass load. The margins of error presented in this study therefore seem to generally agree with those obtained from other SP studies under current sampling and management practices. Interestingly, the TSS loadings for the three CoS sites were all in a similar range (30,000 to 40,000 kg) despite having significant snow volume differences. However, this could be attributed to soil erosion on the smaller sites given the Valley Rd. site is the only facility contained on a concrete surface. This is also likely the reason why the TDS loading was very high at the Valley Rd. facility as there is no opportunity for infiltration as available at other sites. The TDS and COD were strongly correlated (R = 0.962) as were the COD and DOC (R = 0.966). Organics were not significantly present in snow with DOC comprising 0.36–1.03% of TDS loading. Two sites contributed 89% of DOC loading: Valley Rd. (611 kg) and Central Ave. (199 kg). Strong correlations were also observed between COD loading and copper (R = 0.918), chromium (R = 0.980), nickel (R = 0.973), zinc (R = 0.958), and ΣPAHs (R = 0.953) along with a moderate correlation in lead (R = 0.948). The largest metal loading estimates were of manganese (25 kg) and zinc (12 kg) driven by single-event pulses observed at Valley Rd. The next highest loading estimates were copper (2.25 kg with 52% comprising Valley Rd.) and nickel (1.38 kg with 64% comprising Valley Rd.). Loading estimates for copper, manganese, selenium, and ΣPAHs were not found to be significantly correlated to snow volume, though correlations were observed for chromium (R = 0.974), nickel (R = 0.945), lead (R = 0.921), and zinc (R = 0.950). However, each site’s characteristics are quite variable making direct comparisons difficult and potentially erroneous. For example, the TDS, COD, and DOC values at the Central Ave. site were markedly higher than the Wanuskewin Rd. despite each have the same estimated SWE volume.

During the winter in cold-region Canadian cities, precipitation falls as snow and runoff events do not occur for several months, though winter activities continue in the catchment. After a snowfall event, municipalities will plow snow from the roads into roadside banks and apply grit material (road salt or sand). Within hours of deposition, these snowbanks become contaminant sinks for heavy metals and particulates from traffic and winter maintenance activities, as well as transformation products of primarily organic contaminants (Glen and Sansalone 2002; Sillanpää and Koivusalo 2013; Westerlund and Viklander 2011). In this study, while sample DOC contributed to oxygen demand, the majority of COD is likely inorganic: COD has been observed as a major source of contamination in roadway runoff (Huang et al. 2007) and it is not unreasonable that COD in snow would be mainly traffic-derived. Strong COD correlations with reactive metals species Cr, Ni, Pb, and Zn indicate probable road and vehicle origins. However, despite all snow facilities receiving a mixture of roadside snow, the estimated contaminant loads for each facility varied. Contaminant deposition decreases drastically with increased distance from the roadway, with the bulk of pollutants deposited immediately beside the road (Hautala et al. 1995; Kuoppamäki et al. 2014). Hautala et al. (1995) observed that chloride concentrations were significantly greater at 10 m than 30 m from the road edge, while the opposite was true for low-molecular-weight (LMW) polyaromatic hydrocarbons (PAHs). Industrial emissions have additionally been implicated in non-roadside SP contamination (Li et al. 2015). The comparatively low PAH (0.07 kg) to DOC loads (907 kg) (total loading from all sites) are likely due to high particle affinity, but deposition distribution may play some role in inter-site variation. Facilities likely received volumes of non-roadside snow which contributed different contaminant species and concentrations. The Valley Rd. site possessed the greatest contamination load as expected due to lack of on-site infiltration. However, its TSS values were comparable to those observed at facilities with half the estimated SM volume at Wanuskewin Rd. and Central Ave. The Central Ave. site possessed 89-fold higher TDS and 70-fold higher COD than the Wanuskewin Rd. site despite seemingly similar environmental characteristics. The infiltration capacity at the Central Ave. site may have been reduced by high-TDS SM runoff over time, resulting in overall poorer SM quality leaving the site and impacting surrounding environments.

Conclusions

The main objective of this study was to assess urban SM quality from storage facilities. Toxicity results were inconclusive yet contribute to the limited knowledge for urban SP and SM runoff. Snow stored at urban facilities is primarily roadside snow, while cold climate SW studies largely analyze winter road runoff; therefore, this data may reflect conditions experienced in roadside snowpacks, or conditions in aged SM, relative to fresh winter road runoff. Nevertheless, quality data from snow storage facilities is lacking, though previous research has shown the roadside snow taken to these facilities contribute a significant portion of the traffic contaminant burden. As SPs provide an intermediary for the transport of urban SW contaminants to receiving environments, SP treatment strategies must be developed to ensure proper contaminant removals before releasing into receiving environments. Over several seasons, elution and particulate-bound contaminants potentially degrade on-site soils and groundwater if storage facilities are not properly sealed. Wetlands and swales adjacent to snow facilities may also develop characteristics resembling stormwater ponds, including contaminated sediment beds and salt-impacted flows or infiltration.

Observations were generally in agreement with previous literature, with preferential elution of TDS out of the SP as the melt period commenced and contaminant spikes observed following warm, sunny days. Significant chloride and manganese peaks occurred throughout the 2020 melt season, indicating management facilities must manage melt periods featuring contaminant spikes. Though fluctuating, elevated TSS values persisted in SPs and elevated TDS persisted in snowmelt weeks into the spring melt season, indicative of pulsed melt events instead of an elevated baseline throughout the melt period. Remarkably, dissolved concentrations of anthracene exceeded the CCME guideline threshold in virtually all SP and SM samples, while fluorene, benzo[a]pyrene, pyrene, and phenanthrene were found in almost all SP samples and 50% of SM samples. Despite the hydrophobicity of PAHs, traffic-based loading on managed snow is sufficient to harbor toxic potential to receiving water bodies. Increasing thaw-freeze periods may introduce contaminant spikes during winter months as well, extending the melt period requiring attention. Despite the magnitude of contamination reported in certain samples, no significant toxicological trends were observed in either R. subcapitata or V. fischeri.

It is estimated that snow plowed from CoS roads in 2020 contributed a total of 226,000 kg TSS and 119,000 kg TDS at their outlets. A majority (60%) of this loading undergoes primary treatment at the Valley Rd. facility, and the true quality impacting the SSR is yet unknown. Simple contaminant loading calculations using seasonal mean concentrations indicate relatively high TDS and relatively low TSS in snowmelt runoff at the paved Valley Rd. facility. This may indicate infiltration of TDS and erosion of TSS at unpaved sites, though research into on-site soil quality is needed. The traffic-derived COD is strongly correlated with chromium, nickel, lead, and zinc. However, the lack of continuous flow measurement data at snow facilities limits the ability to calculate flow-weighted site mean concentrations, and only a seasonal mean loading estimate could be ascertained from the data. Furthermore, no snowfall depth-SP volume correlations could be calculated due to limited data. This study establishes foundations upon which the aforementioned data can be developed.