Draft Multi-year variations in High Arctic river temperatures in response to climate variability

Water temperature measurements (2004–2016) from two small rivers in the High Arctic were analyzed to determine the effects of climate variability on thermal regime and the sensitivity to climate change. The East and West rivers (unofficial names) drain similar watersheds (11.6 and 8.0 km2, respectively) and are located at the Cape Bounty Arctic Watershed Observatory (CBAWO), Melville Island, Canada (74°55′N, 109°35′W). Differences in seasonal timing of river temperatures were evident when comparing the coldest and warmest years of the study period, and across different discharge conditions. Snowmelt runoff is characterized by uniformly cold water (∼0–1 °C) over a wide range of discharge conditions, followed by warming water temperatures during flow recession. The rivers showed varying sensitivity to mid-summer air temperature conditions in a given year, with warmer years indicating high correlation (r2 = 0.794–0.929), whereas colder years showed reduced correlation (r2 = 0.368–0.778). River temperatures reached levels which are reported to negatively affect fish and other cold-water aquatic species (>18 °C) with greater frequency and duration during the warmest years. These results provide a basis to further enhance prediction of river thermal conditions to assess ecosystem health in a river system and to refine insights into the effects of climate change on High Arctic aquatic ecosystems.


Introduction
Climate models predict that temperature and precipitation will increase globally, with high latitude regions having a greater rate of change (IPCC 2014, Nilsson et al. 2015).Such changes in northern regions can alter hydrological systems and affect water resources in terms of quantity, quality and temperature (Lammers et al. 2007;Webb et al. 2008;IPCC 2014;AMAP 2017).These changes have implications for fish and other aquatic species that are known to be influenced by the frequency, duration and intensity of exposure to high water temperatures (Boulton et al. 1998;Callaghan et al. 2011;Nilsson et al. 2015), particularly for organisms accustomed to relatively cold water (<10 °C) (Reist et al. 2006).Moreover, hydrological systems in cold region settings can be highly variable on a yearly basis due to changes in large scale atmospheric circulation (AMAP 2017).
Although knowledge concerning river temperatures, heat fluxes, and general seasonal regimes within Arctic regions is emerging (Nilsson et al. 2015;Whitefield at al. 2015), much is still unknown about the interannual and long term thermal state of Arctic rivers in the context of a changing climate.Temperatures in temperate hydrological systems have been found to be largely influenced by meteorological controls (Webb and Zhang 1999;Cassie 2006;Webb et al. 2008) but similar investigations in the Arctic are uncommon (e.g., Blaen et al. 2013;King et al. 2016), and unavailable in the Canadian High Arctic.As such, long term records of river temperatures are necessary in order to advance our knowledge of how climate variability effects the seasonal timing, sensitivity and intensity and duration of extreme temperatures within a changing hydrological system.This information is critical for validation for numerical D r a f t models which attempt to predict the effects of climate change for ecosystem monitoring and management (King et al. 2016).
We present a record of water temperature from two similar High Arctic rivers from 2004 to 2016 in order to evaluate long term temperature trends in the context of climate change, particularly recent record warmth (Lamoureux and Lafrenière 2017).The main objectives of this study are to: (1) identify river temperature patterns that result from interannual changes in climate, particularly through a period of notable warming; (2) determine the meteorological and discharge factors that explain variations in water temperature; and (3) determine the probabilities of thermal exceedance for each river to evaluate likely responses to projected future climate.Results from this study provide a framework for understanding the factors that affect the High Arctic river thermal regime and provide insights into likely future thermal conditions in similar river systems.

Study Site
Research was carried out at the Cape Bounty Watershed Observatory (CBAWO), Melville Island, Nunavut, Canada (74°55' N, 109°35' W, Figure 1).CBAWO was established in 2003 in order to investigate climate change influences on hydrological processes, permafrost, soils, vegetation, greenhouse gas emissions, and contaminant cycling in High Arctic rivers and lakes.
The West and East rivers (8.0 km 2 and 11.6 km 2 , respectively, unofficial names) are the focus of this study and their watersheds have similar geology, topography, climate, hydrology, vegetation, soils and geomorphic evolution (Lamoureux and Lafrenière 2017).

D r a f t
The climate of the area is a polar semi-desert with long cold winters and a short melt summer season.Mean January temperature is -31 °C and mean July temperature is 4.0 °C.Precipitation occurs primarily as snow (mean 120 mm) and with sporadic summer rainfall (mean June-July of 29 mm) (Lamoureux and Lafrenière 2017).The landscape is underlain by continuous, thick permafrost (c.500 m) with a seasonal active layer that can reach 70 to 90 cm depth by late summer depending on thaw conditions for a given year (Lamoureux and Lafrenière 2017).Vegetation is characterized by prostrate shrub tundra that is <5 cm in height, hence there is not a canopy that shields the rivers from solar insolation.
Both rivers are gravel bedded with alternating pools and riffles, and in some reaches are braided due to sedimentary load and channel width.Runoff is derived primarily from winter snow, which is extensively redistributed by wind and accumulates to depths of 1-5 m within land concavities and segments of the river channels.In some cases, ice and snow accumulations persist in these river channels, partly or entirely through the summer period (Lamoureux et al. 2006;Bolduc et al. 2018).Initial streamflow typically begins in mid-June with discharge reaching peak flow after a few days.Snowmelt recession and baseflow follows due to rapid depletion of snow, with sporadic pluvial (rainfall) events supplementing discharge until river freeze up occurs in late August or early September (Lewis et al. 2012).Water depths at the gauging stations vary from 0.3-0.4m during snowmelt, to 0.1-0.2m during baseflow, and these are consistent with the flow depths elsewhere in the catchments.River discharge data are reported in Schevers et al. (submitted).Briefly, river stage was measured using either an Omega CP-Level 101 (±0.2%) or an Onset U20 water level logger (±0.3%) at 10 minute intervals (Table 1).In 2012 and 2014, an Unidata capacitive water depth probe logged with an Onset H22 logger deployed for redundancy was utilized when the primary D r a f t logger battery failed.Discharge was calculated from stage measurements using rating curves derived from manual velocity-depth cross-sectional profiles of each river by using either a General Oceanics or a Swoffer 2100 flowmeter (both ±1%) near the outlet of each river.Crosssectional velocity measurements were recorded at 0.1-0.5 m width intervals and captured a wide range of discharge conditions in ice-free channels from the start of snowmelt to baseflow (typically n=10-24 per season).We follow the hydrological period zonation (snowmelt, recession, baseflow, rainfall) used by Schevers et al. (submitted) for consistency.

River temperature and discharge data analysis
In order to identify seasonal and long term trends regarding river temperatures, mean and maximum daily river temperatures and discharge were calculated.Raw river temperature data was also used to calculate thermal exceedance and cumulative thermal probabilities on a log probability plot by ranking the raw data from largest to smallest and then calculating a plotting each position of this data using a relative frequency definition as follows: where PP is the plotting position, n is the ranking positon of river temperature for each river.
Data from the 2005 season was omitted from these calculations due to missing temperature data for most of the mid to late summer season.Multiple linear regressions were performed between river temperatures (dependent variable), discharge and air temperature (independent variables) with SigmaPlot 12.5 to test for correlations amongst these primary variables.All discharge, precipitation, river temperatures and air temperature data used for multiple linear D r a f t regression analysis was collected from June 7 to August 10 which is the typical range for a summer season at CBAWO.
To identify variations in water temperature with respect to air temperature, the thermal response for the rivers was calculated by dividing mean daily water temperatures (T w ) with mean daily air temperatures (T a ) within a given hydrological day (0500 h to 0500 h, daily minimum discharge): Values for each daily thermal response was then sorted into classifications adapted from (Piccolroaz et al. 2016).Slope values ranging from 0 to 0.44 were classified as being resilient, as the influence of snow or inputs of colder water have been known to diminish the influence that changing air temperatures have on river temperatures (Lisi et al. 2015;Piccolroaz et al. 2016;Bolduc et al., 2018).Values ranging from 0.45 to 0.65 were classified as mixed and represent a transitional phase of a river's thermal response between reactivity and resiliency.Slope values > 0.65 were classified as reactive where river temperatures show the greatest response to air temperatures.Examples of these classifications in the East River during the year 2012 are illustrated in Figure 2.

Meteorology and discharge characteristics
Mean daily air temperature during runoff between 2004 and 2016 at the CBAWO was 3.9 °C, with 2004 as the coldest year (1.4 °C) and 2012 the warmest year (6.8 °C) (Figure 3).

D r a f t
Mean total rainfall was 33.7 mm, with 2014 the driest season (12.8 mm) and 2009 the wettest season (52.2 mm).In general, total monthly precipitation was higher during the mid-to latesummer season (July) compared to the early season (June).Major precipitation events occurred in certain years such as 2009 and were associated with increased discharge (Favaro and Lamoureux 2014).
River discharge for both rivers is typical of an Arctic nival runoff regime (Woo, 2012).
Initial discharge typically begins shortly after mean air temperatures rise above 0 °C, with discharge in the warmest year (2012) beginning up to three weeks earlier than in 2009 (Figure 4), which was a cold season similar to 2004 (Figure 3).Overall, the West River exhibited higher discharge compared to the East River, with discharge ranging from 0.01 to 3.48 m 3 /s and 0.01 to 3.05 m 3 /s, respectively.Daily discharge was typically at its minimum in the morning at approximately 1000 h and its daily maximum in the evening at 1800 h.Recession in most years was largely complete by early July and baseflow continued to the end of the measurement period.While discharge during this 12-year period has been influenced by precipitation events (Figure 4), the rejuvenation of discharge was usually of limited volume and short lived (Favaro and Lamoureux 2014).A notable exception was a major rainfall event in late July, 2009, that generated the maximum seasonal instantaneous discharge (Figure 4).

River temperatures
Seasonal mean river temperatures were 4.1 °C for the East River and 4.4 °C for the West River during the study period, with 2004 being the coldest and 2012 being the warmest year for D r a f t both air and river temperatures (Figure 3).Daily air temperature typically reached a minimum at 0400 h and a maximum at 1600 h.River temperature typically lagged air temperature by ~1 hour, with respective daily minimum and maximum values occurring at 0500 h and 1700 h for both rivers.Overall, air temperature and river temperature from 2004-2016 are significantly correlated (West: r = 0.79, n =10498, p < 0.001) and (East: r = 0.80, n = 9510, p < 0.001) although there are substantial differences between seasons (Table 2).
On a seasonal basis, both rivers were dominated by thermal conditions between 0-10 °C, with temperatures close to 0 °C during much of the snowmelt period.Water temperatures begin to increase as flow recession occurs and are characterized by pronounced diel variability (Figure 4).Mean daily river temperatures for the rivers during June were relatively cold (≤ 10 °C) with both rivers warming up in the month of July (Figure 5).During warmer years such as 2007 and 2012, both rivers exhibited higher mean daily river temperatures than the average in the early season (mid to late June), while colder seasons such as 2004 and 2009 delayed the warming of river temperatures into early to mid-July (Figures 3, 5).There was a clear contrast in mean daily river temperatures when comparing warm and cold years, with the coldest year (2004) ranged up 4.3 °C in the East River and 5.9 °C in the West River, while the warmest year (2012) ranged up to 12.2 °C in the East and up to 13.9 °C in the West River (Figure 5).Overall, the West River had higher daily mean temperatures than the East River for all years except between 2005 to 2007.However, while the West River might have had higher mean temperatures, the East River exhibited higher maximum daily temperatures for almost all days throughout the study period, suggesting different thermal controls for each river (Figure 6).The highest recorded river temperature recorded at Cape Bounty was during July 2007 when the D r a f t East River reached a maximum river temperature of 22.3 °C (Figure 6).In contrast, the West River only reached a maximum temperature of 14.7 °C during the same period.

River temperatures and thermal exceedance
Overall, water temperatures in the West River indicate more thermal stability than the East River and fewer occurrences of temperature exceeding 16 °C.The thermal residence time (or duration) of river temperatures above this threshold was 58 hours for the West River and 75 hours for the East River.However, while the East River typically achieved a higher daily thermal maximum, the daily maximum temperature was maintained typically only for 1 hour, while the West River was comparatively more stable by maintaining the daily maximum temperature for ~3 hours.
Analysis of thermal exceedance probabilities was also performed in order to examine the likelihood for each river to reach high to extreme temperature levels (>16 °C) in a High Arctic setting (Figure 7).On a yearly basis, the probability of reaching temperatures above this threshold is low and only occurs during years that have unusually warm seasonal conditions such as 2007 and 2012 (Figure 7).The probability of river temperatures exceeding the 16 °C threshold is more likely for the East River (3% and 5% in 2012 and 2007, respectively) and the West River at 5% (for 2012 only).

Seasonal patterns to water thermal response
To further observe the relationship between air temperatures and river temperatures for each river, a thermal response was calculated between water temperatures (T w ) and air temperatures (T a ) to classify the thermal state of the river on a given hydrological day (Lisi et al. 2015;Piccolroaz et al. 2016).The thermal response of the rivers on a given day was determined by categorizing the thermal response as being either thermally resilient, mixed or reactive (Figure 8).
Both rivers are characterized by a period of early season resilient water temperatures associated with snow melt and low temperatures.This phase of resiliency lasts 7-12 days in most years and coincides with the highest discharge of the season (Figure 8).By contrast, recession and baseflow water temperatures are more reactive, often showing a high correlation between air and water temperature.While individual days in a given season show a return to resiliency or a mixed response, the East River broadly shows sustained reactivity throughout the recession and baseflow period (Figure 8).This thermal reactivity is associated with the highest observed water temperatures in the record (Figures 3, 4).The West River appears to be less consistently thermally reactive during the mid-season (Figure 8).During the years 2004 to 2009, the water temperature shows an initial phase of reactivity for up to 12 days, followed by a return to more frequent resiliency, especially in 2006 and 2007.This return to thermal resiliency is not evident in 2010 to 2016, where mid-and late season water temperature is reactive to air temperature and similar to the East River response (Figure 8).

D r a f t
A comparison of two of the warmest years (2007 and 2012) illustrates these thermal response patterns.The late season resilience in 2007 is associated with intermediate (~4-8 °C) water temperatures and includes runoff from late season major rainfall events that increased discharge (Figure 9).Rainfall appears to moderate the sensitivity of water temperature change during the mid-summer season (July 1 -July 20), and other occasions in the late summer season (July 21 -August 10).In 2012, the reactivity of the water temperatures to air temperatures is evident throughout the low discharge recession and baseflow periods and indicates a strong relationship even during the late season when both air and water temperatures begin to cool (Figure 9).The early season resilience is associated with low river temperatures (~0-2 °C) with equal to or greater air temperature conditions for 2007 (Figure 9A) and 2012 (Figure 9C), and includes high discharge during snowmelt.

Hydrological and environmental controls over water temperature in High Arctic rivers
Streamflow in High Arctic settings is typically sustained by spring snowmelt, sporadic rainfall and the melting of semi-permanent snowbanks, ground ice and where present, glaciers (Woo 2012).These contributions often cause river discharge to be highly seasonal, with a welldefined snowmelt period causing a peak flow at the beginning of the summer season, a subsequent recession in discharge, and baseflow during the mid-to late summer season when contributions from snowmelt become limited.Results from Cape Bounty indicate that water temperature is also highly seasonal, and reflects both hydrological sources and seasonal D r a f t weather.During spring snowmelt, water sources are dominated by cold water inputs and generate consistent low temperatures that are resilient to atmospheric warming, even in the warmest melt seasons such as 2012 (Figure 8).During this phase of runoff, water temperature is nearly constant at ~0 °C, while air temperature typically shows clear diel cycling.The exhaustion of catchment snow results in flow recession and warming water temperatures, marking a transition to a more reactive water thermal regime.This reactive thermal regime appears to be highly responsive to both diel solar inputs and overall air temperature conditions and can result in sharp rises in water temperature in warm mid-summer periods during the month of July.This pattern is similar to lower latitude river systems, including lowland natural rivers in the maritime temperate region (Moore 2006;Leach and Moore 2014) and snow fed rivers in the mountainous alpine region of Switzerland (Uehlinger et al. 2003;Lisi et al. 2015;Toffolon and Piccolroaz 2015;Piccolroaz et al. 2016).Increased thermal reactivity is evident during the warmest years at CBAWO such as 2007 and 2012, when water temperature in both rivers reached temperatures in excess of 16 °C, while muted July water temperatures occurs during cooler summers such as 2004 and 2014 (Figure 6).The shift from a resilient to reactive thermal regime (Figure 8) was noted in all years and in both rivers, and is consistent with the seasonal progression of water temperature observed in other Arctic systems (Lammers et al. 2007;Blaen et al. 2013;Lisi et al. 2015).
The High Arctic water temperature sensitivity to summer climate evident in this study also reflects the nature of the landscape and structure of vegetative cover.While vegetation and topography provides varying degrees of solar shading to river systems in boreal, temperate and alpine regions (Johnson and Jones 2000;Mellina et al. 2002;Moore 2006;Webb et   Both river and air temperatures begin to cool in late July in response to the seasonal decrease in solar insolation intensity due to progressively lower sun angles.Hence, while the rivers generally appear to remain reactive to air temperature and share similar atmospheric forcing, the decrease in available energy from the climate system is reflected in a reduction in water heating and a progressive loss of heat in the rivers.In absence of a water energy balance including radiative fluxes, we interpret the late season water cooling as resulting from low discharge that further reduces thermal mass of surface water and enhances heat exchanges.
Thus river water temperature responses to energy inputs are highly responsive to short term energy balance conditions and seasonal cooling after mid-summer.
Results in this study are limited to a single downstream measurement location and preclude assessment of upstream thermal variability.However, previous work at CBAWO indicates that landscape morphology and hillslope drainage patterns affect the routing of melting snow, ice and precipitation to the rivers and influence river temperature due to temperature differentials between the river and these external water sources (Bolduc et al. 2018).For example, rainfall runoff from tributaries was noted to dampen the diel river temperature range and this pattern was evident downstream as well (Figure 9b, July 21-August 10).

D r a f t
Residual channel snow has also been shown to influence river temperatures in these rivers, and provides cold inputs of meltwater, cooling and dampening overall river temperatures in the channel (Bolduc et al. 2018).Snow accumulates preferentially in areas with incised channels and in the lee of hillslopes aligned relative to the persistent northwest wind direction, resulting in localized winter accumulations in excess of 5 m (Bonnaventure et al. 2016).In these locations, residual channel snow can remain until late summer, and in some years may remain multiple years at CBAWO (Lamoureux et al. 2006).Where present in the channel, residual snowmelt cools river water by up to 4 °C, particularly during baseflow (Bolduc et al. 2018).
At a finer spatial scale, surface tributaries generally contribute relatively warm water to the rivers, while subsurface inflows appear to contribute relatively cool water (Bolduc et al. 2018).Subsurface inflows are consistently cool due to the presence of a frost table at the base of the active layer throughout the melt season in this permafrost setting.Longitudinal temperature profiling of the West River during the 2014 season indicates that subsurface inflows are highly localized and contribute to downstream water cooling at both local scales (up to 20 m) and reach scales (c.100+ m).The volumes of subsurface inflows remain unknown in the High Arctic, but the thermal effects of this water resulted in river reaches where downstream temperature gradients were negative (-0.2 to -0.5 °C/1000 m; Bolduc et al. 2018).
Hyporheic exchanges have been found to occur within lower Arctic settings, where hyporheic (flow in the channel bed material) processes tend to be influenced by discharge, channel morphology and riverbed permeability (Zarnetske et al. 2007;Westhoff et al. 2011).
These exchanges have been found to cool or dampen river temperature variations at hourly D r a f t and seasonal timescales (Zarnetske et al. 2007;Westhoff et al. 2011).The presence of hyporheic exchanges has not been determined for the rivers at CBAWO, however, the potential for these exchanges to influence river temperatures in the High Arctic exists.Notably, hyporheic and subsurface slope inflows appear to be most developed during the late thaw season and coincide with low discharge in the river.Hence, the thermal effects by these cool water exchanges may contribute to the observed overall late-season cooling of river water temperatures and warrant further investigation.

Climate variability and water temperature
In this study we observe that changes in water temperatures closely follow changes in air temperatures under most conditions (Table 2).However, the lack of a consistent interannual relationship between air temperature and water temperature indicates that other factors such as residual snowmelt and precipitation likely also influence river temperatures (Lisi et al. 2015;Toffolon and Piccolroaz 2015;Piccolroaz et al. 2016).The West River channel was characterized by the presence of persistent residual snow c. 500 m upstream of the gauging station in all seasons prior 2009 which likely caused the thermal state of this river to be more thermally-resilient during the mid-to late season compared to the East River during the same seasons.The presence of these cold water contributions prior to 2009 would explain why the West River has a lower temperature maximum and relatively stable mid-range temperatures (Figure 6).Diminished amounts of year over year residual snow were observed in the West River channel, especially after 2009 and likely due to very warm conditions during 2007 and D r a f t 2008.This diminished the resiliency effect of residual channel snow and is evident in thermal response of the West River exhibiting greater resiliency before 2008, a mixed response in 2009, and then similar thermal response patterns to the East River after 2009 (Figure 8).In 2010 and subsequent seasons, the timing of the shift in thermal state from early season resilience (June 7 -June 30) to reactive (July 1 -August 10) of the West and East river is similar, suggesting that the thermal effect of residual channel snow has been largely diminished.Therefore, this multiyear regime change of diminished residual channel snow has caused the West River to be much more susceptible to mid-to late season warm air temperatures, especially during the warmest year in 2012 when maximum river temperature was higher in terms of both duration and intensity.
This multi-year shift from mid-season thermal resiliency to reactivity is further evident when comparing intra-seasonal air temperature and discharge during the two warmest years (2007 and 2012) for the West River (Figure 9).The overall seasonal negative correlation between river temperatures and discharge for the West River in 2012 is due to the thermal influence of cold snowmelt discharge and strong reactive heating in the mid-to late season (Figure 8).By contrast, the thermal response in 2007 is more variable with several instances where rainfall discharge revitalizes discharge in the early, middle and late season and dampens river temperature response to higher air temperatures (Figure 9).Hence, such contributions dampen river temperatures relative to air temperatures, causing the river to be more resilient to mid-to late-season warming.

Implications for climate change and aquatic ecosystems
The multi-year record in this study (2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016) covers a wide range of climatic conditions, including the coldest ( 2004) to the warmest summers (2012) on record since regional observations began in 1949 (Lamoureux and Lafrenière 2017).For this reason, the river temperature data in this study are well-suited to consider the impact of projected climate changes in this region.It has been predicted that the influences of a warming climate (climate change) will not only affect river temperatures, but also influence channel snow and ice conditions (Callaghan et al. 2011;Nilsson et al. 2015).Due to higher predicted air temperatures in high latitude regions (Frey and McClelland 2009;IPCC 2014;Nilsson et al. 2015), the seasonality of snowmelt is predicted to change.The timing of initial peak flow from snowmelt occurs up to three weeks earlier during the warmest years at CBAWO (2007 and 2012, early June), while colder years such as 2004 and 2009 result in later initial flow (mid to late June).
The timing of snowmelt is a critical determinant of when recession begins and water temperatures can begin to rise above ~0 °C, especially given that solar irradiance intensity is at a potential maximum in late June.A shift towards earlier discharge recession would likely increase the potential for warmer water.Projections also suggest that the duration of the summer season will also become longer (Rouse et al. 1997;Nijssen et al. 2001;Lewis and Lamoureux 2010;IPCC 2014), although the impacts on water temperature arising from an extension of the runoff period are poorly understood.It is likely that diminishing solar energy availability and cooler autumn temperatures would constrain water warming under these circumstances.

D r a f t
Fish and other aquatic organisms have been known to be particularly sensitive to water temperatures (Baroudy and Elliott 1994;Rouse el al. 1997;Boulton et al. 1998;Thyrel et al. 1999;Larsson et al. 2005;Reist et al. 2006;Przytulska et al. 2015).The effects that climate change and changing water temperatures have on aquatic species depends on the response of the aquatic system to external climate forcing as well as the resiliency or sensitivity of organisms to changing thermal conditions.For example, fish and other species accustomed to cold water settings such as the High Arctic have been known to be negatively or lethally impacted when subjected to water temperature conditions greater than 21 °C (Baroudy and Elliott 1994;Thyrel et al. 1999;Larsson et al. 2005;Przytulska et al. 2015).The presence of Arctic char has been commonly observed in the West River, especially in late season (July, early August, unpublished data S.F.Lamoureux) and are present in both the main downstream lakes and the headwater lake in the West River (Roberts et al., 2017).Based on river temperature data from CBAWO, temperatures above 21 °C can be considered as an extreme event associated with record warm air temperatures.As such, organisms such as Arctic char which have been acclimatized to colder temperatures or which have a harder time to adapt to rising temperatures will potentially be forced to relocate to areas of thermal refuge in river systems or encounter additional stress or higher mortality rates (Rouse et al. 1997).
Research carried out on 11 char species (including Arctic char) in Norway, Britain and Sweden in order to determine the impacts of warming water temperatures reveal that there is minimal evidence for thermal adaptation to changing conditions between species (Larsson et al. 2005).Short term exposure to stable mid-range water temperatures observed during mid-to late-summer (10-18 °C) in High Arctic rivers would provide greater opportunities of growth for D r a f t aquatic organisms by boosting their metabolism, feeding rate, and growth rate (Baroudy and Elliott 1994;Rouse el al. 1997;Thyrel et al. 1999;Larsson et al. 2005), provided that food sources are not limited by other factors or multi-stressor effects.
In cases of extreme conditions where river temperatures exceed a critical threshold, aquatic wildlife such as Arctic char acclimatized to cold conditions can temporarily tolerate temperatures between 17 to 21 °C for seven days (Larsson et al. 2005).However, when subjected to temperatures ranging from 22 to 25+ °C (lethal temperatures) fish-char species specifically-cannot even tolerate 10 minutes of these conditions (Baroudy and Elliott 1994;Larsson et al. 2005).Analysis of river temperature at this site has demonstrated that river temperature close to or at lethal temperatures is achievable for both rivers during very warm years.Given the strong correlation between river and air temperatures during mid-to late season, predicted climate warming in the Arctic (IPCC 2014) is likely to increase the frequency of potentially lethal season water heating events.While results from previous modelling research at CBAWO suggest that projected warming air temperatures would extend the melt and runoff season into the autumn (Lewis and Lamoureux 2010), results here indicate the potential for stress for aquatic organisms is most likely to be during the month of July, where water temperatures are at their highest, unless significant warming occurs during the month of August.Given the declining solar angle and intensity during August, and dark conditions emerging after mid-August, it is likely that sustained air and water warming will be limited.
The thermal sensitivity of Arctic fish species and the occurrence of high water temperatures in this study highlights the importance of residual channel snow and other subsurface water sources which act to dampen water temperature during July.The decline of D r a f t residual snow influence in the West River observed during this study period indicates that climate warming will likely diminish the influence of residual channel snow during the period of greatest water heating potential.Due to the reduction of channel snow influence, other factors such as precipitation inducing subsurface inflows and thawing ground ice could contribute new inflow pathways of cool water (Bolduc et al. 2018), may result in alternative thermal refugia for aquatic organisms.These results suggest the need to investigate aquatic-thermal interactions in the High Arctic to improve predictive capacity and to identify species at risk.

Conclusion
This research provides insights into long-term river temperature trends and the thermal state of small rivers in the Canadian High Arctic.Further meteorological changes in high latitude regions will influence the probability for water thermal exceedances and the seasonal timing of river temperatures for these two river systems and other similar aquatic systems.Results indicate that air temperature records can be used to identify the potential for future changes in temperature trends in such river systems with consideration for catchment-specific factors such as summer residual channel snowpack.During the period 2004-2016, the duration of warm water temperature extremes is short-lived under current climate conditions.Advancing the understanding of long term river temperature trends in Arctic settings will be useful to quantify and assess the projected changes in aquatic systems for ecosystem and water quality management, and to identify possible aquatic species at risk.2006 2007 2008 2009 2010 2012 2014 2016 2004 2006 2007 2008 2009 2010 2012

D
While meteorological and hydrological data collection has been undertaken at Cape Bounty since 2003, river temperature data are only available after 2004.River measurements are available for 2003-2010 and in alternate years in 2012, 2014, and 2016.Additionally, discharge measurements ceased in late June 2005 during snow melt recession.There have been replacements or upgrades to instrumentation throughout this time period which are indicated in Table 1.Meteorological data for this study was collected each year from the weather station WestMet which was established in 2003 (Figure 1).Initially, air temperature at this weather station was collected at 10 minute intervals with an Onset H8 logger (0.2 °C accuracy) from 2003-2006, and on an hourly basis with an Onset UA-003 logger (0.1 °C accuracy) from 2006-2016.Hourly precipitation was measured with a Davis industrial tipping bucket gauge from 2003-2016 (0.2 mm resolution) and is reported as uncorrected for wind under catch.Both air temperatures and precipitation are measured 1.5 m above ground.River temperatures were measured with an Omega OM-CP 101 water level and temperature data logger in 2004 to 2006 (±0.1 °C) and subsequently an Onset U20 water level and temperature data logger (±0.1 °C) until 2016.Water temperature measurements were taken at 10 minute intervals throughout the study period.Instruments were carefully shielded to prevent direct exposure to solar radiation and lowered as necessary to maintain submersion during measurement periods.Sensors were cleared of floating debris on a daily basis.
vegetation and upland shading are not a factor in this High Arctic setting as there is effectively no vegetation canopy and topography is muted.Despite the continuous sunlight during the summer season, discharge and river temperatures still show strong diel cycles in this Arctic setting due to variations in solar insolation due to the low sun angle during the evening.

Figure 1 :
Figure 1: The Cape Bounty Arctic Watershed Observatory (CBAWO) study site representing the East (11.6 km 2 ) and West (8.1 km 2 ) watersheds.Contour intervals are in meters above sea level.The blue circle within the inset map shows the location of CBAWO in the Canadian High Arctic.Sources: Inset map drawn from Google Map data.Main map modified from Lamoureux and Lafrenière, 2017.

Figure 2 :
Figure 2: Examples of daily thermal response for the East River during the year 2012, classified into (a) resilient, (b) mixed and (c) reactive states when comparing water temperature vs air temperature (Tw/Ta = slope).

Figure 3 :
Figure 3: Annual record of mean air temperatures, total precipitation, and river temperatures for the East and West Rivers in the Summer season (June 7 -August 10) from 2004-2016 located at Cape Bounty, Melville Island.

Figure 4 :
Figure 4: A comparison of air temperatures and West and East River temperatures between (a) 2009 and (c) 2012.Additionally, contrasting discharge for the West and East rivers along with summer precipitation is shown for (b) 2009 and (d) 2012.

Figure 5 :
Figure 5: Mean daily river temperatures for the (a) East and (b) West rivers between the years 2004-2016.

Figure 6 :
Figure 6: Maximum daily river temperatures for the (a) East and (b) West rivers between the years 2004-2016.

Figure 7 :
Figure 7: Comparison of exceedance probabilities for the (a) East River and (b) West River temperatures between 2004 and 2016.Temperature data is based on hourly datasets for each season.

Figure 8 :
Figure 8: Comparison of the mean daily thermal response (Tw/Ta) between river temperatures vs air temperatures (Tw/Ta) for the (a) West River and (b) East River.Each day is calculated within a hydrological day (0500 h to 0500 h).

Figure 9 :
Figure 9: Comparison of mean river temperature (°C) and air temperature (°C) for the West River during (a) 2007 and (b) 2012.Mean river temperature (°C) and mean river discharge The Cape Bounty Arctic Watershed Observatory (CBAWO) study site representing the East (11.6 km2) and West (8.1 km2) watersheds.Contour intervals are in meters above sea level.The blue circle within the inset map shows the location of CBAWO in the Canadian High Arctic.Sources: Inset map drawn from Google Map data.Main map modified from Lamoureux and Lafrenière, Examples of daily thermal response for the East River during the year 2012, classified into (a) resilient, (b) mixed and (c) reactive states when comparing water temperature vs air temperature (Tw/Ta = slope)Annual record of mean air temperatures, total precipitation, and river temperatures for the East and West Rivers in the Summer season (June 7 -August 10) from 2004-2016 located at Cape Bounty, Melville Island.air temperatures and West and East River temperatures between (a) 2009 and (c) 2012.Additionally, contrasting discharge for the West and East rivers along with summer precipitation is shown for (b) 2009 and (d) 2012.Comparison of exceedance probabilities for the (a) East River and (b) West River temperatures between 2004 and 2016.Temperature data is based on hourly datasets for each season.
Comparison of mean river temperature (°C) and air temperature (°C) for the West River during (a) 2007 and (b) 2012.Mean river temperature (°C) and mean river discharge (m3/s) was also compared for the West River during (c) 2007 and (d) 2012.Each point represents mean temperature and discharge for one day.296x209mm (300 x 300 DPI) Page 39 of 39 https://mc06.manuscriptcentral.com/asopen-pubsArctic Science

Table and figure captions:Table 1 :
Summary of instruments for meteorological and hydrological data collection used in this study at Cape Bounty from years 2003 to 2016.

Table 2 :
Multiple regression analysis between river temperatures (dependent variable), discharge levels and air temperature (independent variables) to identify r, r 2 , n and p values.Analysis was performed for each year between the time periods of 2004 to 2016.All p values are < 0.001.Rivers are indicated by WR (West) and ER (East).Multiple regression equations are provided in Supplementary tableS1.