Add a dash of salt? Effects of road salt on oxygen and nutrient fluxes out of streams, stormwater ponds and lakes

Study Sites and Sampling MethodsWe studied Lake Simcoe and its watershed (Ontario, Canada) where Chloride levels have steadily increased (~1 mg Cl- L-1 yr-1) over the past forty years in the lake. To explore a range of waterbodies and anthropogenic impacts within Lake Simcoe and its watershed, we selected seven nearshore lake sites, five streams sampled just upstream of the lake, and four stormwater ponds located in Barrie, Ontario. Streams and stormwater ponds were sampled five times from October 2010 to August 2011 (about once every two months). Lake sites were sampled three times during the ice fee season (October 2010, May 2011, and August 2011).During each sampling period, four to six sites were visited once per week and water was collected from the deepest part of the stream segment. Intact sediment cores in streams were collected from sediment depositional areas within 100 m of the water sampling point. Water samples and sediment cores were collected from stormwater pond sites just up-pond of the outflow structure. For lake sites, water samples were collected at the long-term monitoring locations and sediment cores were sampled nearer to shore at a depth of &lt; 2 m. In addition, two 23 L carboys were filled with whole water at each site for use in salt addition experiments. Finally, salt (NaCl) and salt/sand mixed samples were collected from ten storage depots located around the watershed and used as the salt source for the experiment.At each site, two intact sediment cores were collected using a sediment core affixed to a PVC poll and a one-way value (McCarthy et al. 2016). In situ dissolved oxygen (DO; mg O2 L-1), temperature (°C), and conductivity (µS m-1) measurements were taken at sampling depths using electronic handheld probes (YSI models 55 DO and 30/10 FT; YSI, Yellow Springs, Ohio, USA). Whole water samples were collected for total nutrient (TN and TP), chlorophyll a (CHL), total suspended solid (TSS), and chloride analysis. Water samples were filtered in the lab within 24 hours using 0.2 µm polycarbonate membrane filter (Millipore) with a precombusted glass fiber (Whatman GF/F) pre-filter for dissolved organic matter (DOM), organic carbon (DOC), and nutrients (NO3, TDN and TDP).Flow-through sediment core experimental designSediment cores and water carboys were set up in a flow-through system (McCarthy and Gardner 2003; Juckers et al. 2013). The environmental chamber was maintained using a light/dark and temperature cycle matching that of the collection month. The flow-through systems involved the continuous circulation of water from carboys equipped with air stones, over the sediment water interface of the core, with no air headspace within the core, using a peristaltic pump set at a flow rate of c.a. 1.5 mL min-1. Actual flow rates varied from 0.8 to 2.1 mL min-1 with a mean of 1.3 mL min-1. The system pumped water directly over the sediment surface and outflowed at the top of the core, facilitating water exchange.Water samples were collected to calculate ambient flux prior to dawn of day two (the start of the second light/dark cycle). After sunset of day two, inflowing carboys were switched out with salt/sand spiked carboys. During October and December 2010, salt additions doubled ambient chloride levels (an increase of 40, 100, and 500 mg Cl- L-1 for lake, stream, and stormwater pond sites, respectively). During March, May, and August 2011, salt additions were standardized to 500 mg Cl- L-1. Water samples were collected just before dawn two days after salt additions were made. Samples for TDP and NO3 were processed the same as ambient samples. Samples for DO analysis were measured immediately or preserved with sodium azide for later analysis (Juckers et al. 2013).DO, TDP, and NO3 fluxes were determined by calculating the difference in concentration between the inflow and outflow water and multiplying this difference by the actual flow rate over the sediment core. Rates were then standardized to the surface area of the core (52.8 cm2). All measurements reflect the net response of uptake, release and cycling of O2, N and P across the sediment–water interface, where negative values indicate uptake and positive values indicate release by the sediment.Laboratory chemical analysesChloride was measured on room temperature samples using a Cl- specific ion electrode (Thermo Fisher, model 9617BNWP). DOC (mg C L-1) and TDN (µg N L-1) were measured by combustion after acidification using an OI Aurora TOC analytical analyzer with an external nitrogen detector. TP and TDP (µg P L-1) were measured using the standard colorimetric method after persulfate (1.7% f.c.) autoclave digestions. Following acidification (pH &lt; 2), NO3 was measured by second-derivative UV spectroscopy. CHL (µg L-1) was determined from particles collected on GF/F filters following hot ethanol extraction. TSS (mg L-1) was measured as the mass of particles collected on GF/F filters after drying in an oven at 60°C until constant mass (Williams et al. 2013).DO concentration was determined using two methods for the flow-through sediment core experiment (membrane inlet mass spectrometry (MIMS) and O2 probe). For October and December 2010 events, O2 was measured using MIMS following the recommendations as in Kana (1994) and McCarthy et al. (2016) within 2 months of sample collection (Juckers et al. 2013). For March, May and August 2011 events, O2 was measured immediately after sampling using a YSI 55-DO probe.DOM composition was determined in a prior study using a seven component parallel factor (PARAFAC) model of fully correct absorbance and excitation-emission matrix data (Williams et al. 2013, 2016). PARAFAC components were reported here as percent of total fluorescence intensity and were: C1 (humic-like), C2 and C3 (terrestrial, humic-like), C4 (soil, fulvic-like), C5 (microbial, humic-like), C6 (anthropogenic), C7 (protein-like).Statistical analysisStatistical analyses were carried out in R using Rstudio and the following packages and their requirements: broom, psych, rpart, randomForest, scales and tidyverse (Liaw and Wiener 2002; Wickham 2016; Therneau and Atkinson 2019; Wickham et al. 2019; Revelle 2020; Robinson et al. 2020; Wickham and Seidel 2020). Color palettes were based on colorBlindness (Ou 2020). Pearson bivariate correlation was used to determine covariation between ambient chloride levels and water quality variables across all samples and within waterbody type. A resampling analysis of variance (ANOVA) and post hoc analysis was used to compare ambient O2, N, and P flux by the main effect of waterbody type and sampling event. The effect of salt additions on O2, N, and P flux was determined by paired t-test. A resampling approach was used for comparisons based on overall effect, waterbody type, and event. A normal two-tailed test was used for event by site comparisons because two sediment core replicates was not sufficient to generate a random test population. Regression tree with random forest was used to determine what influenced ambient flux. Regression tree with random forest and classification tree analyses were used to determine environmental drivers of flux responses to salt addition, where the responses were based on site by event paired t-test results (mean paired difference, p-value, and significance difference (α = 0.1)). Classifications were more uptake/less release by sediment, less uptake/more release by sediment, and similar.<br>