Concentrations and Yields of Mercury, Methylmercury, and Dissolved Organic Carbon From Contrasting Catchments in the Discontinuous Permafrost Region, Western Canada

Abstract

Climate change and permafrost thaw may impact the mobilization of terrestrial dissolved organic carbon (DOC), mercury (Hg), and neurotoxic methylmercury (MeHg) into aquatic ecosystems; thus, understanding processes that control analyte export in northern catchments is needed. We monitored water chemistry for 3 years (2019–2021) at a peatland catchment (Scotty Creek) and a mixed catchment (Smith Creek) in the Dehcho (Northwest Territories), within the discontinuous permafrost zone of boreal western Canada. The peatland catchment had higher DOC and dissolved MeHg, but lower total Hg concentrations (mean ± standard deviation; 19 ± 2.6 mg DOC L⁻¹; 0.08 ± 0.04 ng DMeHg L⁻¹; 1.1 ± 0.3 ng THg L⁻¹) than the mixed catchment (12 ± 4.4 mg DOC L⁻¹; 0.05 ± 0.01 ng DMeHg L⁻¹; 3.1 ± 2.2 ng THg L⁻¹). Analyte concentrations increased with discharge at the mixed catchment, suggesting transport limitation and the flushing of near-surface, organic-rich flow paths during wet periods. In contrast, analyte concentrations in the peatland catchment were not primarily associated with discharge. MeHg concentrations, MeHg:THg, and MeHg:DOC increased with water temperature, suggesting enhanced Hg methylation during warmer periods. Mean open water season DOC and total MeHg yields were greater and more variable from the peatland than the mixed catchment (1.1–6.6 vs. 1.4–2.4 g DOC m⁻²; 5.2–36 vs. 6.1–10 ng MeHg m⁻²). Crucial storage thresholds controlling runoff generation likely drove greater inter-annual variability in analyte yields from the peatland catchment. Our results suggest climate change may influence the production and transport of MeHg from boreal-Arctic catchments as temperatures increase, peatlands thaw, and runoff generation is altered.

Key Points

• Comparison of mercury, methylmercury, and dissolved organic carbon concentrations and yields for a peatland and a mixed-landscape catchment

• Production of methylmercury in the peatland catchment was temperature-dependent, and yields were limited by runoff generation thresholds

• Analyte concentrations in the mixed catchment increased with discharge, and had lower inter-annual variability in analyte yields

Plain Language Summary

Boreal rivers deliver dissolved organic carbon (DOC), mercury (Hg), and its neurotoxic form, methylmercury (MeHg), from contributing landscapes to downstream waters. In northern regions, thawing permafrost (i.e., perennially frozen ground) in peatland environments may release Hg, MeHg, and DOC from soils to rivers. Over 3 years, we measured the concentrations of water chemistry in two boreal creeks with differing contributing landscapes in the discontinuous permafrost region of the Dehcho (Northwest Territories, Canada). Our results showed different patterns in analyte concentrations versus yields (i.e., annual analyte mass delivered per unit area) between the sites. At the peatland-dominated Scotty Creek, concentrations of MeHg and DOC were consistently high and not primarily controlled by discharge. Still, runoff and analyte yields varied annually as peatlands stored water in a dry year and discharged water in wet years. At Smith Creek, with a mixed landscape of mountains, peatlands, and forests, concentrations of DOC and Hg increased during high-flow events as runoff contributions from peatlands increased. Over the monitoring period, analyte yields at the mixed catchment varied less due to consistent groundwater inputs and runoff generation from steeper slopes. Changing runoff due to climate change and permafrost thaw will likely alter analyte yields in the region.

1 Introduction

Boreal rivers are important conduits of terrestrially-derived dissolved organic carbon (DOC) that co-transport mercury (Hg) and neurotoxic methylmercury (MeHg) to downstream freshwater environments and the ocean (Fink-Mercier, Lapierre, et al., 2022; Kirk & St. Louis, 2009; Zolkos et al., 2020). Monitoring total Hg (THg) and MeHg concentrations and yields in high-latitude catchments is a research priority for northern communities as permafrost thaw occurring in response to rapid climate warming exposes previously frozen Hg and carbon to biogeochemical and hydrological processes (Dastoor et al., 2022; St. Pierre et al., 2018). Within the Interior Plains of boreal western Canada, extensive peatlands in the discontinuous permafrost zone (i.e., 10%–90% areal extent of perennially frozen ground) are increasingly warming, burning, and thawing due to climate change (Gibson et al., 2018, 2021; Wright et al., 2022); complete permafrost loss is projected by 2050 (Chasmer & Hopkinson, 2017). Since MeHg biomagnifies and bioaccumulates in food webs, local increases in THg and MeHg concentrations may threaten the food security of northern Indigenous communities, while enhanced export of THg and MeHg may accumulate in downstream waterbodies and marine ecosystems (N. Basu et al., 2022; Houde et al., 2022; Moslemi-Aqdam et al., 2023).

Anoxic, acidic, DOC-rich conditions support the microbial methylation of inorganic divalent Hg^II into MeHg, making peatlands ideal methylation sites (Bravo & Cosio, 2020; Gordon et al., 2016; Poulin et al., 2019). Higher in situ methylation efficiency has been linked to the presence of bioavailable Hg^II and labile dissolved organic matter (DOM), whereas MeHg bound to aromatic DOM has been linked to longer-range transport in boreal surface waters (Bravo et al., 2017; Mangal et al., 2022; Shanley et al., 2022). Permafrost peatlands have accumulated large Hg stores bound to organic matter (Lim et al., 2020; Schuster et al., 2018), and non-permafrost peatlands are hotspots of MeHg production across the global boreal-Arctic (Gordon et al., 2016; Poulin et al., 2019; Tarbier et al., 2021). However, few studies have examined concentrations and export of MeHg in northern rivers (Jonsson et al., 2022) and it remains unclear how strongly local hotspots of MeHg production influence the chemistry of catchment outlets.

The species (i.e., THg or MeHg), fraction (i.e., dissolved or particulate), and concentration of Hg drive biotic uptake, where elevated concentrations of dissolved MeHg maximize uptake (Blanchfield et al., 2022; Gagnon & Fisher, 1997). Peatland-dominated catchments predominantly deliver dissolved Hg species bound with DOC (Emmerton et al., 2022; Kirk & St. Louis, 2009), with a higher proportion of MeHg to THg (%MeHg) than sediment-laden rivers carrying particulate-bound Hg species (St. Pierre et al., 2018; Staniszewska et al., 2022). Discharge and changing water sources can drive variability in water chemistry concentrations, where concentration-discharge metrics identify chemodynamic (i.e., high variability in response to discharge) or chemostatic (i.e., low variability or non-systematic in response to discharge) behavior and thus the hydrological or biogeochemical controls on an analyte (N. B. Basu et al., 2010; Pohle et al., 2021). Where a non-systematic response to discharge is observed, additional factors, such as temperature, may drive the change in chemistry of water sources over time; elevated soil temperatures can stimulate the production of both MeHg and DOC (Mattsson et al., 2015; Yang et al., 2016). The presence of discontinuous permafrost in Canada's Interior Plains influences the connectivity of catchment water sources (Wright et al., 2022) and thus may enable patterns of THg and MeHg that contrast other boreal regions, which have shown the highest concentrations during summer and a stronger response of THg than MeHg to changing discharge (Fink-Mercier, Lapierre, et al., 2022; Lam et al., 2022).

Export of THg and MeHg from high latitudes has been quantified primarily in large Arctic rivers, carrying 37–44 Mg THg and 0.8–2.5 Mg MeHg into the Arctic Ocean annually (Soerensen et al., 2016; Sonke et al., 2018; Zolkos et al., 2020). Contributions from sub-catchments vary depending on land cover and hydrology. The yields of THg, MeHg, and DOC (i.e., analyte mass delivered per unit area) in boreal catchments can exhibit high seasonal variability tied to either hydrological or biogeochemical controls. Spring freshet, driven by snowmelt, often dominates seasonal hydrographs and the annual yields of DOC and Hg in northern regions, despite its short period (i.e., days to weeks; Burd et al., 2018; Gandois et al., 2021; Schelker et al., 2011). Conversely, summer storms can drive high-intensity exports of terrestrially-derived THg and DOC from deeper soil layers and mobilize MeHg produced during warmer summertime conditions (Kirk & St. Louis, 2009; Schelker et al., 2011; Shanley et al., 2022). Runoff generation in peatland-dominated catchments is controlled by storage thresholds, where significant runoff is generated only if these thresholds are exceeded. However, low peatland runoff occurs during dry periods when the landscape retains water (Goodbrand et al., 2019). The dry climate of the Interior Plains lends additional sensitivity to runoff generation as even small variability in the balance between precipitation and evapotranspiration can drive large relative variations in annual runoff (Burd et al., 2018). Unique amongst peatland-dominated regions of the boreal-Arctic, the Interior Plains has experienced significant increases in annual basin runoff and runoff ratios in the absence of increasing precipitation, suggesting that ongoing permafrost thaw is altering runoff (Mack et al., 2021), driven by changes in land cover and hydrological connectivity of sub-surface and surface flows (Quinton et al., 2019). Therefore, understanding the relationships between water chemistry concentrations, yields, and changing runoff is essential in the rapidly warming Interior Plains.

Here, we aimed to quantify the inter- and intra-annual variability in THg, MeHg, and DOC concentrations and yields from 3 years of water chemistry observations from two sub-Arctic catchments with differing land cover composition in the Interior Plains (boreal Western Canada). To meet this goal, we monitored analyte concentrations during the open-water season, coupled with DOM composition metrics and analysis of concentration-discharge relationships. We expected the peatland catchment to be chemostatic due to consistent contributions from peatlands, with the potential for a signal of summertime MeHg production peaks, and to deliver consistently high concentrations of MeHg and DOC as observed in other boreal peatland catchments. In contrast, we anticipated chemodynamic behavior from the catchment with a mixed landscape of mountain, forest, and peatlands with terrestrial analyte flushing during high flow events and a greater contribution of particle-bound Hg. Our findings revealed processes that control the concentrations and exports of THg, MeHg, and DOC to predict the influence of current and future climate-driven changes, such as permafrost thaw, on analyte delivery in the Interior Plains.

2 Materials and Methods

2.1 Study Catchments

Samples were collected at the outlets of Scotty Creek and Smith Creek catchments in the Dehcho region, Northwest Territories (Figure 1); Scotty Creek is within the sporadic discontinuous permafrost zone (10%–50% permafrost), while Smith Creek has extensive discontinuous permafrost (50%–90% permafrost; Brown et al., 2002). The catchments are located in the homelands of First Nation and Pehdzeh Ki First Nation, respectively. Scotty Creek is a 129 km² catchment located within the Taiga Plains Mid-Boreal Ecoegion, and Smith Creek is a 154 km² intersecting the Taiga Plains Mid-Boreal Ecoregion, Taiga Cordillera Low Subarctic Ecoregion, and Boreal Cordillera High Boreal Ecoregion, with both catchments draining into the Deh Cho (Mackenzie River; Ecosystem Classification Group, 2009). The regional climate is characterized by long, cold winters and short summers, although Scotty Creek is warmer and wetter than Smith Creek based on mean annual air temperature (−2.5 and −4.7°C, respectively) and mean annual precipitation (415 and 350 mm, respectively; Fick & Hijmans, 2017).

Figure 1

Study catchments. (a) Study site locations in relation to permafrost zones (Brown et al., 2002) and nearby communities, including Ochre River with longer-term discharge records, (b) location within Canada and the Taiga Plains, (c) Smith Creek—Mixed, and (d) Scotty Creek—Peat, with flow paths (ArcGIS derived) and elevation (Natural Resources Canada, 2013, 2022) overlaid by land cover (Hermosilla et al., 2022). See Table S1 in Supporting Information S1 for proportions of land cover at each catchment.

The surficial geology of Scotty Creek is characterized by thin to thick tills and fine-grained glaciolacustrine deposits from glacial retreat that support peatland development, with a bedrock of shale, limestone, and dolomite formed during the Devonian period (Aylsworth et al., 2000; Wheeler et al., 1996). Smith Creek has surficial geology of glaciolacustrine deposits, thick tills, and some areas of exposed shale, limestone, and dolomite bedrock within the Franklin mountains (Aylsworth et al., 2000; Wheeler et al., 1996). Vegetation cover at both sites includes mixed-wood forests of trembling aspen (Populus tremuloides) and white spruce (Picea glauca) in the uplands with peatlands in the lowlands (Ecosystem Classification Group, 2009). Peatlands in this region are mosaics of relatively dry permafrost peat plateaus that support black spruce (Picea mariana), Labrador tea (Rhododendron groenlandicum), and lichens alongside wetter permafrost-free wetlands: bogs with Sphagnum mosses and fens with varying vegetation of sedges and shrubs (Quinton et al., 2019). The land cover composition varies between the sites; Scotty Creek is low-relief and peatland-dominated, while Smith Creek is steeper, with the Franklin mountain range to the east, the Mackenzie River to the west, and is predominantly forested with peatlands in the southwest portion of the catchment (Figure 1, Table S1 in Supporting Information S1). We henceforth refer to Scotty Creek as the “Peat” catchment and Smith Creek as the “Mixed” catchment.

We quantified the Smith Creek catchment characteristics with the slope and hydrology toolsets on ArcGIS 10.8 using the Canadian Digital Elevation Model (Natural Resources Canada, 2013) with corrections in low-lying areas using the National Hydro Network (Natural Resources Canada, 2022). The Scotty Creek catchment boundaries were obtained from the National Hydro Network (Natural Resources Canada, 2022). We further summarized simplified land cover characteristics in Table S1 in Supporting Information S1; the numeric classification used by Hermosilla et al. (2022) corresponds to land cover types as follows: Wetland class 80 + 81; Forest 210 + 220 + 230; Shrub 40 + 50 + 100; Barren 32 + 33; Water 20.

2.2 Sample Collection and Analysis

Sample collection was initiated in 2019, where water samples were taken ∼monthly during the ice-free season (∼April to September or October) for Hg, MeHg, DOC, DOM characterization, nutrients, and major ions (University of Alberta & Dehcho-AAROM, 2022). Dehcho Aboriginal Aquatic Resources and Ocean Management program members collected samples from 2020 to 2021. In total, 24 grab samples were collected through the study at each site (Table 1).

Table 1. Characteristics of (a) Scotty Creek and (b) Smith Creek

Sampling characteristics	(a) Scotty Creek—peat			(b) Smith Creek—mixed
	2019	2020	2021	2019	2020	2021
Total annual rainfall (mm)	259	474	245	305	323	304
Annual air temperature (°C)	−0.9	−2.1	1.4	−2.9	−4.4	−3.1
Monitoring start	24 March	11 June	25 June	24 March	5 June	30 June
Monitoring end	29 September	24 September	5 October	28 September	22 October	16 September
Grab samples (n)	11	7	6	11	7	6
Bulk THg (ng L^−1)	1.21 ± 0.33	1.09 ± 0.29	1.05 ± 0.25	3.57 ± 2.80	2.56 ± 1.55	3.00 ± 1.56
Filtered THg (ng L^−1)	0.93 ± 0.31	ND	ND	1.19 ± 1.14	ND	ND
Bulk MeHg (ng L^−1)	0.09 ± 0.04	0.09 ± 0.02	0.08 ± 0.02	0.08 ± 0.07	0.07 ± 0.02	0.11 ± 0.04
Filtered MeHg (ng L^−1)	0.08 ± 0.04	ND	ND	0.05 ± 0.01	ND	ND
DOC (mg L^−1)	18.1 ± 2.2	21.6 ± 1.5	20.3 ± 1.4	14.1 ± 4.9	11.9 ± 3.5	10.9 ± 2.1
SUVA254 (L mg^−1 C m^−1)	3.4 ± 0.2	3.6 ± 0.2	3.4 ± 0.5	3.5 ± 0.7	3.3 ± 0.3	3.3 ± 0.1
pH	7.6 ± 0.3	7.5 ± 0.2	7.4 ± 0.4	8.1 ± 0.1	8.1 ± 0.1	8.1 ± 0.1
EC (μS cm^−1)	188 ± 24	157 ± 21	211 ± 31	553 ± 197	659 ± 219	736 ± 116

• Note. The total annual rainfall and mean annual air temperature (15 m height) were taken from eddy covariance towers at the sites (Sonnentag, 2022; Sonnentag & Quinton, 2022). The total annual rainfall value is solely precipitation as rain and does not include snowfall. The mean ±1 standard deviation (SD) of grab sample water chemistry includes total mercury (THg), methylmercury (MeHg), dissolved organic carbon (DOC) concentrations, specific absorbance at 254 nm (SUVA₂₅₄), pH, and electrical conductivity (EC). ND, no data.

To understand how different catchment water sources influence stream water chemistry, we sampled two potential sources of streamflow. In July 2021, we sampled a small spring that discharged into the mixed catchment creek, following the same procedures as the creek sampling. Over 3 years of visual observations indicated consistent flows originating from a nearby hillslope and this was assumed to be groundwater exfiltration. During July of 2021, porewater samples were additionally taken from wetlands in the peatland and mixed catchment (see sampling details and data in Thompson et al., 2022).

We collected unfiltered (2019–2021) and filtered (2019 only) samples from the edge of the creeks for total Hg (THg) (2 × 125 mL) and MeHg (2 × 250 mL) analysis following the clean hands-dirty hands protocol (St. Louis et al., 1994). All samples were collected in glass amber bottles (Environmental Supply Company, Inc., Durham, NC, US), certified to be cleaned by the manufacturer. Filtered THg and MeHg samples were obtained by passing one of the two samples through an acid-washed Nalgene 0.45 μm cellulose nitrate filter tower within 24 hr. The filtrate was then transferred to a new bottle. All THg and MeHg samples were preserved with 0.2% and 0.4% trace-metal grade hydrochloric acid (HCl), respectively.

We measured each site's physical parameters, nutrients, major ions, and DOM characteristics. Electrical conductivity (EC), pH, and water temperature were measured on-site with calibrated PT1 and PT2 Ultrapens (Myron L Company, Carlsbad, CA, US). We collected two additional filtered (0.7 μm Grade GF/F, Whatman) 60 mL water samples in acid-washed amber glass bottles, one of which was preserved with 0.6 mL of 2M HCl for DOC and cation analysis, while the other sample remained non-acidified for anion analysis, DOM absorbance, and fluorescence spectroscopy. All water samples were kept cool in transit to the laboratory and remained refrigerated until analyses.

Hg samples were analyzed in the Canadian Association for Laboratory Accreditation-certified Biogeochemical Analytical Service Laboratory (University of Alberta). Water samples were analyzed for THg concentrations on a Tekran 2600 Mercury Analyzer (Tekran Instruments Corporation, Scarborough, ON, Canada). THg was analyzed following EPA Method 1631. Samples were first oxidized with bromine chloride (BrCl) for at least 12 hr, and the presence of excess BrCl was established by pipetting a sample onto potassium iodide starch paper (Hintelmann & Ogrinc, 2003). 10% of the samples were spiked by a known quantity of mercuric chloride (HgCl₂) to assess procedural recoveries (Spex-CertiPrep, US), approximately equivalent to the sample concentration. Spike recoveries ranged from 95% to 105%. The Tekran 2600 was standardized by a 9-point standard curve (0–40 ng L⁻¹) at the start of each analytical day, using certified Brooks Rand HgCl₂ standards (R² > 0.999). Water samples were analyzed for MeHg using isotope dilution, distillation, and analyses on a Tekran 2700 Methylmercury Analyzer (Tekran Instruments Corporation, Scarborough, ON, Canada) coupled to an Agilent 7900 inductively coupled plasma-mass spectrometer (ICP-MS; Agilent Technologies, Inc., Santa Clara, CA, US). MeHg was analyzed following EPA Method 1630. All samples were spiked before distillation with Me²⁰¹Hg as an internal standard to correct for MeHg loss or formation during analysis. Me²⁰²Hg was used as the ambient tracer. Blanks of Milli-Q® water and the reagents were run with samples beginning at the distillation stage and before the analysis by gas chromatograph paired with an inductively coupled plasma mass spectrometer. 20% of samples were duplicated, falling within ±10%. Detection limits were 0.08 ng L⁻¹ for THg and 0.01 ng L⁻¹ for MeHg. Filtered THg and MeHg samples were presumed proportionate to the dissolved fraction of THg and MeHg, while the particulate fraction of THg and MeHg was calculated from the difference between bulk and filtered concentrations.

The Natural Resources Analytical Laboratory (University of Alberta) analyzed samples for DOC, metals, and nutrients. A TOC-L combustion analyzer with a TNM-L module (Shimadzu Corporation, Kyoto, Japan) measured DOC and total dissolved nitrogen concentrations. Colorimetry (Thermo Gallery Plus Beermaster Autoanalyzer, Thermo Fisher Scientific, Waltham, MA, US) determined concentrations of chloride (Cl), nitrate (NO₃-N), nitrite (NO₂-N), ammonium (NH₄-N), orthophosphate (PO₄-P), and sulfate ($${\text{SO}}_{4}^{-2}$$-S). Concentrations of sodium (Na), potassium (K), calcium (Ca), manganese (Mn), iron (Fe), copper (Cu), zinc (Zn), magnesium (Mg), phosphorus (P), and sulfur (S) were measured by inductively coupled plasma optical emission spectroscopy (ICP-OES, iCAP6300 Duo, Thermo Fisher Scientific, Waltham, MA, US).

The absorbance of DOM was measured from 200 to 700 nm (UV-1280, UV-VIS Spectrophotometer, Shimadzu Corporation, Kyoto, Japan) and corrected with Milli-Q water blanks. We used an Aqualog fluorometer (Horiba Ltd., Kyoto, Japan) to measure DOM fluorescence emission-excitation spectra. Samples were scanned from excitation wavelengths of 230–500 nm at 5 nm increments and emission wavelengths of 210–620 nm at 2 nm increments. Corrections for excitation and emission, inner filter effects, and Raman signal calibration were applied before analysis.

2.3 Continuous Data Collection

In 2019, Smith Creek was instrumented with HOBO water level loggers (HOBO MX2001, Onset Computer, Bourne, MA, US), recorded hourly, which was not required for Scotty Creek as it houses a long-term Water Survey of Canada hydrometric station. Both the peatland catchment and mixed catchment were instrumented with Spectro::lysers (s::can, Vienna, Austria) to measure UV-visible absorbance, recorded on 3–12 hr intervals depending on sunlight conditions and duration between site visits. High-frequency UV-visible absorbance was measured to examine the behavior of A₂₅₄ as a proxy for DOC. The Spectro::lyser measured decadal absorbance (cm⁻¹) between 200 and 700 nm at 2.5 nm intervals and was turbidity-corrected at 255 nm by subtracting the average wavelength between 550 and 700 nm from the Spectro::lyser spectra (Burd et al., 2018); 255 nm is presumed proportional to 254 nm. Spectro::lyser data were gap-filled with autosampler data on occasions where debris was caught in the measurement window, or there was insufficient solar panel charge. An autosampler (Teledyne ISCO, Lincoln, NE, US) collected ∼daily water samples to analyze UV-visible absorbance, fluorescence, and DOC. Spectro::lyser and autosampler data were checked by comparing them to grab samples (Table S2 in Supporting Information S1).

Due to COVID-19 constraints, no continuous absorbance measurements were taken from 2020 to 2021, although water level loggers were installed at Smith Creek. Discharge data from the peatland catchment was acquired from Water Survey of Canada records (Environment and Climate Change Canada, 2021), while we estimated discharge from hourly water depth records at the mixed catchment, converted using a stage rating curve and velocity-area calculations from monthly field measurements in 2019 (Turnipseed & Sauer, 2010). Any data gaps in discharge for mixed catchment were filled based on a regression derived from discharge records of the nearby Ochre River (Figure 1a); see Table S3 in Supporting Information S1 for details on the Ochre River catchment characteristics and gap-filling regression (Environment and Climate Change Canada, 2021). Environmental data (mean annual air temperature at 15 m height and total annual rainfall) were taken from meteorological stations within the peatland catchment and mixed catchments (Sonnentag, 2022; Sonnentag & Quinton, 2022).

2.4 Data Analysis

2.4.1 Water Chemistry Analysis

We assessed the DOM aromaticity of grab samples using the specific ultraviolet absorbance at 254 nm (SUVA₂₅₄), where SUVA₂₅₄ values increase with DOM aromaticity (Weishaar et al., 2003). We calculated SUVA₂₅₄ by dividing the decadal absorbance at 254 nm (A₂₅₄, cm⁻¹, corrected for interference by Fe; Weishaar et al., 2003) with concentration of DOC.

We used R (R version 4.1; R Core Team, 2023) package staRdom (Pucher et al., 2019) to calculate the biological index (BIX) and humification index (HIX) and characterized the DOM pool into components with parallel factor analysis (PARAFAC) using MATLAB toolbox drEEM (Mathworks, Natick, MA, US) following protocol from Murphy et al. (2013) and Oliver et al. (2017). We used the grab samples and autosampler samples to construct the PARAFAC model (total n = 137; n grab samples = 48, n autosampler samples = 89). Eight samples were excluded from the model construction due to high leverage, and component intensities were projected for these samples after the model was finalized. We identified five unique components (Figure S1 in Supporting Information S1) that were validated through half-split analysis and found multiple matches for each component on the online fluorescence database, OpenFluor (Murphy et al., 2014). The maximum fluorescence of excitation and emission in Raman units (Fmax) was used to calculate the percent contribution of each component to total fluorescence, as the fluorophores' actual structure is unknown (Oliver et al., 2017). Table S4 in Supporting Information S1 summarizes the characteristics of the absorbance and fluorescence analyses.

We used the R package vegan (Oksanen et al., 2022) to perform a principal component analysis (PCA) to examine relationships among the absorbance and fluorescence indices from grab samples at both creeks. We used the PCA's first principal component (PC1) scores to indicate aromatic quality. We further used a two-sample t-test to compare the overall water chemistry concentrations of selected analytes between the sites (Table S5 in Supporting Information S1). A pairwise t-test was additionally used to compare selected water chemistry analytes at higher (discharge >1 m³ s⁻¹) and lower (discharge <1 m³ s⁻¹) flow.

2.4.2 Concentration-Discharge Analysis

To compare flow annually and between the study catchments, we calculated runoff (mm day⁻¹) by dividing the daily discharge rate by catchment area. To examine the response of Hg, MeHg, and DOC concentration to discharge, we determined chemodynamic (where analyte concentration varies in response to discharge) or chemostatic (a non-systematic or limited variation in analyte concentration in response to discharge) relationships between analytes and discharge (Pohle et al., 2021). We calculated β, the log-log slope between discharge and the analyte concentrations. We used β as an indicator of transport limitation (β > 0), where analyte sources are present on the landscape and become connected as discharge increases, source limitation (β < 0), where finite analytes are depleted or mixed with more dilute waters, or chemostasis (β = 0; Koger et al., 2018; Musolff et al., 2015). In order to determine the stability or variability of the analyte concentration relative to discharge, we plotted β against the ratio of coefficients of variation of analyte concentration over discharge (CV_C/CV_Q = [µ_Qσ_C]/[µ_Cσ_Q]), where µ_C and µ_Q are mean values of concentration and discharge, respectively, and σ_C and σ_Q are standard deviations. In the case of β ≈ 0, the analyte concentrations may vary little with discharge or exhibit large but non-systematic variability. Chemostatic behavior was defined as −0.2 < β < 0.2 and CV_C/CV_Q < 0.5 (Koger et al., 2018; Musolff et al., 2015). We additionally examined three high-flow events per site (via identification of the inflection point on the hydrograph and confirming a >10% increase in discharge) in spring, early summer, and late summer of 2019. The flushing index (FI = [C_Qpeak,norm]–[C_{Qinitial,norm}]) was calculated to indicate whether flushing or diluting was observed during the rising limb. We identified the direction of hysteresis, where clockwise hysteresis indicated flushing (i.e., higher concentrations during the rising limb than the falling limb), anticlockwise hysteresis indicated a delayed transport response, and no hysteresis indicated an intermediate response (Gandois et al., 2021; Pohle et al., 2021; Shogren et al., 2021; Vaughan et al., 2017).

2.4.3 Flux Estimates

We modeled daily bulk THg, MeHg, and DOC fluxes from measured concentrations and discharge using a web-based tool (Park et al., 2015) developed using the USGS constituent-load modeling software LOADEST (Runkel et al., 2004), spanning April–September 2019, 2020, and 2021. Daily constituent fluxes were calculated from daily discharge measurements based on a calibration regression model developed from paired concentration and discharge data; one outlier MeHg concentration value was removed from the mixed catchment regression model. A regression derived from Spectro::lyser A₂₅₄ was used to predict daily DOC concentrations for a subset of our observations (2019) to use within the LOADEST DOC models to ensure hard-to-capture high-flow events were included in our modeling framework (Figure S2 in Supporting Information S1). To contextualize the 2019–2021 yields, we included 2006–2021 discharge data in the LOADEST analyte flux models. Historical Water Survey of Canada discharge data were used for the peatland catchment (station 0ED009), while the mixed catchment discharge was estimated as described above from the combined mixed catchment and nearby Ochre River record (station 10HC008; Environment and Climate Change Canada, 2021). Concentration-discharge relationships can vary over time, so we examined possible scenarios and ranges in Hg, MeHg, and DOC fluxes between 2006 and 2021 rather than directional trends. We used the Adjusted Maximum Likelihood Estimator calibration regressions and selected the model based on the lowest Akaike Information Criterion using the default model (Model 0) and those without a linear time function (acceptable models = 1, 2, 4, 6). Models are nested within Equation 1, where a_i is the i^th model coefficient, ln Q is ln(discharge) minus the center of ln(discharge), dtime is the decimal time minus the center of decimal time, Q is discharge in ft³ s⁻¹, and flux is in mass day⁻¹. The calibration data is centered in the model to eliminate collinearity through methods described by Cohn et al., 1992. The LOADEST output model coefficients and fit (R²) are in Table S6 in Supporting Information S1.

$$$\mathrm{ln}\,\text{load}=\mathrm{a}0+\mathrm{a}1\,\mathrm{ln}\,Q+\mathrm{a}2\,\mathrm{ln}\,{Q}^{2}+\mathrm{a}3\,\mathrm{sin}(2\pi \text{dtime})+\mathrm{a}4\,\mathrm{cos}(2\pi \text{dtime})$$$ (1)

We obtained yields from the sum of daily fluxes within the yearly sampling period normalized to the catchment area. The spring period was defined as 1 April to the end of high freshet flows, while the summer period was defined from the end of freshet to 30 September (Burd et al., 2018).

3 Results

3.1 Meteorological Conditions and Hydrology

The mean annual air temperature over the 3 years was higher at both sites than long-term averages (Table 1). 2020 had the highest total annual rainfall for both sites, especially at the peatland catchment, which had 474 mm. Total annual rainfall in 2019 and 2021 was lower and varied little between sites. Cumulative runoff from the spring and summer of 2019 to 2021 at the peatland catchment was 61, 311, and 142 mm, respectively, while the mixed catchment totaled 87, 146, and 86 mm (Figure 2, Table S7 in Supporting Information S1). Summer runoff, occurring in response to storm events rather than snowmelt, was greater than spring freshet in 2019 and 2020 at both catchments, while freshet dominated the seasonal hydrographs in 2021, at 79% of runoff at the peatland catchment and 51% of runoff at the mixed catchment. During the monitoring period, the mixed catchment exhibited similar runoff patterns as the nearby gauged Ochre River (Table S3 in Supporting Information S1). While the mixed catchment does not have long-term hydrometric data, we utilized the long-term record at Ochre River to compare its variability with the peatland catchment (2006–2021; Table S8 in Supporting Information S1). The longer-term records showed cumulative runoff in the peatland catchment to exhibit far greater variability (mean ±1 standard deviation [SD]: 164 ± 91 mm year⁻¹; coefficient of variation [CV] 56) compared to the proxy for the mixed (236 ± 42 mm year⁻¹; CV 18).

Figure 2

Catchment runoff comparison of 2019–2021 hydrographs with the historical record from 2006 to 2021 at (a) Scotty Creek—Peat, and (b) Smith Creek—Mixed, with Smith Creek's pre-2019 record predicted from Ochre River discharge; see Table S3 in Supporting Information S1. Historical runoff data are maintained by the Water Survey of Canada. The legend in (b) applies to panels (a) and (b). (c, d) Cumulative runoff, separated by spring and summer (April–September 2019–2021). The legend in (d) applies to panels (c) and (d). The mean partitioning and range of runoff from 2006 to 2021 were from discharge records of Scotty Creek, and Ochre River records were used to predict pre-2019 discharge at Smith Creek. The spring period is defined between 1 April and the end of high freshet flows, and summer is defined between the end of freshet to 30 September. Smith Creek flow record in 2019 ranged from 27 May to 28 September; cumulative runoff is based on this period. IQR, interquartile range.

3.2 Water Chemistry and DOM Composition

Relative to the mixed catchment, the peatland catchment had lower EC (mean ± SD: 190 ± 32 vs. 620 ± 200 μS cm⁻¹), pH (7.5 ± 0.3 vs. 8.1 ± 0.1), and bulk THg concentrations (1.13 ± 0.30 vs. 3.13 ± 2.17 ng THg L⁻¹), with similar concentrations of bulk MeHg (0.09 ± 0.03 vs. 0.08 ± 0.04 ng MeHg L⁻¹) and higher concentrations of DOC (19.5 ± 2.6 vs. 11.9 ± 4.3 mg DOC L⁻¹; Figure 3; Table S5 in Supporting Information S1). At the peatland catchment, filtered THg concentrations were lower than the mixed catchment, but filtered MeHg concentrations were higher (0.93 ± 0.31 vs. 1.19 ± 1.14 ng THg L⁻¹; 0.08 ± 0.04 vs. 0.05 ± 0.01 ng MeHg L⁻¹; Table S5 in Supporting Information S1). The proportion of bulk THg as MeHg (%MeHg) was higher at the peatland catchment than at the mixed catchment (8.1 ± 3.5 vs. 3.4 ± 1.8%MeHg). At both sites, MeHg and THg were predominantly in the dissolved fraction, with greater dissolved proportions at the peatland catchment (dissolved MeHg = 88% ± 11%; dissolved THg = 77% ± 16%) than the mixed catchment (dissolved MeHg = 71% ± 22%; dissolved THg = 57% ± 17%).

Figure 3

Seasonal analyte concentrations and yields. Concentrations of filtered (2019) and bulk concentrations and yields (2019–2021) of (a–d) total mercury (THg), (e–h) methylmercury (MeHg), and (i–l) dissolved organic carbon (DOC) in Scotty Creek—Peat and Smith Creek—Mixed. Yields of bulk THg, MeHg, and DOC from April–September 2019–2021 are normalized by catchment area and displayed by season. The mean partitioning and range of analyte yields from 2006 to 2021 were modeled from discharge records of Scotty Creek, and Ochre River records were used to predict pre-2019 discharge at Smith Creek.

When comparing the DOM composition between the two catchments, we found the peatland catchment had higher SUVA₂₅₄, HIX values, and higher contributions of humic-like PARAFAC component C1 (Figure 4; Figure S1 in Supporting Information S1). By contrast, the DOM composition at the mixed catchment spanned a wider range of aromaticity than the peatland catchment (Figure 4).

Figure 4

Seasonal dissolved organic matter (DOM) composition. Principal component analysis (PCA) with shapes indicating high (discharge, Q ≥ 1 m³ s⁻¹) and low (Q ≤ 1 m³ s⁻¹) flow periods as well as the month of the year in Scotty Creek—Peat and Smith Creek—Mixed. DOM indices are described in Table S4 in Supporting Information S1, and PCA loadings are in Table S9 in Supporting Information S1. C1–C5, parallel factor analysis components; where C1, terrestrial humic-like; C2, microbial humic-like; C3, fulvic-like; C4, terrestrial humic-like; C5, protein, tryptophan-like; SUVA, specific ultraviolet absorbance at 254 nm; hix, humification index; bix, biological index.

We sampled a groundwater spring in the mixed catchment in July of 2021 and found low concentrations of DOC (1.22 mg L⁻¹), THg (0.91 ng L⁻¹), and MeHg (0.03 ng L⁻¹) with high EC (1,600 μS cm⁻¹) and ions (266 mg Na L⁻¹; 179 mg Ca L⁻¹; 188 mg SO₄-S L⁻¹), but similar pH to stream measurements (7.8). By comparison, peatland (i.e., peat plateaus, bogs, and fens) porewater sampled at both catchments in 2021 had high mean DOC (44.0 ± 11.7 mg L⁻¹), THg (19.10 ± 10.50 ng L⁻¹), and MeHg concentrations (0.32 ± 0.29 ng L⁻¹) with low values of EC (56 ± 29 μS cm⁻¹) and pH (5.1 ± 0.8) (Thompson et al., 2022).

Due to COVID-19 constraints, early spring sampling occurred only in 2019; at the peatland catchment, MeHg and DOC concentrations were relatively high in March 2019, temporarily declined, increased through early summer, and declined through late summer. By contrast, THg, MeHg, and DOC concentrations at the mixed catchment were relatively low in spring, increased with early summer high flows, and showed a consistent decline toward late summer (Figure 3). DOM composition varied through the year, with changing aromaticity associated with flow conditions and sampling months (Figure 4).

3.3 Concentration-Discharge Relationships

Low-flow, early-season samples at the peatland catchment were associated with a microbial humic-like peak (C2, Figure 4), lower concentrations of DOC, and higher EC values (Figure S3 in Supporting Information S1). In contrast, higher flow periods were associated with higher SUVA₂₅₄ and HIX regardless of the season (Figure 4). The mixed catchment similarly had flow-related water chemistry differences, which were more pronounced than the peatland catchment. Two distinctive regimes were present in the mixed catchment: low flows had higher EC and Na and lower DOC and THg concentrations (Figure S3 in Supporting Information S1) with microbial/protein-like DOM composition during early and late season sampling (microbial humic-like C2 and protein-like C5; Figure 4). In comparison, high flows had lower EC and higher DOC (Figure S3 in Supporting Information S1) with aromatic, terrestrial DOM composition corresponding with humic-like components in early to mid-summer samples (C3 and C4; Figure 4).

We observed chemostatic relationships of THg, MeHg, and DOC at the peatland catchment and chemodynamic flushing responses to increased discharge at the mixed catchment when inputting all grab samples from the monitoring period (Figure 5). We plotted the log-log slope of concentration-discharge (β) and the relative stability of the response (coefficients of variation of analyte concentration over discharge, CV_C/CV_Q). From regressions of untransformed THg, MeHg, and DOC concentrations against discharge for the peatland catchment, only DOC significantly increased with discharge (R² = 0.26, p < 0.001), as observed in the paired high-low discharge analysis (Figure S3 in Supporting Information S1). At the mixed catchment, THg concentrations responded most strongly to discharge (R² = 0.68, p < 0.001), followed by DOC (R² = 0.47, p < 0.001), in contrast to no statistically significant response of MeHg to discharge, which is reflected by its higher CV_C/CV_Q value in Figure 5a.

Figure 5

Concentration-discharge relationships, 2019–2021. Evaluation of relative chemostatic or chemodynamic behavior in Scotty Creek—Peat and Smith Creek—Mixed, for observations of bulk total mercury (THg; n = 24, 24, respectively), methylmercury (MeHg; n = 23, 24), and dissolved organic carbon (DOC; n = 53, 52) by plotting (a) the log-log slope between discharge and the analyte (β) versus coefficients of variation of analyte concentration over discharge (CV_C/CV_Q), and (b–d) untransformed concentration and discharge (Q) values with model formula, adjusted R², and p values of the regressions and a 95% confidence interval surrounding the best-fit line if statistically significant (p < 0.05).

We found that changes in flow increased the proportion of particulate MeHg with discharge at the mixed catchment (R² = 0.64, p < 0.01, Figure S4 in Supporting Information S1). Interestingly, the proportion of particulate Hg significantly decreased with discharge at the peatland catchment (R² = 0.45, p = 0.014) and did not significantly vary with discharge at the mixed catchment (Figure S4 in Supporting Information S1).

As biogeochemical processes governing the chemistry of stream water sources shift through the year, we selected three flow events in 2019 where A₂₅₄ (DOC proxy; Figure S2 in Supporting Information S1) was monitored at high frequency (Figure 6). Corresponding to spring, early summer, and late summer, event-scale analysis of concentration-discharge metrics showed variability throughout the 2019 monitoring period in the response of DOM to changing flow conditions.

Figure 6

Event-scale concentration-discharge relationships, 2019. High-frequency monitoring of absorbance at 254 nm (A₂₅₄) in 2019 plotted with runoff at (a–d) Scotty Creek—Peat and (e–h) Smith Creek—Mixed. Shading identifies the timeframe of the A₂₅₄–discharge (Q) hysteresis plots, normalized per high flow event, during spring, early summer, and late summer of 2019. The arrow indicates the start of the high-flow event. FI, flushing index; β, log-log slope between discharge and A₂₅₄; CW, clockwise hysteresis; ACW, anticlockwise hysteresis; NH, no hysteresis.

At the peatland catchment, the FI and hysteresis response of A₂₅₄ to discharge varied through the open-water season despite β of the full-season and event-scale analyses falling within the chemostatic threshold (−0.2 > β < 0.2). In spring and early summer, FI > 0 indicated A₂₅₄ enrichment, although the response was quicker in spring (clockwise hysteresis) than in early summer (anticlockwise hysteresis). Rapid A₂₅₄ dilution was observed in late summer (FI < 0, clockwise hysteresis).

In all three events at the mixed catchment, we observed FI > 0 and β > 0.2, indicating consistent enrichment of A₂₅₄ and transport limitation (Figure 6). However, the hysteresis responses at each event differed; in spring, anticlockwise hysteresis indicated a slow response from either low connectivity or distant sources. During early summer, a lack of hysteresis indicated an intermediate response due to variable sources and pathways. In late summer, A₂₅₄ at the mixed catchment exhibited a clockwise response, as an indication of a fast flushing response from near sources or high connectivity. Throughout the open water season, we thus observed an increasingly quick enrichment of A₂₅₄ in response to increased discharge at the mixed catchment (Figure 6).

3.4 Relationships Between Biogeochemical Characteristics and Hg Forms

DOC concentrations were associated more strongly with MeHg concentrations at the peatland catchment and with THg concentrations at the mixed catchment (Figure 7). DOM composition additionally had varying associations with Hg forms in the creeks. Filtered THg and MeHg concentrations at the peatland catchment and THg concentrations (bulk and filtered) at the mixed catchment increased with PC1 scores (Figure 7), corresponding to increasing aromatic quality. Both bulk and filtered MeHg increased with A₂₅₄ at the peatland catchment (bulk MeHg-A₂₅₄: R² = 0.20, p = 0.01; filtered MeHg-A₂₅₄: R² = 0.62, p = 0.002), while bulk and filtered THg increased with A₂₅₄ at the mixed catchment (bulk THg-A₂₅₄: R² = 0.76, p < 0.001; filtered THg-A₂₅₄: R² = 0.87, p < 0.001). We further examined relationships between Hg forms and PARAFAC components (Figure S5 in Supporting Information S1). Bulk THg concentrations at the mixed catchment decreased with the microbial-derived, humic-like C2 and increased with the fulvic-like and terrestrial-humic components (C3, C4). Filtered THg concentrations at both sites decreased with the microbially derived C2 and increased with the terrestrial-humic component C4. At the peatland catchment, filtered THg concentrations increased with terrestrial-humic C1 (note that p = 0.05) and at the mixed catchment increased with the fulvic-like component C3. Filtered MeHg concentrations decreased with a microbially derived, humic-like component (C2) and increased with a terrestrial-humic component (C4) at the peatland catchment.

Figure 7

Relationships between geochemical parameters and mercury (Hg) forms. Regressions of dissolved organic carbon (DOC), scores of principle component 1 (PC1) from principal component analysis on absorbance and fluorescence indices (Figure 4, Table S9 in Supporting Information S1), electrical conductivity (EC), and water temperature in Scotty Creek—Peat, and Smith Creek—Mixed, against (a–d) bulk total Hg (THg; n = 24, 24 respectively), (e–h) filtered THg (n = 11, 11), (i–l) bulk methylmercury (MeHg; n = 24, 23), and (m–p) filtered MeHg (n = 11, 11) separated by the site. Water temperature data are from May–September 2019. Model formula, adjusted R², and p values of the regressions displayed, and a 95% confidence interval surrounds the best-fit line if statistically significant (p < 0.05).

Increasing EC corresponded to decreasing concentrations of filtered THg and MeHg at the peatland catchment and bulk and filtered THg at the mixed catchment (Figure 7). However, EC exhibited a much narrower range at the peatland catchment (129–244 µS cm⁻¹) than at the mixed catchment (263–1,064 µS cm⁻¹). In addition, EC significantly decreased with increasing discharge at both sites (R² = 0.41, p < 0.001 at peatland catchment; R² = 0.64, p < 0.001 at mixed catchment).

Bulk and filtered MeHg increased with the water temperature at the peatland catchment, as did filtered THg (Figure 7), with no significant relationships for the same parameters at the mixed catchment. We further explored indicators of MeHg production (Figure S6 in Supporting Information S1), finding strong increases in bulk and filtered %MeHg, MeHg:DOC, and MeHg:A₂₅₄ with the water temperature at the peatland catchment. Conversely, bulk %MeHg decreased with the water temperature at the mixed catchment, while MeHg:DOC and MeHg:A₂₅₄ did not significantly relate to water temperature at the mixed catchment.

3.5 Yields of THg, MeHg, and DOC

Cumulative spring and summer yields from 2019 to 2021 of bulk THg ranged from 76 to 406 ng THg m⁻² at the peatland catchment and 405–645 ng m⁻² at the mixed catchment (Figure 3, Table S7 in Supporting Information S1). Cumulative bulk MeHg yields ranged from 5.21 to 36.3 ng MeHg m⁻² at the peatland catchment and 6.06–10.0 ng MeHg m⁻² at the mixed catchment, while cumulative DOC yield ranged from 1.16 to 6.63 g DOC m⁻² at the peatland catchment and 1.37–2.37 g DOC m⁻² at the mixed catchment. At both sites, 2021 was the only year spring analyte yields comprised ≥half of the cumulative yield.

We used the 2006 to 2021 discharge record from the peatland catchment and mixed catchment (2019–2021: measured; 2006–2018: predicted from the mixed catchment proxy) to estimate scenarios of yield magnitude and seasonal partitioning for the creeks (Figure 3, Table S7 in Supporting Information S1). At the peatland catchment, the mean cumulative yields for spring and summer were estimated to be 209 ng THg m⁻², 18 ng MeHg m⁻², and 3.0 g DOC m⁻². Freshet contributions were highly variable between years; from 2006 to 2021, spring freshet delivered between 15% and 80% of the cumulative analyte yields at the peatland catchment. At the mixed catchment, the mean cumulative yields from 2006 to 2021 were estimated to be 443 ng THg m⁻², 6.6 ng MeHg m⁻², and 1.5 g DOC m⁻². Spring freshet yields of THg and DOC at the mixed catchment typically outsized summer yield when examining longer-term estimates, while MeHg was evenly divided or had higher summer yields.

4 Discussion

Few studies have monitored stream chemistry and quantified yields of Hg forms in small to mid-sized catchments in permafrost regions. To understand processes that impact the concentrations and yields of THg, MeHg, and DOC in the face of ongoing climate change, we conducted 3 years of open-water season sampling within the discontinuous permafrost zone of boreal western Canada, focusing on a peatland-dominated catchment and a mixed-landscape catchment of similar area (∼150 km²). Below, we discuss temporal variabilities of concentrations and yields from the two catchments and link the processes likely responsible for driving variability to climate change and permafrost thaw.

4.1 Comparing Catchment Water Chemistry and DOM Composition

Overall, the water chemistry and DOM composition in the peatland catchment were characteristic of wetland sources (Bourbonniere, 2009), with higher DOM aromaticity and DOC concentrations and lower pH and EC than the mixed catchment. The water chemistry and DOM composition exhibited greater variability in the mixed catchment. Notably, we found much higher EC and THg concentrations with a greater particulate fraction of Hg forms relative to the peatland catchment. The peatland catchment had higher filtered MeHg concentrations than the mixed catchment but similar bulk MeHg concentrations. The THg and MeHg concentrations in the peatland catchment and mixed catchment were relatively low but within previously observed ranges for boreal streams in Canada (0.46–20.46 ng THg L⁻¹; 0.03–4.82 ng MeHg L⁻¹; Fink-Mercier, Lapierre, et al., 2022; Lam et al., 2022; Thompson et al., 2023).

We found stronger associations of MeHg-DOM in the peatland catchment, while THg-DOM was more strongly correlated in the mixed catchment. Similarly, global assessments have found THg-DOM relationships to be typically stronger than MeHg-DOM in surface waters (Lavoie et al., 2019; Wu et al., 2022), although more robust coupling between MeHg and aromatic DOM has been observed in wetland-influenced boreal streams, with some seasonal variability (Fink-Mercier, del Giorgio, et al., 2022; Mangal et al., 2022).

4.2 Hydrological Influences on Catchment Water Chemistry

In the peatland catchment, discharge had no consistent influence on MeHg or THg concentrations, although higher discharge had a moderately positive effect on DOC concentrations and aromaticity. The widespread peatlands were likely responsible for the muted influence of discharge on MeHg, THg, and DOC concentrations. Channel fens constitute much of the stream network in the peatland catchment (Haynes et al., 2022) and are likely to act as biogeochemical reactor sites for MeHg, THg, and DOC in addition to conveying water (Gordon et al., 2016; Poulin et al., 2019; Tarbier et al., 2021). Regardless of flow conditions, analytes from other land covers would then move through the fen network, where they are degraded or transformed and mixed with internal sources, thus reducing the overall variability in analyte concentrations delivered to the outflow (Olefeldt & Roulet, 2012; Poulin et al., 2019; Tarbier et al., 2021; Figure 8). A Siberian watershed influenced by peatlands likewise exhibited overall chemostatic DOC behavior, with a narrow range of DOC concentrations when considering all observations, whereas event-scale analyses elucidated seasonal shifts in hysteresis loops and flushing relationships (Gandois et al., 2021).

Figure 8

Conceptual diagram of analyte delivery from water sources to creeks. Hydrologic pathways and relative concentration trends of a peatland catchment (Scotty Creek) compared to a mixed catchment (Smith Creek) during high and low flow periods. DOC, dissolved organic carbon; THg, total mercury; MeHg, methylmercury.

Examining event-scale patterns of water chemistry from different times of the year can provide further insight into shifting water sources and MeHg mobilization. While MeHg is challenging to monitor at high temporal frequency, we found that dissolved MeHg strongly correlated with A₂₅₄ in the peatland catchment, and we measured A₂₅₄ at high frequency through 2019. The modest and late snowmelt of 2019 rapidly enriched A₂₅₄ at the start of the freshet in the peatland catchment. We interpret this enrichment to signal a shift from groundwater contributions with lower DOM to DOM influenced by surficial organic soils. In contrast, A₂₅₄ dilution from precipitation has been previously observed in peatland catchments during spring freshet with substantial snowmelt (Burd et al., 2018; Gandois et al., 2021). A subsequent high-flow event lasting nearly a month through early summer showed slower A₂₅₄ enrichment through anticlockwise hysteresis. Commonly observed in boreal catchments, anticlockwise hysteresis suggests an initial source close to the creek (e.g., riparian zones) supplied DOM, followed by increasing contributions from peatlands with higher DOC and MeHg concentrations (Gordon et al., 2016; Shogren et al., 2021; Thompson et al., 2022). The rapid dilution of DOM in response to a late summer high-flow event suggested the initial mobilization of nearby high DOM sources and lower DOM contributions after peak flows, possibly due to increasing groundwater influence during late summer (Burd et al., 2018) that likely delivered lower MeHg concentrations. Given the correlation between dissolved MeHg and A_254, high-frequency tracking of A₂₅₄ can plausibly inform MeHg co-transport in peatland-influenced streams where it is not practical to frequently sample MeHg (Thompson, 2021).

The chemodynamic flushing relationship between analyte concentrations and discharge at the mixed catchment suggested consistent transport limitation of THg and DOC, with a less systematic response from MeHg. During low flow periods, groundwater contributions likely controlled the high EC and low concentration inputs of THg, MeHg, and DOC (Figure 8); DOM adsorption in mineral soils can lead to low DOC concentrations and less aromatic DOM quality (Kaiser & Kalbitz, 2012; Kothawala et al., 2012). With permafrost thaw, high-latitude catchments across the Interior Plains are expected to have increased groundwater interaction with surface waters (Wright et al., 2022). We believe that periods of high flow led to increased hydrological connectivity of organic-rich and riparian soils that contributed higher THg and DOC (Figure 8), as shown from chemodynamic flushing responses that aligned with DOC behavior in organic-rich tundra (Shogren et al., 2021) and heterogeneous catchments across Scotland (Pohle et al., 2021).

Event-scale behavior of DOM at the mixed catchment investigated through high-frequency A₂₅₄ monitoring showed DOM enrichment in response to increased flow. In addition, we found increasingly rapid enrichment throughout the summer, which may be linked to enhanced connectivity from high antecedent moisture conditions in the early summer event and the seasonal evolution of subsurface flow paths due to the deepening of the active layer in organic soils (Lafrenière & Lamoureux, 2019). These findings likely reflect patterns of THg concentrations that correlated well with A₂₅₄ in the mixed catchment. Overall, the high-frequency sampling indicated consistent transport limitation during the open water period at the mixed catchment in contrast to the peatland catchment.

In the mixed catchment, erosion was a potential contributor to sediment-bound THg and MeHg. The proportion of particulate MeHg increased with discharge; while THg did not linearly increase with discharge, the catchment had a consistently high particulate fraction of THg relative to the peatland catchment. In catchments with readily erodible sediments, sediment supply has been identified as an important control on concentrations and fluxes of THg and MeHg (e.g., Staniszewska et al., 2022; St. Pierre et al., 2018). The mountainous regions of the mixed catchment could thus deliver sediments, and associated metals, to the catchment outlet.

4.3 Biogeochemical Influences on Catchment Water Chemistry

Stream water temperature, a likely proxy of metabolic activity in peatlands, was indicated to influence the mobilization of MeHg in the peatland catchment. As such, MeHg is likely to be production-limited in the peatland catchment. In 2019, MeHg concentrations increased with rising water temperature, as did %MeHg, MeHg:DOC, and MeHg:A_254, suggesting enhanced summertime microbial methylation (Bravo & Cosio, 2020; Shanley et al., 2022; Wasiuta et al., 2019). In addition, filtered MeHg concentrations increased with DOC concentrations and terrestrially-derived DOM; these results provide further evidence of wetlands as environments where MeHg concentrations strongly associate with aromatic DOC (Branfireun et al., 2020; Lavoie et al., 2019). We inferred that higher intensity MeHg co-transport with aromatic DOM (i.e., higher MeHg:A₂₅₄) occurred during warm summer periods of high MeHg production. During low flow, early-season periods, lower intensity co-transport occurred when DOM quality at the peatland catchment was less aromatic. Therefore, seasonality must be considered when interpreting A₂₅₄ as a proxy for MeHg concentrations (as referenced previously). Boreal Quebec streams had similar trends, where MeHg:DOC relationships were strongest in summer and fall relative to winter and spring (Fink-Mercier, del Giorgio, et al., 2022). Production limitations within the mixed catchment may still be present, but other catchment processes (e.g., transport limitation and flushing) were likely more influential.

4.4 Drivers of Inter- and Intra-Annual Variability in Catchment Yields

Low inter-annual variability in analyte concentrations at the peatland catchment was contrasted by high inter-annual variability in runoff and analyte yields. For example, a particularly high-flow year at the peatland catchment (2020) resulted in a ≥400% increase in cumulative runoff and yields of THg, MeHg, and DOC relative to a low-flow year (i.e., 2019). The predominance of peatlands likely contributed to this variability, where high runoff generation occurred upon the exceedance of storage thresholds, but low runoff occurred during dry periods when the landscape retains water (Goodbrand et al., 2019). By contrast, the cumulative runoff and analyte yields varied by less than two-fold at the mixed catchment, reflecting the lower inter-annual variability in longer-term runoff trends inferred from the nearby gauged Ochre River. Consistent runoff generation at the mixed catchment can be connected to the steeper topographic relief and groundwater inputs from deep flow paths with long residence times that explain the lower inter-annual runoff variability (Utting et al., 2013).

We found freshet analyte yields to be variable, despite the importance of spring freshet highlighted in previous studies of northern catchments (Burd et al., 2018; Gandois et al., 2021; Zolkos et al., 2020). The continental climate of the Interior Plains can result in variable precipitation during winter and summer; the balance of freshet versus summer runoff depends on snow cover and summer storms. This variability was highlighted at both sites by the occurrence of summer storms in 2020 that contributed to high summer yields (52%–79% of cumulative yields) in comparison to the high freshet yields in 2021 (51%–84% of cumulative yields) likely driven by above-average snow pack and high antecedent moisture from the previous year (Sonnentag, 2022; Sonnentag & Quinton, 2022).

The ∼monthly sampling frequency during the open-water season lends uncertainty to our yield estimates, given our findings of DOM shifts in response to short-term events such as rainfall. Still, the grab sample collection spanned a range of flow conditions, including high flows (Figure 3). The absence of samples during the ice-covered season precludes a full annual assessment of THg, MeHg, and DOC yields. Samples collected prior to snowmelt (March 2019) at the peatland catchment had relatively high analyte concentrations whereas the mixed catchment had relatively low analyte concentrations, which may be indicative of winter period concentrations although further data collection is necessary. Still, this is likely to comprise a small fraction of annual analyte yields, as indicated by winter runoff trends in the region. The nearby Jean-Marie River is monitored year-round by the Water Survey of Canada (station 10FB005 with historical data from 1972 to 2019; Environment and Climate Change Canada, 2021) and has a mean ∼6% of annual runoff occurring between November and March with ∼94% of runoff between April and October. Likewise, 5% of annual THg yields are estimated during the winter period for the Mackenzie River (Zolkos et al., 2020), although enhanced winter streamflow coupled with increasing water chemistry yields have been observed in northern streams (Spence et al., 2015).

4.5 Contextualizing Catchment Yields Within the Boreal-Arctic

Our work provides data on THg, and MeHg yields from ∼150 km² northern catchments, adding to limited data from small-medium sized catchments in permafrost regions. In our study region, ongoing permafrost thaw in peatlands is expanding the coverage of thermokarst bogs and fens (Wright et al., 2022) which is altering landscape water chemistry and increasing runoff (Gordon et al., 2016; Haynes et al., 2022; Mack et al., 2021; Quinton et al., 2019). However, we observed modest concentrations and yields of THg, MeHg, and DOC compared to other boreal-Arctic catchments across Canada (Table 2).

Table 2. Comparison of Mean Bulk Total Mercury (THg), Methylmercury (MeHg), and Dissolved Organic Carbon (DOC) Yields and Ratios in Canadian Boreal-Arctic Rivers

Site	Area (km^2)	THg (ng m^−2 year^−1)	MeHg (ng m^−2 year^−1)	DOC (g m^−2 year^−1)	MeHg:THg (ng ng^−1)	THg:DOC (ng g^−1)	MeHg:DOC (ng g^−1)	Citation
Thaw slump, Peel River tributary SE^a	5.5 × 10^0	3.8 × 10^5	1.8 × 10^3	1.8 × 10^0	4.7 × 10^−3	2.1 × 10^5	1.0 × 10^3	St. Pierre et al. (2018) and Zolkos et al. (2019)
Chinusaw, QC	4.4 × 10^1	1.2 × 10^3	7.5 × 10^1	7.8 × 10^0	6.0 × 10^−2	1.6 × 10^2	9.6 × 10^0	de Melo et al. (2022) and Fink-Mercier, Lapierre, et al. (2022)
Peatland-dominated	1.3 × 10^2	2.1 × 10^2	1.8 × 10^1	3.0 × 10^0	8.6 × 10^−2	7.0 × 10^1	6.0 × 10^0	This study
Scotty Creek, NWT^b
Mixed-landscape	1.5 × 10^2	4.4 × 10^2	6.6 × 10^0	1.5 × 10^0	1.5 × 10^−2	2.9 × 10^2	4.4 × 10^0	This study
Smith Creek, NWT^b
Old Crow, YU	1.4 × 10^3	9.0 × 10^2	1.0 × 10^1	2.1 × 10^0	1.1 × 10^−2	4.3 × 10^2	4.8 × 10^0	Staniszewska et al. (2022)
Nottaway, QC	5.5 × 10^4	2.9 × 10^3	6.4 × 10^1	7.1 × 10^0	2.2 × 10^−2	4.1 × 10^2	9.0 × 10^0	de Melo et al. (2022) and Fink-Mercier, Lapierre, et al. (2022)
Churchill, MB	2.8 × 10^5	1.3 × 10^2	1.4 × 10^1	1.4 × 10^0	1.1 × 10^−1	9.3 × 10^1	1.0 × 10^1	Kirk and St. Louis (2009) and Liu et al. (2022)
Mackenzie, NWT	1.8 × 10^6	2.0 × 10^3	1.1 × 10^1	1.0 × 10^0	5.6 × 10^−3	1.9 × 10^3	1.1 × 10^1	Emmerton et al. (2013) and Liu et al. (2022)

• Note. Data are ordered by catchment area.
• ^a Calculated from spot discharge measurements and grab sample concentrations.
• ^b Mean cumulative spring-summer yields 2006–2021, missing autumn and winter; see Table S7 in Supporting Information S1.

Our data in comparison to other boreal-Arctic catchments highlight that key catchment characteristics (i.e., permafrost thaw type, land cover, and climate) differently impact yields of THg, MeHg, and DOC (Table 2; see Tank et al., 2020). The yields and yield ratios (MeHg:THg, THg:DOC, and MeHg:DOC) of the peatland catchment were most similar to the Churchill River (Kirk & St. Louis, 2009), a large wetland-influenced river system. The mixed catchment had analyte yields and yield ratios falling in between the wetland-influenced Churchill and the more sediment-laden Old Crow and Mackenzie rivers (Emmerton et al., 2013; Kirk & St. Louis, 2009; Staniszewska et al., 2022), which aligns with mixed-landscape contributions from peatlands, mountains, and uplands. Both the peatland and mixed catchments had lower analyte yields than the Eastern James Bay rivers of variable size (Chinusaw and Nottaway rivers), likely explained by a much wetter regional climate than the dry Interior Plains (de Melo et al., 2022; Fink-Mercier, Lapierre, et al., 2022). The most dramatic yields of THg and MeHg were observed downstream of retrogressive thaw slumps in the Peel Plateau, related to steep relief, high ice content within the widespread permafrost, and erodible Pleistocene tills (St. Pierre et al., 2018; Tank et al., 2020).

4.6 Changing Climate and Hydrology in the Interior Plains

Continued permafrost thaw and rising air and water temperatures may result in enhanced production of MeHg from Hg^II in thaw wetlands (Gordon et al., 2016; Poulin et al., 2019; Tarbier et al., 2021) and mobilization of DOC (Frey & Smith, 2005; Olefeldt et al., 2014). Increased downstream transport of DOC has negative implications for water treatment processes (Matilainen et al., 2010) and elevated MeHg increases the toxicity of country food (N. Basu et al., 2022). Therefore, continued and expanded monitoring of small and mid-sized catchments that contribute terrestrial Hg and DOC to large Arctic rivers and, eventually, the ocean will be required in the coming years. In addition, understanding lateral DOC transport will be necessary for evaluating the carbon sink-source strength of thawing landscapes (Hugelius et al., 2020; Vonk et al., 2019). The ongoing sampling of the peatland catchment will be particularly essential, as the catchment suffered an intense wildfire in October 2022 (Lamberink, 2022) with potential impacts on runoff generation, MeHg production, and analyte loads (Ackley et al., 2021; Burd et al., 2018; Nelson et al., 2021).

Our study suggests that ongoing climate change could influence MeHg yields by reducing limitations on production (via potential increases in temperature-driven Hg^II methylation), and transport (by increased landscape-derived analyte contributions due to increased runoff). We showed runoff to be a primary driver of analyte yields, which was particularly variable at the peatland catchment in a reflection of the runoff-regulating function of peatlands. Increasing analyte yields may be expected with increasing streamflow in peatland dominated catchments of the Interior Plains due to landscape change and permafrost thaw (Mack et al., 2021). Moreover, despite current inter-annual runoff stability at the mixed catchment, chemodynamic increases of THg and DOC concentrations from increased flows imply a heightened sensitivity of analyte yields to enhanced summer precipitation and streamflow predicted in high latitudes (Lafrenière & Lamoureux, 2019).

5 Conclusions

We used a 3-year data set to show that MeHg, Hg, and DOC concentrations and yields in two northern catchments underlain by discontinuous permafrost were governed by the hydrology and biogeochemical processes of their contributing landscapes. Variability in the water chemistry concentrations of the peatland catchment was not primarily driven by discharge, with a signal of summertime Hg^II methylation as MeHg concentrations increased with water temperature. Instead, the peatland catchment's high inter-annual variability in runoff and analyte yields was likely driven by peatland hydrological function shifting from water storage in drier periods to runoff generation in wetter periods. By contrast, the chemodynamic analyte flushing in the mixed catchment reflected shifts from groundwater influence during low flows to the predominance of peatland and riparian zone inputs during high flows. During the sampling period, annual runoff and analyte yields varied little at the mixed catchment. With ongoing permafrost thaw and changes to temperature and precipitation, THg, MeHg, and DOC yields will likely be sensitive to shifting runoff patterns and biogeochemical controls on DOC and MeHg production.

Acknowledgments

Research for this publication was undertaken on the lands and territories of Treaty 11, 8, and 6 of contemporary Canada. We are grateful for analytical support from Mingsheng Ma, Yuting (Tina) Chen, Wendong (Neil) Liu, Crystal Dodge, Meng (Athena) Yun Hua, Allan Harms, Haley Alcock, and Julia Orlova, as well as field support from Dehcho-AAROM, Erik Umbach, Maya Frederickson, Theresa Wilkes, Gabriel Hould Gosselin, Kate Marouelli, Tarek Chicot, and Donovan Landry. Funding for this study was provided by the Canada Foundation for Innovation, Campus Alberta Innovates Program, NWT Cumulative Impact Monitoring Program (CIMP199), Fonds de recherche du Québec Nature et technologies, Natural Sciences and Engineering Research Council (RGPIN-2016-04688, CGS-M, and PGS-D), Dehcho-AAROM, Weston Family Foundation, Northern Scientific Training Program, and University of Alberta Northern Research Award.

Open Research

Data Availability Statement

Data are available on Mackenzie DataStream. Access: University of Alberta and Dehcho-AAROM (2022), https://doi.org/10.25976/y9yi-no57.