Frozen stocks and hidden fluxes: quantifying mercury release from coastal erosion along the Yukon coast, Canada
Abstract
Ongoing climate warming accelerates permafrost thaw and coastal erosion in the Arctic, mobilizing sediment-bound mercury (Hg) that was previously locked in frozen ground. Its neurotoxic and bioaccumulating form, methylmercury (MeHg), poses serious risks for ecosystems, marine biota, and Arctic Indigenous Peoples. To better understand this risk, this study aimed to quantify the present Hg stock in terrestrial sediments of the Yukon Coastal Plain (YCP), Canada, to estimate the annual Hg flux from coastal erosion, and to evaluate the likely fate of released Hg in the marine environment. Hg concentrations were measured for 158 terrestrial and marine sediment samples. Based on previous studies, the YCP was divided into 44 terrain units, consisting of several uniform depth layers. Layer-specific Hg contents were upscaled from the concentrations measured in terrestrial samples of the respective terrain unit and layer. Where no measurement was available, the Hg concentration was predicted using a random forest model. Calculated stocks per terrain unit were combined with erosion rates to derive Hg fluxes. Marine sediment Hg concentrations and MeHg measurements provided insights into offshore transport and transformation pathways. The total Hg stock of the YCP was estimated at 381,080 kg (271,540 to 501,930 kg), of which about 113 kg (87 to 163 kg) are released annually into the Beaufort Sea. Highest Hg stocks and fluxes occurred in high bluffs at Herschel Island–Qikiqtaruk (HIQ) and the eastern part of the mainland coast. Marine sediments exhibited generally lower Hg concentrations than terrestrial sediments, implying that a substantial fraction of the released Hg is transported offshore, re-emitted into the atmosphere, or incorporated into biogeochemical cycling in the nearshore zone. Sedimentary MeHg concentrations were low, but substantial methylation might occur in the water column, where it can effectively be taken up by phytoplankton, entering the Arctic food web. Since sediment data alone cannot resolve exposure risk, future work should target Hg speciation in water and key biota to understand risks for ecosystems and local communities.
Full text
Frozen stocks and hidden fluxes: quantifying mercury release from coastal erosion along the Yukon coast, Canada Master’s Thesis to attain the academic degree of Master of Science (M.Sc.) in Geoecology Submitted by Katharina Jaspers Institute of Environmental Sciences & Geography University of Potsdam 1st supervisor: Dr. Michael Fritz 2nd supervisor: Dr. Anna Irrgang Potsdam, August 18th, 2025
II Katharina Jaspers [email protected] Student number: 820421 1st reviewer Dr. Michael Fritz Telegrafenberg A45S 14473 Potsdam [email protected] 2nd reviewer Dr. Anna Irrgang Telegrafenberg A45S 14473 Potsdam [email protected]
Abstract III Abstract Ongoing climate warming accelerates permafrost thaw and coastal erosion in the Arctic, mobilizing sediment-bound mercury (Hg) that was previously locked in frozen ground. Its neurotoxic and bioaccumulating form, methylmercury (MeHg), poses serious risks for ecosystems, marine biota, and Arctic Indigenous Peoples. To better understand this risk, this study aimed to quantify the present Hg stock in terrestrial sediments of the Yukon Coastal Plain (YCP), Canada, to estimate the annual Hg flux from coastal erosion, and to evaluate the likely fate of released Hg in the marine environment. Hg concentrations were measured for 158 terrestrial and marine sediment samples. Based on previous studies, the YCP was divided into 44 terrain units, consisting of several uniform depth layers. Layer-specific Hg contents were upscaled from the concentrations measured in terrestrial samples of the respective terrain unit and layer. Where no measurement was available, the Hg concentration was predicted using a random forest model. Calculated stocks per terrain unit were combined with erosion rates to derive Hg fluxes. Marine sediment Hg concentrations and MeHg measurements provided insights into offshore transport and transformation pathways. The total Hg stock of the YCP was estimated at 381,080 kg (271,540 to 501,930 kg), of which about 113 kg (87 to 163 kg) are released annually into the Beaufort Sea. Highest Hg stocks and fluxes occurred in high bluffs at Herschel Island–Qikiqtaruk (HIQ) and the eastern part of the mainland coast. Marine sediments exhibited generally lower Hg concentrations than terrestrial sediments, implying that a substantial fraction of the released Hg is transported offshore, re-emitted into the atmosphere, or incorporated into biogeochemical cycling in the nearshore zone. Sedimentary MeHg concentrations were low, but substantial methylation might occur in the water column, where it can effectively be taken up by phytoplankton, entering the Arctic food web. Since sediment data alone cannot resolve exposure risk, future work should target Hg speciation in water and key biota to understand risks for ecosystems and local communities.
Zusammenfassung IV Zusammenfassung Die anhaltende Klimaerwärmung beschleunigt das Tauen von Permafrost und die damit einhergehende Küstenerosion in der Arktis. Sedimentgebundenes Quecksilber (Hg), das zuvor im gefrorenen Boden gespeichert war, wird damit mobilisiert. Seine neurotoxische und bioakkumulierende Form, Methylquecksilber (MeHg), stellt ein erhebliches Gesundheitsrisiko für Ökosysteme, marine Lebewesen und indigene Völker in der Arktis dar. Diese Studie zielte darauf ab, den Hg-Gehalt der terrestrischen Sedimente des Yukon Coastal Plain (YCP) in Nordwest-Kanada zu erfassen, jährliche Hg-Flüsse infolge der Küstenerosion abzuleiten, und nachzuvollziehen, wohin das freigesetzte Hg gelangt. Hierfür wurden Hg-Konzentrationen in mehr als 200 terrestrischen und marinen Sedimentproben gemessen. Frühere Studien haben das YCP in 44 terrain units unterteilt, die jeweils aus mehreren homogenen Tiefenschichten bestehen. Diese Unterteilung wurde aufgegriffen und schichtspezifische Hg-Gehalte wurden aus den Messwerten abgeleitet. Für Schichten, in denen keine Messungen vorlagen, wurde die Hg-Konzentration mithilfe eines Random Forest-Modells vorhergesagt. Die berechneten Hg-Bestände pro terrain unit wurden mit Erosionsraten kombiniert, um jährliche Hg-Flüsse abzuleiten. Messungen von Hgund MeHgKonzentrationen in marinen Sedimenten lieferten Hinweise auf Hg-Transport ins offene Meer und auf Umwandlungsprozesse unter verschiedenen Umweltbedingungen. Der Gesamt-Hg-Bestand des YCP wurde auf 381.080 kg (271.540 bis 501.930 kg) geschätzt, wovon jährlich etwa 113 kg (87 bis 163 kg) in die Beaufortsee freigesetzt werden. Die höchsten Hg-Bestände und -Flüsse traten in den steilen Kliffbereichen von Herschel Island–Qikiqtaruk sowie im östlichen Teil der Festlandküste auf. Die niedrigeren Hg-Konzentrationen in marinen Sedimenten im Vergleich zu terrestrischen Proben weisen auf einen substanziellen Transport des freigesetzten Hg ins offene Meer, eine Abgabe in die Atmosphäre oder eine Aufnahme in biogeochemische Kreisläufe der küstennahen Zone hin. Die niedrigen sedimentären MeHg-Konzentrationen lassen keine direkten Schlüsse auf Folgen für Ökosysteme und die lokale Bevölkerung zu. Hg könnte in erheblichem Umfang methyliert werden, von Phytoplankton aufgenommen werden und so ins marine arktische Nahrungsnetz gelangen. Weitere Studien sollten die Auswirkungen von
Zusammenfassung V Hg im Wasser und in Schlüsselorganismen wie Phytoplankton oder marinen Säugetieren erforschen.
Table of content VI Table of content Abstract .......................................................................................................................................................... III Zusammenfassung ................................................................................................................................ IV Table of content ........................................................................................................................................ VI List of abbreviations ............................................................................................................................... IX List of figures ............................................................................................................................................... X List of tables ...............................................................................................................................................XII 1 Introduction ........................................................................................................................................... 1 2 Objectives ............................................................................................................................................... 5 3 Scientific background ...................................................................................................................6 3.1 Permafrost and coastal dynamics in the Arctic ............................................................................... 6 3.2 Contaminants in permafrost ........................................................................................................................ 8 3.3 Material fluxes from erosion of the Yukon coast............................................................................ 10 4 Study area ............................................................................................................................................. 11 4.1 General setting ...................................................................................................................................................... 11 4.2 Climate and vegetation .................................................................................................................................. 12 4.3 Landscape evolution ......................................................................................................................................... 13 4.4 Coastal dynamics ................................................................................................................................................ 16 4.5 Sediment characteristics in the Canadian Beaufort Sea .......................................................... 17 5 Material and methods ................................................................................................................. 19 5.1 Field work ................................................................................................................................................................. 19 5.1.1 Sampling strategy.................................................................................................................................... 19 5.1.2 Terrestrial samples ................................................................................................................................... 19 5.1.3 Marine sediment samples .................................................................................................................. 21 5.2 Laboratory work ................................................................................................................................................... 21 5.2.1 Sample preparation ................................................................................................................................ 21 5.2.2 Mercury, Carbon, and Nitrogen ...................................................................................................... 22 5.2.3 Methylmercury.......................................................................................................................................... 23
Table of content VII 5.2.3.1 Extraction ................................................................................................................................................... 23 5.2.3.2 Analysis ...................................................................................................................................................... 24 5.2.3.3 Processing ................................................................................................................................................ 25 5.2.4 Grain size ....................................................................................................................................................... 26 5.3 Data analysis ......................................................................................................................................................... 27 5.3.1 Significance and correlation ............................................................................................................ 27 5.3.2 Modeling mercury concentrations ............................................................................................. 28 5.3.2.1 Data preparation .................................................................................................................................. 28 5.3.2.2 Random forest model...................................................................................................................... 28 5.3.2.3 Prediction and uncertainty ........................................................................................................... 32 5.3.2.4 Simpler approaches........................................................................................................................... 32 5.3.3 Estimating stocks & fluxes ................................................................................................................. 32 6 Results ................................................................................................................................................... 34 6.1 Geochemical parameters ............................................................................................................................ 34 6.1.1 Total mercury ............................................................................................................................................ 34 6.1.2 Methylmercury.......................................................................................................................................... 36 6.1.3 Carbon and nitrogen .............................................................................................................................38 6.1.4 Grain size distribution .......................................................................................................................... 42 6.2 Correlation matrices ........................................................................................................................................ 44 6.2.1 Terrestrial samples ................................................................................................................................. 44 6.2.2 Marine samples ....................................................................................................................................... 45 6.3 Mercury stocks & fluxes from the Yukon coast .............................................................................. 46 6.3.1 Model evaluation .................................................................................................................................... 46 6.3.2 Predictions .................................................................................................................................................. 49 6.3.3 Stocks ............................................................................................................................................................... 51 6.3.4 Fluxes .............................................................................................................................................................. 53 7 Discussion ........................................................................................................................................... 58 7.1 Assessment of outcomes and uncertainties ....................................................................................58 7.1.1 Model behavior and performance................................................................................................58 7.1.2 Evaluation and contextualization of mercury stocks and fluxes ............................... 61 7.2 Mercury in the terrestrial coastal domain .......................................................................................... 63 7.2.1 Terrestrial mercury concentrations in the Arctic context ............................................. 63 7.2.2 Spatial variability of mercury stocks ........................................................................................... 64 7.2.3 Drivers of mercury fluxes from coastal erosion .................................................................. 66
Table of content VIII 7.3 Fate of mercury in the marine environment ................................................................................... 67 7.4 Methylmercury—transformations, pathways, potential risk ................................................. 69 7.4.1 Spatial patterns and environmental controls on methylmercury.......................... 69 7.4.2 Nearshore methylmercury dynamics ....................................................................................... 70 7.4.3 Potential risks and implications for Arctic food webs ..................................................... 72 7.5 Mercury in the One Health framework ................................................................................................ 73 8 Conclusion .......................................................................................................................................... 76 9 References .......................................................................................................................................... 78 Appendix .................................................................................................................................................... XIII Acknowledgements ......................................................................................................................... XVII Tools used in this thesis ................................................................................................................ XVIII Selbstständigkeitserklärung ........................................................................................................ XIX
List of abbreviations IX List of abbreviations C carbon CCME Canadian Council of Ministers of the Environment DOM Dissolved Organic Matter g gram Hg total mercury HIQ Herschel Island–Qikiqtaruk ICP-MS Inductively Coupled Plasma Mass Spectrometer kg kilogram LIS Laurentide Ice Sheet M molar MeHg methylmercury µg microgram n number of samples used NaTeB sodium tetraethyl-borate OC Organic Carbon OM Organic Matter rpm revolutions per minute SOC Soil Organic Carbon t ton Tg teragram TN Total Nitrogen TOC Total Organic Carbon WMAC-NS Wildlife Management Advisory Council–North Slope YC24 Yukon Coast 2024 YCP Yukon Coastal Plain yr year
Introduction 4 Previous studies indicate that substantial amounts of Hg are present at the YCP and in the Canadian Beaufort Sea, with recently increasing Hg concentrations in some marine biota (Leitch 2006; Kirk et al. 2012; Emmerton et al. 2013; Braune et al. 2015). The annual Hg contribution from the Mackenzie River was estimated to be more than three times higher than from coastal erosion (Leitch 2006). Still, it remains highly uncertain how much Hg is mobilized from coastal erosion, and what happens with it once it gets released. Despite the expected lower Hg contribution from coastal erosion compared to the one from the Mackenzie River, closing this knowledge gap is important, especially in areas where erosion is the dominant lateral material transport pathway and rivers are small or absent. Hg coming from coastal permafrost can have different chemical and physical characteristics than Hg coming from riverine sources, and might behave differently in marine environments (Du et al. 2024). For the Yukon coast, very few Hg measurements of terrestrial and marine sediments exist, and it is unclear whether Hg released from coastal erosion is of concern in this area and what happens with the released Hg entering the nearshore zone. If buried quickly in nearshore sea floor sediments, it may pose little risk to Arctic food webs. However, if transported and transformed into MeHg in the marine water column, or frequently remobilized in shallow water, it has the potential to affect Arctic ecosystems and human health through dietary exposure.
Objectives 5 2 Objectives The aim of this thesis is to quantify Hg stocks in terrestrial permafrost of the YCP and to estimate Hg fluxes resulting from coastal erosion. Additionally, this study discusses the potential fate of released Hg in the marine environment. To achieve this, Hg concentrations were measured in terrestrial and marine sediment samples, which were obtained during the Yukon Coast expedition (July 2024) and prior field campaigns. To scale up from point measurements to the landscape scale, the terrain segmentation approach from Couture et al. (2018) was used, which divided the YCP into 44 distinct terrain units, each characterized by relatively homogenous properties and depth-layers structures. For layers and terrain units where no measurements were available, a random forest model was applied to fill data gaps by predicting Hg concentrations based on other available environmental parameters. This allowed for the completion of the data set and enabled the calculation of Hg stocks for all terrain units. Estimated stocks were combined with coastal erosion rates to calculate annual Hg fluxes entering the nearshore zone. MeHg concentrations were measured for a certain sample set to gain insight into possible transformation processes and hotspots of methylation. The potential fate of released Hg was discussed based on marine sediment measurements and insights from the literature. Ultimately, this thesis aims to assess whether the released Hg is likely to be buried, transformed, or transported—and what implications this may have for ecosystems and humans.
Scientific background 6 3 Scientific background 3.1 Permafrost and coastal dynamics in the Arctic Permafrost extends over about 15 % of the land masses of the Northern Hemisphere (Obu et al. 2019). In some locations, hundreds of meters thick deposits have accumulated over millennia (Alexeeva & Erbajeva 2000; Schirrmeister et al. 2002; French 2007). Low soil temperatures in permafrost regions strongly suppress microbial activity, limiting decomposition and mineralization of the organic material (OM) coming from flora and fauna. As a result, vast amounts of organic carbon (OC) have accumulated and been preserved in permafrost and the active layer (the layer on top of permafrost which is subject to annual freeze-thaw cycles) (Hugelius et al. 2014). With accelerated thawing of permafrost and melting of the contained ground ice, increasing quantities of this stored carbon are being mineralized and evade into the atmosphere as carbon dioxide or methane, two potent greenhouse gases (Schuur et al. 2015). This creates a self-reinforcing feedback loop, known as the permafrost carbon feedback, where warming accelerates emissions, which in turn intensify warming and further permafrost degradation (Friedlingstein et al. 2001; Schuur et al. 2015). The Arctic Ocean is currently acting as a carbon sink, estimated to sequester about 52 to 226 Tg carbon from atmospheric carbon dioxide annually (Vonk, Fritz et al. 2025). This capacity is increasingly offset by the release of terrestrial OC into the Arctic Ocean—estimated at 4.9-14 Tg yr-1 from coastal erosion and 47.71 (18.5-127) Tg yr-1 from rivers (Vonk, Fritz et al. 2025). The majority of Arctic permafrost coasts is subject to coastal erosion (Irrgang et al. 2022). The combination of ongoing thermal denudation (erosion due to the influence of sensible heat, water, or solar radiation) and thermal abrasion (combined effect of mechanical and thermal energy of water; Aré 1988) leads to permafrost degradation and the formation of thermokarst features in coastal areas (Günther et al. 2013). One such landscape feature resulting from thermal denudation is called retrogressive thaw slump (RTS), a slope failure forming in hilly terrain due to thawing of ice-rich permafrost or melting of massive ground ice (Nesterova et al. 2024). It is characterized by an eroding and retreating headwall cutting into the undisturbed tundra (Fig. 2; Lantuit & Pollard 2005). The eroded
Scientific background 7 material often accumulates in front of the relatively steep headwall, where it forms mud pools together with water from melting ground ice (Nesterova et al. 2024). Usually there are one or more creeks, called thaw streams, through which water and sediments are transported from the mud pools into the ocean. Parts of the RTS floor can stabilize (temporarily) when no more ground ice is present or exposed to warmer air or solar irradiation (Nesterova et al. 2024). RTSs are known to free large amounts of sediment and organic matter (OM) (Weege 2016). Figure 2: Characteristics of retrogressive thaw slumps (modified from Lantuit & Pollard 2005). Permafrost thaw in coastal regions causes the release of sediments, carbon, nutrients (e.g., nitrogen, phosphorus; Reyes & Lougheed 2015; Hugelius et al. 2020; Vonk, Fritz et al. 2025; Wagner et al. 2025), ancient DNAs (Vorobyova et al. 1997), and contaminants (e.g., heavy metals, legacy pollutants; Schuster et al. 2018; Miner et al. 2021) into Arctic nearshore waters. The material input changes chemical and physical properties of the water column, contributing to ocean acidification (Semiletov et al. 2016) and influencing primary productivity (Juma et al. 2025). All these processes alter the ecosystem services that are provided by Arctic coasts, such as providing habitat for marine biota, and ground for housing, subsistence activities, and cultural heritage for Indigenous Peoples (Malinauskaite et al. 2019; Bringloe et al. 2022; Liew et al. 2022). Future Arctic projections show an increase in coastal erosion rates from historical averages of 0.9 ± 0.4 m yr-1 (1850-1950) to 1.6 ± 0.5 m yr-1 (low-emission scenario) or
Scientific background 8 2.6 ± 0.8 m yr-1 (high-emission scenario) by the end of the 21st century (Nielsen et al. 2022). This will set an increasing number of coastal communities and settlements at risk, potentially requiring relocation and increasing social, economic, and psychological costs, including food security and infrastructure loss (IPCC 2022). 3.2 Contaminants in permafrost Permafrost soils have accumulated not only large amounts of OM, but also of heavy metals (Perryman et al. 2020). Very little is known about the total stocks of heavy metals in permafrost. Recent studies have looked at arsenic, cadmium, chromium, cobalt, copper, manganese, mercury, nickel, lead, and zinc, among others (Ji et al. 2020; Perryman et al. 2020). Concentrations of these elements have been found to be elevated in some permafrost soils compared to soils of lower latitudes (Perryman et al. 2020; Miner et al. 2021). Even if concentrations would be low, the total amount in the large permafrost domain could have the potential to significantly set ecosystems at risk of pollution, when released at high rates. This study focuses on Hg, a heavy metal with high density, high thermal and electrical conductivity, and low solubility in water. Hg in the environment occurs predominantly as elemental mercury Hg(0), oxidized mercury Hg(II), and methylmercury MeHg (CH3Hg+). Elemental Hg is liquid under standard conditions and is the dominant form of gaseous Hg in the atmosphere from volcanic activities (Siegel & Siegel 1984) and rock weathering (Wen et al. 2021). Natural emissions are increased by one-third from anthropogenic emissions, including coal combustion, artisanal and small-scale gold mining, and cement and metal production (UNEP 2013; Outridge et al. 2018). Dastoor et al. (2022) estimated that more than 98 % of the total Hg emissions occur in regions other than the Arctic but can be transported into the Arctic. In the atmosphere, Hg(0) has a residence time of about one year, allowing for long-range transport away from local sources and into remote regions such as the Arctic (Lindqvist & Rodhe 1985; Travnikov 2005). The major deposition pathway of atmospheric Hg(0) into terrestrial systems in the Arctic tundra has been found to be the uptake by vegetation (Jiskra et al. 2019). Due to elevated Hg concentrations in the atmosphere since the beginning of the 20th century, Hg concentrations are often higher in the upper active layer than in the
Scientific background 9 underlying permafrost (Olson et al. 2018; Lim et al. 2020). In terrestrial reservoirs and natural waters, Hg(II) is the predominant form of Hg (Skyllberg 2012). This oxidized Hg shows a high affinity for dissolved organic matter (DOM), particles, and sulfur (Skyllberg 2012). Under anoxic conditions and the presence of microorganisms that carry the hgcAB gene pair (sulfate reducing bacteria, iron reducing bacteria, methanogens, among others), Hg(II) can be transformed into MeHg, a neurotoxic compound that bioaccumulates in food webs, with increasing concentrations towards higher trophic levels (Jonsson et al. 2022). MeHg production in Arctic soils is generally limited by cold temperatures and low microbial activity (Yang et al. 2016; Tarbier et al. 2021), and MeHg concentrations in permafrost soils usually make up for only a small fraction (less than 5 %) of total Hg (Olson et al. 2018; Tarbier et al. 2021). Hg stored in Arctic soils can be mobilized and transported into aquatic (terrestrial and marine) systems via riverine transport and coastal erosion (Dastoor et al. 2022). Arctic rivers drain large catchments underlain by permafrost, and deliver Hg bound to DOM and mineral particles into the Arctic Ocean (Schuster et al. 2011), predominantly in the form of Hg(II) (Dastoor et al. 2022). Riverine inputs peak in spring and early summer, when 90 % of the total annual Hg flux coming with Arctic rivers is exported into the oceans (Zolkos et al. 2020). Coastal erosion releases Hg that has been previously locked in frozen sediments directly into the nearshore marine environment. This process primarily introduces particulate-bound Hg(II), often associated with clay surfaces and OM (Brigatti et al. 2005; Rutkowski et al. 2021). Since a lot of this material is derived from deeper (older) sediments, it may contain Hg that has been isolated from biogeochemical cycling for centuries or millennia (Rutkowski et al. 2021). In marine waters, the released Hg(II) can be deposited and buried on the sea floor (particle-bound Hg), where it may remain relatively stable (Braune et al. 2015). Some fractions of Hg(II) can be transformed into MeHg by microbial communities in sediments or in the water column (Jonsson et al. 2022). MeHg production has been observed in Arctic marine environments, especially in shelf seas where terrestrial inputs are high and primary productivity supports microbial activity (Chételat et al. 2022; Jonsson et al. 2022). MeHg produced in the ocean can be demethylated by photo-reduction or biotic demethylation to Hg(0) and can evade back into the atmosphere (Driscoll et al. 2013; Dastoor et al. 2022). Another fraction can enter the
Scientific background 10 base of the marine food web via uptake by phytoplankton and microbes (Watras et al. 1998; Braune et al. 2015). It then biomagnifies through trophic levels, resulting in elevated concentrations in fish, seabirds, and marine mammals, posing risks to both wildlife and Indigenous Peoples that rely on traditional diets (Loseto et al. 2008; AMAP 2021). 3.3 Material fluxes from erosion of the Yukon coast The Yukon coast is characterized by high rates of coastal erosion (Obu et al. 2017; Irrgang et al. 2018). Along the approximately 300 km long coastline, more than 200,000 m2 land are lost each year, releasing about 1.83 × 106 m3 of sediment into the Arctic Ocean (Couture et al. 2018). This annually released sediment is estimated to contain about 35.5 × 106 kg of carbon (Couture et al. 2018). The Hg flux from the Yukon coast has been estimated by Leitch (2006) at approximately 250 kg. These fluxes make the Yukon coast a large contributor of permafrost-derived material entering the Arctic marine environment. The consequences of this release with all its components (carbon, nutrients, contaminants, microbes) are only partially understood. For the Inuvialuit and Gwich’in communities living in this region, subsistence harvesting of marine mammals, fish, and seabirds make up an important part of their diet and culture (Donaldson et al. 2010; AMAP 2015). Because these marine biota rank high in the food web, they can accumulate high levels of MeHg through biomagnification, potentially posing health risks to the people consuming large amounts of such organisms (Dietz et al. 2013). Understanding the sources and pathways of Hg is therefore critical for risk assessments and the development of effective protection and management strategies for both ecosystems and Indigenous Peoples (Grandjean et al. 2010).
Study area 11 4 Study area 4.1 General setting The YCP in the northwest of Canada is a 10 to 40 km wide lowland that gently slopes northward towards the Beaufort Sea (Welsh & Rigby 1971). It is bordered by the Arctic Coastal Plain of Alaska in the west and extends eastward to the Mackenzie Delta. To the south, the YCP is framed by the British, Barn and Richardson mountains (Welsh & Rigby 1971). The island HIQ is part of the YCP and lies approximately 3 km off the Yukon mainland coast on the Canadian Beaufort Sea shelf (Fig. 3). The island covers an area of about 116 km2 and rises to a maximum elevation of 181 m above sea level. Once connected to the mainland, HIQ became an island within the last 1600 years (Burn 2009). Today, the very shallow Workboat Passage separates it from the Yukon mainland coast. When referring to YCP and the Yukon coast in this thesis, HIQ is always included. In 1987, HIQ was designated as a Territorial Park under the Inuvialuit Final Agreement (1984) and the Government of Yukon’s Parks Act with the aim to “protect and sustain the ecological integrity and heritage values for generations to come” (WMAC-NS 2019) and to conserve wildlife, habitat and traditional indigenous use (IFA 1984). At the same time, the Ivvavik National Park of Canada was established, followed recently by the Aullaviat/Anguniarvik Traditional Conservation Area (2023), which together encompass large parts of northwest Canada, including the entire YCP (Government of Canada 2025). The YCP borders the Canadian Beaufort Sea, which is a marginal sea of the Arctic Ocean. Its shallow parts form the Canadian Beaufort Sea shelf with an approximate width of 150 km and up to 80 m water depth (Hill et al. 1991). The Mackenzie Trough (northeast of HIQ) and Herschel Basin (southeast of HIQ) are the only deeper parts of the shelf. Further offshore, the Canadian Beaufort Sea shelf slopes into the adjacent Beaufort Basin which reaches depths greater than 3600 m (Lemenkova 2020). The entire region lies in the continuous permafrost zone (permafrost coverage > 90 %; Obu et al. 2019), and parts of the shelf area are underlain by subsea permafrost (Overduin et al. 2019).
Study area 12 Figure 3: Location of the study area in northwestern Canada. The Yukon coast is divided into 44 terrain units (Couture et al. 2018), which are classified into 8 sections in this study, based on characteristics of the respective terrain units (Tab. 1). The limit of the Last Glacial Maximum is taken from Ehlers et al. (2011). 4.2 Climate and vegetation The climate at the Yukon coast is characterized by continental Arctic air masses during the long, cold winters, and maritime Arctic air during the short, cool summers (Rampton 1982). From 1991 to 2020, the average temperature was -24 °C in February and 8.2 °C in July at the Komakuk Beach weather station (Environment & Climate Change Canada (ECCC) 2025). Compared to the previous period from 1961 to 1990, this represents an increase of 2 °C in February and 0.6 °C in July (ECCC 2025). Rising air temperatures have been confirmed on HIQ, where Burn & Zhang (2009) reported a 2.5 °C rise in mean annual temperature over the 20th century. This local increase is assumed to reflect a broader regional trend beyond the Yukon coast (Burn & Zhang 2009). During the 20th century, mean annual ground temperatures rose both at the top of permafrost and at 20 m depth by 2.6 °C and 1.9 °C, Basemap: ESRI Satellite CRS: EPSG 32607
Study area 13 respectively (Burn & Zhang 2009). The most recent precipitation data from ECCC are available for the period 1971 to 2000, showing a total precipitation of 161.3 mm yr-1 for Komakuk Beach, of which 83.5 mm was rain (ECCC 2025). The two major wind directions along the Yukon coast are northwest and east (Radosavljevic et al. 2022; Brembach et al. 2023). The YCP lies approximately 100 km north of the tree line, with Arctic Tundra as the main vegetation type (Scudder 1997). The region is primarily covered by tussock tundra with cotton grass (Eriophorum spp.) as the dominant vegetation type (Raynolds & Walker 2022). Other common plant types include dwarf shrubs, sedges, grasses, mosses, and lichens (Raynolds & Walker 2022). The vegetation communities vary locally and are strongly influenced by local water regime and microtopography (Wolter et al. 2016). An increase in the abundance and extent of shrubs in tundra areas has been reported for large parts of the Arctic (Tape et al. 2006; Mekonnen et al. 2021) and is expected for the YCP as well (Wolter et al. 2016). 4.3 Landscape evolution The YCP is primarily composed of Pleistocene sediments, which are overlain by a layer of unconsolidated Holocene deposits that vary in thickness (up to > 15 m) and extend into the Beaufort Sea shelf (Rampton 1982; Fritz et al. 2012). The region’s geology is strongly affected by its location at the interface of the unglaciated Beringian landscape to the west and the coverage by the Laurentide Ice Sheet (LIS) to the east (Rampton 1982; Fritz et al. 2012). During the Late Wisconsinan, between approximately 23,000 and 18,000 years ago, the LIS reached its maximum extent (Fig. 3; Rampton 1982). While areas west of Firth River, such as Komakuk Beach, remained ice-free, the ice sheet’s influence is still visible in landscape features such as morainic ridges, kame terraces, and meltwater channels in the eastern region (Rampton 1982; Fritz et al. 2012). HIQ itself was formed by the LIS, which pushed material from the adjacent Herschel Basin (Mackay 1959; Fritz et al. 2012). This is evident today from the presence of icethrust terrestrial and marine sediments on the island (Fritz et al. 2011; Fritz et al. 2012). As the ice sheet retreated and the Holocene began, rising summer temperatures triggered permafrost thawing, leading to the truncation of late
Material and methods 20 samples. The same approach was used for obtaining samples from mud pools and stabilized zones in Slump D. Seven mud pool samples were collected from four sites in Slump D, including surface samples and samples from 20 cm below surface, along with two surface mud pool samples from the mainland coast. From stabilized zones in Slump D—characterized by very dry and partly vegetated areas—six samples were collected from three sites, again from both the surface and 20 cm depth. Finally, samples from sediments transported with three thaw streams in Slump D were collected by holding plastic bottles directly into the flowing water. Those samples (from mud pools, stabilized zones, and thaw streams) are sometimes referred to as “disturbed active layer” samples. The sampling locations are depicted in Figure 4. Figure 4: Sample locations and types of the YC24 samples obtained during the YC24 expedition in July 2024. The images show typical examples of the respective type, and the sampling strategy. Photos by AWI/Esther Horvath (coring) and Katharina Jaspers (all others).
Material and methods 21 5.1.3 Marine sediment samples All YC24 marine sediment samples were collected along a transect starting at the outflow of Slump D and extending perpendicularly to the coastline. Two samples from 0 and 15 cm depth below surface were taken from the beach at Slump D (i.e., with 0 m distance to shore). From there, samples were taken every 100 m during the first kilometer and every 500 m during the second kilometer of the transect, resulting in 12 additional sampling sites. At seven of these sites, a van Veen grab sampler was used to obtain surface sediment samples. Using a spoon, the marine material was transferred into Whirl-Paks®. At three of these locations, a sample ring was taken from the material in the grabber to determine DBD, assuming the sample remained undisturbed during collection. At the other five locations, an UWITEC gravity corer with an inner diameter of 60 mm was used to gain sediment material. The corer, which sinks into the sediment by its own weight, contained liners that were closed and stored under a dark cover after retrieval. Back on the island, the cores were split into 0-2 cm (surface), 2-7 cm, and the deeper segments. The sample material was then transferred into Whirl-Paks®. In total, 24 marine sediment samples were collected, of which 20 were volumetric. 5.2 Laboratory work 5.2.1 Sample preparation All analyses were conducted at the Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research in Potsdam, Germany (AWI). Only the MeHg measurements were carried out at the University of Stockholm, Sweden (Fig. 5). Back from field work, all 76 frozen samples were freeze-dried with a temperature of 35 °C at the heating plates. This temperature was selected to minimize the risk of MeHg degradation during the drying process. For freeze-drying, the sample bags were opened, covered with a cloth, and placed in the freeze-dryer until they were dry, which took between two to four days per run. Samples were weighed before and after drying to determine DBD for volumetric samples. After drying, the samples were split into three subsamples (Fig. 5). One subsample was grinded for Hg, Total Organic Carbon (TOC), Total Inorganic Carbon (TIC), and Total Nitrogen (TN) analyses using a planetary mill. Another subsample was grinded for MeHg
Material and methods 22 analysis using a coffee grinder to allow for better temperature control. The remaining subsample was left unprocessed for grain size analysis. Unless otherwise stated, all concentrations are reported as per dry weight and data are reported as mean ± standard deviation. Figure 5: Workflow of sample analysis. Samples were taken at the YCP, transported frozen to AWI Potsdam, and analyzed for Hg, TOC/TIC, TN, grain size distribution and MeHg (at University (U) of Stockholm). 5.2.2 Mercury, Carbon, and Nitrogen Hg, TOC, TIC, and TN concentrations were analyzed by weighing in 50 mg of the milled samples. Several standards were analyzed along with the samples for quality control. Duplicate measurements were performed for each sample, and when the
Material and methods 23 relative standard deviation of the duplicates exceeded 5 %, an additional two 50 mg subsamples were analyzed. Hg concentrations were determined using a Milestone Direct Mercury Analysis System (DMA-80 evo, MWS-GmbH). TOC and TIC were measured with an ELEMENTAR soli TOC cube, and TN concentrations were obtained using an ELEMENTAR rapid MAX N exceed. The atomic C/N ratio was calculated using Equation (1): 𝐶/𝑁 𝑟𝑎𝑡𝑖𝑜 =𝑇𝑂𝐶 𝑇𝑁 ×1.167 . (1) 5.2.3 Methylmercury The process of MeHg determination can be divided into the extraction procedure (extracting MeHg from the sediment) and the actual analysis (measuring MeHg concentrations). The extraction and measurements were conducted at the University of Stockholm with a total of three days of extraction and four days of analysis. 5.2.3.1 Extraction About 0.5 g of the grinded sample material was weighed in into new 50 ml Falcon tubes, along with three times 0.05 g of Certified Reference Material (CRM, MeHg concentration according to manufacturer: 75 ± 4 µg kg-1) for each of the four analysis days. Additionally, three empty Falcon tubes (blanks) were measured alongside the samples per analysis day to monitor potential contamination and background levels. An internal standard (IS; isotopically enriched Me200Hg standard) was added to each tube: for samples and blanks, this was 100 µl of IS with a Me200Hg concentration of 1.3 ppb (first day) / 1 ppb (all other days), and for the CRMs, this was 250 µl of IS with a Me200Hg concentration of 13.9 ppb. Milli-Q® water was added, and the tubes were shaken to ensure that all sample particles were in contact with the liquid. Then, the tubes were left to equilibrate for one hour. The chemical extraction was initiated by adding 10 ml KBr (1.4 KBr in 10 % H2SO4), 2 ml CuSO4 (1 M CuSO4 in 250 ml), and 10 ml DCM (Dicholoromethane CH2Cl2). The tubes were left to react for 45 minutes. They were then placed on a sample rotor (Fisherbrand™ Multi-Purpose
Material and methods 24 Tube Rotator) for 45 minutes at 65 rpm, followed by centrifugation at 3000 rpm for 5 minutes. This resulted in a separation of two liquid mixtures: a lower clear layer containing DCM and the extracted MeHg, and an upper greenish layer containing KBr and CuSO4 (Appendix, Fig. 27). The lower layer was transferred into new Falcon tubes using glass Pasteur pipettes. To remove DCM from the extracted MeHg, 10 ml of Milli-Q® water were added to the tubes containing the pipetted liquid, which were placed in a warm water bath at 40 °C and purged from the DCM with nitrogen gas (N2). This process evaporated the DCM, trapping the extracted MeHg in the Milli-Q® water. The tubes were stored in a freezer until analysis, which took place between one and five days later. 5.2.3.2 Analysis MeHg content was measured with a Tekran® Model 2700 Methyl Mercury Analyzer connected to an Inductively Coupled Plasma Mass Spectrometer, thermo scientific iCAP TQ (ICP-MS). For analysis, samples, blanks, and CRMs were acidified by adding 225 µl acetate buffer (4 molar (M) solution; 2 M of sodium acetate and 2 M of glacial acetic acid dissolved in Milli-Q® water, final volume brought to 1 l) and ethylated adding 1 % sodium tetraethyl-borate (NaTeB). Calibration blanks and standards were created according to Table 2. All mixtures were prepared in brown glass vials, which were closed immediately after adding NaTeB to prevent MeHg loss due to evaporation. The samples were gently shaken and placed in the Tekran® analyzer, connected to the ICP-MS, for measurement.
Material and methods 25 Table 2: Components added prior to MeHg analysis. All numbers are given in ml. The working solutions Std10ppt and Std500ppt are MeHg standard diluted 20 times (Std10ppt) and 1000x (Std500ppt) with Milli-Q® water, glacial acetic acid, and HCl. Label Milli-Q® water Acetate buffer Std10ppt Std500ppt IS sample NaTeB Milli-Q® water 30 Calblank (3x) 29.745 0.225 0.03 Std0.02 29.685 0.225 0.06 0.03 Std0.05 29.595 0.225 0.15 0.03 Std0.1 (3x) 29.445 0.225 0.30 0.03 Std0.5 29.715 0.225 0.03 0.03 Std2 29.625 0.225 0.12 0.03 Std4 29.505 0.225 0.24 0.03 Samples & blanks 27.745 0.225 2.0 0.03 CRM (3x) 29.645 0.225 0.1 0.03 Calblank + IS (3x) 29.705 0.225 0.04 0.03 Std2 + IS (3x) 29.585 0.225 0.12 0.04 0.03 5.2.3.3 Processing The results from the Tekran® analyzer were used for external calibration to verify the concentration of the standards. From the data from the ICP-MS, the abundances of the different Hg isotopes in the IS were calculated. By considering the ratios of Me²⁰⁰Hg and Me²⁰²Hg in both the IS and the natural environment, the ICP-MS signals were deconvoluted to distinguish between naturally occurring and spiked mercury isotopes. Using the known volume and concentration of the IS, the mass of Me200Hg per ICP-MS count was determined. This ratio was then applied to calculate the concentration of MeHg per sample. To assess precision, triplicate analyses were performed for four selected samples representing high and low TOC content as well as high and low Hg concentrations. Deviations between the triplicates ranged from 6.2 to 41.3 % (relative standard deviation). The
Material and methods 26 concentrations of the blanks were below 0.001 ng g-1. The CRMs showed a mean concentration of 76.9 ± 3.6 ng g-1. 5.2.4 Grain size Prior to grain size analysis, the sample material was treated to maintain only mineral components by removing all organic material. For this, 100 ml of 3 % H2O2 and 2-3 drops of ammonia (NH3) were added to 7-10 g of freeze-dried sample material. The samples were shaken for 15 minutes each day on a plate shaker, and every one to two days, 10 ml of 30 % H2O2 was added. The pH value was monitored and maintained between 6.5 and 8 by adding either acetic acid (CH3COOH) or NH3 as needed. Once the pH value remained stable for several days and the liquid above the dispersed sediment appeared clear, the excess liquid was removed by centrifuging and decanting. The samples were then frozen and freeze-dried with 50 °C at the heating plates. Approximately 1.6 to 5.2 g of dried material was weighed in into plastic bottles, tetrasodium pyrophosphate was added, and the bottles were filled with Milli-Q® water. They were placed on an overhead shaker for at least 12 h prior to analysis. Grain size analysis was conducted with a Laser particle counter Mastersizer 3000 (MALVERN). To prevent damage on the instrument from large particles, all material > 1 mm was sieved before measurement. To account for particles > 1 mm, an additional sieving was done with a 2 mm sieve, and the weights of the 1-2 mm and the > 2 mm fraction were recorded. Since the laser provides results as volumetric percentages, the weights from the larger fractions were integrated into the data set by assuming a particle density of 2.65 g cm-3. Soil Specific Area (SSA) was given directly from the laser output and was only available for particles < 1 mm. Size distribution, sorting parameters, and median grain sizes were calculated over all grain sizes with the package GRADISTAT (Blott & Pye 2001).
Material and methods 27 5.3 Data analysis 5.3.1 Significance and correlation Whenever data sets were compared in this study, it was tested whether observed differences were significant or not. Significant differences indicate that it is unlikely that the observed differences have occurred by random chance alone. Assessing significance is important to increase the probability that conclusions are based on actual patterns in the data, not on random variance (Berry 1986). Appropriate and commonly used statistical tests were applied based on data characteristics. Normality was assessed with the Shapiro-Wilk test, and variance homogeneity with the F-test. If both conditions were met, a Student’s t-test was applied. When data were not normally distributed, the non-parametric Mann-Whitney U test was used. When data were normally distributed but variances differed, Welch’s t-test (numerical values) or Levene’s test (factor groupings included) was applied. All tests and further calculations were conducted in R (R Core Team 2025). For all statistical tests, functions from the stats package (version 4.3.3; R Core Team 2025) and car package (version 3.1-3; J. Fox & Weisberg 2019) were used. Whether a correlation or difference was significant or not was indicated by the p-value. The significance level was chosen at 0.05, and p-values below 0.001 are reported as p < 0.001. Linear correlations between selected parameters from the YC24 samples were calculated for the terrestrial and marine data set respectively. The parameters included TOC, TIC, TN, Hg, MeHg, sand content, silt content, clay content, depth, DBD, and SSA. Known autocorrelations between some parameters (e.g., TOC and TN, sand and silt content) were acknowledged. Pearson’s correlation coefficient r and the associated significance level p were computed with the function rcorr from the R package Hmisc (version 5.2-2; Harrell 2025). The r-value is a proxy for the strength of a linear correlation between two parameters, with a range from -1 (strong negative linear correlation) to +1 (strong positive linear correlation). An rvalue of 0 indicates no linear relationship between two parameters. For visualization, correlation matrices were created using ggplot (R package ggplot2, version 3.5.0; Wickham 2016).
Material and methods 28 5.3.2 Modeling mercury concentrations 5.3.2.1 Data preparation The terrain units from Couture et al. (2018) used in this study cover the entire Yukon coastline and extend approximately 2 km inland. Each terrain unit is further divided vertically into one up to 48 layers, each assumed to have uniform physical and geochemical characteristics and a consistent Hg concentration. To calculate onshore Hg stocks and fluxes, the aim was to determine a Hg concentration for each terrain unit and layer. In addition to the results from own samples collected during the YC24 expedition, 217 samples from the AWI Potsdam archive were analyzed for their Hg content in the same way as described in section 5.2.2, thereby extending the data set for this study. More Hg concentrations were harvested from other studies. All available Hg concentrations were assigned to their respective terrain unit and layer, depending on sample location and depth. Where multiple data points existed for a single layer, the arithmetic mean from all measurements was used. Samples YC24_T2, YC24_T3, and YC24_T4—taken from mud pools, stabilized zones, and thaw streams within Slump D—were excluded, as they derive from disturbed sites and are not considered representative for their terrain unit. 5.3.2.2 Random forest model From all data collected, Hg concentrations could be assigned to 158 out of 508 layers from 21 out of 44 terrain units (Fig. 6, step 1). For layers lacking measured values, Hg concentrations were estimated using a random forest model. This type of model is an ensemble learning method that combines the outcomes of many decision trees and thereby improves prediction accuracy and controls overfitting (Breiman 2001). Each tree is trained on a random subset of data and the final prediction is made by averaging the results (Breiman 2001). In this case, predictor variables comprised latitude, longitude, mean depth, minimum depth, maximum depth, TOC content, DBD, grain size class, slope score, and land cover (Tab. 3). Those predictor variables were chosen based on the Boruta algorithm using the Boruta function (R package Boruta, version 8.0.0; Kursa & Rudnicki 2010) in R (Fig. 6, step 2). This algorithm evaluates the importance of potential predictor variables by comparing shuffled values of a variable with the original order of the values (Kursa & Rudnicki 2010). In
Material and methods 29 this case, only one potential predictor (whether a layer belonged to the active layer or permafrost) was excluded by the algorithm. Predictor values were obtained from different sources according to Table 3. The order of importance of the predictors was assessed using the FeatureImp function from the package iml (version 0.11.4; Molnar et al. 2018), which measures the importance “as the factor by which the model’s prediction error increases when the feature is shuffled” (Molnar et al. 2018). Error bars represent the variation in the estimated variable importance due to random permutations. Table 3: Random forest model predictors, their source, and how the data were prepared for the model. A more detailed description of the data preparation can be found in the text below. Data Reference Data preparation Latitude, longitude, mean depth, minimum depth, maximum depth, TOC content, DBD Couture et al. 2018 Adding coordinates where missing; including own TOC measurements Grain size class, slope score CanCoast 2.0 data set (Manson et al. 2019) Assignment according to sample location (grain size) and longest coastline section (slope score) Land cover (LANDC), 21 classes Bartsch et al. 2019 Determining main land cover classes and their coverage (in %) Since not all data were following the same separation as the one given by the terrain units, data preparation was necessary. Where no coordinates were given by Couture et al. (2018), a center point of the respective terrain unit was chosen for latitude and longitude values. TOC values from Couture et al. (2018) were replaced where own TOC measurements were available. The grain size class, where the sample points of a terrain unit fell in, was assigned to the respective terrain unit. The assigned slope score of a terrain unit represents the longest part of the respective terrain unit. The determination was done by intersecting the CanCoast slope layer (polyline) with the terrain unit layer (polygon) in QGIS (version 3.42.0). The length of each segment of the lines was calculated, and by using the Group
Results 36 6.1.2 Methylmercury MeHg concentrations were determined for all YC24 samples. Across the terrestrial samples, MeHg concentrations ranged from 0.064 µg kg-1 (YC24_T7_6, cliff at the Yukon coast, 110 cm depth) to 1.24 µg kg-1 (YC24_T2_2, mud pool from Slump D, 20 cm depth), with a mean of 0.29 ± 0.25 µg kg-1 (n = 52). Concentrations generally decreased with depth, with the highest concentrations found in the upper 30 cm (mean = 0.38 ± 0.28 µg kg-1, n = 26), followed by an average concentration of 0.22 ± 0.21 µg kg-1 between 30 and 100 cm (n = 17), and lowest mean concentrations below 100 cm depth (0.14 ± 0.04 µg kg-1, n = 9). Differences between the upper 30 cm and the other depth ranges were significant (p < 0.001). When comparing the three different sampling regions, average MeHg concentrations were similar between Slump D on HIQ (0.28 ± 0.23 µg kg-1, n = 38) and the Yukon mainland coast (0.30 ± 0.297 µg kg-1, n = 14), and lower in marine sediments (0.21 ± 0.08 µg kg-1, n = 24), but those differences were not statistically significant (p > 0.05). Within Slump D, the highest MeHg concentrations were found in mud pools (mean = 0.41 ± 0.32 µg kg-1, n = 9), followed by sediments in thaw streams (0.27 ± 0.04 µg kg-1, n = 3) and stabilized zones (0.27 ± 0.08 µg kg-1, n = 6) (Fig. 9). Differences were not significant. Samples from the undisturbed tundra on top of Slump D and from the headwall of Slump D exhibited significantly lower concentrations (mean = 0.23 ± 0.21 µg kg-1, n = 22) than those found within the RTS, with three times higher mean concentrations in the active layer (0.49 ± 0.43 µg kg-1, n = 4) than in permafrost samples (0.17 ± 0.06 µg kg-1, n = 18) (p = 0.002). For the mainland coast samples, similar depth-related significant trends were observed, with undisturbed active layer samples showing more than three times higher mean MeHg concentrations (0.61 ± 0.36 µg kg-1, n = 4) compared to permafrost (0.18 ± 0.19 µg kg-1, n = 8).
Results 37 Figure 9: Terrestrial sediment MeHg concentrations from different parent material types and locations. Stabilized zones and thaw stream samples originated from Slump D only, while active layer, permafrost, and mud pools samples came from Slump D and the mainland coast. In marine sediments (n = 24), MeHg concentrations were generally lower than in terrestrial samples with a range from 0.08 µg kg-1 (YC24_T5_2, beach at Slump D, 15 cm depth) to 0.38 µg kg-1 (YC24_MC1_4_2-7cm, 800 m distance to shore, 2-7 cm depth) and a mean of 0.21 ± 0.08 µg kg-1. Higher MeHg concentrations were found with increasing distance to shore (Fig. 10), with a mean concentration of 0.12 ± 0.07 µg kg-1 (n = 5) within the first 200 m and an about twofold significantly higher mean concentration of 0.27 ± 0.05 µg kg-1 (n = 11) beyond 600 m offshore (p = 0.002). Concentrations also tended to increase with sediment depth. Subsurface samples (> 2 cm depth) showed a (not significantly, p = 0.15) higher mean concentration (0.23 ± 0.09 µg kg-1, n = 11) than surface samples (0.19 ± 0.07 µg kg-1, n = 13). Figure 10: Marine sediment MeHg concentrations from grab samples and short sediment cores along a 2000 m offshore transect from the southeast coast of HIQ in front of Slump D. The water depth is compressed for improved data visualization. 1100 cm water depth
Results 38 The proportion of MeHg relative to total Hg (%MeHg) ranged from 0.07 % (YC24_TC2_7, permafrost from undisturbed tundra at Slump D, 72 cm depth) to 2.85 % (YC24_T7_1, active layer from Yukon coast, 20 cm depth) for terrestrial samples, with a mean of 0.53 ± 0.59 % (n = 52). Marine sediment samples showed a smaller range from 0.15 % (YC24_T5_2, marine beach, 0 m offshore) to 0.89 % (YC24_MC1_1_2-4cm, 4 m water depth, 200 m offshore) with a mean of 0.38 ± 0.14 %, which did not differ significantly from the terrestrial mean. Differences in %MeHg between active layer and permafrost samples were significant (p = 0.006), with an almost four times higher mean %MeHg in active layers (1.21 ± 1.03 %, n = 8) than in permafrost samples (0.31 ± 0.32 %, n = 26). 6.1.3 Carbon and nitrogen TOC concentrations in terrestrial samples ranged from 0.5 % (Komakuk Beach, 4 m depth) to 49.42 % (Phillips Bay, 55 cm depth), with a mean of 9.23 % across the full data set (all terrain units and layers, n = 586). For the YC24 subset, the TOC range was similar, although the mean was lower at 6.58 % (n = 52). Total Inorganic Carbon (TIC) concentrations were considerably lower and ranged from 0.11 to 1.89 % (n = 35), with 17 samples having TIC concentrations below the detection limit of 0.1 %. Excluding those, the mean TIC concentration was 0.78 ± 0.46 %. TOC contents were double as high in the active layer (mean = 14.15 ± 13.84 %, n = 136) than in permafrost (mean = 7.75 ± 10.08 %, n = 450). This significant trend was also observed, but not significant, within the YC24 subset, where undisturbed active layer samples showed higher TOC values (mean = 11.51 ± 13.09 %, n = 8) than permafrost samples (mean = 8.16 ± 8.01 %, n = 26). In contrast, when including disturbed unfrozen samples from Slump D (i.e., mud pools, stabilized zones, thaw streams), this trend reversed and TOC concentrations in permafrost (8.16 %) exceeded those in the active layer (mean = 5 ± 8.24 %, n = 26), though not significantly (p = 0.071). TN concentrations (in YC24 samples) varied from 0.11 % (YC24_T1_4, 200 cm depth) to 2.72 % (YC24_T8_1, 20 cm depth), with a mean of 0.46 % (n = 52). TN values showed a similar pattern as TOC concentrations: in undisturbed YC24 active layer samples, TN concentrations were higher (mean = 0.82 ± 0.86 %, n = 8) than in permafrost samples (mean = 0.51 ± 0.44 %, n = 26), but including disturbed active layer samples resulted in a lower TN mean in non-permafrost samples (0.4 ± 0.54 %, n = 26)
Results 39 compared to permafrost samples. In both cases, the differences were not significant (p > 0.05). TOC and TN concentrations were generally enriched in surface layers (Fig. 11). Samples from the upper meter had a mean TOC concentration of 12.87 ± 12.7 % (n = 268) and a mean TN concentration of 0.48 ± 0.54 % (n = 43), while samples below one meter averaged 6.17 ± 9.08 % TOC (n = 318; significantly lower) and 0.34 ± 0.21 % TN (n = 9; no significant difference). Spatial differences in TOC and TN distribution were evident. The mainland coast exhibited a more than twofold higher mean TOC concentration (10.96 ± 12.34 %, n = 422) than HIQ (mean = 4.81 ± 6.63 %, n = 164), a trend consistent across both the full data set and the YC24 samples (but significant only for the full data set). Along the mainland coast, TOC concentrations varied regionally: the eastern part (> -138.4°W) had a significantly lower mean TOC concentration (8.35 ± 9.27 %, n = 166) compared to the western part (< -139.5°W; mean = 13.35 ± 13.94 %, n = 111) (Fig. 12). TN concentrations were also higher (but not significantly higher) at the mainland coast (mean = 0.68 ± 0.78 %, n = 14) than at Slump D (mean = 0.37 ± 0.31 %, n = 38). The TN distribution along the YCP could not be assessed due to all YC24 samples originating from the western unit. Figure 11: Vertical distribution of Hg, MeHg, TOC, and TN in terrestrial sediments from active layer, permafrost, mud pools, stabilized zones, and thaw streams.
Results 40 Figure 12: TOC distribution across terrain units and layers of the YCP from west (W) to east (E).
Results 41 The atomic TOC/TN (C/N) ratio in terrestrial sediments ranged from 1.5 (YC24_TC1_2, tundra on top of Slump D, 33 cm depth) to 20.9 (YC24_T1_1, Slump D headwall, 20 cm depth) with one outstanding C/N ratio for sample YC24_TC1_2 (top of permafrost, Slump D headwall, 33 cm depth) with a ratio of 60.1. On average, Slump D exhibited higher (but not significantly higher) C/N ratios (16.6 ± 7.7, n = 38) than the Yukon coast (12 ± 5.7, n = 14). Within Slump D, the C/N ratios across different landscape types (mud pools, stabilized zones, thaw streams) were relatively consistent, with average values between 11.6 and 15.3. This was significantly lower than the C/N ratios of the undisturbed tundra around Slump D (18.5 ± 9.6, n = 22). Overall, terrestrial C/N ratios were significantly lower in surficial samples (0-30 cm; mean = 13.8 ± 3.3, n = 26) than in deeper layers (> 30 cm; mean = 17.3 ± 9.6, n = 26), and in non-permafrost samples (13.8 ± 3.4, n = 26) than in permafrost samples (17.3 ± 9.6, n = 26). Marine sediment samples (here: YC24 samples and data from literature and other publications) exhibited generally low TOC concentrations with a mean of 1.72 ± 1.84 % (n = 87), ranging from 0.24 % (Whale Bay, 0-2 cm sediment depth) to 15.32 % (Nunaluk Spit, 30 m offshore, 0-2 cm sediment depth). Deeper sediments showed significantly higher TOC levels (mean = 1.94 ± 1.05 %, n = 28) compared to surface sediments (mean = 1.61 ± 2.12 %, n = 59) (p = 0.001), but surface sediments demonstrated a considerably broader range (0.24 % to 15.32 %) than deeper samples (0.54 % to 6.8 %). Along the mainland coast, marine TOC concentrations were more than ten times higher in the eastern region east of -138.4°W (90.31 ± 35.2 %, n = 7) compared to the western part west of -139.5°W (5.66 ± 8.36 %, n = 3). The difference was significant (p = 0.017), but the database was very small. No trend in TOC concentration in relation to the distance from shore was found. TN concentrations, available for the marine YC24 samples and data from Petzold (2024) (n = 31), were comparable between surface sediments (mean = 0.09 %, n = 15) and deeper layers (mean = 0.12 %, n = 13). Although distance to shore rarely influenced TN levels, it is noteworthy that the eleven samples with TN values below detection limit (i.e., < 0.1 %) were collected within the first 600 m offshore (Fig. 13). Those eleven samples also exhibited the highest C/N ratios averaging 93.7 (assuming TN = 0.05 %). The four highest and lowest C/N ratios (24.34 to 32.16 and 11.1 to 12.6, respectively) originated from these samples, assuming TN = 0.05 %. Excluding those values, the ratio in marine samples ranged from 13.2 to 19.7. The
Results 42 mean C/N ratio of marine samples (17 ± 2, n = 20; excluding highest and lowest values) was significantly higher (p = 0.008) than the one in terrestrial samples (15.6 ± 7.3, n = 52). Within marine samples, the ratio was slightly but not significantly (p = 0.072) higher in surface sediments (17.8 ± 1.7, n = 11) than below 2 cm sediment depth (16.3 ± 2, n = 12). No trend was observed with distance to shore (Fig. 13). Figure 13: Distribution of Hg, TOC, TN, C/N ratio, MeHg, and SSA in marine sediments along the transect in front of Slump D (n = 24). Connected dots represent surface sediments (02 cm depth), while transparent dots show values of samples of greater depth (up to 37 cm). 6.1.4 Grain size distribution Grain size distribution was available for all terrestrial and marine YC24 samples. Terrestrial sediment samples predominantly consisted of medium to coarse silt (Fig. 14), with median grain sizes ranging from 5.97 µm (YC24_T1_6, Slump D headwall, 100 cm depth) to 59.07 µm (YC24_T7_1, mainland coast bluff, 20 cm depth).
Results 43 Across landscape types, grain size distributions tended to be consistent within their sampling location, even when samples were taken at different depths (Appendix, Fig. 31). Notable exceptions included site YC24_T7 (Yukon coast bluff), where samples from 20 cm (YC24_T7_1) and 65 cm depth (YC24_T7_4) exhibited a pronounced increase in volume of grains around 63 µm, in contrast to other YC24_T7 samples. At site YC24_T1 (Slump D headwall) grain size profiles varied slightly with depth: higher sand and gravel fractions were observed in the active layer (mean sand + gravel proportion = 20.6 %, n = 2), at 50 cm depth (34.2 %), and at 200 cm depth (45.4 %), while the intermediate section (between 50 and 200 cm) showed a finer composition (mean sand + gravel proportion = 10.9 %, n = 2). Specific surface area (SSA) was used as a proxy for grain size distribution. SSA values were similar between permafrost and non-permafrost samples, indicating no distinct textural shift across the thermal boundary. Spatial differences, however, were evident. Samples from Slump D contained significantly finer material than those from the Yukon coast. This was reflected in significantly (p < 0.001) higher SSA values at Slump D (mean = 1543.7 ± 228.1 m2 kg-1, n = 38) compared to the Yukon coast (mean = 914.2 ± 231 m2 kg-1, n = 14). Within the slump, mud pools and stabilized zones showed similar grain size characteristics and SSA values, whereas thaw streams contained slightly and significantly finer sediments, with SSA values averaging at 1819 ± 109 m2 kg-1 (n = 3), compared to approximately 1588 ± 154 m2 kg-1 (n = 13) on the slump floor (mud pools and stabilized zones). Grain size distribution in marine sediment samples exhibited a wide range (Fig. 14), with median grain size spanning from 5.23 µm (YC24_MC1_2, 8 cm sediment depth, 6 m water depth, 400 m offshore) to 561.53 µm (YC24_T5_2, marine beach, 15 cm sediment depth). Coarse fractions (sand and gravel) were predominantly found near the beach and in the nearshore area, while finer sediments became increasingly dominant with greater distance from the coast and increasing water depth. Samples collected beyond 600 m offshore had a significantly (p = 0.038) lower mean median grain size of 8.98 ± 2.21 µm (n = 11), in contrast to those within the first 200 m, which averaged 240.15 ± 216.14 µm (n = 5). Trends with sediment depth were apparent but not significant (p = 0.084): surface sediments showed higher median grain sizes (mean = 107.18 ± 163.18 µm, n = 14) compared to subsurface layers below 2 cm depth (mean = 23.16 ± 39.82 µm, n = 10).
Results 44 Figure 14: Grain size distribution of YC24 samples of different types from Slump D, the mainland coast, and marine sediment samples. 6.2 Correlation matrices Correlation patterns were investigated for the YC24 samples (n = 76), where most data were available for. It is acknowledged that there are autocorrelations between several parameters, such as those for grain sizes (sand, silt, clay, SSA) or between TOC and TN (both forming organic material). 6.2.1 Terrestrial samples In terrestrial samples (YC24 samples, n = 52), very few strong and significant (p < 0.05) linear correlations were found (Fig. 15). Strongest correlations occurred between autocorrelated parameters, namely SSA & clay (r = 0.99), TOC & TN (r = 0.98), and sand & silt (r = -0.85). Other than that, DBD correlated significantly with TN (r = -0.75) and consequently with TOC (r = -0.74), as well as with Hg concentrations (r = -0.51). Apart from this, Hg only showed significant correlations with the autocorrelated parameters clay (r = 0.54) and SSA (r = 0.55), and with TIC (r = -0.75). MeHg concentrations did not correlate significantly and strongly with any of the analyzed parameters.
Results 45 Figure 15: Correlation matrix of field and geochemical parameters in terrestrial sediment samples (n = 52). The number in each box gives the respective linear correlation coefficient r. Grey values indicate that the correlation was not significant (p > 0.05). 6.2.2 Marine samples In strong contrast to results from terrestrial samples, many parameters in marine sediments (n = 24) were strongly and statistically significantly correlated, with most correlations being positive (Fig. 16). Sand content and TIC were notable exceptions, showing mostly negative correlations. Hg in marine samples showed particularly strong positive correlations with clay (r = 0.95), SSA (r = 0.95), and TN (r = 0.89). Hg and MeHg were positively correlated with one another (r = 0.69), and MeHg generally correlated significantly with most parameters, which had not been the case in terrestrial samples. Depth did not show any significant correlations with other parameters, and correlations with DBD were generally weaker and mostly negative.
Results 52 and Hg stock was observed, reflecting the increase in soil volume with increasing bluff elevation (r = 0.93, p < 0.001). Figure 22: Hg stocks [g m-2] of the Yukon coast. Hg stocks were compared between study sections using uniform volumes of one cubic meter to account for different bluff heights of the different terrain units (Fig. 23). Stocks ranged from 0.02 g m-3 (Workboat Passage E) to 0.15 g m-3 (Simpson Point). From these calculations, stocks were highest at section 2, and lowest in sections 1 and 4, which were less affected by the Last Glacial Maximum than other sections. The variability of Hg stocks across sections was considerable. Whether differences were significant was tested by performing multiple pairwisecomparisons between the means of the groups, using the Tukey ‘Honest Significant Difference’ method (function TukeyHSD, package stats, version 4.3.3, R Core Team 2025) as normality and homogeneity of variances were given (confirmed with Shapiro-Wilk and Levene tests). Only section 2 differed significantly from sections 1, 3, 4, 5, and 8. Large variability was obvious within sections, indicating substantial heterogeneity that was not covered by the division of sections. Basemap: ESRI Satellite CRS: EPSG 32607
Results 53 Figure 23: Hg stocks [g m-3] across the eight sections of the Yukon coast. For more details about the sections, see Table 1. To place these results in a broader context, total Hg stocks were also estimated using alternative methods. Applying a single Hg:TOC ratio, as done for instance by Schuster et al. (2018) resulted in a total stock estimate of 147,390 kg Hg— approximately 2.5 times lower than the stock estimated from the random forest model. Splitting the training data into three subsets based on TOC content gave a higher estimate of 232,150 kg (1.5 times lower than the random forest estimate). 6.3.4 Fluxes The mean coastal erosion rate along the YCP was 0.71 m yr-1, varying considerably across the coast (Tab. 5). Maximum erosion rates exceeding 1 m yr-1 were observed west of HIQ, at the terrain unit Herschel Island North (1.2 m yr-1), and in parts of the eastern coast, including Stokes Point (3.58 m yr-1), Kay Point (2.3 m yr-1), and King Point SE (1.1 m yr-1). Some coastal segments were stable (Malcolm River fan with barrier island and Kay Point Spit) or showed accretion (Shingle Point E, Stokes Point SE) (Tab. 5). Hg fluxes per terrain unit resulting from coastal erosion ranged from -4.7 kg yr-1 to 43.2 kg yr-1 (Fig. 24), with negative values indicating accumulation. The mean Hg flux amounted to 2.59 kg yr-1. Normalized fluxes per meter of coastline ranged
Results 54 from -0.39 g m-1 yr-1 to 2.57 g m-1 yr-1, with a mean of 0.38 g m-1 yr-1. In both cases, the highest fluxes were recorded at Herschel Island N and Herschel Island W, while the highest accumulation took place at Shingle Point E. Overall, Hg release was lower south and southwest of HIQ, and higher in the eastern parts of the study area and at HIQ itself (Fig. 24). The total Hg flux for the entire YCP was estimated at 113 kg yr-1 (87 to 163 kg yr-1) considering accumulation (negative) and equilibrium (neutral) at two terrain units, respectively. Figure 24: Hg fluxes [g m-1 yr-1] from the Yukon coast. To compare Hg fluxes between sections, the mass of Hg per cubic meter of eroding material was calculated by considering the bluff height and annual change rate. Figure 25 illustrates the distribution of these fluxes across sections. Values ranged from -0.14 (Shingle Point E) to 0.22 g m-3 yr-1 (Simpson Point). The highest Hg release per cubic meter was observed for section 2, where the mean value was significantly higher than in sections 1 and 4. No significant differences were found between the other sections, as determined by a one-way ANOVA test, following the procedure described in chapter 6.3.2. Although the sections were defined based on similar Basemap: ESRI Satellite CRS: EPSG 32607
Results 55 physical characteristics, the observed variability in Hg flux per cubic meter suggests that these groupings do not reflect uniform behavior with respect to Hg release. Figure 25: Hg fluxes [g m-3 yr-1] across the eight sections of the Yukon coast. For more details about the sections, see Table 1. To assess the broader implications for total Hg flux estimates, two simplified approaches were tested. Applying a single Hg:TOC ratio, as done for instance by Schuster et al. (2018), resulted in a total flux estimate of 56.5 kg Hg yr-1–half the estimate from the random forest model. Subdividing the training data into three subsets gave a higher total Hg flux of 83.7 kg yr-1, which remains 1.4 times lower than the random forest model-based estimate.
Results 56 Table 5: Section allocation, characteristics, stocks, and fluxes per terrain unit. Change rate and bluff height from Couture et al. (2018). Terrain unit Section Change rate [m yr-1] Bluff height [m] Hg stock [g m-2] Hg stock [g m-3] Hg flux [g m-1 yr-1] Hg flux [g m-3 yr-1] Clarence Lagoon W 1 -1.6 7.1 0.168 0.024 0.260 0.023 Clarence Lagoon 1 -0.8 1.0 0.065 0.065 0.056 0.070 Clarence Lagoon E 1 -0.9 2.2 0.086 0.039 0.080 0.040 Komakuk Beach W2 1 -1.1 6.7 0.213 0.032 0.242 0.033 Komakuk Beach W1 1 -1.6 6.8 0.137 0.020 0.214 0.020 Komakuk Beach 1 -1.3 6.2 0.142 0.023 0.191 0.024 Malcolm River fan 1 -0.8 1.9 0.073 0.038 0.059 0.039 Malcolm River fan with barrier island 1 0 1.0 0.091 0.091 0 NA Nunaluk Spit 2 -1.2 1.0 0.093 0.093 0.095 0.079 Avadlek Spit 2 -0.7 1.2 0.143 0.119 0.135 0.161 Herschel Island W 3 -0.9 27.7 1.932 0.070 2.079 0.083 Herschel Island N 3 -1.2 36.0 2.033 0.056 2.587 0.060 Simpson Point 2 -0.5 1.0 0.147 0.147 0.108 0.216 Herschel Island E 3 -0.4 21.8 1.201 0.055 0.490 0.056 Herschel Island S 3 -0.3 11.0 0.879 0.080 0.333 0.101 Workboat Passage W 4 -0.5 6.4 0.193 0.030 0.091 0.028 Workboat Passage E 4 -0.4 2.3 0.046 0.020 0.019 0.021 Catton Point 2 -0.3 1.4 0.161 0.115 0.073 0.174 Whale Bay W 4 -1.0 1.7 0.040 0.024 0.039 0.023 Whale Bay 2 -0.5 1.0 0.119 0.119 0.062 0.124
Results 57 Whale Bay E 4 -0.2 8.2 0.385 0.047 0.071 0.043 Roland Bay NW 4 -0.3 5.9 0.149 0.025 0.049 0.028 Roland Bay W 5 -0.4 8.4 0.378 0.045 0.169 0.050 Roland Bay E 5 -0.7 8.4 0.702 0.084 0.450 0.077 Stokes Point W 5 -0.9 12.5 0.874 0.070 0.784 0.070 Stokes Point 2 -3.6 1.9 0.223 0.117 0.779 0.114 Stokes Point SE 5 0.5 8.3 0.578 0.070 -0.060 -0.014 Phillips Bay NW 5 -0.4 13.8 1.037 0.075 0.407 0.074 Phillips Bay W 2 -0.5 1.0 0.122 0.122 0.083 0.166 Phillips Bay 5 -0.6 4.7 0.268 0.057 0.256 0.091 Babbage River Delta 6 -1.0 3.2 0.177 0.055 0.202 0.063 Kay Point Spit 2 0 1.0 0.091 0.091 0 NA Kay Point 7 -2.3 8.5 0.501 0.059 1.122 0.057 Kay Point SE 3 -0.2 26.1 2.093 0.080 0.360 0.069 King Point NW 8 -0.2 32.7 2.601 0.080 0.616 0.094 King Point Lagoon 2 -0.3 1.0 0.121 0.121 0.044 0.147 King Point 8 -0.6 7.4 0.216 0.029 0.225 0.051 King Point SE 8 -1.1 10.2 0.695 0.068 0.749 0.067 Sabine Point W 8 -0.7 20.6 0.789 0.038 0.578 0.040 Sabine Point 8 -0.8 20.6 1.064 0.052 0.867 0.053 Sabine Point E 8 -0.4 19.0 0.494 0.026 0.185 0.024 Shingle Point W 8 -0.2 21.1 1.691 0.080 0.299 0.071 Shingle Point E 8 0.2 13.7 1.515 0.111 -0.391 -0.143 Running River 8 -0.7 22.6 1.944 0.086 1.405 0.089
Discussion 58 7 Discussion 7.1 Assessment of outcomes and uncertainties 7.1.1 Model behavior and performance Spatial parameters were the most influential predictors for modeling Hg distribution along the Yukon coast. The model included commonly used predictors such as TOC, depth below surface, land cover, grain size, and slope (Grigal 2002; Olson et al. 2018; Bishop et al. 2020; Rutkowski et al. 2021). In this study’s data set, TOC and depth below surface had the best vertical resolution with individual values per layer. Grain size, widely recognized as an important predictor of Hg due to its influence on sorption capacity for heavy metals (Salonen & Korkka-Niemi 2007), was only available as a general surficial classification from the CanCoast data set without vertical resolution, likely limiting its explanatory power. Slope was selected as a relevant parameter by the Boruta algorithm, but its reliability is limited by the dynamic morphology of the Yukon coast under sea-ice influence and ongoing coastal erosion (Irrgang et al. 2022). Substantial geomorphological changes occur due to bluff erosion and lateral sediment transport along the shore, that affect slope characteristics and thereby limit its reliability as a static predictor. Among all available predictors, spatial variables (depth and longitude) showed the highest predictive importance, indicated by RMSE increases when omitting the respective predictor (increases of 5.96, 5.3, 3.13, and 2.93 µg kg-1 for mean depth, maximum depth, minimum depth, and longitude, respectively; Fig. 18). The importance of spatial data is due to both spatial autocorrelation of near things, that “are more related than distant things” (Tobler’s First Law of Geography), and the relatively higher precision of these spatial data compared to the other environmental predictors. The strong influence of spatial variables likely compensates for the limited availability and resolution of other key parameters, stressing the need for high-resolution environmental data in future models. The model predicted a strikingly uniform pattern of Hg concentrations at greater depths, which likely reflects the lack of depth-resolved environmental input data. Low variability was observed in model outcomes for deeper layers (approximately below 1 m depth; Fig. 20). This uniformity could reflect real-world patterns in stable
Discussion 59 permafrost zones but seems unusually homogeneous (Olson et al. 2018; Rutkowski et al. 2021) and is likely an artifact of missing vertical detail in predictors. Variables representing the hydrological regime and grain size, which control vertical matter redistribution and sorption processes (Grigal 2002; Chakraborty et al. 2014), were missing or only available for the surface and thus could not capture the subsurface heterogeneity of YCP soils (Rampton 1982; Fritz et al. 2012). This limitation is especially relevant, as hydrological and pedogenic processes vary with depth and can strongly affect Hg mobility and accumulation. Depth-specific environmental data and predictors are essential for a realistic estimation of material concentrations across the entire vertical soil profile, especially where bluffs are high. Model predictions revealed several unexpected patterns that illustrate the limitations of this model approach when training data are scarce. For example, sandy marine beaches and spits with low cliff heights (~ 1 m) and low TOC content (~ 2 %), belonging to section 2, were predicted to have the highest Hg stocks across all sections (Fig. 23). This finding is counterintuitive, as such landforms and their characteristics do not provide typical conditions for high Hg accumulation (Rutkowski et al. 2021; Du et al. 2024). The relatively elevated Hg stocks of section 2 likely reflect the disproportionate influence of the only available training data point for this section with an above-average measured Hg value of 90.87 µg kg-1, along with an overall scarcity of low-Hg concentration training samples in the data set (29 out of 158 data points were below 50 µg kg-1). The highest measured Hg concentration of 271.8 µg kg-1 was consistently assigned to surficial layers characterized by a high TOC content of 25 %, a shallow depth (≤ 10 cm), and a DBD of 310 kg m-3. These predictions were accompanied by wide confidence intervals (Fig. 19), indicating considerable uncertainty. This was probably driven by the large variability within the 11 measured surface samples, where Hg concentrations ranged from 37.97 to 271.8 µg kg-1. Outlier influence and surface variability are two factors driving model uncertainty, which must be addressed by creating a more balanced training data set across the full Hg spectrum and all terrain units to reduce uncertainties. Key spatial patterns are captured by the random forest algorithm, making it a suitable modeling approach for Hg predictions at the Yukon coast. In this study, the model explained 42.2 % of the observed variance in measured Hg concentrations, outperforming similar models for Soil Organic Carbon (SOC) and nitrogen in the
Discussion 60 same region, which reported R2 values between 10 and 35 % (Wagner et al. 2025). Additionally, the RMSE of 22.67 µg kg-1 was lower than the standard deviation of measured concentrations (29.91 µg kg-1), indicating that the model provided better predictions than a simple mean-based calculation (James et al. 2013). The random forest algorithm has previously been used in similar contexts. Wagner et al. (2025) applied the model to estimate SOC and nitrogen stocks of the YCP, using predictors such as elevation, land cover, vegetation height, and soil wetness. Suleymanov et al. (2023) found random forest to provide the most accurate Hg predictions in urban and peri-urban soils in Ufa, Russia. The modeling approach of this study followed the framework proposed by Vaysse & Lagacherie (2017), which has since been adopted by Xie et al. (2024) and Wagner et al. (2025). These applications collectively support the validity of using random forest for spatial Hg modeling, including datasparse and dynamic regions such as the permafrost-affected Yukon coast. The weak correlations between Hg and TOC in this study made the use of Hg:TOC ratios an unreliable method for predicting Hg concentrations from TOC content along the Yukon coast. Across the terrestrial data set with measured Hg concentrations, the correlation between Hg and TOC was not statistically significant (r = 0.14, p = 0.085), indicating that TOC alone cannot explain the observed variability in Hg content. When applied for upscaling, the total Hg stock was substantially underestimated, amounting to only about one quarter of the estimate derived from the random forest model. While many other studies have found more or less strong linear relationships between Hg and TOC and used this for estimating Hg contents (e.g., Olson et al. 2018; Schuster et al. 2018; Lim et al. 2020), others report similarly weak correlations as found in this study (Xu et al. 2014; Du et al. 2024; Giest et al. 2025), suggesting that the TOC-Hg relationship is not universally robust. Dividing the data into subsets based on TOC content, a method found in previous studies to tweak the correlations (Schuster et al. 2018; Tarbier et al. 2021), did not improve the correlations here either (Tab. 4). The limited explanatory power of TOC may reflect site-specific differences in sediment properties and biogeochemical conditions. Differences in OM composition and origin influence its Hg binding capacity and could explain the differences in correlations between different studies (Douglas et al. 2012; Bishop et al. 2020). For example, Hg tends to bind preferentially to thiol and other sulfur-containing functional groups within OM, which vary in abundance across different organic
Discussion 61 sources (Skyllberg et al. 2006). Sediment texture is another factor influencing Hg retention as it determines SSA and mineral sorption capacity, thus influencing heavy metal accumulation and considerably controlling Hg distribution (Brigatti et al. 2005; Salonen & Korkka-Niemi 2007). Terrestrial sediments were mainly composed of medium to coarse silt, consistent with previous findings (Fritz et al. 2012), though high heterogeneity and wide ranges in grain size distribution have been observed, particularly on HIQ (Fig. 14; Rampton 1982; Fritz et al. 2012). The positive correlation between Hg and clay content in the YC24 data set (r = 0.54, p = 3.5 × 10-5) of this study supports the role of fine particles in Hg retention, while clay content does not fully explain spatial variation in Hg distribution either. The association between Hg and TOC or SSA is clearly context-dependent and should not be generalized without considering site-specific environmental and geochemical factors. 7.1.2 Evaluation and contextualization of mercury stocks and fluxes Estimated total Hg stocks and fluxes from the Yukon coast fall within the range of existing regional and pan-Arctic assessments, though substantial uncertainties remain due to high uncertainties within model predictions. This study estimated total Hg stocks at the Yukon coast of approximately 381,080 kg (95 % prediction interval = 271,540 to 501,930 kg), with a total flux of 113 kg yr-1 (87 to 163 kg yr-1). Of this annual flux, 65 kg is estimated to originate from HIQ, while only 48 kg comes from the mainland coast. Compared to the only other regional assessment by Leitch (2006), who estimated Hg fluxes of 85 kg yr-1 from HIQ and 165 kg yr-1 from the mainland coast, fluxes from this study are lower, despite higher measured Hg concentrations in the present study. The difference occurs primarily due to the use of updated, terrain unit-specific erosion masses from Couture et al. (2018) in this study (total of 0.84 × 109 kg yr-1), which is about one third compared to the amounts used by Leitch (total of 2.47 × 109 kg yr-1). At the pan-Arctic scale, Outridge et al. (2008) estimated a total Hg input from coastal erosion of 47.3 t yr-1. The Yukon coast’s contribution of 113 kg corresponds to 0.24 % of this total estimate. In terms of sediment, the total sediment flux from coastal erosion into the Arctic Ocean has been estimated at 430 Tg yr-1 (Stein & MacDonald 2004; Wegner et al. 2015), of which 0.42 % originate from the Yukon coast (assuming a total sediment flux of 1.8 Tg;
Discussion 68 especially in its methylated form—is more relevant for phytoplankton and marine biota than Hg buried in sediments (Jonsson et al. 2022). Despite the proximity to the Mackenzie Delta, eastern YCP sediments did not show elevated Hg levels, indicating that either sediment deposition from the plume is limited or that the material is rapidly transported into other areas. In addition to the lateral transport, ocean-atmosphere exchange continuously modifies Hg concentrations in surface waters, with a net flux from sea to air and sea-ice cover limiting this process seasonally (Dastoor et al. 2022). The multiple transformation and transportation pathways of Hg in the water column are complex and can vary locally (Braune et al. 2015). This complexity complicates source attribution and raises concern that Hg and especially MeHg could increasingly enter Arctic food webs. The spatial distribution of marine sediment Hg concentrations showed a different pattern than the terrestrial sediment Hg concentrations, suggesting that lateral transport and regional coastal processes shape Hg distribution along the Beaufort coast. No linear correlation was found between terrestrial Hg fluxes and Hg concentrations in adjacent marine nearshore sediments from the same terrain unit (r = -0.19, p = 0.54), indicating that high terrestrial input does not directly translate into elevated marine Hg concentrations within the first 500 m offshore which is supported by Hg measurement of Petzold (2024). The spatial pattern of elevated terrestrial Hg concentrations on HIQ and at the eastern mainland coast, however, was reflected in adjacent marine sediments. This points to the influence of winds, coastal currents, and sediment resuspension (Gimsa et al. 2024). The Beaufort Undercurrent, flowing from west to east along the mainland shore, could transport coastal material towards the Mackenzie Delta, while freshwater inputs from smaller rivers of the YCP influence surficial currents and sediment transport locally. Transformations, resuspension and particle association in the water column further influence the regional fate of Hg. As a result, marine Hg inventories cannot be inferred directly from terrestrial source data without taking nearshore dynamics into account.
Discussion 69 7.4 Methylmercury—transformations, pathways, potential risk 7.4.1 Spatial patterns and environmental controls on methylmercury MeHg concentrations in sediments of the YCP were low and within typical ranges for Arctic terrestrial and nearshore environments. Terrestrial MeHg concentrations ranged from 0.064 µg kg-1 to 1.24 µg kg-1, with a mean of 0.29 ± 0.25 µg kg-1. This is comparable to values reported for other Arctic soils (Jonsson et al. 2022), and substantially lower than peak levels of 6.7 µg kg-1 found in Arctic wetland soils (Varty et al. 2021). Marine sediment MeHg levels were even lower, ranging from 0.08 µg kg-1 to 0.38 µg kg-1, consistent with measurement by Fox et al. (2014), who reported a mean value of 0.37 ± 0.36 µg kg-1 in nearshore Beaufort Sea sediments. In both marine and terrestrial samples, %MeHg was low, ranging from 0.07 to 2.85 %. Such low percentages are typical in sediments, but are in strong contrast to marine biota, where MeHg can constitute up to 99 % of total Hg (Loseto et al. 2008). No critical thresholds for MeHg concentrations in soils exist, but the Canadian Council of Ministers of the Environment (CCME) gave a threshold of 6,600 µg kg-1 for total Hg in agricultural and residential soils (CCME 1999). If about one percent of this (assuming %MeHg = 1 % = 66 µg kg-1) would be a critical threshold for MeHg, sediment concentrations of the Yukon coast would clearly stay below this threshold. Higher Hg delivery through atmospheric deposition, permafrost thawing, and coastal erosion could increase MeHg stocks in soils of the Yukon coast, but it seems unlikely that this would result in concerningly high concentrations in soils. MeHg concentrations increased from permafrost to active layer to mud pools, and declined slightly in thaw stream sediments, reflecting the progression from frozen, low-activity soils to water-saturated zones. Intact permafrost showed low MeHg concentrations (0.17 ± 0.11 µg kg-1), consistent with limited microbial activity under frozen conditions (Rivkina et al. 2004). Higher MeHg levels were measured in active layer samples (0.55 ± 0.37 µg kg-1, n = 8), where seasonal thaw allows for microbial activity. The underlying permafrost acts as a barrier for vertical water transport, restricting drainage and potentially creating favorable conditions for anaerobic microbes known to mediate Hg methylation (Bravo & Cosio 2020). Mud pools, located at the base of RTS headwalls and cliffs, probably provided the most anoxic environments due to persistent waterlogging caused by ice melting and limited drainage (Blodau 2002). MeHg concentrations were elevated here, with a mean of
Discussion 70 0.41 ± 0.32 µg kg-1 (n = 9), which could reflected anaerobic conditions that drive microbial Hg methylation (Jonsson et al. 2022; Cardinal et al. 2025). Sediments in thaw streams that transport material from mud pools into the ocean showed lower MeHg levels (0.27 ± 0.04 µg kg-1, n = 3), suggesting that oxic conditions in the streams could promote demethylation (Ullrich et al. 2001). These patterns are derived from 46 samples which may not fully represent their respective landforms. It is unclear how stable MeHg concentrations in these environments are and how quickly methylation and demethylation processes alter MeHg sediment concentrations (Baptista-Salazar et al. 2022; Jonsson et al. 2022; Fabre et al. 2024). A significant part of MeHg can be stored in refractory pools, meaning it is not readily available for demethylation. This has been observed for sediments from lakes and brackish seawater sites, as well as soils from forests and marsh environments (Baptista-Salazar et al. 2022). When this happens in mud pools or wetlands (known for promoting MeHg formation), MeHg could be protected from demethylation during transport into other environments (Baptista-Salazar et al. 2022). It remains unknown how large the refractory pool in our samples is, so that the measurements of this study might represent a temporal snapshot rather than stable background levels. Factors such as temperature, OM availability and quality, and discharge variability may influence short-term MeHg dynamics outside of refractory pools (Schartup et al. 2015; Z. Yang et al. 2016; Seelen et al. 2023). Hydrological regime and redox conditions seem to influence MeHg distribution in bluffs and RTSs, but the interplay of production, transformation, and transport processes is complex, and measured concentrations likely reflect transient conditions rather than steadystate behavior. 7.4.2 Nearshore methylmercury dynamics MeHg concentrations in marine sediments increased with distance from shore and were strongly related to TOC and SSA distribution. Nearshore sediments within the first 500 m from shore were mostly sandy, and silt content increased with distance from shore, similar to findings from Gimsa et al. (2024). This distribution reflects wave-induced winnowing in the very nearshore zone, with finer particles with higher SSA staying in suspension over longer distances. TOC contents of marine sediments were generally low (< 3 %) but increased with distance from shore (r =
Discussion 71 0.7, p < 0.001). This TOC distribution is consistent with the resuspension and deposition zones classified by Jong et al. (2020), who found that terrestrial carbon is mostly deposited in the deposition zone beyond the first few hundred meters offshore. Marine sediment MeHg content rose from 0.08 µg kg-1 (beach) to a maximum of 0.38 µg kg-1 (800 m offshore) and was strongly correlated with TOC (r = 0.89, p < 0.001) and SSA (r = 0.74, p < 0.001). Two mechanisms could drive this offshore increase and explain the strong correlations: (1) Particle-bound MeHg produced in terrestrial sediments is carried offshore with fine, organic-rich particles and settles before extensive water-column transformation takes place. (2) High SSA and TOC content provide substrates and microenvironments conducive to microbial Hg methylation at the sea floor. Generally, the MeHg production in marine sediments can be reduced compared to production in terrestrial sediments (Lehnherr 2014). There are several possible pathways and processes that influence MeHg content in the nearshore zone: coastal erosion releases mostly Hg(II) which can be methylated in reducing zones (Hollweg et al. 2010; Lehnherr et al. 2011; Dastoor et al. 2022). Storm-induced wave attacks cause rapid erosion events and introduce episodic, high-intensity sediment inputs to the ocean along the Alaskan Beaufort Sea coast (Barnhart et al. 2014). These pulse events can temporarily raise Hg concentrations in coastal waters by a factor of 10 to 30 (found at cliffs of the Polish Baltic shore; Bełdowska et al. 2016), potentially increasing MeHg production and uptake by primary producers in biologically active zones. Given the complexity of sediment transport and redox dynamics, it remains challenging to distinguish the contributions of terrestrial inputs from local MeHg production in the Arctic Ocean. Marine sediments exhibited a strong positive correlation between Hg and MeHg concentrations, in contrast to weak and variable relationships in terrestrial sediments, suggesting more consistent or simplified controls on MeHg content in marine settings than on land. In marine sediments, Hg and MeHg concentrations were significantly correlated (r = 0.69, p < 0.001), showing that areas with higher Hg levels tend to support higher MeHg concentrations. This pattern could reflect a relatively uniform depositional environment with stable redox conditions and fine and organic-rich particles that simplify controls on MeHg production (Lehnherr et
Discussion 72 al. 2012; Cossa et al. 2014). In contrast, terrestrial sediments showed no significant Hg-MeHg correlation (r = -0.14, p = 0.33) and lacked significant correlations with TOC (p = -0.04, r = 0.77) and SSA (r = 0.08, p = 0.58). The weak correlations reflect the heterogenous mosaic of soil moisture regimes, OM quality, redox conditions, and microbial communities on land (Lehnherr et al. 2012; Tarbier et al. 2021; Cardinal et al. 2025), as terrestrial samples originated from different landforms (active layer, permafrost, RTS floor). Though direct comparisons are missing, similar differences in correlation patterns under marine and terrestrial conditions can be derived from other studies. Liu et al. (2015) found positive correlations of MeHg with TOC and TN for marine sediments from Kongsfjorden, Svalbard, while correlations of MeHg with TOC and Hg in freshwater lake sediments in Ny-Ålesund were not found (Gopikrishna et al. 2020). Tighter Hg-MeHg coupling in marine sediments implies that marine MeHg concentrations may be more readily predicted from Hg inputs alone, whereas identifying terrestrial MeHg hotspots will require high-resolution environmental data to capture multiple interacting controls. 7.4.3 Potential risks and implications for Arctic food webs Current MeHg concentrations in terrestrial and marine sediments do not indicate a direct risk to the local population or wildlife, but potential impacts through bioaccumulation in the food web cannot be excluded. Measured MeHg levels were not elevated compared to other Arctic regions which suggests no exceptional processes are occurring in soils at the Yukon coast. However, low sedimentary values are not indicative of ecological or human health risks and offer limited insight into actual MeHg dynamics (Ullrich et al. 2001; Fox et al. 2014). Arctic vegetation, particularly mosses and lichens, commonly bioaccumulate MeHg to levels two orders of magnitude higher than in underlying soils (e.g., median MeHg concentration of 0.18 µg kg-1 in soils, and 66.8 µg kg-1 in lichens; St. Pierre et al. 2015), providing a pathway into terrestrial food webs. Caribou, which consume large amounts of lichen, can ingest significant MeHg loads, even though their kidney tissue has not shown elevated levels with Hg concentrations up to 6.2 µg g-1, of which 75 to 90 % is MeHg (Gamberg et al. 2015). The marine food web is more prone to bioaccumulation as the water column has been identified as an effective place for methylation, accounting for 47 ± 62 % of MeHg formation from inorganic Hg(II)
Discussion 73 in polar marine waters (Lehnherr et al. 2011). Additionally, the number of steps in the marine food web is higher than on land, providing more levels of Hg biomagnification (Cohen 1994; Gamberg et al. 2015). St. Pierre et al. (2018) reported a strong increase in MeHg water concentrations of two orders of magnitude from discharge of an RTS. Together with benthic-pelagic coupling (where sedimentderived MeHg is remobilized into overlying waters through resuspension of sediment-bound MeHg), water-column methylation and discharged RTS water supply MeHg to phytoplankton (Kirk et al. 2012). Beginning in phytoplankton, MeHg bioaccumulates and biomagnifies in zooplankton, fish, and marine mammals (Watras et al. 1998; Campbell et al. 2005), while the latter are key subsistence species for Indigenous Peoples (AMAP 2021). Neither water-column MeHg nor MeHg levels in marine biota were measured in this study, pointing to an integrated watersediment-biota analysis in a next step, including seasonal water-column profiles and fish tissue sampling. With rising Hg inputs from increasing coastal erosion, long-term monitoring programs for MeHg levels in water, phytoplankton, zooplankton, benthos, and fish tissues are needed to anticipate and mitigate future risks to Arctic food security. 7.5 Mercury in the One Health framework With ongoing climate warming, it becomes increasingly important to address the interconnectedness of environmental, animal, and human health in a One Health framework (Fig. 26). Hg in a changing Arctic is a powerful example of how these systems interact: human-induced climate warming and increased Hg emissions from coal combustion and gold mining lead to enriched Hg stocks in Arctic soils, which are increasingly mobilized by permafrost thaw and coastal erosion (AMAP 2021; IPCC 2022). The previously frozen as well as the deposited atmospheric Hg is mobilized and released into aquatic and marine systems, where it may be transformed into MeHg, depending on Hg availability, redox conditions, and microbial activity (Ullrich et al. 2001; Lehnherr et al. 2011). This MeHg accumulates in fish and marine mammals, many of which are central to the diets, traditions, and well-being of Indigenous Peoples in the Arctic (AMAP 2021). Therefore, protecting environmental systems such as permafrost is not only a matter of ecosystem and landscape preservation, but is also directly linked to animal and human health.
Discussion 74 Figure 26: The One Health framework in the Arctic (from Westerveld et al. 2023). This study quantified Hg stocks and fluxes originating from the erosion of the Yukon coast, and provides important baseline data on Hg inputs from terrestrial sources. Significant knowledge gaps remain in understanding the complete picture of Hg cycling in the region. In particular, other fluxes such as atmospheric deposition, material transport via rivers and ocean currents, and biotic interactions have not been quantified and characterized so far. Future research should focus on MeHg concentrations, exposures, and transformation pathways, which will require high-resolution sampling of soils, vegetation, and biota, potentially during different seasons. Integrating Indigenous Knowledge through community engagement and two-way knowledge exchange not only improves study set-up and scientific understanding but also ensures relevance, acceptance, and accountability of the findings (Houde et al. 2022). Indigenous Knowledge has always been a holistic approach, including entire systems, and the concept of One Health is increasingly integrated into conservative research efforts. This study contributes to the One Health framework by providing data on environmental Hg occurrences and
Discussion 75 processes, building a foundation for further assessments on its effects on animal and human health in the Yukon coastal region. A comprehensive, interdisciplinary, and inclusive research approach is needed to fully comply to the One Health framework.
Conclusion 76 8 Conclusion This study aimed to quantify Hg stocks and fluxes along the Yukon coast, a region characterized by ongoing coastal erosion driven by permafrost thaw and increased wave action, with erosion rates reaching up to 22 m per year. To estimate Hg stocks and fluxes, sediment concentrations were determined using a combination of measurements from terrestrial and marine sediment samples as well as predictive modeling with a random forest algorithm. The obtained Hg concentrations were combined with sediment volume and observed erosion rates to calculate total and local stock and flux estimates. MeHg concentrations were analyzed for certain samples and compared across different landforms and along a marine transect in front of an RTS. Measured and modeled Hg concentrations were within typical ranges for the Arctic permafrost environments and clearly remained below the critical thresholds of 6,600 µg kg-1 given by the CCME for agricultural and residential soils (CCME 1999). Based on collected Hg concentrations, a total Hg stock of 381,080 kg (271,540 to 501,930 kg) was estimated for the YCP, with approximately 113 kg (87 to 163 kg) released into the Arctic Ocean annually due to coastal erosion. The magnitude of Hg release alone does not reflect the environmental relevance it may pose. The fate of Hg in the marine environment depends on its transport, transformation, and potential bioavailability. This study supports previous findings that a certain fraction of the eroded Hg is likely transported beyond the nearshore zone, may evade into the atmosphere, or bioaccumulate in marine food webs, as marine sediments showed lower Hg concentrations than terrestrial soil. Of particular concern is how much Hg is converted into the neurotoxic MeHg in the water column, which bioaccumulates in the food web. Although MeHg concentrations in both terrestrial and marine sediments were low, sediment data alone are insufficient to characterize MeHg dynamics in the broader marine environment. Consequently, this study does not allow for distinct conclusions regarding the risk to local communities from MeHg exposure through sea food consumption. No elevated MeHg levels in sediments were found, which suggests no unusual methylation activity in the immediate coastal zone.
Conclusion 77 Future research should focus on expanded soil sampling along the Yukon coast to refine stock and flux estimates and reduce current uncertainties. However, more critical is to put efforts on quantifying MeHg concentrations in sea water and in key species consumed by Indigenous Peoples in the region, as these pathways are most relevant for ensuring human and ecological health.
References 84 Lemenkova, P. (2020). Mapping Beaufort Sea Topography and Geophysical Settings Using HighResolution Geospatial Data and GMT. Geographical Information 24(1), 4–18. https://doi.org/10.17846/GI.2020.24.1.4-18 Liao, L., Selim, H. M. & Delaune, R. D. (2009). Mercury Adsorption‐Desorption and Transport in Soils. Journal of Environmental Quality 38(4), 1608–1616. https://doi.org/10.2134/jeq2008.0343 Liew, M., Xiao, M., Farquharson, L., Nicolsky, D., Jensen, A., Romanovsky, V., Peirce, J., Alessa, L., McComb, C., Zhang, X. & Jones, B. (2022). Understanding Effects of Permafrost Degradation and Coastal Erosion on Civil Infrastructure in Arctic Coastal Villages: A Community Survey and Knowledge Co-Production. Journal of Marine Science and Engineering 10(3), 422. https://doi.org/10.3390/jmse10030422 Lim, A. G., Jiskra, M., Sonke, J. E., Loiko, S. V., Kosykh, N. & Pokrovsky, O. S. (2020). A revised pan-Arctic permafrost soil Hg pool based on Western Siberian peat Hg and carbon observations. Biogeosciences 17(12), 3083–3097. https://doi.org/10.5194/bg-17-3083-2020 Lindqvist, O. & Rodhe, H. (1985). Atmospheric mercury—a review. Tellus B 37 B(3), 136–159. https://doi.org/10.1111/j.1600-0889.1985.tb00062.x Liu, Y., Chai, X., Hao, Y., Gao, X., Lu, Z., Zhao, Y., Zhang, J. & Cai, M. (2015). Total mercury and methylmercury distributions in surface sediments from Kongsfjorden, Svalbard, Norwegian Arctic. Environmental Science and Pollution Research 22(11), 8603–8610. https://doi.org/10.1007/s11356-014-3942-0 Loseto, L. L., Stern, G. A., Deibel, D., Connelly, T. L., Prokopowicz, A., Lean, D. R. S., Fortier, L. & Ferguson, S. H. (2008). Linking mercury exposure to habitat and feeding behaviour in Beaufort Sea beluga whales. Journal of Marine Systems 74(3-4), 1012–1024. https://doi.org/10.1016/j.jmarsys.2007.10.004 Mackay, J. R. (1959). Glacial ice-thrust features of the Yukon coast. Geographical Bulletin 13, 5. Mahbub, K. R., Bahar, M. M., Megharaj, M. & Labbate, M. (2018). Are the existing guideline values adequate to protect soil health from inorganic mercury contamination? Environment International 117. https://doi.org/10.1016/j.envint.2018.04.037 Malinauskaite, L., Cook, D., Davðsdóttir, B., Ögmundardóttir, H. & Roman, J. (2019). Ecosystem services in the Arctic: a thematic review. Ecosystem Services 36, 100898. https://doi.org/10.1016/j.ecoser.2019.100898 Manson, G. K., Couture, N. J. & James, T. S. (2019). CanCoast 2.0: data and indices to describe the sensitivity of Canada’s marine coasts to changing climate. https://ostrnrcandostrncan.canada.ca/entities/publication/caff0f1b-6adb-4470-be2c-d4461cf29793*, last checked on 29. June 2025 Manson, G. K. & Solomon, S. M. (2007). Past and future forcing of Beaufort Sea coastal change. Atmosphere - Ocean 45(2), 107–122. https://doi.org/10.3137/ao.450204 Manson, G. K., Solomon, S. M., Forbes, D. L., Atkinson, D. E. & Craymer, M. (2005). Spatial variability of factors influencing coastal change in the Western Canadian Arctic. Geo-Marine Letters 25(2-3), 138–145. https://doi.org/10.1007/s00367-004-0195-9 Mason, R. P., Reinfelder, J. R. & Morel, F. M. M. (1995). Bioaccumulation of Mercury and Methylmercury. Water, Air, and Soil Pollution 80, 915–921. McDonald, B. C. & Lewis, C. P. (1973). Geomorphic and sedimentologic processes of rivers and coast, Yukon Coastal Plain. Geological Survey of Canada, Environmental-Social Program Northern Pipelines. Meinshausen, N. (2024). quantregForest: Quantile Regression Forests (version 1.3-7.1) [Computer software]. Mekonnen, Z. A., Riley, W. J., Berner, L. T., Bouskill, N. J., Torn, M. S., Iwahana, G., Breen, A. L., MyersSmith, I. H., Criado, M. G., Liu, Y., Euskirchen, E. S., Goetz, S. J., Mack, M. C. & Grant, R. F. (2021). Arctic tundra shrubification: a review of mechanisms and impacts on ecosystem carbon balance. Environmental Research Letters 16(5), 53001. https://doi.org/10.1088/1748-9326/abf28b Miner, K. R., D’Andrilli, J., Mackelprang, R., Edwards, A., Malaska, M. J., Waldrop, M. P. & Miller, C. E. (2021). Emergent biogeochemical risks from Arctic permafrost degradation. Nature Climate Change 11, 809–819. https://doi.org/10.1038/s41558-021-01162-y
References 85 Molnar, C., Bischl, B. & Casalicchio, G. (2018). iml: An R package for Interpretable Machine Learning. Journal of Open Source Software 3(26), 786. https://doi.org/10.21105/joss.00786 Mu, C., Zhang, F., Chen, X., Ge, S., Mu, M., Jia, L., Wu, Q. & Zhang, T. (2019). Carbon and mercury export from the Arctic rivers and response to permafrost degradation. Water Research 161, 54–60. https://doi.org/10.1016/j.watres.2019.05.082 Mulligan, R. P. & Perrie, W. (2019). Circulation and structure of the Mackenzie River plume in the coastal Arctic Ocean. Continental Shelf Research 177, 59–68. https://doi.org/10.1016/j.csr.2019.03.006 National Snow and Ice Data Center (2025). Sea Ice Spatial Comparison Tool. https://nsidc.org/sea-icetoday/sea-ice-tools/sea-ice-spatial-comparison-tool*, last checked on 29. June 2025 Nesterova, N., Leibman, M., Kizyakov, A., Lantuit, H., Tarasevich, I., Nitze, I., Veremeeva, A. & Grosse, G. (2024). Review article: Retrogressive thaw slump characteristics and terminology. The Cryosphere 18(10), 4787–4810. https://doi.org/10.5194/tc-18-4787-2024 Nielsen, D. M., Pieper, P., Barkhordarian, A., Overduin, P., Ilyina, T., Brovkin, V., Baehr, J. & Dobrynin, M. (2022). Increase in Arctic coastal erosion and its sensitivity to warming in the twenty-first century. Nature Climate Change 12(3), 263–270. https://doi.org/10.1038/s41558-022-01281-0 Obrist, D., Agnan, Y., Jiskra, M., Olson, C. L., Colegrove, D. P., Hueber, J., Moore, C. W., Sonke, J. E. & Helmig, D. (2017). Tundra uptake of atmospheric elemental mercury drives Arctic mercury pollution. Nature 547(7662), 201–204. https://doi.org/10.1038/nature22997 Obu, J., Lantuit, H., Grosse, G., Günther, F., Sachs, T., Helm, V. & Fritz, M. (2017). Coastal erosion and mass wasting along the Canadian Beaufort Sea based on annual airborne LiDAR elevation data. Geomorphology 293, 331–346. https://doi.org/10.1016/j.geomorph.2016.02.014 Obu, J., Westermann, S., Bartsch, A., Berdnikov, N., Christiansen, H. H., Dashtseren, A., Delaloye, R., Elberling, B., Etzelmüller, B., Kholodov, A., Khomutov, A., Kääb, A., Leibman, M. O., Lewkowicz, A. G., Panda, S. K., Romanovsky, V., Way, R. G., Westergaard-Nielsen, A., Wu, T., et al., Zou, D. (2019). Northern Hemisphere permafrost map based on TTOP modelling for 2000–2016 at 1 km2 scale. Earth-Science Reviews 193, 299–316. https://doi.org/10.1016/j.earscirev.2019.04.023 Ogle, D. H., Doll, J. C., Wheeler, A. P. & Dinno, A. (2025). FSA: Simple Fisheries Stock Assessment Methods (version 0.9.6) [Computer software]. Olson, C., Jiskra, M., Biester, H., Chow, J. & Obrist, D. (2018). Mercury in Active-Layer Tundra Soils of Alaska: Concentrations, Pools, Origins, and Spatial Distribution. Global Biogeochemical Cycles 32(7), 1058–1073. https://doi.org/10.1029/2017GB005840 Outridge, P. M., MacDonald, R. W., Wang, F., Stern, G. A. & Dastoor, A. P. (2008). A mass balance inventory of mercury in the Arctic Ocean. Environmental Chemistry 5(2), 89–111. https://doi.org/10.1071/EN08002 Outridge, P. M., Mason, R. P., Wang, F., Guerrero, S. & Heimbürger-Boavida, L. E. (2018). Updated Global and Oceanic Mercury Budgets for the United Nations Global Mercury Assessment 2018. Environmental Science and Technology 52(20), 11466–11477. https://doi.org/10.1021/acs.est.8b01246 Overduin, P. P., Deimling, T. S. v., Miesner, F., Grigoriev, M. N., Ruppel, C., Vasiliev, A., Lantuit, H., Juhls, B. & Westermann, S. (2019). Submarine Permafrost Map in the Arctic Modeled Using 1-D Transient Heat Flux (SuPerMAP). Journal of Geophysical Research: Oceans 124(6), 3490–3507. https://doi.org/10.1029/2018JC014675 Pelletier, B. R. (1984). Marine science atlas of the Beaufort Sea: sediments. Energy, Mines and Resources Canada. Perryman, C. R., Wirsing, J., Bennett, K. A., Brennick, O., Perry, A. L., Williamson, N. & Ernakovich, J. G. (2020). Heavy metals in the Arctic: Distribution and enrichment of five metals in Alaskan soils. PLoS ONE 15(6). https://doi.org/10.1371/journal.pone.0233297 Petzold, P. (2024). Exploring the Impact of Wind Forcing on Organic Carbon Pathways in the Nearshore Zone of Herschel Island-Qikiqtaruk, Canada [Master's thesis]. University of Potsdam, Germay. Pickart, R. S. (2004). Shelfbreak circulation in the Alaskan Beaufort Sea: Mean structure and variability. Journal of Geophysical Research: Oceans 109(4). https://doi.org/10.1029/2003JC001912
References 86 R Core Team. (2025). R: A Language and Environment for Statistical Computing (version 4.3.3) [Computer software]. R Foundation for Statistical Computing. https://www.R-project.org/*, last checked on 29. June 2025 Radosavljevic, B., Lantuit, H., Knoblauch, C., Couture, N. J., Herzschuh, U. & Fritz, M. (2022). Arctic Nearshore Sediment Dynamics—An Example from Herschel Island—Qikiqtaruk, Canada. Journal of Marine Science and Engineering 10(11), 1589. https://doi.org/10.3390/JMSE10111589/S1 Ramage, J. L., Irrgang, A. M., Herzschuh, U., Morgenstern, A., Couture, N. J. & Lantuit, H. (2017). Terrain controls on the occurrence of coastal retrogressive thaw slumps along the Yukon Coast, Canada. Journal of Geophysical Research: Earth Surface 122(9), 1619–1634. https://doi.org/10.1002/2017JF004231 Rampton, V. N. (1982). Quaternary geology of the Yukon Coastal Plain. Geological Survey of Canada. https://searchworks.stanford.edu/view/1079504*, last checked on 29. June 2025 Rantanen, M., Karpechko, A. Y., Lipponen, A., Nordling, K., Hyvärinen, O., Ruosteenoja, K., Vihma, T. & Laaksonen, A. (2022). The Arctic has warmed nearly four times faster than the globe since 1979. Communications Earth and Environment 3(1), 168. https://doi.org/10.1038/s43247-022-00498-3 Raynolds, M. & Walker, D. (2022). Raster Circumpolar Arctic Vegetation Map. Mendeley Data V2. Reyes, F. R. & Lougheed, V. L. (2015). Rapid nutrient release from permafrost thaw in arctic aquatic ecosystems. Arctic, Antarctic, and Alpine Research 47(1), 35–48. https://doi.org/10.1657/AAAR0013099 Rivkina, E., Laurinavichius, K., McGrath, J., Tiedje, J., Shcherbakova, V. & Gilichinsky, D. (2004). Microbial life in permafrost. Advances in Space Research 33(8), 1215–1221. https://doi.org/10.1016/j.asr.2003.06.024 Rutkowski, C., Lenz, J., Lang, A., Wolter, J., Mothes, S., Reemtsma, T., Grosse, G., Ulrich, M., Fuchs, M., Schirrmeister, L., Fedorov, A., Grigoriev, M., Lantuit, H. & Strauss, J. (2021). Mercury in Sediment Core Samples From Deep Siberian Ice-Rich Permafrost. Frontiers in Earth Science 9, 718153. https://doi.org/10.3389/FEART.2021.718153/BIBTEX Salonen, V. P. & Korkka-Niemi, K. (2007). Influence of parent sediments on the concentration of heavy metals in urban and suburban soils in Turku, Finland. Applied Geochemistry 22(5), 906–918. https://doi.org/10.1016/j.apgeochem.2007.02.003 Schartup, A. T., Balcom, P. H., Soerensen, A. L., Gosnell, K. J., Calder, R. S. D., Mason, R. P., Sunderland, E. M. & St. Louis, V. L. (2015). Freshwater discharges drive high levels of methylmercury in Arctic marine biota. Proceedings of the National Academy of Sciences of the United States of America 112(38), 11789–11794. https://doi.org/10.1073/pnas.1505541112 Schirrmeister, L., Oezen, D. & Geyh, M. A. (2002). 230Th/U dating of frozen peat, Bol’shoy Lyakhovsky Island (northern Siberia). Quaternary Research 57(2), 253–258. https://doi.org/10.1006/qres.2001.2306 Schuster, P. F., Schaefer, K. M., Aiken, G. R., Antweiler, R. C., Dewild, J. F., Gryziec, J. D., Gusmeroli, A., Hugelius, G., Jafarov, E., Krabbenhoft, D. P., Liu, L., Herman-Mercer, N., Mu, C., Roth, D. A., Schaefer, T., Striegl, R. G., Wickland, K. P. & Zhang, T. (2018). Permafrost Stores a Globally Significant Amount of Mercury. Geophysical Research Letters 45(3), 1463–1471. https://doi.org/10.1002/2017GL075571 Schuster, P. F., Striegl, R. G., Aiken, G. R., Krabbenhoft, D. P., Dewild, J. F., Butler, K., Kamark, B. & Dornblaser, M. (2011). Mercury export from the Yukon River Basin and potential response to a changing climate. Environmental Science and Technology 45(21), 9262–9267. https://doi.org/10.1021/es202068b Schuur, E. A. G., McGuire, A. D., Schädel, C., Grosse, G., Harden, J. W., Hayes, D. J., Hugelius, G., Koven, C. D., Kuhry, P., Lawrence, D. M., Natali, S. M., Olefeldt, D., Romanovsky, V. E., Schaefer, K., Turetsky, M. R., Treat, C. C. & Vonk, J. E. (2015). Climate change and the permafrost carbon feedback. Nature 520(7546), 171–179. https://doi.org/10.1038/nature14338 Schwarzkopf, K. (2023). Distribution and origin of organic matter in marine surface sediments on the Canadian Beaufort Shelf [Master's Thesis]. University of Tübingen, Germany. Scudder, G. G. E. (1997). Environment of the Yukon. In H. V. Danks & J. A. Downes (eds.), Insects of the Yukon (13–57). Biological Survey of Canada (Terrestrial Arthropods).
References 87 Seelen, E., van Liem-Nguyen, Wünsch, U., Baumann, Z., Mason, R., Skyllberg, U. & Björn, E. (2023). Dissolved organic matter thiol concentrations determine methylmercury bioavailability across the terrestrial-marine aquatic continuum. Nature Communications 14(1), 6728. https://doi.org/10.1038/s41467-023-42463-4 Semiletov, I., Pipko, I., Gustafsson, Ö., Anderson, L. G., Sergienko, V., Pugach, S., Dudarev, O., Charkin, A., Gukov, A., Bröder, L., Andersson, A., Spivak, E. & Shakhova, N. (2016). Acidification of East Siberian Arctic Shelf waters through addition of freshwater and terrestrial carbon. Nature Geoscience 9, 361–365. https://doi.org/10.1038/NEGO2695 Siegel, S. M. & Siegel, B. Z. (1984). First estimate of annual mercury flux at the Kilauea main vent. Nature 309, 146–147. Skyllberg, U. (2012). Chemical Speciation of Mercury in Soil and Sediment. In G. Liu, Y. Cai & N. J. O’Driscoll (eds.), Environmental Chemistry and Toxicology of Mercury (1. edition, 219–258). Wiley Online Books. https://doi.org/10.1002/9781118146644.ch7 Skyllberg, U., Bloom, P. R., Qian, J., Lin, C. M. & Bleam, W. F. (2006). Complexation of mercury(II) in soil organic matter: EXAFS evidence for linear two-coordination with reduced sulfur groups. Environmental Science and Technology 40(13), 4174–4180. https://doi.org/10.1021/es0600577 St. Pierre, K. A., St. Louis, V. L., Kirk, J. L., Lehnherr, I., Wang, S. & La Farge, C. (2015). Importance of open marine waters to the enrichment of total mercury and monomethylmercury in lichens in the canadian high arctic. Environmental Science and Technology 49(10), 5930–5938. https://doi.org/10.1021/acs.est.5b00347 St. Pierre, K. A., Zolkos, S., Shakil, S., Tank, S. E., Louis, V. L. S. & Kokelj, S. V. (2018). Unprecedented Increases in Total and Methyl Mercury Concentrations Downstream of Retrogressive Thaw Slumps in the Western Canadian Arctic. Environmental Science and Technology 52(24), 14099– 14109. https://doi.org/10.1021/acs.est.8b05348 Stein, R. & MacDonald, R. W. (2004). The Organic Carbon Cycle in the Arctic Ocean. Springer Berlin Heidelberg. https://doi.org/10.1007/978-3-642-18912-8 Stern, G. A., MacDonald, R. W., Outridge, P. M., Wilson, S., Chételat, J., Cole, A., Hintelmann, H., Loseto, L. L., Steffen, A., Wang, F. & Zdanowicz, C. (2012). How does climate change influence arctic mercury? Science of the Total Environment 414, 22–42. https://doi.org/10.1016/J.SCITOTENV.2011.10.039 Stroeve, J. C., Serreze, M. C., Holland, M. M., Kay, J. E., Malanik, J. & Barrett, A. P. (2012). The Arctic’s rapidly shrinking sea ice cover: A research synthesis. Climatic Change 110(3-4), 1005–1027. https://doi.org/10.1007/s10584-011-0101-1 Suleymanov, A., Suleymanov, R., Kulagin, A. & Yurkevich, M. (2023). Mercury Prediction in Urban Soils by Remote Sensing and Relief Data Using Machine Learning Techniques. Remote Sensing 15(12). https://doi.org/10.3390/rs15123158 Tanski, G., Couture, N. J., Lantuit, H., Eulenburg, A. & Fritz, M. (2016). Eroding permafrost coasts release low amounts of dissolved organic carbon (DOC) from ground ice into the nearshore zone of the Arctic Ocean. Global Biogeochemical Cycles 30(7), 1054–1068. https://doi.org/10.1002/2015GB005337 Tanski, G., Wagner, D., Knoblauch, C., Fritz, M., Sachs, T. & Lantuit, H. (2019). Rapid CO2 Release From Eroding Permafrost in Seawater. Geophysical Research Letters 46(20), 11244–11252. https://doi.org/10.1029/2019GL084303 Tape, K. E., Sturm, M. & Racine, C. (2006). The evidence for shrub expansion in Northern Alaska and the Pan‐Arctic. Global Change Biology 12(4), 686–702. https://doi.org/10.1111/j.13652486.2006.01128.x Tarbier, B., Hugelius, G., Sannel, A. B. K., Baptista-Salazar, C. & Jonsson, S. (2021). Permafrost Thaw Increases Methylmercury Formation in Subarctic Fennoscandia. Environmental Science and Technology 55(10), 6710–6717. https://doi.org/10.1021/acs.est.0c04108 Terhaar, J., Lauerwald, R., Regnier, P., Gruber, N. & Bopp, L. (2021). Around one third of current Arctic Ocean primary production sustained by rivers and coastal erosion. Nature Communications 12(1), 169. https://doi.org/10.1038/s41467-020-20470-z
References 88 Timmermans, M.-L. & Toole, J. M. (2023). The Arctic Ocean’s Beaufort Gyre. Annual Reviews of Marine Science 15, 47. https://doi.org/10.1146/annurev-marine-032122Travnikov, O. (2005). Contribution of the intercontinental atmospheric transport to mercury pollution in the Northern Hemisphere 39, 7541–7548. https://doi.org/10.1016/j.atmosenv.2005.07.066 Turner, C. K. & Lantz, T. C. (2018). Springtime in the Delta: the Socio-Cultural Importance of Muskrats to Gwich’in and Inuvialuit Trappers through Periods of Ecological and Socioeconomic Change. Human Ecology 46(4), 601–611. https://doi.org/10.1007/s10745-018-0014-y Ullrich, S. M., Tanton, T. W. & Abdrashitova, S. A. (2001). Mercury in the aquatic environment: A review of factors affecting methylation. Critical Reviews in Environmental Science and Technology 31(3), 241–293. https://doi.org/10.1080/20016491089226 UNEP (2013). Global Mercury Assessment 2013: Sources, Emissions, Releases and Environmental Transport. UNEP Chemicals Branch. van Everdingen, R. O. (1976). Geocryological terminology. Canadian Journal of Earth Sciences 13(6), 862–867. https://doi.org/10.1139/e76-089 Varty, S., Lehnherr, I., St. Pierre, K. A., Kirk, J. & Wisniewski, V. (2021). Methylmercury Transport and Fate Shows Strong Seasonal and Spatial Variability along a High Arctic Freshwater Hydrologic Continuum. Environmental Science and Technology 55(1), 331–340. https://doi.org/10.1021/acs.est.0c05051 Vaysse, K. & Lagacherie, P. (2017). Using quantile regression forest to estimate uncertainty of digital soil mapping products. Geoderma 291, 55–64. https://doi.org/10.1016/j.geoderma.2016.12.017 Vonk, J. E., Fritz, M., Speetjens, N. J., Babin, M., Bartsch, A., Basso, L. S., Bröder, L., Göckede, M., Gustafsson, Ö., Hugelius, G., Irrgang, A. M., Juhls, B., Kuhn, M. A., Lantuit, H., Manizza, M., Martens, J., O’Regan, M., Suslova, A., Tank, S. E., et al., Zolkos, S. (2025). The land–ocean Arctic carbon cycle. Nature Reviews Earth and Environment 6(2), 86–105. https://doi.org/10.1038/s43017-024-00627-w Vorobyova, E., Soina, V., Gorlenko, M., Minkovskaya, N., Zalinova, N., Mamukelashvili, A., Gilichinsky, D., Rivkina, E. & Vishnivetskaya, T. (1997). The deep cold biosphere: facts and hypothesis. FEMS Microbiology Reviews 20(3-4), 277–290. https://doi.org/10.1111/j.1574-6976.1997.tb00314.x Wagner, J., Martin, V., Speetjens, N. J., A’Campo, W., Durstewitz, L., Lodi, R., Fritz, M., Tanski, G., Vonk, J. E., Richter, A., Bartsch, A., Lantuit, H. & Hugelius, G. (2023). High resolution mapping shows differences in soil carbon and nitrogen stocks in areas of varying landscape history in Canadian lowland tundra. Geoderma 438, 116652. https://doi.org/10.1016/j.geoderma.2023.116652 Wagner, J., Wolter, J., Ramage, J., Martin, V., Richter, A., Speetjens, N. J., Vonk, J. E., Lodi, R., Bartsch, A., Fritz, M., Lantuit, H. & Hugelius, G. (2025). Regional synthesis and mapping of soil organic carbon and nitrogen stocks at the Canadian Beaufort coast PREPRINT. https://doi.org/10.5194/egusphere2025-1052 Watras, C. J., Backa, R. C., Halvorsena, S., Hudson’, R. J. M., Morrisona, K. A. & Wente’, S. P. (1998). Bioaccumulation of mercury in pelagic freshwater food webs. The Science of the Total Environment 219, 183–208. https://doi.org/10.1016/S0048-9697(98)00228-9 Weege, S. (2016). Climatic drivers of retrogressive thaw slump activity and resulting sediment and carbon release to the nearshore zone of Herschel Island, Yukon Territory, Canada [Doctoral Thesis]. University of Potsdam, Germany. Wegner, C., Bennett, K. E., Vernal, A. d., Forwick, M., Fritz, M., Heikkilä, M., Łącka, M., Lantuit, H., Laska, M., Moskalik, M., O’Regan, M., Pawłowska, J., Promińska, A., Rachold, V., Vonk, J. E. & Werner, K. (2015). Variability in transport of terrigenous material on the shelves and the deep Arctic Ocean during the Holocene. Polar Research 34(1). https://doi.org/10.3402/polar.v34.24964 Welsh, S. L. & Rigby, J. K. (1971). Botanical and physiographic reconnaissance of Northern Yukon. Brigham Young University Science Bulletin 14(2). Wen, J., Wu, Y., Li, X., Lu, Q., Luo, Y., Duan, Z. & Li, C. (2021). Migration characteristics of heavy metals in the weathering process of exposed argillaceous sandstone in a mercury-thallium mining area. Ecotoxicology and Environmental Safety 208, 111751. https://doi.org/10.1016/j.ecoenv.2020.111751 Western Arctic (Inuvialuit) Claims Settlement Act (1984). Indian Affairs and Northern Development Canada.
References 89 Westerveld, L., Kurvits, T., Schoolmeester, T., Mulelid, O. B., Eckhoff, T. S., Overduin, P. P., Fritz, M., Lantuit, H., Alfthan, B., Sinisalo, A., Miesner, F., Viitanen, L. -K. & NUNATARYUK consortium (2023). Arctic Permafrost Atlas. GRID-Arendal. https://doi.org/10.61523/KPJI4549 Wetterich, S., Kizyakov, A. I., Opel, T., Grotheer, H., Mollenhauer, G. & Fritz, M. (2023). Ground-ice origin and age on Herschel Island (Qikiqtaruk), Yukon, Canada. Quaternary Science Advances 10, 100077. https://doi.org/10.1016/j.qsa.2023.100077 Wickham, H. (2016). ggplot2: Elegant Graphics for Data Analysis (version 3.5.0) [Computer software]. Springer-Verlag. https://ggplot2.tidyverse.org*, last checked on 29. June 2025 WMAC-NS (2019). Herschel Island-Qikiqtaruk Territorial Park Management Plan 2019. Wildlife Management Advisory Council–North Slope. Wolter, J., Lantuit, H., Fritz, M., Macias-Fauria, M., Myers-Smith, I. & Herzschuh, U. (2016). Vegetation composition and shrub extent on the Yukon coast, Canada, are strongly linked to ice-wedge polygon degradation. Polar Research 35(2016). https://doi.org/10.3402/polar.v35.27489 Xie, W., Yu, Q., Fang, W., Zhang, X., Geng, J., Tang, J., Jing, W., Liu, M., Ma, Z., Yang, J. & Bi, J. (2024). Data-driven approaches linking wastewater and source estimation hazardous waste for environmental management. Nature Communications 15(1). https://doi.org/10.1038/s41467-02449817-6 Xu, J., Kleja, D. B., Biester, H., Lagerkvist, A. & Kumpiene, J. (2014). Influence of particle size distribution, organic carbon, pH and chlorides on washing of mercury contaminated soil. Chemosphere 109, 99–105. https://doi.org/10.1016/J.CHEMOSPHERE.2014.02.058 Yang, D., Shi, X. & Marsh, P. (2015). Variability and extreme of Mackenzie River daily discharge during 1973-2011. Quaternary International 380-381, 159–168. https://doi.org/10.1016/j.quaint.2014.09.023 Yang, Z., Fang, W., Lu, X., Sheng, G. P., Graham, D. E., Liang, L., Wullschleger, S. D. & Gu, B. (2016). Warming increases methylmercury production in an Arctic soil. Environmental Pollution 214, 504–509. https://doi.org/10.1016/j.envpol.2016.04.069 Zhang, Y., Jacob, D. J., Dutkiewicz, S., Amos, H. M., Long, M. S. & Sunderland, E. M. (2015). Biogeochemical drivers of the fate of riverine mercury discharged to the global and Arctic oceans. Global Biogeochemical Cycles 29(6), 854–864. https://doi.org/10.1002/2015GB005124 Zolkos, S., Krabbenhoft, D. P., Suslova, A., Tank, S. E., McClelland, J. W., Spencer, R. G. M., Shiklomanov, A., Zhulidov, A. V., Gurtovaya, T., Zimov, N., Zimov, S., Mutter, E. A., Kutny, L., Amos, E. & Holmes, R. M. (2020). Mercury Export from Arctic Great Rivers. Environmental Science and Technology 54(7), 4140–4148. https://doi.org/10.1021/acs.est.9b07145
Appendix XIII Appendix The data used for this thesis and the R code of model predictions can be found in the following GitHub folder: https://github.com/kathijsp/master-thesis MeHg extraction Figure 27: MeHg extraction with two separate layers of two liquid mixtures: a lower clear layer containing DCM and the extracted MeHg, and an upper greenish layer containing KBr and CuSO4. The sediment has settled at the bottom of the tube.
Appendix XIV Thermokarst landforms at the Yukon coast Figure 28: Ice-wedge polygons (A) and thermokarst lakes (B) at the Yukon mainland coast, and coastal erosion at cliffs of HIQ (C). (A) (B) (C)
Appendix XV Figure 29: Slump D on HIQ. Figure 30: Active layer detachments on HIQ.
Appendix XVI Grain size distribution Figure 31: Grain size distribution for all YC24 samples. Cores , headwall, mud pool, stabilized zone, and thaw stream samples were taken from Slump D. T7 samples come from an exposed bluff at the mainland coast. Beach and marine samples are from the transect in front of Slump D. Mostly, samples are sorted by depth. Detailed information about samples can be found on GitHub.