Controls on stream hydrochemistry dynamics in a high Arctic snow-covered watershed

Arctic streams are highly sensitive to climate change due to warmer air temperature and increased precipitation associated with an encroaching low Arctic climatic zone into currently high ‐ Arctic coastal areas. Increases in nivation processes and permafrost degradation will lead to potential changes in stream physicochemical habitat, although these impacts are poorly understood. To address this gap, physicochemical habitat characteristics in streams around Zackenberg in Northeast Greenland National Park were investigated during the summers of 2013 to 2016. Streams with different sized snowpacks represented both low and high snowfall conditions leading to different nivation processes. Streams with larger snowpacks displayed lower channel stability, with higher channel mobility, suspended sediment and solute concentrations. Suspended sediment concentration was identified as a key driver of stream solute concentrations, and varying snowpack levels caused high interannual variability in solute concentrations. Winter snowpack size was confirmed to be an important driver of stream physicochemical habitat in an Arctic region with low glacial cover. We predict climate change will strongly impact stream hydrochemistry in this region through increased nivation processes alongside active layer thickening and solifluction, thereby increasing stream suspended sediment and solute concentrations. These findings indicate that hydrochemistry was principally a function of erosion, with variation being determined by spatial and temporal patterns in erosional processes, and as such, alternative methods to fingerprint water sources should be considered in this region. of

During spring snowmelt, sheet floods-where water moves in sheets instead of streams-can occur due to the frozen active layer preventing infiltration and causing large sediment deposition. Within a small area, permafrost landscapes can have diverse geomorphology, hydrology, and permafrost conditions (Haeberli, 2013). Both nivation processes and localized permafrost thaws can release large amounts of sediment and solutes to streams (Chin et al., 2016;Christiansen 1998b;Kokelj et al., 2015;Messenzehl, Hoffmann, & Dikau, 2014) and alter landscapes (Kokelj & Jorgenson, 2013). Given the low solute load of meltwater, the majority of solutes in stream water come from weathering processes (Holland, 1978), including through the erosion of suspended sediment in turbulent streamflow (Chin et al., 2016). In the Arctic, as in other cold climate regions, sub-zero temperature causes frost cracking and wedging, increasing both physical and chemical weathering (Bluth & Kump, 1994;Peters, 1984).
Although global climate models predict air temperature in Greenland by 2100 to increase by up to 5-7°C (IPCC, 2013), decreased continentality caused by declining sea ice may cause air temperature in the coastal northeast region to increase even higher with a potential 60% rise in precipitation as both snow and rain (Stendel, Christiansen, & Petersen, 2008). During this time, the number of thaw days in Northeast Greenland is expected to increase from 80 to 248 per year (Stendel et al., 2008), and upper permafrost layers could be at risk of degradation (Daanen et al., 2011;Hollesen, Elberling, & Jansson, 2011;Westermann et al., 2015). Deeper snowpacks can warm permafrost to a depth of 18 m (Rasmussen et al., 2017) due to greater insulation. Increased stream sediment and nutrient fluxes are expected through increased nivation processes and solifluction, identified as important periglacial processes impacting the region's geomorphology, (Christiansen, 1998a), and through permafrost degradation, thaw slumps, and rain-induced erosion events. Increased snow depth and active layer depth will increase nutrient and carbon run-off contributions to streams. Stream physicochemical processes influence stream nutrient spiralling, in-stream processing efficiency, suspended sediment concentration, and water temperature thereby significantly impacting on stream biota and their functioning (Berkman & Rabeni, 1987;Bilotta & Brazier, 2008;Chin et al., 2016;Milner, Brown, & Hannah, 2009;Prowse et al., 2006), The aim of this project was to understand the variability in channel stability in Northeast Greenlandic streams in relation to their snowpack size and its influence on physicochemical processes and stream hydrochemistry and to put this into the context of a changing climate. This was done by investigating the relationship between changing hydrology, the interaction with hydrogeology and the influence on hydrochemistry dynamics. Seven streams were selected in close proximity to Zackenberg research station in Northeast Greenland sourced from small, seasonal to large, perennial snowpacks, to represent low and high snowfall conditions, respectively. Streams were characterized in terms of their geomorphological and physiochemical characteristics. The hypotheses tested were (1) streams sourced from large snowpacks will have reduced channel stability due to the increased influence of nivation processes and spring floods on the stream bed and banks; and (2) lower channel stability will lead to higher solute concentrations due to high erosion and suspended sediment concentrations. The findings were placed in the context of a changing climate to understand snowmelt stream hydrochemistry dynamics in Arctic streams might shift in the future.
The Zackenberg research station (74°28′ N, 20°34′ W) is located within the Northeast Greenland National Park in the high Arctic climatic zone (Figure 1). Mean annual air temperature is −9.1°C with July the warmest month with a mean of 5.8°C and February the coldest month with a mean of −22.4°C. Annual mean precipitation is 261 mm of which approximately 10% falls as rain .
The valley floor was deglaciated 8,000 ybp with only a few small high altitude glaciers remaining in the area. Lying in the 1,300 km long East Greenland Caledonian belt (Higgins, Soper, & Leslie, 2000), geologically the area is divided into two parts, with crystalline (gneiss and granite) to the west and cretaceous and tertiary sandstones, conglomerates, black shale, and basalts to the east. The valley floor and low altitude slopes have a layer of loose soils that are well developed in some places but are generally vulnerable to erosion (Hasholt & Hagedorn, 2000;Mernild, Liston, & Hasholt, 2007).
The area is a zone of continuous permafrost, with depth modelled to be 200-300 m in the main valley and 300-500 m in the mountains (Christiansen, Sisgaard, Humlum, Rasch, & Hansen, 2008) with an active layer between 0.4 and 0.8 m (Hollesen et al., 2011;Westermann et al., 2015). Altitude varies between 0 and 1,450 m a. s.l with the glacial plateaux occurring above 1,000 m with wide horizontal valleys caused by glacial erosion below (Mernild, Liston et al., 2007). Periglacial features can be found in the area including ice wedges in the east and rock glaciers in the west, separated by the different geologies . Active layer sliding has been observed on Aucellabjerg.
Vegetation is divided by the geological areas; the western crystalline area is dominated by Vaccinium uliginosum (Bog bilberry) heath and is found with fens, whereas the easten sedimentary side is characterized by Cassiope tetragona (Arctic white heather) heath with Salix arctica (Arctic willow; Bay, 1998). Grasslands, fens, and snowbeds are common in the east, and the mountains are often unvegetated apart from the lower altitude slopes dominated by mountain avens (Dryas; Bay, 1998).
The run-off regime for the wider Zackenberg area is described as glacionival (Hasholt & Hagedorn, 2000). The principal streams included in this study were Kaerelv, Graenseelv, Aucellaelv, Unnamed1, and Palnatokeelv (Figure 1), which were all located in the sedimentary catchment and were sourced predominantly from snowmelt; however, Aucellaelv and Palnatokeelv had small high-altitude glaciers within their catchment. Kaerelv and Graenseelv were sourced from small, seasonal snowpacks and Aucellaelv and Palnatokeelv were sourced from large perrenial snowpacks. Unnamed1 was sourced primarily by a seasonal snowpack but also received contributions from a larger snowpack located nearby.
Lindemanelv and Unnamed2 were also included to represent contrasting physicochemical conditions. Lindemanelv received glacial meltwater, and Unnamed2 was located in the crystalline catchment ( Figure 1). The floodplains of streams Aucellaelv and Palnatokeelv consisted largely of stones, pebbles, and silt and lacked vegetation.
Kaerelv and Graenseelv floodplains are largely vegetated. The floodplain for Unnamed1 is vegetated at Sites A and B but consists of stones and pebbles at Site C. Table 1 shows site characteristics.

| Sampling framework
Air temperature (°C), precipitation (mm), and snow depth (cm) data were obtained from a weather station maintained by the Greenland Ecosystem Monitoring Programme , within 5 km of all sites. Air temperature and snow depth were recorded every 30 min whereas precipitation was recorded hourly.
Sampling took place over three early summer field seasons, from June 26 to July 17, 2013, July 1 to July 22, 2014, and July 6 to July  starting at around 9 a.m., with a space of no more than~90 min between sites. Samples were collected where water was assumed to be completely mixed.
Channel stability was calculated using the whole Pfankuch Index in 2015 (Pfankuch, 1975). Although this method is not quantitative, it is a reliable assessment of stream channel stability when the same observers undertake similtaneous assessments at streams targeted for comparison (Peckarsky et al., 2014), as was the case in this study.
Stream discharge was calculated using the velocity-area method using a flow metre (μP-TAD from Höntzsch instruments, Germany) wetted width and average depth and was measured on sampling days in 2015 and on sample day or on numerous days in 2014, highlighting stage variation throughout the season and stream sensitivity to rain events. Suspended sediment was measured in 2014 and collected from Site C in the lower reaches. Samples were collected where water was well mixed and were collected in 1-L containers before being passed through a preweighed Whatman Glass Fibre Filter paper. Filter papers with sediment were dried at 60°C for 48 hr and then reweighed to calculate sediment weight.
Electrical conductivity was measured continuously at three sites for 11 days in July 2014 using gauging stations which were installed at the streams. Data were recorded on Campbell Scientific CR1000 data loggers and EC sensors which scanned every 10 s and recorded data every 15 min.  to a lack of available imagery from the summer months of during the field campaign, imagery was used from August 2012.
Crustal proportions could not be calculated due to the lack of Cl − data; however, this was presumed to not be a problem for these ratios due to the small proportion of these solutes that originates in the sea (Mg 2+ : 0.06, Ca 2+ : 0.02, and K + : 0.02) as opposed to the snowpack (Holland, 1978).

| Data analysis
Normality of data was tested using Levene's test and residual plots. Nonnormally distributed data were natural log transformed before analysis. One-way analyses of variance (ANOVAs) were undertaken for air temperature, precipitation, and snow depth variables to characterize differences in weather conditions between the field seasons and Pfankuch Index and suspended sediment between streams to determine significant differences in channel stability. To analyse variation in stream hydrochemistry between years, sites, and stream, both one way and two-way ANOVAs were used, and significant results then underwent Tukey post hoc tests. Differences between streams were not analysed in 2013 due to the lack of repeated samples. Pearson product-moment correlation coefficient was employed to ascertain the relationship between Pfankuch stability index with stream hydrochemistry, suspended sediment concentration, and conductivity and to test for the relationship between stream conductivity and precipitation.

| Snowpack size and stream channel stability
The area of principal snowpacks varied. Kaerelv, Graenseelv, and Unnamed1 were sourced from snowpacks with an area of 0.01 km 2 or under; Aucellaelv was sourced from a snowpack of 0.06 km 2 and Palnatokeelv from a snowpack of 0.08 km 2 .
Discharge at streams Kaerelv, Graenseelv, and Unnamed1 increased downstream in 2015 (    where discharge varied from 181 l s −1 on a dry day to 622 l s −1 following a rainstorm event.
Channel stability was classified as either fair or poor channel stability at all sites except Kaerelv Sites B and C which were designated as good (

| Spatial variation in stream hydrochemistry dynamics
Conductivity varied markedly between streams but not within streams (Table 4) (Pfankuch, 1975), discharge, and suspended sediment concentration for each stream and longitudinal site (A-C) Note. For channel stability, "excellent" represents highly stable channels and "poor" represents highly unstable channels. Significant differences for channel stability are shown; significant differences in suspended sediment are shown.
Marked differences were found between streams in hydrochemistry dynamics (Table 5). There was no significant correlation between channel stability and hydrochemistry, Na + , K + , Si, NO 3 − , and NH 4 + were all significantly positively correlated with suspended sediment concentration (Table S1).      Graenseelv, and Unnamed1) (Figure 3), paralleling an increase in downstream discharge.

| Interannual variation in stream hydrochemistry dynamics
Large differences in hydrochemistry were found between years. Conductivity was highly variable, with the highest conductivity recorded in  and Ca 2+ concentration (3,152.8 μEq L −1 ) were recorded in Aucellaelv.    (Table S2). There were significant differences in NH 4 + between years but no significant differences between streams in 2015. In 2014, Aucellaelv had higher NH 4 concentrations than all other streams (Table 5) 0.29 μEq L −1 ). Si was negligible in snow, below the detection limit (<0.4 mg L −1 ; Table 5). Soil water collected in 2016 had very high solute concentrations including highest Si concentrations measured (between 2.61 and 4.55 mg L −1 ; Table 5). This is thought to be due to the later sampling date compared with previous years, means water will have had a longer soil residence time before entering the stream.

| DISCUSSION
Winter snowfall, rainfall, suspended sediment concentration, and underlying geology have been identified as the principal drivers of hydrochemistry dynamics in streams in the Zackenberg valley. Soil water is shown to have little influence on stream hydrochemistry during early summer; however, limited results show that soil water may have more influence towards the end of the summer season, when active layer is thickest.

| Stream channel stability and suspended sediment characterization
Stream channel stability in this region is dependent upon a number of geomorphological and hydrological variables. The largest snowpacks accumulate on south facing lee slopes, where northerly winds blow winter snow and sediment into moraine ridges and fluvial terraces (Christiansen, 1998b) which then melt to feed streams during the summer months. Larger snowpacks can cause greater geomorphological disturbance through nivation processes, which are emphasized on loose unconsolidated sedimentary soils due to increased water infiltration and greater surface area in contact with melted snow, increasing stream suspended sediment load, and reducing channel stability (Christiansen, 1998b;Hasholt & Hagedorn, 2000).
Sheet floods occur during spring snowmelt around the streams at Zackenberg, typically until June due to the frozen active layer preventing infiltration. These flood waters carry high sediment loads leading to sediment deposition (Cable, Christiansen, Westergaard-Nielsen, Kroon, & Elberling, 2017;Christiansen, 1998a). Aucellaelv and Palnatokeelv have higher spring discharges than streams sourced from smaller snowpacks and carry a larger sediment load. This can lead to increased downstream disturbance, leading to bare ground where vegetation is unable to colonize and through this process, create a supply of loose sediments that can enter the water column throughout the summer season. Highest suspended sediment concentrations at upstream sites in Aucellaelv have been found previously (Hasholt & Hagedorn, 2000) highlighting sediment deposition on alluvial cones. Alongside spring floods as a source of sediment, the dirty snow on Aucella mountain in 2013 indicates aeolian sediment transport into snowpacks during winter , and although infrequent, rainstorm events are known to influence sediment flux by driving increased erosion in sparsely vegetated areas in the sedimentary region (Rasch, Elberling, Kakobsen, & Hasholt, 2000).
Significantly lower channel stability in streams sourced from perennial snowpacks than smaller seasonal snowpacks allowed acceptance of Hypothesis 1. However, as channel stability was not signifi- Within the wider Zackenberg river catchment, the predominantly glacier sourced streams overlying crystalline bedrock carry very little sediment to the Zackenberg river. The streams in the sedimentary catchment, which account for only 10-20% of total catchment area and include Aucellaelv, Palnatokeelv, Unnamed1, and Lindemanelv, account for 90% of the sediment transported to the main Zackenberg river (Jakobsen, 1992), with average annual suspended sediment fluxes between 43,000-61,000 ty −1 (Ladegaard-Pedersen et al., 2017). Whilst glacial streams are known to be highly turbid, especially in the early melt season (Gurnell, 1987;Milner & Petts, 1994), it is the snowmelt streams of the Zackenberg drainage basin that transport the most sediment due to their underlying sedimentary material. This situation highlights the importance of characterizing geology into studies of Arctic streams and the important influence of nivation processes and permafrost degradation for sediment transport.

| Spatial variation in channel stability and stream hydrochemistry dynamics
The geological division in the Zackenberg valley between the sedimentary eastern hills and the crystalline western hills caused the differing solute concentrations in Unnamed2 compared with the other study streams. The eastern slopes, which sourced all other study streams, are modified by a combination of cryogenic, nival, fluvial, aeolian, and mass movement processes, which lead to loose, fine-grained sediment entering stream channels in this region. Unnamed2 on the western slopes which are dominated by gravitational processes, as thus such sediment transported by streams in this region is coarser and less likely to reach as far downhill . Nivation processes and permafrost degradation have limited influence in this catchment (Christiansen & Humlum, 1993). This lack of loose, fine sediment, and erosional processes leads to reduced solute load in Unnamed2 compared with the other streams.
Of the streams within the sedimentary region, Aucellaelv and Palnatokeelv, with larger snowpacks, also overlie large areas of solifluction and have notable nivation hollows along their stream banks, which could be responsible for the large sediment load within the stream channels. Kaerelv, Graenseelv, and Unnamed1 are largely overlying alluvial fans, peat bogs, and lateral moraines, similar in Aucellaelv and Palnatokeelv in their lower reaches . The higher suspended sediment concentration in Aucellaelv compared with other streams is likely the cause of the significant difference in solute load for most cations through instream weathering processes of suspended sediment through turbulent stream flow (Chin et al., 2016). Similar to Aucellaelv, the higher suspended sediment concentrations recorded in Palnatokeelv and Lindemanelv was likely due to weathering of rock-derived sediment from nivation processes and permafrost degradation. Although high suspended sediment concentration is a characteristic feature of streams receiving glacial inputs, given the timing of this field campaign in early July during the peak snowmelt period, and the small size of glaciers located in this catchment, glacial inputs were thought to be minimal during the sampling period.
The higher levels of Ca 2+ and Mg 2+ in Aucellaelv are probably derived from black shales in this region (Hasholt & Hagedorn, 2000).
Differences between the study streams in terms of weathering processes were not found but did corroborate findings from Hasholt and Hagedorn (2000) that silicate weathering is the dominant weathering process in the region as shown in the low K + : Si ratios and the low carbonate dissolution in the Ca 2+ : Mg 2+ ratios. This is typical of nonglacierized Arctic catchments, where carbonates and evaporates typical of glacierized catchments (Bluth & Kump, 1994) have been used up, and due to the increased contact with rock, longer residence times, and interaction with the active layer (Anderson, Drever, Frost, & Holden, 2000;Blaen et al., 2013;Fortner, Tranter, Fountain, Lyons, & Welch, 2005).

| Interannual variation in stream hydrochemistry dynamics
Interannual variation in solute concentration was principally due to

| Water sources and their impact on stream hydrochemistry
Water source is a known driver of hydrochemistry, and previous studies have shown variation in solute concentration throughout the summer period due to changes in water sources (e.g., Rasch et al., 2000). Conductivity was highest during the first few days after spring ice break due to the high dissolved load washing out of the first summer snowmelt event (Mernild, Sigsgaard et al., 2007). During the main field campaigns in July each year, the dominant water source for all sites was snow melt. Given the shallow active layer depth and snowmelt pools that had formed nearby, soil water input was probably low during this time period, with the soil water sample collected most likely recently leached snowmelt. Palnatokeelv, Lindemanelv, and Aucellaelv also receive glacial meltwater contributions which is known to lead to reduced channel stability, increased sediment load, and more extreme physicochemical habitat for biota (Milner & Petts, 1994). The relative minimal glacial inputs into these systems during the field campaign mean that they can be classified as nival systems following the classification of Brown, Hannah, and Milner (2003). Given the low solute concentration of snow and the shallow active layer, stream hydrochemistry during July is likely a function of nivation processes causing localized erosion and varying suspended sediment concentrations. However, towards the end of summer, as snowpacks decline and active layer thickness increases, streams receive larger soil water inputs, with the largest contributions during August (Blaen et al., 2013;Rasch et al., 2000). The high solute concentrations measured in Kaerelv, Graenseelv, and Aucellaelv in August 2016 reflect this. During this time, the interaction between stream water and soil water and access to previously frozen solutes from the thicker active layer were key drivers of later summer stream hydrochemistry dynamics. The large spatial and temporal variation in Si concentration shows that in these systems hydrochemistry cannot be used for fingerprinting water source as is traditionally used (e.g., Tranter et al., 1996), but rather, its variation is a product of the spatial and temporal variation of erosion.
As such, alternative methods would have to be implemented within this region to determine basin-scale water sources.

| Regional implications of climate change and conclusions
Northeast Greenland has been predicted to be warmer, wetter, and windier by the end of the century (Stendel et al., 2008), directly influencing stream systems in the region. Active layer thickness on Aucellabjerg and the valley bottom is predicted to increase by 8-12 cm, causing active layer detachments and slides to become frequent processes , leading to an increase in sediment, solutes, and soil water entering streams. Winter precipitation is expected to increase by 40-60% (Stendel et al., 2008). This could lead to larger spring floods increasing sedimentation along stream banks, higher water levels, and increased sediment and solute load in streams due to increased nivation processes and permafrost slumping. The predicted increase in summer precipitation is highly likely to increase weathering processes and so increase stream solute loads Rasch et al., 2000). These climatic changes are expected to cause stream systems to have reduced channel stability and increased suspended sediment concentration, with consequences for stream hydrological and ecological dynamics.
The impacts of these climatic changes are predicted to cause low stability stream systems to become increasingly widespread. This study shows least stable streams and those with highest suspended sediment concentration to have the highest nutrient content. All streams in this study are known to be nutrient limited with respect to primary production (Docherty, Riis, Hannah, Rosenhøj Leth, & Milner, In Press). Increased N and P nutrient inputs into nutrientpoor Arctic streams can increase primary productivity, providing the base of the food web for increased macroinvertebrate diversity and abundance. However, increased nutrient input through nivation processes and permafrost degradation is accompanied by increased suspended sediment inputs, and evidence shows a negative correlation between suspended sediment content and macroinvertebrate abundance (Chin et al., 2016), counteracting the positive impacts of additional nutrient inputs. High suspended sediment concentration causes reduced light penetration through the water column and this combined with high channel mobility can reduce primary producer growth (Ryan, 1991), reducing food availability for macroinvertebrates. Previous studies have found an increase in suspended sediment to be correlated with decreases in macroinvertebrate density, abundance, and richness (Nuttall & Bielby, 1973;Quinn, Davies-Colley, Hickey, Vickers, & Ryan, 1992;Shaw & Richardson, 2001;Wagener & LaPerriere, 1985) and an increase in invertebrate drift (Bilotta & Brazier, 2008;Doeg & Milledge, 1991;Rosenberg & Wiens, 1978). Suspended sediment can cause gills and guts to become clogged (Alabaster & Lloyd, 1982;Bilotta & Brazier, 2008), can smother macroinvertebrate eggs (Jones et al., 2012), and can impede respiration and feeding in Chironomidae, being especially damaging to those that produce silk tubes (Chin et al., 2016;Gray & Ward, 1982). Species-types tolerant of harsh environments such as Diamesa spp. are expected to be more common in these environments. Further research is needed within the Arctic region to fully understand these process changes to their impact on benthic communities.