An Investigation of Snow and Ice Phenology in the Arctic from 1997 – 2019

Alicia Loreta Adele Dauginis

A thesis submitted in conformity with the requirements for the degree of Master of Science Department of Geography and Planning University of Toronto

An Investigation of Surface Snow and Ice Cover in the Arctic from 1997 – 2019

Alicia Loreta Adele Dauginis

Master of Science

Department of Geography and Planning University of Toronto

2020 Abstract

This study investigates the use of finer-resolution satellite observations for monitoring snow and ice phenology in the Arctic. The primary data used were the Interactive Multisensor Snow and Ice

Mapping System (IMS) snow and ice products from 1997 – 2019. Snow and ice phenology were examined by detecting the first and last dates of snow and ice on and off. The largest trends in earlier ice-off dates and later ice-on dates were detected in the Alaska/Russia region, while earlier snow-onset trends were detected across Canada and Eurasia. Sea ice freeze in the Canadian Arctic is shifting earlier, while freeze onset across Eurasia continues to become later, likely due to strong warming across the region. In the Canadian Arctic, significant correlations were identified between snow and ice on/off parameters, and at the pan-Arctic scale, lake ice phenology parameters showed significant correlations to snow and sea ice parameters during the melt season.

Acknowledgments

I would like to thank my supervisor Dr. Laura Brown for her leadership and guidance over the past three years (from undergraduate to Master’s) at the University of Toronto Mississauga. Dr. Brown has provided me with multiple opportunities to attend conferences, involvement in numerous projects related to my research, and experience doing field work related to snow and ice. Dr. Brown has been an excellent mentor and supervisor over the past three years, and without her guidance and knowledge this work would have not been possible. I would also like to thank the University of Toronto Mississauga Graduate Expansion Fund for providing me with funding to attend conferences and workshops related to my research over the past two years. Additionally, I would like to thank current and recent members of the Cryosphere Research Lab at the University of Toronto Mississauga including Alexis Robinson and Justin Murfitt. I would also like to thank Stephen Howell and Michael Brady from Environment Canada for their valuable input and technical support.

iii

Table of Contents

Acknowledgments...... iii

Table of Contents ...... iv

List of Tables ...... vi

List of Figures ...... vii

List of Appendices ...... x

Chapter 1 ...... 1

General Introduction ...... 1

1.1 Motivation ...... 1

1.2 Significance...... 2

1.3 Literature Review...... 3

1.3.1 Arctic Amplification ...... 3

1.3.2 Observed Changes in Snow Cover ...... 4

1.3.3 Observed Changes in Sea Ice ...... 7

1.3.4 Observed Changes in Lake Ice ...... 9

1.3.5 Remote Sensing of Snow ...... 10

1.3.6 Remote Sensing of Sea Ice ...... 13

1.3.7 Remote Sensing of Lake Ice ...... 15

1.3.8 IMS ...... 18

1.4 Objectives and Thesis Structure...... 20

1.5 References ...... 22

Chapter 2 ...... 39

Sea ice and Snow Phenology in the Canadian Arctic Archipelago from 1997 – 2018 ...... 39

2.1 Introduction ...... 39

2.2 Study Area ...... 42

2.3 Data and Methodology ...... 44

2.3.1 Data ...... 44

2.3.2 Methodology ...... 45

2.4 Results and Discussion ...... 47

2.4.1 Temporal Variability and Links Between Sea Ice and Snow Phenology ...... 47

2.4.2 Regional Variability and Links in Sea Ice and Snow Phenology ...... 62

2.5 Conclusion ...... 68

2.6 References ...... 71

Chapter 3 ...... 82

Pan-Arctic Sea Ice, Lake Ice, and Snow Phenology from 1997 – 2019 ...... 82

3.1 Introduction ...... 82

3.2 Data and Methodology ...... 86

3.2.1 Study Regions ...... 86

3.2.2 Data ...... 87

3.2.3 Methodology ...... 88

3.3 Results and Discussion ...... 90

3.3.1 Trends and Correlations ...... 90

3.3.2 Regional Variability ...... 96

3.4 Conclusions ...... 106

3.5 References ...... 108

Chapter 4 ...... 117

Conclusions ...... 117

References ...... 120

Appendices ...... 121

List of Tables

Table 2.1. Sea ice and snow phenology parameters and definitions used in this study...... 46

Table 2.2. Spearman rank correlation coefficient (ρ) for 24 km and 4 km IMS products (2004 - 2018). Bold indicates significance at 95%...... 46

Table 2.3. Earliest and latest sea ice and snow phenology years and respective dates (day of year) detected by the 24 and 4 km IMS products. ** indicates that the 4 km product identified the same year as the 24 km product from 2004 – 2018...... 49

Table 2.4. Spearman rank correlations (ρ) for snow and sea ice phenology dates, 2-m temperature (T), and open water and snow-free duration from 1997 - 2018. Bold indicates significance at 95%...... 50

Table 3.1. Sea ice, lake ice, and snow phenology parameters and definitions in this study...... 88

Table 3.2. Pan-Arctic Spearman rank correlations (ρ) for snow and ice phenology dates using the 24 km (1997 – 2019) and 4 km (2004 – 2019) IMS products. * represents statistically significant correlations at the 95% confidence level...... 93

Table 3.3. Pan-Arctic Spearman rank correlations (ρ) for snow and sea ice phenology dates and monthly 2-m temperature from 1997 - 2019 using 24 km IMS. Note that correlations from August - December are from 1997 – 2018. Months were selected for each phenology parameter based on mean phenology dates in Figure 3. * represents statistically significant correlations at the 95% confidence level...... 94

Table 3.4. Regional Spearman rank correlations (ρ) for snow and ice phenology dates and monthly ERA-Interim 2-m temperature from 2004 - 2019 using 4 km IMS. For sea ice, ‘Canadian Arctic’ includes Baffin Bay, Hudson Bay, and the CAA; ‘Alaska/Russia’ includes Beaufort, Chukchi, and Bering seas; ‘Eurasian Arctic’ includes East Siberian, Laptev, Kara, Bering, and Greenland seas. Note that correlations from August – December are from 2004 – 2018. Months were selected for each phenology parameter based on mean phenology dates in Figure 3. * represents statistically significant correlations at the 95% confidence level...... 100

List of Figures

Figure 2.1. Map of the Canadian Arctic Archipelago study area and specific locations referred to throughout. Base map created using ESRI ArcMap version 10.6 and Natural Resources Canada Digital Elevation Model (Natural Resources Canada 2002, updated 2019), and Statistics Canada Boundary Files (Statistics Canada, 2011)...... 42

Figure 2.2. Mean 24 km (1997 – 2018) solid line and 4 km (2004 – 2018) dotted line IMS first open water and continuous open water (a), freeze onset and continuous ice cover (b), first snow- off and final snow-off (c), and first snow-on and final snow-on dates (d) for the Canadian Arctic Archipelago. Sen’s slope and significance are indicated for the 24 km product. IMS data were downloaded from the U.S. National Ice Center (2008)...... 49

Figure 2.3. Mean ERA-Interim daily 2-metre temperature anomalies for June, July, August, September, and October, 1997 – 2018 for the Canadian Arctic Archipelago. ERA-Interim data were downloaded from ECMWF (2011)...... 51

Figure 2.4. Sea ice and open water conditions in the Canadian Arctic Archipelago during September 2014 on September 1, 6, 20, and 27 from the 4 km IMS product. Base map created using ESRI ArcMap version 10.6 and Natural Earth land vectors (Natural Earth 2020). IMS data were downloaded from the U.S. National Ice Center (2008)...... 53

Figure 2.5. Final freeze timing for 2006 for 24 km (left) and 4 km (right) highlighting the effect of the improved resolution on sea ice detection. White areas in the northwest portion of the study area indicate regions where no ice-off or -on date was detected that year. Base map created using ESRI ArcMap version 10.6 and Natural Earth land vectors (Natural Earth 2020). IMS data were downloaded from the U.S. National Ice Center (2008)...... 54

Figure 2.6. Duration from first snow-off to final snow-off for (left) 2011 and (right) 2012 using the 4 km IMS product. White regions indicate either no snow free conditions detected or lakes are present. Base map created using ESRI ArcMap version 10.6 and Natural Earth land vectors (Natural Earth 2020). IMS data were downloaded from the U.S. National Ice Center (2008). ... 56

vii

Figure 2.7. a) Daily snowfall and mean air temperature at Eureka, Resolute, and Cambridge Bay in August 2013 (locations indicated by symbols) and b) composite images for the first and second half of August 2013 for 2m Air temperature and Sea Level Pressure anomalies (1981 – 2010 climatology) from ERA-Interim. Temperature and snowfall data were obtained from Environment and Climate Change Canada Historical Data (2020) and ERA-Interim data were downloaded from ECMWF (2011). Base map created using ESRI ArcMap version 10.6 and Natural Earth land vectors (Natural Earth 2020)...... 58

Figure 2.8. a) Mean annual air temperature and total annual snowfall and b) Mean monthly air temperature and total snowfall for the month of September at Eureka, Resolute, and Cambridge Bay from 1997 – 2018. Temperature and snowfall data were obtained from the Environment Canada Second Generation of Homogenized Temperature (Vincent et al. 2012) and Adjusted Precipitation for Canada (Mekis and Vincent, 2011) datasets...... 59

Figure 2.9. Open water duration and snow free duration in the Canadian Arctic Archipelago from 1997 – 2018. Sen’s slope of the trend and significance are indicated for the 24 km IMS product. IMS data were downloaded from the U.S. National Ice Center (2008)...... 61

Figure 2.10. Trends in 4 km IMS (2004 – 2018) first open water and first snow-off (a), continuous open water and final snow-off (b), freeze onset and first snow-on (c), and continuous ice cover and final snow-on (d) dates in the Canadian Arctic Archipelago. White areas indicate regions where no ice or snow on or off trends were computed. Regions with significant clustering between sea ice and snow phenology trend at the 95% confidence level are also shown. Base map created using ESRI ArcMap version 10.6 and Natural Earth land vectors (Natural Earth 2020). IMS data were downloaded from the U.S. National Ice Center (2008). ... 63

Figure 2.11. Short-term trends in ERA-Interim 2-m temperature (2004 – 2018) during June (a), July (b), August (c), September (d), and October (e) in the Canadian Arctic Archipelago. Base map created using ESRI ArcMap version 10.6 and Natural Earth land vectors (Natural Earth 2020). ERA-Interim data were downloaded from ECMWF (2011)...... 64

Figure 2.12. 4 km IMS continuous open water dates for (left) 2004 (colder year) and (right) 2011 (warmer year) in the Canadian Arctic Archipelago. White regions represent no detected ice break

viii

up. Base map created using ESRI ArcMap version 10.6 and Natural Earth land vectors (Natural Earth 2020). IMS data were downloaded from the U.S. National Ice Center (2008)...... 65

Figure 3.1. Map of the study regions...... 86

Figure 3.2. Comparison of 24 km (left) and4 km (right) IMS lake ice first open water in 2017. 89

Figure 3.3. Mean 4 km IMS (2004 – 2019) sea ice first open water (FOWS), first snow-off

(first_soff), and lake ice first open water (FOWL) (a), sea ice water clear of ice (WCIS), final snow-off (final_soff), and lake ice water clear of ice (WCIL) (b), sea ice freeze onset (FOS), first snow-on (first_son), and lake ice freeze onset (FOL) (c), and sea ice continuous ice cover (CICS), final snow-on (final_son), and lake ice continuous ice cover (CICL) (d)...... 91

Figure 3.4. Mean 24 km (1997 – 2019) and 4 km (2004 – 2019) IMS sea ice first open water

(FOWS) and water clear of ice (WCIS) (a), first snow-off (first_soff) and final snow-off (final_soff)

(b), lake ice first open water (FOWL) and water clear of ice (WCIL) (c), sea ice freeze onset

(FOS) and continuous ice cover (CICS) (d), fist snow-on (first_son) and final snow-on (final_son)

(e), and lake ice freeze onset (FOL) and continuous ice cover (CICL) (f). Sen’s slope and significance are indicated for each phenology parameter. Note that for lake ice, only the 4 km IMS product was used in this study...... 92

Figure 3.5. Pan-Arctic open water duration for oceans (1997 – 2019), snow-free duration (1997 – 2019) over land, and open water duration for lakes (2004 – 2019). Sen’s slope of the trend and significance are shown...... 95

Figure 3.6. Trends in 4 km IMS (2004 – 2019) sea ice first open water (FOWS), first snow-off

Figure 3.7. Trends in monthly ERA-Interim 2-m temperature from 2004 – 2019 in January (a), February (b), March (c), April (d), May (d), June (e), July (f), August (g), September (h), October (i), November (j), and December (k). Note that data from August – December are from 2004 – 2018...... 99

List of Appendices

Appendix 1- IMS data sources...... 121

x 1

Chapter 1 General Introduction 1.1 Motivation

Surface snow and ice cover in the Arctic are useful indicators of local and regional climate variability and change (Lei et al. 2012; Petty et al. 2018). Changes in the duration and extent of snow and ice cover have direct feedbacks to the climate system as they strongly influence the surface albedo of the planet (Rahmstorf, 2010; Derksen et al. 2012). There are also important linkages between Arctic ecosystems and snow and ice cover, as several species (e.g. seals, polar bears) rely on snow and ice for feeding, transportation, and habitat (Dersken et al. 2012). Additionally, the traditional ways of life of many Northern residents depend on snow and ice cover for sources of food, transportation, and economic activities (Derksen et al. 2012). Recent assessments have shown that dramatic changes are taking place in the Arctic as a result of increasing temperatures (e.g. Hernandez-Henriquez et al. 2015; Johannessen et al. 2016). These changes include reductions in sea ice extent, decreases in snow cover duration, and earlier melt onset in Arctic and sub-Arctic lakes (Serreze and Stroeve, 2015; Surdu et al. 2016; Mudryk et al. 2018). Reductions in snow and ice cover are exacerbated by positive feedbacks associated with changes in albedo, therefore monitoring snow and ice cover is critical for assessing the impacts of climate variability and change.

Monitoring Arctic snow and ice cover is challenging due to limited in situ data and large gaps and biases in surface observing networks; therefore, studies on Arctic snow and ice cover largely rely on the use of satellite observations (Brown et al. 2010; Brown and Duguay, 2011). Arctic snow and ice cover are monitored primarily through satellite-based microwave data, as it provides information regardless of solar illumination and cloud cover (Brown et al. 2014). Microwave measurements have been used for estimating snow and ice melt and freeze onset (e.g. Howell et al. 2006; Markus et al. 2009; Wang et al. 2011) at various spatial resolutions ranging from 6.25 km to 25 km (Brown et al. 2014). Although microwave observations have been successfully applied in past snow and ice applications, these datasets suffer from inherent wavelength specific uncertainties and land contamination (Howell et al. 2006; Brown et al. 2014).

The Interactive Multisensor Snow and Ice Mapping System (IMS) provides an alternative approach to snow and ice monitoring, with daily imagery available at a higher spatial resolution (24 km, 4 km, 1 km) compared to conventional passive microwave datasets. IMS has mainly been used in snow cover monitoring (e.g. Brubaker et al. 2005; Chen et al. 2012; Yu et al. 2017) and lake ice monitoring (Brown and Duguay, 2012; Duguay et al. 2012, 2013, 2014, 2015; Duguay and Brown, 2018) yet has received little attention in sea ice applications despite being advantageous over automated algorithms for examining sea ice phenology (Brown et al. 2014). The higher spatial resolution of IMS can therefore improve monitoring capabilities of surface snow and ice cover across the Arctic and provide insight into how northern regions are responding to warming temperatures.

1.2 Significance

Multiple studies have successfully investigated changes in snow (e.g. Dery and Brown, 2007; Brown et al. 2010; Callaghan et al. 2011; Derksen and Brown, 2012; Najafi et al. 2016) and ice cover (e.g. Howell et al. 2006; Maslanik et al. 2007; Serreze et al. 2007; Stroeve et al. 2012; Brown et al. 2014) across the Arctic using satellite-based observations. Most analyses are generally conducted at the pan-Arctic or hemispheric scale using coarse resolution satellite observations, which can fail to capture information regarding regional variability across the Arctic. The Canadian Arctic Archipelago (CAA), in particular, is often omitted from such analyses, as passive microwave sensors are unable to resolve finer-scale changes in snow and ice cover in this region. Sea ice and snow cover have not been explored together in the CAA in previous literature in regards to the application of IMS. Previous work in the CAA is mainly focused on sea ice monitoring using synthetic aperture radar (SAR) data (e.g. Howell et al. 2008a, b; Howell et al. 2009b; Howell et al. 2013; Howell et al. 2015), and this thesis presents the first comparison of both sea ice, lake ice, and snow phenology together using IMS.

Recognizing the cryosphere as an integrated system in both regional and global scale climate regimes is critical in identifying evidence of responses to climate variability and change, as well as the relationships between components of the cryosphere (Derksen et al. 2012). To gain a better understanding of how other regions of the Arctic are responding to changes in temperature, this thesis also uses IMS to examine sea ice, lake ice, and snow phenology at the pan-Arctic scale. Traditional boundaries between subdisciplines, various datasets used to examine cryospheric

components, and challenges associated with conducting combined field campaigns that span multiple elements of the cryosphere have limited the analyses of multiple elements of the cryosphere (Derksen et al. 2012) as accomplished in this thesis. By examining multiple components of the cryosphere and utilizing finer resolution satellite observations, we can establish a better understanding of how snow and ice cover have responded to warming over recent years.

1.3 Literature Review

The following three sections present a literature review that ties together the warming taking place in the Arctic (1.3.1), with the observed changes in snow and ice conditions (1.3.2 – 1.3.4), followed by an examination of the remote sensing options used for monitoring these changes (1.3.5 – 1.3.8 and subsections).

1.3.1 Arctic Amplification

The past two decades have seen substantial advances in understanding Arctic Amplification, the process whereby temperature variability and trends in the Arctic are larger than those for the Northern Hemisphere or the globe as a whole (Serreze and Barry, 2011). This concept was formally recognized in 1896 by a Swedish scientist (Svante Arrhenius) who argued that changes in atmospheric carbon dioxide concentrations could alter the Earth’s surface temperatures, and that temperature changes would be larger in polar latitudes (Arrhenius, 1896; Serreze and Barry, 2011). Studies of the instrumental record, reconstructions of past climates from proxy records, and experiments with global climate models confirm that Arctic Amplification is a characteristic feature of the global climate system (e.g. Serreze et al. 2007; Serreze and Barry, 2011; Walsh et al. 2011; Ding et al. 2014; van Wijngaarden, 2015a).

Recent Arctic Amplification is best expressed during the fall and winter months, and is much weaker during spring and summer (Serreze et al. 2009; Serreze and Barry, 2011). Seasonal variations in Arctic warming are important as Arctic regions are dominated by seasonal snow and ice cover and are therefore sensitive to small temperature perturbations, however the spatial distribution of warming is not uniform across the Arctic (Serreze et al. 2009; Serreze and Barry, 2011; Walsh et al. 2011; Ding et al. 2014; van Wijngaarden, 2015b). Accelerated warming spanning northern Eurasia to western North America has been identified in multiple studies (e.g. Comiso, 2003; Lawrence et al. 2008; Serreze and Barry, 2011; van Wijngaarden, 2015a; van

Wijngaarden, 2015b). van Wijngaarden (2015a) reports that Siberia, Alaska, and western Canada have experienced greater warming than eastern Canada, Greenland, and northern Europe over the past century. Comiso (2003) reports similar findings, noting positive temperature anomalies in northern Canada, Alaska, and the Beaufort Sea and smaller anomalies in the eastern Bering Sea and parts of Russia. Serreze et al. (2009) and Screen and Simmonds (2010) highlight the strong relationship between sea ice decline and positive surface air temperature (SAT) anomalies over the Arctic Ocean. Staring in the late 1990s, SAT anomalies were seen to turn positive over the Arctic Ocean in autumn, growing in subsequent years (Serreze et al. 2009; Serreze and Barry, 2011). Development of the autumn warming pattern aligns with reductions in September sea ice extent, with the strongest warming found over the Arctic Ocean where sea ice loss is the greatest (Serreze et al. 2009; Screen and Simmonds, 2010; Serreze and Barry, 2011). Amplified warming at higher latitudes has also been linked to reductions in snow mass and extent (e.g. Najafi et al. 2016; Pullianien et al. 2020), as well as delayed ice-onset and earlier ice-off in high-Arctic lakes (Surdu et al. 2016).

1.3.2 Observed Changes in Snow Cover

Snow cover is a defining feature of terrestrial Arctic landscapes and plays a major role in the climate system, with coverage persisting for 8 - 10 months of the year (Dery and Brown, 2007; Derksen et al. 2009; Callaghan et al. 2011). The amount and timing of snow cover are closely linked to temperature and moisture regimes, which have been changing in the Arctic and have thus driven significant changes in the snow regime (Dery and Brown, 2007; Callaghan et al. 2011; Derksen et al. 2012; Najafi et al. 2016). The snow-albedo feedback invokes amplified warming in polar regions through decreased snow cover extent which in turn reduces surface albedo and increases surface temperatures (Dery and Brown, 2007). In response to amplified Arctic warming, decreases in snow cover extent, snow cover duration, and earlier melt onset dates have been well documented (Dery and Brown, 2007; Brown et al. 2010; Callaghan et al. 2011; Derksen and Brown, 2012; Najafi et al. 2016).

In recent decades, northern hemisphere snow cover has shown a dramatic negative trend (Stroeve et al. 2007; Derksen and Brown, 2012; Stroeve et al. 2012; Thackeray and Fletcher, 2016). Arctic snow cover extent has declined by 3.4% decade-1 in May and 15.2% decade-1 in June from 1981 – 2019 (Mudryk et al. 2019). Using a nearly 100-year time series (1922 – 2018), Mudryk et al.

(2020) show strong negative snow extent trends in the early winter and spring and negative snow mass trends from November through May in the Northern Hemisphere. Decreases in snow cover extent are most pronounced during the spring season and are consistent with observed rates of change in snow cover duration (Brown et al. 2010; Callaghan et al. 2011; Derksen et al. 2012). Snow cover duration anomalies have been shown to exhibit a negative trend from 1967 – 2007, with stronger negative trends observed during the spring (Derksen et al. 2012). In high latitude regions of Canada, decreasing snow cover fraction trends are consistent with reductions in annual snow cover duration (2 – 4 days per decade) across the pan-Arctic (Brown et al. 2017; Derksen et al. 2019). In 2019, earlier than average snowmelt over the eastern (Baffin Island, NU) and western (northwest Canada and Alaska) sectors of the North American Arctic combined to cause the 5th and 3rd lowest snow cover extent in May and June since 1967 (Mudryk et al. 2019).

Multiple studies have shown that warming across the Arctic is a key driver of the lengthening snow-free season and decreased snow cover extent (Brown et al. 2010; Derksen et al. 2012; Hernandez-Henriquez et al. 2015; Thackeray and Fletcher, 2016). The enhanced snow-albedo feedback over the northern hemisphere plays a significant role in the poleward retreat of spring snow cover extent, with the greatest potential impacts to the surface radiation budget at high latitudes (60 – 70ºN), where the largest snow cover extent declines have been observed (Hernandez-Henriquez et al. 2015; Thackeray and Fletcher, 2016). In the spring, positive snow- albedo feedbacks impart a stronger signal on snow cover duration, as high net radiation is coupled with thin snow cover, thus positive air temperature anomalies correlate to earlier snowmelt (Peng et al. 2013; Hernandez-Henriquez et al. 2015). With increasing temperatures, surface albedo decreases as surface snow cover area and duration decreases, which in turn decreases the surface albedo and induces greater absorption of solar radiation at the surface, amplifying the response to increasing temperatures (Peng et al. 2013). Surface net radiation changes driven by increasing air temperatures, decreasing snow surface albedos, and increasing atmospheric vapour pressures has decreased late spring and early summer snow cover across the Arctic (Derksen and Brown, 2012; Hernandez-Henriquez et al. 2015). Brown and Robinson (2011) report that air temperature could explain approximately 50% of the interannual variability of spring snow cover extent in the northern hemisphere. Similarly, Brown et al. (2010) report linear decreases in spring snow cover extent where variability is predominately correlated with air temperature. Though previous studies suggest that changes in snow cover and duration is a direct response to atmospheric warming, the

snow cover response is a complex interplay of atmospheric and surface processes (Callaghan et al. 2011; Peng et al. 2013).

Atmospheric warming affects precipitation mechanisms, which are influenced by surface fluxes, moisture advection cloud processes, stability and orographic lifting of air masses (Hernandez- Henriquez et al. 2015). Warmer atmospheric conditions carry more water vapour with efficient precipitation-generating mechanisms and are expected to increase winter precipitation in the northern hemisphere high latitude regions (Callaghan et al. 2011; Hernandez-Henriquez et al. 2015). The amount of snow accumulating on the surface is influenced by the amount, type, and timing of precipitation, as well as blowing snow transport, sublimation, and vegetation interception (Callaghan et al. 2011). Evolution of the high-latitude snowpack has the additional complexity of being strongly dependent on blowing snow processes, local scale terrain, and vegetation (Liston, 1999; King et al. 2008; Callaghan et al. 2011). In Arctic tundra regions, high winds, low temperatures, and low snowfall amounts produce shallow snowfall (~30 – 40 cm) with a wind- hardened surface layer overlying a less dense depth hoar (Derksen et al. 2009; Callaghan et al. 2011). In forested Arctic regions, snow is less dense than tundra regions and cover is more uniform as trees act as windbreaks and shade the snow from incoming solar radiation during the spring (McKay and Gray, 1981; Callaghan et al. 2011). Physiographic and climate factors influencing regional-scale snow cover include elevation, amount of vegetation cover, spatial distribution of freezing temperatures, and location of major cyclones that bring moisture into the Arctic (Liston, 2004; Callaghan et al. 2011). Air temperature and elevation exert strong influences on the distribution of snow cover duration, with marked east-west increases in snow cover in response to the modification of winter air masses over the cold, dry continental interiors (Callaghan et al. 2011). Highest snow accumulation in the Arctic occurs in coastal mountain regions and considerably more moisture reaches the western sector of the Eurasian Arctic than North American, where coastal mountains block moisture brought by westerly winds from the Pacific Ocean (Liston, 2004; Callaghan et al. 2011; Zhang et al. 2019). Cyclonic systems with a maritime origin (e.g. from the Pacific to northwestern North America) bring heavier precipitation than storms spawned in continental regions (e.g. from the eastern side of the Western Cordillera). Winds blowing from the dry continental interior (e.g. southeastern Siberia) result in low winter precipitation with limited snow accumulation.

1.3.3 Observed Changes in Sea Ice

As Arctic temperatures have been increasingly warmer over recent decades, the influence of temperature and associated feedbacks with sea ice changes have prompted investigation into the relationship between sea ice and temperature. The relationship between temperature and sea ice extent and melt onset have become a central focus of many Arctic based climate change studies (e.g. Comiso et al. 2008; Stroeve et al. 2008; Stroeve et al. 2012; Parkinson and Comiso, 2013; Serreze and Stroeve, 2015). Sea ice plays an important role in the surface energy budget by modulating the exchange of energy between the atmosphere-ocean interface, primarily through the ice-albedo feedback (Perovich et al. 2007; Stroeve et al. 2012; Brown et al. 2014). The ice-albedo feedback, similar to the snow-albedo feedback, occurs when air temperatures over sea ice increase and promote melt, resulting in a decreased surface albedo, driving increases in shortwave radiation absorption which in turn further increase surface temperatures and accelerate melt (Curry et al. 1995; Brown et al. 2014). Increased surface air temperatures and subsequent increases in ice- albedo feedback magnitude have been linked to decreasing sea ice extent, a transition from thick multi-year ice (MYI) to thinner seasonal first-year ice (FYI), and earlier melt onset dates (Howell et al. 2006; Maslanik et al. 2007; Serreze et al. 2007; Stroeve et al. 2012; Brown et al. 2014).

Since the start of the satellite record in 1979, Arctic sea ice extent has been exhibiting a decreasing trend, though empirical estimates and causes of sea ice decline remain under debate (Serreze et al. 2007; Stroeve et al. 2012; Parkinson and Comiso, 2013; Serreze and Stroeve, 2015). Numerous studies have attempted to quantify the rate sea ice extent decline using satellite-based passive microwave measurements, which are available continuously since 1979 (Johannessen et al. 2004; Stroeve et al. 2012; Serreze and Stroeve, 2015). These studies report varying rates of decline, with – 10%, – 10.7%, – 12.4% and – 13.3% per decade amongst some of the results reported, with variability attributed to differing spatial and temporal scales of each study (Stroeve et al. 2008; Tivy et al. 2011; Stroeve et al. 2012; Serreze and Stroeve, 2015; Mahmud et al. 2016). Minimum Arctic sea ice extents occur during September following the end of the summer melt season, with the record minimum observed in 2012 (3.4 million km2), and second lowest minimums recorded in 2007, 2016, and 2019 (~ 4.2 million km2) since the start of the satellite record (Stroeve et al. 2012; Parkinson and Comiso, 2013; Serreze and Stroeve, 2015; Simmonds, 2015; NSIDC, 2019a). The dramatic decline in September sea ice extent during 2007 occurred after years of shrinking and thinning of ice cover linked to both natural variability and external forcing, making the ice

increasingly vulnerable to anomalous atmospheric events (Maslanik et al. 2007; Stroeve et al. 2008; Stroeve et al. 2012). A number of studies have pointed out that a key driver of the 2007 September minimum was an atmospheric pattern featuring unusually high sea level pressure over the western Beaufort Sea and northern Canada, favouring northward ice drift, warmer temperatures, and strong melt (Maslanik et al. 2007; Stroeve et al. 2008; Stroeve et al. 2012; Serreze and Stroeve, 2015). Similar patterns were observed during the 2012 minimum, which was well below the 2007 extent, with a greater expanse of seasonal ice and patterns favouring melt contributing to a larger decay in sea ice (Parkinson and Comiso, 2013; Serreze and Stroeve, 2015). Arctic sea ice extent is expected to continue decreasing, eventually leading to an essentially seasonally ice-free Arctic Ocean (Serreze and Stroeve, 2015). However, given the shortness of the available sea ice extent time series (1979 – present), decreasing trends may not be sustained given the natural variability and anthropogenic forcing in the coupled ice-ocean-atmosphere system (Stroeve et al. 2012).

It is well recognized that recent Arctic sea ice decline has both natural and anthropogenic drivers, but their relative importance remains poorly understood (Min et al. 2008; Kay et al. 2011; Ding et al. 2017). Understanding the sources and magnitude of sea ice variability is important in determining the proportion of sea ice decline that can be attributed to natural and anthropogenic forcing; however, this is difficult given the short satellite observation period (Day et al. 2012; Stroeve et al. 2012). Earlier studies indicated that sea ice decline prior to the 1990s was driven in part by dominant patterns of wintertime atmospheric circulation over the high latitude northern hemisphere, referred to as the Arctic Oscillation (AO) or the North Atlantic Oscillation (NOA) (Hu et al. 2002; Deser and Teng, 2008; Ding et al. 2017). However, since the early 1990s, the apparent link between the NOA/AO and sea ice decline diminished as sea ice extent continued to decline despite a reversal in NOA/AO trends (Day et al. 2012; Ding et al. 2017). Other research demonstrates that observed and modelled Arctic sea ice loss cannot result from natural variability alone, and that sea ice decline is in part attributable to anthropogenic forcing (Min et al. 2008; Kay et al. 2011; Kirchmeier-Young et al. 2017). Gregory et al. (2002) found that simulations including only natural forcing exhibited no significant trends with modeled sea ice extent trends. Similarly, Notz and Marotzke (2012) showed that trends and extreme minima in Arctic sea ice extents are inconsistent with natural variability alone, and that observed trends are strongly correlated with atmospheric carbon dioxide concentrations, a main component of anthropogenic forcing.

1.3.4 Observed Changes in Lake Ice

Lakes comprise approximately 2% of the Earth’s land surface, with the majority of lakes located in the Northern Hemisphere (Brown and Duguay, 2010). In Arctic and sub-Arctic regions of North America, lakes cover up to 15 to 40% of the surface, depending on the location (Duguay et al. 2003; Brown and Duguay, 2010). Given the large spatial coverage, ice regimes of these lakes play important roles in biological, chemical, and physical processes of cold region freshwater (Duguay et al. 2015a). Their frequency and size influence the magnitude and timing of landscape-scale evaporative and sensible heat inputs to the atmosphere, and are important for regional climate and meteorology (Rouse et al. 2008b; Duguay et al. 2015a; Murfitt et al. 2018). Shallow lakes warm quickly during the spring and have high evaporation rates until they refreeze in the fall, whereas large lakes take substantially longer to warm, but stay ice-free (or partly ice-free) into the early winter (Rouse et al. 2005; Duguay et al. 2015a). The duration of lake ice cover controls the seasonal heat budget of lake systems, thus determining the magnitude and timing of evaporation (Rouse et al. 2008a; Duguay et al. 2015a). The presence (or absence) of lake ice cover during winter months also affects both regional climate and weather events (Rouse et al. 2008a; Brown and Duguay, 2010; Duguay et al. 2015a).

Climate-driven changes in lake ice phenology have occurred widely throughout the Northern Hemisphere (Sharma et al. 2019), with trends and variability largely related to changes in surface air temperatures (e.g. Duguay et al. 2015a; Smejkalova et al. 2016). Broad scale spatial patterns in these trends are also related to major atmospheric circulation patterns originating from the Pacific and Atlantic Oceans, such as the El Nino-La Nina/Southern Oscillation, Pacific North American pattern, Pacific Decadal Oscillation, and the North Atlantic Oscillation/Arctic Oscillation (Duguay et al. 2015a, b). Long-term trends from ground-based observations reveal increasingly later ice freeze-up and earlier ice break-up dates, which are closely related to trends in air temperature (Duguay et al. 2015a). Despite these overall trends, annual and spatial variability is evident. For example, in High Arctic lakes, freeze-up in 2013/2014 occurred earlier than the 2004 – 2013 average by approximately 1 – 3 weeks for many regions of the Arctic (Duguay et al. 2015b). This is in sharp contrast with the 2012/2013 freeze-up period for lakes in western Russia and southern Finland, where freeze-up occurred approximately 1 – 2 months later (Duguay et al. 2014, 2015b). Earlier ice break-up in 2014 by approximately 2 – 6 weeks occurred over much of Scandinavia, western Russia, and southwestern Alaska and Yukon (Duguay et al. 2015b).

Surdu et al. (2016) find that water clear of ice dates in the Canadian High Arctic are earlier (9 – 24 days for polar oases and 2 – 20 days earlier for polar deserts), and that some lakes may be transitioning from a perennial to seasonal ice regime. Similarly, lakes of the North Slope of Alaska show earlier ice break-up by 17.7 – 18.6 days and a decrease in ice cover duration by approximately 24 days (Surdu et al. 2014). Paquette et al. (2015) use RADARSAT 1 and 2 imagery combined with aerial photographs and field records to show rapid contraction of ice cover extent on Ward Hunt Lake (off the north coast of Ellesmere Island) after at least 50 years of relative stability. From 1953 – 2007, 3.5 to 4.3 m of perennial ice covered 65 – 85% of Ward Hunt Lake during the summer, however by 2012 there were 26 days of open water and a significant increase in melting degree days (Paquette et al. 2015). Mueller et al. (2009) report similar results for five lakes in the northern region of Ellesmere Island, noting a new regime of frequent summer ice loss beginning in 1998. Lehnherr et al. (2018) note that full ice-off has become more frequent in recent decades on Lake Hazen (northern Ellesmere Island), resulting in the progressive loss of multi-year ice cover. Over the 1985 – 2017 period, Lake Hazen was shown to be completely ice-free during summer months in all but seven years (1985 – 1987, 1992, 1996, 2005, and 2009) (Michelutti et al. 2020). Reductions in lake ice thickness have also been reported (Mueller et al. 2009; Derksen et al. 2012; Surdu et al. 2014; Surdu et al. 2016), however there are very few observations of thickness reported in the literature as space-based detection is limited to either very large lakes (due to the resolution limitations of passive microwave, e.g. Kang et al. 2012) or very small tundra lakes that freeze to bed (where active microwave can be used infer ice thickness, e.g. Surdu et al, 2014). Much of what is known about changing or projected changes to Arctic ice thickness is obtained through modelling (e.g. Brown and Duguay, 2011; Surdu et al. 2014).

1.3.5 Remote Sensing of Snow

Remotely sensed observations provide a means by which to monitor snow mass and extent at various spatial and temporal resolutions. Snow cover monitoring has largely relied on the use of optical and microwave datasets. Some of the main datasets used in optical remote sensing of snow include the Moderate Resolution Imaging Spectroradiometer (MODIS) snow products (daily, 8- day, and monthly) and Advanced Very High Resolution Radiometer (AVHRR). The MODIS snow product has shown higher accuracy compared to AVHRR, which often significantly underestimates snow-covered areas due to lower spatial resolution and less available spectral bands (Salomonson and Appel, 2004; Dietz et al. 2012). Compared to the daily MODIS snow

cover product, the 8-day product helps reduce issues of cloud interference, as pixels are only assigned as cloud-covered if they were obstructed for all eight days (Riggs and Hall, 2016; Young et al. 2018). MODIS snow products have been successfully applied in snow mapping applications (e.g. Brubaker et al. 2005; Dong and Menzel, 2016; Young et al. 2018), and are overall in good agreement with available satellite and ground-based snow observations (Parajka and Bloschl, 2008). Parajka and Bloschl (2006) found that the accuracy of the MODIS snow product over Austria was approximately 95% when compared with snow depth data from 754 weather stations. However, Parajka and Bloschl (2006) also highlighted that approximately 63% of the study region was cloud-covered, and cloud cover was even larger during winter months. The National Oceanic and Atmospheric (NOAA) Administration weekly snow cover extent charts have also been widely used in mapping Arctic snow cover (e.g. Dery and Brown, 2007; Brown et al. 2010; Hernandez- Henriquez et al. 2015). The charts were derived from manual interpretation of visible satellite imagery up to May 1999, where the charting method shifted to automated generation of the weekly charts following the introduction of the 24 km IMS product (Brown et al. 2010; Brown and Robinson, 2011). The NOAA weekly snow cover extent charts provide data spanning from late 1966 to present, making this product advantageous in examining long-term trends in climatological and hydrological applications; however, the ~190 km spatial resolution can fail to capture spatial variability compared to finer-resolution datasets (Brown et al. 2007). During summer months, the NOAA snow charts have been shown to overestimate snow cover extent in the Canadian Arctic, with this bias attributed to elevation effects and frequent cloud cover (Wang et al. 2005; Brown et al. 2007; Estilow et al. 2015). Brown et al. (2007) show that the NOAA dataset is unable to capture interannual snow cover variability over the central Canadian Arctic, and recommend caution be taken when using this product during summer months where snow cover is controlled by small regions with frequent cloud cover.

Alternative approaches in snow mapping include the use of microwave remotely sensed data, as they provide information regardless of cloud cover and polar darkness (Brown et al. 2014). Products commonly used in snow mapping include the Special Sensor Microwave/Imager (SSM/I) and Scanning Multichannel Microwave Radiometer (SMMR) (e.g. Cho et al. 2017), Advanced Microwave Scanning Radiometer for the Earth Observing System (AMSR-E), Advanced Microwave Scanning Radiometer 2 (AMSR-2) (e.g. Larue et al. 2018; Gao et al. 2019), and QuikSCAT (prior to its demise). Passive microwave sensors do not carry an illumination source,

and rather measure energy that is naturally emitted (NRCAN, 2015b). This emitted energy is related to temperature and moisture properties of the emitting surface or object (NRCAN, 2015a). Since microwave wavelengths are relatively long (ranging from 1 mm to 1 m), the available energy is quite small (compared to optical wavelengths), thus the field of view must be large enough to detect enough energy in order to record a signal; therefore, most passive microwave sensors are characterized by low spatial resolution (NRCAN, 2015a). Passive microwave data are well suited for snow cover monitoring due to the all-weather imaging capabilities, large swath width with frequent overpass times, and relatively long available time series (since 1979); however, the coarse spatial resolution (25 km) hinders their applications and reduces accuracy (Derksen et al. 2004; Gao et al. 2010; De Lannoy et al. 2012). There are well-documented uncertainties in using passive microwave measurements to retrieve snow water equivalent and snow cover extent due to differences in snow and surface cover properties (e.g. snow depth, snow grain size, topography, vegetation), which influence microwave emission and backscatter (Brown et al. 2010; Park et al. 2012; Tedesco et al. 2015). AMSR-E has mainly seen applications in mapping snow depth and snow water equivalent (e.g. Comiso et al. 2003; Derksen, 2008; Dai et al. 2012; Mortimer et al. 2020). In a recent study comparing multiple snow water equivalent products, Mortimer et al. (2020) found that AMSR-E snow water equivalent products showed stark differences in climatological snow water equivalent patterns compared to other products. The AMSR-E products exhibited low spatial and temporal correlations with snow water equivalent anomalies, and higher RMSE and bias compared to the seven other datasets (Mortimer et al. 2020).

Active microwave data are useful in snow applications as they are sensitive to many snow parameters (e.g. density, depth, grain size, liquid water content) (Shi, 2008). Active microwave sensors are generally divided into two categories: imaging and non-imaging, and they illuminate target surfaces or objects by providing their own source of microwave radiation to measure and record the reflected radiation (NRCAN, 2015a). The most common form of imaging active microwave sensors is Radio Detection and Ranging (radar), which transmits a microwave signal towards a target and detects the backscattered signal (NRCAN, 2015a). Target features are discriminated by the strength of the backscatter signal and the location of objects is determined based on the time lag between signal transmission and reflectance (NRCAN, 2015a). Non-imaging sensors include altimeters and scatterometers, and in most cases take measurements in one linear dimension, whereas imaging sensors use two-dimensional representations (NRCAN, 2015a).

Many spaceborne active microwave sensors have collected data in the C- and X-bands, which are not optimal for snow retrieval given the low sensitivity of backscattering coefficients to snow depth (Tedesco et al. 2015). The higher sensitivity of Ku-band data to snow depth has allowed for the development of algorithms to estimate snow depth over ice sheets and land (Tedesco et al. 2015). In an assessment of spring snow cover duration over the northern Canada, Brown et al. (2007) show that QuikSCAT scatterometer data (Ku-band) and IMS best captured spatial variability in snow cover duration compared to NOAA snow charts and passive microwave datasets. Compared to IMS, QuikSCAT was shown to overestimate snow cover duration over the Canadian Arctic Archipelago by ~10 days, which may be an effect of temporal variability in QuikSCAT measurements (Brown et al. 2007). Techniques using synthetic aperture radar (SAR) data have been widely used to estimate snow water equivalent (e.g. Rott et al. 2009; Zhu et al. 2018). SAR imaging can provide observations at spatial resolutions less than 100 m, which is a significant improvement compared to coarse passive microwave data (Thompson and Kelly, 2019). By exploiting frequency-dependent sensitivity to snow and underlying soil properties, multi-frequency SAR observations (e.g. X- and Ku-band) can be used to retrieve snow water equivalent (e.g. Zhu et al. 2018). Zhu et al. (2018) established a model for X- and Ku-band radar backscatter and a retrieval algorithm for snow water equivalent and applied these to SnowSAR data and ground measurements. Promising performance was demonstrated with a RMSE of snow water equivalent retrievals under 30 mm and a correlation coefficient above approximately 0.64 (Zhu et al. 2018).

1.3.6 Remote Sensing of Sea Ice

Remote sensing of sea ice largely relies on the use of passive (e.g. Spreen et al. 2008; Serreze and Stroeve, 2015) and active microwave remote sensing (e.g. Mahmud et al. 2016; Howell et al. 2019), since these sensors are able to penetrate the frequent cloud cover and can detect sea ice year-round. Widely used passive microwave datasets in sea ice monitoring include SMMR and SSM/I (e.g. Lynch et al. 2017; Crawford et al. 2018), AMSR-E, and AMSR-2. The most commonly used techniques to detect melt onset are based on high temporal resolution (1 day) passive microwave brightness temperatures that use thresholds to identify the shift from dry winter sea ice conditions to melt onset (Howell et al. 2019). Similar to monitoring snow cover, passive microwave brightness temperatures allow for long term analyses (since 1979), but only at a 25 km spatial resolution (Howell et al. 2019). The coarse spatial resolution makes it difficult to resolve

small leads and polynyas and leads to issues near coastal areas due to pixel-based land contamination (Howell et al. 2006; Brown et al. 2014). Johnson and Eicken (2016) note that strong brightness temperature contrasts across pixels can result in falsely high estimates of sea ice concentration, particularly during the summer when there is open water near coastal areas. Methods have been applied to the NASA Team sea ice concentration algorithms (e.g. Cavalieri et al. 1999) to reduce the effects of land contamination at grid points in the vicinity of land and open water, however these problems are not entirely removed (Johnson and Eicken, 2016; NSIDC, 2019b). Comparing the 4 km IMS and NASA Team products in coastal areas, Brown et al. (2014) show that IMS generally detected earlier open water dates compared to NASA Team and noted that some land contamination can remain in the NASA Team data, despite improvements to the algorithm. AMSR-E showed less of a difference than NASA Team data compared to IMS in coastal regions, presumably due to resolution differences between AMSR-E (8.9 km) and NASA Team data (25 km), though differences in the algorithms used to produce each product may also be a factor (Brown et al. 2014).

Active microwave algorithms, specifically applied to QuikSCAT, have been largely used to estimate sea ice melt and freeze onset (e.g. Howell et al. 2008a; Howell et al. 2010; Mortin et al. 2014). Compared to passive microwave measurements (25 km) the QuikSCAT data offer measurements at an improved spatial resolution (4.45 km), but the sensitivity to surface scattering can lead to uncertainties during transient weather events (Yu et al. 2009; Howell et al. 2010; Brown et al. 2014). Howell et al. (2010) developed and evaluated water clear of ice algorithms using QuikSCAT and AMSR-E measurements and found that the QuikSCAT algorithm was able to better represent the ice edge and ice clearing processes, however transient clearing events were best represented by AMSR-E. Compared to IMS, QuikSCAT showed more extensive regions near the coastline where open water is detected later by IMS, though Brown et al. (2014) note that reconstructing grid resolution to 4.45 km from 8 – 10 km (effective resolution) may result in land contamination. The late detection of open water by QuikSCAT may also be attributed to weather, as the transition from ice to open water is often influenced by wind, which ultimately produces backscatter values over water similar to land as a result of wind roughened surfaces (Brown et al. 2014). Active microwave algorithms using synthetic aperture radar (SAR) have also been applied in high resolution retrieval of sea ice melt and freeze events (e.g. Yackel et al. 2001; Kwok et al. 2003). SAR estimates provide the highest spatial resolution (i.e. 100 m) compared to other sea ice

products, however the moderate temporal resolution, narrow swath width, and limited image availability across the Arctic limits the application of SAR to smaller geographic regions (Brown et al. 2014; Howell et al. 2019). The RADARSAT Constellation Mission (managed by the Canadian Space Agency) launched in July 2019 is currently the primary Canadian SAR mission. RADARSAT data has been used extensively to monitor sea ice in the Canadian Arctic (e.g. Howell et al. 2015; Howell et al. 2019). Mahmud et al. (2016) successfully utilized RADARSAT-1 and RADARSAT-2 to detect melt onset in the northern CAA, however noted that representative spatiotemporal melt onset estimates were limited by the temporal resolution of the imagery. Improvements to the temporal resolution of RADARSAT imagery have been addressed with the launch of RCM, as the three-satellite configuration will allow for daily revisits of Canada as well as daily access to the Arctic up to four times per day (CSA, 2019). Howell et al. (2019) explore a multi-sensor approach by combining RADARSAT-2 and Sentinel-1 SAR imagery as a more robust approach for melt onset detection over sea ice. Multi-sensor melt onset estimates were compared to Advanced Scatterometer (ASCAT) and passive microwave melt onset estimates and revealed that the multi-sensor product is capable of detecting much more detail with respect to the spatial distribution of melt (Howell et al. 2019). Multi-sensor melt onset detection was shown to be in good agreement with passive microwave and ASCAT data in regions with homogenous sea ice cover, however the multi-sensor data showed noticeable improvements within narrow channels and regions with more heterogeneous sea ice (Howell et al. 2019).

1.3.7 Remote Sensing of Lake Ice

Optical remote sensing (e.g. Landsat, AVHRR, MODIS) can provide valuable information on lake ice cover due to the moderate spatial resolution of optical sensors and potential for long-term analysis. Landsat provides the potential for long-term analyses of lake ice trends at a 30 m spatial resolution, with multispectral images acquired every 16 days. Landsat imagery provides an advantage over AVHRR data, because although AVHRR provides more frequent temporal coverage, the areal extent of lakes may be smaller than the pixel resolution of AVHRR (Duguay et al. 2003). Nitze et al. (2017) were able to detect approximately 80,000 individual lakes (larger than 1 hectare) using Landsat in the northern regions of Alaska and Siberia, highlighting the advantage of Landsat’s 30 m spatial resolution. Despite the resolution advantages of this sensor, Landsat imagery is often used only as a validation or supplementary data source in lake ice studies, as the 16-day temporal resolution does not provide adequate temporal coverage to capture rapid

changes in ice cover (e.g. Cook and Bradley, 2010; Geldsetzer et al. 2010). The daily MODIS snow product has a wide variety of applications including validation and incorporation into hydrological modelling (e.g. Parajka and Bloschl, 2008; Brown et al. 2010; Brown and Duguay, 2012). MODIS surface reflectance products coupled with MODIS 8-day snow cover composites have been used for identifying lake ice-on and off dates (Mishra et al. 2011). Kropacek et al. (2013) used the 8-day MODIS snow product to examine lake ice phenology over the Tibetan Plateau and compared 8-day composites to Landsat and SAR imagery. Overall accuracy for open water during break-up and freeze-up was calculated using root mean square error of time differences in days, yielding a value of 9.6 days (Kropacek et al. 2013). These results are somewhat comparable to Brown and Duguay (2012), who find a mean absolute difference between the daily MODIS snow product and IMS of 9 days during the ice-on season, and 3 days during the ice-off season. Caution should be taken when using a product that is assimilated from imagery spanning multiple days (i.e. MODIS 8-day snow product), as this can introduce uncertainty and inaccuracy into phenology estimates. Optical imagery is limited to the spring and summer months in high-latitude regions as there is no source of illumination during late fall and winter due to polar darkness. Instead, microwave remote sensing provides an alternative approach with fewer restrictions (i.e. imaging under cloudy and dark conditions) to monitor lake ice (Surdu et al. 2014).

Previous studies have shown the utility of using brightness temperatures from the SMMR at 37 GHz (e.g. Barry and Maslanik, 1993) and SSM/I at 85 GHz (e.g. Walker et al. 2000; Schertzer et al. 2003) for examining lake ice phenology; however, identifying spatial variability in ice phenology using SMMR and SSM/I is difficult due to their coarse spatial resolution (Kang et al. 2012). Additionally, the 85 GHz channel is susceptible to considerable atmospheric interference, and the 25 km spatial resolution can result in large differences in water/land brightness temperatures (Cavalieri et al. 1999; Howell et al. 2009b). AMSR-E brightness temperatures have been used to determine ice phenology on two of the largest lakes in northern Canada (Kang et al. 2012). Kang et al. (2012) compared AMSR-E ice phenology dates with QuikSCAT and IMS and found that AMSR-E freeze onset dates were approximately one week earlier than QuikSCAT, and that AMSR-E and IMS show a difference of approximately one week on average, with AMSR-E detecting ice-on later. Differences between AMSR-E and QuikSCAT may be due to the sensitivity of AMSR-E to fractional ice cover within pixels, whereas QuikSCAT is more sensitive to changes in surface roughness over open water areas (Kang et al. 2012). Overall, the AMSR-E 18.7 GHz

channel provides a viable means for monitoring of ice phenology on large northern lakes using brightness temperature measurements (Kang et al. 2012; Kang et al. 2014). The ice phenology algorithm developed by Kang et al. (2012) may be applicable to other large lakes in the northern hemisphere and to historical records of SMMR and SMM/I data.

Synthetic aperture radar (SAR) systems (active microwave) generally have a much higher spatial resolution (1 m to about 100 m) compared to passive microwave data, and therefore have the potential to further improve our understanding of lake ice phenology (Duguay et al. 2015a). Active microwave data provide the potential for obtaining ice phenology over a wide range of lake sizes (Duguay et al. 2002; Howell et al. 2009a; Geldsetzer et al. 2010; Brown and Duguay, 2012). Active microwave sensors used in lake ice applications include QuikSCAT, European Remote Sensing 1 and 2 (ERS-1 and ERS-2), Envisat Advanced Synthetic Aperture Radar (ASAR), Sentinel-1, and RADARSAT-1/2. QuikSCAT has been used successfully to derive freeze onset, melt onset and ice-off dates on Great Bear Lake and Great Slave Lake in northern Canada (Howell et al. 2009a). The high temporal resolution of QuikSCAT allows for more precise timing of ice phenology estimates, which can be challenging to obtain over large lakes (Howell et al. 2009a). Unfortunately, QuikSCAT data are no longer available as its mission ended in November 2009. ERS-1/2 have been used to monitor ice formation, thickening, and freezing of shallow Arctic and sub-Arctic lakes in Alaska (Jeffries et al. 1994; Morris et al. 1995) and northern Manitoba (Duguay et al. 1999). A number of studies have used calibrated ERS-1 data and quantitative analysis of backscatter intensity to examine different periods of ice growth (e.g. Jeffries et al. 1994; Morris et al. 1995). RADARSAT-1 SAR imagery has also been used for monitoring ice growth and decay in shallow sub-Arctic (tundra and forest) lakes in northern Manitoba (Duguay et al. 2002). A limitation of these two SAR systems (i.e. ERS-1/2 and RADARSAT-1) is the insufficient temporal resolution, thus determining the precise dates of ice phenological changes is difficult (Howell et al. 2009a; Antonova et al. 2016). C-band radar (3.75 – 7.5 cm) is the most common band used to examine changes in lake ice cover for high-latitude regions using SAR imagery (e.g. Mueller et al. 2009; Cook and Bradley, 2010; Surdu et al. 2014; Surdu et al. 2015; Surdu et al. 2016). In C- band SAR, the high contrast between ice and open water, representing the amount of backscatter returned to the sensor, allows for the detection of summer ice minimum and water clear of ice timing (Morris et al. 1995; Duguay et al. 2002; Geldsetzer et al. 2010; Surdu et al. 2016). Robust determination of the timing of freeze up using SAR is limited by the low backscatter contrast

between newly formed floating ice and open water (Cook and Bradley, 2010; Surdu et al. 2016). Additionally, the C-band backscatter from open water is sensitive to wind speed and direction, as well as the local radar incidence angle and polarization combination (Geldsetzer and Van Der Sanden, 2013; Duguay et al. 2002; Surdu et al. 2015; Surdu et al. 2016).

1.3.8 IMS

As previously discussed, multiple satellite-based data are available for monitoring Arctic snow and ice cover at various spatial and temporal resolutions. Optical sensors provide both moderate spatial and temporal resolutions, however their utility is limited by frequent cloud cover and polar darkness in high latitude regions. Alternatively, microwave-based measurements provide information on snow and ice cover regardless of solar illumination and extensive cloud cover, though coarse spatial resolution (passive microwave), land contamination, insufficient temporal resolution, and inconsistent viewing geometries present new challenges. An alternative approach in snow and ice mapping is the application of the Interactive Multisensor Snow and Ice Mapping System (IMS).

The IMS product provides daily snow and ice cover maps of the Northern Hemisphere produced by analysts at 1 km (2014 – present), 4 km (2004 – present), and 24 km (1997 – present) spatial resolutions (National Ice Center, 2008; Brown et al. 2014). Discrete values are assigned to land, snow covered land, water, and ice by utilizing multi-sourced datasets. IMS is derived from a variety of data products including optical and passive satellite imagery, as well as in situ observations (Appendix 1). Of note when using IMS is the operational nature of the product compared to the more consistent long-term snow and ice climatology possible with existing products; the inclusion of more data sources over time does change the nature of the IMS product, however, as shown by Brown et al. (2014) this does not preclude the use of IMS for examining changes over time. Additionally, IMS records could potentially exhibit lag effects in the timing of melt or freeze events if no data are available to the analyst and changes cannot be noted (Helfrich et al. 2007).

Geostationary data looping is the primary tool for determination of snow cover in the IMS product (Ramsay, 1998; Helfrich et al. 2007). For geostationary satellite imagery, a sequence loop of images is available over the entire coverage region for that particular satellite (Helfrich et al. 2019). The geostationary looping represents an estimated 60% of the snow analysis examination areas

during the winter, and 30% during the summer (Helfrich et al. 2007). During summer months, polar orbiting satellites’ visible channels characterize approximately 65% of the snow analysis (Helfrich et al. 2007). Visible imagery is preferred for mapping snow extent; however, analysts will use microwave data in the event that optical imagery is unavailable due to cloud occlusion or low solar illumination angles (Helfrich et al. 2007). Snow extent data derived from microwave sensors have well documented snow misidentification errors due to signal obstructions, snow grain size, and land cover influences that influence microwave emission and scatter (Chang et al. 1996; Foster et al.1999; Foster et al. 2004; Helfrich et al. 2007; Derksen, 2008; Brown et al. 2010). The accuracy of snow detection with the 4 km IMS product increases with snow depth and snow cover extent. Chen et al. (2012) show that the average rate of agreement between IMS and snow depth records ranges from 79% to 100%, improving later in the season when snow cover extent increases. Compared to 4 km IMS, the 24 km IMS product holds onto snow too long over large areas in the Arctic in June, and maps higher snow cover fractions in the Arctic during the spring melt period (Brown et al. 2010). Changes in snow cover are influenced by local-scale processes and microclimate factors, which may not be captured by IMS given its spatial resolution (24 km in particular). IMS can, however, provide an overview of snow cover conditions at an improved spatial resolution over passive microwave imagery. Additionally, the 1-day temporal resolution of IMS may better capture changes in snow cover throughout the snow advance/retreat seasons, compared to the finer resolution (500 m) 8-day MODIS snow product. To improve the performance of IMS, external data sources from meteorological observations and automated snow cover maps are often used to validate an area having snow or snow obscured with clouds (Helfrich et al. 2007). Often, the IMS analyst must use a consensus of several data sources to provide an optimal approach to determining the presence of snow (Helfrich et al. 2007).

Ice cover analysis relies on a different approach than snow cover, relying less on high albedo, stagnate cover, and meteorological conditions (Helfrich et al. 2007). Newly formed ice often has a low albedo until the ice thickens and becomes more opaque (Wadhams, 2000; Helfrich et al. 2007). Furthermore, ice is a dynamic surface, making it more difficult to distinguish from clouds (Helfrich et al. 2007). The prominence of low-level stratus clouds over polar regions precludes the use of visible imagery as the primary source for ice observations (Helfrich et al. 2007). Most high latitude regions are verified as being ice-covered using microwave-based data, representing approximately 30 to 35% of the winter and autumn ice cover input. In regions where observations

are unavailable, ice climatology is alternatively used for estimating ice cover (Helfrich et al. 2007). IMS analysts attempt to identify whether each pixel contains more than 50% ice cover (Helfrich et al. 2007). IMS cannot be used to detect initial melt onset of sea ice (i.e. wet snow over ice), as only the discrete cover types of snow, ice, land and water are noted in the product. The higher spatial resolution of IMS compared to the current passive microwave estimates contributes to the improvements of ice cover estimates by reducing land contamination and allowing coastal regions to be more accurately represented (Brown et al. 2014).

1.4 Objectives and Thesis Structure

The main gaps that have been identified in the literature are that although snow and ice phenology have been studied using multiple types of satellite-based measurements in Arctic regions, there has been little attention given to the IMS product, which offers an improved spatial resolution over passive microwave products, a higher temporal resolution compared to optical and active microwave data, and coverage of the entire Arctic region. Additionally, the improved spatial resolution and daily image availability make IMS suitable for examining sea ice, lake ice, and snow phenology together.

The overall goal of this work was to examine both long term and recent spatiotemporal changes in snow and ice phenology utilizing finer-resolution satellite observations. This was achieved through two main objectives: 1) investigating the changes in sea ice and snow phenology in the Canadian Arctic Archipelago utilizing the 24 km and 4 km IMS products and 2) utilizing and expanding the results from objective 1 to examine the larger scale changes and relationships in sea ice, lake ice, and snow phenology at the pan-Arctic scale.

To address objectives 1 and 2, this manuscript-based thesis has been structured into two primary content chapters. Chapter 1 serves as this introduction, while Chapter 2 addresses objective 1 by detailing the utility of the 24 and 4 km IMS products to examine sea ice and snow phenology in the Canadian Arctic Archipelago, and discusses trends and spatial variability in sea ice and snow melt and freeze. Chapter 3 addresses objective 2 through the application of IMS products for detecting sea ice, lake ice, and snow phenology at the pan-Arctic scale and discusses trends for the pan-Arctic as a whole as well as regional differences in melt/freeze shifts.

Chapter 2 of this thesis was submitted to the journal Arctic Science (manuscript ID: AS-2020- 0024) on July 7th, 2020 and the manuscript was accepted to the journal on September 11th, 2020 (manuscript ID: AS-2020-0024.R1; see https://www.nrcresearchpress.com/doi/10.1139/AS-2020- 0024#.X3JI-2hKhhE). The research objectives were designed together with Laura Brown and I carried out all of the processing and analysis for this work. Laura Brown provided insight regarding the methodology and interpretation of results, as well as valuable revision and comments.

Chapter 3 is currently being prepared for submission to The Cryosphere with co-author Laura Brown. I carried out all of the processing and analysis for this work, with feedback and revisions contributed by Laura Brown.

Finally, Chapter 4 provides an overall conclusion of the research findings and suggests further work this research can lead to.

1.5 References

Arrhenius, S. 1896. XXXI. On the influence of carbonic acid in the air upon the temperature of the ground. The London, Edinburgh, and Dublin Philosophical Magazine and Journal of Science, 41(251), 237-276.

Antonova, S., Duguay, C. R., Kääb, A., Heim, B., Langer, M., Westermann, S., & Boike, J. 2016. Monitoring bedfast ice and ice phenology in lakes of the Lena river delta using TerraSAR-X backscatter and coherence time series. Remote Sens. 8(11), 903.

Barry, R.G. and Maslanik, J.A. 1993. Monitoring lake freeze-up/break-up as a climatic index. In Proc. of Snow Watch ’92, WDC-A Glaciological Data Report 25, 66–79.

Benson, B. J., Magnuson, J. J., Jensen, O. P., Card, V. M., Hodgkins, G., Korhonen, J., ... and Granin, N. G. 2012. Extreme events, trends, and variability in Northern Hemisphere lake-ice phenology (1855–2005). Clim. Change. 112(2), 299-323.

Brown, R., Derksen, C., and Wang, L. 2007. Assessment of spring snow cover duration variability over northern Canada from satellite datasets. Remote Sens. Environ. 111(2-3), 367-381.

Brown, R., Derksen, C. and Wang, L. 2010. A multi‐data set analysis of variability and change in Arctic spring snow cover extent, 1967–2008. Geophys. Res. 115: D16111. doi: 10.1029/2010JD013975.

Brown, L. C., and Duguay, C. R. 2010. The response and role of ice cover in lake-climate interactions. Prog. Phys. Geo. 34(5), 671-704.

Brown, R. D., and Robinson, D. A. 2011. Northern Hemisphere spring snow cover variability and change over 1922-2010 including an assessment of uncertainty. Cryosphere, 5(1), 219.

Brown, L.C. and Duguay, C.R. 2011. The fate of lake ice in the North American Arctic. Cryosphere, 5: 869-892. doi: 10.5194/tc-5-869-2011.

Brown, L.C. and Duguay, C.R. 2012. Modelling lake ice phenology with sub-grid cell variability. Adv. Meteorol. 2012, 529064.

Brown, L.C., Howell, S.E., Mortin, J. and Derksen, C. 2014. Evaluation of the Interactive Multisensor Snow and Ice Mapping System (IMS) for monitoring sea ice phenology. Remote Sens. Environ. 147: 65-78. doi: https://doi.org/10.1016/j.rse.2014.02.012.

Brown, R., D. Vikhamar Schuler, O. Bulygina, C. Derksen, K. Luojus, L. Mudryk, L. Wang, and D. Yang. 2017. Arctic terrestrial snow cover, Chapter 3 in: Snow, Water, Ice and Permafrost in the Arctic (SWIPA) 2017 Assessment, Arctic Monitoring and Assessment.

Brubaker, K. L., Pinker, R. T., and Deviatova, E. 2005. Evaluation and comparison of MODIS and IMS snow-cover estimates for the continental United States using station data. J. Hydrometeorol. 6: 1002-1017. doi: https://doi.org/10.1175/JHM447.1

Callaghan, T.V., Johansson, M., Brown, R.D., Groisman, P.Y., Labba, N., Radionov, V., … and Golubev, V.N. 2011. The changing face of Arctic snow cover: a synthesis of observed and projected changes. Ambio, 40: 17-31.

Canadian Space Agency (CSA). 2019. What is the RCM? Available online: https://www.asc- csa.gc.ca/eng/satellites/radarsat/what-is-rcm.asp

Cavalieri, D. J., Parkinson, C. L., Gloersen, P., Comiso, J. C., and Zwally, H. J. 1999. Deriving long‐term time series of sea ice cover from satellite passive‐microwave multisensor data sets. J. Geophys. Res. 104(C7), 15803-15814.

Chang, A.T.C., Foster, J.L. and Hall, D.K. 1996. Effects of forest on the snow parameters derived from microwave measurements during the BOREAS winter field campaign. Hydrol. Process, 10: 1565-1574. doi: https://doi.org/10.1002/(SICI)1099-1085(199612)10:12<1565::AID- HYP501>3.0.CO;2-5.

Chen, C., Lakhankar, T., Romanov, P., Helfrich, S., Powell, A., and Khanbilvardi, R. 2012. Validation of NOAA-Interactive Multisensor Snow and Ice Mapping System (IMS) by comparison with ground-based measurements over continental United States. Remote Sens. 4, 1134-1145.

Cho, E., Tuttle, S. E., and Jacobs, J. M. 2017. Evaluating consistency of snow water equivalent retrievals from passive microwave sensors over the north central US: SSM/I vs. SSMIS and AMSR-E vs. AMSR2. Remote Sens. 9(5), 465.

Comiso, J. C., Cavalieri, D. J., and Markus, T. 2003. Sea ice concentration, ice temperature, and snow depth using AMSR-E data. IEEE Geosci. Remote Sens. 41(2), 243-252.

Comiso, J. C., Parkinson, C. L., Gersten, R., and Stock, L. 2008. Accelerated decline in the Arctic sea ice cover. Geophys. Res. Lett. 35(1).

Cook, T. L., and Bradley, R. S. 2010. An analysis of past and future changes in the ice cover of two High-Arctic lakes based on synthetic aperture radar (SAR) and Landsat imagery. Arct. Antarct. Alp. Res. 42(1), 9-18.

Crawford, A. D., Horvath, S., Stroeve, J., Balaji, R., and Serreze, M. C. 2018. Modulation of sea ice melt onset and retreat in the Laptev Sea by the timing of snow retreat in the West Siberian Plain. J. Geophys. Res. 123(16), 8691-8707.

Curry, J. A., Schramm, J. L., and Ebert, E. E. 1995. Sea ice-albedo climate feedback mechanism. J. Clim. 8(2), 240-247.

Dai, L., Che, T., Wang, J., and Zhang, P. 2012. Snow depth and snow water equivalent estimation from AMSR-E data based on a priori snow characteristics in Xinjiang, China.Remote Sens. Environ. 127, 14-29.

De Lannoy, G. J., Reichle, R. H., Arsenault, K. R., Houser, P. R., Kumar, S., Verhoest, N. E., and Pauwels, V. R. 2012. Multiscale assimilation of Advanced Microwave Scanning Radiometer–EOS snow water equivalent and Moderate Resolution Imaging Spectroradiometer snow cover fraction observations in northern Colorado. Water Resour. Res. 48(1).

Derksen, C., Brown, R., and Walker, A. 2004. Merging conventional (1915–92) and passive microwave (1978–2002) estimates of snow extent and water equivalent over central North America. J. Hydrometeor. 5(5), 850-861.

Derksen, C. 2008. The contribution of AMSR-E 18.7 and 10.7 GHz measurements to improved boreal forest snow water equivalent retrievals. Remote Sens. Environ. 112: 2701-2710. doi: https://doi.org/10.1016/j.rse.2008.01.001.

Derksen, C. and R. Brown. 2012. Snow [in “Arctic Report Card 2012”]. Jeffries, M. O., J. A. Richter-Menge and J. E. Overland. (Eds.). Available online at http://www.arctic.noaa.gov/reportcard.

Derksen, C., Smith, S. L., Sharp, M., Brown, L., Howell, S., Copland, L., ... and Bernier, M. 2012. Variability and change in the Canadian cryosphere. Clim. Change, 115(1), 59-88.

Derksen, C., Burgess, D., Duguay, C., Howell, S., Mudryk, L., Smith, S., Thackeray, C. and Kirchmeier-Young, M. 2019: Changes in snow, ice, and permafrost across Canada; Chapter 5 in Canada’s Changing Climate Report, (ed.) E. Bush and D.S. Lemmen; Government of Canada, Ottawa, Ontario, p.194–260.

Day, J. J., Hargreaves, J. C., Annan, J. D., and Abe-Ouchi, A. 2012. Sources of multi-decadal variability in Arctic sea ice extent. Environ. Res. Lett. 7(3), 034011.

Déry, S.J. and Brown, R.D. 2007. Recent Northern Hemisphere snow cover extent trends and implications for the snow‐albedo feedback. Geophys. Res. Lett. 34: L22504. doi:10.1029/2007GL031474.

Deser, C., and Teng, H. 2008. Evolution of Arctic sea ice concentration trends and the role of atmospheric circulation forcing, 1979–2007. Geophys. Res. Lett. 35(2).

Dietz, A. J., Kuenzer, C., Gessner, U., and Dech, S. 2012. Remote sensing of snow–a review of available methods. Int. J. Remote Sens. 33(13), 4094-4134.

Ding, Q., Wallace, J. M., Battisti, D. S., Steig, E. J., Gallant, A. J., Kim, H. J., and Geng, L. 2014. Tropical forcing of the recent rapid Arctic warming in northeastern Canada and Greenland. Nature, 509(7499), 209.

Ding, Q., Schweiger, A., L’Heureux, M., Battisti, D. S., Po-Chedley, S., Johnson, N. C., ... and Steig, E. J. 2017. Influence of high-latitude atmospheric circulation changes on summertime Arctic sea ice. Nat. Clim. Change.

Dong, C. and Menzel, L. 2016. Improving the accuracy of MODIS 8-day snow products with in situ temperature and precipitation data. J. Hydrol. 534, 466-467.

Duguay, C. K., Rouse, W. R., Lafleur, P. M., Boudreau, L. D., Crevier, Y., and Pultz, T. J. 1999. Analysis of multi-temporal ERS-1 SAR data of subarctic tundra and forest in the northern Hudson Bay Lowland and implications for climate studies. Can. J. Remote Sens. 25(1), 21-33.

Duguay, C. R., Pultz, T. J., Lafleur, P. M., and Drai, D. 2002. RADARSAT backscatter characteristics of ice growing on shallow sub‐Arctic lakes, Churchill, Manitoba, Canada. Hydrol. Proc. 16(8), 1631-1644.

Duguay, C. R., Flato, G. M., Jeffries, M. O., Ménard, P., Morris, K., and Rouse, W. R. 2003. Ice‐ cover variability on shallow lakes at high latitudes: model simulations and observations. Hydrol. Process. 17(17), 3465-3483.

Duguay, C., Brown, L., Kang, K., and Kheyrollah Pour, H. 2012. [The Arctic] Lake Ice [in “State of the Climate 2011]. Bull. Amer. Meteor. Soc. 93(7), S138-S140.

Duguay, C., L. Brown, Kang K-K., and Kheyrollah Pour, H. 2013. [The Arctic] Lake ice [In “State of the Climate in 2012”]. Bull. Amer. Meteor. Soc. 94(8): S124-S126.

Duguay, C., L. Brown, Kang, K.-K., and Kheyrollah Pour, H. 2014. [The Arctic] Lake ice [In “State of the Climate in 2013”]. Bull. Amer. Meteor. Soc. 95 (7).

Duguay, C. R., Bernier, M., Gauthier, Y., and Kouraev, A. 2015a. Remote sensing of the cryosphere: Remote sensing of lake and river ice: 1st edn. John Wiley & Sons, Ltd, UK.

Duguay, C., L. Brown, Kang, K.-K., and Kheyrollah Pour, H. 2015b. [The Arctic] Lake ice [In “State of the Climate in 2014”]. Bull. Amer. Meteor. Soc. 96 (7), S144-S145.

Duguay C and Brown L. 2018. Lake Ice [in Arctic Report Card 2018], https://arctic.noaa.gov/Report-Card/Report-Card-2018/ArtMID/7878/ArticleID/785/Lake-Ice

Estilow, T. W., Young, A. H., and Robinson, D. A. 2015. A long-term Northern Hemisphere snow cover extent data record for climate studies and monitoring. Earth Syst. Sci. Data, 7(1), 137.

Foster, J. L., Hall, D. K., Chang, A. T., Rango, A., Wergin, W., and Erbe, E. 1999. Effects of snow crystal shape on the scattering of passive microwave radiation. IEEE Geosci. Remote Sens. 37(2), 1165-1168.

Foster, J. L., Sun, C., Walker, J. P., Kelly, R., Dong, J., and Chang, A. 2004. Mapping random and systematic errors of satellite-derived snow-water equivalent observations in Eurasia. In Remote Sensing for Agriculture, Ecosystems, and Hydrology, 5568, 150-159.

Gao, Y., Xie, H., Lu, N., Yao, T., & Liang, T. (2010). Toward advanced daily cloud-free snow cover and snow water equivalent products from Terra–Aqua MODIS and Aqua AMSR-E measurements. Journal of Hydrology, 385(1-4), 23-35.

Gao, S., Li, Z., Chen, Q., Zhou, W., Lin, M., and Yin, X. 2019. Inter-Sensor Calibration between HY-2B and AMSR2 Passive Microwave Data in Land Surface and First Result for Snow Water Equivalent Retrieval. J. Sens. 19(22), 5023.

Geldsetzer, T. 2010. Mapping and monitoring lake ice using SAR satellites. In GRIP Project – Fresh Surface Water Mapping and Monitoring Using SAR Satellites. Available online: ftp://ftp.ccrs.nrcan.gc.ca/ad/Lake_ice/Documents/Mapping%20and%20monitoring%20lake%20i ce%20using%20SAR%20satellites.pdf

Geldsetzer, T., and Van Der Sanden, J. J. 2013. Identification of polarimetric and nonpolarimetric C-band SAR parameters for application in the monitoring of lake ice freeze-up. Can. J. Remote Sens. 39(3), 263-275.

Gregory, J. M., Stott, P. A., Cresswell, D. J., Rayner, N. A., Gordon, C., and Sexton, D. M. H. 2002. Recent and future changes in Arctic sea ice simulated by the HadCM3 AOGCM. Geophys. Res. Lett. 29(24).

Helfrich, S. R., McNamara, D., Ramsay, B. H., Baldwin, T., and Kasheta, T. 2007. Enhancements to, and forthcoming developments in the Interactive Multisensor Snow and Ice Mapping System (IMS). Hydrol. Process, 21: 1576-1586. doi: 10.1002/hyp.6720.

Hernandez-Henriquez, M. A., Déry, S. J., and Derksen, C. 2015. Polar amplification and elevation- dependence in trends of Northern Hemisphere snow cover extent, 1971–2014. Environ. Res. Lett. 10: 044010. doi: 10.1088/1748-9326/10/4/044010.

Howell, S. E., Tivy, A., Yackel, J. J., and Scharien, R. K. 2006. Application of a SeaWinds/QuikSCAT sea ice melt algorithm for assessing melt dynamics in the Canadian Arctic Archipelago. J. Geophys. Res. 111: C07025. doi:10.1029/2005JC003193.

Howell, S. E., Tivy, A., Yackel, J. J., Else, B. G., and Duguay, C. R. 2008a. Changing sea ice melt parameters in the Canadian Arctic Archipelago: Implications for the future presence of multiyear ice. J. Geophys. Res. 113: C09030. doi:10.1029/2008JC004730.

Howell, S. E., Tivy, A., Yackel, J. J., and McCourt, S. 2008b. Multi‐year sea‐ice conditions in the western Canadian Arctic Archipelago region of the northwest passage: 1968–2006. Atmos. - Ocean, 46: 229-242. doi: 10.3137/ao.460203.

Howell, S. E., Brown, L. C., Kang, K. K., and Duguay, C. R. 2009a. Variability in ice phenology on Great Bear Lake and Great Slave Lake, Northwest Territories, Canada, from SeaWinds/QuikSCAT: 2000–2006. Remote Sensing of Environment, 113(4), 816-834.

Howell, S.E., Duguay, C.R. and Markus, T., 2009b. Sea ice conditions and melt season duration variability within the Canadian Arctic Archipelago: 1979–2008. Geophys. Res. Lett. 36: L10502. doi:10.1029/2009GL037681.

Howell, S.E., Tivy, A., Agnew, T., Markus, T. and Derksen, C., 2010. Extreme low sea ice years in the Canadian Arctic Archipelago: 1998 versus 2007. J. Geophys. Res. 115: C10053. doi:10.1029/2010JC006155.

Howell, S. E., Wohlleben, T., Dabboor, M., Derksen, C., Komarov, A., and Pizzolato, L. 2013. Recent changes in the exchange of sea ice between the Arctic Ocean and the Canadian Arctic Archipelago. J. Geophys. Res. 118: 3595-3607. doi:10.1002/jgrc.20265.

Howell, S.E., Derksen, C., Pizzolato, L. and Brady, M. 2015. Multiyear ice replenishment in the Canadian Arctic Archipelago: 1997–2013. J. Geophys. Res. 120: 1623-1637. doi: 10.1002/2015JC010696.

Howell, S. E., Small, D., Rohner, C., Mahmud, M. S., Yackel, J. J., and Brady, M. 2019. Estimating melt onset over Arctic sea ice from time series multi-sensor Sentinel-1 and RADARSAT-2 backscatter. Remote Sens. Environ. 229: 48-59. doi: https://doi.org/10.1016/j.rse.2019.04.031.

Hu, A., Rooth, C., Bleck, R., and Deser, C. 2002. NAO influence on sea ice extent in the Eurasian coastal region. Geophys. Res. Lett. 29(22).

Jeffries, M. O., Morris, K., Weeks, W. F., and Wakabayashi, H. 1994. Structural and stratigraphic features and ERS 1 synthetic aperture radar backscatter characteristics of ice growing on shallow lakes in NW Alaska, winter 1991–1992. J. Geophys. Res. 99(C11), 22459-22471.

Johannessen, O. M., Bengtsson, L., Miles, M. W., Kuzmina, S. I., Semenov, V. A., Alekseev, G. V., ... and Hasselmann, K. 2004. Arctic climate change: Observed and modelled temperature and sea‐ice variability. Tellus A, 56(4), 328-341.

Johannessen, O.M., Kuzmina, S.I., Bobylev, L.P. and Miles, M.W. 2016. Surface air temperature variability and trends in the Arctic: new amplification assessment and regionalisation. Tellus A, 68: 28234. doi: 10.3402/tellusa.v68.28234.

Johnson, M., and Eicken, H. 2016. Estimating Arctic sea-ice freeze-up and break-up from the satellite record: A comparison of different approaches in the Chukchi and Beaufort Seas. Elem Sci Anth. 4.

Kang, K. K., Duguay, C. R., and Howell, S. E. L. 2012. Estimating ice phenology on large northern lakes from AMSR-E: algorithm development and application to Great Bear Lake and Great Slave Lake, Canada. Cryosphere, 6(2), 235-254.

Kay, J. E., Holland, M. M., and Jahn, A. 2011. Inter‐annual to multi‐decadal Arctic sea ice extent trends in a warming world. Geophys. Res. Lett. 38(15).

King, J., Pomeroy, J., Gray, D. M., Fierz, C., Föhn, P., Harding, R., ... & Plüss, C. 2008. Snow- atmosphere energy and mass balance. In: Snow and Climate: Physical Processes, Surface Energy Exchange and Modelling (pp. 70-124). Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA.

Kirchmeier-Young, M. C., Zwiers, F. W., and Gillett, N. P. 2017. Attribution of extreme events in Arctic Sea ice extent. J. Clim. 30(2), 553-571.

Kropáček, J., Maussion, F., Chen, F., Hoerz, S., and Hochschild, V. 2013. Analysis of ice phenology of lakes on the Tibetan Plateau from MODIS data. Cryosphere, 7(1), 287-301.

Kwok, R., Cunningham, G. F., and Nghiem, S. V. 2003. A study of the onset of melt over the Arctic Ocean in RADARSAT synthetic aperture radar data. J. Geophys. Res. 108(C11).

Larue, F., Royer, A., Sève, D. D., Roy, A., and Cosme, E. 2018. Assimilation of passive microwave AMSR-2 satellite observations in a snowpack evolution model over northeastern Canada. Hydrol. Earth Syst. Sci. 22(11), 5711-5734.

Lawrence, D. M., Slater, A. G., Tomas, R. A., Holland, M. M., and Deser, C. 2008. Accelerated Arctic land warming and permafrost degradation during rapid sea ice loss. Geophys. Res. Lett. 35(11).

Lehnherr, I., Louis, V.L.S., Sharp, M., Gardner, A.S., Smol, J.P., Schiff, S.L., Muir, D.C., Mortimer, C.A., Michelutti, N., Tarnocai, C. and Pierre, K.A.S. 2018. The world’s largest High Arctic lake responds rapidly to climate warming. Nat. Commun. 9: 1290. doi: 10.1038/s41467- 018-03685-z

Lei, R., Leppäranta, M., Cheng, B., Heil, P., and Li, Z. 2012. Changes in ice-season characteristics of a European Arctic lake from 1964 to 2008. Clim. Change, 115(3-4), 725-739.

Liston, G. E. 2004. Representing subgrid snow cover heterogeneities in regional and global models. J. Clim. 17(6), 1381-1397.

Lynch, A. H., Serreze, M. C., Cassano, E. N., Crawford, A. D., and Stroeve, J. 2016. Linkages between Arctic summer circulation regimes and regional sea ice anomalies. J. Geophys. Res. 121(13), 7868-7880.

Mahmud, M. S., Howell, S. E., Geldsetzer, T., and Yackel, J. 2016. Detection of melt onset over the northern Canadian Arctic Archipelago sea ice from RADARSAT, 1997–2014. Remote Sens. Environ. 178: 59-69. doi: https://doi.org/10.1016/j.rse.2016.03.003.

Markus, T., and Cavalieri, D. J. 2009. The AMSR-E NT2 sea ice concentration algorithm: Its basis and implementation. J. Remote Sens. Soc. of Japan, 29(1), 216-225.

Maslanik, J. A., Fowler, C., Stroeve, J., Drobot, S., Zwally, J., Yi, D., and Emery, W. 2007. A younger, thinner Arctic ice cover: increased potential for rapid, extensive sea‐ice loss. Geophys. Res. Lett. 34: L24501. doi:10.1029/2007GL032043.

McKay, G. A., and Gray, M. D. 1981. The distribution of snow cover. Handbook of Snow, 153- 190. 1st edn, Pergamon Press, Toronto, ON, CA.

Michelutti, N., Douglas, M. S., Antoniades, D., Lehnherr, I., St. Louis, V. L., St. Pierre, K., ... and Smol, J. P. 2020. Contrasting the ecological effects of decreasing ice cover versus accelerated glacial melt on the High Arctic's largest lake. P. Roy. Soc. B-Biol. Sci. 287(1929), 20201185.

Min, S. K., Zhang, X., Zwiers, F. W., and Agnew, T. 2008. Human influence on Arctic sea ice detectable from early 1990s onwards. Geophys. Res. Lett. 35(21).

Mishra, V., Cherkauer, K. A., Bowling, L. C., and Huber, M. 2011. Lake ice phenology of small lakes: Impacts of climate variability in the Great Lakes region. Global Planet. Change, 76(3-4), 166-185.

Morris, K., Jeffries, M. O., and Weeks, W. F. 1995. Ice processes and growth history on Arctic and sub-Arctic lakes using ERS-1 SAR data. Polar Rec. 31(177), 115-128.

Mortimer, C., Mudryk, L., Derksen, C., Luojus, K., Brown, R., Kelly, R., and Tedesco, M. 2020. Evaluation of long-term Northern Hemisphere snow water equivalent products. Cryosphere, 14(5), 1579-1594.

Mortin, J., Howell, S. E., Wang, L., Derksen, C., Svensson, G., Graversen, R. G., and Schrøder, T. M. 2014. Extending the QuikSCAT record of seasonal melt–freeze transitions over Arctic sea ice using ASCAT. Remote Sens. Environ. 141, 214-230.

Mudryk, L.R., Derksen, C., Howell, S., Laliberté, F., Thackeray, C., Sospedra-Alfonso, R., Vionnet, V., Kushner, P.J. and Brown, R. 2018. Canadian snow and sea ice: historical trends and projections. Cryosphere, 12: 1157-1176.

Mudryk, L., Brown, R., Derksen, C., Luojus, K., Decharme, B., & Helfrich, S. 2019. Terrestrial snow cover [in “Arctic Report Card 2019”]. J. Richter-Menge, M. L. Druckenmiller, and M. Jeffries, Eds.

Mueller, D. R., Van Hove, P., Antoniades, D., Jeffries, M. O., and Vincent, W. F. 2009. High Arctic lakes as sentinel ecosystems: Cascading regime shifts in climate, ice cover, and mixing. Limnol. Oceanogr. 54, 2371-2385.

Murfitt, J., Brown, L., and Howell, S. 2018. Evaluating RADARSAT-2 for the Monitoring of Lake Ice Phenology Events in Mid-Latitudes. Remote Sens. 10(10), 1641.

Najafi, M. R., Zwiers, F. W., and Gillett, N. P. 2016. Attribution of the spring snow cover extent decline in the Northern Hemisphere, Eurasia and North America to anthropogenic influence. Clim. Change, 136: 571-586. doi: 10.1007/s10584-016-1632-2.

National Ice Center. 2008. IMS Daily Northern Hemisphere Snow and Ice Analysis at 1 km, 4 km, and 24 km Resolutions, Version 1. Boulder, Colorado USA. NSIDC: National Snow and Ice Data Center. doi: http://dx.doi.org/10.7265/N52R3PMC

National Snow and Ice Data Center (NSIDC). 2019a. Arctic sea ice reaches second lowest minimum in satellite record. Available online: https://nsidc.org/arcticseaicenews/2019/09/arctic- sea-ice-reaches-second-lowest-minimum-in-satellite-record/

National Snow and Ice Data Center (NSIDC). 2019b. Sea ice concentration documentation. Available online: https://nsidc.org/data/nsidc-0051

National Snow and Ice Data Center (NSIDC). 2020. Remote sensing: Active microwave. Available online: https://nsidc.org/cryosphere/seaice/study/active_remote_sensing.html

Natural Resources Canada (NRCAN). 2015a. Microwave remote sensing. Available online: https://www.nrcan.gc.ca/earth-sciences/geomatics/satellite-imagery-air-photos/satellite-imagery- products/educational-resources/9371

Natural Resources Canada (NRCAN). 2015b. Passive vs. active remote sensing. Available online: https://www.nrcan.gc.ca/node/14639

Nitze, I., Grosse, G., Jones, B. M., Arp, C. D., Ulrich, M., Fedorov, A., and Veremeeva, A. 2017. Landsat-based trend analysis of lake dynamics across northern permafrost regions. Remote Sens. 9(7), 640.

Notz, D., and Marotzke, J. 2012. Observations reveal external driver for Arctic sea‐ice retreat. Geophys. Res. Lett. 39(8).

Paquette, M., Fortier, D., Mueller, D. R., Sarrazin, D., and Vincent, W. F. 2015. Rapid disappearance of perennial ice on Canada's most northern lake. Geophys. Res. Lett. 42(5), 1433- 1440.

Parajka, J., and Blöschl, G. 2006. Validation of MODIS snow cover images over Austria. Hydrol. Earth Syst. Sc. 3(4), 1569 – 1601.

Parajka, J., and Blöschl, G. 2008. The value of MODIS snow cover data in validating and calibrating conceptual hydrologic models. J. Hydrol. 358(3-4), 240-258.

Park, H., Yabuki, H., and Ohata, T. 2012. Analysis of satellite and model datasets for variability and trends in Arctic snow extent and depth, 1948–2006. Polar Sci. 6(1), 23-37.

Parkinson, C. L., and Comiso, J. C. 2013. On the 2012 record low Arctic sea ice cover: Combined impact of preconditioning and an August storm. Geophys. Res. Lett. 40(7), 1356-1361.

Peng, S., Piao, S., Ciais, P., Friedlingstein, P., Zhou, L., and Wang, T. 2013. Change in snow phenology and its potential feedback to temperature in the Northern Hemisphere over the last three decades. Environ. Res. Lett. 8(1), 014008.

Perovich, D. K., Nghiem, S. V., Markus, T., and Schweiger, A. 2007. Seasonal evolution and interannual variability of the local solar energy absorbed by the Arctic sea ice–ocean system. J. Geophys. Res. 112(C3).

Petty, A. A., Stroeve, J. C., Holland, P. R., Boisvert, L. N., Bliss, A. C., Kimura, N., and Meier, W. N. 2018. The Arctic sea ice cover of 2016: A year of record-low highs and higher-than- expected lows. Cryosphere, 12(2), 433-452.

Pulliainen, J., Luojus, K., Derksen, C., Mudryk, L., Lemmetyinen, J., Salminen, M., ... & Norberg, J. (2020). Patterns and trends of Northern Hemisphere snow mass from 1980 to 2018. Nature, 581(7808), 294-298.

Rahmstorf, S. 2010. A new view on sea level rise. Nat. Clim. Change. 4(4), 44-45.

Ramsay, B. H. 1998. The interactive multisensor snow and ice mapping system. Hydrol. Process, 12: 1537-1546.

Riggs, G. A., Hall, D. K., and Román, M. O. 2016. MODIS Snow Products User Guide for Collection 6 (C6), available at: http://modis-snow-ice.gsfc.nasa.gov/?c=userguides

Rott, H., Yueh, S. H., Cline, D. W., Duguay, C., Essery, R., Haas, C., ... and Nagler, T. 2010. Cold regions hydrology high-resolution observatory for snow and cold land processes. P. IEEE, 98(5), 752-765.

Rouse, W. R., Oswald, C. J., Binyamin, J., Spence, C., Schertzer, W. M., Blanken, P. D., ... and Duguay, C. R. 2005. The role of northern lakes in a regional energy balance. J. Hydrometeorol. 6(3), 291-305.

Rouse, W. R., Binyamin, J., Blanken, P. D., Bussières, N., Duguay, C. R., Oswald, C. J., ... and Spence, C. 2008a. The influence of lakes on the regional energy and water balance of the central Mackenzie River Basin. In Cold region atmospheric and hydrologic studies. The Mackenzie GEWEX experience (pp. 309-325). Springer, Berlin, Heidelberg.

Rouse, W.R., Blanken, P.D., Duguay, C.R.; Oswald, C.J., and Schertzer, W.M. 2008b. Climate- lake interactions. In W. Rouse (Ed.), Cold Region Atmospheric and Hydrologic Studies (139 – 160). New York, NY, USA; Springer.

Salomonson, V. V., and Appel, I. 2004. Estimating fractional snow cover from MODIS using the normalized difference snow index. Remote Sens. Environ. 89(3), 351-360.

Schertzer, W. M., Rouse, W. R., Blanken, P. D., and Walker, A. E. 2003. Over-lake meteorology and estimated bulk heat exchange of Great Slave Lake in 1998 and 1999. J. Hydrometeorol. 4(4), 649-659.

Screen, J.A. and Simmonds, I. 2012. Declining summer snowfall in the Arctic: Causes, impacts and feedbacks. Clim. Dyn. 38: 2243-2256.

Serreze, M. C., Holland, M. M., and Stroeve, J. 2007. Perspectives on the Arctic's shrinking sea- ice cover. Science, 315: 1533-1536. doi: 10.1126/science.1139426.

Serreze, M. C., Barrett, A. P., Stroeve, J. C., Kindig, D. N., and Holland, M. M. 2009. The emergence of surface-based Arctic amplification. Cryosphere, 3(1), 11-19.

Serreze, M. C., and Barry, R. G. 2011. Processes and impacts of Arctic amplification: A research synthesis. Global Planet. Change, 77(1-2), 85-96.

Serreze, M. C., and Stroeve, J. 2015. Arctic sea ice trends, variability and implications for seasonal ice forecasting. Philos. Trans. R. Soc. A. 373(2045), 20140159.

Sharma, S., Blagrave, K., Magnuson, J. J., O’Reilly, C. M., Oliver, S., Batt, R. D., ... and Woolway, R. I. 2019. Widespread loss of lake ice around the Northern Hemisphere in a warming world. Nat. Clim. Change, 9(3), 227-231.

Shi, X. 2008. Active remote sensing systems and applications to snow monitoring: Chapter 3 in Advances in land remote sensing: System, modeling, inversion, and application. 1st edn. Springer Netherlands.

Smejkalova, T., Edwards, M. E., and Dash, J. 2016. Arctic lakes show strong decadal trend in earlier spring ice-out. Sci. Rep. 6, 38449.

Spreen, G., Kaleschke, L., and Heygster, G. 2008. Sea ice remote sensing using AMSR‐E 89‐GHz channels. J. Geophys. Res. 113: C02S03. doi:10.1029/2005JC003384.

Stroeve, J., Serreze, M., Drobot, S., Gearheard, S., Holland, M., Maslanik, J., ... and Scambos, T. 2008. Arctic sea ice extent plummets in 2007. Eos Trans. AGU. 89(2), 13-14.

Stroeve, J. C., Serreze, M. C., Holland, M. M., Kay, J. E., Malanik, J., and Barrett, A. P. 2012. The Arctic’s rapidly shrinking sea ice cover: a research synthesis. Clim. Change, 110: 1005-1027. doi: 10.1007/s10584-011-0101-1

Surdu, C. M., Duguay, C. R., Brown, L. C., and Fernández Prieto, D. 2014. Response of ice cover on shallow lakes of the North Slope of Alaska to contemporary climate conditions (1950–2011): radar remote-sensing and numerical modeling data analysis. Cryosphere, 8(1), 167-180.

Surdu, C. M., Duguay, C. R., Pour, H. K., and Brown, L. C. 2015. Ice Freeze-up and Break-up Detection of Shallow Lakes in Northern Alaska with Spaceborne SAR. Remote Sens. 7(5), 6133- 6159.

Surdu, C. M., Duguay, C. R., & Fernández Prieto, D. 2016. Evidence of recent changes in the ice regime of lakes in the Canadian High Arctic from spaceborne satellite observations. Cryosphere. 10(3), 941-960.

Tedesco, M., Derksen, C., Deems, J.S., & Foster, J. L. 2015. Remote sensing of the cryosphere: Remote sensing of snow depth and snow water equivalent: 1st edn. John Wiley & Sons, Ltd, UK

Thackeray, C. W., and Fletcher, C. G. 2016. Snow albedo feedback: Current knowledge, importance, outstanding issues and future directions. Prog. Phys. Geo. 40(3), 392-408.

Thompson, A., and Kelly, R. 2019. Observations of a coniferous forest at 9.6 and 17.2 GHz: Implications for SWE retrievals. Remote Sens. 11(1), 6.

Tivy, A., Howell, S.E., Alt, B., McCourt, S., Chagnon, R., Crocker, G., Carrieres, T. and Yackel, J.J. 2011. Trends and variability in summer sea ice cover in the Canadian Arctic based on the Canadian Ice Service Digital Archive, 1960–2008 and 1968–2008. J. Geophys. Res. 116: C03007. doi:10.1029/2009JC005855. van Wijngaarden, W. A. 2015a. Arctic temperature trends from the early nineteenth century to the present. Theor. Appl. Climatol. 122, 567-580. van Wijngaarden, W. A. 2015b. Temperature trends in the Canadian Arctic during 1895–2014. Theor. Appl. Climatol. 120, 609-615.

Wadhams P. 2000. Ice in the Ocean. Gordon and Breach Science Publishers: London, England; 351.

Walker, A. E., Silis, A., Metcalf, J. R., Davey, M. R., Brown, R. D., and Goodison B. E. 2000. Snow cover and lake ice determination in the MAGS region using passive microwave satellite and conventional data. In Proc. 5th Scientific Workshop, Mackenzie GEWEX Study, Edmonton, Alberta, Canada, 39–42.

Walsh, J. E., Overland, J. E., Groisman, P. Y., amd Rudolf, B. 2011. Ongoing climate change in the Arctic. Ambio, 40(1), 6-16.

Wang, L., Sharp, M., Brown, R., Derksen, C., & Rivard, B. 2005. Evaluation of spring snow covered area depletion in the Canadian Arctic from NOAA snow charts. Remote Sens. Environ. 95: 453-463. doi:10.1016/j.rse.2005.01.006.

Wang, L., Wolken, G.J., Sharp, M.J., Howell, S.E.L., Derksen, C., Brown, R.D., Markus, T. and Cole, J. 2011. Integrated pan‐Arctic melt onset detection from satellite active and passive microwave measurements, 2000–2009. J. Geophys. Res. 116: D22103. doi:10.1029/2011JD016256.

Yackel, J. J., Barber, D. G., and Papakyriakou, T. N. 2001. On the estimation of spring melt in the North Water polynya using RADARSAT‐1. Atmos.-Ocean, 39(3), 195-208.

Young, K.L., Brown, L. and Labine, C., 2018. Snow cover variability at Polar Bear Pass, Nunavut. Arct. Sci. 4: 669-690. doi: dx.doi.org/10.1139/as-2017-0016.

Yu, P., Clausi, D. A., and Howell, S. E. 2009. Fusing AMSR-E and QuikSCAT imagery for improved sea ice recognition. IEEE Geosci. Remote Sens. 47(7), 1980-1989.

Yu, L., Liu, T., and Zhang, S. 2017. Temporal and Spatial Changes in Snow Cover and the Corresponding Radiative Forcing Analysis in Siberia from the 1970s to the 2010s. Adv. Meteorol. 2017.

Zhang, X., Flato, G., Kirchmeier-Young, M., Vincent, L., Wan, H., Wang, X., … and Kharin, V.V. 2019. Changes in Temperature and Precipitation Across Canada; Chapter 4. In Bush, E. and Lemmen, D.S. (Eds.) Canada’s Changing Climate Report (pp. 112 – 193). Government of Canada, Ottawa, Ontario.

Zhu, J., Tan, S., King, J., Derksen, C., Lemmetyinen, J., and Tsang, L. 2018. Forward and inverse radar modeling of terrestrial snow using SnowSAR data. IEEE Geosci. Remote Sens. 56(12), 7122-7132.

38 39

Chapter 2 Sea ice and Snow Phenology in the Canadian Arctic Archipelago from 1997 – 2018

Abstract

The multiple islands and narrow channels that form the Canadian Arctic Archipelago (CAA) complicate snow/ice monitoring, as coarse resolution satellite observations are unable to resolve smaller-scale changes in snow/ice cover. We present the first study showing the utility of the Interactive Multisensor Snow and Ice Mapping System (IMS) 24 km (1997–2018) and 4 km (2004–2018) products to investigate changes in sea ice and snow phenology together in the CAA. While ice break-up and snow retreat are shifting earlier (p>0.05), on par with other Arctic regions, the final summer clearing of ice is shifting later. This, combined with trends towards earlier ice freeze and snow fall (p<0.05), result in shorter open water and snow free seasons in the CAA. Spatial links between sea ice and snow are evident as significant clusters of trends were identified for all phenology parameters. The western regions were dominated by shifts toward shorter snow/ice seasons, while eastern regions tended to exhibit longer cover. Our research highlights the considerable regional and interannual variability in the timing of sea ice and snow advance/retreat within the CAA and emphasizes how the ice and snow dynamics in this complex region are responding to ongoing changing climate conditions.

2.1 Introduction

Arctic sea ice and snow cover play a critical role in the surface energy budget by modulating the exchange of energy between the atmosphere and ocean through various feedback mechanisms (Perovich et al. 2007; Stroeve et al. 2012; Brown et al. 2014). When surface air temperatures increase and promote snow and ice melt, surface albedo decreases and drives increases in shortwave radiation absorption, which in turn further increase surface temperatures and accelerate melt (Curry et al. 1995; Brown et al. 2014). Recent increases in surface air temperatures have resulted in reduced sea ice extent, a transition from thick multi-year ice (MYI) to thinner seasonal first-year ice (FYI), and earlier melt onset dates (Howell et al. 2006; Maslanik et al. 2007; Serreze et al. 2007; Stroeve et al. 2012; Brown et al. 2014). Significant declines in MYI have occurred in the Canadian Arctic Archipelago (CAA) and Beaufort Sea compared to previous studies (e.g. Tivy

et al. 2011; Derksen et al. 2012), with the rate of decline from 1968 – 2016 almost doubling compared to the previously identified trend for 1968 – 2008 (Mudryk et al. 2018). However, MYI loss in the Beaufort Sea is greater than the north facing coast of the CAA, which still contains some of the thickest sea ice in the world, which is more resistant to melting than thinner ice elsewhere (Laliberte et al. 2016; Mudryk et al. 2018). The expected persistence of summer ice extent in the CAA and Greenlandic regions of the Arctic through the mid-century (Laliberte et al. 2016) has resulted in the area being formally recognized as the “Last Ice Area” (WWF, 2018; Moore et al. 2019). Howell and Brady (2019) recently reported that ice area fluxes from the Arctic Ocean into the CAA have increased by 103 km2/year from 1997 – 2018, suggesting that the CAA may become a larger outlet for Arctic Ocean ice area loss.

Decreases in snow cover extent and duration on land, as well as earlier melt onset dates, have also been reported (Dery and Brown, 2007; Brown et al. 2010; Callaghan et al. 2011; Derksen and Brown, 2012; Najafi et al. 2016). Northern Hemisphere springtime (March to June) snow cover extent was found to decrease by approximately 3.3 % per decade from 1981 to 2010 (Thackeray et al. 2016). Declines in snow water equivalent (SWE) have also been reported across the Northern Hemisphere from 1981 to 2010, however evidence of regional increases in winter snow season accumulation (therefore higher SWE) have been identified in the western CAA (Mudryk et al. 2015; Mudryk et al. 2018). Exceptions to widespread snow loss patterns exist in parts of the Arctic, where increasing temperatures increase the moisture capacity of air, which drive increased precipitation (Bintania and Andry, 2017; Thackeray et al. 2019). Analyzing snow cover changes in remote areas such as the CAA can be challenging given the large degree of spatial variability due to the complex physiography associated with the region and limited observational networks in place.

Monitoring changes in Arctic sea ice and snow cover has primarily relied on the use of satellite- based microwave data (e.g. QuikSCAT, Advanced Microwave Scanning Radiometer (AMSR- E/AMSR2), RADARSAT-1/2, Scanning Multi-Channel Microwave Radiometer (SMMR), Special Sensor Microwave/Imager (SSM/I), NASA and Canadian Ice Services sea ice products) (e.g. Comiso and Nishio, 2008; Howell et al. 2008a; Spreen et al. 2008; Howell et al. 2012; Wang et al. 2013) as they provide information regardless of solar illumination and extensive cloud cover (Brown et al. 2014). Algorithms applied to microwave measurements have been used for estimating sea ice and snow melt and freeze onset at various spatial resolutions ranging from 6.25

to 25 km (e.g. Howell et al. 2006; Markus et al. 2009; Wang et al. 2011, Brown et al. 2014). Additionally, multisensor approaches using microwave data have been used to estimate snow parameters over sea ice, such as combining synthetic aperture radar (SAR) and Moderate Resolution Imaging Spectroradiometer (MODIS) data to estimate snow thickness over first year ice (e.g. Zheng et al. 2017) and merging SAR (European Remote Sensing Satellite-1 (ERS-1) and RADARSAT-1) data for measuring snow-covered first year ice (e.g. Yackel et al. 2007). Though microwave observations have been successfully applied in snow and ice applications, single sensor microwave datasets suffer from inherent wavelength specific uncertainties and coarse spatial resolutions (Brown et al. 2014). Retrieval algorithms typically have trouble during the ice growth period in the fall when more new and young ice are present due to specular reflection and surface roughness changes (Howell et al. 2008a). Additionally, significant land contamination can occur between narrow channels when resolving sea ice conditions (Howell et al. 2006). Deriving snow cover information from passive microwave brightness temperatures is complicated by the physical structure of the snowpack and regional land cover differences that influence microwave emission and scatter (Wang et al. 2005). Other techniques include scatterometer backscatter observations which provide a higher spatial resolution (~ 2 to 5 km), however are only available from 2000 onward (Howell et al. 2019). Spaceborne active microwave SAR sensors offer increased spatial resolution (20 to 100 m), however inconsistent viewing geometries and limited image availability across the Arctic present challenges in monitoring snow and ice over large areas (Howell et al. 2006; Howell et al. 2019).

In the Canadian Arctic Archipelago, where narrow channels and multiple islands dominate the landscape, using coarse resolution data can be problematic as these sensors cannot resolve smaller- scale changes in snow and ice cover. An alternative approach in snow and ice mapping includes the use of the National Ice Center Interactive Multisensor Snow and Ice Mapping System (IMS). IMS utilizes a variety of multi-sourced datasets such as visible imagery, passive microwave data, and ancillary data (Ramsay 1998; Helfrich et al. 2007; Brown et al. 2014). Throughout this manuscript we imply snow cover to be snow cover on land, as IMS does not distinguish between snow-covered and snow-free sea ice. IMS has been primarily used in snow cover applications (e.g. Brubaker et al. 2005; Chen et al. 2012; Yu et al. 2017), as the daily temporal resolution and ability to penetrate cloud cover provide improved monitoring capabilities over visible imagery (e.g. MODIS Snow Cover product). IMS has received little attention in sea ice applications,

however, has been found to be advantageous over several automated algorithms available for monitoring sea ice phenology (Brown et al. 2014). The higher spatial resolution of IMS compared to the current passive microwave sea ice estimates contributes to the improvement of sea ice estimates by reducing land contamination and allowing coastal regions to be more accurately represented (Brown et al. 2014). The inclusion of microwave imagery in the IMS product reduces issues of cloud cover obstructions and polar darkness associated with visible imagery in the IMS product (Helfrich et al. 2007; Brown et al. 2014). This paper will examine spatial changes in sea ice and snow (on land) phenology in the Canadian Arctic Archipelago from 1997 – 2018, with a focus on improved spatial resolution of satellite-based observations. The objectives of this paper are to 1) provide an overview of spatial and temporal changes in sea ice and snow phenology and 2) investigate links between sea ice and snow phenology in the context of a warming climate.

2.2 Study Area

High Arctic Canada is comprised of multiple islands, waterways, and narrow channels, and is bordered by the Arctic Ocean, Beaufort Sea, Baffin Bay, and the Canadian mainland (Figure 2.1). The CAA is considered to be the world’s second largest high-Arctic landmass, covering an area of approximately 1.4 million km2 (Hund 2014). Islands located north of 74.5˚ N in the CAA (north of the Parry Channel) are collectively known as the Queen Elizabeth Islands (QEI) (Melling, 2002). Sea ice and snow cover duration persists for most of the year in the CAA. Sea ice within the CAA is predominantly landfast for six to eight months of the year, hindering wind driven sea ice movement in the CAA (Melling 2002; Howell et al. 2008b). Ice begins to break up in July and refreezes in October, though trends towards earlier melt onset and later freeze up have been reported (e.g. Markus et al. 2009). During the melt season ice becomes mobile and travels southeast across the CAA and thick MYI from the Arctic Ocean basin is transported into the CAA (Howell et al. 2008b). Arctic Ocean-CAA ice exchange from 1997 – 2018 has significantly increased and is associated with a longer open water season combined with thinning sea ice which facilitates faster ice speed and earlier (later) break-up (freeze-up) (Howell and Brady, 2019). The total minimum ice coverage at the end of the melt season exhibits interannual variability depending on thermodynamic and dynamic interactions (Howell et al. 2006). FYI generally dominates southern regions of the CAA; however, recently southerly regions are experiencing more MYI inflow due to increases in open water area providing more capacity for ice exchange into the CAA (Howell et al. 2006; Howell et al. 2013).

Snow cover over land and perennial sea ice surfaces typically persists for up to nine to ten months each year in the High Arctic (Mudryk et al. 2018). The snow melt season is short, and the duration of snow-free land cover typically persists for less than two months. Most winter precipitation occurs in a solid form due to the presence of saturated clouds and cooler temperatures (Dolant et al. 2017). During summer months, snowfall declines have been identified over the CAA and Arctic Ocean and are associated with changes in precipitation form (snow turning into rain) and lower- atmospheric warming (Screen and Simmonds 2012). Total precipitation has increased in the Canadian Arctic during all seasons as warmer temperatures increase atmospheric moisture content (Vincent et al. 2015; Zhang et al. 2019).

2.3 Data and Methodology 2.3.1 Data

The IMS product provides daily snow and ice cover maps of the Northern Hemisphere subjectively produced by analysts at 1 km (2014 – present), 4 km (2004 – present), and 24 km (1997 – present) spatial resolutions (U.S. National Ice Center 2008; Brown et al. 2014). Discrete values are assigned to land, snow covered land, water, and ice by utilizing a variety of multi-sourced datasets (for a complete list of data sources, see U.S. National Ice Center 2008). Visible imagery is preferred for mapping snow extent; however, analysts will use microwave data in the event that optical imagery is unavailable due to cloud occlusion or low solar illumination angles (Helfrich et al. 2007; Brown et al. 2010). Snow extent data derived from microwave sensors have well documented snow misidentification errors due to signal obstructions, snow grain size, and land cover influences that affect microwave emission and scatter (Chang et al. 1996; Foster et al. 2005; Helfrich et al. 2007; Derksen 2008; Brown et al. 2010). Alternitavely, analysts rely more on snow climatology to estimate high latitude snow cover during the winter months than microwave data (Helfrich et al. 2007). The accuracy of snow detection with the 4 km IMS product increases with snow depth and snow cover extent (Brubaker et al. 2005; Somnez et al. 2014). Ice cover analysis relies on a different approach than snow cover, where changes in winter ice cover are primarily noted using AVHRR or MODIS observations; though microwave-based retrievals and ice climatology are used when visible imagery is unavailable (Helfrich et al. 2007). Most high latitude regions are verified as being ice-covered using microwave-based data, representing approximately 30 to 35% of the winter and autumn ice cover input (Helfrich et al. 2007). For a complete description of the IMS product, see Helfrich et al. (2007) and National Ice Center (2008).

Temperature data were derived from the European Centre for Medium-Range Weather Forecasts (ECMWF) reanalysis data (ERA-Interim) and were compared to changes in sea ice and snow phenology (ECMWF 2011). Daily 2-m temperature was used to calculate monthly averages from 1997 – 2018 and 2004 – 2018. Although many reanalysis datasets are available for use, such as JRA-55 (Kobayashi et al. 2015), NCEP (Kanamitsu et al. 2002), ERA-40 (Uppala et al. 2005), ERA-5 (Hersbach et al. 2020), ERA-Interim provides coverage of the entire Canadian Arctic at a a sufficent spatial resolution (0.75º, ~79 km) to examine the large-scale climate patterns relevant for sea ice and snow cover changes at the scale being examined for this work. Monthly surface temperature and precipitation data at Cambridge Bay, Resolute, and Eureka were obtained from

the Environment Canada Second Generation of Homogenized Temperature (Vincent et al. 2012, see: https://www.canada.ca/en/environment-climate-change/services/climate-change/science- research-data/climate-trends-variability/adjusted-homogenized-canadian-data/surface-air- temperature.html) and Adjusted Precipitation for Canada datasets (Mekis and Vincent 2011, see: https://www.canada.ca/en/environment-climate-change/services/climate-change/science- research-data/climate-trends-variability/adjusted-homogenized-canadian-data/precipitation.html). Historical daily climate and weather data available from the Environment Canada website (Environment and Climate Change Canada, 2020) were used to supplement monthly homogoenized datasets and to validate the timing of snow-on and -off.

2.3.2 Methodology

Both the 24 km and 4 km IMS products were used to detect sea ice and snow phenology dates by comparing consecutive IMS images following the methodology of Brown et al. (2014). An algorithm was developed to iterate through each pixel of each consecutive pair of daily IMS images and note the date of the first and last transition from ice/snow to water/land for each pixel (and vice versa for the freeze parameters). The phenology parameters used and their subsequent definitions can be found in Table 2.1. The IMS products were used to identify both long-term (24 km, 1997 – 2018) and recent (4 km, 2004 – 2018) changes in sea ice and snow phenology. The 4 km product offers considerable improvements in spatial resolution compared to the 24 km product, therefore making finer-scale analysis of recent changes in snow and ice conditions possible, while the 24 km product provides an overview of long-term trends since 1997. Despite the lower spatial resolution of the 24 km product, the 24 and 4 km IMS products are generally in good agreement and are significantly correlated from 2004 – 2018 (Table 2.2). Differences between phenology dates detected by the IMS products are likely due to the improved ability of the 4 km product to resolve smaller-scale sea ice features, such as leads and polynyas, and improved detection of sea ice parameters along coastal regions (Brown et al. 2014). In addition to spatial resolution, interannual and regional variability in sea ice and snow conditions will inherently affect phenology parameters detected by both IMS products, particularly for sea ice. As sea ice in the CAA does not always entirely melt or clear out of the channels, open water parameters detected using IMS are limited to pixels in which ice breaks-up and/or clears from the channels during the melt season. In regions where ice did not clear in a given year, no open water parameters would be detected.

Table 2.1. Sea ice and snow phenology parameters and definitions used in this study. Parameter Definition First open water The first change from ice to water for a given pixel Continuous open water The last change from ice to water, signalling ice-free conditions for the remainder of the season Freeze onset The first detection of ice for a given pixel Continuous ice cover The date of the last change from water to ice, signalling ice-covered conditions for the remainder of the season First snow-off The first change from snow-covered land to snow-free land for a given pixel Final snow-off The last change from snow-covered to snow-free land, signalling snow-free conditions for the remainder of the season First snow-on The first change from snow-free land to snow-covered land Final snow-on The last change from snow-free to snow-covered land, signalling snow-covered conditions for the remainder of the season

Table 2.2. Spearman rank correlation coefficient (ρ) for 24 km and 4 km IMS products (2004 - 2018). Bold indicates significance at 95%.

ρ p-value First open water 0.9588 < 0.000 Continuous open water 0.9294 < 0.000 Freeze Onset 0.9941 < 0.000 Continuous ice cover 0.9882 < 0.000 First snow-off 0.9852 < 0.000 Last snow-off 0.9647 < 0.000 First snow-on 0.9941 < 0.000 Last snow-on 0.9529 < 0.000

Before calculating the non-parametric Spearman rank coefficients (ρ) for phenology parameters and monthly temperatures, the time series data were detrended using the “Pracma” package in R (https://CRAN.R-project.org/package=pracma) to ensure the relationships were driven by variability in the parameters rather than shared trends (Pizzolato et al., 2014). To detrend the time series data, the linear trend was removed by computing the least-squares fit of a straight line and subtracting the resulting function from the data (Borchers, 2019).

To evaluate temperature and changes in sea ice and snow phenology, 2-m temperature and phenology parameters were analyzed using the “zhang” method of climate trend analysis, available in the “zyp” package in R (Bronaugh and Werner 2019). The “zhang” method of trend analysis has been used successfully for representing trends in temperature and precipitation (Zhang et al.

2000) and lake ice phenology events (Murfitt and Brown 2017). This method is appropriate for spatial trends as it accounts for autocorrelation and employs non-parametric tests. If a significant linear trend is present in the data it is removed and the autocorrelation is computed repeatedly until the difference in the estimates of the slope and autoregressive model in two consecutive iterations is smaller than 1% (Bronaugh and Werner, 2019). The resulting time series is analyzed with the Mann-Kendall test, and Sen’s slope is then computed to indicate the amount of increase or decrease for each pixel over the time period examined (and significance of this trend) (Bronaugh and Werner, 2019). A complete description of the “zhang” method is described in Wang and Swail (2001).

For the 2004 – 2018 period, the trend detection method was applied to temperature and phenology parameters to detect regional changes and identify areas of warming (cooling) and earlier (later) snow and ice-off (snow and ice-on) dates. Phenology trends for each pixel were computed if at least 13 of the 15 years in the data record identified an ice/snow off date. Regions that experience only occasionally melt (e.g. sea ice to the north of Bathurst Island) are therefore excluded from the analysis and considered No Data, resulting in the spatial extent for phenology trends representing only the area that experienced melt in 13 or more years. The subsequent extent of the analysis hence represents a smaller geographic region than a single given year might experience from between 2004 – 2018.

In order to examine the spatial dependence of phenology parameters, local Moran’s I was computed for correpsonding melt and freeze parameters (Moran, 1950). Moran’s I provides an index of spatial autocorrelation that differs from typical correlation indices as it considers the correlation between variables with respect to location (Howell et al. 2006). The resulting indices provide information on whether variables with similar values are located near each other, or whether neighbouring locations have dissimilar values (Howell et al. 2006).

2.4 Results and Discussion 2.4.1 Temporal Variability and Links Between Sea Ice and Snow Phenology 2.4.1.1 Sea Ice Phenology

Mean open water and freeze dates for the 24 km (1997 – 2018) and 4 km (2004 – 2018) IMS products in the CAA are shown in Figure 2.2a and b. The earliest and latest open water and freeze timing detected by both IMS products are shown in Table 2.3. Overall, there is good agreement between the detected annual freeze and thaw timing identified by the two products (see Figure 2.2, 1-2 day mean difference in dates, mainly attributed to resolution). The 4 km IMS product identified the same earliest / latest years as the 24 km product, looking at the time series only after 2004 (with the exception of continuous ice cover in 2006). The following examination of the phenology will focus mainly on the 24 km product as it has the longer time series.

Mean open water dates range from July 6 (day 187) to August 26 (day 238) and freeze dates range from September 30 (day 273) to October 26 (day 299). While the date of first open water detection in the CAA is shifting earlier, sea ice in the CAA is exhibiting trends towards later continuous open water and earlier freeze dates during the 1997 – 2018 period (Figure 2.2a, b). First and continuous open water both show statistically significant negative correlations with 2-m temperature in July (ρ = -0.50 and -0.64, p < 0.05; Table 2.4). In August and September, correlation coefficients are weaker than July (p > 0.05). Freeze onset and continuous ice cover are both significantly correlated to 2-m temperatures in September (Table 2.4; ρ = 0.68 and 0.58, p < 0.05), with weaker correlations to October 2-m temperatures (ρ = 0.41 and 0.39, p > 0.05).

Figure 2.2. Mean 24 km (1997 – 2018) solid line and 4 km (2004 – 2018) dotted line IMS first open water and continuous open water (a), freeze onset and continuous ice cover (b), first snow-off and final snow-off (c), and first snow-on and final snow-on dates (d) for the Canadian Arctic Archipelago. Sen’s slope and significance are indicated for the 24 km product. IMS data were downloaded from the U.S. National Ice Center (2008).

24 km IMS 4 km IMS 24 km IMS 4 km IMS Earliest Earliest Latest Latest First open water 1998 (187) 2010 (198) ** 2002 (223) 2004 (222) ** Continuous open water 1998 (208) 2011 (220) ** 2018 (233) 2018 (236) Freeze onset 2018 (278) 2018 (277) 1998 (299) 2012 (293) ** Continuous ice cover 2014 (279) 2014 (280) 1998 (300) 2006 (301) First snow-off 2012 (174) 2012 (173) 2004 (202) 2004 (202) Final snow-off 2011 (175) 2011 (175) 2013 (205) 2013 (205) First snow-on 2013 (236) 2013 (236) 1998 (268) 2007 (259) ** Final snow-on 2013 (246) 2013 (245) 2006 (268) 2006 (267)

Table 2.4. Spearman rank correlations (ρ) for snow and sea ice phenology dates, 2-m temperature (T), and open water and snow-free duration from 1997 - 2018. Bold indicates significance at 95%.

ρ p-value First open water and July 2-m T -0.5020 0.019 First open water and August 2-m T -0.2196 0.325 First open water and September 2-m T -0.2569 0.247 Continuous open water and July 2-m T -0.6431 0.002 Continuous open water and August 2-m T -0.2851 0.198 Continuous open water and September 2-m T -0.1485 0.508 Freeze onset and September 2-m T 0.6883 0.001 Freeze onset and October 2-m T 0.4150 0.056 Continuous ice cover and September 2-m T 0.5855 0.005 Continuous ice cover and October 2-m T 0.3970 0.068 First snow-off and June 2-m T -0.7459 0.000 First snow-off and July 2-m T -0.8035 0.000 Final snow-off and June 2-m T -0.6928 0.000 First snow-off and July 2-m T -0.7357 0.000 First snow-on and August 2-m T 0.6680 0.001 First snow-on and September 2-m T 0.6894 0.001 Final snow-on and August 2-m T 0.7787 0.000 Final snow-on and September 2-m T 0.8464 0.000 First open water and First snow-off 0.4082 0.060 Continuous open water and Final snow-off 0.6364 0.002 Freeze onset and First snow-on 0.4455 0.039 Continuous ice cover and Final snow-on 0.5686 0.007 Open water duration and snow-free duration 0.6070 0.003

2.4.1.1.1 Ice-off

First open water is shifting 2.1 d decade-1 earlier and continuous open water is shifting 3.5 d decade-1 later (though, neither are statistically significant) (Figure 2.2a). 1998 represents the earliest first open water and continuous open water dates detected over the 22-year record (Table 2.3), consistent with low ice conditions and anomalously warm air temperatures in the region (Figure 2.3; Howell et al. 2006; Howell et al. 2009; Mudryk et al. 2018). Air temperatures in 1998 were 1.8ºC higher in June, 0.9 ºC higher in July, 0.8 ºC higher in August, and 3.0°C higher in September, compared to the 1997 – 2018 average (Figure 2.3).

The latest first open water dates were detected in 2002 (24 km product) and in 2004 (4 km product), both occurring in early August, with only 3 days difference between the two years. Temperature anomalies are slightly more negative for June through September in 2004 compared to 2002. However, the later opening of the Cape Bathurst and Lancaster Sound polynyas in 2002 compared to 2004 (July 24 (205) in 2002 vs. June 21 (173) in 2004) contributed to the overall later mean date in 2002.

2018 represents the latest continuous open water dates (Table 2.3), consistent with negative temperature anomalies identified in the CAA from June to September (Figure 2.3). The longest duration from first open water to continuous open water was also observed in 2018 (25 days for 24 km IMS; 29 days for 4 km IMS), indicating that the CAA remained ice covered longest during the 2018 melt season over the 22-year record. Occurrences of later open water dates in 2002 and 2004 coincide with a period of heavy MYI conditions from 2000 – 2004 in the CAA due to dynamic ice import and in situ growth of MYI (Howell et al. 2008a). When sea ice in the CAA is

not landfast, it is exchanged with the Arctic Ocean to the north and west, a process that primarily occurs during the summer months (Howell et al. 2013). The export of MYI from the Arctic Ocean into the CAA provides the CAA with an additional source of MYI that complements FYI aging, maintaining relatively stable sea ice conditions within the CAA, even during the summer months (Howell et al. 2015). The weak trend toward earlier first open water (2.1 d decade-1) can be explained by this presence of stable, replenished ice cover during summer months. Contrarily, continuous open water dates exhibit a negative (later) trend (3.5 d decade-1), implying increased ice mobilization within the High Arctic channels. Howell and Brady (2019) show that the mean ice area flux between the Arctic Ocean and QEI gates has significantly increased since 1997 by 103 km2 per year. Earlier first open water dates may be facilitating the capacity for ice exchange to occur, subsequently resulting in later continuous open water dates identified within the CAA.

2.4.1.1.2 Ice-on

Both freeze onset and continuous ice cover are shifting earlier, 5.4 d decade-1 earlier and 3.9 d decade-1 earlier respectively (p < 0.05 for both), with much shorter duration between freeze onset and complete freeze than identified in the earliest/latest open water dates (Figure 2.2b).

2018 was identified as the year of earliest freeze onset and 2014 as the year of earliest continuous ice cover (Table 2.3). Heavier ice conditions during these years can be attributed to both cooler than usual temperatures (Figure 2.3), particularly in the early summer, and slightly delayed ice-off dates (continuous open water in particular). Sea ice in 2014 was likely influenced by a heavier than usual concentration of MYI already in the CAA from a large replenishment in 2013 (Howell et al. 2015). The heavy end-of-summer ice conditions and short open water duration during September favoured earlier 2014 freeze, as highlighted in Figure 2.4. The Western Parry Channel, M’Clintock Channel, and QEI were almost entirely ice-covered on September 1 (244) and then began to break- up on September 6 (249). The three regions began to see ice cover advance after September 20 (263) and were completely ice-covered again by September 27 (270). Howell and Brady (2019) note that the QEI exchange gates did not collapse (with the exception of the Peary Channel), resulting in negligible ice exchange between the Arctic Ocean and CAA during 2014. Despite a low ice exchange year, ~90% of the total ice area flux into the CAA through the M’Clure Strait in 2014 was MYI, which further impedes melt and facilitates earlier freeze (updated from Howell et al. 2013, Howell 2018 personal communication).

The latest freeze onset and continuous ice cover dates detected were in 1998 (24 km product), corresponding to the earliest open water dates detected by the 24 km IMS product (Table 2.3). The latest freeze dates detected by the 4 km product were in 2012 (freeze onset) and 2006 (continuous ice cover), with the 4 km product identifying a slightly later complete freeze in 2006 than 1998 (Table 2.3). Some of the largest temperature anomalies were identified in August and September during these years (and October in 2006) (Figure 2.3) and correspond to strong positive surface air temperature anomalies discussed by Howell et al. (2015) during 1998, 2006, 2008, 2010, 2011, and 2012 in the CAA. The mean freeze onset date in 1998 is 8 to 9 days later than in 2006, while the continuous ice cover dates are closer (i.e. within 3 to 5 days of each other). The duration from onset to total ice cover is ~ 6 (24 km IMS) to 10 (4 km IMS) days (mean) in 2006, but in 1998 it is ~ 1 day (mean). Although continuous ice cover resumed at a comparable mean date in 1998 and 2006, the anomalously high air temperatures in October 2006 reflect the longer time required for ice to completely freeze over in 2006. Temperature anomalies were near 0°C in October 1998, therefore the latest freeze dates detected in 1998 can be attributed to anomalously warm air temperatures during the summer months combined with restricted MYI inflow from the Arctic Ocean (Howell et al. 2010; Howell et al. 2013). While the 2006 final freeze date was one of the latest in the time series, the offset of +3.5 days between the 24 km and 4 km products was a result of several regions that opened/re-froze after first freeze, that the 24 km imagery could not resolve (e.g. the flaw lead near the western side of the study region), highlighting how finer-resolution satellite data can improve estimates of sea ice parameters (Figure 2.5).

2.4.1.2 Snow Phenology

Mean snow-off and snow-on dates in the CAA are shown in Figure 2.2c and d, with the earliest and latest dates detected by both IMS products shown in Table 3. Overall, there is slightly better agreement between the annual snow-on and off dates identified by the two products compared to sea ice (see Figure 2, 1-day average difference vs. 1 – 2 day average difference for sea ice). Mean snow-off dates range from June 21 (172) to July 24 (205) and snow-on dates range from August 24 (236) to September 25 (268). Snow cover in the CAA is exhibiting trends towards earlier snow- off and earlier snow-on during the 1997 – 2018 period (Figure 2.2c, d). Snow-on and -off dates show stronger relationships to 2-m temperatures compared to sea ice, with the lowest correlation coefficient being 0.66 and all values being statistically significant (Table 2.4). First and final snow- off show strong negative correlations with June (ρ = -0.74 and -0.69, p < 0.05) and July (ρ = -0.80 and -0.73, p < 0.05) 2-m temperatures (Table 2.4). First and final snow-on trends also show statistically significant correlations to 2-m temperatures in August and September (ρAugust = 0.66 and 0.77; ρSeptember = 0.68 and 0.84; p < 0.05 for both months).

2.4.1.2.1 Snow-off

First snow-off is shifting 6.2 d decade-1 earlier (p > 0.05) and final snow-off is shifting 0.2 d decade-1 later (p > 0.05) (Figure 2.2c). Trends toward an earlier snow-free season are consistent with reductions in Arctic snow mass trends in the Northern Hemisphere, which have decreased by 49± 49 Gt decade-1 since 1980 (Pulliainen et al. 2020).

The earliest first snow-off date occurred in 2012 (Table 2.3). Although the earliest first open water date was detected in 1998, temperature anomalies in June and July 2012 exceed those in 1998 (Figure 2.3). 2012 was characterized by a negative North Atlantic Oscillation phase, which is associated with enhanced southerly air flow into the Arctic which contributes to warmer temperature anomalies and rapid snowpack ablation (Derksen and Brown 2012).

The earliest final snow-off dates occurred in 2011 (Table 2.3). Figure 2.2c shows that the mean first snow-off and mean final snow-off dates in 2011 are the same, indicating that no subsequent snowfall occurred during the melt season (note that this does not imply that all of the snow in the CAA melted on the same day, as the values used here are mean values for the entire study area). Although 2012 saw snow cover come off earlier than 2011, the time to snow-free conditions in 2012 was much longer (17 days) in 2012 compared to 2011. Within the study area, the mean duration from first to final snow-off in 2011 was 0.4 days, with 97% of the area experiencing complete snow retreat on the first snow-off day (i.e. no difference between the first and final snow- off dates). In 2012 only 69% of the snow retreated on the first snow-off day, which can be attributed to a snow-fall event that occurred in mid-August over parts of Banks, Melville, Somerset, Baffin, Cornwallis, and Devon Islands (Figure 2.6).

The latest first snow-off date was detected in 2004 (Table 2.3), with the mean first snow-off date for the southern CAA on July 15 (197), which is considerably later than other years which also exhibit cooler temperatures. In 2013 and 2014, the mean first snow-off dates detected in the southern CAA were June 28 (179, 2013) and July 4 (185, 2014), approximately two weeks earlier than in 2004. Mean first snow-off dates in the northern CAA were also later in 2004 (10 to 11 days) compared to 2013 and 2014, though the interannual variability is not as large in this region. The lowest temperature anomaly over the 22-year record during July occurred in 2004 (Figure 2.3; -1.82ºC), which may explain the later snow-off dates detected in 2004.

The latest final snow-off date occurred in 2013 (Table 2.3) and are consistent with cold spring temperatures. Trishchenko and Wang (2018) report cold temperatures and short melting periods over the CAA in July 2013 and suggest that such conditions could lead to incomplete melt and maintain the snowpack over the entire warm season. Young et al. (2018) note the persistence of a late-lying snowbed at Nanuit Itillinga (formerly Polar Bear Pass, Bathurst Island, NU) until July 9 (190) in 2013 due to a cold spring, whereas in 2012 (which was exceptionally warm) the snowbed completely disappeared by July 1 (183). Spatial patterns of snow cover extent across Bathurst

Island and Cornwallis Island in 2013 derived from the 500-m MODIS snow product show that snow cover began to retreat on July 4 (185) and was almost entirely snow-free on July 21 (202) (Young et al. 2018). The final snow-off dates from the 4 km IMS product yield similar results, with final snow-off dates ranging from July 3 to July 29 (184 – 210). This illustrates the utility of the 4 km IMS product, despite the coarse resolution compared to the 500-m MODIS snow product.

2.4.1.2.2 Snow-on

First snow-on is shifting 5.4 d decade-1 earlier (p < 0.05) and final snow-on is shifting 2.6 d decade- 1 earlier (p > 0.05) (Figure 2.2d). Trends toward an earlier snow-on season are consistent with increases in precipitation identified over northern Canada across all seasons (Mekis and Vincent 2011; Vincent et al. 2015). Trends in snowfall ratio, which reflect the combined effect of temperature and precipitation, have also increased in northern Canada since 1948 (Vincent et al. 2015).

The earliest first and final snow-on dates occurred in 2013 in early September (Table 2.3). Although overall snowfall recorded in 2013 in August (Figure 2.7a) and September (Figure 2.8b) was generally low relative to other years, cooler conditions (Figure 2.3) likely inhibited snowmelt and allowed the early snow cover to persist during the snow-on period. Early snowfall in 2013 is consistent with Young et al. (2018) who found earliest snow onset in 2013 on Bathurst and Cornwallis Islands from 2000 – 2013. IMS imagery (not shown) illustrates that in the CAA, approximately 89% of the first snow-on days in 2013 occurred between August 17 – 31 (229 – 243). Daily snowfall and air temperature measurements recorded at Eureka, Resolute, and Cambridge Bay reveal that in 2013 temperatures were higher during the first half of August, followed by a noticeable decline during the last half, accompanied by increased snowfall at each location (Figure 2.7a). This aligns with the anomalously low temperature and pressure for the second half of August across much of the CAA (Figure 2.7b).

The latest first snow-on date detected was in 1998 (Table 2.3). First snow-on in 1998 occurred 13 ± 6 days later than all other years over the 22-year IMS record, however it should be noted that first snow-on in 2002 (September 23, 266) is only two days earlier than in 1998 (September 25, 268). Higher annual air temperatures and snowfall were recorded in 1998 at each station location compared to 2002, with the exception of snowfall in Eureka which was 44 cm in 1998 compared to 52.3 cm in 2002 (Figure 2.8a). During September in 1998, snowfall amounts recorded at Cambridge Bay and Eureka were near 0, while 11 cm of total snowfall during September were recorded in Resolute (Figure 2.8b).

The latest final snow-on date occurred in 2006 in late September. Temperature anomalies were positive during September (Figure 2.3) and total annual snowfall recorded at Cambridge Bay and Eureka was relatively low (Figure 2.8a). Total September snowfall at each station was also low (0.2 cm at Cambridge Bay, 1 cm at Resolute, and 4.2 cm at Eureka) (Figure 2.8b). MODIS imagery and field measurements recorded by Young et al. (2018) also show that snow cover was late in 2006 at Nanuit Itillinga (formerly Polar Bear Pass, Bathurst Island, NU) and that some small areas were still snow-free on the island by the end of September. IMS imagery shows that in 2006 most of the islands north of the Parry Channel were snow-covered earlier than regions to the south, which did not become completely snow-covered until late September to early October.

2.4.1.3 Links in Sea Ice and Snow Phenology

All sea ice and snow phenology dates show a positive and statistically significant correlation to their corresponding phenology parameter, with the exception of first open water and snow-off (Table 2.4). The weakest relationship was identified between first open water and first snow-off (ρ = 0.40, p > 0.05), while the strongest relationship was identified for continuous open water and final snow-off (ρ = 0.63, p < 0.05). Freeze onset and first snow-on dates showed a weaker relationship (ρ = 0.44, p < 0.05) compared to continuous ice cover and final snow-on dates (ρ = 0.56, p < 0.05). It is interesting to note that the strongest relationships were observed for the final snow/ice on or off phenology parameter (i.e. the last occurrence of melt or freeze), suggesting the ‘last occurrence’ parameters may best capture the cumulative climatic effects driving both snow/ice phenology parameters.

Despite evidence of long-term warming in the Arctic (e.g. Box et al. 2019; Zhang et al. 2019), we demonstrate that the Canadian Arctic Archipelago, as a whole, is responding differently to

warming compared to other regions of the Arctic. While showing substantial annual variability, this region continues to maintain relatively persistent ice conditions over the 1997 – 2018 period. The positive trend (p > 0.05) identified for continuous open water dates (3.9 d decade-1) and the significant negative trend in freeze onset dates (-5.4 d decade-1, p < 0.05), result in open water duration decreasing by 8 d decade-1 (p < 0.05) from 1997 – 2018 (Figure 2.9). While first open water will be later than sea ice melt onset, the weak negative trend in first open water dates (-2.91 d decade-1, p > 0.05) is consistent with findings from previous studies, highlighting that no significant trends in melt onset dates were found in the CAA (e.g. Mahmud et al. 2016; Marshall et al. 2019); while other Arctic regions (e.g. Baffin Bay, Eastern Greenland, Barents Sea, Beaufort Sea, and Chukchi Sea) have shown significantly earlier melt onset by 2.3 to 6.9 d decade-1 (e.g. Stroeve et al. 2014).

Multiple studies have shown that warming across the Arctic is a key driver of the lengthening snow-free season and decreased snow cover extent at the pan-Arctic scale (e.g. Brown et al. 2010; Derksen et al. 2012; Hernandez-Henriquez et al. 2015; Thackeray and Fletcher 2016). However, taking into account final snow-off and first snow-on dates, snow-free duration in the CAA has decreased by 4.2 d decade-1 (p > 0.05) from 1997 – 2018 (Figure 2.9). This shift toward a shorter snow-free season reflects changes in snow onset (i.e. earlier snow-on) identified. Although larger (insignificant) trends towards earlier snow-off dates were identified in the CAA, occurrences of snowfall events after first snow-off likely minimize the magnitude of trends detected in the final snow-off dates compared to the initial snow free conditions (-0.2 d decade-1, p > 0.05).

Further linkages between sea ice and snow phenology will be examined in Section 2 using the 4 km IMS product as the higher spatial resolution can resolve finer-scale changes in snow and ice cover and provides considerably improved detail over the 24 km product in the complex network of islands and channels of the CAA (e.g. Figure 2.5).

2.4.2 Regional Variability and Links in Sea Ice and Snow Phenology

2.4.2.1 Melt Season

With the acknowledgment that short timespans and both interannual and spatial variability can inflate trends, the spatial variability in sea ice and snow phenology, from 2004 to 2018, was calculated using the 4 km IMS product (Figure 2.10). The largest trends towards earlier open water (first open water) are in the Amundsen Gulf, part of the Western Parry Channel, and Prince Regent Inlet, with median decreases ranging from 3 to 26 days over the study years (Figure 2.10a). Earlier first open water dates were identified along the eastern coast of the Boothia Peninsula, where the mean decrease was 20 days and the median decrease was 22 days over the 15-year record. Many of the larger shifts in first open water timing identified in these regions can be tied to the presence of polynyas, which typically become ice free earlier than surrounding areas. Snow cover on the Boothia Peninsula follows a similar pattern to the adjacent sea ice, with significant clustering toward earlier snow and ice off shifts identified in this region (Figure 2.10a).

The largest trends toward earlier first snow-off dates are on Victoria Island, with a mean and median of 13 and 14 days earlier, while final snow-off changes are similarly early. Similar trends toward earlier open water dates are also apparent in the western CAA, where the median trend through the Western Arctic Waterway (adjacent to Victoria Island) is 13 days earlier. Figure 2.10a shows significant clustering of earlier first snow-off and open water shifts throughout the Western Arctic Waterway and on Victoria Island. These shifts in timing are also consistent with strong warming in June, July, and August over the region (Figure 2.11a).

On Banks Island there is a trend toward later final snow-off dates, particularly in the northern and eastern regions where the median final snow-off shift is 5 days later (Figure 2.10b). Similar clusters are identified in final snow-off and continuous open water trends in the Banks Island region, where significant clustering toward later snow and ice off are present along the northern and eastern coast (Figure 2.10b). This may be related to later ice break-up in the western High Arctic, as the eastern regions typically break up earlier than the west, whereas some western regions do not experience any break-up during summer (see Figure 2.12) (Melling 2002; Tivy et al. 2011).

Figure 2.12. 4 km IMS continuous open water dates for (left) 2004 (colder year) and (right) 2011 (warmer year) in the Canadian Arctic Archipelago. White regions represent no detected ice break up. Base map created using ESRI ArcMap version 10.6 and Natural Earth land vectors (Natural Earth 2020). IMS data were downloaded from the U.S. National Ice Center (2008).

Within the QEI, a trend towards later first open water and first snow-off dates dominates the northeastern regions, though there is considerable regional variability. Overall, first open water and first snow-off trends show significant clustering toward later ice/snow off in this region, with some areas of earlier ice/snow off identified on the northern coast of Ellesmere Island (Figure 2.10a). Sea ice within the Ellesmere Island region shows a shift in timing towards later first open water dates, with the mean and median of 13 and 15 days later. First snow-off patterns on Ellesmere are variable, with areas in the south shifting towards later snow-off while areas around the northern coastline are shifting toward earlier snow-off. Earlier snow-off dates in the northern region correspond with increasing air temperatures identified in June and July (Figure 2.11a, b). Land surface temperatures during the summer over the Lake Hazen watershed, located on northern Ellesmere Island, have increased by 2.6ºC from 2000 – 2012 (Lehnherr et al. 2018). To the west of Ellesmere and Axel Heiburg, Ellef Ringnes and Amund Ringnes Islands are dominated by later snow-off dates, with first snow-off dates getting later by an average of approximately 11 days (Figure 2.10a). In the south-eastern QEI region, first open water and first snow-off dates also show positive trends, with snow coming off later on Cornwallis and Devon Islands, and sea ice coming off later in the Jones Sound and part of the Eastern Parry Channel.

Variability in first open water trends in the Eastern Parry Channel can be identified in Figure 2.10a, where significant clusters of earlier and later ice-off dates are present. First open water trends are getting later north of Somerset Island (median = 13 days) and earlier north of Prince of Wales and Baffin Islands (median = 9 and 13 days, respectively). This may be due to the presence of the Viscount Melville Sound and Lancaster Sound Polynyas, as polynyas generally have low average sea ice thickness, making them more susceptible to earlier melt (Hannah et al. 2009). The cluster of earlier open water dates north of Prince of Wales also correspond to a cluster of earlier first snow-off shifts identified on the northern region of the island.

Unlike first open water, a shift towards later continuous open water dates is shown throughout the entire Eastern Parry Channel (Figure 2.10b). Median dates in this region are high, at 45 days later, as this region is a source of MYI transport into Baffin Bay throughout the open water season, leading to highly variable open water dates (Tang et al. 2004). Continuous open water is also becoming later in the Prince Regent Inlet (median = 10 days), which is where earlier first open water dates were detected (Figure 2.10a, b). Larger ice area fluxes from the Arctic Ocean into the CAA from 1997 – 2018 may be responsible for the later continuous open water dates in these regions, as larger fluxes are associated with longer flow duration and faster moving ice (Howell and Brady, 2019).

Trends in final snow-off timing appear to be largely positive (later) compared to first snow-off dates throughout the QEI region over the 15-year record. In the eastern QEI, significant clustering of later final snow-off and continuous open water trends was identified (Figure 2.10b). On Devon and Cornwallis Islands, the median first snow-off shift was 7 days, while the median shift in final snow-off was 15 days. In the western QEI (e.g. Melville, Bathurst, Prince Patrick), larger positive (later) trends in final snow-off dates are identified where first snow-off trends were either negative (earlier) or negligible (no change), suggesting that snowfall likely occurs in this region after the first snow-off date.

2.4.2.2 Freeze Season

Sea ice freeze and snow-on trends (Figure 2.10c, d) exhibit similar spatial patterns to open water and snow-off for the most part (Figure 2.10a, b), with the regions showing earlier melt also displaying later freeze and vice versa. Generally, later freeze/snow-on trends (earlier open water/snow-off) were detected in the south-western CAA, and earlier freeze/snow-on trends (later

open water/snow-off) were detected in the eastern CAA. Compared to snow and ice off shifts, there appears to be more spatial variability observed in the clusters of snow-on and freeze in the eastern QEI, whereas during the melt season, this region was largely dominated by clustering toward later melt.

In the western CAA, temperature trends are positive (warmer) during September and October, though not as strong in October (Figure 2.11d, e). In the Western Arctic Waterway, a small shift of 1 day later (mean) for freeze onset was detected (Figure 2.10c). Later freeze was identified in Franklin Bay and Darnley Bay (median = 7 and 8 days, respectively) and in Coronation Gulf (median = 9 days). Mean shifts in continuous ice cover timing (1 day, Figure 2.10d) in the Western Arctic Waterway are similar to freeze onset, as sea ice typically refreezes rapidly following the sea ice extent minimum in September (Onarheim et al. 2018).

The central region of Victoria Island shows a small trend toward later snow-on (1 to 4 days; Figure 2.10c) and is predominantly cooling by 0.5 to 1.5ºC. Trends toward earlier first snow-on (median = 9 days) dominate the majority of Victoria Island (particularly coastal regions), however temperature trends are predominantly warming in these regions. Significant clusters of earlier freeze onset and first snow-on dates were also identified in these regions. Stronger warming (0.3 to 2.6ºC) is present over Banks Island in September, yet the region shows significant clustering toward earlier snow-on and even larger negative (earlier) first snow-on trends (median = 20 days) compared to Victoria Island. The shift toward earlier first snow-on in regions exhibiting strong warming is likely explained by both changes in temperature and precipitation (as more water vapour would be present in the warmer air and hence the potential for precipitation). Earlier snow- on dates detected in the CAA are likely reflecting changes in the Arctic hydrological cycle. Increases in precipitation have been identified across all seasons in northern Canada from 1948 – 2012 (Vincent et al. 2015). Global climate models show that projected increases in Arctic precipitation over the twenty-first century are due mainly to intensification of local surface evaporation resulting from retreating sea ice and Arctic warming (Bintanja and Selten 2014). The associated increase in Arctic precipitation may result in net accumulation over ice sheets (Singarayer et al. 2006; Bintanja and Selten 2014), however recent projections show a decrease in average annual Arctic snowfall and a shift towards a rain-dominated Arctic, particularly during summer months (Bintanja and Andry 2017).

Freeze onset and continuous ice cover trends are largely negative (earlier) in the eastern CAA (Figure 2.10c, d). A trend towards cooler temperatures is present over Baffin Island in September (Figure 2.11d) followed by wide-spread cooling over the eastern CAA in October (Figure 2.11e). In the south-eastern CAA, the median shift toward earlier freeze onset is 22 days. In the Baffin Inlets and Eastern Parry Channel, median freeze onset trends detected were 20 and 26 days earlier while median continuous ice cover trends detected were 18 and 23 days earlier. The timing of sea ice freeze shifts shows distinct east-west differences compared to the open water season, where the shifts of first and continuous open water are highly variable across the study area (Figure 2.10a, b).

Overall, snow-on generally progresses from the northeast to the southwest, however considerable spatial variation is present in some years. The variation observed during the snow-on season may be attributed to the strong local and regional controls on snow cover. Snowfall typically begins to accumulate towards the end of August and is influenced at the local scale by the amount and type of precipitation, blowing snow transport and sublimation, and variability in terrain (Callaghan et al. 2011). At the regional scale, physiographic and climatic factors influencing snow cover are elevation, spatial distribution of freezing temperatures, and location of typical cyclone tracks carrying moisture to the region (Callaghan et al. 2011). In exposed regions of the Arctic, such as Bathurst Island, snow depth and properties can exhibit strong local variation (Callaghan et al. 2011; Young et al. 2018). At a larger scale, snow onset patterns on Bathurst Island appear to be influenced by sea ice cover, with snow onset progressing from north (where sea ice typically persists throughout the melt season) to south (where open water is typically present until early October) during September (Young et al. 2018). Within the southern CAA, some regions (e.g. Banks Island, Victoria Island, Baffin Island) exhibit earlier first snow-on dates (Figure 2.10d) and later final snow-on dates (Figure 2.10c). Warmer temperatures identified in September (Figure 2.11d) during the snow onset period may result in earlier snowfall accumulation due to the greater amounts of moisture present in the warmer air; however, snow cover may be unable to persist under warm conditions which facilitate melt.

2.5 Conclusion

This is the first paper to show the utility of finer resolution satellite observations to investigate changes in sea ice and snow phenology together in the Canadian Arctic Archipelago, which is

anticipated to be the “Last Ice Area” in the Arctic, from 1997 – 2018. Using the Interactive Multisensor Snow and Ice Mapping System, we examined the spatial and temporal changes in sea ice and snow phenology at both the 24 km (1997 – 2018) and 4 km (2004 – 2018) resolutions in relation to temperature conditions. The CAA is a unique sea ice region as there is considerable spatial variability due to portions of open water (where ice can circulate freely) contrasted with narrow channels and multiple islands (where landfast ice dominates for most of the year) (Derksen et al. 2019). Compared to other Arctic regions, declines in sea ice area from 1968 – 2016 are not as large (-4.8% decade-1 in the CAA, -8.3% decade-1 in the Beaufort Sea, -11.4% decade-1 in Baffin Bay, -10.8% decade-1 in Hudson Bay) (Derksen et al. 2019). We present recent trends from 1997 – 2018 within the CAA towards later continuous open water (-3.5 d decade-1) and earlier freeze onset (-5.4 d decade-1) suggesting that ice cover duration in the CAA is increasing, and earlier first snow-on trends (-5.4 d decade-1) indicate longer snow cover duration. The strongest correlations were observed between continuous open water and final snow-off (ρ = 0.63, p < 0.05) and continuous ice cover and final snow-on (ρ = 0.56, p < 0.05) dates. Statistically significant clusters of earlier (later) snow and ice off trends, as well as earlier (later) snow and ice on trends were identified for all phenology parameters, indicating linkages between snow and ice phenology in the CAA. Regionally, significant clustering of earlier open water and snow-off trends were identified in the Western Arctic Waterway and on Banks and Victoria Islands. Snow cover appears to be persisting on the ground for a longer duration during the melt season in the western CAA regions, as indicated by earlier first snow-off and later final snow-off shifts. Earlier snow-on and freeze-onset were identified in the Queen Elizabeth Islands, though considerable variability in snow-on dates was apparent throughout the eastern QEI. Less interannual variability was apparent during the freeze season, as freeze is largely governed by cooling atmospheric temperatures; whereas during the melt season, the amount of total absorbed solar energy and snow-ice-albedo feedbacks are strongly linked to the timing of melt.

Changes in the cryosphere have direct feedbacks to the global climate system; therefore, understanding the processes and interactions of the cryosphere in the context of a warming climate is essential for characterizing the current state of the cryosphere and identifying patterns of variability and change. Recent assessments have indicated that dramatic changes are occurring as a result of rising surface air temperatures (Lemke et al. 2007; Derksen et al. 2012; Hernandez- Henriquez et al. 2015; Johannessen et al. 2016), such as declines in Arctic sea ice extent, decreases

in snow cover extent and duration, mass loss from glaciers and ice caps, later freeze-up and earlier break-up of rivers and lakes, and warming of permafrost (Lemke et al. 2007; Brown et al. 2010; Romanovsky et al. 2010; Brown and Duguay 2011; Gardner et al. 2011; Tivy et al. 2011; Derksen et al. 2012). Identifying changes in snow and ice conditions utilizing finer-scale satellite observations will contribute to improved snow and ice forecasting, climate monitoring, and understanding climate variability and change in high-latitude regions.

2.6 References

Bintanja, R. and Selten, F.M. 2014. Future increases in Arctic precipitation linked to local evaporation and sea-ice retreat. Nature, 509: 479-482. doi: 10.1038/nature13259.

Bintanja, R. and Andry, O. 2017. Towards a rain-dominated Arctic. Nat. Clim. Change, 7: 263- 267. doi: 10.1038/nclimate3240.

Brochers, H. W. 2019. Package ‘pracma’. Available online: https://cran.r- project.org/web/packages/pracma/pracma.pdf

Box, J.E., Colgan, W.T., Christensen, T.R., Schmidt, N.M., Lund, M., Parmentier, F.J.W., Brown, R., Bhatt, U.S., Euskirchen, E.S., Romanovsky, V.E. and Walsh, J.E. 2019. Key indicators of Arctic climate change: 1971–2017. Environ. Res. Lett. 14: 045010. doi: https://doi.org/10.1088/1748-9326/aafc1b.

Bronaugh, D., and Werner, A. (2019). “zyp”. Available at: https://cran.r- project.org/web/packages/zyp/zyp.pdf

Brown, R., Derksen, C. and Wang, L. 2010. A multi‐data set analysis of variability and change in Arctic spring snow cover extent, 1967–2008. Geophys. Res. 115: D16111. doi: 10.1029/2010JD013975.

Brown, L.C. and Duguay, C.R. 2011. The fate of lake ice in the North American Arctic. Cryosphere, 5: 869-892. doi: 10.5194/tc-5-869-2011.

Callaghan, T.V., Johansson, M., Brown, R.D., Groisman, P.Y., Labba, N., Radionov, V., Barry, R.G., Bulygina, O.N., Essery, R.L., Frolov, D.M. and Golubev, V.N. 2011. The changing face of

Arctic snow cover: a synthesis of observed and projected changes. Ambio, 40: 17-31. doi: 10.1007/s13280-011-0212-y.

Comiso, J. C., and Nishio, F. 2008. Trends in the sea ice cover using enhanced and compatible AMSR‐E, SSM/I, and SMMR data. J. Geophys. Res. 113: C02S07. doi: 10.1029/2007JC004257.

Curry, J. A., Schramm, J. L., and Ebert, E. E. 1995. Sea ice-albedo climate feedback mechanism. J. Clim. 8: 240-247.

Derksen, C. and R. Brown. 2012. Snow. In: Arctic Report Card 2012. Jeffries, M. O., J. A. Richter- Menge and J. E. Overland. (Eds.). Available online at http://www.arctic.noaa.gov/reportcard.

Derksen, C., Smith, S.L., Sharp, M., Brown, L., Howell, S., Copland, L., Mueller, D.R., Gauthier, Y., Fletcher, C.G., Tivy, A. and Bernier, M. 2012. Variability and change in the Canadian cryosphere. Clim. Change, 115: 59-88. doi: 10.1007/s10584-012-0470-0.

Derksen, C., R. Brown, L. Mudryk, and K. Luojus. 2015. Arctic: Terrestrial Snow. In: State of the Climate in 2014. J. Blunden and D. S. Arndt (Eds.). Bull. Amer. Meteor. Soc. 97: 145-147. doi: :10.1175/2016BAMSStateoftheClimate.1.

Derksen, C., Burgess, D., Duguay, C., Howell, S., Mudryk, L., Smith, S., Thackeray, C. and Kirchmeier-Young, M. 2019. Changes in snow, ice, and permafrost across Canada: Chapter 5. In

E. Bush & D.S. Lemmen (Eds.), Canada’s Changing Climate Report (pp. 194 – 260). Ottawa, Ontario: Government of Canada.

Déry, S.J. and Brown, R.D. 2007. Recent Northern Hemisphere snow cover extent trends and implications for the snow‐albedo feedback. Geophys. Res. Lett. 34: L22504. doi:10.1029/2007GL031474.

Dolant, C., Langlois, A., Brucker, L., Royer, A., Roy, A., and Montpetit, B. 2017. Meteorological inventory of rain-on-snow events in the Canadian Arctic Archipelago and satellite detection assessment using passive microwave data. Phys. Geogr. 39: 428-444. doi: 10.1080/02723646.2017.1400339.

European Centre for Medium-range Weather Forecast (ECMWF). 2011. The ERA-Interim reanalysis dataset, Copernicus Climate Change Service (C3S) (accessed May 2020). Available online: https://www.ecmwf.int/en/forecasts/datasets/archive-datasets/reanalysis-datasets/era- interim

Foster, J.L., Sun, C., Walker, J.P., Kelly, R., Chang, A., Dong, J. and Powell, H. 2005. Quantifying the uncertainty in passive microwave snow water equivalent observations. Rem. Sens. Environ. 94:187-203. doi:10.1016/j.rse.2004.09.012.

Gardner, A.S., Moholdt, G., Wouters, B., Wolken, G.J., Burgess, D.O., Sharp, M.J., Cogley, J.G., Braun, C. and Labine, C. 2011. Sharply increased mass loss from glaciers and ice caps in the Canadian Arctic Archipelago. Nature, 473: 357-360. doi: 10.1038/nature10089.

Hannah, C. G., Dupont, F., and Dunphy, M. 2009. Polynyas and tidal currents in the Canadian Arctic Archipelago. Arctic, 62: 83-95.

Hersbach, H., Bell, B., Berrisford, P., Hirahara, S., Horányi, A., Muñoz‐Sabater, J., Nicolas, J., Peubey, C., Radu, R., Schepers, D. and Simmons, A. 2020. The ERA5 global reanalysis. Q. J. R. Meteorol. Soc. doi: https://doi.org/10.1002/qj.3803.

Howell, S.E., Duguay, C.R. and Markus, T., 2009. Sea ice conditions and melt season duration variability within the Canadian Arctic Archipelago: 1979–2008. Geophys. Res. Lett. 36: L10502. doi:10.1029/2009GL037681.

Howell, S. E., Assini, J., Young, K. L., Abnizova, A., and Derksen, C. 2012. Snowmelt variability in polar bear pass, Nunavut, Canada, from QuikSCAT: 2000–2009. Hydrol. Process, 26: 3477- 3488. doi: https://doi.org/10.1002/hyp.8365.

Howell, S.E., Derksen, C., Pizzolato, L. and Brady, M. 2015. Multiyear ice replenishment in the Canadian Arctic Archipelago: 1997–2013. J. Geophys. Res. 120: 1623-1637. doi: 10.1002/2015JC010696.

Howell, S.E. and Brady, M. 2019. The dynamic response of sea ice to warming in the Canadian Arctic Archipelago. Geophys. Res. Lett. 46: 13119-13125. doi: 10.1029/2019GL085116.

Hund, A.J. 2014. Antarctica and the Arctic Circle: A Geographic Encyclopedia of the Earth's Polar Regions. ABC-CLIO, Santa Barbara, California.

Kanamitsu, M., Ebisuzaki, W., Woollen, J., Yang, S.‐K., Hnilo, J. J., Fiorino, M., and Potter, G. L. 2002. NCEP–DOE AMIP‐II reanalysis (R‐2). Bull. Am. Meteorol. Soc. 83: 1631–1643. https://doi.org/10.1175/BAMS‐83‐11‐1631.

Kobayashi, S., Ota, Y., Harada, Y., Ebita, A., Moriya, M., Onoda, H., Onogi, K., Kamahori, H., Kobayashi, C., Endo, H. and Miyaoka, K. 2015. The JRA-55 reanalysis: General specifications and basic characteristics. J. Meteorol. Soc. Jpn. 93: 5–48. https://doi.org/10.2151/jmsj.2015‐001.

Laliberté, F., Howell, S.E.L. and Kushner, P.J. 2016. Regional variability of a projected sea ice‐ free Arctic during the summer months. Geophys. Res. Lett. 43: 256-263. doi: 10.1002/2015GL06855.

Lemke, P., Ren, J., Alley, R.B., Allison, I., Carrasco, J., Flato, G., Fujii, Y., Kaser, G., Mote, P., Thomas, R.H. and Zhang, T. 2007. Observations: changes in snow, ice and frozen ground. In: Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Solomon, S., D. Qin, M. Manning, Z. Chen, M. Marquis, K.B. Averyt, M. Tignor and H.L. Miller (Eds.). Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA.

Markus, T., Stroeve, J. C., and Miller, J. 2009. Recent changes in Arctic sea ice melt onset, freezeup, and melt season length. J. Geophys. Res. 114: C12024. doi:10.1029/2009JC005436.

Marshall, S., Scott, K. A., and Scharien, R. K. 2019. Passive Microwave Melt Onset Retrieval Based on a Variable Threshold: Assessment in the Canadian Arctic Archipelago. Remote Sens. 11: 1304. doi: 10.3390/rs11111304.

Mekis, É., and Vincent, L. A. 2011. An overview of the second generation adjusted daily precipitation dataset for trend analysis in Canada. Atmos.-Ocean, 49:163-177. doi: https://doi.org/10.1080/07055900.2011.583910.

Melling, H. 2002. Sea ice of the northern Canadian Arctic Archipelago. J. Geophys. Res. 107: 3181. doi:10.1029/2001JC001102.

Moore, G. W. K., Schweiger, A., Zhang, J., and Steele, M. 2019. Spatiotemporal variability of sea ice in the Arctic's Last Ice Area. Geophys. Res. Lett. 46, 11237-11243. doi: https://doi.org/10.1029/2019GL083722.

Moran, P. A. 1950. Notes on continuous stochastic phenomena. Biometrika, 37(1/2), 17-23.

Mudryk, L. R., Derksen, C., Kushner, P. J., and Brown, R. 2015. Characterization of Northern Hemisphere snow water equivalent datasets, 1981–2010. J. Clim. 28: 8037-8051. doi: https://doi.org/10.1175/JCLI-D-15-0229.1.

Murfitt, J., and Brown, L. C. 2017. Lake ice and temperature trends for Ontario and Manitoba: 2001 to 2014. Hydrol. Process, 31: 3596-3609. doi: https://doi.org/10.1002/hyp.11295.

Natural Resources Canada. 2002, updated 2019. Digital elevation model of Canada – Canada3D, 2001. Available online: https://open.canada.ca/data/en/dataset/042f4628-94b2-40ac-9bc1- ca3ac2a27d82

Natural Earth. 2020. Coastline vector. Available online: https://www.naturalearthdata.com/downloads/10m-physical-vectors/10m-coastline/

Natural Earth. 2020. Land vector. Available online: https://www.naturalearthdata.com/downloads/10m-physical-vectors/10m-land/

Onarheim, I. H., Eldevik, T., Smedsrud, L. H., and Stroeve, J. C. 2018. Seasonal and regional manifestation of Arctic sea ice loss. J. Clim. 31: 4917-4932. doi: https://doi.org/10.1175/JCLI-D- 17-0427.1.

Pizzolato, L., Howell, S. E., Derksen, C., Dawson, J., and Copland, L. 2014. Changing sea ice conditions and marine transportation activity in Canadian Arctic waters between 1990 and 2012. Clim. Change, 123(2), 161-173.

Pulliainen, J., Luojus, K., Derksen, C., Mudryk, L., Lemmetyinen, J., Salminen, M., Ikonen, J., Takala, M., Cohen, J., Smolander, T. and Norberg, J. 2020. Patterns and trends of Northern Hemisphere snow mass from 1980 to 2018. Nature, 581: 294-298. doi: https://doi.org/10.1038/s41586-020-2258-0.

Ramsay, B. H. 1998. The interactive multisensor snow and ice mapping system. Hydrol. Process, 12: 1537-1546.

Romanovsky, V.E., Smith, S.L. and Christiansen, H.H. 2010. Permafrost thermal state in the polar Northern Hemisphere during the international polar year 2007–2009: a synthesis. Permafrost and Periglac. Process, 21: 106-116. doi: 10.1002/ppp.689.

Screen, J.A. and Simmonds, I. 2012. Declining summer snowfall in the Arctic: Causes, impacts and feedbacks. Clim. Dyn. 38: 2243-2256.

Serreze, M. C., Holland, M. M., and Stroeve, J. 2007. Perspectives on the Arctic's shrinking sea- ice cover. Science, 315: 1533-1536. doi: 10.1126/science.1139426.

Singarayer, J. S., Bamber, J. L., and Valdes, P. J. 2006. Twenty-first-century climate impacts from a declining Arctic sea ice cover. J. Clim. 19: 1109-1125. doi: https://doi.org/10.1175/JCLI3649.1.

Sönmez, I., Tekeli, A. E., and Erdi, E. 2014. Snow cover trend analysis using interactive multisensor snow and ice mapping system data over Turkey. Int. J. Climatol. 34: 349-2361. doi: 10.1002/joc.3843.

Spreen, G., Kaleschke, L., and Heygster, G. 2008. Sea ice remote sensing using AMSR‐E 89‐GHz channels. J. Geophys. Res. 113: C02S03. doi:10.1029/2005JC003384.

Statistics Canada. 2011. Boundary Files, 2011 Census: Statistics Canada Catalogue no. 92-160- G. Available online: https://www12.statcan.gc.ca/census-recensement/2011/geo/bound- limit/bound-limit-2011-eng.cfm

Stroeve, J. C., Markus, T., Boisvert, L., Miller, J., and Barrett, A. 2014. Changes in Arctic melt season and implications for sea ice loss. Geophys. Res. Lett. 41: 1216-1225. doi: 10.1002/2013GL058951.

Tang, C. C., Ross, C. K., Yao, T., Petrie, B., DeTracey, B. M., and Dunlap, E. 2004. The circulation, water masses and sea-ice of Baffin Bay. Prog. Oceanogr. 63: 183-228. doi:10.1016/j.pocean.2004.09.005.

Thackeray, C. W., Fletcher, C. G., Mudryk, L. R., and Derksen, C. 2016. Quantifying the uncertainty in historical and future simulations of Northern Hemisphere spring snow cover. J. Clim. 29: doi: 8647-8663. https://doi.org/10.1175/JCLI-D-16-0341.1.

Thackeray, C. W., Derksen, C., Fletcher, C. G., and Hall, A. 2019. Snow and climate: Feedbacks, drivers, and indices of change. Curr. Clim. Change Rep. 5: 322-333.

Trishchenko, A. P., and Wang, S. 2018. Variations of climate, surface energy budget, and minimum snow/ice extent over Canadian arctic landmass for 2000–16. J. Clim. 31: 1155-1172. doi: 10.1175/JCLI-D-17-0198.1

U.S. National Ice Center. 2008. Updated daily. IMS daily northern hemisphere snow and ice analysis at 1 km, 4 km, and 24 km resolutions, version 1. [v1.1, 1.2, and 1.3]. Boulder, Colorado

USA. NSIDC: National Snow and Ice Data Center. doi: https://doi.org/10.7265/N52R3PMC.

[May 2020].

Uppala, S.M., Kållberg, P.W., Simmons, A.J., Andrae, U., Bechtold, V.D.C., Fiorino, M., Gibson, J.K., Haseler, J., Hernandez, A., Kelly, G.A. and Li, X. 2005. The ERA‐40 re‐analysis. Q. J. R. Meteorol. Soc. 131: 2961–3012. doi: https://doi.org/10.1256/qj.04.176.

Vincent, L. A., X. L. Wang, E. J. Milewska, H. Wan, F. Yang, and V. Swail, 2012. A second generation of homogenized Canadian monthly surface air temperature for climate trend analysis, J. Geophys. Res. 117: D18110, doi:10.1029/2012JD017859.

Vincent, L.A., Zhang, X., Brown, R.D., Feng, Y., Mekis, E., Milewska, E.J., Wan, H. and Wang, X.L. 2015. Observed trends in Canada’s climate and influence of low-frequency variability modes. J. Clim. 28: 4545-4560. doi: https://doi.org/10.1175/JCLI-D-14-00697.1.

Wang, X. L., and Swail, V. R. 2001. Changes of extreme wave heights in Northern Hemisphere oceans and related atmospheric circulation regimes. J. Clim. 14: 2204-2221. doi: https://doi.org/10.1175/1520-0442(2001)014<2204:COEWHI>2.0.CO;2.

Wang, L., Sharp, M., Brown, R., Derksen, C., and Rivard, B. 2005. Evaluation of spring snow covered area depletion in the Canadian Arctic from NOAA snow charts. Remote Sens. Environ. 95: 453-463. doi:10.1016/j.rse.2005.01.006.

Wang, L., Derksen, C., Brown, R., and Markus, T. 2013. Recent changes in pan‐Arctic melt onset from satellite passive microwave measurements. Geophys. Res. Lett. 40: 522-528. doi:10.1002/grl.50098.

WWF. 2018. Last Ice Area. Available online at http://www.wwf.ca/conservation/arctic/lia/

Yackel, J.J., Barber, D.G., Papakyriakou, T.N., Breneman, C. First-Year Sea Ice Spring Melt Transitions in the Canadian Arctic Archipelago from Time-Series Synthetic Aperture Radar Data, 1992–2002. Hydrol. Process. 2007, 21, 253–265.

Young, K.L., Brown, L. and Labine, C., 2018. Snow cover variability at Polar Bear Pass, Nunavut. Arct. Sci. 4: 669-690. doi: dx.doi.org/10.1139/as-2017-0016.

Yu, L., Liu, T., and Zhang, S. 2017. Temporal and Spatial Changes in Snow Cover and the Corresponding Radiative Forcing Analysis in Siberia from the 1970s to the 2010s. Adv. Meteorol. 2017. doi: https://doi.org/10.1155/2017/9517427.

Zhang, X., Vincent, L. A., Hogg, W. D., and Niitsoo, A. 2000. Temperature and precipitation trends in Canada during the 20th century. Atmos.-Ocean, 38: 395-429. doi: https://doi.org/10.1080/07055900.2000.9649654.

Zhang, X., Flato, G., Kirchmeier-Young, M., Vincent, L., Wan, H., Wang, X., Rong, R., Fyfe, J., Li, G., Kharin, V.V. 2019: Changes in Temperature and Precipitation Across Canada; Chapter 4. In Bush, E. and Lemmen, D.S. (Eds.) Canada’s Changing Climate Report (pp. 112 – 193). Government of Canada, Ottawa, Ontario.

Zheng, J., Geldsetzer, T., and Yackel, J. Snow Thickness Estimation on First-Year Sea Ice Using Microwave and Optical Remote Sensing with Melt Modelling. Remote Sens. Environ. 2017, 199, 321–332.

81 82

Chapter 3 Pan-Arctic Sea Ice, Lake Ice, and Snow Phenology from 1997 – 2019

Abstract

Arctic snow and ice cover are useful indicators of climate variability and change. Dramatic reductions in snow and ice cover duration and extent since the start of the satellite era have been linked to amplified warming in the Arctic. To monitor these changes, passive microwave data are most commonly used, however the coarse resolution (25 km) limits the ability of passive microwave sensors to resolve finer-scale changes. Here, we use the National Ice Center Interactive Multisensor Snow and Ice Mapping System (IMS) 24km (1997 – 2019) and 4 km (2004 – 2019) snow and ice products to examine changes in sea ice, snow, and lake ice phenology across the pan- Arctic. Snow-off and ice-off dates are occurring earlier across the pan-Arctic, with the largest trends detected for sea ice first open water dates (-7.72 d decade-1, p < 0.05). Lake ice open water parameters were significantly correlated with snow and sea ice melt parameters from 2004 – 2019, with stronger relationships identified during the melt season compared to the freeze season. Ice- on dates are becoming earlier, with significant positive trends detected for sea ice continuous ice cover (10.60 d decade-1, p < 0.05). Snow onset is also becoming earlier across the pan-Arctic, though no significant trends in snow cover duration were detected as snow-off is also shifting earlier. Regionally, changes in snow-off and lake ice-off dates are larger in North America than Eurasia, which can be linked to strong warming over North America since 2004. The Bering and Chukchi Seas showed the largest trends toward earlier melt and later freeze across the pan-Arctic, consistent with earlier (later) snow and lake ice-off (-on) in west and southwest Alaska. The marked regional variability in snow and ice phenology parameters across the pan-Arctic highlights the complex relationships between snow, ice, and climate, and warrants continuous monitoring to understand how different regions of the Arctic are responding to rapid warming.

3.1 Introduction

The cryosphere is the second largest component of the global climate system after the ocean, exerting significant effects on the Earth’s energy balance, atmospheric circulation, and heat transport (Lemke et al. 2007; Callaghan et al. 2011; Derksen et al. 2012). The relevance for climate

variability and change is based on physical properties, such as high surface reflectivity (albedo) and latent heat associated with phase changes, both of which have a strong impact on the surface energy balance (Lemke et al. 2007).The extent and duration of snow and ice cover have direct feedbacks to the climate system as they strongly influence planetary albedo (Rahmstorf, 2010; Derksen et al. 2012). Seasonal snow and ice cover are also important for Arctic ecosystems as they rely on snow and ice cover for feeding, transportation, and habitat (Dersken et al. 2012). Additionally, the traditional ways of life of many Northern residents depend on snow and ice cover for sources of food, transportation, and economic activities (Derksen et al. 2012). Recent assessments reveal strong linkages between decreasing snow and ice cover and increasing temperatures in the Arctic (e.g. Hernandez-Henriquez et al. 2015; Johannessen et al. 2016; Druckenmiller and Ritcher-Menge, 2020). Reductions in sea ice extent, decreases in snow cover duration, and earlier melt onset in Arctic and sub-Arctic lakes have been reported (Serreze and Stroeve, 2015; Surdu et al. 2016; Mudryk et al. 2018). Arctic surface air temperatures in 2019 were the second highest in the 120-year (1900 – present) observational record (Druckenmiller and Richter-Menge, 2020), and are projected to continue to increase well into the twenty-first century (Overland, 2020). Though the Arctic is undergoing sustained change, observations are often marked by regional differences tied in part to global connections via the atmosphere and ocean (Druckenmiller and Richter-Menge, 2020). For example, in the Canadian Arctic Archipelago (CAA), the effect of warming on sea ice dynamics may be counterintuitive as warming could result in increased ice import from the Arctic Ocean into the CAA (Melling, 2002; Howell and Brady, 2019). Therefore, monitoring Arctic snow and ice cover is critical to improve our understanding of this complex and variable region in the context of climate variability and change.

Monitoring Arctic snow and ice cover largely relies on the use of satellite observations, as ground- based observations are constrained by limited in situ data, large gaps and biases in surface observing networks, and limited geographic coverage (Brown et al. 2010; Brown and Duguay, 2011). Satellite-based microwave data are most commonly used in snow and ice monitoring as they provide information regardless of solar illumination and cloud cover (Brown et al. 2014). Microwave measurements have been used to estimate snow (both on land and on sea ice) and ice melt and freeze onset (e.g. Howell et al. 2006; Yackel et al. 2007; Markus et al. 2009; Wang et al. 2011; Zheng et al. 2017) at various spatial resolutions ranging from 6.25 to 25 km (Brown et al. 2014). The Special Sensor Microwave/Imager (SSM/I) and Scanning Multichannel Microwave

Radiometer (SMMR) passive microwave datasets have been widely used in snow and sea ice mapping (e.g. Cho et al. 2017; Lynch et al. 2017; Crawford et al. 2018). Passive microwave data are well-suited for snow and ice monitoring due to all-weather imaging capabilities and long available records (since the late 1970s), though the coarse resolution (25 km) limits their application and reduces the accuracy of estimates (Derksen et al. 2004; Gao et al. 2010; De Lannoy et al. 2012). There are well-documented uncertainties in using passive microwave measurements to retrieve snow water equivalent and snow cover extent due to differences in snow and surface cover properties (e.g. snow depth, snow grain size, topography, vegetation), which influence microwave emission and backscatter (Brown et al. 2010; Park et al. 2012; Tedesco et al. 2015). The coarse spatial resolution also limits the ability of the sensors to resolve small leads and polynyas and can result in errors near coastal areas due to pixel-based land contamination (Howell et al. 2006; Brown et al. 2014). Johnson and Eicken (2016) note that strong brightness temperature contrasts across pixels can result in falsely high estimates of sea ice concentration, particularly during the summer when there is open water near coastal areas. SMMR and SSM/I are less commonly used in lake ice applications as the spatial resolution limits analyses to large lakes only. Additionally, the 85 GHz channel is susceptible to considerable atmospheric interference, and the 25 km spatial resolution can result in large differences in water/land brightness temperatures (Cavalieri et al. 1999; Howell et al. 2009).

Optical remote sensing data have also been used to monitor Arctic snow and ice cover (e.g. Nitze et al. 2017; Young et al. 2018) as it provides an improved spatial resolution (e.g. 500 m Moderate Resolution Imaging Spectroradiometer Snow Product) compared to passive microwave data. The use of optical imagery is limited to the spring and summer months in high-latitude regions as there is no source of illumination during late fall and winter due to polar darkness. Additionally, the poor temporal resolution of some optical data (e.g. 16 days for Landsat, 8-day MODIS snow product) can introduce uncertainty and inaccuracy into estimates of snow conditions on the Earth’s surface. Active microwave data have been used successfully in snow (e.g. Brown et al. 2007), sea ice (e.g. Mortin et al. 2014), and lake ice (e.g. Howell et al. 2009) applications. Active microwave algorithms using synthetic aperture radar (SAR) provide high resolution (20 to 100 m) retrieval of snow and ice parameters (e.g. Surdu et al. 2016; Zhu et al. 2018; Howell and Brady, 2019). SAR estimates of snow and ice cover provide the highest spatial resolution compared to other products, however the moderate temporal resolution, narrow swath width, and limited image availability

across the Arctic limits the application of SAR to smaller geographic regions (Brown et al. 2014; Howell et al. 2019). Multisensor approaches exploiting advantages of microwave and optical sensors have been used to estimate snow thickness on first year sea ice (e.g. Zheng et al. 2017) and to resolve leads and polynyas at an improved spatial resolution (e.g. Ludwig et al. 2019). The all- weather capabilities of microwave data combined with high temporal resolution of optical imagery can improve estimates of snow and ice parameters in the Arctic.

An alternative approach to snow and ice mapping is the use of the National Ice Center Interactive Multisensor Snow and Ice Mapping System (IMS) product. IMS is created using a variety of multi- sourced datasets (e.g. optical imagery, microwave data, ancillary data) and provides daily maps of snow and ice cover at 24 km, 4 km, and 1 km spatial resolutions (Ramsay 1998; Helfrich et al. 2007). The daily temporal resolution and all-weather monitoring capabilities make IMS suitable in snow cover applications (e.g. Brubaker et al. 2005; Chen et al. 2012; Yu et al. 2017) and lake ice monitoring on large lakes (e.g. Brown and Duguay, 2012; Duguay et al. 2012, 2013, 2014, 2015; Duguay and Brown, 2018). Though not commonly used in sea ice applications, Brown et al. (2014) show that IMS is advantageous over several automated algorithms for monitoring sea ice phenology. IMS is also able to improve sea ice estimates by reducing land contamination and better represent coastal regions compared to passive microwave estimates (Brown et al. 2014), and to resolve finer-scale details between narrow ocean channels (Dauginis and Brown, 2020). This work will expand on the work of Dauginis and Brown (2020) and will examine changes in sea ice, lake ice, and snow phenology from 1997 – 2019 across the pan-Arctic. The objectives of this paper are to 1) assess changes in sea ice, lake ice, and snow phenology from 1997 – 2019 across the pan- Arctic and 2) analyze regional changes in snow and ice phenology during more recent years (2004 – 2019) across the pan-Arctic.

3.2 Data and Methodology 3.2.1 Study Regions

Figure 3.1. Map of the study regions.

In this study, Arctic regions north of 57º were considered when examining pan-Arctic snow and ice phenology in the first section of the results (Figure 3.1). For the second section of the results, a regional approach was taken. For snow and lake ice, phenology parameters were considered on a hemispheric scale (i.e. North America and Eurasia). For sea ice, phenology parameters were examined at three regional scales: Canadian Arctic, Alaska/Russia, and Eurasian Arctic. ‘Canadian Arctic’ includes Baffin Bay, Hudson Bay, and the CAA; ‘Alaska/Russia’ includes the Beaufort,

Chukchi, and Bering seas; ‘Eurasian Arctic’ includes the East Siberian, Laptev, Kara, Barents, and Greenland seas. These regions were grouped based on their similar ice-phenology characteristics, as each region exhibits similar ice on and off dates. The Canadian Arctic has been shown to exhibit earlier freeze trends during recent years (e.g. Dauginis and Brown, 2020), and has shown weaker trends toward earlier melt onset compared to other Arctic regions (e.g. Mahmud et al. 2016; Marshall et al. 2019). Sea ice in the Alaska/Russia region has shown large reductions in sea ice extent over the past decade, which has been linked to strong warming and large sea surface temperature anomalies in this area (Druckenmiller and Richter-Menge, 2020; Perovich et al. 2020).

3.2.2 Data

Snow and ice data were obtained from the National Ice Center Interactive Multisensor Snow and Ice Mapping System (IMS) snow and ice product. IMS is an operational product used to map daily snow and ice cover over the Northern Hemisphere at 1 km (2014 – present), 4 km (2004 – present), and 24 km (1997 – present) spatial resolutions (https://www.natice.noaa.gov/ims/). Analysts use a variety of multi-sourced datasets (for a complete list of data sources, see National Snow and Ice Data Center, https://nsidc.org/data/g02156) to subjectively produce maps with discrete values assigned to land, snow-covered land, water, and ice. Snow mapping primarily relies on visible imagery; however, if visible imagery is unavailable due to cloud occlusion or low solar illumination, microwave data is used instead (Helfrich et al. 2007; Brown et al. 2010). As misidentification errors associated with microwave data can occur, analysts rely more on snow climatology compared to microwave data to estimate high latitude snow cover during winter months (Chang et al. 1996; Foster et al. 2005; Helfrich et al. 2007; Derksen 2008; Brown et al. 2010). Ice cover analysis primarily relies on AVHRR or MODIS observations, however microwave-based retrievals and ice climatology are used when visible imagery is unavailable, with microwave retrievals representing approximately 30-35% of the ice cover input (Helfrich et al. 2007).

Temperature data are from the European Centre for Medium-Range Weather Forecasts (ECMWF) global atmospheric reanalysis (ERA-Interim) and were compared to changes in snow and ice phenology (www.ecmwf.int/en/research/climate-reanalysis/era-interim.). ERA-Interim provides coverage of the entire Arctic at a spatial resolution of approximately 79 km (0.75º) and has been widely used in Arctic studies (e.g. Mortin et al. 2016; Bunzel et al. 2018). Daily 2-m temperature

was used to calculate mean monthly temperature trends from 2004 – 2019. In September 2019, ERA-Interim was officially phased out due to the release of ERA5 (Hersbach et al. 2020), therefore mean monthly temperature trends from September to December reflect temperature trends from 2004 – 2018.

3.2.3 Methodology

Table 3.1. Sea ice, lake ice, and snow phenology parameters and definitions in this study.

Parameter Definition

FOWS First open water (sea ice) The first change from ice to water for a given pixel First open water (lake ice) FOWL

WCIS Water clear of ice (sea ice) The last change from ice to water, signaling ice-free conditions for the remainder of Water clear of ice (lake ice) the season WCIL

FOS Freeze onset (sea ice) The first detection of ice for a given pixel Freeze onset (lake ice) FOL

CICS Continuous ice cover (sea ice) The date of the last change from water to ice Continuous ice cover (lake ice) CICL first_soff First snow-off The first change from snow-covered land to snow-free land for a given pixel final_soff Final snow-off The last change from snow-covered to snow-free land, signalling snow-free conditions for the remainder of the season first_son First snow-on The first change from snow-free land to snow-covered land Final snow-on The last change from snow-free to snow-covered land final_son

The 24 km and 4 km IMS products were used to examine changes in snow and ice phenology dates across the pan-Arctic following the methodology of Brown et al. (2014) and Dauginis and Brown (2020). For each pixel, consecutive days of IMS imagery were compared to determine the first and last changes between snow/ice and land/water to determine the timing of the snow/ice on and off parameters examined. The phenology parameters used in this study and their definitions can be found in Table 3.1. The 24 km IMS product was used to examine trends in mean snow and sea ice phenology dates across the pan-Arctic from 1997 – 2019. For lake ice, only the 4 km IMS product (2004 – 2019) was used since the 24 km product can only detect very large lakes (Figure 3.2). In addition to detecting more lakes, the 4 km IMS product can also provide more detailed information on lake ice phenology within each lake, as shown in Figure 3.2. The non-parametric Sen’s slope was calculated for the phenology time series and reported per decade. Mean 4 km IMS snow and sea ice phenology dates are generally in good agreement with the 24 km IMS product, with the differences mainly attributed to the improved ability of the 4 km product to resolve smaller-scale features and changes in the ice cover extent than the 24 km product can detect (e.g. leads, polynyas, near-shore conditions, and changes at the ice edges) (Brown et al. 2014; Dauginis and Brown,

2020). Interannual and regional variability in snow and ice conditions will inherently affect phenology parameters detected by both IMS products, particularly for sea ice, which may not entirely clear out of some regions in a particular season, leading to no ice -off or -on phenology detected for that year (Dauginis and Brown, 2020).

Figure 3.2. Comparison of 24 km (left) and4 km (right) IMS lake ice first open water in 2017.

To investigate the relationship between phenology parameters and temperature, correlations between variables were examined using Spearman’s rank correlation coefficient (ρ) as this method describes the overall strength of the relationship between two variables and does not require data to follow independent normal distributions (non-parametric) (Hauke and Kossowski, 2011). Datasets were detrended prior to correlation analysis to ensure relationships were not a result of a shared trend, but rather driven by actual relationships between variability in phenology parameters and temperature (Pizzolato et al. 2014). Data were detrended using the “pracma” package in R (https://CRAN.R-project.org/package=pracma) which removes the linear trend from a given dataset by computing the least-squares fit of a straight line to the data and subtracting the resulting function from the data (Borchers, 2019). The detrended data were then used to calculate Spearman correlation coefficients between phenology parameters and temperature.

To evaluate spatial trends in snow and ice phenology and temperature, 4 km IMS phenology dates and ERA-Interim 2-m temperature data were analyzed using the “zhang” method of trend analysis, available in the “zyp” package in R (Bronaugh and Werner 2019). This method of trend analysis was proposed by Zhang et al. (2000) and has been successfully used to represent trends in temperature and precipitation (Zhang et al. 2000) and lake ice phenology (Murfitt and Brown 2017). The “zhang” method is suitable for analyzing spatial trends in this study as it employs non-parametric tests and accounts for autocorrelation. The linear trend is removed from the time series if it is significant and the autocorrelation is computed (Bronaugh and Werner 2019). The autocorrelation computation repeats until the differences in the estimates of the slope and autoregressive model in two consecutive iterations is smaller than 1% (Bronaugh and Werner 2019). The Mann-Kendall test is applied to the resulting time series and the Sen’s slope of the trend is computed (Bronaugh and Werner 2019). The final result is the Sen’s slope (amount of increase or decrease) at each location over the given time period, as well as the significance of each trend (Bronaugh and Werner 2019). Pixels with less than 14 years of phenology data (e.g. regions where ice off only occurs occasionally) are treated as No Data, meaning the spatial extent of the trend examination represents the geographic region where snow/ice off has occurred in at least 14 of the last 16 years.

3.3 Results and Discussion

3.3.1 Trends and Correlations

Mean snow, sea ice, and lake ice phenology dates across the pan-Arctic are shown in Figure 3.3 (4 km IMS, 2004 – 2019). Sea ice open water dates range from mid-March to late September and snow-off dates range from mid-March to late August (Figure 3.3a, b). Lake ice-off ranges from late March to late August (Figure 3.3a, b). Sea ice freeze dates range from mid-September to late April and snow onset dates range from mid-August to late January (Figure 3.3c, d). Lake ice-on dates range from early September to early February (Figure 3.3c, d).

Figure 3.3. Mean 4 km IMS (2004 – 2019) sea ice first open water (FOWS), first snow-off (first_soff), and lake ice first open water (FOWL) (a), sea ice water clear of ice (WCIS), final snow-off (final_soff), and lake ice water clear of ice (WCIL) (b), sea ice freeze onset (FOS), first snow-on (first_son), and lake ice freeze onset (FOL) (c), and sea ice continuous ice cover (CICS), final snow-on (final_son), and lake ice continuous ice cover (CICL) (d).

Figure 3.4. Mean 24 km (1997 – 2019) and 4 km (2004 – 2019) IMS sea ice first open water (FOWS) and water clear of ice (WCIS) (a), first snow-off (first_soff) and final snow-off (final_soff) (b), lake ice first open water (FOWL) and water clear of ice (WCIL) (c), sea ice freeze onset (FOS) and continuous ice cover (CICS) (d), fist snow-on (first_son) and final snow-on (final_son) (e), and lake ice freeze onset

(FOL) and continuous ice cover (CICL) (f). Sen’s slope and significance are indicated for each phenology parameter. Note that for lake ice, only the 4 km IMS product was used in this study.

Mean snow, sea ice, and lake ice phenology dates for the 24 km (1997 – 2019) and 4 km (2004 – 2019) IMS products are shown in Figure 3.4. Overall, the Pan-Arctic shows a shift toward a longer snow and ice-free season (Figure 3.4) from 1997 – 2019, with trends toward earlier snow-off and ice-off and later freeze detected. While the annual variability is similar between the 24 km and 4 km mean phenology dates, a small offset of 3.5 days later for melt and 3.4 days earlier for freeze (average) is evident in the sea ice phenology as a result of the resolution differences mentioned previously, with the timing particularly influenced by the geographic extent of the melt detected by the 4 km product at the northern ice limit. The overall agreement between the products is < 1 day for the snow phenology dates.

rho (ρ) First melt

first_soff FOWL FOWS 0.375 (24 km) 0.618* (4 km) 0.382 (4 km)

FOWL 0.547* (4 km) -

Final melt

final_soff WCIL WCIS 0.462* (24 km) 0.715* (4 km) 0.644* (4 km)

WCIL 0.512* (4 km) - First Freeze

first_son FOL

FOS 0.147 (24 km) 0.232 (4 km) 0.082 (4 km)

FOL -0.268 (4 km) - Final Freeze

final_son CICL

CICS 0.365 (24 km) 0.371 (4 km) 0.241 (4 km)

CICL 0.194 (4 km) -

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

FOWS - - -0.278 -0.506* -0.361 -0.435* -0.586* -0.444* -0.773* - - -

WCIS - - -0.333 -0.529* -0.291 -0.437* -0.603* -0.372 -0.678* - - -

FOS 0.368 0.225 0.170 0.641* - - - - 0.452* 0.515* 0.228 -0.116

CICS 0.333 0.277 -0.039 0.460* - - - - 0.324 0.380 0.146 0.000 first_soff - - -0.210 -0.050 -0.201 -0.390 -0.566* -0.211 - - - - final_soff - - 0.083 -0.051 -0.060 -0.242 -0.487* -0.223 - - - - first_son 0.218 ------0.044 0.411 0.299 0.122 0.175 final_son 0.275 ------0.052 0.359 0.385 -0.081 -0.018 *Trends that are statistically significant (α = 0.05).

Sea ice open water dates both show significant negative (earlier) trends (Figure 3.4a), with a larger negative trend detected for first open water (-7.72 d decade-1) compared to water clear of ice (- 3.31 d decade-1). Sea ice melt parameters were most strongly correlated with temperature in September (ρ = -0.77 and -0.67, p < 0.05; Table 3.3). Snow-off dates show similar trends, with both first snow-off (-4.9 d decade-1, p < 0.05) and final snow-off (-3.21 d decade-1, p > 0.05) becoming earlier (Figure 3.4b), and both snow retreat parameters significantly correlated with 2- m temperatures in July (ρ = -0.56 and -0.48, p < 0.05; Table 3.3). Sea ice first open water and first snow-off dates are weakly correlated when using the 24 km IMS (ρ = 0.37, p > 0.05) and 4 km IMS (ρ = 0.38, p > 0.05) products (Table 3.2). Stronger (statistically significant) correlations were identified for sea ice water clear of ice and final snow-off dates for both IMS products (ρ = 0.46 and 0.64, p < 0.05, for 24 and 4 km IMS products) (Table 3.2).

During the melt season, lake ice first open water and water clear of ice dates were significantly correlated with snow and sea ice melt parameters from 2004 – 2019 (Table 3.2). Stronger relationships were identified between lake ice and sea ice melt season parameters (ρFOW Sea Ice and

Lake Ice = 0.61 and ρWCI Sea Ice and Lake Ice = 0.71, p < 0.05) compared to lake ice and snow (ρfirst snow-off and FOW Lake Ice = 0.54 and ρfinal snow-off and WCI Lake Ice = 0.51, p < 0.05). Trends for lake ice first open water and water clear of ice dates from 2004 – 2019 are negative (-0.76 and -0.02 d decade-1),

though neither were statistically significant. We acknowledge that the 16-year time series (Figure 3.4c) does not provide a comparative time-span to the other trends examined; however, it should be noted that the direction of the trends is negative (earlier) and therefore follows similar patterns observed in snow and sea ice trends during the 1997 – 2019 melt season.

Sea ice freeze onset shows a slightly positive (later) trend (0.36 d decade-1, p > 0.05), while the continuous ice cover trend is much larger and statistically significant (10.6 d decade-1 p < 0.05) (Figure 3.4d). Both first and final snow-on trends are negative, indicating that the pan-Arctic is seeing earlier snow onset over the 1997 – 2019 study period. First snow-on is becoming earlier by 2.79 d decade-1 (p < 0.05) while final snow-on is becoming slightly earlier by 0.64 d decade-1 (p > 0.05) (Figure 3.4e). Snow-on dates show small positive correlations with sea ice freeze parameters, though none are statistically significant (Table 3.2). Lake ice freeze onset and continuous ice cover exhibit trends toward later freeze (4.97 and 4.44 d decade-1, p > 0.05; Figure 3.4f), and although caution should be taken with the short time-span, it should again be noted that lake ice freeze dates show an overall shift toward later freeze. No significant correlations were detected between lake ice/sea ice and lake ice/snow parameters during the freeze season, though similar to the melt season, stronger correlations were detected between lake ice and sea ice freeze compared to lake ice-on and snow-on (Table 3.2).

Overall, snow and ice cover are coming off earlier across the pan-Arctic, while trends during the freeze season vary for sea ice, lake ice, and snow. Earlier sea ice water clear of ice dates (-3.31 d decade-1, p < 0.05) contribute to longer open water duration detected across the pan-Arctic (4.8 d decade-1, p > 0.05; Figure 3.5). Smaller (non-significant) trends were detected in lake ice parameters, with the resulting open water duration in Arctic lakes increasing by 6.9 d decade-1 from 2004 – 2019 (p > 0.05) (Figure 3.5). No trend in snow-free duration was identified (-0.2 d decade-1, p > 0.05; Figure 3.5), despite first snow-off trending significantly earlier (Figure 3.4b).

Examining snow and ice cover at the pan-Arctic scale provides important information on how the cryosphere is responding to climate change as a whole, however the large degree of spatial variability warrants further investigation into snow and ice conditions at regional scales. For example, Dauginis and Brown (2020) demonstrate that the CAA is responding differently to warming compared to other regions of the Arctic; their findings show later summer clearing of ice and earlier freeze and snow onset since 2004, consistent with findings from previous studies that showed no significant trends toward earlier sea ice melt onset dates in the CAA (e.g. Mahmud et al. 2016; Marshall et al. 2019). Other Arctic regions (e.g. Baffin Bay, Eastern Greenland, Barents Sea, Beaufort Sea, and Chukchi Sea) have shown significantly earlier sea ice melt onset by 2.3 to 6.9 d decade-1 (e.g. Stroeve et al. 2014). The response of snow cover to changes in climatic and hydrologic regimes also varies regionally, with northern Canada and eastern Siberia experiencing increased snowfall, while Scandinavia and regions around the Greenland ice sheet are experiencing increasing rainfall (Box et al. 2019). Additionally, ice cover duration in Arctic lakes since 2004 shows interannual and regional variability, with lakes in western Russia showing anomalies ranging from 59 days shorter to 57 days longer, while smaller anomalies were identified in Canadian Lakes (Duguay and Brown, 2018). Therefore, the following section will examine regional variability in sea ice, lake ice, and snow phenology from 2004 – 2019 using the 4 km IMS product as the higher spatial resolution (compared to the 24 km product) allows finer-scale changes in snow and ice cover to be detected.

3.3.2 Regional Variability

3.3.2.1 Melt

Short-term trends in sea ice, snow, and lake ice phenology from 2004 – 2019 are presented in Figure 3.6. Overall, sea ice, snow, and lake ice show tendencies toward earlier melt, with the

exception of 1) Eurasian snow-off parameters, which show little change from 2004 – 2019 compared to other Arctic regions and 2) sea ice first open water in the Canadian Arctic. Sea ice within the Canadian Arctic shows a trend toward later first open water (median = 2 days), though there is considerable spatial variability (Figure 3.6a), while water clear of ice shows a trend towards earlier timing (median = 7 days) (Figure 3.6b). The largest shifts in sea ice open water parameters were detected in the Alaska/Russia region, followed by Eurasia and Canada. Shifts in snow-off and lake ice-off are larger over the North American Arctic (medianFirst snow-off = 8 days earlier, medianFOW Lake Ice = 4 days earlier) compared to the Eurasian Arctic (medianFirst snow-off = 0 days, medianFOW Lake Ice = 1 day earlier).

Sea ice first open water dates in the Canadian Arctic show considerable regional variability, with earlier FOWS in the CAA (median = 7 days), later FOWS in Hudson Bay (median = 6 days), and a median of 0 days in Baffin Bay, due to the northern and southern portions having opposite trend directions (Figure 3.6a). Within Baffin Bay, trends in open water timing are predominantly earlier to the north (between Greenland and northern Baffin Island), while later shifts dominate the southern regions through Davis Strait (Figure 3.6a). Significant negative correlations between sea ice first open water dates and temperature (June through September) were identified in the

Canadian Arctic (ρ = -0.597, -0.547, -0.738, and -0.557; Table 3.4), indicating that earlier sea ice- off dates in the Canadian Arctic are related to air temperature during the melt season. Warming trends were identified over the northern region of Baffin Bay in July and August ranging from 0.01 to 3ºC (Figure 3.7g, h) where shifts toward earlier ice-off were detected (Figure 3.6a, b). Trends towards earlier WCIS were detected throughout the Canadian Arctic (median = 7 days).

Additionally, WCIS dates and 2-m temperature are significantly correlated in July (ρ = -0.55, p < 0.05) and August (ρ = -0.68, p < 0.05) in the Canadian Arctic (Table 3.4). The median temperature increase during July over Hudson Bay is 0.55ºC, with majority of the region showing warming trends (Figure 3.7g). Strong warming was identified in the northern part of Baffin Bay during July and August (Figure 3.7g, h) where earlier WCIS shifts were identified. In the CAA, shifts toward earlier WCIS are mostly confined to the southern channels, where temperature increases are larger in August and September (Figure 3.7h, i).

Figure 3.6. Trends in 4 km IMS (2004 – 2019) sea ice first open water (FOWS), first snow-off

(first_soff), and lake ice first open water (FOWL) (a), sea ice water clear of ice (WCIS), final snow-off

(final_soff), and lake ice water clear of ice (WCIL) (b), sea ice freeze onset (FOS), first snow-on

(first_son), and lake ice freeze onset (FOL) (c), and sea ice continuous ice cover (CICS), final snow-on

(final_son), and lake ice continuous ice cover (CICL) (d).

Snow and Lake Ice Eurasia Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec first_soff - - 0.123 -0.623* -0.905* -0.35 -0.682* 0.241 - - - - final_soff - - 0.144 -0.573* -0.773* -0.244 -0.558* 0.167 - - - - first_son 0.870* ------0.058 0.353 0.771* -0.203 -0.078 final_son 0.838* ------0.35 0.621* 0.835* -0.207 0.032

FOWL - - -0.238 -0.429 -0.191 0.155 -0.044 0.3 - - - -

WCIL - - -0.17 -0.541* -0.417 -0.035 -0.355 -0.147 - - - -

FOL -0.288 0.361 ------0.392 -0.317 0.003 0.235

CICL -0.179 0.417 ------0.325 -0.239 -0.039 0.167 North America Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec first_soff - - -0.488 -0.385 -0.573* -0.241 -0.317 -0.258 - - - - final_soff - - -0.367 -0.205 -0.5 0.044 -0.364 -0.211 - - - - first_son 0.247 ------0.476 0.621* 0.439 -0.160 0.364 final_son 0.158 ------0.461 0.442 0.585* 0.05 0.028

FOWL - - -0.3 -0.455 -0.309 -0.423 -0.297 -0.770* - - - -

WCIL - - -0.252 -0.405 -0.329 -0.452 -0.279 -0.776* - - - -

FOL 0.026 0.017 ------0.082 0.196 0.464 -0.017

CICL 0.126 0.052 ------0.010 0.192 0.453 0.078 Sea Ice Canadian Arctic Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

FOWS - - -0.152 -0.373 -0.461 -0.597* -0.547* -0.738* -0.557* - - -

WCIS - - -0.117 -0.361 -0.632* -0.752* -0.55* -0.638* -0.557* - - -

FOS 0.232 0.664* 0.508* 0.291 - - - - 0.514 0.732* 0.671* 0.367

CICS -0.067 0.564* 0.305 0.367 - - - - 0.6* 0.55* 0.792* 0.789* Alaska/Russia Arctic Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

FOWS - - -0.326 -0.111 -0.15 0.047 -0.326 -0.464 0.314 - - -

WCIS - - -0.358 0.073 0.067 0.317 -0.020 -0.208 0.457 - - -

FOS 0.002 0.405 0.426 -0.061 - - - - 0.021 -0.385 0.225 0.125

CICS 0.379 0.505* 0.629* -0.047 - - - - -0.114 -0.046 0.303 0.089 Eurasian Arctic Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

FOWS - - -0.379 -0.464 -0.111 -0.102 0.111 0.094 -0.125 - - -

WCIS - - -0.232 -0.611* -0.238 -0.511* -0.247 -0.241 -0.096 - - -

FOS 0.220 0.302 0.041 0.25 - - - - 0.532* 0.571* 0.192 0.257

CICS 0.176 0.167 -0.111 0.285 - - - - 0.378 0.564* 0.203 0.239 *Trends that are statistically significant (α = 0.05).

100

Looking at the mainland areas of the Canadian Arctic (The Canadian Territories: Nunavut, Northwest and Yukon), snow-off trends are predominantly earlier over the 2004 – 2019 period, though the trends in first_soff are of larger magnitude than final_soff. Median first snow-off shifts in Nunavut (NU), Northwest Territories (NWT) and Yukon (YT) are 11, 8, and 6 days earlier, respectively. Lake ice first open water is shifting earlier in Nunavut (median = 3 days) and the NWT (median = 6 days), with larger shifts detected in the NWT likely related to strong warming over the region in June (Figure 3.6f). Lake ice-off dates in northern Quebec (bordered by Hudson

Bay and Baffin Bay) are shifting later, with median FOWL shifting 9 days later and WCIL shifting 9 days later. No statistically significant correlations between lake ice-off dates and temperature in June and July were identified over North America (Table 3.4), however Figure 3.7g shows a widespread cooling pattern over northern Quebec in July from 2004 – 2019. Looking at some of the larger lakes, Nettilling Lake and Amadjuak Lake (Baffin Island, NU) show FOWL is shifting 3 days (median values) earlier for both lakes. From 2004 – 2018, these lakes have shown positive ice cover duration anomalies for 7 of the last 14 ice seasons, with most of the longer ice cover duration anomalies observed during the last six seasons (Duguay and Brown, 2018). Great Slave

Lake and Great Bear Lake (NWT) FOWL is shifting 6 and 4 days (median values) earlier, consistent with negative ice cover duration anomalies (shorter ice cover duration) for 9 of the last 14 years identified by Duguay and Brown, 2018.

Sea ice in the Alaska/Russian coastal region also shows large trends toward earlier ice-off, with

FOWS shifting 29 days earlier in the Beaufort Sea, 27 days earlier in Chukchi Sea, and 31 days earlier in the Bering Sea (median values). The Chukchi and Bering Seas have shown larger sea surface temperature warming trends in August compared to the Arctic-wide August mean, and September sea ice extent in the Chukchi Sea was well below the 1981 – 2010 median in 2012, 2018, and 2019 (Druckenmiller and Richter-Menge, 2020; Perovich et al. 2020). No significant correlations between sea ice-off dates and temperature were identified in these regions (Table 3.4), however warming patterns are present over the Bering/Chukchi Seas for almost all months since 2004 (Figure 3.6). First snow-off trends across Alaska (median = 4 days earlier) are smaller compared to Canada, though western Alaska shows strong trends toward earlier snow-off. Strong warming over western Alaska from 2004 – 2019 during April (Figure 3.7d) may contribute to earlier snowmelt in the region. The median lake ice first open water date has shifted 19 days earlier in Alaska, though southwestern Alaska shows some of the largest trends toward earlier FOWL

101

across the pan-Arctic (Figure 3.6a). The median FOWL trend in Lake Clark is 45 days earlier while WCIL is 44 days earlier. On nearby Iliamna Lake, FOWL and WCIL are shifting 29 days and

32 days earlier respectively. Focussing on the more northern regions of Alaska, first_soff trends across North Slope Alaska (NSA) are towards earlier dates (median = 8 days), though final_soff trends are considerably smaller (median = 1 day). Lake ice-off across NSA is shifting later, with both first open water and water clear of ice shifting 3 days (median) later. Arp et al. (2013) found that the Arctic Coastal Plain (northern Alaska) and Beringia (western Alaska) areas experienced the latest ice-out timing from 2007 – 2012 compared to other lakes across Alaska, as climatology in these regions is influenced by sea-ice conditions along the Arctic Ocean coast. Though long- term trends (1950 – 2011) indicate earlier ice break-up and shorter ice seasons in NSA (Surdu et al. 2014), the trends toward later ice-off in northern Alaska identified in this study from 2004 – 2019 (Figure 3.6a) may be reflecting interannual variability and the complex responses of lake ice to changes in temperature, sea ice, and snow cover conditions. The far-eastern regions of Russia show opposite responses central to Alaska, as first_soff trends are earlier (median = 1 day) and final_soff trends show no change (median = 0 days) on the Chukchi Peninsula. No lakes were detected by the 4 km IMS product in this region.

Trends in sea ice-off parameters were much larger compared to snow and lake ice-off in northeastern Eurasia, with FOWS in the East Siberian and Laptev Seas shifting 12 and 11 days (median values) earlier (Figure 3.6a). Warming patterns over the East Siberian and Laptev Seas were detected in July (medianEast Siberian = 0.01ºC and medianLaptev = 0.02ºC) and August (medianEast

Siberian = 0.29ºC and medianLaptev = 2.25ºC), though no significant correlations between sea ice first open water dates and temperature were detected in Eurasia (Table 3.4). Earlier ice-off dates in these regions are consistent with large reductions in September sea ice extent in the East Siberian and Laptev Seas from 1979 – 2016 (Onarheim et al. 2018). Larger sea ice-off trends were detected in west Eurasia, with earlier FOWS detected in the Kara (median = 14 days), Barents (median = 18 days), and Greenland (median = 13 days) Seas. The earlier ice-off shifts detected in this study across the Eurasian Arctic are consistent with Bliss and Anderson (2018), who report negative (earlier) trends in sea ice melt onset across Eurasia from 1979 – 2017 of -9.45 d decade-1 (East Siberian), -7.3 d decade-1 (Laptev), -8.19 d decade-1 (Kara), -8.47 d decade-1 (Barents), and -2.37 d decade-1 (Greenland). Small/no shifts in snow-off and lake ice-off timing were detected in northeastern Eurasia. Median shifts in snow-off were 1 day earlier (first snow-off) and 0 days for

102

final snow-off (Figure 3.6a, b). Median lake ice-off shifts were 0 days (FOWL) and 1 day later

(WCIL). In northwestern Eurasia, later snow-off trends were detected in northwest Russia

(medianfirst snow-off = 7 days and medianfinal snow-off = 3 days) and in Scandinavia near the Gulf of

Bothnia (Figure 3.6a, b). Lake ice in Lake Ladoga (northwest Russia) shows earlier FOWL (median

= 29 days) and WCIL (median = 9 days) shifts. Lake Onega (northwest Russia) also shows earlier

FOWL (median = 5 days) and WCIL (median = 6 days) shifts, though not as large as Lake Ladoga. From 1955 – 2015, total ice cover duration in Lake Onega decreased by 50 days, though decreases were mostly attributed to delayed freeze (Filatov et al. 2019). Earlier break-up dates have been detected in 40 lakes across Finland from 1963 – 2014 (Kuusisto, 2015), however our recent short- term trends show that lake ice-off is becoming later (median = 2 days) in Finnish lakes nearby to Lake Ladoga and Onega. Mean 4 km IMS imagery shows that the average break-up dates range from mid-April to mid-May in this region, though temperature shifts are only negative (cooler) in southwestern Finland during April and positive (warmer) over all of Finland during May (Figure 3.7d, e).

3.3.2.2 Freeze

During the freeze season, sea ice and lake ice show trends toward later freeze (Figure 3.6c, d). Sea ice freeze onset in the Canadian Arctic is shifting earlier (median = 11 days), while sea ice within the Alaska/Russia region shows the largest delays in freeze onset (median = 8 days) across the pan-Arctic (Figure 3.6c). Unlike the melt season (where lake ice-off trends were larger over North America compared to Eurasia), overall shifts toward later lake ice freeze onset are larger across Eurasia (median = 8 days) than North America (median = 2 days) (Figure 3.6c). Over both regions, earlier first_son (medianNorth America = 8 days and medianEurasia = 9 days) and final_son (medianNorth

America = 3 days and medianEurasia = 7 days) trends were detected, with the exception of southwestern Alaska which showed later final snow-on trends (Figure 3.6c, d).

Earlier sea ice freeze and snow-on trends were detected across the Canadian Arctic. Sea ice shows freeze onset shifting earlier by 6 days in the CAA, and11 days in Hudson Bay and Baffin Bay

(median values) (Figure 3.6c). Snow-on is shifting earlier in northern Quebec (medianfirst snow-on =

8 days, medianfinal snow-on = 16 days) and corresponds to earlier FOL (median = 4 days) and FOS identified in this region. FOS, CICS, and final_son dates are each significantly correlated with

October 2-m temperatures in North America/Canada, with the strongest correlation between FOS

103

(ρ = 0.732, p < 0.05), followed by final_son (ρ = 0.585, p < 0.05) and CICs (ρ = 0.55, p < 0.05) (Table 3.4). A large cooling pattern can be seen over eastern Canada in October which may contribute to delayed snow and ice-on dates in this region (Figure 3.7j). Earlier snow-on dates in the Canadian Arctic are consistent with observed increases in precipitation across all seasons in Canada from 1948 – 2012 (Vincent et al. 2015). Global climate models project increases in Arctic precipitation over the twenty-first century due to enhanced local surface evaporation resulting from sea ice loss; however, recent projections show a shift toward a rain-dominated Arctic, particularly during summer months (Bintanja and Selten, 2014; Bintanja and Andry, 2017). Nettilling Lake shows earlier freeze onset (median = 3 days) and continuous ice cover (median = 1 day), though there is considerable variability in freeze-up as the east shows trends toward earlier freeze and west shows trends toward later freeze. Amadjuak Lake and Lake Hazen show trends toward later freeze onset (median = 3 days for both lakes), consistent with later ice formation across Arctic lakes from 2002 – 2015 (Du et al. 2017; Derksen et al. 2019). Increases in mean monthly lake surface temperatures in August have been reported to delay freeze-up by 0.3 d decade-1 in Lake Hazen from 2000 – 2012 (Lehnherr et al. 2018). Sea ice in northwest Canada shows similar trends, with FOS and CICS shifting 4 and 5 days (median values) later in the Western Arctic Waterway. Total summer (June to October) ice in this region has decreased by 6 – 15% from 1968 – 2016 (Derksen et al. 2019). First snow-on across the Canadian Territories is shifting earlier

(medianNunavut = 6 days, medianNWT = 14 days, medianYukon = 6 days), though delayed snow onset can be identified along north and northwest regions of Canada (south of the Western Arctic Waterway near the Nunavut – NWT border) (Figure 3.6c, d). Ice-on dates are shifting later in Great

Slave Lake (medianLake Ice FO = 3 days and medianLake Ice CIC = 1 day) and in Great Bear Lake

(medianLake Ice FO = 8 days and medianLake Ice CIC = 6 days).

Delayed sea ice freeze was identified throughout the Alaska/Russia region, though there is considerable variability in snow onset and lake ice freeze (Figure 3.6c, d). Median shifts toward

FOS (5 days) and CICS (8 days) in the Beaufort Sea are consistent with multi-year ice losses and lengthening of the open water season in this region (Galley et al. 2016). The Chukchi and Bering

Seas both show trends toward later freeze onset (medianChukchi Sea = 10 days and medianBering Sea =

23 days) and continuous ice cover (medianChukchi Sea = 20 days and medianBering Sea = 43 days), with the Bering Sea representing the region with the largest delay in sea ice freeze across the pan-Arctic. During the ice cover season in 2017/2018 the Bering Sea ice extent was lower than any previous

104

winter in the reconstructed or observed record, attributed to warmer sea surface temperatures, delayed freeze, and frequent storms (Thoman et al. 2020). In 2019 the Bering Sea also had extremely low ice cover during the winter and may have acted as a precursor to low summer ice conditions in the Chukchi Sea (Perovich et al. 2020). Sea ice did not completely freeze over in the Chukchi Sea until December 24 in 2019 (approximately a month later than average), with only 2007 and 2016 showing similar freeze patterns since satellite observations began in 1979 (Perovich et al. 2020). Strong warming trends observed over the Bering and Chukchi Seas from October through January (1 – 6 ºC shift) likely contribute to the delayed freeze detected (Figure 3.7j, k), though no significant correlations between CICS and 2-m temperatures from October to January were identified in this region. First_son trends over Alaska are becoming earlier (median = 2 days), while final_son is shifting later (median = 5 days). The largest trends toward later snow-on were detected in west and southwest Alaska, with first_son shifting 2 days later and final_son shifting 19 days later (median values). In southwest Alaska, Lake Clark and Iliamna Lake show large shifts toward later freeze onset, at 34 days and 44 days later respectively, consistent with Duguay and Brown (2018) who report that 2017/2018 freeze-up in west Alaska was approximately two weeks later than the 2004 – 2017 mean. Wendler et al. (2017) report a 17% increase in mean snowfall across Alaska from 1946 – 2014, with the largest increases occurring in west and southwest Alaska. More snowfall here may be tied to warming in this region identified during almost all months from 2004 – 2019 (Figure 3.7), as warmer air is able to hold more moisture which can thus facilitate increases in precipitation (Thackeray et al. 2019). Increases in snowfall can reduce lake ice thickness by insulating the lake surface and inhibiting deep ice formation (Brock, 2016), which may contribute to earlier ice-off and later ice-on dates in west/southwest Alaska. First snow-on in the NSA region is shifting earlier (median = 2 days), whereas final snow-on is shifting later

(median = 5 days). Lake ice within the NSA region shows trends toward earlier FOL and CICL (median = 9 days for both).

In Eurasia, sea ice freeze onset is shifting later, though the trends in freeze (Figure 3.6c, d) are smaller in magnitude than the trends for melt (Figure 3.6a, b). FOS in the East Siberian and Laptev seas is shifting later (medianEast Siberian Sea = 4 days and medianLaptev Sea = 7 days), where strong warming shifts are present (over the East Siberian and Laptev Seas) in September and October

(Figure 3.7i, j). Snow onset over eastern Russia is shifting earlier (medianfirst snow-on = 10 days and medianfinal snow-on = 7 days), where increasing snowfall has been previously documented

105

(Kononova, 2012). Reanalysis datasets show the strongest increases in annual total precipitation during the cold season (October to May) in regions north of 50 degrees (Box et al. 2019). Lake ice shows similar patterns to snow onset, with earlier FOL and CICL detected over northeast Russia.

In northwest Russia and Scandinavia, first snow-on (medianNorthwest Russia = 8 days and medianScandinavia = 15 days) and final snow-on (medianNorthwest Russia = 12 days and medianScandinavia

= 12 days) are shifting earlier. Lakes within this region show opposite patterns, with later FOL and

CICL identified in Lake Ladoga, Lake Onega, and in Scandinavian lakes. Using data from 1890 – 2015, Karetnikov et al. (2017) show that the number of winters with complete freeze over of Lake Ladoga decreased after 1950 and that the ice season has become shorter. Lake Onega shows later freeze onset (median = 28 days) and continuous ice cover (median = 15 days) shifts. Across Scandinavia, freeze onset is shifting 34 days (median value) later and continuous ice cover shifting 20 days (median value) later. Large freeze-up anomalies in this region were also identified through previous lake ice research (e.g. Duguay and Brown, 2018), for example the 2017/2018 freeze season showed delayed freeze up by approximately 2 – 5 weeks compared to the 2004 – 2018 mean. Sea ice in the western Eurasian Arctic also shows delayed freeze onset, with median shifts of 6 and 10 days in the Kara and Barents Seas detected. FOS and CICS are significantly correlated with October 2-m temperature (Table 3.4), and warming patterns were detected over northern Eurasia in October (Figure 3.7k).

3.4 Conclusions

This paper examined sea ice, snow, and lake ice phenology across the pan-Arctic using the Interactive Multisensor Snow and Ice Mapping System (IMS) snow and ice products. Using IMS, we were able to examine both long-term snow and ice on/off trends (1997 – 2019) at a 24 km spatial resolution, as well as more recent shifts in snow and ice phenology (2004 – 2019) at an improved resolution of 4 km. Our results show that the Arctic is shifting toward a longer snow and ice-free season, with trends toward earlier snow/ice-off and later freeze detected. Sea ice showed the largest trends toward later melt and earlier freeze, with FOWS timing becoming earlier by 7.72 -1 -1 d decade and CICS becoming later by 10.60 d decade . Lake ice and snow-off parameters are also shifting earlier, though trends were not as large as those detected for sea ice. Lake ice-off showed significant correlations with snow-off and sea ice-off during the melt season, while no significant correlations were found between any snow/lake ice/sea ice parameters during the freeze season. This likely reflects the strong influence of surface air temperature on snow and ice melt,

106

whereas during the freeze season, precipitation plays an important role in determining the timing of snow onset and lake size/volume is an important determinant for freeze timing.

Sea ice in the Canadian Arctic is disappearing earlier, though during the freeze season ice-on timing is shifting earlier (9 to 11 days), showing opposite trends compared to other regions across the pan-Arctic. Snow onset also shows similar shifts to sea ice over Canada, with the timing of snow-on moving earlier by 6 to 14 days across the Canadian Territories. Unlike sea ice and snow, lake ice-on dates are shifting later across Canada (up to 3 days), though shifts are smaller during the freeze season compared to the melt season. The largest trends toward earlier melt were detected in the Alaska/Russia region, with FOWS becoming 29 days earlier in the Beaufort Sea, 27 days earlier in the Chukchi Sea, and 31 days earlier in the Bering Sea. Earlier lake ice-off dates were also detected near the Bering Sea, with medians of 45 (FOWL) and 44 (WCIL) days earlier in Lake

Clark and 29 (FOWL) and 32 (WCIL) days for Iliamna Lake. Delays in freeze were also observed in this region, with median CICS in the Chukchi and Bering Seas shifting by 20 and 43 days later. This is consistent with delayed snow onset over land and delayed freeze onset in lakes across southwestern Alaska. These phenology trends are likely related to strong temperature increases, as trends in 2-m temperature are positive during most months from 2004 – 2019 in this region No shift in first snow-off dates were detected in Eurasia (median = 0 days) and only a small shift toward earlier final snow-off (median = 1 day) was identified though larger trends toward earlier snow-off were detected in northwest Eurasia compared to the east. Lake ice shows a similar east- west pattern, with Lake Ladoga and Lake Onega showing earlier FOWL (medians = 29 and 5 days) compared to FOWL in northeast Eurasia (median = 0). Snow onset is also getting earlier over Eurasia, consistent with previous studies that have documented increases in total annual precipitation during cold seasons over the Arctic (Box et al. 2019).

By examining multiple components of the cryosphere together, we can better understand how warming affects snow and ice cover and how these components are related. As the Arctic continues to experience unprecedented change as a response to increasing temperatures, continuous monitoring of changes in snow and ice cover is essential to improve our understanding of climate variability and change.

107

3.5 References

Alexeev, V. A., Arp, C. D., Jones, B. M., and Cai, L. 2016. Arctic sea ice decline contributes to thinning lake ice trend in northern Alaska. Environ. Res. Lett. 11(7), 074022.

Arp, C. D., Jones, B. M., and Grosse, G. 2013. Recent lake ice‐out phenology within and among lake districts of Alaska, USA. Limnol. Oceanogr. 58(6), 2013-2028.

Bintanja, R. and Selten, F.M. 2014. Future increases in Arctic precipitation linked to local evaporation and sea-ice retreat. Nature, 509: 479-482. doi: 10.1038/nature13259.

Bintanja, R. and Andry, O. 2017. Towards a rain-dominated Arctic. Nat. Clim. Change, 7: 263- 267. doi: 10.1038/nclimate3240.

Bliss, A. C., and Anderson, M. R. 2018. Arctic sea ice melt onset timing from passive microwave‐ based and surface air temperature‐based methods. J. Geophys. Res. 123(17), 9063-9080.

Box, J. E., Colgan, W. T., Christensen, T. R., Schmidt, N. M., Lund, M., Parmentier, F. J. W., ... and Walsh, J. E. 2019. Key indicators of Arctic climate change: 1971–2017. Environ. Res. Lett. 14(4), 045010.

Brock, B. W. 2016. Shrinking sea ice, increasing snowfall and thinning lake ice: A complex Arctic linkage explained. Environ. Res. Lett. 11(9), 091004.

Borchers, H. W. 2019. Package ‘pracma’. Available online: https://cran.r- project.org/web/packages/pracma/pracma.pdf

Bronaugh, D., and Werner, A. 2019. “zyp”. Available at: https://cran.r- project.org/web/packages/zyp/zyp.pdf

Brown, R., Derksen, C., and Wang, L. 2007. Assessment of spring snow cover duration variability over northern Canada from satellite datasets. Remote Sens. Environ. 111(2-3), 367-381.

Brown, R., Derksen, C. and Wang, L. 2010. A multi‐data set analysis of variability and change in Arctic spring snow cover extent, 1967–2008. Geophys. Res. 115: D16111. doi: 10.1029/2010JD013975.

108

Brown, L.C. and Duguay, C.R. 2011. The fate of lake ice in the North American Arctic. Cryosphere, 5: 869-892. doi: 10.5194/tc-5-869-2011.

Bunzel, F., Notz, D., and Pedersen, L. T. 2018. Retrievals of Arctic sea‐ice volume and its trend significantly affected by interannual snow variability. Geophys. Res. Lett. 45(21), 11-751.

Chang, A.T.C., Foster, J.L. and Hall, D.K. 1996. Effects of forest on the snow parameters derived from microwave measurements during the BOREAS winter field campaign. Hydrol. Process, 10: 1565-1574.

Dauginis, A. and Brown, L. C. 2020. Sea ice and snow phenology in the Canadian Arctic Archipelago from 1997 – 2018. Arct. Sci. Manuscript submitted on July 7th 2020, ID: AS-2020- 0024.

De Lannoy, G. J., Reichle, R. H., Arsenault, K. R., Houser, P. R., Kumar, S., Verhoest, N. E., and Pauwels, V. R. 2012. Multiscale assimilation of Advanced Microwave Scanning Radiometer–EOS

109

snow water equivalent and Moderate Resolution Imaging Spectroradiometer snow cover fraction observations in northern Colorado. Water Resour. Res. 48(1).

Derksen, C. and R. Brown. 2012. Snow [in “Arctic Report Card 2012”]. Jeffries, M. O., J. A. Richter-Menge and J. E. Overland. (Eds.). Available online at http://www.arctic.noaa.gov/reportcard.

Derksen, C., Burgess, D., Duguay, C., Howell, S., Mudryk, L., Smith, S., Thackeray, C. and Kirchmeier-Young, M. 2019. Changes in snow, ice, and permafrost across Canada; Chapter 5 in Canada’s Changing Climate Report, (ed.) E. Bush and D.S. Lemmen; Government of Canada, Ottawa, Ontario, p.194–260.

Druckenmiller, M. L. and J. Richter-Menge. 2020. Overview [in “State of the Climate in 2019”]. Bull. Amer. Meteor. Soc. 101(8), S245 – S246.

Du, J., Kimball, J. S., Duguay, C., Kim, Y., and Watts, J. D. 2017. Satellite microwave assessment of Northern Hemisphere lake ice phenology from 2002 to 2015. Cryosphere, 11, 47 – 63.

Duguay, C., Brown, L., Kang, K., and Kheyrollah Pour, H. 2012. [The Arctic] Lake Ice [in “State of the Climate 2011]. Bull. Amer. Meteor. Soc. 93(7), S138-S140.

Duguay, C., L. Brown, Kang K-K., and Kheyrollah Pour, H. 2013. [The Arctic] Lake ice [In “State of the Climate in 2012”]. Bull. Amer. Meteor. Soc. 94(8): S124-S126.

Duguay, C., L. Brown, Kang, K.-K., and Kheyrollah Pour, H. 2014. [The Arctic] Lake ice [In “State of the Climate in 2013”]. Bull. Amer. Meteor. Soc. 95 (7).

110

Duguay, C. R., Bernier, M., Gauthier, Y., and Kouraev, A. 2015a. Remote sensing of the cryosphere: Remote sensing of lake and river ice: 1st edn. John Wiley & Sons, Ltd, UK.

Duguay, C., L. Brown, Kang, K.-K., and Kheyrollah Pour, H. 2015b. [The Arctic] Lake ice [In “State of the Climate in 2014”]. Bull. Amer. Meteor. Soc. 96 (7), S144-S145.

Duguay C and Brown L. 2018. Lake Ice [in Arctic Report Card 2018], https://arctic.noaa.gov/Report-Card/Report-Card-2018/ArtMID/7878/ArticleID/785/Lake-Ice

Filatov, N., Baklagin, V., Efremova, T., Nazarova, L., and Palshin, N. 2019. Climate change impacts on the watersheds of Lakes Onego and Ladoga from remote sensing and in situ data. Inland Waters, 9(2), 130-141.

Galley, R. J., Babb, D., Ogi, M., Else, B. G. T., Geilfus, N. X., Crabeck, O., ... and Rysgaard, S. 2016. Replacement of multiyear sea ice and changes in the open water season duration in the Beaufort Sea since 2004. J. Geophys. Res. Lett. 121(3), 1806-1823.

Hauke, J., and Kossowski, T. 2011. Comparison of values of Pearson's and Spearman's correlation coefficients on the same sets of data. Quaestiones Geographicae, 30(2), 87-93.

Hersbach, H., Bell, B., Berrisford, P., Hirahara, S., Horányi, A., Muñoz‐Sabater, J., ... and Simmons, A. 2020. The ERA5 global reanalysis. Q J R Meteorol Soc. 146(730), 1999-2049.

111

Howell, S.E. and Brady, M. 2019. The dynamic response of sea ice to warming in the Canadian Arctic Archipelago. Geophys. Res. Lett. 46: 13119-13125. doi: 10.1029/2019GL085116.

Johnson, M., and Eicken, H. 2016. Estimating Arctic sea-ice freeze-up and break-up from the satellite record: A comparison of different approaches in the Chukchi and Beaufort Seas. Elem Sci Anth. 4.

Karetnikov, S., Leppäranta, M., and Montonen, A. 2017. A time series of over 100 years of ice seasons on Lake Ladoga. J. Great Lakes Res. 43(6), 979-988.

Kononova, N. K. 2012 The influence of atmospheric circulation on the formation of snow cover on the north eastern Siberia. Ice and Snow, 1, 38–53 (In Russian, English summary)

Kuusisto, E. 2015. Trends of breakup dates in Finnish lakes in 1963-2014. In 20th International Northern Research Basins Symposium and Workshop Kuusamo, Finland–August 16–21, 2015 (p. 35).

Lehnherr, I., Louis, V.L.S., Sharp, M., Gardner, A.S., Smol, J.P., Schiff, S.L., Muir, D.C., Mortimer, C.A., Michelutti, N., Tarnocai, C. and Pierre, K.A.S. 2018. The world’s largest High

112

Arctic lake responds rapidly to climate warming. Nat. Commun. 9: 1290. doi: 10.1038/s41467- 018-03685-z

Lemke, P., Ren, J., Alley, R.B., Allison, I., Carrasco, J., Flato, G., … and Zhang, T. 2007. Observations: changes in snow, ice and frozen ground. In: Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Solomon, S., D. Qin, M. Manning, Z. Chen, M. Marquis, K.B. Averyt, M. Tignor and H.L. Miller (Eds.). Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA.

Ludwig, V., Spreen, G., Haas, C., Istomina, L., Kauker, F., and Murashkin, D. 2019. The 2018 North Greenland polynya observed by a newly introduced merged optical and passive microwave sea-ice concentration dataset. Cryosphere, 13(7).

Markus, T., Stroeve, J. C., and Miller, J. 2009. Recent changes in Arctic sea ice melt onset, freezeup, and melt season length. J. Geophys. Res. 114: C12024. doi:10.1029/2009JC005436.

Melling, H. 2002. Sea ice of the northern Canadian Arctic Archipelago. J. Geophys. Res. 107(C11), 2-1.

113

Mortin, J., Svensson, G., Graversen, R. G., Kapsch, M. L., Stroeve, J. C., and Boisvert, L. N. 2016. Melt onset over Arctic sea ice controlled by atmospheric moisture transport. Geophys. Res. Lett. 43(12), 6636-6642.

Onarheim, I. H., Eldevik, T., Smedsrud, L. H., and Stroeve, J. C. 2018. Seasonal and regional manifestation of Arctic sea ice loss. J. Clim. 31(12), 4917-4932.

Overland, J. E. 2020. Less climatic resilience in the arctic. Weather. Clim. Extremes, 100275.

Park, H., Yabuki, H., and Ohata, T. 2012. Analysis of satellite and model datasets for variability and trends in Arctic snow extent and depth, 1948–2006. Polar Sci. 6(1), 23-37.

Perovich, D., Meier, W., Tschudi, M., Wood, K., Farrell, S., Hendricks, S., … and Webster, M. 2020. Sea ice [in “State of the Climate in 2019”]. Bull. Amer. Meteor. Soc. 101(8), S251 – S256.

Przybylak, R., and Wyszyński, P. 2020. Air temperature changes in the Arctic in the period 1951– 2015 in the light of observational and reanalysis data. Theor. Appl. Climatol. 139(1-2), 75-94.

Rahmstorf, S. 2010. A new view on sea level rise. Nat. Clim. Change. 4(4), 44-45.

Ramsay, B. H. 1998. The interactive multisensor snow and ice mapping system. Hydrol. Process, 12: 1537-1546.

Serreze, M. C., and Stroeve, J. 2015. Arctic sea ice trends, variability and implications for seasonal ice forecasting. Philos. Trans. R. Soc. A. 373(2045), 20140159.

114

Surdu, C. M., Duguay, C. R., and Fernández Prieto, D. 2016. Evidence of recent changes in the ice regime of lakes in the Canadian High Arctic from spaceborne satellite observations. Cryosphere. 10(3), 941-960.

Tedesco, M., Derksen, C., Deems, J.S., and Foster, J. L. 2015. Remote sensing of the cryosphere: Remote sensing of snow depth and snow water equivalent: 1st edn. John Wiley & Sons, Ltd, UK

Thoman, R., Bhatt, U., Bieniek, P., Brettschneider, B., Brubaker, M., Danielson, S., … and Walsh, J. (2020). The record low Bering Sea ice extent in 2018 [in “Context, impacts, and an assessment of the role of anthropogenic climate change in Explaining extreme events of 2018 from a climate perspective”]. Bull. Amer. Meteor. Soc. 101(1), 53 – 59.

Thackeray, C. W., Derksen, C., Fletcher, C. G., and Hall, A. 2019. Snow and climate: Feedbacks, drivers, and indices of change. Curr. Clim. Change Rep. 5(4), 322-333.

Wendler, G., Gordon, T., and Stuefer, M. 2017. On the precipitation and precipitation change in Alaska. Atmosphere, 8(12), 253.

115

Yackel, J.J., Barber, D.G., Papakyriakou, T.N., and Breneman, C. 2007. First-year sea ice spring melt transitions in the Canadian Arctic Archipelago from time-series synthetic aperture radar data, 1992–2002. Hydrol. Process. 21, 253–265.

Young, K.L., Brown, L. and Labine, C., 2018. Snow cover variability at Polar Bear Pass, Nunavut. Arct. Sci. 4: 669-690. doi: dx.doi.org/10.1139/as-2017-0016.

Zheng, J., Geldsetzer, T., and Yackel, J. 2017. Snow thickness estimation on first-year sea ice using microwave and optical remote sensing with melt modelling. Remote Sens. Environ. 199, 321–332.

Zhu, J., Tan, S., King, J., Derksen, C., Lemmetyinen, J., and Tsang, L. 2018. Forward and inverse radar modeling of terrestrial snow using SnowSAR data. IEEE Geosci. Remote Sens. 56(12), 7122-7132.

116

Chapter 4 Conclusions

This study has explored the use of the Interactive Multisensor Snow and Ice Mapping System (IMS) 24 km and 4 km snow and ice products for detecting sea ice, lake ice, and snow phenology at the regional and pan-Arctic scales. This work will be facilitated in the future by the more recent release of the 1 km IMS product (released in 2014), as more data becomes available for longer- term analyses, and as new data become incorporated into the IMS product to provide improved estimates of snow and ice cover.

Firstly, this work examined sea ice and snow phenology in the Canadian Arctic Archipelago (CAA) from 1997 – 2018 using the 24 and 4 km IMS products. Methods that have been traditionally used to monitor snow and ice cover in the Arctic are typically limited by the coarse resolution of satellite microwave data (e.g. e.g. Howell et al. 2006; Markus et al. 2009; Wang et al. 2011). As a result, pan-Arctic studies do not always include the CAA in their analyses as these microwave data have difficulty resolving sea ice features within the narrow channels of the CAA. This study demonstrated that the IMS products can be used for both longer-term (1997 onwards) and recent (2004 onwards) analyses of snow and ice cover in the CAA, particularly the 4 km product as it is able to resolve finer scale changes in sea ice and snow cover which are not always visible in coarser-resolution satellite data.

Examining changes in sea ice and snow phenology in the CAA is important as this region has responded differently to climate warming over recent years compared to other Arctic regions (e.g. Dauginis and Brown, 2020), and is now formally recognized as the “Last Ice Area” (WWF, 2018; Moore et al. 2019). Results from Chapter 2 showed that the final clearing of ice is becoming later at 3.5 d decade-1 and freeze onset is becoming earlier at5.4 d decade-1, suggesting that ice cover duration in the CAA is increasing. Similarly, earlier first snow-on trends (5.4 d decade-1earlier) indicate that snow cover duration is getting longer. The strongest relationships were identified between continuous open water and final snow-off (ρ = 0.63, p < 0.05) and continuous ice cover and final snow-on (ρ = 0.56, p < 0.05) dates. This suggests that the ‘last occurrence’ (i.e. last change from ice (snow) to water (snow-free land) and vice versa) parameters may better capture the cumulative climatic effects driving both snow/ice phenology parameters compared to ‘first

117

occurrence’ parameters. Regional analyses using the 4 km IMS product showed significant clustering toward earlier snow and ice-off dates in the western CAA in the Western Arctic Waterway and on Banks and Victoria Islands. This is consistent with stronger warming trends identified in the western CAA from 2004 – 2018. In the Queen Elizabeth Islands (QEI), earlier sea ice freeze and snow-on trends were detected, though snow onset dates showed considerable variability over the eastern QEI regions. Overall, there was less interannual variability observed during the freeze season compared to the melt season, as freeze is largely governed by cooling temperatures. During the melt season, there are strong linkages between the amount of solar energy absorbed, snow-ice-albedo feedbacks, and snow and ice melt. Additionally, the unique sea ice characteristics present within the CAA facilitate a large degree of interannual variability during the melt season, as sea ice does not always entirely melt each year in northern regions.

Chapter 3 expands on the work presented in Chapter 2 through the application of IMS products for detecting sea ice, lake ice, and snow phenology at the pan-Arctic scale. The inclusion of lakes in Chapter 3 is possible due to the number of large lakes the 4 km IMS product can resolve across the pan-Arctic, as very few large lakes are present across the CAA (and none that can be resolved with 24 km). Using IMS, this chapter examined both long-term snow and ice on/off trends (1997 – 2019) at a 24 km spatial resolution, as well as more recent trends in snow and ice phenology (2004 – 2019) at an improved resolution of 4 km. Results from Chapter 3 show that snow and ice- off trends are becoming earlier, indicating that the pan-Arctic is shifting toward a longer snow and ice-free season. The largest trends were identified for sea ice parameters, with first open water becoming earlier by 7.72 d decade-1 and continuous ice cover becoming later by 10.60 d decade-1. Snow-off and lake ice-off dates are also shifting earlier, though trends were not as large as those identified for sea ice melt parameters. Both snow-on parameters revealed that the pan-Arctic is seeing earlier snow onset (1997 – 2019), though only first snow-on trends were statistically significant (2.79 d decade-1 earlier, p < 0.05). Significant relationships were identified between lake ice and sea ice/snow parameters during the melt season, however no significant correlations were detected during the freeze season. Regionally, the Canadian Arctic, sea ice freeze and snow onset are shifting earlier (as identified in Chapter 2), while other regions in the pan-Arctic are shifting later, providing further evidence that the Canadian Arctic is responding differently to warming compared to other Arctic regions. Sea ice and lake ice within the Alaska/Russia region showed the largest trends toward earlier melt and later freeze, consistent with strong warming

118

trends detected over Alaska since 2004. Eurasia showed smaller shifts toward earlier snow-off compared to North America, with northwest Eurasia showing larger shifts toward earlier snow-off compared to northeast Eurasia. Similar spatial patterns were detected in lake ice-off dates, with Lake Ladoga and Lake Onega (northwest Eurasia) showing the largest trends toward earlier open water across the entire Eurasian Arctic. Snow onset trends over Eurasia are similar to those detected over North America, with snow onset becoming earlier over both regions, consistent with previous studies that have documented increases in total annual precipitation over the Arctic during the winter (Box et al., 2019).

Overall, this work highlights 1) how finer resolution satellite data can improve estimates of snow and ice cover in the Arctic, 2) the complex relationships between sea ice, lake ice, snow cover, and climate, and 3) the large degree of regional and interannual variability in snow and ice cover. Continued research should be conducted using finer-resolution satellite data to account for changes in snow and ice cover which are not visible in coarse resolution imagery, as this can affect estimates of snow and ice parameters and result in inaccurate estimates related to timing of snow and ice melt/freeze. Additionally, work analyzing multiple components of the cryosphere together (versus analyzing one component independently) can provide useful information regarding the relationship(s) between multiple snow/ice parameters. As most literature generally focuses on assessing snow or ice independently, it is difficult to understand how different snow and ice are related and how they influence each other. Furthermore, the responses of snow and ice cover will vary greatly regionally, highlighting the importance of studying snow and ice not only at a pan- Arctic scale, but also at a regional scale. Different regions will be characterized by unique snow, ice, and climate regimes (for example, the CAA as shown in Chapters 2 and 3), therefore examining snow and ice cover at a pan-Arctic scale only provides limited information on how snow and ice are responding to recent climatic changes. As the 1 km IMS product launched in 2014, this will provide useful information regarding snow and ice cover at an improved spatial resolution in upcoming years compared to traditional microwave remote sensing.

119

References

Box, J.E., Colgan, W.T., Christensen, T.R., Schmidt, N.M., Lund, M., Parmentier, F.J.W., Brown,

R., Bhatt, U.S., Euskirchen, E.S., Romanovsky, V.E. and Walsh, J.E. 2019. Key indicators of

Arctic climate change: 1971–2017. Environ. Res. Lett. 14: 045010. doi: https://doi.org/10.1088/1748-9326/aafc1b.

Dauginis, A. and Brown, L. C. 2020. Sea ice and snow phenology in the Canadian Arctic

Archipelago from 1997 – 2018. Arct. Sci. Revised manuscript submitted on Sept 2nd 2020, ID:

AS-2020-0024.R1.

Howell, S. E., Tivy, A., Yackel, J. J., and Scharien, R. K. 2006. Application of a

SeaWinds/QuikSCAT sea ice melt algorithm for assessing melt dynamics in the Canadian Arctic

Archipelago. J. Geophys. Res. 111: C07025. doi:10.1029/2005JC003193.

Markus, T., Stroeve, J. C., and Miller, J. 2009. Recent changes in Arctic sea ice melt onset, freezeup, and melt season length. J. Geophys. Res. 114: C12024. doi:10.1029/2009JC005436.

Wang, L., Wolken, G.J., Sharp, M.J., Howell, S.E.L., Derksen, C., Brown, R.D., Markus, T. and

Cole, J. 2011. Integrated pan‐Arctic melt onset detection from satellite active and passive microwave measurements, 2000–2009. J. Geophys. Res. 116: D22103. doi:10.1029/2011JD016256.

WWF. 2018. Last Ice Area. Available online at http://www.wwf.ca/conservation/arctic/lia/

120 121

Appendices

Appendix 1- IMS data sources.

IMS data sources as of 2015 (Source: NSDIC, https://nsidc.org/data/g02156 Table 11) V1: V2: V3: Sensor or Source Platform or Orgization 1998 - 2004 - 2014 - 2004 2014 present ACNFS sea ice area fraction and NIC x sea ice thickness AMSR-2 GCOM-W x NOAA POES Satellites (15 - 18), Aqua, AMSU x x x EUMETSAT MetOp-A ASCAT EUMETSAT MetOp-A x x ATMS (MIRS based) S-NPP x Automated snow detection layers NESDIS and NCEP x x x NOAA POES Satellites (14 - 19), AVHRR x x x EUMETSAT MetOp-A Canadian snow analysis Environment Canada x GFS daily snow depth NCEP x GMS Imager JMA GMS-5 (Himawari 5) x x GOES Imager NOAA GOES Satellites (9, 10, 11, 13) x x x Hourly surface weather reports METAR x MODIS Aqua and Terra x x MTSAT-1R Imager JMA MTSAT-1R (Himawari 6) x MTSAT-2 Imager JMA MTSAT-2 (Himawari 7) x MVIRI MFG x x Various radar published from Europe, Radar Japan, China, South Korea, Canada, or x U.S. SAR Radarsat-2 x SAR (C-band) Sentinel-1A x SEVIRI MSG x SNODAS NOHRSC x x x SSM/I DMSP Satellites x x x SSMIS DMSP Satellites x x U.S. Air Force Snow and Ice USAF x x x Analysis Product Various weather reports, ice charts, In situ data from U.S. and other foreign x and snow depth reports countries VIIRS Binary Snow Cover EDR NASA Goddard x VIIRS Sea Ice Characterization NASA Goddard x EDR VIIRS (visible channels 1,2,3, IR S-NPP Satellites x channel 15, day/night bands) Weekly sea ice analysis and ice NIC x x x edge