Assessment of Snowmelt Quality Discharging from a Cold-Climate Urban Landscape During Spring Melt

Stormwater results from precipitation events and melting snow running off urban landscapes and typically being released into receiving water bodies with little to no treatment. Despite evidence of its deleterious impacts, snowmelt (SM) management and treatment are limited, partly due to a lack of quality and loading data. This study examines snowmelt quality during the spring for a cold-climate, semi-arid Canadian city (Saskatoon, Saskatchewan). Four snow storage facilities receiving urban snow plowed from roads in mixed-land-use urban catchments (228 km 2 ) were sampled including snow piles (ve events) and SM (twelve events) runoff in 2019 and 2020. Samples were analyzed for pH, EC, TDS, TSS, COD, DOC, metals, chloride, PAHs, and Raphidocelis subcapitata and Vibrio scheri toxicity. Notable event-specic TSS spikes occurred on April 13th, 2019 (3,513 mg/L) and April 24th, 2019 (3,838 mg/L), and TDS, chloride, and manganese on March 26th, 2020 (15,000 mg/L, 5,800 mg/L, 574 mg/L), April 17th, 2020 (5,200 mg/L, 2,600 mg/L, 882 mg/L), and April 23rd, 2020 (5,110 mg/L, 2,900 mg/L, 919 mg/L), though chloride remained elevated through May 1st, 2020 samples (1,000 mg/L). Additionally, at two sites sampled April 13th, 2019 pulses of aluminum (401 mg/L) and PAHs (pyrene, phenanthrene, anthracene; 71 µg/L, 317 µg/L, 182 µg/L) were detected. The EC 50 for R. subcapitata and V. scheri was observed, if at all, above expected toxicity thresholds. 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 (R 2 = 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 conguration of four replicates of negative control and a ve-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 uorescence as a proxy for algae cell count, growth inhibition was measured as a function of uorescence 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 uorescence and cell counts were read using the Tecan Spark® multifunction plate reader (Tecan Trading AG, Switzerland). Sample concentration in wells ranged from 100–6.25%, both inoculated and uninoculated to account for background uorescence. Initially, the EC 50 at 72 hours was calculated; as the EC 50 was not observed at the full concentration for all samples, the EC 10 was additionally calculated (see Statistical Analysis section). 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 EC 10 values of R. subcapitata and V. scheri against individual SW parameters and metals and PAH species. when sand is used as an anti-slip agent (often in combination with NaCl or MgCl 2 , CaCl 2 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.


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 ashiness of SW discharging from these landscapes, in conjunction with signi cantly decreased SW quality ( The form in which contaminants are found is known to effect potential toxic impacts by affecting bioavailability to aquatic organisms; in SM, elevated concentrations of Cd, Cu, Mn, Ni, and Zn have been associated with elevated chloride ion concentrations (Lazur et al. 2020; Mayer et al. 2011a; Reinosdotter and Viklander 2007). 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). Most heavy metal burdens in urban SM (>90%) were in the particulate phase 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 ( 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 rst, while heavier or particle-bound contaminants may require high volumes or intensities to ow into sewers (Mayer et al. 2011b). This phenomenon is referred to as the ' rst ush', and though it is used in SW to refer to a single-event phenomenon, the mechanisms of pollutant build-up and wash-off also in uence spring melt runoff characteristics. As snow melting occurs on warm days contaminants are preferentially eluted from the SP ; 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 (Gal 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 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 plowed snow is signi cantly more burdened than undisturbed SPs, containing up to 50% of areal pollutant mass (Kuoppamäki et  There are four snow storage facilities within the CoS, all of which were included in the present study ( Figure 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 ow 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 2nd, 2019 and two SP samples were obtained on March 7th, 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, labelled, 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.

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/m 3 ) was calculated by packing snow from the April 2nd, 2019, and April 13th, 2019, snow events into a plastic container, estimating the volume, weighing, melting the snow at room temperature, measuring the volume, and then using the difference in volume to calculate density. The biological analyses comprised the enumeration of fecal coliforms, while twotoxicity assays were conducted including Raphidocelis subcapitata algae and Vibrio scheri bacteria.

Physicochemical Analyses
The TSS concentration was measured via vacuum ltration using Whatman TM 934-AH TM glass micro bre lters 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 Te on lter using a 12-mL Luer-Lock syringe. Approximately 40 mL of ltered sample was placed in a glass vial and analyzed using a Lotix combustion TOC analyzer (Teledyne Tekmer, 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 metals analysis, samples were acidi ed with 0.02 N nitric acid and vacuum-ltered through a 0.45-µm nitrocellulose lter. A 100-mL sample volume was passed through the lter and the ltrate 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 20x prior to analysis, and using a custom calibration standard (SCP Science) for blanks and standards of 10, 50, and 100 ppb. The certi ed 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 (HNO 3 ) (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 Scienti c (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-ltered (Whatman TM 934-AH TM glass micro bre lters) to remove high TSS concentrations prior to solid-phase extraction (SPE) to eliminate clogging of the lter. After ltration, 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-labelled internal standard mix (500 mg/L of acenaphthene-d 10 , chrysene-d 12 , and phenanthrene-d 10 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 minutes. 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 Scienti c 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-5ms (60 m x 250 µm I.D., lm thickness 0.1 µm) fused silica capillary column. Both instruments were operated in full scan mode, and data were analyzed using an isotope-dilution work ow, i.e., areas of target compounds were normalized to the areas of recovered deuterium-labelled standards. A seven-point calibration curve, along with extraction and solvent blanks were run with each batch of samples. Limits of detection and quanti cation are included in Table S3.

Biological and Toxicity Analyses
The rst toxicity assessment was the inhibition of the luminescence of V. scheri 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 one hour 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. scheri 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 multipipettor, plates were inoculated with 8 µL of the bacteria solution. Luminescence readings were taken immediately, after 5 minutes, and after 15 minutes, with plates remaining on ice between measurements. Due to the con guration of the plate, samples were run as sevenstep dilutions while the phenol was a six-step dilution series. Values for both samples and phenol are reported as the EC 50 and EC 10 relative to the negative control at 15 minutes.
The second bioassay tested the 72-hr 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 (R 2 = 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 con guration of four replicates of negative control and a ve-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 uorescence as a proxy for algae cell count, growth inhibition was measured as a function of uorescence 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 uorescence and cell counts were read using the Tecan Spark® multifunction plate reader (Tecan Trading AG, Switzerland). Sample concentration in wells ranged from 100-6.25%, both inoculated and uninoculated to account for background uorescence. Initially, the EC 50 at 72 hours was calculated; as the EC 50 was not observed at the full concentration for all samples, the EC 10 was additionally calculated (see Statistical Analysis section).

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 tra c 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 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 re ect 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-tra c related land use impacts, it nonetheless represents a signi cant 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 sitemean-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-speci c physicochemical and contaminant concentrations, seasonal loadings were estimated for the 2020 SM season (Table 3) as presented in the Results and Discussion. 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).

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 EC 10 values of R. subcapitata and V. scheri 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. scheri) observed in the negative control. Within GraphPad Prism, four-parameter logistic regression was used to t dose-response curves to the growth rate and luminescence inhibition data, respectively, to obtain the EC 50 and its 95% con dence interval (CI) for each analyzed sample. As the EC 50 was not observed for many samples, the EC 10 and its 95% CI was also interpolated from the curve. The EC x is de ned 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) Toxicity 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 (Section 3.1.1); DOC, COD, and TSS (Section 3.1.2); and metals and PAHs (Section 3.1.3). The toxicity section presents the two toxicity assays R. subcapitata (Section 3.2.1) and V. scheri (Section 3.2.2). The nal 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). Measurements of chloride were performed on select samples (Table S8) A two-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 2nd, 2019, and March 7th, 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  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 signi cant difference related to site nor event was observed for any samples. The highest TSS was observed on April 13th, 2019 SP. Elevated TSS values in SM were observed on April 17th, 2020 (USask, 1,800 mg/L), and April 28th, 2020 (Central Ave, 2,982 mg/L), coinciding with warm days in which higher ow 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 ndings that large particles are deposited onsite and not transported with meltwater (Viklander 1999 (2011), who noted that over 90% of the metals 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 (Figure 2 with concentrations of individual elements included in Tables S9-S10). Concentrations of dissolved aluminum, iron, lead, and zinc were generally higher in SP samples, though SP concentrations were comparable to those in summer SW obtained as part of a parallel study (Popick et

R. subcapitata Bioassay
Signi cant toxicity was generally not observed in either SP or SM samples, with the calculation of an EC 10 necessary to quantify toxic response for many samples (Table 2). It is assumed that chloride might have played a signi cant 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-speci c sample characteristics, though toxic responses were observed for all tested April 17th, 2020 samples.
The most toxic response for which a 95% CI could be calculated was the USask SM sample on this date (EC 50 at 43.7 percent 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 17th, 2020 Central Ave. sample generated the second-lowest EC50 threshold (54.3 percent 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 23rd, 2020 Central Ave. sample has similar quality measurements but no detected EC 50 50 and EC 10 for R. subcapitata and 15-minute luminescence inhibition for V. scheri (Microtox). Results are presented as percent of total sample concentration required to read the endpoint. EC 50 and EC 10 values were generated using GraphPad Prism as a dose-response curve (see Methods) of, in the case of R. subcapitata, the initial (t = 0) uorescence observed after 72 hours, and in the case of V. scheri, the initial (t = 0) luminescence observed after 15 minutes. EC 10 values were interpolated from relating 10% luminescence inhibition to sample concentration on a standardized curve (see Methods).  (Mayer et al. 2011a). When testing the impact of the three road salts on rainbow trout, Hintz and Relyea (2017) found CaCl 2 was the most harmful to sh growth, followed by NaCl and then MgCl 2 . Conversely, emphasizing the intraspecies variability of toxicity, Hopkins et al. (2013) found MgCl 2 and NaCl caused comparable developmental deformities in rough-skinned newts.

V. scheri 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 di culties in calculating IC 50  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. 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 tra c and winter maintenance activities, as well as transformation products of primarily organic contaminants (Glen

Conclusions
The main objective of this study was to assess urban SM quality from storage facilities. Comprehensive analysis of conventional SW parameters and toxicological results for R. subcapitata and V. scheri determined herein have contributed to the limited database for urban SP and SM runoff worldwide. Snow stored at urban facilities is primarily roadside snow, while cold-climate SW studies largely analyze winter road runoff; therefore, this data may re ect 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 signi cant portion of the tra c 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.
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. Signi cant chloride and manganese peaks occurred throughout the 2020 melt season, indicating management facilities must manage melt periods featuring contaminant spikes. Though uctuating, 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. 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 signi cant toxicological trends were observed in either R. subcapitata or V. scheri.
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 in ltration of TDS and erosion of TSS at unpaved sites, though research into on-site soil quality is needed. The tra c-derived COD is strongly correlated with chromium, nickel, lead, and zinc.
However, the lack of continuous ow-measurement data at snow facilities limits the ability to calculate ow-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.
Declarations 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-speci c and error bars indicate 95% CIs.