The Impacts of Climate Change on Potential Permafrost Distributions from the Subarctic to the High Arctic Regions in Canada

Andrew Tam

A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy Department of Geography University of Toronto

Andrew Tam

Doctor of Philosophy

Department of Geography University of Toronto

2014 Abstract

A climate change impact assessment on the potential permafrost distributions is presented in four research studies that was conducted using various locations along a geographical south-to-north study transect from 52.2°N to 82.5°N within Canada. The transect begins in Lansdowne House,

Ontario, and ends at Alert, Nunavut, with intermediate locations at Big Trout Lake, Peawanuck,

Fort Severn, Rankin Inlet, Resolute Bay, and Eureka. The first study established the climatic potential for permafrost at Peawanuck from 1959-2011 using the Stefan Frost (Fs) Number index and Stefan equation for active layer thicknesses. Fs and the Stefan depths demonstrated favourable potential for permafrost; however, freezing degree-days were observed to be declining. The second study examined active layers developments at five locations from 2004-

2011 using the Xie-Gough Algorithm for multilayered soil profiles. Climate conditions for potential permafrost distributions were assessed and compared using Fs. The third study explored the changes in the potential permafrost using Fs under future climate warming scenarios projected by an ensemble of Global Climate Models for 2011-2100. Climate change projections within the transect indicate warming above the 1971-2000 mean air temperature baseline by +1.5

ii to +2.4°C for 2011-2040; +2.6 to +4.1°C for 2041-2070; and +3.3 to +7.1°C for 2071-2100. For this century, Fs projections indicate that climate conditions will remain supportive for continuous permafrost distributions at Resolute Bay, Eureka, and Alert. By 2040, conditions for Rankin Inlet indicate change from continuous to discontinuous permafrost. For Peawanuck, conditions by

2100 are projecting to be suitable for sporadic permafrost. The fourth study focuses at

Peawanuck and three other locations within northern Ontario, and assessed the behaviour of palsa formation and occurrence in the context of climate change for the 2020s, 2050s, and 2080s.

By the end of this century, warming projections support palsa occurrence; however, conditions will no longer support new palsa formation.

iii

Acknowledgments

First and foremost, I would like to express my sincerest gratitude to my supervisor, Dr. William

Gough, for his continuous support of my research endeavors, for providing me with professional knowledge, and for his guidance during my pursuant of a Doctoral degree.

I would like to recognize my PhD committee: Drs. George Arhonditsis, Carl Mitchell, and

Mathew Wells. I would like to thank each member for their support, professional guidance, and motivation during my entire graduate experience.

I would like to acknowledge and express my gratitude to Dr. Changwei Xie, Cold and Arid

Regions Environmental and Engineering Research Institute, Chinese Academy of Sciences, for sharing his expertise in permafrost research.

I would like to thank my parents (Stephen & Lily) and my brother and sister-in-law (Charles &

Angela), for their love, encouragement, and support. I would also like to thank all my friends and supporters in Trenton and Toronto whom continue to inspire me in my pursuit of happiness.

This thesis is dedicated to the late Mr. Donald (Don) Kovanen (1954-2012) of the Department of

National Defence. Don was always supportive of my scientific pursuits, and he was instrumental in enabling my exploration of the Great Canadian High Arctic.

Portions of this work were funded and supported by the Department of Physical and

Environmental Sciences at the University of Toronto Scarborough, the Wildlife Research and

Development Section of the Ontario Ministry of National Resources, the Ontario Ministry of

Environment, and the Department of National Defence at Trenton, Alert, and Eureka.

Table of Contents

Contents

Acknowledgments ...... iv

Table of Contents ...... v

List of Tables ...... x

List of Figures ...... xi

Statement of Co-authorship ...... xiv

Chapter 1 Introduction and background ...... 1

1 Chapter 1 ...... 1

1.1 Introduction ...... 1

1.2 Permafrost ...... 2

1.2.1 Permafrost distributions ...... 3

1.2.2 Freezing process ...... 4

1.2.3 Palsas ...... 5

1.3 Permafrost characterization ...... 7

1.3.1 Degree-Days ...... 7

1.3.2 Frost Number ...... 8

1.3.3 Stefan Depths ...... 9

1.3.4 Stefan Frost Number ...... 11

1.3.5 XG-Algorithm ...... 12

1.3.6 TTOP Model ...... 14

1.3.7 The Kudryavtsev Equation ...... 15

1.3.8 Active layer thickness measurements ...... 16

1.3.9 Soil samples and analyses ...... 18

1.4 Climate change ...... 21

1.4.1 Climate modelling and future scenarios ...... 21

1.4.2 Climate change impacts ...... 24

1.5 Research aim, objectives, and questions ...... 27

1.6 Chapter 2 ...... 28

1.7 Chapter 3 ...... 29

1.8 Chapter 4 ...... 29

1.9 Chapter 5 ...... 30

1.10 Study sites ...... 31

1.10.1 Permafrost research at adjacent study sites ...... 33

1.11 References ...... 38

Chapter 2 An assessment of potential permafrost at Peawanuck, Ontario, from 1959-2011 ...... 44

2 Chapter 2 ...... 44

2.1 Abstract ...... 44

2.2 Introduction ...... 45

2.3 Methodology ...... 48

2.3.1 Study Area and Geography ...... 48

2.3.2 Soil Characterization ...... 50

2.3.3 Climate Data ...... 51

2.3.4 The Stefan Equation ...... 53

2.3.5 The Stefan Frost Number ...... 54

2.4 Results ...... 55

2.4.1 Stefan Frost Numbers for 1959 to 2011 ...... 55

2.4.2 Stefan Depths from 1959 to 2011 ...... 60

2.5 Discussion ...... 61

2.5.1 Stefan Frost Number Index for Peawanuck ...... 61 vi

2.5.2 Stefan Depths of Freezing and Thawing ...... 64

2.6 Conclusion ...... 65

2.7 Acknowledgements ...... 66

2.8 References ...... 66

Chapter 3 An application of the Stefan Frost Number and XG-Algorithm in the Canadian Subarctic and Arctic Regions from 2004 to 2011 ...... 69

3 Chapter 3 ...... 69

3.1 Abstract ...... 69

3.2 Introduction ...... 70

3.3 Methodology ...... 73

3.3.1 Description of the study locations ...... 73

3.3.2 Freezing and thawing degree-days ...... 78

3.3.3 The Frost Number and the Stefan Frost Number ...... 78

3.3.4 The XG-Algorithm ...... 80

3.3.5 Algorithm input parameters ...... 82

3.3.6 Error Calculations ...... 83

3.4 Results ...... 84

3.4.1 Frost Number and Stefan Frost Number ...... 84

3.4.2 Active Layer Simulations ...... 87

3.5 Discussion ...... 89

3.5.1 Field Validation XG-Algorithm ...... 89

3.5.2 Addressing the Research Question ...... 92

3.6 Conclusion ...... 94

3.7 Acknowledgements ...... 95

3.8 References ...... 96

vii

Chapter 4 An assessment of potential permafrost along a south-to-north transect in Canada under predicted climate warming scenarios from 2011 to 2100 ...... 99

4 Chapter 4 ...... 99

4.1 Abstract ...... 99

4.2 Introduction ...... 100

4.3 Methodology ...... 105

4.3.1 Study locations ...... 105

4.3.2 Global Climate Models and the Localizer Tool ...... 108

4.3.3 Frost Number and Stefan Frost Number ...... 109

4.3.4 Site-specific conditions ...... 111

4.3.5 Regression analysis ...... 112

4.4 Results ...... 113

4.4.1 Climate baseline 1961-2011 ...... 113

4.4.2 Future climate scenarios 2011-2100 ...... 114

4.4.3 Stefan Frost Numbers 1971-2100 ...... 121

4.5 Discussion ...... 122

4.5.1 Future climate conditions for permafrost distributions ...... 122

4.5.2 Limitations and errors ...... 125

4.6 Conclusion ...... 126

4.7 Acknowledgements ...... 127

4.8 References ...... 128

viii

Chapter 5 The fate of Hudson Bay Lowlands palsas in a changing climate ...... 132

5 Chapter 5 ...... 132

5.1 Abstract ...... 132

5.2 Introduction ...... 133

5.2.1 Study Area ...... 135

5.3 Methods ...... 137

5.3.1 Climate Data ...... 137

5.3.2 Data Analysis ...... 138

5.4 Results ...... 140

5.4.1 Climate data analysis ...... 140

5.4.2 Climate projections ...... 143

5.4.3 MAAT threshold ...... 144

5.4.4 Number of Days below -10oC per year ...... 147

5.5 Discussion ...... 148

5.6 Conclusion ...... 150

5.7 Acknowledgements ...... 151

5.8 References ...... 151

Chapter 6 Discussion and Conclusion ...... 154

6 Chapter 6 ...... 154

6.1 Discussion and conclusion ...... 154

6.2 Recommendations for further research ...... 163

6.3 References ...... 166

List of Tables

Table 2-1. Description of sample locations at Winisk and Peawanuck, Ontario...... 50

Table 2-2. Climate Data from Environment Canada for Peawanuck and Winisk, Ontario...... 52

Table 2-3. Soil properties and theoretical scenarios of Kf/Ku ratios for Peawanuck, Ontario...... 53

Table 2-4. Trends in degree-days for Peawanuck, Ontario, from 1959 to 2011...... 59

Table 3-1. Study location details ...... 74

Table 3-2. Model Input Parameters ...... 83

Table 3-3. Statistical analysis of active layer depth trends from 2004 to 2011...... 89

Table 3-4. Systematic and root mean square errors for the XG-Algorithm at two observation locations ...... 90

Table 4-1. Description of Study Locations ...... 107

Table 4-2. Summary of Climate Baseline Analysis ...... 114

Table 5-1. Criteria for palsas for the Hudson Bay Lowlands...... 138

Table 5-3. Observed climate and downscaled climate projections for Big Trout Lake and Lansdowne House...... 144

Table 5-4. Projected MAAT values for the four weather stations...... 145

Table 5-5. Projected number of days below -10oC per year for Big Trout Lake and Lansdowne House...... 147 x

List of Figures

Figure 1-1. Photograph of a piece of permafrost extracted from the base of a test pit near Alert, Nunavut, and showing the presence of ice in the ground material. Photo taken by A. Tam...... 3

Figure 1-2. Photograph of a developing “baby” palsa near Peawanuck, Ontario. Photo taken by W.A. Gough...... 5

Figure 1-3. Active layer thickness measurement near Peawanuck, Ontario, using a graduated stainless steel probe. Photo taken by W.A. Gough...... 16

Figure 1-4. Backhoe excavation of a soil test pit for active layer thickness measurement near Alert, Nunavut. Photo taken by A. Tam...... 17

Figure 1-5. Base of an excavated test pit showing direct visible presence of ice. Photo taken by A. Tam...... 18

Figure 1-6. Extracting intact samples of permafrost from the base of an excavated test pit using a jackhammer near Alert, Nunavut. Photo taken by A.Tam...... 19

Figure 1-7. Photograph of soil samples collected in plastic bags. Photo taken by A. Tam...... 20

Figure 1-8. The Localizer Tool projection output for mean air temperature anomaly at Eureka, Nunavut, under using A2 emission scenario (mean of 20 models) for the period of 2041-2070 with baseline using 1971-2000...... 23

Figure 1-9. Photograph of a damaged concrete floor caused by deepening active layer and shifting permafrost. A Canadian 25 cent quarter is used for scale; located in the mid-right section of the photo. Photo taken by A.Tam...... 25

Figure 1-10. Map of all study locations along the south-to-north transect...... 32

Figure 2-1. Location map containing the study area and Peawanuck, Ontario...... 49

Figure 2-2. Landscape of the Hudson Bay Lowlands near Peawanuck, Ontario. Photo taken by W.A. Gough...... 51 xi

Figure 2-3. Stefan Frost Number Index results for Peawanuck peat, silt, and clay for a) 1959- 1977; and, b) 1995-2011; showing the 0.67 threshold for continuous and 0.50 for unfavourable permafrost distributions...... 56

Figure 2-4. Stefan Frost Number Index results for three thermal conductivity scenarios with

Kf/Ku ratios at 1.0, 1.5 and 2.0 for a) 1959-1977 and b) 1995-2011; showing the 0.67 threshold for continuous and 0.50 for unfavourable permafrost distributions...... 58

Figure 2-5. Stefan thawing (Xu) and freezing (Xf) depths are shown for a) Peat, b) Silt, and c)

Clay soil compositions at Peawanuck, Ontario, from 1959 to 2011 with ΔX (Xf –Xu)...... 60

Figure 3-1. The geographical south to north transect map containing the five study locations within the Canadian Subarctic, Low Arctic, and High Arctic ...... 72

Figure 3-2. Site condition at local weather stations taken at A) Peawanuck, Ontario (Hudson Bay Lowlands); B) Resolute Bay, Nunavut; C) Eureka, Nunavut; and D) Alert, Nunavut...... 73

Figure 3-3. Site-specific ground material profiles for A) Peawanuck, Ontario (Hudson Bay Lowlands, HBL); B) Rankin Inlet, Nunavut (RAN); C) Resolute Bay, Nunavut (RES); D) Eureka, Nunavut (ERK); and E) Alert, Nunavut (ALR)...... 77

Figure 3-4. Results for A) using Frost Number (F) for all five study locations; B) using Stefan Frost Number (Fs) for all five study locations; and, C) Stefan Frost Number for the HBL using three conditions of Kf/Ku ratios from 1.0, 1.6 and 2.0; with ±1-standard deviation error bars and showing the 0.67 threshold for continuous permafrost distribution...... 86

Figure 3-5. Output active layer thickness depths from XG-Algorithm for: A) Peawanuck, Ontario (Hudson Bay Lowlands); B) Rankin Inlet, Nunavut; C) Resolute Bay, Nunavut; D) Eureka, Nunavut; and E) Alert, Nunavut. F) Change in active layer depths for all five study locations. . 88

Figure 4-1. Map of Study Location in northern Canada...... 106

Figure 4-2. a) Baseline climate data for all five study locations from local observation weather stations; b) Projected MAAT for A1B, A2 and B1 emission scenarios from 2011 to 2100; Stefan Frost Number results for permafrost potential from baseline and projected MAAT under A1B (c), A2 (d), and B1 (e) emission scenarios...... 115 xii

Figure 4-3. Projection of MAAT change rates for the period of 2041-2070 for Alert, Nunavut (left, a-c) and Eureka, Nunavut (right, d-f), under IPCC A1B, A2, and B1 (top to bottom) emission scenarios. The ‘+’ shows the locations of the study sites...... 117

Figure 4-4. Projection of MAAT change rates for the period of 2041-2070 for Resolute Bay, Nunavut (left, a-c) and Rankin Inlet, Nunavut (right, d-f), under IPCC A1B, A2, and B1 (top to bottom) emission scenarios. The ‘+’ shows the locations of the study sites...... 119

Figure 4-5. Projection of MAAT change rates for the period of 2041-2070 for Peawanuck, Ontario (a-c), under IPCC A1B, A2, and B1 (top to bottom) emission scenarios. The ‘+’ shows the location of the study site...... 120

Figure 5-1. The Study Area – Hudson Bay Lowlands of northern Ontario, Canada...... 136

Figure 5-2. Temperature trends for Big Trout Lake for the period of 1951 – 2010; due to incomplete data, the following years could not be applied: 1990, 1992-4, 1996-7, 2006, and 2008-10...... 141

xiii

Statement of Co-authorship

At the time of this thesis submission, Chapters 2 to 5 are research manuscripts that are currently submitted and under review in peer-reviewed journals. The fourth manuscript has been accepted for publication. These manuscripts were written with the intent to serve as standalone research studies. For each manuscript, I was the primary author responsible for research project planning, study design, field logistics, data collection and analyses, and writing. My supervisor,

Dr. William Gough, provided supervision and guidance throughout the entire process of this research; Dr. Changwei Xie, visiting professor at the University of Toronto Climate Lab provided contributions to all manuscripts; Mr. Slawomir Kowal, colleague at the University of

Toronto Climate Lab provided contributions to the fourth manuscript. All co-author contributions to the each manuscript are described below.

Chapter 2

Chapter 2 is co-authored by Drs. William Gough and Changwei Xie. Both co-authors provided guidance, expert advice and editorial input.

Citation: Tam, A., Gough, W.A., and Xie, C. (2014). An assessment of potential permafrost at

Peawanuck, Ontario, from 1959-2011.

xiv

Chapter 3

Chapter 3 is co-authored by Drs. William Gough and Changwei Xie. Both co-authors provided guidance, expert advice and editorial input.

Citation: Tam, A., Gough, W.A., and Xie, C. (2014). An application of the Stefan Frost Number and XG-Algorithm in the Canadian Subarctic and Arctic Regions from 2004 to 2011.

Chapter 4

Chapter 4 is co-authored by Drs. William Gough and Changwei Xie. Both co-authors provided guidance, expert advice and editorial input.

Citation: Tam, A., Gough, W.A., and Xie, C. (2014). An assessment of potential permafrost along a south-to-north transect in Canada under projected climate warming scenarios from 2011 to 2100.

Chapter 5

Chapter 5 is co-authored by Dr. William Gough, Dr. Changwei Xie, and Mr. Slawomir Kowal.

The co-authors provided expert advice and editorial input. This paper will be part of a special edition of AAAR which will feature the Hudson Bay Lowlands. Dr. Gough took the initial lead on this paper. I refined the analysis and re-wrote the manuscript and addressed the concerns of the reviewers. Dr. Xie and Mr. Kowal determined the southern extent of palsas in the HBL in an aerial survey. The Statistical Downscaling Model (SDSM) data obtained for this work was completed by Joyce Zhang, and funded through a third party contract with the Ontario Ministry of Environment.

Citation: Tam, A., Gough, W.A., Kowal, S., and Xie, C. (2014). The fate of Hudson Bay

Lowlands palsas in a changing climate. (In press). Arctic, Antarctic, and Alpine Research.

xvi 1

Chapter 1 Introduction and background 1 Chapter 1

1.1 Introduction

Permafrost underlies nearly 25% of the land mass within the northern hemisphere;

Canada possesses the world`s second largest extent of permafrost, underlying 50% of Canada’s land mass (Pukonen, 1998; Smith et al., 2005; French, 2007; Duan and Naterer, 2009; Dobinski,

2011; Derksen et al., 2012; McClymont et al., 2013). Permafrost is a terrestrial component of the periglacial domain that forms a part of the cryosphere (French, 2007). With observations and projections of climate change, current permafrost distributions are expected to change (Smith et al., 2005). Changes in permafrost distributions will have impacts, such as shifting ground conditions, on the environment and infrastructures built in Canada’s northern communities

(Smith et al., 2005; Zhang et al., 2008; Rinke et al., 2012; Throop et al., 2012; Slater and

Lawrence, 2013). This research presents potential permafrost distributions, as followed in

Anisimov et al. (1997), and acknowledges the existence of a time lag response between any temperature changes at the surface and the underlying permafrost layer (Slater and Lawrence,

2013). Change in potential permafrost distributions will take additional time due to the thermal inertia of ice adjusting to a new thermal equilibrium (Slater and Lawrence, 2013). The resulting shifts can include changes from the current permafrost distributions, such as from continuous to discontinuous distributions, discontinuous to sporadic, and ultimately to an absence of permafrost (Anisimov et al., 1997; Slater and Lawrence, 2013). As current permafrost

2 characterizations are typically generalized by the consideration of climatological properties over vast geography, this research demonstrated the inclusion of the site-specific soil thermal properties in the application of various permafrost tools. These tools include the Stefan equation, the Xie-Gough Algorithm, the Frost Number Index, and the Stefan Frost Number that may be coupled with the results from Global Climate Models (GCMs) to project future climates conditions for potential permafrost from 2011 to 2100.

1.2 Permafrost

Permafrost is defined as ground material that remains below 0ºC in a perennial frozen state for at least two consecutive years, a definition based solely on temperature (Gough and

Leung, 2002; Smith and Burgess, 2002; Shur and Jorgenson, 2007; French and Shur, 2010;

Derksen et al., 2012; McClymont et al., 2013). Permafrost is classified under the Cryosol soil orders due to the presence of cryoturbation and ice formation (Bockheim et al., 2006; Juma,

2006). Permafrost may exist in unconsolidated ground materials such as soils and gravels, and even in bedrock as shown in Figure 1-1. Typically, permafrost exists with the presence of ice formations as interstitial cementing particles or as bodies of ice (Duan and Naterer, 2009).

Ground material with little soil moisture content can exist as dry permafrost. The soil layer above the permafrost is known as the active layer that experiences seasonal freezing and thawing

(Smith and Burgess, 2002; Sazonova et al., 2004; French, 2007; French and Shur, 2010;

Dobinski, 2011; Slater and Lawrence, 2013). The active layer thickness is typically thinner in the higher latitudes and altitudes; thicker active layers are typical for lower latitudes and altitudes

(Dobinski, 2011). The permafrost table is located at the interface between the base of the active layer and the top of the permafrost layer (Dobinski, 2011).

Figure 1-1. Photograph of a piece of permafrost extracted from the base of a test pit near Alert,

Nunavut, and showing the presence of ice in the ground material. Photo taken by A. Tam.

1.2.1 Permafrost distributions

The distribution of permafrost can be assessed horizontally using the terms continuous and discontinuous. Continuous permafrost is considered when the permafrost layer is free of any interruptions. The degree of interruption in the permafrost distribution can be considered discontinuous, sporadic, isolated and permafrost absent. The typical areal distributions of permafrost are 90-100% for continuous, 50-90% for discontinuous, 10-50% for sporadic, and 0-

10% for isolated (French, 2007). Permafrost distributions that are not supported by current climatic conditions are known as relict (French, 2007).

1.2.2 Freezing process

Depending on the soil type, soil thermal conductivity, and soil moisture content, different soils will have different rates of freezing and thawing (French, 2007; Akinyemi et al., 2011;

Dobinski, 2011; Meurth and Mauser, 2012; McClymont et al., 2013). When temperatures descend to the freezing point and below, heat energy is released from the ground and into the surface air. As the active layer begins to freeze from the top downwards into the subsurface, the soil moisture content located in the soil pore space, the gravitational and hygroscopic waters, also begins to freeze and establishes a freezing plane that has a temperature of 0°C; this process is also known as the zero-curtain effect (Hinkel et al., 2001; French, 2007; Dobinski, 2011).

Below the freezing plane, liquid water will migrate toward the plane due to electrostatic and osmotic forces, known as cryosuction, reducing the soil moisture content in the adjacent materials (Harris, 1986). As the moisture freezes, latent heat of fusion is released at 3.35 x 105 J kg-1; once the heat loss to the surface exceeds the latent heat of fusion, the freezing plane will begin to descend into the active layer. At the base of the active layer, at the permafrost table, up freezing may also occur to freeze the ground material. The soil temperature gradient within the active and permafrost layers is dependent on soil thermal conductivity (Akinyemi et al., 2011).

During the freezing process, soil thermal conductivity serves as an important control as the frozen soil thermal conductivity is typically greater than the unfrozen soil thermal conductivity because the conductivity of ice is four times greater than that of water (Nixon and McRoberts,

1973; French, 2007; Shur and Jorgenson, 2007; Tam, 2009).

1.2.3 Palsas

Palsas are typically characterized as surface features that are dome-shaped mounds and are composed of a frozen core consisting of sediments and ice as shown in Figure 1-2 (Seppälä,

1986).

Figure 1-2. Photograph of a developing “baby” palsa near Peawanuck, Ontario. Photo taken by

W.A. Gough.

Alternating ice lens layers contribute to the dome-shape of the mounds. Palsas have been observed within subarctic regions of the world at northern Canada, Alaska, Iceland, Northern

Scandinavia, and Siberia (Kershaw and Gill, 1979; Seppälä, 1986; Zuidhoff and Kolstrup, 2000;

Gurney, 2001; Hinkel et al., 2001; Lewkowicz and Coultish, 2004; Vallée and Payette, 2007;

Tsuyusaki et al., 2008; Kujala et al., 2008; Kirpotin et al., 2009; Tam, 2009; Thibault and

Payette, 2009; Cyr and Payette, 2010; Saemundsson et al., 2012; Pengerud et al., 2013). Typical heights of palsas range from 0.4 to 10 meters with diameters ranging from a few metres to tens of metres (Brown, 1973; Seppälä, 1986; Kuhry, 2008; Kujala et al., 2008). The development of palsas have been observed in soils rich in organic material, such as peat, and where climate conditions have a mean annual air temperature of -2oC or lower (Brown, 1973; Seppälä, 1986;

Parviainen and Luoto, 2007; Kujala et al., 2008; Tam; 2009). Within northern Quebec, Canada, the observed mean annual air temperature for palsa presence is 0oC (Cyr and Payette, 2010). The observation of palsas provides surface evidence of the state of the underlying permafrost within a location, primarily as an indicator of discontinuous and sporadic permafrost distributions

(Seppälä, 1986; Dredge, 1992; Vallée and Payette, 2007; Kujala et al., 2008; Kirpotin et al.,

2009).

1.3 Permafrost characterization

Air temperatures and ground properties are key variables in characterizing the potential permafrost distribution and in determining the extent of permafrost formation and degradation

(French, 2007).

1.3.1 Degree-Days

Degree-days (DD) indices are based on temperature measurements calculated from the accumulation of daily temperatures above and below a critical set threshold in units of degree-

Celsius-days (Juliussen and Humlum, 2007). Thawing degree-days (TDD) is calculated by the sum of the daily temperature above the 0ºC threshold temperature for a given period with units in

ºC·days. For the freezing degree-day (FDD, ºC·days) calculation, the daily temperature below the 0ºC threshold is applied. The FDD is expressed as a positive value in this thesis. These degree-days techniques have been applied in engineering applications of relating climate conditions with ground freezing and thawing actions (French, 2007). All air temperature data applied in the degree-days techniques were provided by Environment Canada’s National Climate

Data and Information Archive.

1.3.2 Frost Number

An effective method to determine the distribution of permafrost was provided by Nelson and Outcalt in 1983 with further computational procedures outlined in Nelson (1986) and Nelson and Outcalt (1987). The Frost number (F) uses climatological data to provide an index classification. Continuous permafrost has a Frost number threshold of F ≥ 0.67. Discontinuous permafrost with F < 0.67; sporadic at F < 0.6; and, permafrost absent at F < 0.5 (Nelson and

Outcalt, 1987). F is a dimensionless ratio between the freezing and thawing degree-days accumulations, represented as:

FDD F  . (1-1) FDD  TDD

In Nelson (1986), the F was tested at 57 locations in central Canada within Manitoba,

Saskatchewan, Alberta, Northwest Territories, and Nunavut (then part of the Northwest

Territories). The F computational assessment used 30-year mean monthly climate data from

1951-1980 to calculate the FDD and TDD indices which are shown on isarithmic maps for central Canada (Nelson, 1986). The assessment demonstrated that F is able to define areal extents of permafrost that were also in close agreement with empirical delineations within the study locations (Nelson, 1986).

1.3.3 Stefan Depths

The active layer is the upper subsurface layer that is developed during seasonal thawing during the warm seasons, and completely freezes during the cool seasons (French and Shur,

2010). Simulations of active layer thicknesses can provide indications of changes that may affect the permafrost. The original Stefan equation is an analytical solution, derived by Josef Stefan, for the moving boundary problem of freezing on polar ice caps (Crepeau, 2007). Stefan applied a simple conservation of energy model, L·ρ·dh, where L is the latent heat of fusion of ice (J/kg), ρ is the soil bulk density (kg/m3), dh is the observed ice thickness, a length scale (m), to represent the heat loss per unit area in an amount with Fourier’s law of heat conduction, K·ΔT/h(t)·dt, where K represents the soil thermal conductivity (W/(m·k)); ΔT is the difference between the water-ice and air-ice interface temperatures (°C), and t, time in seconds (Crepeau, 2007). Stefan formulated that the square of ice thickness was a function of linear time and temperature, or alternately, ice thickness as a function of the square root of time:

( ) . (1-2)

Stefan expressed that equation 1-2 was an oversimplification of the problem; however, comparisons between the model results with experimental data yielded rough agreement, with some errors being attributed to the difficulty in obtaining accurate soil thermal conductivity values at the time (Crepeau, 2007). Stefan further experimented by including time-dependent temperature and accounted for the boundary change at the air-ice surface interface into a heat diffusion equation using advanced mathematics; the results of this experiment yielded an

10 approximate solution that was identical, and supportive, to the equation 1-2 formulation

(Crepeau, 2007). The conclusions from Stefan’s experimentation revealed that by including time-dependant temperature, the model results were closer to field observations, in comparison to the rough agreement from earlier linear temperature approach, that were made available from field data collected from early British and German polar expeditions (Crepeau, 2007).

The common form of the Stefan equation was expressed in Jumikis (1977) and Lunardini

(1981) to readily predict the freezing and thawing depths (X) in an assumed homogenous soil layer (Nelson and Outcalt, 1987; Broadridge and Pincombe, 1995; Woo et al., 2004; Hayashi et al., 2007; Hughes and Braithwaite, 2008; Zhang et al., 2008; Duan and Naterer, 2009; Xie and

Gough, 2013). This formulation applies the site-specific characteristics of soil thermal properties and available air temperature data with the assumption of negligible effects from sensible heat

(Xie and Gough, 2013):

2K  DD 2K  DD X  ( )0.5  ( )0.5 . (1-3) QL L   

DD represents the degree-days that can be expressed as FDD for freezing and TDD for thawing;

QL is the volumetric latent heat of soil composed of the L; the moisture content, ω; and ρ.

Jumikis (1977) and Lunardini (1981) have applied the Stefan equation for multilayered soils, and more recently in Xie and Gough (2013).

1.3.4 Stefan Frost Number

The Stefan Frost Number (Fs) is a dimensionless ratio, similar to equation 1-1 in formulation, for predicting permafrost distributions for a location by including the subsurface information by using the Stefan equation (equation 1-3) instead of degree-days accumulations

(Nelson, 1986; Nelson and Outcalt 1987).

(1-4)

The Stefan equation (1-2) includes subsurface information in the depths (X) of thawing (t) and freezing (f) formulations by using the soil thermal conductivities that may change respectively to the frozen and unfrozen soil states (Nelson et al. 1997; Woo et al. 2004). The Fs threshold values for predicting permafrost distributions are identical to those of the F (Nelson, 1986;

Nelson and Outcalt 1987).

In Nelson (1986), Fs computations were demonstrated in central Canada using uniform

(homogenous) and non-uniform (heterogeneous) soil properties to account for different K values of different soils. As K is higher in the frozen state of soils, which is attributed to the presence of water occupying the pore fraction of soils, the Kf/Ku ratio was introduced to account for this effect (Nelson, 1986). The Kf/Ku ratio modifies the potential freezing and thawing depths in Fs.

For example, when the FDD/TDD ratio is equal for a uniformed soil, the Kf/Ku ratio can result in deeper freezing depths to provide a possible explanation for the presence of frozen ground. The results from this assessment demonstrated that permafrost distributions are sensitive to soil properties and can also be obtained without detailed data inputs (Nelson, 1986).

1.3.5 XG-Algorithm

The application of the Stefan equation was intended to determine the freezing and thawing depth for a single homogenous soil layer, with the fundamental assumption that soil begins to freeze at 0°C temperature. For multilayered soil profiles, the application of the Stefan equation can lead to misrepresentations associated with arithmetic averaging of the soil physical properties within the multilayered ground (Xie and Gough, 2013). Jumikis (1977) and Lunardini

(1981) had provided an algorithm (JL-Algorithm) for multilayered soils, which has been widely applied, using the differential form of the squared Stefan equation (Xie and Gough, 2013). Xie and Gough (2013) determined a problem within the mathematical derivation of the JL-Algorithm where the parameters K and QL remained constant in the different soil layers. For multilayered soils, the soils from each individual layer are expected to have different physical properties in which K and QL should be different, unless a special condition exists where each soil layer has identical properties. This has implications as the soil temperature gradient will vary within different soils physical properties, which will also affect the accuracy of the freezing and thawing depth calculations (Xie and Gough, 2013). The findings in Xie and Gough (2013) suggested that the JL-Algorithm should not be applied in multilayered soils, where each soil layer has different soil physical properties, as the identical physical properties being considered in the algorithm will produce the same results as the original Stefan equation under homogenous soils. Xie and Gough (2013) then proposed a simple algorithm (XG-Algorithm) capable of determining the freezing and thawing front in multilayered soil profiles that contain heterogeneous physical properties with varying thickness depths with the underlying assumption that each horizontal layer is homogenous. The XG-Algorithm established an iterative approach

13 in calculating individual Stefan freezing and thawing depths within a multilayered soil profile using ratios between soil physical properties, soil thermal conductivity, and the potential freezing and thawing depths of the previous layers. As shown in Xie and Gough (2013) for a given surface freeze/thaw index, the thaw/freeze depth of two soil types A and B, with respective suffix of a and b, in the same locality can be calculated by Stefan equation (1-3):

2ka  DD 0.5 2ka  DD 0.5 X a  ( )  ( ) , (1-5) QLa L a  a

2kb  DD 0.5 2kb  DD 0.5 X b  ( )  ( ) . (1-6) QLb L b  b

The XG-Algorithm establishes the relationship between the equations (1-5) and (1-6) to produce a ratio between the physical properties (P) of both layer types A and B:

X a 2ka  DD/(L  a a ) 0.5 ka  b b 0.5 Pab   ( )  ( ) . (1-7) X b 2kb  DD/(L  b b ) kb  a a

The ratio Pab, equation 1-7, can then be applied to determine the freeze/thaw depth of Soil Type

B (Xb) when Xa is known:

X X  a b P ab . (1-8)

1.3.6 TTOP Model

The temperature at the top of the permafrost, TTOP, model is provided in Smith and

Riseborough (1996) as a tool for determining permafrost distributions based on ground temperatures and by establishing a climate-permafrost relationship. The TTOP model applies the difference between the TDD and FDD accumulations with a ratio of the Ku to Kf, and further includes n-scaling factors between summer air and surface thawing indices, nt, to account for vegetation thermal effects; and nf, a scaling factor between winter air and surface freezing indices to represent the thermal effects of snow cover (Smith and Riseborough, 1996; Henry and

Smith, 2001; Smith and Riseborough, 2002). The TTOP model formulation is shown in Smith and Riseborough (1996) as:

( ) (1-9)

This research could not apply the TTOP Model due to limited available data on the surface vegetation layers, snow cover, and ground temperature observations at all study locations.

1.3.7 The Kudryavtsev Equation

The Kudryavtsev equation can be applied to determine the seasonal depths of thawing or freezing in the formulation as shown in Anisimov et al. (1997):

( ) ( ) ( ) ( )

( ) ( ) (1-10)

Where Z represents the depth of thawing or freezing in metres; As is the annual amplitude of the surface temperature (°C); Tz is the mean annual temperature at the depth of seasonal thawing

(°C); the variables AZ and Zc are further described in Anisimov et al. (1997); K and C are the thermal conductivity (W/m°C) and the volumetric heat capacity (J/m3°C) of the soil; P is the period of the annual temperature cycle (1 year in seconds); and QL is the volumetric latent heat of fusion (J/m3). The Kudryavtsev equation increases the complexity of permafrost modelling by requiring additional data inputs then the requirement for the Stefan equation (Anisimov et al.,

1997). This permafrost tool incorporates a greater detail of the temperature and physical properties of the ground materials and accounts for the seasonal thermal effects by the presence of snow and vegetation covers (Anisimov et al., 1997). A validation exercise for the Kudryavtsev equation was conducted by Anisimov et al. (1997), the errors of outputs were below 10% between the output results compared with actual data collected from sites in Russia and Alaska.

The Kudryavtsev equation is applied in most permafrost modelling techniques; however, due to the intensive data requirements, this approach could not be applied for this research due to limited available data at all study locations.

1.3.8 Active layer thickness measurements

This research conducted active layer thickness measurements at the Peawanuck, Ontario, and Alert, Nunavut, using mechanical probing methods such as applying an active layer thickness probe consisting of a graduated stainless steel rod (Figure 1-3) and by direct measurements of active layer thickness from excavated test pits (Figure 1-4).

Figure 1-3. Active layer thickness measurement near Peawanuck, Ontario, using a graduated stainless steel probe. Photo taken by W.A. Gough.

Figure 1-4. Backhoe excavation of a soil test pit for active layer thickness measurement near

Alert, Nunavut. Photo taken by A. Tam.

Test pits were excavated by backhoe, where accessible during the summer season, to about 2-3 m in length and 1 m in width with depths varying until frozen material were observed by the direct presence of ice particles (Figure 1-5). In the event that direct presence of ice was not visibly detected or shallow bedrock had been reached, the test pit would be re-filled and re- compacted appropriately. A new test pit in an adjacent location would be selected for excavation.

Within the test pits, the active layer thickness was measured using a hand measuring tape from the top surface until the upper limit of frozen ground, where a distinct ice band can be visibly seen within the exposed soil test pit.

Figure 1-5. Base of an excavated test pit showing direct visible presence of ice. Photo taken by

A. Tam.

1.3.9 Soil samples and analyses

Soil samples were collected from excavated test pits at specific depths of 5 cm from the surface and at the base of the pits. These soil samples were retrieved from the test pit walls after an additional removal of 5 cm horizontal of ground material by hand tools to remove any thermal impacted soils by the backhoe excavation process and from prolonged exposure to the environment. The soil samples from the top and bottom of the pits were placed in labelled glass soil sampling jars (500 mL).

Figure 1-6. Extracting intact samples of permafrost from the base of an excavated test pit using a jackhammer near Alert, Nunavut. Photo taken by A.Tam.

Large samples of frozen soils were extracted using an electric jackhammer connected to a portable diesel generator and hand pick axes (Figure 1-6). The large pieces of intact frozen soil were collected as samples and placed into “Whirlpak” plastic bags and sealed (Figure 1-7).

Figure 1-7. Photograph of soil samples collected in plastic bags. Photo taken by A. Tam.

Soil samples were packaged shipped by air in sealed thermal containers with ice packs to maintain cold temperatures. The soil samples that were retrieved from field visits in 2007-2008 from Winisk and Peawanuck were analyzed for soil composition and gravimetric soil moisture content (%, g/g) at the University of Toronto at Scarborough Soil Laboratory in Toronto, Ontario

(Tam, 2009). Soil samples retrieved from 2009 to 2012 from the arctic locations were sent to

AGAT Laboratories in Mississauga, Ontario, for gravimetric soil moisture content, soil thermal conductivity, and soil texture analyses.

1.4 Climate change

Future projections indicate climate warming during the 21st century (Sazonova et al.,

2004; Zhang et al., 2008; Rinke et al., 2012; Slater and Lawrence, 2013). For the Canadian

North, climate warming is expected to range between +2.0 and +5.0 °C by 2100 (Smith et al.,

2005; Zhang et al., 2008). Air temperature increases are projected to degrade permafrost as the winter seasons are expected to shorten, which will reduce the freezing degree-days accumulations while increasing the thawing degree-days. The Intergovernmental Panel on

Climate Change (IPCC) 4th Assessment Report (AR4) projected that mean global temperatures are expected to warm from +1.1 to +6.4 °C by 2100, with the greatest temperature increase for high latitude regions (Solomon et al., 2007; Zhou et al., 2009). Slater and Lawrence (2013) reported the projected mean temperature changes for the arctic to range from +2.2 to +7.8 °C by

2099, using radiative forcing targets from the IPCC 5th Assessment Report (AR5).

1.4.1 Climate modelling and future scenarios

Projections of future climates may be produced using General Circulation Models

(GCMs) under IPCC AR4 greenhouse gas emission scenarios. Future projections from climate models are typically compared to 30-year climate baselines that were obtained from local observation weather stations. Within this research, the baseline period of 1971-2000 was selected and three future periods were 2011-2040, 2041-2070, and 2071-2100. As there are many different GCMs developed around the world, projections from different climate models will vary. To account for this variation, a multi-mean ensemble approach can be applied by

22 combining the outputs from a variety of available GCMs to produce a mean projected value.

However, GCMs function by producing climate projections based on sophisticated calculation grids containing climate and ocean parameters, the resolution of GCM grids are coarse, such as hundreds by hundreds of kilometres. An example of a multi-mean ensemble approach is the

Localizer Tool produced by the Environment Canada, the University of Toronto, and the

University of Prince Edward Island Climate Sections (UTSC, 2013). The multi-model ensemble projections have a resolution grid of 200 km by 200 km, and produces projections under three

IPCC AR4 scenario experiments: A1B (mean of 24 model), A2 (20 models), and B1 (21 models) as shown in Figure 1-8 (UTSC, 2013). Statistical downscaling model (SDSM) provides an alternate method to enhance spatial and temporal resolutions if local climate data are available.

SDSM statistically links coarse resolution projections to the local scales using site-specific climate data. SDSM data for this research was made available by Zhang (2011).

Figure 1-8. The Localizer Tool projection output for mean air temperature anomaly at Eureka,

Nunavut, under using A2 emission scenario (mean of 20 models) for the period of 2041-2070 with baseline using 1971-2000.

For IPCC AR4 greenhouse gas emission scenarios, such as: A1B, A2, B1, and B2, various future world conditions were considered such as future atmospheric CO2 concentrations, socioeconomics, and global and regional policy management strategies. For example, A1B is representative of a future climate under moderate emissions where atmospheric CO2 concentrations have reached and stabilized at 720 ppm by 2100; for A2, high emissions lead to a continuous atmospheric CO2 concentration increase that reaches 860 ppm by 2100; and, B1 is the low emissions scenario that reflects an environmental conscious and friendly future where atmospheric CO2 concentrations reach 620 ppm (TGICA, 2007; Lewis and Lamoureux, 2010;

Etzelmüller et al., 2011). Additionally, B2 reflects low emissions with a regional approach and a slower but steady growth in population, with emphasis on environmental policies to reduce greenhouse gas emissions, where CO2 concentrations reach 600 ppm by 2100 (Gough and Leung,

2002). At the time of this research, the final draft Report of the Working Group I contribution to the IPCC 5th Assessment Report "Climate Change 2013: The Physical Science Basis" was recently released in fall 2013 with the final report being expected to be released later in 2014

(unavailable at the current time). As the IPCC AR4 relied on projected emission scenarios of atmospheric [CO2], the IPCC AR5 applies projected scenarios using Representative

Concentration Pathway (RCP) radiative forcing targets, such as RCP2.6, RCP4.5, RCP6.0 and

RCP8.5 scenarios. The value in these four RCP scenarios represents the respective radiative forcing in W/m2 and considers future emitted greenhouse concentrations (Slater and Lawrence,

2013).

1.4.2 Climate change impacts

As temperature increases, permafrost may enter a state of degradation that results in the thickening of the active layer. In continuous permafrost areas, thickening of the active layer can reduce the underlying permafrost layer thickness (French, 2007). For discontinuous permafrost regions, thickening of the active layer may result in the ultimate disappearance of the permafrost layer. Climate change in permafrost regions will impact human infrastructures (buildings, transportation, and pipelines) built in Canadian northern communities that are dependent on ground stability (Lewkowicz, 2007; Zhou et al., 2009; Cannone et al., 2010; Smith and

Riseborough, 2010; Berthling and Etzelmüller, 2011; Etzelmüller et al., 2011). Variation in the

25 temperatures can promote repeated freezing and thawing processes that can provide mechanical and thermal stresses that can damage infrastructure foundations as shown in Figure 1-9 and rock

(French, 2007; Berthling and Etzelmüller, 2011).

In the continuous and discontinuous permafrost regions, building foundations are typically elevated and supported on pillars to dissipate heat energy. Most constructed pile foundations are extended to the bedrock for increased stability (French, 2007). However, with the presence of a thick permafrost layer, some pile foundations may not be extended into the

26 bedrock and may rest within the permafrost layer. Thickening of the active layer and melting of the permafrost, under such a situation, may lead to subsidence and sediments movement that may shift the pile foundations and compromise the building structure (French, 2007; Smith and

Riseborough, 2010; Berthling and Etzelmüller, 2011). Transportation networks, such as airfields and seasonal ice roads, connect northern remote communities; however, these networks are dependent on the stability of the ground. Frozen ground provides greater bearing and load capacities that allows for greater shipments and transportation of freight (Smith and

Riseborough, 2010; Dobinski, 2011). With climate change, reductions in the duration of the frozen season and thickness of ice or frozen ground materials can limit the supply of materials into a community (Muller, 2008; Wen et al., 2010). Further load restrictions on vehicles and aircrafts will also have negative impacts for northern communities. Pipelines and utility conduits are built to deliver water, sewage, fuel, and electrical cables in most northern communities. The impacts of frost heaving and subsidence beneath or adjacent to pipelines and conduits can destabilize the foundations and compromise the integrity of the structures by increasing stress and tension. The shifting of ground increases the risk of spills and unintentional releases of materials that can be hazardous to humans and the environment (McCarthy et al., 2004; French,

2007; Smith and Riseborough, 2010).

1.5 Research aim, objectives, and questions

The aim of this research is to assess the potential for Canadian permafrost under changing climate conditions along a geographical south-to-north transect. The primary objective of this research is to identify site-specific characteristics at the study locations to establish present day climate conditions for potential permafrost distributions. The second objective of this research is to determine the climate trends at each location and to assess the potential permafrost distributions using a combination of temperature analysis and permafrost modelling techniques.

The third objective is to determine future climate change projections for the locations within the study transect and assess the impacts of climate change on the potential permafrost.

The main research questions for this thesis are:

1. Can the potential for permafrost be assessed using available permafrost tools at

Peawanuck, Ontario, in the Subarctic region of the Hudson Bay Lowlands? If so, can the presence of permafrost at this location be rationalized by the asymmetric frozen and unfrozen soil thermal conductivities?

2. Can the current distribution of permafrost be readily assessed within the Canadian

Subarctic and Arctic regions based on climatological and ground parameters, and can the active layer thickness profiles at these locations be simulated and compared to determine permafrost change?

3. Can the potential permafrost distributions be determined at five study locations in

Canada using the Stefan Frost Number equation and can future changes to the potential be predicted using GCM results under climate warming scenarios?

4. Can climate criteria for palsas in the HBL be determined by analyzing climate data from available weather stations at the northern and southern palsa limits? If so, what is the projected fate of palsas in the HBL using climate change projections for the region as has been done elsewhere (e.g. Fronzek et al. (2006))?

1.6 Chapter 2

Chapter 2 focuses on assessing the climatic potential for permafrost at Peawanuck,

Ontario, from 1959 to 2011, by applying the Stefan Frost Number Index from Nelson and

Outcalt (1987) and the Stefan equation (Duan and Naterer, 2009). The Stefan Frost Number determines the difference between the freezing and thawing degree-days accumulations with the inclusion of a soil thermal properties parameter, the ratio between the frozen and unfrozen soil thermal conductivities, to enhance the freezing effect. This enhancement of the freezing effect provides a rationalization for the presence of permafrost at Peawanuck during recent changing climate conditions. The Stefan freezing and thawing depths were determined using the Stefan equation to demonstrate changes in the active layer thicknesses. This chapter provides insight into the climatic potential for permafrost, and for the first time, provides Stefan Frost Number values for Peawanuck, Ontario.

1.7 Chapter 3

Chapter 3 examines the development of the active layers at five sites within a geographical south-to-north study transect from 55°N to 82.5°N within northern Canada from

2004 to 2011. For active layer thickness simulations, this research applied the XG-Algorithm of the Stefan equation that was introduced in Xie and Gough (2013). A validation exercise was conducted for the XG-Algorithm by comparing simulated active layer thickness results at two observation sites. In addition, climate conditions suitable for potential permafrost distributions were assessed and compared with the results from the Frost Number and Stefan Frost Number indices (Nelson and Outcalt, 1987). This chapter details the importance of the negative thermal offset factor that emphasizes the impact of the freezing degree-days by a ratio of the frozen to unfrozen soil thermal conductivities for the presence of permafrost under changing climate.

1.8 Chapter 4

Chapter 4 explores the projected changes in the potential permafrost distributions at the five study sites within the geographical south-to-north study transect under future climate warming scenarios projected by an ensemble of GCMs. This ensemble of climate models produced mean changes to surface air temperatures that were applied to project future warming under IPCC A1B, A2, and B1 emissions scenarios for the entire 21st century. Results from the future climate scenarios were applied to the Stefan Frost Number to assess climatic conditions for potential permafrost distributions for the future periods of 2011 to 2040, 2041 to 2070, and

2071 to 2100.

1.9 Chapter 5

Chapter 5 provides a climate change assessment on the formation and occurrence of palsas at four locations within the HBL, the southernmost extent of the geographical south-to- north study transect. Climatological criteria for palsa formation and occurrence in the HBL were based upon palsa literature from Fennoscandian and neighboring Quebec. Climate projections were produced from two models, the Canadian coupled climate model CGCM2 and British

Hadley Centre model HadCM3, using two IPCC emission scenarios, A2 and B2. For two locations, a statistical downscaling model (SDSM) was available for projections. The SDSM results in this research were funded and produced by Joyce Zhang in 2011, a third party contractor with the Ontario Ministry of Environment; the data is available online from: http://www.utsc.utoronto.ca/~gough/stn_results.htm. The HBL criteria were applied to the future climate scenarios to assess for palsa formation and occurrence for the future periods of the

2020s, 2050s, and 2080s.

1.10 Study sites

The study sites are located along a south-to- north geographical transect within northern

Canada that begins in the Subarctic region (52.2°N) within the Hudson Bay Lowlands at

Lansdowne House, Ontario, and ends at the northern tip of Ellesmere Island within the Canadian

High Arctic (82.5°N) at Canada’s northernmost inhabited place in the world, Alert, Nunavut.

The study sites are identified in Figure 1-10 from south-to-north. For this study, the subarctic region of the Hudson Bay Lowlands of northern Ontario contains the following locations:

Lansdowne House (52o12’27”N, 87o54’06”W); Big Trout Lake (53o45’0”N, 90o0’0"W);

Peawanuck (55°0’30” N; 85°25’20” W); and Fort Severn (55o59’25”N, 87o37’59”W). Within the low arctic region in Nunavut, this study selected the Hamlet of Rankin Inlet (62°48’35” N;

92°05’58” W) which is located on the western shore of Hudson Bay within the Kivalliq Region.

For the high arctic locations in Nunavut, the following locations were selected: the Hamlet of

Resolute Bay, which is located on Cornwallis Island within the Qikiqtaaluk Region (74°41’51”

N; 94°49’56” W); Eureka, which is located on the west coast of Ellesmere Island within the

Qikiqtaaluk Region (79°59’20” N; 85°56’30” W); and Alert, which is located on the northeastern tip of Ellesmere Island adjacent to the Lincoln Sea of the Arctic Ocean within the

Qikiqtaaluk Region (82°30’1” N; 62°20’37” W). All maps in this research were produced in

Manifold 8 Geographical Information System, using Lambert conformal conic projection.

Figure 1-10. Map of all study locations along the south-to-north transect.

1.10.1 Permafrost research at adjacent study sites

The potential permafrost distributions in this thesis were assessed at eight locations along a geographical north-to-south transect from 52.2°N to 82.5°N. Previous permafrost research within the study transect are limited, especially in the subarctic region of northern Ontario. For the Hudson Bay Lowlands at and around Peawanuck in northern Ontario (adjacent to Polar Bear

Provincial Park), recent research is limited to a few studies, such as Gough and Leung (2002),

Tam (2009), and Tam et al., (2014; Chapter 5). The focus of Gough and Leung (2002) was to investigate the climatological conditions in supporting permafrost along the northern Ontario and

Hudson Bay coastal region by calculating and comparing the Frost Number classification for permafrost distributions with direct field observations. Gough and Leung (2002) demonstrated that the Frost Number threshold values indicated discontinuous permafrost distributions (F =

0.61) which was in contrast to field observations indicating continuous permafrost. Gough and

Leung (2002) first proposed an explanation that possible calculation errors in the thawing degree-days affected the Frost number results; they later concluded that these possible errors could not account for the inconsistency. A second hypothesis was proposed that the inconsistent permafrost distribution could be explained by the asymmetric thermal properties of frozen and unfrozen soils, the 'thermal offset' phenomenon between different thermal conductivities which became the primary focus in Tam (2009). This was further investigated with detailed soil properties obtained from field work in a single homogenous soil layer for Chapter 2 and in a multilayered approach in Chapter 3 of this thesis. In focusing on the impacts of climate change on surface features, Chapter 4 assesses the projected change in potential permafrost distributions and Chapter 5 presents a themed study on the fate of palsas, frozen peat mounds, at four locations within subarctic northern Ontario.

As direct research in this subarctic region is limited, other recent studies were reviewed to the west of this study location at Wapusk National Park near Churchill, Manitoba, such as

Kershaw and McCulloch (2007), Dyke and Sladen (2010), Zhang et al. (2012), and Zhang

(2013). Kershaw and McCulloch (2007) provided characterizations of vegetation and snowpack formations in northern Manitoba. These surface layers have profound thermal insulating abilities, the heat loss potentials that can influence ground temperature affecting the underlying permafrost thermal equilibrium. The role of organic carbon within peatlands and the formation of peat plateaus were investigated by Dyke and Sladen (2010). The role of permafrost in the formation of peat plateaus play an important role for the local ecosystem, serving as habitats for wildlife, and the presence of peat further provides an additional thermal insulating layer between the air and ground temperatures (Dyke and Sladen, 2010). Polar bears have been observed using elevated peat plateaus and peat banks as habitat dens, and caribou have been observed utilizing the plateaus for winter foraging; these activities may impact the thermal insulating peat cover

(Dyke and Sladen, 2010). The focus of Dyke and Sladen (2010) was to investigate the sensitivity of peat plateau terrain under climate warming scenarios, providing insight on changes to the permafrost and on the relationship between the air and ground temperatures with surface environmental conditions. Zhang et al. (2012) provided insight on the thickening of the active layer during the 20th century using a process-based model, The Northern Ecosystem Soil

Temperature (NEST) model, to conduct modelling and mapping of the permafrost. Zhang et al.

(2012) further measured thaw depths in 2007 by probing the ground and collected additional data on surface vegetation within northern Manitoba. The modelling results from Zhang et al. (2012) indicated that the active layer thickness will increase by 37% under future climate change scenarios. Zhang (2013) suggested that permafrost will persist at Wapusk National Park,

35 occupying 65-81% of the land area, by the end of the 21st century, in which these results are consistent to the fate of permafrost presented within this work in Chapters 4 and 5.

For palsa and permafrost research to the west and east of the Hudson Bay Lowlands of northern Ontario, studies by Dredge (1992), Vallée and Payette (2007), Thibault and Payette

(2009), and Cyr and Payette (2010) were reviewed. Dredge (1992) provided a detailed review of palsa and permafrost presence in the Hudson Bay Lowlands of northern Manitoba from

Churchill to Gillam. Dredge (1992) highlighted the importance of having high soil thermal conductivity of wet and icy peat to allow greater freezing depths during the winter season, which can also support and augment the underlying permafrost layer. The role of insulating snow cover was also detailed in Dredge (1992). To the east of the Hudson Bay Lowlands of northern

Ontario, some work in northern Quebec palsa and permafrost were reviewed, such as Vallée and

Payette (2007), Thibault and Payette (2009), and Cyr and Payette (2010). The northern Quebec palsas were considered in establishing the climatological criteria for the formation and presence of palsas in the northern Ontario, as presented in Chapter 5. From Cyr and Payette (2010), the important threshold criterion for the palsa presence was observed at a mean annual air temperature of 0oC. The observation of palsas provides surface evidence of the state of the underlying permafrost by serving as a physical indicator of discontinuous and sporadic permafrost distributions (Seppälä, 1986; Dredge, 1992;Vallée and Payette, 2007; Kujala et al.,

2008; Kirpotin et al., 2009).

For the arctic region, some recent permafrost investigations have been conducted between the 55°N to 82.5°N latitudes, such as Lewkowicz (2007), Cannone et al. (2010),

Romanovsky et al. (2010), Smith et al. (2010), Rinke et al. (2012), and Throop et al. (2012) at

36 similar study locations of Baker Lake (near Rankin Inlet), Resolute Bay, Eureka, and Alert in

Nunavut. The research by Lewkowicz (2007) and Cannone et al. (2010) focused on the impacts of active layer detachments on vegetation near Eureka, Nunavut. These landslides can be triggered by a surge of soil water from rainfall events, rapid snow melt, and rapid melting of the active layer or the top of the permafrost that increases the pore water pressures and decreases strength and friction between the active layer and permafrost base, which permits mass movement over low angled slopes (Lewkowicz, 2007; Cannone et al., 2010). The research from

Lewkowicz (2007) and Cannone et al. (2010) provided ground characteristics and insight on active layer thicknesses that can be measured from newly exposed ground failure fronts, near

Eureka, Nunavut. Romanovsky et al. (2010) and Smith et al. (2010) provided a summary of the

North American contributions from the International Polar Year, primarily the network of soil temperature boreholes within the Permafrost Observatory Network to improve knowledge of the permafrost thermal state. This network of boreholes overlap with three study locations from this work within the 56° to 82.5°N latitudes at Baker Lake (near Rankin Inlet), Resolute Bay, and

Alert; however, the mean annual ground temperature records for these locations could not be applied in this thesis as the data were not continuous and did not span the entire time period to complement the available 50-year air temperature data records, as required in this research.

Throop et al. (2012) conducted thermal monitoring at 10 locations within northern Canada, from the 60° to 83°N latitudes, and with 9 locations located within the 60° to 70°N latitudes.

Overlapping locations from this study were reviewed for Baker Lake (near Rankin Inlet) and

Alert for soil characteristics and physical properties, primarily the soil moisture contents and soil thermal conductivities (Throop et al., 2012). Some results from Throop et al. (2012) indicated that mean annual air temperature is the primary determinant of permafrost temperatures that provides insight on thermal state of the permafrost; and the important roles of soil moisture

37 content and the latent heat effects on the thawing and freezing process of different ground materials. Romanovsky et al. (2010) and Smith et al. (2010) highlighted the importance and need for additional global permafrost monitoring to generate long-term permafrost temperature data series in order to determine any changes to the permafrost distributions and thermal states. The studies further illustrated the requirement for permafrost studies within Canada to address the limited number of monitoring stations, including weather stations, within the polar network. The contributions from this thesis provides an attempt to address some of the data gap issues in permafrost research, specifically along this study transect, where historical and continuous active layer and permafrost measurements are either limited or non-existent.

1.11 References

Akinyemi, OD., Sauer, TJ., and Onifade, YS. 2011. Revisiting the block method for evaluating thermal conductivities of clay and granite. International Communications in Heat and Mass Transfers 38, 1014-1018.

Anisimov, O., Shiklomanov, N., and Nelson, F. 1997. Global warming and active-layer thickness: results from transient general circulation models. Global and Planetary Change 15, 61-77.

Berthling, I., and Etzelmüller, B. 2011. The concept of cryo-conditioning in landscape evolution. Quaternary Research 75, 378-384.

Bockheim, J., Mazhitova, G., Kimble, J., and Tarnocai, C. 2006. Controversies on the genesis and classification of permafrost-affected soils. Geoderma 137, 33-39.

Broadridge, P. and Pincombe, B. 1996. The Stefan Solidification Problem with Nonmonotonic Nonlinear Heat Diffusivity. Mathl. Comput. Modelling 23, 87-98.

Brown, R. 1973. Influence of climatic and terrain factors on ground temperatures at three locations in the permafrost region of Canada. Ottawa: National Research Council of Canada. NRC 13574 pp. 35-68.

Cannone, C., Lewkowicz, AG., and Guglielmin, M. 2010. Vegetation colonization of permafrost-related landslides, Ellesmere Island, Canadian High Arctic. Journal of Geophysical Research 115, G04020, doi:10.1029/2010JG001384.

Crepeau, J. 2007. Josef Stefan: His life and legacy in the thermal science. Experimental Thermal and Fluid Science 31, 795-803.

Cyr, S., and Payette, S. 2010. The origin and structure of wooded permafrost mounds at the arctic treeline in eastern Canada. Plant Ecology & Diversity 3, 35-46.

Derksen, C., Smith, SL., Sharp, M., Brown, L., Howell, S., Copland, L., Mueller, DR., Gauthier, Y., Fletcher, CG., Tivy, A., Bernier, M., Bourgeois, J., Brown, R., Burn, CR., Duguay, C., Kushner, P., Langlois, A., Lewkowicz, AG., Royer, A., and Walker, A. 2012. Variability and change in the Canadian cryosphere. Climatic Change 115, 59-88.

Dobinski, W. 2011. Permafrost. Earth-Science Reviews 108, 158-169.

Dredge, LA., 1992. Field guide to the Churchill Region, Manitoba: glaciations, sea level changes, permafrost landforms, and archeology of the Churchill and Gillam areas. Geological Survey of Canada: Ottawa, doi:10.4095/134064

Duan, X., and Naterer, G. 2009. Heat conduction with season freezing and thawing in an active layer near a tower foundation. International Journal of Heat and Mass Transfer 52, 2068-2078.

Dyke, LD., and Sladen, WE. 2010. Permafrost and peatland evolution in the northern Hudson Bay Lowlands, Manitoba. Arctic 63, 429-441.

Etzelmüller, B., Schuler, TV., Isaksen, K., Christiansen, HH., Farbrot, H., and Benestad, R. 2011. Modeling the temperature evolution of Svalbard permafrost during the 20th and 21st century. The Cryosphere 5, 67-79.

French, H. 2007. The Periglacial Environment. (3rd Ed.). John Wiley & Sons: West Sussex

French, H., and Shur, Y. 2010. The principles of cryostratigraphy. Earth-Science Reviews 101, 190-206.

Fronzek, S., Luoto, M., and Carter, TR. 2006. Potential effect of climate change on the distribution of palsa mires in Subarctic Fennoscandia. Climate Research 32(1), 1-12.

Gough, WA., and Leung, A. 2002. Nature and fate of Hudson Bay permafrost. Regional Environmental Change 2, 177-184.

Gurney, SD. 2001. Aspects of the genesis, geomorphology and terminology of palsas: perennial cryogenic mounds. Progress in Physical Geography 25, 249-260.

Harris, SA. 1986. The Permafrost Environment. (1st Ed.). Barnes & Noble Books: Totowa, New Jersey.

Hayashi, M., Goeller, N., Quinton, W., and Wright, N. 2007. A simple heat-conduction method for simulating the frost-table depth n hydrological model. Hydrological Processes 21, 2610-2622.

Hinkel, K., Paetzold, F., Nelson, F., and Bockheim, J. 2001. Patterns of soil temperature and moisture in the active layer and upper permafrost at Barrow, Alaska: 1993–1999. Global and Planetary Change 29, 293–309.

Hughes, P., and Braithwaite, R. 2008. Application of a degree-day model to reconstruct Pleistocene glacial climates. Quaternary Research 69, 110-116.

Juliussen, H., and Humlum, O. 2007. Towards a TTOP Ground Temperature Model for Mountainous Terrain in Central-Eastern Norway. Permafrost and Periglacial Processes 18, 161-184.

Juma, NG. 2006. The Pedosphere and its Dynamics: A Systems Approach to Soil Science. Salman Production: Edmonton.

Jumikis, AR. 1977. Thermal Geotechnics. Rutgers Univ. Press: New Brunswick, NJ.

Kershaw, GP., and Gill D. 1979. Growth and decay of palsas and peat plateaus in the Macmillan Pass – Tsichu River area, Northwest Territories, Canada. Canadian Journal of Earth Science 16, 1362-1374.

Kershaw, GP., and McCulloch, J. 2007. Midwinter snowpack variation across the arctic treeline, Churchill, Manitoba, Canada. Arctic, Antarctic, and Alpine Research 39, 9-15.

Kirpotin, SN., Polishchuk, Y., and Bryksina, N. 2009. Abrupt changes of thermokarst lakes in Western Siberia: impacts of Climatic warming on permafrost melting. International Journal of Environmental Studies 66(4), 423-431.

Kuhry, P. 2008. Palsa and peat plateau development in the Hudson Bay Lowlands, Canada: timing, pathways and causes. Boreas 37, 316-327.

Kujala, K., Seppälä, M., and Holappa, T. 2008. Physical properties of peat and palsa formation. Cold Regions Science and Technology 52, 408-414.

Lewis, T., and Lamoureux, SF. 2010. Twenty-first century discharge and sediment yield predictions in a small high Arctic watershed. Global and Planetary Change 71, 27-41.

Lewkowicz, AG. 2007. Dynamics of active‐layer detachment failures, Fosheim Peninsula, Ellesmere Island, Nunavut, Canada. Permafrost Periglacial Process 18, 89–103.

Lewkowicz, AG., and Coultish, TL. 2004. Beaver Damming and Palsa Dynamics in a Subarctic Mountainous Environment, Wolf Creek, Yukon Territory, Canada. Arctic, Antarctic, and Alpine Research 36(2), 208-218.

Lunardini, VJ. 1981. Heat Transfer in Cold Climates. Van Nostrand Reinhold Publishers: New York.

McCarthy, K., Walker, L., and Vigoren, L. 2004. Subsurface fate of spilled petroleum hydrocarbons in continuous permafrost. Cold Regions Science and Technology 38, 43-54.

McClymont, AF., Hayashi, M., Bentley, LR., Christensen, BS. 2013. Geophysical imaging and thermal modeling of subsurface morphology and thaw evolution of discontinuous permafrost. Journal of Geophysical Research: Earth Surface 118, 1826-1837.

Meurth, M., and Mauser, W. 2012. Rigorous evaluation of a soil heat transfer model for mesoscale climate change impact studies. Environmental Modelling and Software, doi:10.1016/j.envsoft.2012.02.017

Muller, S. 2008. Frozen in Time: Permafrost and Engineering Problems, HM French and FE Nelson (eds). American Society of Civil Engineers: Reston, VA.

Nelson, F. 1986. Permafrost Distributions in Central Canada: Application of a Climate-Based Predictive Model. Annals of the Association of American Geographers 76, 550-569.

Nelson, F., and Outcalt, S. 1987. A Computational Method for Prediction and Regionalization of Permafrost. Arctic and Alpine Research 19, 279-288.

Nelson, F., Shiklomanov, N., Mueller, G., Hinkel, K., Walker, D., and Bockheim, J. 1997. Estimating active layer thickness over a large region: Kuparuk River Basin, Alaska, USA. Arctic and Alpine Research 29, 367-378.

Nixon, J., and McRoberts, E. 1973. A study of factors affecting the thawing of frozen soils. Can. Geotech. J. 20, 439-452.

Parviainen, M., and Luoto, M.. 2007. Climate envelopes of mire complex types in Fennoscandia. Geografiska Annaler, Series A: Physical Geography, 89(2), 137-151.

Pengerud, A., Cécillon, L., Johnsen, LK., Rasse, DP., and Strand, LT. 2013. Permafrost Distribution Drives Soil Organic Matter Stability in a Subarctic Palsa Peatland. Ecosystems 16, 934-947.

Pukonen, J. 1998. Soil thermal properties and heat transfer processes near Ny-Alesund, northwestern Spitsbergen, Svalbard. Polar Research 17, 165-179.

Rinke, A., Matthes, H., Christensen, JH., Kuhry, P., Romanovsky, VE., and Dethloff, K. 2012. Arctic RCM simulations of temperature and precipitation derived indices relevant to future frozen ground conditions. Global and Planetary Change 80-81, 136-148.

Romanovsky, VE., Smith, SL., and Christiansen, HH. 2010. Permafrost thermal state in the polar Northern Hemisphere during the International Polar Year 2007–2009: a synthesis. Permafrost Periglacial Process 21, 106–116.

Saemundsson, T., Arnalds, Kneisel, C., Jonsson, HP., and Decaulne, A. 2012. The Orravatnsrustir palsa site in Central Iceland – Palsas in an Aeolian sedimentation environment. Geomorphology 167-168, 13-20.

Sazonova, TS., Romanovsky, VE., Walsh, JE., and Sergueev, DO. 2004. Permafrost dynamics in the 20th and 21st centuries along the East Siberian transect. Journal of Geophysical Research 109, D01108, DOI: 10.1029/2003JD003680

Seppälä, M., 1986: The Origin of Palsas. Physical Geography 68, 141-147.

Shur, Y., and Jorgenson, M. 2007. Patterns of Permafrost Formation and Degradation in Relation to Climate and Ecosystems. Permafrost and Periglacial Process 18, 7–19.

Slater, AG., and Lawrence, DM. 2013. Diagnosing Present and Future Permafrost from Climate Models. Journal of Climate 26, 5608-5623.

Smith, MW., and Riseborough, DW. 1996. Permafrost monitoring and detection of climate change. Permafrost and Periglacial Processes 7, 301-310.

Smith, SL., and Burgess, MM. 2002. A Digital Database of Permafrost Thickness in Canada. Geological Survey of Canada, Open File 4173, P79.

Smith, SL., and Riseborough, DW. 2010. Modelling the thermal response of permafrost terrain to right-of-way disturbance and climate warming. Cold Regions Science and Technology 60, 92-103.

Smith, SL., Burgess, MM., Riseborough, DW., and Nixon, FM. 2005. Recent Trends from Canadian Permafrost Thermal Monitoring Network Sites. Permafrost and Periglacial Process 16, 19-30.

Smith, SL., Romanovsky, VE., Lewkowicz, AG., Burn, CR., Allard, M., Clow, GD., Yoshikawa, K., and Throop, J. 2010. Thermal state of permafrost in North America - A contribution to the International Polar Year. Permafrost Periglacial Process 21, 117–135.

Solomon, S., Qin, D., Manning, M., Alley, RB., Berntsen, T., Bindoff, NL., Chen, Z., Chidthaisong, A., Gregory, JM., Hegerl, G.C., Heimann, M., Hewitson, B., Hoskins, BJ., Joos, F., Jouzel, J., Kattsov, V., Lohmann, U., Matsuno,T., Molina, M., Nicholls, N., Overpeck, J., Raga, G., Ramaswamy, V., Ren, J., Rusticucci, M., Somerville, R., Stocker, TF., Whetton, P., Wood RA., and Wratt, D. 2007. Technical Summary. In: Solomon, S., Qin, D., Manning, M., Chen, Z., Marquis, M., Averyt, KB., Tignor, M., and Miller, HL. (eds.). Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA.

Tam, A. 2009. Master’s Thesis: “Permafrost in Canada’s Subarctic Region of Northern Ontario”. University of Toronto, Department of Geography.

Tam, A., Gough, WA., and Xie, C. 2014b. The fate of Hudson Bay Lowlands palsas in a changing climate. Arctic, Antarctic, and Alpine Research (accepted)

TGICA. 2007. “General Guidelines on the use of Scenario Data for Climate Change Adaptation Assessment, Version 2.” Intergovernmental Panel on Climate Change. Accessed Online: May 1st, 2013 from: http://www.ipcc-data.org/guidelines

Thibault, S., and Payette, S. 2009. Recent permafrost degradation in bogs of the James Bay area, Northern Quebec. Permafrost and Periglacial Processes 20, 383-389.

Throop, J., Lewkowicz, AG., and Smith, SL. 2012. Climate and ground temperature relations at sites across the continuous and discontinuous permafrost zones, northern Canada. Can. J. Earth Sci. 49, 865-876.

Tsuyuzaki, S., Sawada, Y., Kushida, K., and Fukuda, M. 2008. A preliminary report on the vegetation zonation of palsas in the Arctic National Wildlife Refuge, northern Alaska, USA. Eco Res 23, 787-793.

UTSC. 2013. Localizer. University of Toronto Scarborough. Accessed Online: May 22nd, 2013 from: http://climatechange.utsc.utoronto.ca/

Vallée, S., and Payette, S. 2007. Collapse of permafrost mounds along a subarctic river over the last 100 Years (Northern Quebec). Geomorphology 90, 162-170.

Wen, Z., Sheng, Y., Ma, W., Wu, Q., Wang, D., Sun, Z., and Li, G. 2012. Probabilistic analysis of the replacement thickness of the in-cuts roadbed in permafrost regions. Cold Regions Science and Technology 64, 57-67.

Woo, MK., Arain, MA., Mollinga, M., and Yi, S. 2004. A two-directional freeze and thaw algorithm for hydrologic and land surface modeling. Geophysical Research Letters 31, DOI: 10.1029/2004GL019475

Xie, C., and Gough, WA. 2013. A simple thaw-freeze algorithm for multilayered soils using the Stefan equation. Permafrost and Periglacial Processes, DOI: 10.1002/ppp.1770

Zhang, JYJ. 2011. Ontario Climate Change SDSM Downscaling Results. Accessed Online: September 1st, 2012 from: http://www.utsc.utoronto.ca/~gough/stn_results.htm

Zhang, Y. 2013. Spatio-temporal feature of permafrost thaw projected from long-term high- resolution modeling for a region in the Hudson Bay Lowlands in Canada. Journal of Geophysical Research: Earth Surface 118, 542-522.

Zhang, Y., Chen, W., and Riseborough, DW. 2008. Transient projections of permafrost distributions in Canada the 21st century under scenarios of climate change. Global and Planetary Change 60, 443-456.

Zhang, Y., Li, J., Wang, X., Chen, W., Sladen, W., Dyke, L., Dredge, L., Poitevin, J., McLennan, D., Stewart, H., Kowalchuk, S., Wu, W., Kershaw, GP., and Brook, RK. 2012. Modelling and mapping permafrost at high spatial resolution in Wapusk National Park, Hudson Bay Lowlands. Can. J. Earth Sci. 49, 925-937.

Zhou, F., Zhang, A., Li, R., and Hoeve, E. 2009. Spatio-temporal simulation of permafrost geothermal response to climate change scenarios in a building environment. Cold Regions Science and Technology 56, 141-151.

Zuidhoff, FS., and Kolstrup, E. 2000. Changes in Palsa Distribution in Relation to Climate Change in Laivadalen, Northern Sweden, Especially 1960-1997. Permafrost Periglacial Processes 11, 55-69.

Chapter 2 An assessment of potential permafrost at Peawanuck, Ontario, from 1959-2011 2 Chapter 2

2.1 Abstract

Based on previous Frost number calculations and field investigations in and around

Peawanuck, Ontario, inconsistencies in the permafrost states were identified. This study applied the Stefan Frost (Fs) Number Index that combines both climatological parameters and soil thermal properties to assess the climatic potential for permafrost presence. The Stefan equation was applied to simulate the freezing and thawing depths of the active layer to provide insight on the state of permafrost based on available climate records spanning 1959 to 2011. An analysis on the freezing degree-days accumulation revealed a statistically significant decreasing trend at this location since 1959. Comparisons using various soil thermal properties were conducted for the

Fs and Stefan depth simulations to rationalize the existence of permafrost even during the recent period of recent changing climate conditions. Both Fs and the Stefan depths demonstrated favourable potential for permafrost at Peawanuck; however, with climate warming changes being observed in this region of the Hudson Bay Lowlands of northern Ontario, future changes to the potential permafrost are likely.

Keywords: Permafrost; Soil thermal conductivity; Active layer; Stefan depths; Stefan Frost

Number; Hudson Bay Lowlands.

2.2 Introduction

Permafrost is defined as ground material that remains below 0ºC for at least two consecutive years (Gough and Leung, 2002; French, 2007; Shur and Jorgenson, 2007). The active layer is located above the permafrost, and the variation in thicknesses is driven by seasonal freezing and thawing cycles (French, 2007). Changes in the climatological and thermal factors of the soils can affect the seasonal freezing and thawing cycles causing impacts on the underlying permafrost layer (Burn and Smith, 1988; Bockheim et al., 2006; French, 2007).

Continuous permafrost underlies the northern region of the Hudson Bay Lowlands (HBL), located in Northern Ontario, Canada (Natural Resources Canada, 2006). The following research questions are proposed: “Can the potential for permafrost be assessed using available permafrost tools at Peawanuck, Ontario, in the Subarctic region of the Hudson Bay Lowlands? If so, can the presence of permafrost at this location be rationalized by the asymmetric frozen and unfrozen soil thermal conductivities?”

This work applies climatological and soil thermal properties observed at Peawanuck,

Ontario, to assess the potential for permafrost through the Stefan Frost (Fs) Number index and to demonstrate active layer thickness calculations through the application of various soil thermal conductivities in the Stefan equation. The Stefan equation is capable of determining the soil freezing and thawing depths based on the soil thermal conductivity and degree-days accumulations (Duan and Naterer, 2009, Pang et al., 2009). Nelson (1986) successfully employed the Fs for various locations in Manitoba, Saskatchewan, Alberta, Northwest

Territories, and Nunavut (formerly part of the Northwest Territories). This approach can provide estimates of the active layer thickness for Peawanuck, Ontario, which may provide additional

46 insight on impacts to the underlying permafrost layer (Muller, 2008; Zhang et al., 2008; Pang et al., 2009). Fs provides a dimensionless ratio of the Stefan equation, mainly between the freezing and thawing degree-days (FDD, TDD) and the asymmetric soil thermal properties based on a ratio between the frozen and unfrozen soil thermal conductivities, Kf and Ku respectively (Nelson and Outcalt, 1987). As permafrost presence is associated with the accumulation of FDD for a region, it is important to include site-specific soil thermal properties and soil moisture content.

The Kf/Ku ratio further reflects the presence of water occupying the pore fraction of soils as represented by the soil moisture content, which is an important factor influencing the soil thermal conductivities. An increase in soil moisture content can increase the soil thermal conductivity (Nixon and McRoberts, 1973). This can be beneficial to permafrost during cold winters where high Kf values promote deeper freezing depths or during dry summers by reducing the Ku and the extent of the thawing depths. As shown in Nixon and McRoberts (1973), a possible range of Kf/Ku is 1.0 to 3.7, in which Kf is typically greater than Ku based on the thermal properties of ice having greater thermal conductivity than unfrozen water. For greater example, soils with a greater frozen to unfrozen soil thermal conductivity ratio will have an amplifying effect on the FDD accumulation, thus supporting permafrost presence. Conversely, in regions where the soil thermal conductivity ratio is less, the amplifying effect on FDD is muted. If the combined soil thermal conductivity ratio and FDD accumulations becomes less than the TDD accumulations, this condition may lead to a net thawing and thus a reduction of permafrost areas.

Historical permafrost measurements in this region are limited; however, climate data from 1959 are available for Winisk, Ontario, and after 1986 for Peawanuck, Ontario (Tam,

2009). Gough and Leung (2002) previously researched HBL permafrost by calculating the Frost

Number threshold classification for permafrost distributions in comparison to field observations

47 of the permafrost state. The Frost Number (F) was introduced by Nelson and Outcalt (1987) as a nuanced ratio of FDD and TDD:

FDD0.5 F  ( ) . (2-1) TDD0.5  FDD0.5

Using the Frost Number, continuous permafrost distributions can be classified with F ≥ 0.67 threshold value; discontinuous permafrost distributions with F < 0.67; sporadic permafrost at F <

0.6; and for no permafrost, F < 0.5. Gough and Leung (2002) demonstrated that the Frost

Number threshold values indicated discontinuous permafrost distributions (F = 0.61) in the HBL at Peawanuck, Ontario; however, this was contrary to field observations indicating continuous permafrost. Gough and Leung (2002) first proposed an explanation that possible calculation errors in the thawing degree-days affected the Frost number results. They concluded that these possible errors could not account for the inconsistency. They then proposed a second hypothesis that the inconsistent permafrost distribution could be explained by the asymmetric thermal properties of frozen and unfrozen soils, the 'thermal offset' phenomenon between different thermal conductivities. In general, the soil thermal conductivity is higher for frozen soils resulting in an asymmetric difference in thermal properties between frozen and unfrozen soils due to the presence of ice (Burn and Smith, 1988; French, 2007). The physical properties and type of the ground material affects the susceptibility to freezing and thawing (French, 2007).

2.3 Methodology

2.3.1 Study Area and Geography

The study area is located at Peawanuck, Ontario, (55°0'30" N; 85°25'20"W) in the

Hudson Bay Lowlands of Northern Ontario (Figure 2-1). Fens and bogs were observed in the surrounding area along the Winisk River. The terrestrial ecozone for the study area is the Hudson

Plains, located north of the Boreal Plains and south of the Southern Arctic ecozones (Natural

Resources Canada, 2007).

Figure 2-1. Location map containing the study area and Peawanuck, Ontario.

2.3.2 Soil Characterization

Soil characterization from field samples collected at Peawanuck, Ontario, revealed high soil moisture content ranging from 20% to 60% and high organic soil matter content (Tam,

2009). Characterization of the soil samples revealed the presence of soil organic matter,

Sphagnum, and partially decomposed plant materials as the major composition of the samples.

Gravimetric soil moisture content was higher (~60%) for samples containing abundant organic materials, clayey, and silty soil composition located inland within the terrestrial HBL (Tam,

2009). Table 2-1 provides additional site conditions based on soil samples collected along the shore of the Hudson Bay, at Winisk and Peawanuck, Ontario. The surface conditions were primarily organic, with thicker peat layers further inland away from the Hudson Bay shore

(Figure 2-2). Near the shores, the soil samples consisted of sandy and gravelly soils with trace soil organic content with low gravimetric soil moisture content (~20%; Tam, 2009). This can be attributed to the poor water retaining ability of sand and gravel with the lack of abundant organic material and clay minerals in the sandy soils (Eyles and Miall, 2007).

Table 2-1. Description of sample locations at Winisk and Peawanuck, Ontario. Sample Locations Latitude Longitude Observations Peawanuck, On. Peat surface layer (0.1-0.3 m). Subsurface soils were highly organic with silty and clayey 55°15.405 -85°12.394 textures. Winisk, On. Grassy surface with presence of sand and gravel (0.1-0.2 m). Subsurface composed of highly 55°15.623 -85°12.639 organic soils with clayey textures.

Figure 2-2. Landscape of the Hudson Bay Lowlands near Peawanuck, Ontario. Photo taken by W.A. Gough.

2.3.3 Climate Data

Climate data were acquired for the HBL at the communities of Winisk and Peawanuck

(Table 2-2) from Environment Canada’s National Climate Data and Information Archive

(Environment Canada, 2012). Prior to 1986, daily temperature data were obtained from the weather station in Winisk, Ontario; after the destructive 1986 flood event, the weather station was relocated to Peawanuck (Figure 2-1). For the Peawanuck climate data, incomplete or missing observed daily temperature data were omitted between 1978 and 1994; the record could not be reconstructed from the neighboring climate data records at Fort Severn, Ontario, as the

52 data for this station is limited to 2006 onwards. The range of observed daily minimum and maximum temperatures (Tmin, Tmax) for the summer month of July and winter month of

January are shown in Table 2 for the specified periods where data are available for Peawanuck and Winisk, Ontario. Using the Peawanuck climate data, the TDD and FDD were calculated.

Statistical analysis was conducted using MINITAB 14 Software; this included an assessment for normal distribution and serial autocorrelation prior to applying linear regression (R) for the calculated TDD and FDD. Normal distribution was determined using a probability density plot set at 95% confidence interval. Significance for autocorrelation was set at a 5% threshold level using eight lags. The numbers of lags were determined automatically using MINITAB default settings based on the number of data entries divided by four. For TDD, the first lag (of eight lags) was significant with the remainder being not significant; for FDD, the first two lags were significant out of eight lags.

Table 2-2. Climate Data from Environment Canada for Peawanuck and Winisk, Ontario. Location Peawanuck, ON Winisk, ON Relief Elevation (m) 52.7 9.2 Annual Tmean (ºC) -3.1 -2.5 Daily Tmin January (ºC) -29 to -25 -35 to -25 Daily Tmax January (ºC) -24 to -20 -24 to -20 Daily Tmin July (ºC) 6 to 10 4 to 8 Daily Tmax July (ºC) 16 to 20 15 to 22 Years of Data Availability 1995 to 2011 1959 to 1964; 1969 to 1977

2.3.4 The Stefan Equation

The Stefan equation is capable of predicting the depths (X) of freezing and thawing depths in an assumed homogenous soil layer (Duan and Naterer, 2009). The Stefan equation is presented as:

1 0.5 1 0.5 X  [(2K  DD)QL ]  [(2K  DD)(L  ) ] , (2-2)

where K represents the soil thermal conductivity; DD is the Degree-Day Index; and QL is the volumetric latent heat of soil composed of the latent heat of fusion of ice, L; the moisture content, ω; and the soil bulk density, ρ. The input parameters based on the soil properties are listed in Table 2-3 (Tam, 2009). The Kf/Ku ratios for peat, silt and clay soils in Table 2-3 were calculated based on observed soil thermal conductivities (Tam, 2009). Three additional theoretical scenarios were developed to capture the range of the Kf/Ku ratios for the three observed soil types and to consider the thermal offset effects, including “no effect” using scenario 1.

Table 2-3. Soil properties and theoretical scenarios of Kf/Ku ratios for Peawanuck, Ontario.

Soil Type Soil Thermal Soil Thermal Kf /Ku Ratio Soil Gravimetric Conductivity Conductivity Bulk Soil (frozen, Kf ) (unfrozen Ku) Density Moisture (ρ) Content (ω) W·(m°C)-1 W·(m°C)-1 dimensionless kg.m-3 % Peat (Organic) 1.38 0.79 1.75 680 20-60 Silt 1.38 1.11 1.24 1450 15-40 Clay 2.08 1.73 1.20 1300 15-50 Scenario 1 - - 1.00 - - Scenario 2 - - 1.50 - - Scenario 3 - - 2.00 - -

2.3.5 The Stefan Frost Number

The Stefan Frost Number (Fs) is similar to the F formulation in equation 2-1 but includes the Stefan equation, which contains the soil properties and the degree-day variables (Nelson and

Outcalt, 1987). Fs can be summarized by the following dimensionless ratio of the Stefan depths

(X) of thawing (t) and freezing (f):

(2-3)

Furthermore, the Fs formulation can be simplified to solely consider the degree-days and the freezing and thawing soil thermal conductivity ratio values:

FDD  K / K  Fs  ( f u ) . (2-4) FDD  K f / Ku  TDD

The Fs formulation shows how the FDD can be influenced by the ratio of the frozen and unfrozen soil thermal conductivities (Nelson et al., 1997; Woo et al., 2004). The Fs threshold values for permafrost distributions are the same as the F thresholds.

2.4 Results

2.4.1 Stefan Frost Numbers for 1959 to 2011

Temporal trends in Fs based on three soil types were conducted for the periods of 1959-

1977 (Figure 2-3a) and of 1995-2011 (Figure 2-3b). The three soil types are peat, silt, and clay with corresponding observed soil thermal conductivity values listed in Table 2-3. The minimum and maximum Fs results for 1959-1977 were observed in 1977 and 1964, respectively, for peat at

0.63 to 0.7; for silt at 0.59 to 0.69; and, for clay at 0.58 to 0.68. For the period of 1995-2011, the minimum and maximum Fs results were observed in 1999 and 2004, respectively, for peat at

0.59 to 0.69; for silt at 0.55 to 0.65; and, for clay at 0.54 to 0.65.

Figure 2-3. Stefan Frost Number Index results for Peawanuck peat, silt, and clay for a) 1959-

1977; and, b) 1995-2011; showing the 0.67 threshold for continuous and 0.50 for unfavourable permafrost distributions.

Three additional scenarios were conducted for the periods of 1959-1977 (Figure 2-4a) and 1995-2011 (Figure 2-4b) to demonstrate the enhanced freezing effects, the negative thermal offset phenomenon, that occurs from the Kf/Ku ratio scenarios affecting FDD in Fs. The three scenarios provide insight on the impacts of thermal offset on Fs result. For comparison, no thermal offset is represented in the first scenario by a Kf/Ku ratio = 1.0, where Fs = F. The second scenario of a Kf/Ku ratio = 1.5 represents conditions close to the Peawanuck field values for peat, silt, and clay that were based on observations listed in Table 2-3. The final scenario of a Kf/Ku ratio = 2.0 represents a strong thermal offset effect on the soil thermal conductivities. Fs = F results under the first scenario for the period from 1959-1977 ranged from 0.56 to 0.66 indicating conditions for discontinuous and sporadic permafrost. Under the second scenario, Fs ranged from

0.61 to 0.71, indicating conditions for continuous and discontinuous permafrost. For the third scenario, Fs ranged from 0.64 to 0.74, indicating conditions for continuous and discontinuous permafrost. For the period from 1995-2011, the Fs results under the first scenario ranged from

0.52 to 0.63, indicating conditions for discontinuous and sporadic permafrost. The second scenario ranged from 0.57 to 0.67, indicating conditions for continuous to sporadic permafrost.

For the third scenario, Fs ranged from 0.61 to 0.70, indicating conditions for continuous and discontinuous permafrost.

Figure 2-4. Stefan Frost Number Index results for three thermal conductivity scenarios with

Kf/Ku ratios at 1.0, 1.5 and 2.0 for a) 1959-1977 and b) 1995-2011; showing the 0.67 threshold for continuous and 0.50 for unfavourable permafrost distributions.

Climate trend analysis from 1959 to 2011 are shown in Table 2-4, three segmented periods from

1959-1964, 1969-1977, and 1995-2011 are shown based on the available data. Gaps between the periods were not included due to missing and unavailable data. Both TDD and FDD showed weak to moderate positive correlations from 1959 to 1977; however, there was no statistical significance. For the period from 1995 to 2011, both TDD and FDD were observed to have a moderate negative correlation with statistical significance set at p-values < 0.05 (Table 2-4). The mean TDD had increased from 1286.4 (1959-1964) to 1689.1 (1995-2011); and the mean FDD had decreased from 4332.4 (1959-1964) to 3729.0 (1995-2011). For the span of data availability from 1959-1964, there were no significant change in TDD; however, FDD was observed to have an overall moderate negative correlation with statistical significance set at p-values < 0.05.

Table 2-4. Trends in degree-days for Peawanuck, Ontario, from 1959 to 2011. Year 1959-1964 1969-1977 1995-2011 1959-2011 Mean TDD 1286.4 2068.0 1630.7 1689.1 Standard deviation 94.6 190.2 191.9 323.6 TDD TDD Corr. (R) 0.36 0.35 0.03 -0.003 TDD p-value 0.48 0.36 0.91 0.99 Mean FDD 4332.4 4215.9 3258.0 3729.0 Standard deviation 227.4 437.0 619.0 719.0 FDD FDD Corr. (R) 0.77 0.04 -0.79 -0.78 FDD p-value 0.07 0.91 0.00* 0.00* *significance at p-value < 0.05.

2.4.2 Stefan Depths from 1959 to 2011

Figure 2-5. Stefan thawing (Xu) and freezing (Xf) depths are shown for a) Peat, b) Silt, and c)

Clay soil compositions at Peawanuck, Ontario, from 1959 to 2011 with ΔX (Xf –Xu).

Over the 50-year period shown in Figure 5, the Stefan freezing depths have shown a significant decreasing trend, R = -0.78, for the soil types of peat (Figure 2-5a); silt (Figure 2-5b); and, clay (Figure 2-5c) with p-value = 0.00. The Stefan thawing depths have slightly increased

(R = 0.025; p-value 0.89) since 1959 for all soil types. The difference in Stefan depths for freezing, Xf, and thawing, Xu, layers were calculated to determine the permafrost state as ΔX

(Figure 2-5). A decreasing trend in the difference between the Stefan freezing and thawing depths are shown with negative correlations of -0.709 for peat, p-value <0.05; silt, R = -0.685, p- value <0.05; and, clay, R = -0.683, p-value <0.05.

2.5 Discussion

2.5.1 Stefan Frost Number Index for Peawanuck

The presence of permafrost depends on the climatological and soil thermal parameters

(French, 2007; Shur and Jorgenson, 2007). Continuous and discontinuous permafrost distributions can be easily determined using climatological parameters, such as the F calculation as shown in Nelson and Outcalt (1987); however, as shown in Gough and Leung (2002), discrepancies were identified between F results with field observations for the HBL. This study applied another permafrost tool, the Fs, as shown in Nelson (1986) and Nelson and Outcalt

(1987) to assess potential permafrost based on climate and soil thermal properties at Peawanuck,

Ontario, for the first time. In Nelson (1986), the Fs were calculated for other regions of northern

Canada. Similar to the F, the Fs applies both FDD and TDD but with the addition of the frozen and unfrozen soil thermal conductivity ratio to account for the asymmetric soil freezing process driven by the thermal offsetting phenomenon. In general, frozen soils have greater thermal

62 conductivities by a ratio of 1.5 at 20-25% soil moisture content than unfrozen soils (Nixon and

McRoberts, 1973; Kujala et al., 2008; Kokelj et al., 2010). For Peawanuck, field conditions had thermal conductivity ratios of 1.20 (Clay), 1.24 (Silt), and 1.75 (Peat) with corresponding gravimetric soil moisture contents of 37, 27, and 70%, respectively (Table 3-3). Greater F and Fs values are expected at areas typically dominated by the FDD accumulations, this provides favourable permafrost potential. As the Fs thresholds for potential permafrost are the same as F,

Fs > 0.5 indicate conditions favourable for permafrost presence, albeit at the sporadic distribution. Unfavourable potential for permafrost is typical where the FDD accumulations and soil thermal conductivity ratio have been reduced or where TDD accumulations have increased; the F and Fs values would be < 0.5. The Fs calculations in this study demonstrated both the enhancing or reducing effects on the FDD accumulations due to variations in the frozen and unfrozen soil thermal conductivity ratio, the dependency of the ratio is based on the site-specific soil type and on the state (frozen and unfrozen) that can be also be modified by the soil moisture content of the study location. For Peawanuck, soil samples identified the ground materials to be peat, silt, and clay compositions.

The overall temporal trend using the three soil types for Fs at Peawanuck (1959-2011) showed continuing support for permafrost presence; however, based on the individual soil types, the permafrost distribution ranges from continuous for primarily peat to discontinuous and sporadic, for the silt and clay soils. For the same time period, climate trends for Peawanuck also demonstrated very weak negative correlations, but not statistically significant, for TDD; for

FDD, negative correlations that were statistically significant were observed. This further demonstrates the importance of peat, soil organic matter, for supporting permafrost. The assumption of homogeneity was considered when assessing Fs for the three soil types. To adjust

63 for the homogeneity, three theoretical scenarios were developed to capture the range of the three observed soil Kf/Ku ratios to demonstrate the enhanced freezing effects, the thermal offset phenomenon. This range in Kf/Ku ratio variation can also be representative of different or a combination (heterogeneity) of soil types and ground materials. The first scenario applied a

Kf/Ku ratio = 1.0, this assessed the potential permafrost without the asymmetric freezing and thawing processes, the results yielded low Fs values that ranged within the sporadic potential for permafrost with a few years reaching the discontinuous threshold. By removing the thermal offset, using Kf/Ku ratio = 1.0, Fs formulation becomes F where the results are based solely on the FDD and TDD ratio. The second scenario of Kf/Ku ratio = 1.5 represents conditions reflective of the soils observed at Peawanuck. Fs values in the second scenario had greater values than the first scenario, however the potential for permafrost was observed primarily in the discontinuous range with some sporadic distributions. The second scenario demonstrates the Kf/Ku ratio amplification on the FDD such that potential permafrost remains suitable within the discontinuous rather than the sporadic distribution. The final third scenario applied a Kf/Ku ratio

= 2.0 representing a strong thermal offset effect on the soil. As expected, the third scenario yielded the greater Fs values that raised the potential permafrost past the discontinuous permafrost threshold with some years passing the continuous permafrost threshold. This demonstrates that the potential for permafrost, shown by the Fs, can be reflective of the climate trends of the FDD and TDD; however, the influence of the soil thermal properties, by the Kf/Ku ratio, which includes variation of soil moisture content, on the FDD accumulation had demonstrated an important role in determining the presence of permafrost at Peawanuck.

2.5.2 Stefan Depths of Freezing and Thawing

In locations south of the subarctic, it is typical to have a negative difference in the Stefan depth values, condition of Xf < Xu, where the frozen soil of the winter season is completely thawed during the following summer season (French, 2007; Duan and Naterer, 2009). When the seasonal Stefan depth of thawing does not exceed the seasonal Stefan depth of freezing (Xf >

Xu), the remaining positive difference in the Stefan depth values represent the theoretical thickness of a frozen layer of soil that has persisted from the previous summer thawing season

(French, 2007; Duan and Naterer, 2009). Once the additional frozen soil persists through two subsequent freezing cycles without thawing, the resulting classification of permafrost may be applied (French, 2007; Duan and Naterer, 2009). In regions with continuous permafrost, such as in the Arctic and parts of the Subarctic, positive differences in the Stefan depth values can lead to the thickening of the permafrost layer (French, 2007; Duan and Naterer, 2009). With climate warming trends being experienced in the Canadian North, negative differences in the Stefan depth values (Xu > Xf) are expected to result in a thicker active layer that degrades the permafrost layer (French, 2007; Muller, 2008; Duan and Naterer, 2009). For Peawanuck, the

Stefan depth calculations where shown to be favourable for sustaining current permafrost presence where Stefan depths of freezing exceeded thawing depths. Since 1959, the data indicates a statistically significant decreasing trend in FDD and where little variation is shown in the TDD trend, however this was not statistically significant. As expected with continuous warming in the future, FDD should decreased with an increase in TDD such that the scenarios where TDD exceeded FDD may occur to favoring the negative Stefan depth condition (Xf < Xu).

However, the thawing process and negative impacts to the underlying permafrost layer may be delayed by a time lag response between the changing climate conditions. The presence of an

65 organic peat and vegetation surface layer may serve as a thermal insulator, with low soil thermal conductivities, against the warming summer air temperatures (Thie, 1974; Hinkel et al., 2001;

Cheng et al., 2004; Spielvogel et al., 2004; Martini et al., 2006; French, 2007; Zhang et al., 2008;

Pang et al., 2009). During the winter season, the presence of a snow layer above the soil can provide an additional thermal insulating effect that can prolong thawing of active layer in spring; however, this insulating layer can also negatively impact the permafrost layer by preventing beneficial heat loss from the soil column to the atmosphere during winter. These topics, time lag responses and thermal insulation, are beyond the scope of this work (Anisimov et al., 1997;

Cline, 1997; Cheng et al., 2004; Osterkamp, 2005; French, 2007; Zhang et al., 2008).

2.6 Conclusion

This research provides insight into the climatic potential for permafrost at Peawanuck,

Ontario, within the Canada’s subarctic region of northern Ontario, by using available permafrost tools such as the Frost Number, the Stefan equation, and the Stefan Frost Number. As soil thermal properties are not components in the Frost number formulation, the Stefan Frost Number was applied. The Stefan Frost Number formulation was able to capture the asymmetric freezing and thawing process through the Kf/Ku ratio modification on the FDD variable. Active layer simulations using the Stefan equation have demonstrated that active layer depths have thickened over the last 50 years, provide insight on possible impacts to the underlying permafrost (Burn and Smith, 1988; Anisimov et al., 1997; Riseborough et al., 2008). Both the Stefan equation and

Fs tools have demonstrated that potential permafrost was favourable at Peawanuck during the periods of available climate data, 1959-1977 and 1994-2011. The Kf/Ku ratio amplification on the

FDD provided insight that with warming climate trends expected in the future, there is

66 possibility for permafrost potential to remain favourable. The organic peat material and silty- clayey soils of this study location, with greater soil moisture contents, may enhance the Kf/Ku ratio during the winter, to favour the freezing process. The upper peat layer may provide additional insulating and thermal offsetting effects against warming air temperatures during the summer seasons; however, further reductions to the FDD accumulations in the future will negatively affect the Stefan freezing depths and potential permafrost.

2.7 Acknowledgements

Portions of this study was funded and supported by the Wildlife Research and

Development Section of the Ontario Ministry of Natural Resources, in which we gratefully thank

Dr. Martyn Obbard and Kevin Middel, and the Department of Physical and Environmental

Sciences at the University of Toronto Scarborough.

2.8 References

Anisimov, O., Shiklomanov, N., and Nelson, F. 1997. Global warming and active-layer thickness: results from transient general circulation models. Global and Planetary Change 15, 61-77.

Bockheim, J., Mazhitova, G., Kimble, J., and Tarnocai, C. 2006., Controversies on the genesis and classification of permafrost-affected soils. Geoderma 137, 33-39.

Burn, CR., and Smith, CAS. 1988. Observation of the “Thermal Offset” in Near- Surface Mean Annual Ground Temperatures at Several Sites near Mayo, Yukon Territory, Canada. Arctic 41. 91-166.

Cheng, G., Zhang, J., Sheng, Y., and Chen, J. 2004. Principle of thermal insulation for permafrost protection. Cold Regions Science and Technology 40, 71-79.

Cline, D. 1997. Snow surface energy exchanges and snowmelt at a continental, midlatitude Alpine site. Water Resources Research 33, 689-701.

Duan, X., and Naterer, G. 2009. Heat conduction with season freezing and thawing in an active layer near a tower foundation. International Journal of Heat and Mass Transfer 52, 2068-2078.

Environment Canada. 2012. “Hourly Data Report: Peawanuck (AUT) Ontario.” Government of Canada. Accessed online: 01 October 2012 from: http://climate.weatheroffice.gc.ca/climateData/hourlydata_e.html.

Eyles, N. and Miall, A. 2007. Canada Rocks: The Geologic Journey. 1st Ed. Markham: Fitzhenry & Whiteside Limited.

French, M. 2007. The Periglacial Environment. 3rd Ed. West Sussex: John Wiley & Sons.

Gough, WA., and Leung, A. 2002. Nature and fate of Hudson Bay permafrost. Regional Environmental Change 2, 177-184.

Kokelj, SV., Riseborough, D., Coutts, R., and Kanigan, JCN. 2010. Permafrost and terrain conditions at northern drilling-mud sumps: Impacts of vegetation and climate change and the management implications. Cold Regions Science and Technology 64, 46-56.

Kujala, K., Seppälä, M., and Holappa, T. 2008. Physical properties of peat and palsa formation. Cold Regions Science and Technology 52, 408-414.

Martini, I., Cortizas, M., and Chesworth, W. 2006. Peatlands: Evolution and Records of Environmental and Climate Changes. Elsevier B.V. pp 53-72.

Muller, S. 2008. Frozen in Time: Permafrost and Engineering Problems, HM French and FE Nelson (eds). Reston, VA: American Society of Civil Engineers.

Natural Resources Canada. 2006. “Physical Components of Watersheds: Permafrost.” Government of Canada. Accessed online: 01 October 2012 from: http://atlas.nrcan.gc.ca/site/english/maps/environment/hydrology/watershed1/.

Natural Resources Canada. 2007. “Ecological Framework: Terrestrial Ecozones.” Government of Canada. Accessed online: 01 October 2012 from: http://atlas.nrcan.gc.ca/site/english/maps/environment/ecology/framework/ terrestrialecozones/.

Nelson, F. 1986. Permafrost Distribution in Central Canada: Application of a Climate-Based Predictive Model. Annals of the Association of American Geographers 76, 550-569.

Nelson, F., and Outcalt, S. 1987. A Computational Method for Prediction and Regionalization of Permafrost. Arctic and Alpine Research 19, 279-288.

Nixon, J., and McRoberts, E. 1973. A study of factors affecting the thawing of frozen soils. Canadian Geotechnical Journal 20, 439-452.

Osterkamp, T. 2005. The recent warming of permafrost in Alaska. Global and Planetary Change 49, 187-202.

Pang, Q., Cheng, G., Li, S., and Zhang, W. 2009. Active layer thickness calculation over the Qinghai-Tibet Plateau. Cold Regions Science and Technology 57, 23-28.

Riseborough, D., Shiklomanov, N., Etzelmüller, B., Gruber, S., and Marchenko, S. 2008. Recent Advances in Permafrost Modelling, Permafrost and Periglacial Processes 19, 137–156.

Shur, Y., and Jorgenson, M. 2007. Patterns of Permafrost Formation and Degradation in Relation to Climate and Ecosystems. Permafrost and Periglacial Processes 18, 7–19.

Spielvogel, S., Knicker, H., and Kogel-Knabner, I. 2004. Soil organic matter composition and soil lightness. Journal of Plant Nutrition and Soil Science 167, 545-555.

Tam, A. 2009. Permafrost in Canada's Subarctic Region of Northern Ontario. Master’s Thesis. Department of Geography, University of Toronto, Toronto.

Thie, J. 1974. Distribution and Thawing of Permafrost in the Southern Part of the Discontinuous Permafrost Zone in Manitoba. Arctic 27, 165-248.

Woo, MK., Arain, MA., Mollinga, M., and Yi, S. 2004. A two-directional freeze and thaw algorithm for hydrologic and land surface modeling. Geophysical Research Letters 31. DOI: 10.1029/2004GL019475.

Zhang, Y., Wang, S., Barr, A., and Black, T. 2008. Impacts of snow cover on soil temperature and its simulation in a boreal aspen forest. Cold Regions Science and Technology 52, 355-370.

Chapter 3 An application of the Stefan Frost Number and XG-Algorithm in the Canadian Subarctic and Arctic Regions from 2004 to 2011 3 Chapter 3

3.1 Abstract

An assessment on the potential permafrost distributions at five study locations in northern

Canada from 2004 to 2011 was conducted using the Stefan Frost Number equation. Continuous permafrost distributions were verified for Nunavut at Rankin Inlet, Resolute Bay, Eureka, and

Alert. The importance of the ratio between the frozen and unfrozen soil thermal conductivities was further investigated to rationalize the apparent discrepancy between the observed continuous permafrost and the climate suggested discontinuous permafrost distributions within the Hudson

Bay Lowlands at Peawanuck, Ontario. As an indicator for permafrost change, active layer thicknesses along this geographical south to north study transect were simulated using the XG-

Algorithm, with available climate data and site-specific physical properties. Over the last eight years, the active layer thicknesses at each study location have increased, with the greatest thickening observed in the High Arctic locations. XG-Algorithm simulation results were compared with observed field measurements at two locations in a validation exercise; the overall model systematic error was +0.5% for the Peawanuck, Ontario; and +1.2% for Alert, Nunavut. In addition, soil physical properties and active layer thicknesses measurements collected from the remote study locations are provided.

Key words: Active layer; Stefan equation; Thermal offset; Climate change; Permafrost

3.2 Introduction

Warming of the climate system is unequivocal, and many natural systems are being affected by regional climate change, particularly temperature increases (IPCC, 2007). Numerous studies show that global warming is amplified in cold regions, such as in the polar and in high altitude regions. Research indicates that Arctic average temperatures have risen at almost twice the rate as the rest of the world in the past few decades (Hassol, 2004). Permafrost is an important part of the cryosphere, and the active layer is a key indicator of climate change. The active layer is located above the permafrost, and the thickness of this layer varies throughout the season because of the freezing and thawing interactions driven by the climate and ground conditions (Smith and Burgess, 2002; French, 2007). The variation in active layer thickness can lead to impacts on the underlying permafrost layer that may affect local ground stability

(Sazonova et al., 2004). Understanding the response of the active layer to both freezing and thawing can assist planners, engineers, and decision makers in assessing adaptable methods to cope with future climate change (Sazonova et al., 2004). Given the importance of the active layer, the research question for this study is, "Can the current distribution of permafrost be readily assessed within the Canadian Subarctic and Arctic regions based on climatological and ground parameters, and can the active layer thickness profiles at these locations be simulated and compared to determine permafrost change?

In this study, the potential permafrost distribution was assessed based on recent air temperature data and soil properties obtained at five locations along a geographical south to north transect within northern Canada, as shown in Figure 3-1. The study locations range from the Subarctic through the Low Arctic, and into the High Arctic regions; permafrost distribution

71 at the five locations is classified as continuous permafrost (Brown, 1973; Gough and Leung,

2002; Natural Resources Canada, 2006; Tam, 2009). To assess the distributions of permafrost for the specific locations, the Frost Number (F) and the Stefan Frost Number (Fs) equations, as shown in Nelson and Outcalt (1987), were applied and compared for all locations. Fs is a modification of F by incorporating the Stefan equation, which includes the soil physical and thermal conductivity variables; F is based on the climatic variables of the freezing and thawing degree days. The inclusion of the soil properties in Fs is particularly important as the soil thermal conductivity is different for frozen and unfrozen soils, which causes the ‘negative thermal offset’ phenomenon, a discrepancy in freezing and thawing depths (Romanovsky and Osterkamp, 1995;

Burn and Smith, 1987; Gough and Leung, 2002).

The active layer profiles for the five study locations were simulated using climate data obtained from local weather stations with site-specific physical parameters that were applied in a algorithm using the Stefan equation, the XG-Algorithm (Xie and Gough, 2013). During the last several decades, the Stefan equation was widely used in permafrost research for spatial active- layer characterization by estimating ground properties empirically, using air temperature records and active-layer data obtained from representative locations (Nixon and McRoberts, 1973;

Nelson, 1986; Broadridge and Pincombe, 1995; Rees, 2006; Hayashi et al., 2007; Hughes and

Braithwaite, 2008; Riseborough et al., 2008; Xie and Gough, 2013). This paper formally demonstrates an application of the XG-Algorithm, a modification of the Stefan equation, as proposed by Xie and Gough (2013) for calculating thawing and freezing depths within multilayered ground materials. The XG-Algorithm was proposed as a mathematical correction to the Jumikis (1977) and Lunardini (1981) methods that resulted in large systematic errors in simulation results when applied in multi-layered systems. The XG-Algorithm can be used to

72 determine active layer thickness employing available climate data (temperature), and ground material parameters such as: soil moisture content, soil thermal conductivity, soil bulk density, and soil depth, and layer thickness data.

Figure 3-1. The geographical south to north transect map containing the five study locations within the Canadian Subarctic, Low Arctic, and High Arctic

3.3 Methodology

3.3.1 Description of the study locations

For the five study locations, the southern-most location is situated in the subarctic region within the Hudson Bay Lowlands (HBL) of Northern Ontario. Moving northward into the Low

Arctic region, climate data were selected from Rankin Inlet, Nunavut (RAN), and three locations within the Canadian High Arctic region at Resolute Bay (RES), Eureka (ERK) and Alert (ALR),

Nunavut (Figure 3-2).

Figure 3-2. Site condition at local weather stations taken at A) Peawanuck, Ontario (Hudson Bay Lowlands); B) Resolute Bay, Nunavut; C) Eureka, Nunavut; and D) Alert, Nunavut.

Climate data were obtained through the Environment Canada’s National Climate Data and Information Archive for the five study locations, and the details of the weather stations are provided in Table 3-1.

Table 3-1. Study location details Locations Latitude Longitude Average annual WMO ID* Surface condition air temperature (°C) Peawanuck, 55°0’30” N 85°25’20” W –3.3°C 71434 Dense peat Ontario material with low lying vegetation Rankin 62°48’35” N 92°05’58” W –11°C 71083 Gravel and Inlet, sedimentary Nunavut materials with limited vegetation Resolute 74°41’51” N 94°49’56” W –16.4°C 71924 Gravel and Bay, sedimentary Nunavut materials Eureka, 79°59’20” N 85°56’30” W –19.7°C 71917 Sandy clay loam Nunavut and clay compositions Alert, 82°30’1” N 62°20’37” W –18°C 71355 Sandy clay loam Nunavut with shallow depths to fractured shale bedrock * World Meteorological Organization (WMO) identification number.

Active layer thickness measurements were obtained by field measurements at HBL,

ERK, and ALR locations. Physical parameters of ground material were collected within the HBL inclusively from August 21 to 27, 2008, and active layer thickness measurements were collected and used to provide algorithm validation. Active layer thickness measurements for ERK were collected on June 4th, 2011, and for ALR, numerous measurements were collected from August

2007 to August 2012.

Site conditions show that the HBL landscape is predominately overlain by an upper layer of organic peat material that coincides with suitable conditions for thermal offsetting effects that favours the presence of permafrost (Burn and Smith, 1987; French, 2007; Gough and Leung,

2002; Tam, 2009). The thermal offset is influenced by modifications to the soil moisture and thermal conductivity properties. For example, in warm and relatively dry summers, the upper peat layer serves as a thermal insulating layer by reducing the unfrozen soil thermal conductivity value (typical values range from 0.50 to 1.00 W/(m·k); Xie and Gough, 2013). During the wet fall season when temperatures reach freezing, the presence of greater soil moisture and ice enhances the soil thermal conductivity for freezing; in comparison, the frozen soil thermal conductivity is typically 1.5 to 2.0 times greater than the unfrozen soil thermal conductivity values (Nixon and McRoberts, 1973; Burn and Smith, 1987; Gough and Leung, 2002; Kujala et al., 2007; Shur and Jorgenson, 2007; Tam, 2009). In the Low Arctic region at RAN, site conditions reveal that tundra becomes the dominant feature where vegetation density becomes sparsely distributed which reduces the upper soil organic layer and soil thermal offset effect.

Similarly for RES, ERK and ALR, the inherently nutrient-poor tundra prevents the natural development of any significant vegetation layer; the subsurface materials are mainly mineral properties with relatively shallow bedrock. Variation in the site-specific characteristics revealed that in RAN and RES the top layer of tundra was dominantly mineral composing mostly of gravel materials. For ERK, this site exhibited dominant clay and sandy compositions due to the lack of gravel deposits. At ALR, the ground material was primarily composed of sandy clay loam, however, numerous outcrops of the shallow and fractured shale bedrock were observed.

For all study locations, access to historical active layer measurement data were not available.

Observed site-specific characteristics were applied to best represent the ground material profiles as shown in Figure 3-3. Three-layered XG-Algorithm approach with the third layer being a mineral subsurface layer or shallow bedrock was applied for the majority of the study locations. For the HBL, a thick upper organic layer (0.3 m) was specifically applied based on in- situ observations. For RAN and RES, a layer of gravel material above the natural tundra was incorporated to the simulation to represent local land practices (Nuna Burnside Engineering and

Environmental Ltd., 2010; Department of Economic Development and Transportation, 2009).

Specifically at ERK where soils are predominately clayey and sandy with little gravel, a dominant clay layer was incorporated for this region (Defence Construction Canada, 2010a). For

ALR, a two-layer soil profile was selected to best represent a known tundra location containing a shallow subsurface layer and immediate shale bedrock (Defence Construction Canada, 2010b).

3.3.2 Freezing and thawing degree-days

Climate data was applied to determine the Thawing and Freezing Degree-Day Indices

(TDD and FDD, respectively). For all study locations, daily average air temperatures (°C) within the time range from August 2003 to August 2012 were obtained to calculate the TDD and FDD

Indices to represent the period from 2004 to 2011. For the TDD index, the summation of the daily average air temperature for each day greater than 0°C was applied; and for FDD, the threshold for each day less than 0°C was applied (Juliussen and Humlum, 2007).

3.3.3 The Frost Number and the Stefan Frost Number

The Frost Number (F) is shown in Nelson and Outcalt (1987) as:

FDD0.5 F  ( ) . (3-1) TDD0.5  FDD0.5

F is a ratio between the square root function of the FDD and TDD balance. The presence of continuous permafrost was determined when F ≥ 0.67 threshold. If F < 0.67, discontinuous permafrost distributions are likely to be observed. If F < 0.6 sporadic permafrost is likely and <

0.5 no permafrost can be supported by the climate. This study further applies the Stefan Frost

Number calculation (Fs), also shown in Nelson and Outcalt (1987), as a relative tool for determining permafrost distributions. The FDD is weighted by ratio of the frozen and unfrozen

soil thermal conductivities, Kf and Ku, respectively, in order to account for the amplifying effect on FDD by the often substantially higher frozen soil conductivity when Kf > Ku.

(3-2)

The Stefan equation (X) provides a useful method for predicting the depth of thawing (t)/freezing

(f) in soils where little site-specific information is available (Nelson et al., 1997; Woo et al.,

2004). The Stefan equation is represented by equation (3-3):

2k  DD 2k  DD X  ( )0.5  ( )0.5 , (3-3) QL L    where k represents the soil thermal conductivity (W/(m·k)); DD is the Degree-Day Index; and

QL = L·ω·ρ; the volumetric latent heat of soil composed of: L, latent heat of fusion of ice (J/kg);

ω is the gravimetric moisture content (%); and ρ is the soil bulk density (kg/m3).

The Fs can be simplified to solely consider the climatic component and the freezing and thawing soil thermal conductivity ratio value:

FDD  K f / Ku  Fs  ( ) . (3-4) FDD  K f / Ku  TDD

The permafrost distribution boundary thresholds for Fs thresholds are identical to F for continuous, discontinuous and no permafrost; however, the significant difference between F and

Fs calculation is the accounting of the thermal offset phenomenon on the FDD variable using the ratio of the frozen to unfrozen soil thermal conductivities (Burn and Smith, 1987; Gough and

Leung, 2002). For Canadian Subarctic, Fs is expected to result in the range: 0.5 – 0.67; for the

Arctic, Fs is expected to be ≥ 0.67. However, in warmer climate areas at lower latitudes with

80 greater TDD and lower FDD values, Fs is expected to be <0.5, indicating a negative permafrost potential. It is possible for permafrost to continue to exist as relict if Fs drops below 0.5 due to permafrost’s substantial thermal inertia, however, observations of this effect are beyond the scope of this study.

3.3.4 The XG-Algorithm

Xie and Gough (2013) presented a simple algorithm, XG-Algorithm, to determine the freezing/thawing front in a multilayered ground surface where each layer contained different physical parameters and varying thickness depths. The XG-Algorithm can improve simulation accuracy by avoiding misrepresentations of ground material parameters in multilayered ground associated with the arithmetic averaging. In accordance with Xie and Gough (2013), for a given surface freeze/thaw index, the thaw/freeze depth of two soil types (types A and B; indicated below by a suffix, a or b) in the same locality can be calculated by the Stefan equation (3-3):

2ka  DD 0.5 2ka  DD 0.5 X a  ( )  ( ) , (3-5) QLa L a  a

2kb  DD 0.5 2kb  DD 0.5 X b  ( )  ( ) . (3-6) QLb L b  b

The XG-Algorithm establishes the relationship between the equations (3-5) and (3-6) to produce a ratio between the physical parameters (P) of both layers types A and B:

X a 2ka  DD/(L  a a ) 0.5 ka  b b 0.5 Pab   ( )  ( ) . (3-7) X b 2kb  DD/(L  b b ) kb  a a

The ratio Pab, equation 3-6, can be applied to determine the freeze/thaw depth in Soil Type B

(Xb) when Xa is known:

X X  a b P ab . (3-8)

In the multilayered calculations, the XG-Algorithm requires real soil layer depth (Z) to produce the freeze/thaw depths (X). For a three-layered ground profile where layer 1 (Z1), layer 2 (Z2) and layer 3 (Z3) and a specific Z exists, the Stefan equation is first applied:

2k1  DD 0.5 X 1  ( ) . (3-9) 1 1  L

If the condition X1 ≤ Z1 occurs, the freezing/thawing depth is contained within the real depth of layer 1 and the entire process is completed. If the condition X1 > Z1 occurs, the XG-Algorithm for multilayered thawing/freezing is applied to determine the freezing/thawing depth into layer 2:

X1  X1  Z1 . (3-10)

For this example, an assumption has been made that layer 2 is located below layer 1 and has different physical parameters. Similar to equation (3-8), the freezing/thawing depth in layer 2 can be determined:

X X  1 2 P 12 . (3-11)

The ratio method for determining P12 is analogous to equation (3-7). In an iterative process, if the conditions of X2 ≤ Z2 occurs, the process is completed which results in a two-layered

thawing/freezing algorithm with a freezing/thawing depth of Z1 + X2. If the condition X2 > Z2 occurs, the XG-Algorithm is applied to determine the freezing/thawing depth into layer 3:

ΔX2 = X2 – Z2, (3-12).

The iterative approach of the XG-Algorithm in determining the freezing/thawing depth can be summarized by:

in X  (Zi )  X n1 i1 . (3-13)

3.3.5 Algorithm input parameters

Input parameters for observed site-specific ground material conditions are listed in Table

3-2 for the application of the XG-Algorithm to simulate active layer thickness development and freezing processes. In determining the Kf and Ku values for the Kf/Ku ratios of the study locations, a generalization of the observed soil profiles of Figure 3-3 was assumed by calculating the mean K value based on the respective compositions and state from Table 3-2.

Table 3-2. Model Input Parameters Soil Component Soil Thermal Soil Thermal Soil Bulk Soil Moisture Conductivity Conductivity Density Content (frozen, kf ) (unfrozen ku) (ρ) (ω) Type W/(m·k) W/(m·k) kg/m3 % Peat (Organic) 1.38 0.79 680 70 Peat w/ Silt 0.46 0.37 800 40 Sandy-silt w/ Peat 2.48 1.52 1250 15 Sandy Clay Loam 1.99 1.54 1500 25 Silt 1.38 1.11 1450 27 Clay 2.08 1.73 1300 37 Gravel (Fill) 2.00 1.50 1900 10 Gravel, pebbles, sand, silt 2.36 1.35 1700 20 Granite (Bedrock) 2.20 2.20 2700 10 Shale (Bedrock) 2.40 2.20 2500 10

3.3.6 Error Calculations

Two error calculations are presented in this research to assess the difference between the simulation results produced from the XG-Algorithm with available observed active layer thickness measurements collected during field visits at two locations, HBL and ALR. The two analyses are the systematic error, as followed in Xie and Gough (2013), and Root Mean Square

Error (RMSE) to determine the total error, including non-systematic error, as shown in Hinzman et al. (1998). For systematic error, a simple measure was produced using the percentage difference between the calculated depth values and the observed field measurement values (Xie and Gough, 2013). The RMSE method accounts for the square root of the mean squared difference between the simulated (predicted) and observation values where RMSE values can range from zero to infinity (Hinzman et al., 1998). For RMSE, a value of zero indicated a perfect fit between predicted and observation (Hinzman et al., 1998). However, due to the squared

84 function in RMSE, over-predictions and under-predictions cannot be distinguished (Hinzman et al., 1998).

3.4 Results

3.4.1 Frost Number and Stefan Frost Number

F and Fs were calculated for the five study locations and shown in Figures 3-4a and 3-

34b. Using the continuous permafrost distribution threshold of F and Fs = 0.67, four of the five study locations were consistently above the continuous permafrost threshold. Error bars at ±1 standard deviation were applied to demonstrate additional variability. For Fs in Figure 3-4b, the thermal offset, Kf/Ku ratio, was 1.6 for HBL, RAN had a Kf/Ku of 1.3, RES at 1.4, ERK at 1.2, and ALR at 1.2.

Results for RAN, RES, ERK, and ALR from 2004 to 2011 showed values greater than

0.67 indicating continuous permafrost distributions. For HBL, F values suggested discontinuous permafrost distribution; this was similar for Fs, however, when including +1 standard deviation,

2004, 2008, and 2009 are near or exceed the 0.67 threshold. A further analysis was conducted for

HBL under three thermal offset conditions as shown in Figure 3-4c. Under condition 1, the thermal offset effect was removed, where the ratio Kf/Ku = 1.0; for condition 2, thermal offset is present at observed site-specific ground material parameters for the HBL (Figure 3-3), Kf/Ku =

1.6; and for condition 3, a larger Kf/Ku value was selected at 2.0. For condition 1, without thermal offset, Fs results were less than the 0.67 threshold, indicating discontinuous permafrost distributions for the years 2004-2011. The minimum Fs value under condition 1 was observed at

0.522 (2010) and a maximum of 0.610 (2004) with one-standard deviation of 0.03; discontinuous permafrost distributions. For condition 2 with the thermal offset present, the HBL Fs results were below the continuous permafrost threshold of 0.67; however, in considering one-standard deviation of 0.03, the following years were near or exceeded the continuous threshold: 2004

(0.664), 2008 (0.644), and 2009 (0.654). For all other years in condition 2, Fs results were greater than 0.5 for discontinuous permafrost. Under condition 3, a larger thermal offset value was applied, three of the HBL Fs results exceeded the continuous permafrost threshold in 2004

(0.689), 2008 (0.670), and 2009 (0.680). In considering one-standard deviation of 0.03 for condition 3, the following additional years are near or exceed the continuous permafrost threshold: 2005 (0.640), 2006 (0.629), 2007 (0.646), and 2011 (0.646).

3.4.2 Active Layer Simulations

XG-Algorithm results for the five study locations within the study transects produced the active layer thickness developments for the period of 2004 to 2011 in Figure 3-5, using the model input parameters from Tables 3-1 and 3-2. Specific active layer thickness developments and the subsequent 1-directional freezing of the active layers were produced for the year 2011 to demonstrate the variability between study locations (Figure 3-5). The HBL was observed to have the greatest active layer depth ranging from 1.5 to 2.0 m. Active layer thickness was the shallowest at RES and ALR with maximum thickness near 1 m in depth. As shown in Table 3-3, the active layer thickness was observed to have deepened over the period from 2004 to 2011, with greatest negative correlation trends at RES and ALR showing statistical significance with p- values of 0.011 and 0.001, respectively.

In 2011, the active layer thickness development began earliest in the HBL region in late-

April after snow disappearance; RAN and ERK in late-May; RES and ALR in mid-June.

Freezing of the surface soil layer began in late-October for the HBL; mid-October for RAN; and, early-September for RES, ERK and ALR (Figure 3-5). The surface layer at HBL was unfrozen for 6 months; complete freezing of the soil column was reached in the following January. For

RAN, the surface was unfrozen for 4.5 months and the subsurface remained unfrozen for an additional 2 months before completely freezing by mid-December. The duration of unfrozen surface soils at RES was almost 3 months; complete freezing of the active layer was reached within 2 months by the end of October. At ERK, the surface was unfrozen for 3 months; complete freeze up was reached in 1.7 months (mid-October). For ALR, the surface layer was

88 unfrozen for 2.5 months with the complete freezing of the active layer reach in 1.2 months by early-October.

Table 3-3. Statistical analysis of active layer depth trends from 2004 to 2011.

Active Layer Depth Trends 2004-2011 Site HBL RAN RES ERK ALR Corr. (r) -0.23 -0.68 -0.83 -0.69 -0.92 p-value 0.585 0.077 0.011* 0.062 0.001* *significance at alpha = 0.05.

3.5 Discussion

3.5.1 Field Validation XG-Algorithm

As active layer thickness records are unavailable for this study, for example HBL, RAN,

RES, and ALR are not part of the Circumpolar Active Layer Monitoring (CALM) Program

Network and there are no data presented for ERK for the 2004-2011 period. The validations of the XG-Algorithm results were compared to summer field observations that were collected from site visits. The XG-Algorithm outputs for certain days were compared with the real observed field measurements at that same calendar day. For example and as shown in Table 3-4, the XG-

Algorithm results for August 21, 2008 in the HBL was 135.2 cm; field observation for that same day was 132 cm, yielding a systematic error of +2.4%. In this study, systematic errors for the

XG-Algorithm are shown in Table 3-4 as percentage difference between the calculated depth values and the observed field measurement values. The average systematic error for the XG-

Algorithm in HBL was +0.5%, with a range from -1.9 to +2.5%; for ALR, the average systematic error for the XG-Algorithm was +1.2%, ranging from -5.4 to +9.1%. Furthermore,

RMSE values are shown in Table 3-4 between the calculated depth (simulated) and the observed field measurement values for both locations; the RMSE at HBL was 2.4 cm; and for ALR,

RMSE was 4.7 cm. As the RMSE applied a squares method, over-predications and under-

90 predictions cannot be distinguished; however, in comparison between the two locations, the results for HBL showed the least difference between XG simulated (predicted) and observed active layer thickness depths.

Table 3-4. Systematic and root mean square errors for the XG-Algorithm at two observation locations Observed XG-Algorithm Systematic Root Mean Square Location Date Depth (cm) Depth (cm) Errors (%) Error (cm) HBL 21-Aug-08 132 135.2 +2.4 - HBL 22-Aug-08 133.8 137.2 +2.5 - HBL 23-Aug-08 136.4 138.4 +1.5 - HBL 24-Aug-08 137.7 139.0 +0.9 - HBL 25-Aug-08 139.5 139.8 +0.2 - HBL 26-Aug-08 144 141.5 -1.7 - HBL 27-Aug-08 145.8 143.0 -1.9 - Average HBL 138.5 139.2 +0.5 RMSE = 2.4 ALR 16-Aug-07 60 64.1 +6.8 - ALR 18-Aug-08 80 87.3 +9.1 - ALR 18-Aug-09 85 80.4 -5.4 - ALR 27-Aug-11 88 89.9 +2.2 - ALR 21-Aug-12 95 91.2 -4.0 - Average ALR 81.6 82.6 +1.2 RMSE = 4.7

Greater range in systematic errors was observed for ALR in the High Arctic. Possible explanations for the inconsistencies include errors in the measured depths due to sampling errors and location specific details that the XG-Algorithm did not parameterize. For two sample locations within the ALR region that were underestimated by the XG-Algorithm, test pits were excavated for active layer thickness measurements, the depth of the test pits were reached when ice particles were observed to confirm ≤ 0°C soil temperatures. Thermal contamination may have resulted during the excavation process where the presence of ice can melt due to friction and exposure to the warmer atmospheric conditions (French, 2007). These influences may have

91 contributed to deeper active layer thickness measurements. The variability of local surface features (such as vegetation, water bodies, and slope) may affect the active layer depth. The input of additional heat energy from a local moving water source produced from spring melt or rock salinity could have thickened the active layer (Wang et al., 2009; Stotler et al., 2009) while only heat obtained from surface was considered in XG-Algorithm. The local presence of water may have increased the soil moisture content and enhanced the soil thermal conductivity of the soil resulting in greater thawing susceptibility. For the sample locations within the ALR region that were overestimated by the XG-Algorithm, the presence of shallow fractured shale bedrock prevented test pits from reaching maximum depths, although ice particles were observed at this site, the soil profile was observed to be two-layered composed primarily of sandy clay loam underlain by shale bedrock. Variation in the soil physical properties is a possible explanation as fractured shale may affect the soil thermal conductivity and bulk density values. For the HBL region, the XG-Algorithm performed well in comparison to the measured active layer thicknesses, as shown by low systematic errors below ±10% and by the lower RMSE value in comparison to ALR. Statistical correlation between the XG-Algorithm results and the observed active layer measurement was shown to be 0.991 (R2 = 0.983) with statistical significance at p<0.05. Overall, the XG-Algorithm demonstrated the ability to produce active layer thicknesses within the set input parameters of soil physical properties and field observations; the overall systematic error values in the validation exercise were within 1-standard deviation of the observed measurements (Xie and Gough, 2013).

3.5.2 Addressing the Research Question

For this study, the question was proposed: “Can the current distribution of permafrost be readily assessed within the Canadian Subarctic and Arctic regions based on climatological and ground parameters, and can the active layer thickness profiles at these locations be simulated and compared to determine permafrost change?" Five locations along a geographical south-to- north transect were selected. The study locations are located in the continuous permafrost zone

(Natural Resources Canada, 2006). The HBL was selected as the southernmost location in the geographical transect for study as this subarctic region is near the transition area between discontinuous and sporadic permafrost; ALR was selected as the northernmost location given the availability of climate data and soil properties. Climate data and soil properties were applied in the F and Fs equations to assess the distribution of permafrost. The F was first attempted, however, to account for the negative thermal offset phenomenon, the Fs was applied as this formulation included a physical component rather than a solely climatological relationship. The

Fs accounts for the differences in the thermal properties of the active layer, specifically the differences in soil thermal conductivities in the frozen and unfrozen ground states as demonstrated in Figure 3-4c. This study further simplified the Fs to include solely the freezing and thawing degree-days and the frozen and unfrozen soil thermal conductivity ratio variables, respectively, to better assess the distribution of permafrost. As demonstrated in Figure 3-4c, the amplifying influence of the frozen soil thermal conductivity on the FDD provides a reasonable explanation for the presence of continuous permafrost distribution in the HBL, even as climate conditions suggests unfavourable warming with discontinuous permafrost distributions.

Conveniently, both F and Fs share the same threshold values for permafrost distributions.

The XG-Algorithm was applied to simulate the active-layer thickness depths. For validation of the XG-Algorithm depths, site-specific active layer thickness depths were measured within three of the five study locations for the years 2008 to 2012. The application of the XG-Algorithm was successfully shown to be a robust model in simulating the active layer thickness profiles within a

Canadian Subarctic location and a High Arctic location. The systematic errors of the XG-

Algorithm results were demonstrated to be within 1-standard deviation of the observed measurements and that the overall average systematic error were +0.5% for the HBL and +1.2% for ALR. In considering the total error, the RMSE further supported that HBL showed the least difference between simulated and observed active layer thicknesses with a value of 2.4 cm where

ALR had a RMSE value of 4.7 cm. The XG-Algorithm demonstrated the ability to produce active layer thicknesses that allowed comparisons to be made between the active layer thicknesses, duration of unfrozen surface soils, the timings of the onset of thawing and freezing, and the timing in complete freezing of the active layer. Consecutive seasonal thawing and deepening of the active layer can impact the permafrost state by degrading the top of permafrost table that can lead to reductions of the permafrost layer thickness (Burn and Smith, 1987; Duan and Naterer, 2009). The active layer thickness results for all five study locations (HBL, RAN,

RES, ERK and ALR) demonstrated negative correlations for changes, thickening of the active layer; however, only two locations (RES and ALR) showing statistical significance, p<0.05

(Table 3-3).

Although the thermal insulating effects from surface vegetation and peat layers are not directly covered in the scope of this study, these important factors may also contribute to temporarily preserving continuous permafrost under climatically unfavourable conditions. It is widely known that the Canadian High Arctic may be more susceptible to climate change and

94 climate warming, as most ground lacks the protective and insulating vegetated and organic layers due to the harsh polar environment (Sazonova et al., 2004). With increasing active layer depths being observed over the last 8 years, and further expected in the coming future, it is not unreasonable to also conclude the possibility that there may be a future shift of permafrost distributions, specifically at the southern extent of this study transect, particularly the subarctic.

Further research using Global Climate Models and future climate warming scenarios will be applied to this study transect to determine any climate change indicators for permafrost, such as those discussed in French (1999), as: (1) an increase in active layer thickness, (2) increases in permafrost degradation, and (3) possible slope and active layer failures.

3.6 Conclusion

The application of Frost Number, the Stefan Frost Number, and the XG-Algorithm demonstrated the benefits of applying available climate data and site-specific physical parameters for remote locations. The active layer thicknesses can be produced within the geographical south to north study transects of the Hudson Bay Lowlands, the Low and High

Arctic regions. Overall, the XG-Algorithm performed well when validated with field observations yielding overall average systematic errors of +0.5% and +1.2%, and RMSE of 2.4 cm and 4.7 cm, for the HBL and ALR locations, respectively. Active layer thicknesses produced using the XG-Algorithm has indicated that depths have been deepening. The application of the

Fs further indicated that the HBL in the Subarctic region may have been experiencing unfavourable climate conditions for the presence of permafrost in recent years, and as demonstrated, the frozen and unfrozen soil thermal conductivities provides an explanation for the

95 presence of the observed continuous permafrost distributions, whereas the climate if solely considered would favour discontinuous distributions. For the Canadian Arctic, Fs results for

RAN, RES, ERK, and ALR have indicated favorable conditions for continuous permafrost distributions. Although the active layers in the study locations are expected to continue to thicken with warming climate, few impacts in the immediate future are anticipated given the overall cooler climate of the High Arctic and thick underlying permafrost layer.

3.7 Acknowledgements

Portions of this study was funded and supported by the Wildlife Research and

Development Section of the Ontario Ministry of Natural Resources and the Department of

Physical and Environmental Sciences at the University of Toronto Scarborough.

3.8 References

Broadridge, P., and Pincombe, B. 1996. The Stefan Solidification Problem with Nonmonotonic Heat Diffusivity. Mathl. Comput. Modelling 23, 87-98.

Burn, C., and Smith, C. 1987. Observation of the “Thermal Offset” in Near-Surface Mean Annual Ground Temperatures at Several Sites near Mayo, Yukon Territory, Canada. Arctic 41, 99-104.

Defence Construction Canada. 2010a. “Operation and Maintenance Plan CFS Eureka (ERK), Nunavut.” Nunavut Water Board Public Registry. Accessed Online: 01 December 2012 from: ftp://nunavutwaterboard.org/1%20PRUC/3%20MUNICIPAL/3B/3BC%20- %20Camp/3BC-ERK1015%20(DND%20Eureka)/.

Defence Construction Canada. .2010b. “Operation and Maintenance Plan CFS Alert (ALT), Nunavut.” Nunavut Water Board Public Registry. Accessed Online: 01 December 2012 from: ftp://nunavutwaterboard.org/1%20PRUC/3%20MUNICIPAL/3B/3BC%20- %20Camp/3BC-ALT1015%20(Alert)/.

Department of Economic Development and Transportation. 2009. “Resolute Bay Airport Operation and Maintenance Manual Sewage Treatment Facility.” Nunavut Water Board Public Registry. Accessed Online: 01 December 2012 from: ftp://nunavutwaterboard.org/1%20PRUC/3%20MUNICIPAL/3B/ 3BM%20-%20Municipality/3BM-YRB0308/.

Duan, X., and Naterer, G. 2009. Heat conduction with season freezing and thawing in an active layer near a tower foundation. International Journal of Heat and Mass Transfer 52, 2068-2078.

French, HM. 1999. Past and present permafrost as an indicator of climate change. Polar Research 18, 269-274.

French, HM. 2007. The Periglacial Environment. (3rd Ed.). John Wiley & Sons: West Sussex.

Gough, WA., and Leung, A. 2002. Nature and fate of Hudson Bay permafrost. Regional Environmental Change 2, 177-184.

Hassol,SJ. 2004. Impacts of a warming Arctic, Arctic climate impact assessment. Cambridge University Press; New York: 2004.

Hayashi, M., Goeller, N., Quinton, W., and Wright, N. 2007. A simple heat-conduction method for simulating the frost-table depth in hydrological model. Hydrological Processes 21, 2610-2622.

Hinzman, LD., Goering, DJ., and Kane, DL. 1998. A distributed thermal model for calculating soil temperature profiles and depth of thaw in permafrost regions. Journal of Geophysical Research 103, 28,975-28,991.

Hughes, P., and Braithwaite, R. 2008. Application of a degree-day model to reconstruct Pleistocene glacial climates. Quaternary Research 69, 110-116.

IPCC. 2007. “Climate Change 2007: Synthesis Report. Summary for Policymakers.” A report of the Intergovernmental Panel on climate change.

Juliussen, H., and Humlum, O. 2007. Towards a TTOP Ground Temperature Model for Mountainous Terrain in Central-Eastern Norway. Permafrost and Periglacial Processes 18, 161-184.

Jumikis, AR. 1977. Thermal Geotechnics. Rutgers Univ. Press: New Brunswick, NJ.

Kujala, K., Seppälä, M., and Holappa, T. 2007. Physical properties of peat and palsa formation. Cold Regions Science and Technology 54, 408-414.

Lunardini, VJ. 1981. Heat Transfer in Cold Climates. Van Nostrand Reinhold Publishers: New York.

Natural Resources Canada. 2006. "Physical Components of Watersheds: Permafrost." Government of Canada. Accessed Online: November 20th, 2011 from: http://atlas.nrcan.gc.ca/site/english/maps/environment/hydrology/watershed1

Nelson, F. 1986. Permafrost Distribution in Central Canada: Applications of a Climate-Based Predictive Model. Annals of the Association of American Geographers 76, 550-569.

Nelson, F., and Outcalt, SI. 1987. A Computational Method for Prediction and Regionalization of Permafrost. Arctic and Alpine Research 19, 279-288.

Nixon, J., and McRoberts, E. 1973. A study of factors affecting the thawing of frozen soils. Can. Geotech. J. 20, 439-452.

Nuna Burnside Engineering and Environmental Ltd. 2010. “Water Supply Facility Operation and Maintenance (O&M) Plan, Hamlet of Rankin Inlet.” Nunavut Water Board Public Registry. Accessed Online: 01 December 2012 from: ftp://nunavutwaterboard.org/1%20PRUC/3%20MUNICIPAL/3B/ 3BM%20-%20Municipality/3BM-GRA0207/.

Rees, W. 2006. Remote Sensing of Snow and Ice.Taylor & Francis/CRC Press: Boca Raton, FL.

Riseborough, D., Shiklomanov, N., Etzelmüller, B., Gruber, S., and Marchenko, S. 2008. Recent advances in permafrost modelling. Permafrost and Periglacial Process 19, 137-156.

Romanovsky, VE., and Osterkamp, TE. 1995. Interannual Variation of the Thermal Regime of the Active Layer and Near-Surface Permafrost in Northern Alaska. Permafrost and Periglacial Processes 6, 313-335.

Shur, Y., and Jorgenson, M. 2007. Patterns of Permafrost Formation and Degradation in Relation to Climate and Ecosystems. Permafrost and Periglac. Process. 18, 7–19.

Smith, S., and Burgess, M. 2002. A Digital Database of Permafrost Thickness in Canada. Geological Survey of Canada, Open File 4173, P79.

Stotler, RL., Frape, SK., Ruskeeniemi, T., Ahonen, L., Onstott, TC., and Hobbs, MY. 2009. Hydrogeochemistry of groundwaters in and below the base of thick permafrost at Lupin, Nunavut, Canada. Journal of Hydrology 373, 80-95.

Tam, A. 2009. “Permafrost in Canada’s Subarctic Region of Northern Ontario”. Master’s Thesis. University of Toronto, Department of Geography, Toronto.

Wang, G., Hongchang, H., and Taibin, L. 2009. The influence of freeze-thaw cycles of active soil layer on surface runoff in a permafrost watershed. Journal of Hydrology 375, 438-449.

Woo, MK., Arain, MA., Mollinga, M., and Yi, S. 2004. A two-directional freeze and thaw algorithm for hydrologic and land surface modeling. Geophysical Research Letters 31. DOI: 10.1029/2004GL019475.

Xie, C., and Gough, WA. 2013. A simple thaw-freeze algorithm for multilayered soils using the Stefan equation. Permafrost and Periglacial Processes, DOI: 10.1002/ppp.1770.

Chapter 4 An assessment of potential permafrost along a south-to-north transect in Canada under predicted climate warming scenarios from 2011 to 2100 4 Chapter 4

4.1 Abstract

Potential permafrost distributions at five study locations were assessed using future climate warming projections in a geographical south-to-north study transect from 55°N to 82°N within northern Canada. An ensemble of climate models produced mean changes to surface air temperatures that were applied to project future warming under IPCC A1B, A2, and B1 emissions scenarios for the entire 21st century. Validation of the multi-model ensemble air temperature means showed differences ranging from -0.1 to -0.2°C between the modeled and observed local baseline climate record. Results from the future climate scenarios were applied to the Stefan Frost Number to assess climatic conditions for permafrost distributions based on the ratio between freezing and thawing degree-days accumulations with consideration to site-specific soil thermal properties. Within the study transect, climate change projections indicate warming above the 1971-2000 mean air temperature baseline by a minimum of +1.5°C and a maximum of

+2.4°C for the period 2011-2040; +2.6 to +4.1°C for 2041-2070; and +3.3 to +7.1°C for 2071-

2100. Stefan Frost Number results projected that climate conditions will remain supportive for continuous permafrost distributions within the Canadian High Arctic region for this century. By

2040 in the Low Arctic, projections indicate shifts in the potential from continuous to discontinuous permafrost. For the southernmost extent of this study in the subarctic region of

100 northern Ontario, climate conditions for the remainder of this century are expected to be suitable for sporadic permafrost.

Keywords: Permafrost; Climate change; Climate modelling; Degree-days; Stefan Frost

4.2 Introduction

Permafrost is defined as ground material that remains below 0°C for at least two consecutive years (French, 2007; Dobinski, 2011; Waller et al., 2012). The impacts of climate warming and the response of permafrost can be evaluated using General Circulation Models

(GCMs) using future climate projections (Anisimov et al., 1997; Gough and Leung, 2002;

Sazonova et al., 2004; Woo et al., 2004; Zhang et al., 2008; Lewis and Lamoureux, 2010; Slater and Lawrence, 2013). All GCMs indicate projections of climate warming during the 21st century

(Sazonova et al., 2004; Zhang et al., 2008; Rinke et al., 2012; Slater and Lawrence, 2013). For the Canadian Arctic, warming is expected to range between 2.0 to 5.0 °C by 2100 (Smith et al.,

2005; Zhang et al., 2008; Throop et al., 2012). The Intergovernmental Panel on Climate Change

(IPCC) 2007 based, on atmospheric [CO2] emission scenarios, projected mean global climate to warm within the range of +1.1 to +6.4 °C by 2100, with the greatest temperature increase for northern latitudes (Solomon et al., 2007; Zhou et al., 2009). Mean temperature changes for the arctic by 2099, using Representative Concentration Pathway (RCP) radiative forcing targets such as RCP2.6, RCP4.5, RCP6.0 and RCP8.5 scenarios in the fifth phase of the Coupled Model

Intercomparison Project, were projected to be +2.2, +3.8, +4.5 and +7.8 °C, respectively (Slater and Lawrence, 2013). Possible impacts from climate change in permafrost regions will be changes to local land features and infrastructure that are dependent on ground stability (Zhou et

101 al., 2009; Etzelmüller et al., 2011; Zhang, 2013). Current northern infrastructure is designed for cooler climate conditions; the onset of warming can deepen the active layer, thus degrading the underlying permafrost by reducing soil strengths, ground slumping, and failures (Anisimov and

Nelson, 1996; Smith et al., 2005; French, 2008; Zhou et al., 2009; Dobinski, 2011; Zhang, 2013).

The potential for permafrost were assessed at five study locations, along a geographical north-to-south transect from 55°N to 82.5°N, under three future climate projections produced from ensembles of GCMs using the Localizer Tool provided by Environment Canada, the

University of Toronto, and the University of Prince Edward Island Climate Sections (UTSC,

2013). Previous permafrost research at the five study locations are limited, however there have been some recent research conducted between the 55°N to 82.5°N latitudes, such as Throop et al.

(2012) at similar study locations of Alert and Baker Lake (near Rankin Inlet) in Nunavut. For the

Hudson Bay Lowlands at Peawanuck in northern Ontario (adjacent to Polar Bear Provincial

Park), research is limited (Gough and Leung, 2002; Tam, 2009; Tam et al., 2014a; Tam et al.,

2014b); however, other recent studies were conducted by Zhang et al. (2012) and Zhang (2013) to the west of this study location at Wapusk National Park near Churchill, Manitoba. Derksen et al. (2012) summarized key research conducted during the International Polar Year on permafrost at various locations, including the Canadian high Arctic, northern Manitoba, and northern

Quebec.

For this study, climate change scenarios relied upon the available emissions scenarios provided by the IPCC 4th Assessment Report built into the Localizer Tool (TGICA, 2007;

UTSC, 2013). These models are initialized in 1900 and produce a simulation from that point to

2100. The first century is forced by greenhouse gas concentrations from observations and the

102 latter century using a range of emission scenarios. The Localizer Tool was applied to determine multi-model ensemble mean projected changes in air temperatures that are ‘bias-corrected’ to specific geographical locations under three such scenario experiments: A1B, A2, and B1

(Fenech, 2009; UTSC, 2013). The resolution of the Localizer Tool is approximately 200 km by

200 km (UTSC, 2013). The ‘bias-correction’ in the Localizer Tool refers to the correction between the difference of the model biases in projected climate changes from the period of 1971 to 2000 and future periods of 2011 to 2100 with actual baseline climate data obtained from local observation stations with at least 70% data availability (Fenech, 2009; UTSC, 2013). Once bias- corrected, the average ensemble change for the future periods are applied to the observed local baseline climate values (Fenech, 2009; UTSC, 2013). The A1B scenario experiment produced mean annual air temperatures (MAAT) from an ensemble of 24 climate models; A2 applied 20 models; and, 21 models for B1 (UTSC, 2013). The scenario A1B represents a future with moderate emissions where atmospheric [CO2] reaches 720 ppm and stabilizes by 2100; A2, high emissions where atmospheric [CO2] continues to increasing and reaches 860 ppm by 2100; and,

B1, low emissions reflecting a more ecologically friendly future where atmospheric [CO2] reaches 620 ppm (TGICA, 2007; Lewis and Lamoureux, 2010; Etzelmüller et al., 2011).

The climate projections were applied to assess climate change impacts to the potential for permafrost based on calculating threshold values of the Stefan Frost Number (Fs). This provides insight of either favourable or unfavourable climate conditions for permafrost distributions. The distributions of permafrost are generally referred to as continuous, discontinuous, sporadic and isolated, and permafrost absent (Gough and Leung, 2002; French, 2007; Zhang et al., 2008;

Dobinski, 2011). Climatological conditions may affect the permafrost distributions because of the active layer becoming thicker with climate warming (Anisimov and Nelson, 1996; Wang et

103 al., 2009). Thickening of the active layer can thaw portions of the underlying permafrost, and extensive thickening of the active layer can completely thaw areas with thin or relict permafrost

(Anisimov and Nelson, 1996; French, 2007; Dobinski, 2011; Waller et al., 2012). The following research question was proposed: “Can the potential permafrost distributions be determined at five study locations in Canada using the Stefan Frost Number equation and can future changes to the potential be predicted using GCM results under climate warming scenarios?” This study acknowledges that a lag time difference exists between the climate conditions and changes to permafrost distributions within the subsurface; the intent of this study is to provide information on the climatological conditions suitable for permafrost distribution at each location rather than establishing definite time frames for distribution change (Anisimov et al., 1997).

For a quick assessment of the climatological conditions for permafrost distributions, the

Frost Number (F) Index (Nelson and Outcalt, 1987) can be applied where climate records are available. Anisimov and Nelson (1996) have demonstrated changes to the permafrost distributions using F under climate warming projected by GCMs. The F analysis determines a ratio between the freezing and thawing degree-days accumulations, which was solely based on regional temperature data (Nelson and Outcalt, 1987). Furthermore, Slater and Lawrence (2013) applied a similar simple permafrost model, using a modified F approach, to provide insight on future permafrost distributions in current permafrost regions. To incorporate ground thermal properties into the ratio of freezing and thawing degree-days to account for soil complexities,

Tam et al. (2014a) investigated the Fs equation from Nelson and Outcalt (1987), with simplifications that emphasized a modifying ratio between the frozen and unfrozen soil thermal conductivities on the freezing degree-days accumulations. The modification of the freezing degree-days parameters was rationalized based on frozen soil thermal conductivity values being

104 typically 1.5 to 2 times greater than the unfrozen soil thermal conductivity, under similar soil moisture conditions (Nixon and McRoberts, 1973; Burn and Smith, 1987; Gough and Leung,

2002; Côté and Konrad, 2005; Kujala et al., 2007; Shur and Jorgenson, 2007; Duan and Naterer,

2009; Tam, 2009; Kokelj et al., 2010). This asymmetric difference in soil thermal properties amplifies the soil freezing process, also referred as the “negative thermal offset” in Romanovsky and Osterkamp (1995). Similar to F, the Fs threshold values, calculated from site-specific conditions, can be applied as an indicator for permafrost distributions (Nelson and Outcalt,

1987).

This study further acknowledges the role and influence of a winter snow layer; however, this factor is omitted from the scope of this research due to lack of continuous snow data. The presence of a snow layer above the soil surface may provide a thermal insulating effect that can affect the timing of spring thawing, and in preventing additional heat loss from the soil during winter, which may have impacts on the F and Fs results (Slater and Lawrence, 2013). For example, during the winter season the presence of a snow layer above the soil can provide an additional thermal insulating effect, and higher albedo, that can prolong thawing of active layer in spring when air temperatures are warmer. This same insulating layer can also negatively impact the permafrost layer by preventing beneficial heat loss from the soil column to the atmosphere during winter (French, 2007). Nonetheless, the contributions from this study provides an attempt to address a greater data gap issue in permafrost research in Canada, specifically along this study transect where historical and continuous active layer and permafrost measurements are either limited or non-existent. Readily accessible data in this study, such as climatological data (collected from Environment Canada weather stations) and soil physical data

(collected from field visits), permitted the use of simple permafrost tools to calculate local

105 potential permafrost distribution, and for this study, the Fs is most readily applicable. More comprehensive tools have been developed and are available, such as the temperature at the top of permafrost (TTOP) model (Henry and Smith, 2001; Smith and Riseborough, 2002) and the

Kudryavtsev equation (Anisimov et al., 1997), however, these tools require significantly more data inputs and field measurements that limits the applications.

4.3 Methodology

4.3.1 Study locations

Five study locations were selected along a geographical north-to-south transect (Figure 4-

1) that begins near the northern tip of the Arctic Archipelago at Alert, Nunavut (82°30’1” N;

62°20’37” W), and ends in the subarctic region of northern Ontario at Peawanuck, Ontario

(55°0’30” N; 85°25’20” W). As northern Canada is remotely inhabited with sparse weather monitoring stations, the following intermediate locations were selected based on geography within the study transect and available climate records at Rankin Inlet (62°48’35” N; 92°05’58”

W), Resolute Bay (74°41’51” N; 94°49’56” W) and Eureka (79°59’20” N; 85°56’30” W),

Nunavut.

106

Figure 4-1. Map of Study Location in northern Canada.

107

Baseline climate data at the five locations were collected from weather observation stations provided by Environment Canada’s National Climate Data and Information Archive

(Environment Canada, 2013). Mean annual air temperatures, precipitation and observations from field visits on soil properties and permafrost distributions for the five locations are presented in

Table 4-1. Continuous permafrost distribution at Peawanuck was observed during the 2007-2009 field visits (Tam, 2009).

Table 4-1. Description of Study Locations Locations Mean Annual Precipitation Soil Types Current Air Temperature (mm) Permafrost (°C) Distribution Alert, NU. -19.0 to -17.0 100 to 250 Sandy clay loam Continuous Eureka, NU. -20.0 to -18.0 50 to 150 Sandy clay loam Continuous Resolute Bay, NU. -17.0 to -15.0 150 to 250 Gravel and sandy Continuous clay loam Rankin Inlet, NU. -11.0 to -10.0 250 to 400 Gravel and sandy Continuous clay loam Peawanuck, ON. -4.0 to -2.0 450 to 550 Peat and silt Continuous

The longest available baseline climate record representing 50 years were obtained for

Alert and Resolute Bay, Nunavut, from 1961-2011. Peawanuck, Rankin Inlet, and Eureka had records spanning from 16, 29, and 49 years, respectively. The baseline climate record for

Peawanuck, Ontario, was the shortest record due to significant missing and unavailable data.

Climate data prior to 1986 were obtained from the abandoned community of Winisk, Ontario.

Winisk was destroyed by a flood event in 1986 which resulted in the relocation of the community approximately 30 km south, at present-day Peawanuck, Ontario (Tam, 2009).

108

4.3.2 Global Climate Models and the Localizer Tool

Multi-model ensemble means produced from GCMs in the Localizer Tool were applied to project future climate conditions using the IPCC AR4 emission scenario experiments: A1B,

A2, and B1. The Localizer Tool selected the closest weather observation station with at least

70% data availability (UTSC, 2013). At Peawanuck, Ontario, there was less than 70% of the climate data availability; the Localizer Tool selected the nearest geographical station to satisfy this criteria. Manitouage, Ontario, (49°8’0” N; 85°50’0” W) was selected; however, as this location was not representative of the Hudson Bay Lowlands and located further south, the

Localizer Tool was manually shifted to Moosonee, Ontario, (51°16’20” N; 80°38’35” W) which is situated along the south-western shores of James Bay. The ensemble of models produced mean air temperatures for the baseline period of 1971-2000, and the three future periods of 2011-2040,

2041-2070, and 2071-2100. The multi-model ensemble results were interpolated to a common resolution and grid projection of 200 km by 200 km for comparison (UTSC, 2013). Future projections were produced by applying the change in mean air temperature between the model baseline and projection values, and then adding the change to the observed baseline values

(UTSC, 2013). The results of the projections were applied to assess climate change impacts to the climatological conditions for permafrost distributions based on the threshold values along the five study locations using the Fs.

109

4.3.3 Frost Number and Stefan Frost Number

The Frost Number (F) index is a dimensionless ratio, shown in Nelson and Outcalt

(1987), between the square root function of the Freezing and Thawing Degree-Day Indices (FDD and TDD, respectively).

FDD0.5 F  ( ) . (4-1) TDD0.5  FDD0.5

TDD was calculated as an annual accumulation of the daily temperature above the 0ºC threshold temperature with units in ºC·days. For FDD, the daily temperature below the 0ºC threshold was applied and expressed as a positive value. These degree-days techniques have been applied in engineering applications of relating climate conditions with ground freezing and thawing actions

(French, 2007). The F threshold value for continuous permafrost is F ≥ 0.67; discontinuous permafrost is F < 0.67; F < 0.6 for sporadic; and, F < 0.5 for permafrost absent (Nelson and

Outcalt, 1987).

The Stefan Frost Number (Fs) is a dimensionless ratio, similar to F in formulation, for determining potential permafrost distributions for a location by considering the subsurface information using the Stefan equation with degree-days accumulations (Nelson and Outcalt,

1987; Anisimov et al., 1997).

(4-2)

110

Where the Stefan equation (4-3) provides a useful method for incorporation the subsurface information, particularly the depths (X) of thawing (t) and freezing (f) in soils, inherently the soil thermal conductivities (K, in units W/(m·k)) in respective frozen and unfrozen (u) states (Nelson et al., 1997; Woo et al., 2004).

2K  DD 2K  DD X  ( )0.5  ( )0.5 (4-3) QL L   

Where QL = L·ω·ρ; the volumetric latent heat of soil composed of: L, latent heat of fusion of ice

(J/kg); ω is the moisture content (%); and ρ is the soil bulk density (kg/m3). DD represents the

Degree-Day Index, in which FDD and TDD are applied.

As shown in Tam et al. (2014a), the Fs can be simplified to solely consider the climatic component and the freezing and thawing soil thermal conductivity ratio value. The ratio of the frozen and unfrozen soil thermal conductivities (Kf/Ku) modifies the FDD accumulations.

FDD  K f / Ku  Fs  ( ) . (4-4) FDD  K f / Ku   TDD

The Fs thresholds are the same as the F thresholds (Nelson and Outcalt, 1987).

111

4.3.4 Site-specific conditions

In order to apply the Fs, the ratios between the frozen and unfrozen soil thermal conductivities were calculated for the study locations. The mean frozen and unfrozen soil thermal conductivities were applied and calculated based on the observed soil types and layers

(Tam et al., 2014a). Field measurements to collect soil data were conducted for Peawanuck,

Ontario, from 2006 to 2008; Eureka, Nunavut in 2011; and, Alert, Nunavut, in 2011 and 2012.

At the northern most location, Alert, Nunavut, the ground material was composed of sandy clay loam to a depth of 50 to 100 cm where fractured shale bedrock was observed during site excavations at 90 to 100 cm depth (Defence Construction Canada, 2010a); the Kf /Ku (2.19 / 1.87

W/(m·k)) ratio was 1.17. Silt and clay are the dominant material at Eureka, Nunavut, where the upper layer consisted of sandy clay loam (Defence Construction Canada, 2010b); the Kf/Ku (2.15

/ 1.82 W/(m·k)) ratio was 1.18. At Resolute Bay, Nunavut, vegetation becomes sparsely distributed on the surface, where gravel and sand become dominant in the upper 30 to 50 cm, followed by sandy clay loam and silt ground materials (Department of Economic Development and Transportation, 2009); the Kf/Ku (1.91 / 1.33 W/(m·k)) ratio was 1.44. For Rankin Inlet,

Nunavut, the surface layer is vegetated with grasses and lichen, typical of a tundra ecosystem

(Nuna Burnside Engineering and Environmental Ltd., 2010). Although the organic material layer within this Low Arctic Region is not as extensive as the Hudson Bay Lowlands, the organic presence does influence the soil thermal conductivity and soil moisture content (Tam, 2009). At

Rankin Inlet, soil profiles were dominant with gravel along the Hudson Bay shores with sandy clay loam; the Kf/Ku (1.79 / 1.38 W/(m·k)) ratio was 1.30. Site conditions for Peawanuck,

Ontario, consists of an upper peat layer that ranges from 10-30 cm in thickness followed by silty,

112

clayey and gravelly ground material; the calculated Kf/Ku (0.92 / 0.58 W/(m·k)) ratio was 1.59.

Wetlands and bogs dominate the Hudson Bay Lowlands landscape to provide an abundance of organic material content that can enhance the regional soil thermal conductivity. Along the shore of the Hudson Bay, there is greater presence of sand and gravel (Tam, 2009).

4.3.5 Regression analysis

A statistical software package, MINITAB 14, was applied to conduct various statistical analyses (Minitab, 2003). Probability density plots using 95% confidence intervals were scrutinized visually to determine normality of the baseline MAAT data for the study locations; normal distributed data were considered for the five locations. For determining autocorrelation of the time series for MAAT, the software was set to a 5% significance limit with default lag settings, where the numbers of lags were automatically determined by default settings. From the autocorrelation analysis of the baseline MAAT, the first two lags for Rankin Inlet (of 8 lags) and

Alert (of 14 lags) were observed to be significant with the remainder being not significant; the first three lags in Eureka (of 13 lags) were significant; only lag number 4 (of 13 lags) for

Resolute Bay was significant; and there were no significance at Peawanuck (of 7 lags).

Regression analysis of the MAAT was then conducted to determine the rate of temperature change for specific periods. The model Localizer Tool produced an ensemble MAAT for the period of 1971 to 2000; a difference comparison was conducted with the observed MAAT to assess performance and validation of the Localizer Tool. As climate baseline ranged from 1961-

2011 for this study, specific periods were selected for analysis. For Alert, Eureka, and Resolute

Bay, the period of 1971 to 2000 was chosen for both the Localizer Tool and observed MAAT to

113 be assessed. For Rankin Inlet and Peawanuck, incomplete MAAT data resulted in the comparison using the period of 1982-2000 and 1971-1990, respectively.

4.4 Results

4.4.1 Climate baseline 1961-2011

Model performance was assessed by comparing MAAT from the five study locations between the historical observational climate data with the model outputs for the same periods

(Figure 4-2a). Overall, the model output from the Localizer Tool was capable of reproducing historical climate data with mean differences ranging from -0.1 to -0.2°C in the Arctic region and the subarctic region in northern Ontario (Table 4-2). At the Arctic locations, climate-warming trends were observed up to a maximum of +0.339°C/year, at Rankin Inlet, Resolute Bay, and

Alert within the last 50 years of record (Table 4-2). Slight climate-cooling trends from -0.017 to -

0.061°C/year were observed in the subarctic region (Table 4-2).

114

Table 4-2. Summary of Climate Baseline Analysis Temp. Model Change Rate Observed Model Difference Location Period (°C/year) R-sq. MAAT (°C) MAAT (°C) (°C) Alert, Nu. 1961-2000 +0.022 0.09 -17.9 - - 1971-2000 +0.046 0.22 -17.9 -18.0 -0.1 2000-2011 +0.339 0.69 -16.7 - - Eureka, Nu. 1961-1992 -0.007 0.00 -19.8 - - 1961-2010 +0.121 0.36 -19.1 - - 1971-2000 +0.072 0.28 -19.6 -19.7 -0.1 Resolute Bay, 1961-1997 +0.016 0.04 -16.5 - - Nu. 1971-2000 +0.339 0.17 -16.3 -16.4 -0.1 1997-2011 +0.167 0.28 -14.8 - - Rankin Inlet, 1982-1996 +0.109 0.32 -11.3 - - Nu. 1982-2000 +0.339 0.62 -11.6 -11.8 -0.2 1996-2011 +0.137 0.14 -6.2 - - Peawanuck, 1961-2009 -0.017 0.02 -3.1 - - On. 1971-1990 -0.051 0.02 -0.9 -1.1 -0.2 1995-2009 -0.061 0.03 -3.7 - -

4.4.2 Future climate scenarios 2011-2100

Projections were produced by the Localizer Tool for Alert, Eureka, Resolute Bay, Rankin

Inlet, and Peawanuck under emission scenarios A1B, A2, and B1 (Figure 4-2b). The ‘bias- corrected’ projections applied local climate change rates to the observed baseline temperatures from 1971-2000 (Fenech, 2009; UTSC, 2013). As shown in Figure 4-2b, climate warming is projected under all emission scenarios, with the greatest amount of warming to be experienced under scenario A2 high emissions of atmospheric [CO2] for all study locations. Overall for the study transect, climate change projections indicate warming above the 1971-2000 baseline by a minimum of +1.5°C and a maximum of +2.4°C for the period 2011-2040; +2.6 to +4.1°C for

2041-2070; and +3.3 to +7.1°C for 2071-2100.

115

Figure 4-2. a) Baseline climate data for all five study locations from local observation weather stations; b) Projected MAAT for A1B, A2 and B1 emission scenarios from 2011 to 2100; Stefan

Frost Number results for permafrost potential from baseline and projected MAAT under A1B

(c), A2 (d), and B1 (e) emission scenarios.

116

4.4.2.1 Alert, Nunavut

Beginning at the northernmost location at Alert, Nunavut, climate scenarios A1B, A2, and B1, for mid-century (2041-2070) projects change rates of +4.0°C (Figure 4-3a), +3.7°C

(Figure 4-3b), and +3.0°C (Figure 4-3c), respectively. By 2100, A1B projects climate change of

+5.8°C, +6.6°C under A2 scenario, and +3.9°C under B1 scenario, with MAAT for Alert expected to reach -14.1 to -11°C.

4.4.2.2 Eureka, Nunavut

At Eureka, Nunavut, projected mid-century change rates under A1B, A2, and B1 scenarios are expected to be +3.8°C (Figure 4-3d), +3.7°C (Figure 4-3e), and +2.9°C (Figure 4-

3f), respectively. By the end of the century at Eureka, A1B scenario projects a change of +5.5°C,

+6.5°C for A2 scenario, and +3.7°C for B1 scenario, with MAAT projected to range from -16 to

-13.2°C.

117

Figure 4-3. Projection of MAAT change rates for the period of 2041-2070 for Alert, Nunavut

(left, a-c) and Eureka, Nunavut (right, d-f), under IPCC A1B, A2, and B1 (top to bottom) emission scenarios. The ‘+’ shows the locations of the study sites.

118

4.4.2.3 Resolute Bay, Nunavut

For Resolute Bay, Nunavut, the projected mid-century change rates under A1B, A2, and

B1 scenarios are +4.0°C (Figure 4-4a), +3.9°C (Figure 4-4b), and +3.0°C (Figure 4-4c), respectively. By 2100, A1B scenario projects a change of +5.9°C under A1B, +7.1°C under A2, and +3.9°C under B1, with MAAT ranging from -12.5 to -9.3°C.

4.4.2.4 Rankin Inlet, Nunavut

At Rankin Inlet, Nunavut, mid-century projection rates under A1B, A2, and B1 scenarios are +3.6°C (Figure 4-4d), +3.4°C (Figure 4-4e), and +2.7°C (Figure 4-4f), respectively. By the end of the century for Rankin Inlet, A1B scenario projects a change of +5.3°C, +6.1°C under A2, and +3.5°C under B1, with MAAT from -8.3 to -5.7 °C.

4.4.2.5 Peawanuck, Ontario

At the southernmost location within the subarctic region of northern Ontario at

Peawanuck, Ontario, mid-century change under A1B, A2, and B1 scenarios are +3.5°C (Figure

4-5a), +3.2°C (Figure 4-5b), and +2.6°C (Figure 4-5c), respectively. By the end of this century,

Peawanuck is projected for +4.6°C in the A1B scenario, +5.3°C in the A2 scenario, and +3.3°C in the B1 scenario, with MAAT ranging from +0.2 to +2.2°C.

119

Figure 4-4. Projection of MAAT change rates for the period of 2041-2070 for Resolute Bay,

Nunavut (left, a-c) and Rankin Inlet, Nunavut (right, d-f), under IPCC A1B, A2, and B1 (top to bottom) emission scenarios. The ‘+’ shows the locations of the study sites.

120

121

4.4.3 Stefan Frost Numbers 1971-2100

The Fs values were calculated from the observation period of 1971-2000 and then from the ‘bias-corrected’ projections for the periods of 2011- 2040, 2041-2070, and 2071-2011 under the A1B (Figure 4-2c), A2 (Figure 4-2d) and B1 (Figure 4-2e) emission scenarios. From the baseline study from 1971-2000, climate conditions for continuous permafrost distributions were favourable for Alert, Eureka, Resolute Bay, and Rankin Inlet with Fs values of 0.86, 0.83, 0.85, and 0.69, respectively. At Peawanuck, current climate conditions were favourable for discontinuous permafrost with Fs at 0.62 although site observations indicate continuous permafrost; this issue is further discussion below regarding a possible time lag response between climate and permafrost distribution change. By the end of the century, the projected Fs values for

Alert, Eureka and Resolute Bay under the A1B, A2, and B1 emission scenarios are projected to range from 0.71 (A2) to 0.78 (B1) providing climate conditions suitable for maintaining continuous permafrost distributions. For Rankin Inlet, changes in the climate conditions are projected to occur within the 2011-2040 period where the projected Fs values of 0.65 (A1B) to

0.66 (A2, B1) suggest unfavourable conditions to continuous permafrost but favouring discontinuous permafrost. By the end of the century for Rankin Inlet, Fs values are projected to range from 0.58 (A2) to 0.62 (B1), maintaining the climate condition for discontinuous and possibly sporadic permafrost as the Fs value for sporadic threshold is 0.60. At Peawanuck,

Ontario, the projected Fs values, within the near future period of 2011-2040, are projected to show climate conditions that are favourable for sporadic permafrost at 0.58 (A1B) to 0.59 (A2,

B1). By 2100, climate conditions further maintain sporadic permafrost with projections at

Peawanuck with Fs values at 0.51 (A2) to 0.55 (B1).

122

4.5 Discussion

4.5.1 Future climate conditions for permafrost distributions

The findings of this work address our research question. Firstly, the multi-model ensemble means produced from GCMs in the Localizer Tool were compared to the baseline climate of 1971-2000. As demonstrated, there were minimal discrepancies between the observed and modelled MAATs which provide favourable support for the GCMs applied at four of the five study locations. The greatest discrepancy was shown for Peawanuck, which could be attributed to the poor climate data set containing many missing values and that it is possible for current climate conditions to have shifted away from the baseline conditions applied in the models. This change in climate conditions may suggest a lag time response in permafrost distribution shift, a possible explanation for the presence of relict continuous permafrost being observed during field visits while Fs values indicate climate conditions favouring discontinuous permafrost. The quality of data is further discussed as other locations had better climate records containing little to no data gaps, such as at Alert, Eureka, and Resolute Bay weather stations.

After GCM validation, future climate projections were produced for three future time periods under three emission scenario experiments: A1B, A2, and B1. Since the beginning of the 21st century, the MAAT has been increasing in the Canadian North as shown in Table 2, with the exception of Peawanuck, Ontario, were climate records indicate very slight cooling trends.

Overall, greater climate change rates for the Canadian High Arctic are projected under the A1B moderate emission scenario for the beginning and middle century periods; by the end of this

123 century at 2100, A2 high emissions scenario projects the greatest climate change rate. B1 scenario results projected less intense MAAT increases; however, all locations were projected to increase in air temperatures. For the High Arctic locations at Alert, Eureka, and Resolute Bay,

Nunavut, by the end of 2040, the MAAT is expected to increase by +1.7 (A2) to +2.4°C (A1B); by mid-century, a further increase of +3.0 (B1) to +4.1°C (A1B); and by 2100, a further projected increase by +3.7 (B1) to +7.1°C (A2). Further south at Rankin Inlet, Nunavut, similar increases in MAAT are projected; however, the A2 scenario projects the greatest temperature increase with +6.1°C by the end of 2100. For the southernmost location of the study transect, at

Peawanuck, Ontario, the MAAT at the end of 2040 is projected to increases by +1.6°C (A1B,

B1) and by the end of 2100, +5.3°C (A2) from baseline. As this study applied the emission scenarios from the IPCC 4th Assessment Report in the Localizer Tool, we acknowledge that the final draft Report of the Working Group I contribution to the IPCC 5th Assessment Report

"Climate Change 2013: The Physical Science Basis" was recently released in fall 2013 with the final report being expected to be released in 2014. The IPCC 4th Assessment Report relied on projected emission scenarios of atmospheric [CO2]. The IPCC 5th Assessment Report applies projected scenarios using RCP radiative forcing targets, such as RCP2.6, RCP4.5, RCP6.0 and

RCP8.5 scenarios where the values in these scenarios represent the respective radiative forcing in

W/m2 (Slater and Lawrence, 2013). The radiative forcing values of the RCPs also account for future emitted greenhouse concentrations (Slater and Lawrence, 2013). As both assessment reports have indicated future warming conditions for the arctic, the projections produced and applied in this study, based on the 4th assessment report, remain a rational and relevant approach for assessing climate change impacts (Rinke et al., 2012).

124

Applying these results to potential permafrost distributions, the Stefan Frost Number equation provided an assessment of the climate conditions based on the FDD and TDD accumulations weighted by the site-specific soil thermal properties at each location, using each emission scenario, from the 2011 to 2100. Based on the baseline climate data from 1971-2000, climate conditions were favourable for continuous permafrost at Alert, Eureka, Resolute Bay, and Rankin Inlet; climate conditions were favourable for discontinuous permafrost at

Peawanuck. The Fs results based on the projected climate conditions suggest that the continuous permafrost distributions in the High Arctic will be maintained at the three High Arctic locations at Alert, Eureka, and Resolute Bay for the remainder of this century by the continuous low temperatures that will provide sufficient FDD accumulations for permafrost formation. For

Rankin Inlet, climate conditions favourable for continuous permafrost distributions are projected to change by the end of 2040 with all three scenarios, A1B, A2, and B1, for Fs indicating discontinuous and possibly sporadic permafrost. This projection is consistent with the increase in

MAAT, particularly, with the projected increase in TDD accumulations with a concurrent decrease in FDD. In the subarctic region at Peawanuck, Fs results project climate conditions favourable for maintaining sporadic permafrost by the end of 2040 and until the end of this century. For permafrost at Peawanuck, an important factor to consider is the role of the soil thermal conductivity (K). The presence of organic peat materials within the Hudson Bay

Lowlands can provide thermal insulating effects, reduced K values under dry summer conditions, or enhanced freezing conditions with increased K values, during wet fall/winter seasons, which may support the presence of permafrost even under changing climate conditions (Tam, 2009).

This study also acknowledges that changes to the soil thermal properties, such as the frozen and unfrozen states of soil, and soil moisture content can affect the results produced from the Fs, and may contribute to a lag time response between the climate conditions and shifts in permafrost

125 distribution (Anisimov et al., 1997). In addition, a reduction in the ratio between the frozen and unfrozen soil thermal conductivities may increase unfavourable conditions for permafrost and hasten the change in distribution; alternatively, an increase in the ratio can favour the development or persistence of permafrost (Côté and Konrad, 2005; Duan and Naterer, 2009).

However, the scope of this study focuses on the climatological conditions; additional research for the timing of permafrost distribution shifts and lag time changes are recommended.

4.5.2 Limitations and errors

Mean air temperatures were projected from an ensemble of GCMs using the Localizer

Tool for this study. The Localizer Tool compared the model output results of the baseline period with available observation data from the local weather station records provided by Environment

Canada to correct for any local biases prior to conducting the future projections (UTSC, 2013). A difference analysis was performed to determine model performance of the Localizer Tool that was shown in Table 4-2. For the four Arctic locations, the model and observed MAAT demonstrated little differences; a possible explanation is the data quality was higher and available with most data periods beginning from 1961 to present day, with the exception of

Rankin Inlet where the climate data is only available from 1982 to present day. This study highlights the importance of data quality and data availability at observation stations that can affect climate-modeling performance. For the subarctic Region at Peawanuck, Ontario, poor data quality and gaps within the observed weather record (>70%) triggered the Localizer Tool to automatically select a further southern location. To correct for this selection, an alternative location at Moosonee, Ontario, was manually chosen based on similar geography, distance to

126

Peawanuck, and >70% data availability. For the baseline climate data at Peawanuck, the available climate records of Winisk and Peawanuck, Ontario, were combined to represent a longer period. Prior to a community relocation in 1986 caused by severe flooding, climate data in the area was collected approximately 30 km north at the now abandoned location of Winisk,

Ontario. Data quality of the Winisk and Peawanuck records were not completed and contained decades of gaps that could have created variation in the regression analysis and interpolation

(Shiklomanov et al., 2007); for example, only the following periods contained completed yearly data at Winisk: 1959-1964 and 1969-1977, and for Peawanuck: 1994-2011. Finally, this study further indicates that current climate conditions are changing and that a lag time response in permafrost distributions may be evident at Peawanuck, where “relict” continuous permafrost is being observed while climate conditions have already shifted to favour discontinuous permafrost.

4.6 Conclusion

The application of the Stefan Frost Number equation with projections from multi-model ensembles of GCMs under three emission scenarios was capable of providing insight into the future permafrost potentials within the study transect. Climate warming trends were projected under IPCC A1B, A2, and B1 scenarios for all locations. The greatest warming change from baseline temperatures projected by 2100, under the A2 scenario, for the Canadian High Arctic is expected at Resolute Bay at +7.1°C; for the Low Arctic at Rankin Inlet, +6.1°C; and, for the subarctic regions of northern Ontario at Peawanuck, +5.3°C. For the High Arctic at Resolute

Bay, Eureka and Alert, Nunavut, climate conditions are expected to remain favourable in

127 maintaining the continuous permafrost distributions. At Rankin Inlet, Nunavut, changes to the climate conditions by the end of this century projects a shift towards discontinuous and possibly to sporadic permafrost distributions. For Peawanuck, Ontario, climate conditions by 2100 are expected to remain favourable in supporting sporadic permafrost.

4.7 Acknowledgements

Portions of this study was funded and supported by the Wildlife Research and

Development Section of the Ontario Ministry of Natural Resources and the Department of

Physical and Environmental Sciences at the University of Toronto Scarborough.

128

4.8 References

Anisimov, OA., and Nelson, FE. 1996. Permafrost distribution in the Northern Hemisphere under scenarios of climatic change. Global and Planetary Change 14, 59-72.

Anisimov, OA., Shiklomanov, NI., and Nelson, FE. 1997. Global warming and active-layer thickness: results from transient general circulation models. Global and Planetary Change 15, 61-77.

Burn, CR., and Smith, CAS. 1988. Observation of the “Thermal Offset” in Near- Surface Mean Annual Ground Temperatures at Several Sites near Mayo, Yukon Territory, Canada. Arctic 41, 91-166.

Côté, J., and Konrad, J. 2005. A generalized thermal conductivity model for soils and construction materials. Canadian Geotechnical Journal 42, 443-458.

Defence Construction Canada. 2010a. Operation and Maintenance Plan CFS Alert (ALT), Nunavut. Nunavut Water Board Public Registry. Accessed online: 01 December 2012 from: ftp://nunavutwaterboard.org/1%20PRUC/3%20MUNICIPAL/3B/ 3BC%20-%20Camp/3BC-ALT1015%20(Alert)/.

Defence Construction Canada. 2010b. Operation and Maintenance Plan CFS Eureka (ERK), Nunavut. Nunavut Water Board Public Registry. Accessed online: 01 December 2012 from: ftp://nunavutwaterboard.org/1%20PRUC/3%20MUNICIPAL/3B/ 3BC%20-%20Camp/3BC-ERK1015%20(DND%20Eureka)/.

Department of Economic Development and Transportation. 2009. Resolute Bay Airport Operation and Maintenance Manual Sewage Treatment Facility. Nunavut Water Board Public Registry. Accessed online: 01 December 2012 from: ftp://nunavutwaterboard.org/1%20PRUC/3%20MUNICIPAL/3B/ 3BM%20-%20Municipality/3BM-YRB0308/.

Dobinski, W. 2011. Permafrost. Earth-Science Reviews 108, 158-169.

Duan, X., and Naterer, G. 2009. Heat conduction with season freezing and thawing in an active layer near a tower foundation. International Journal of Heat and Mass Transfer 52, 2068-2078.

Environment Canada. 2013. Climate Data Online - Canada’s National Climate Archive. Government of Canada. Accessed 01 April 2013 from: http://climate.weatheroffice.gc.ca/climateData/canada_e.html.

129

Fenech, A. 2009. Rapid assessment of the impacts of climate change (RAICC): An integrated approach to understanding climate change in the Halton Region of Ontario, Canada. Doctor’s Thesis, University of Toronto, Department of Geography.

French, H. 2007. The Periglacial Environment. (3rd Ed.). John Wiley & Sons: West Sussex.

French, H. 2008. Recent Contributions to the Study of Past Permafrost. Permafrost and Periglacial Processes 19, 179-194.

Gough, WA., and Leung, A. 2002. Nature and fate of Hudson Bay permafrost. Regional Environmental Change 2, 177-184.

Kujala, K., Seppälä, M., and Holappa, T. 2008. Physical properties of peat and palsa formation. Cold Regions Science and Technology 52, 408-414.

Lewis, T., and Lamoureux, SF. 2010. Twenty-first century discharge and sediment yield predictions in a small high Arctic watershed. Global and Planetary Change 71, 27-41.

Minitab (2003). Minitab Release 14 Statistical Software. Minitab Inc. Accessed 01 Nov 2013. http://www.minitab.com/.

Nelson, F., and Outcalt, S. 1987. A Computational Method for Prediction and Regionalization of Permafrost. Arctic and Alpine Research 19, 279-288.

Nixon, J., and McRoberts, E. 1973. A study of factors affecting the thawing of frozen soils. Can. Geotech. J. 20, 439-452.

Nuna Burnside Engineering and Environmental Ltd. 2010. Water Supply Facility Operation and Maintenance (O&M) Plan, Hamlet of Rankin Inlet. Nunavut Water Board Public Registry. Accessed online: 01 May 2013 from: ftp://nunavutwaterboard.org/1%20PRUC/3%20MUNICIPAL/3B/ 3BM%20-%20Municipality/3BM-GRA0207/.

130

Shiklomanov, NI., Anisimov, OA., Zhang, T., Marchenko, S., Nelson, FE., and Oelke, C. 2007. Comparison of model-produced active layer fields: Results from northern Alaska. Journal of Geophysical Research 112. F02S10, DOI: 10.1029/2006JF000571.

Shur, Y., and Jorgenson, M. 2007. Patterns of Permafrost Formation and Degradation in Relation to Climate and Ecosystems. Permafrost and Periglacial Process 18, 7–19.

Slater, AG., and Lawrence, DM. 2013. Diagnosing Present and Future Permafrost from Climate Models. Journal of Climate 26, 5608-5623.

Smith, SL., Burgess, MM., Riseborough D., and Nixon, FM. 2005. Recent Trends from Canadian Permafrost Thermal Monitoring Network Sites. Permafrost and Periglacial Process 16, 19-30.

Solomon, S., Qin, D., Manning, M., Alley, RB., Berntsen, T., Bindoff, NL., Chen, Z., Chidthaisong, A., Gregory, JM., Hegerl, GC., Heimann, M., Hewitson, B., Hoskins, BJ., Joos, F., Jouzel, J., Kattsov, V., Lohmann, U., Matsuno, T., Molina, M., Nicholls, N., Overpeck, J., Raga, G., Ramaswamy, V., Ren, J., Rusticucci, M., Somerville, R., Stocker, TF., Whetton, P., Wood, RA., Wratt, D. 2007. Technical Summary. In: Solomon, S., Qin, D., Manning, M., Chen, Z., Marquis, M., Averyt, KB., Tignor, M., Miller, HL. (eds). Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA.

Tam, A. 2009. Permafrost in Canada’s Subarctic Region of Northern Ontario. Master’s Thesis, University of Toronto, Department of Geography, Toronto.

Tam, A., Gough, and WA., Xie, C. 2014a. An application of the Stefan Frost Number and XG- Algorithm in the Canadian Subarctic and Arctic Regions from 2004 to 2011. Theoretical and Applied Climatology (submitted).

Tam, A., Gough, WA., and Xie, C. 2014b. The fate of Hudson Bay Lowlands palsas in a changing climate. Arctic, Antarctic, and Alpine Research (accepted)

TGICA. 2007. General Guidelines on the use of Scenario Data for Climate Change Adaptation Assessment, Version 2. Intergovernmental Panel on Climate Change. Accessed online: 01 May 2013 from: http://www.ipcc-data.org/guidelines/.

131

UTSC. 2013. Localizer. University of Toronto Scarborough. Accessed online: 22 May 2013 from: http://climatechange.utsc.utoronto.ca/.

Waller, RI., Murton, JB., and Kristensen, L. 2012. Glacier-permafrost interactions: Process, products and glaciological implications. Sedimentary Geology 255-256, 1-28.

Wang, G., Hongchang, H., and Taibin, L. 2009. The influence of freeze-thaw cycles of active soil layer on surface runoff in a permafrost watershed. Hydrology 375, 438-449.

Woo, MK., Arain, MA., and Mollinga, M., Yi, S. 2004. A two-directional freeze and thaw algorithm for hydrologic and land surface modeling. Geophysical Research Letters 31. DOI: 10.1029/2004GL019475.

132

Chapter 5 The fate of Hudson Bay Lowlands palsas in a changing climate 5 Chapter 5

5.1 Abstract

The climatological conditions for the presence of palsas in the Hudson Bay Lowlands

(HBL) in Ontario Canada are examined using data from four climate stations, Big Trout Lake,

Lansdowne House, Peawanuck and Fort Severn. These stations sandwich the existing region where palsas occur. The criteria for the formation and occurrence of palsas were taken from the literature on Fennoscandian and neighboring Quebec palsas were applied to the HBL. Thermal thresholds set at -2oC and 0oC mean annual air temperature, and number of days below -10oC per year were met for the two more northerly locations; the two southerly locations were on the edge of the thresholds. Climate projections from two models under two emission scenarios for the

2020s, 2050s and 2080s indicated that by the 2080s all four locations would fail the -2oC threshold for palsa formation but at three locations the 0oC threshold for palsa presence was met for some projection scenarios. Over the next century, it is likely that the climate conditions will continue to be capable of supporting existing palsas; however, by the end of the century the threshold criteria for new palsa formation will not be met for most of the HBL.

133

5.2 Introduction

A palsa is a geomorphological formation that often appears as a dome-shaped circular mound that has a permafrost core and alternating layers of segregated ice lenses with frozen peat at the surface (van Everdingen, 1998; Pissart, 2002). The surface of the palsa consists of an active layer that experiences summer thawing and winter freezing. Palsas range in size from 0.5 to 12 m high, meters to tens of meters in diameter, and are located primarily in bogs within the subarctic regions of the world (Seppälä, 1986; Zuidhoff and Kolstrup, 2000; Lewkowicz and

Coultish, 2004; Vallée and Payette, 2007; Kujala et al., 2008; Thibault and Payette, 2009; Cyr and Payette, 2010; Saemundsson et al., 2012). Specifically, palsas are present in the northern hemisphere in areas such as northern Canada, Alaska, Iceland, Northern Scandinavia, and

Siberia (Seppälä, 1986; Zuidhoff and Kolstrup, 2000; Gurney, 2001, Hinkel et al., 2001, Kirpotin et al., 2009). In this work, we examined the behavior of palsas in the Hudson Bay Lowlands

(HBL) of northern Ontario, Canada, a region of palsas that has not been studied in detail (Tam,

2009) in the context of climate change.

Palsas are underlain primarily by discontinuous permafrost and provide surface evidence of the underlying permafrost (Seppälä, 1986; Kujala et al., 2008; Kirpotin et al., 2009).

Therefore, studying the state and location of palsas in fact provides an insight into the state and location of the permafrost itself which covers vast regions of the northern hemisphere (Luoto et al., 2004; Kujala et al., 2008; Tam, 2009). The literature indicates that the mean annual air temperature (MAAT) of -2oC or lower is necessary for palsa formation; however, they have been

134 observed at greater temperatures (Seppälä, 1986; Parviainen and Luoto, 2007; Kujala et al.,

2008). For Canada, the MAAT observed for palsa presence is 0 oC in northern Quebec (Cyr and

Payette, 2010). In the majority of locations, including the HBL, palsas are located at the southernmost limit of discontinuous permafrost where the -2oC palsa formation threshold is located, and are therefore very sensitive to any type of climatic changes.

The thermal thresholds reflect the thermodynamic heat loss necessary for the maintenance of the ice lens. In addition, the material covering the lens must have seasonally varying thermal conductivity to minimize heat gained in the warmer seasons and accentuate heat loss in the colder seasons. Peat provides the ideal ground cover for palsas in the HBL and elsewhere. Peat has a particularly low thermal conductivity when it is dry and unfrozen (0.37–

0.79 W/mK); the conductivity increases with moisture (0.46–1.38 W/mK) and dramatically increases when frozen (Gough and Leung, 2002; Kujala et al., 2008). Thus, a climate that has low precipitation in the summer and higher precipitation in advance of the winter season is ideal for palsa formation and this scenario tends to be the case in Fennoscandia (Parviainen and Luoto,

2007). In addition, the dome shape of the palsa enables wind to reduce the snow cover on palsas, hence minimizing the thermal insulation of snow cover allowing for more winter heat loss.

As a result of the considerations outlined above, other climate indicators have been found to be important. These include winter temperature (Seppälä, 1986) with a constraint of 120 days or more of the year lower than -10oC for palsas in Sweden. Furthermore, Parviainen and Luoto

(2007) suggest that 500 mm/year in precipitation with a relative minima during the summer and

135 winter months is optimum for palsa formation and occurrence. Other factors include July and

January temperatures, continentality, vegetation, hydrology, and thawing degrees days, however, we acknowledge that these measures are not necessarily independent.

In this paper we examine specific research questions including: 1) Can climate criteria for palsas in the HBL be determined by analyzing climate data from available weather stations at the northern and southern palsa limits? 2) If so, what is the projected fate of palsas in the HBL using climate change projections for the region as has been done elsewhere (e.g. Fronzek et al.

(2006))?

5.2.1 Study Area

The study area is situated within the HBL in a region that lies between Fort Severn and

Peawanuck, Ontario, near the Hudson Bay coast, and Big Trout Lake and Lansdowne House,

Ontario (Figure 5-1). Tam (2009) identified the location of palsas in the Peawanuck area and further south, and within 20 km south of Fort Severn. Two of the authors (Kowal & Xie) observed palsas in an aerial survey during the summer of 2011 (54o12’N, 88o54’W), 80 km northeast of Big Trout Lake. As reported by Tam (2009), this cross section of palsas is part of a wide band of palsas extending from the Manitoba border to James Bay. Big Trout Lake is 314 km southwest of Peawanuck and 290 km south-south-west from Fort Severn. Since the Big Trout

Lake / Fort Severn transect is roughly perpendicular to the palsa band, the palsa band is estimated to be just over 220 km in width.

136

Figure 5-1. The Study Area – Hudson Bay Lowlands of northern Ontario, Canada.

137

5.3 Methods

5.3.1 Climate Data

Climate data from Big Trout Lake (53o50’N, 89o52’W), Lansdowne House (52o14’N,

87o53’W), Peawanuck (54o59’N, 85o26’W) and Fort Severn (56o01’N, 87o35’W) were examined using palsa criteria developed for other palsa regions (Seppälä, 1986; Parviainen and Luoto,

2007; Kujala et al., 2008). Big Trout Lake and Lansdowne House are situated to the south of the

HBL palsa regions (Figure 5-1). Peawanuck and Fort Severn are at the northern extent of the palsa region, both located close to the Hudson Bay coast. Daily temperature (daily minimum, maximum, and mean) and precipitation data of varying temporal spans were available for all four locations (Big Trout Lake, 1951-2010; Lansdowne House, 1953-2010; Peawanuck, 1986-2010;

Fort Severn, 2006 -2010) (Environment Canada, 2012). However, due to significant data gaps in the minimum, maximum, and mean temperatures, the datasets from 1990, 1992-1994, 1996-

1997, 2006, and 2008-2010 could not be applied in this work. Projection data for Big Trout Lake and Lansdowne House were available using two climate models and statistical downscaling. For

Peawanuck and Fort Severn, climate model projection data are used directly as described below.

138

5.3.2 Data Analysis

Garnered from the palsa literature, the following threshold criteria for palsa formation

and occurrence were examined and compiled in Table 5-1 for mean annual air temperature ≤ -

2oC (Kujala et al., 2008); ≤ 0oC (Cyr and Payette, 2010); 120 days or more per year below -10oC

(Seppälä, 1986). The following optimum ranges were compiled: annual precipitation, 497 ± 78.8

mm; winter precipitation, 89.6 ± 27.1 mm; summer precipitation, 184 ± 16.4; July temperature,

11.1 ± 1.05 oC; January temperature, -15.8 ± 2.84 oC; Continentality, 26.9 ± 3.53 oC (Parviainen

and Luoto, 2007). Continentality was calculated as the difference between the mean temperature

of the warmest and coldest months.

Table 5-1. Criteria for palsas for the Hudson Bay Lowlands. Criterion MAAT Number Winter Summer Annual July January Continentality T (oC) of days < Prec. Prec. Prec. T (oC) T (oC) T (oC) -10oC/yr (mm) (mm) (mm) Thresholds - 120 89.6 ± 184.0 ± 497.0 ± 11.5 ± -15.8 ± 26.9 ± 3.53 and 2.0oC*; days/yr 27.1 16.4 78.8 1.05 2.84 Optimum 0oC** Ranges *,** indicates Mean Annual Air Temperature (MAAT) criteria at the -2oC and 0oC thresholds, respectively, have been met.

The HBL relevant palsa thresholds were used when examining the statistically

downscaled projection data for Big Trout Lake and Lansdowne House. These projections were

produced in a two-step process for Big Trout Lake and Lansdowne House. First, projection data

were obtained from coarse resolution climate models for grid boxes corresponding to their

139 locations. A statistical downscaling model (SDSM) was then used to statistically link locally collected observational data (1961-1990) to simulations for the same period (Wilby and Dawson,

2012). This enabled the linking of larger scale climate variables with their local manifestation.

These relationships were then used to downscale the coarse resolution climate projections to these localities, providing better spatial and temporal resolution. This technique has two steps; the first is a correlation analysis between site-specific observations and large scale flow taken from globally-gridded atmospheric and surface observations. For each location, the relevant large-scale flow variables are used to build a statistical model of the local conditions. In the second step, projection data of the large scale flow is used to reverse the process and generate a local projection for the given location. A weather generator is also used to achieve the required temporal resolution. This latter technique was not applied to Peawanuck and Fort Severn due to a lack of climatological baseline data for the 1961 – 1990 period. The climate models used were

CGCM2 and HadCM3 with resolutions at 3.75° by 3.75°and 3.75° by 2.5° latitude and longitude, respectively. The former (CGCM2) is the Canadian coupled climate model (Flato and

Boer, 2001; Kim et al., 2002, 2003) and the latter (HadCM3) is the British Hadley Centre model

(Collins et al., 2001). Two emission scenarios from the Intergovernmental Panel on Climate

Change’s (IPCC) Special Report on Emissions Scenarios (SRES) were used, A2 and B2

(Nakićenović, 2000; Nakićenović and Swart, 2000). A2 represents a world of rapid economic growth, increasing population, regionalization of economic and environmental policy and often referred to as “business as usual” with high emissions of atmospheric carbon dioxide reaching

860 ppm by 2100 (Lewis and Lamoureux, 2010). B2 is also regional with a slower but steady growth in population, however with more emphasis on environmental policy designed to reduce greenhouse gas emissions; in this scenario, carbon dioxide concentrations reaches 600 ppm by

2100 (Gough and Leung, 2002). As presented below the downscaling worked well for

140 temperature but was not successful for precipitation and thus not used for climate projection of precipitation. For the purposes of model evaluation, the climate model output was used directly.

For Peawanuck and Fort Severn, in the absence of downscaled results, we have also used projections from the climate models directly, acknowledging these have coarse resolution and that the results therefore cover a broader area and were not linked to surface observations. These were done for the A2 and B2 emission scenarios.

5.4 Results

5.4.1 Climate data analysis

The historical trends in temperature are presented in Figure 5-2 for Big Trout Lake and are representative of the HBL region located between Fort Severn and Lansdowne House. The averaged minimum daily temperature (Tmin), the averaged maximum daily temperature (Tmax) and the averaged mean daily temperature (Tmean) are shown in Figure 5-2 with increasing trends. Seasonally (not shown), in all seasons except for the fall are warming similarly to the results of Gagnon and Gough (2002).

141

Figure 5-2. Temperature trends for Big Trout Lake for the period of 1951 – 2010; due to incomplete data, the following years could not be applied: 1990, 1992-4, 1996-7, 2006, and 2008-10.

Table 5-2 summarizes the climate data analysis examining palsa climate ranges. The -2oC and 0oC thresholds were met for the HBL palsas as was the 120 days below -10oC per year threshold. However, for Lansdowne House and Big Trout Lake only the winter precipitation optimum range were met; there was insufficient precipitation data for Peawanuck and Fort

Severn to make a determination. In addition, the July temperature optimum range was met in

Fort Severn, but the three other locations exceeded this range. Otherwise, the HBL tends to have

142

considerably more precipitation, greater continentality, with warmer temperatures in July and

colder in January.

Table 5-2. Evaluation of four weather/climate stations in the HBL with respect to identified criteria for palsas. Lansdowne House and Big Trout Lake are located south of the southern extent of palsas whereas Peawanuck and Fort Severn are located along the Hudson Bay coast within 20 km of the observed palsas (Tam, 2009). Location MAAT Number Winter Summer Annual July January Continentality T (oC) of days < - Prec. Prec. (mm) Prec. T (oC) T (oC) T (oC) 10oC/yr (mm) (mm) Fort Severn -4.6*,** 125A N/A N/A N/A 12.4 A -22.9 35.3 Peawanuck -3.7*,** 119 N/A N/A N/A 14.0 -22.9 36.9

Big Trout -2.7*,** 120A 72.4 A 253.0 609.1 16.2 -23.7 39.9 Lake Lansdowne -1.3** 112 80.5 A 291.0 699.5 17.2 -22.3 39.5 House

*,** indicates that criteria (within one standard deviation) has been met. A indicates the palsa threshold has been met, in reference to Table 5-1.

We note that although Peawanuck and Fort Severn are located at the northern edge of the

palsa occurrence within the HBL in northern Ontario, it may be incorrect to use climate data

from these stations as indicators of the northward extent of potential palsa development. In this

instance the physical boundary of land and sea (Hudson Bay) prevents a potentially more

northerly extent of palsas and we note that palsas do exist further north in Quebec. We observe

that the HBL sites exhibit much greater continentality than does Fennoscandia with slightly

higher values of continentality for the in-land sites (Lansdowne House and Big Trout Lake).

Winter extrema are similar for all four sites which is a reflection of the complete ice cover for

143

Hudson Bay during January. The marine influence on temperature is more clearly seen in the

July temperatures for Peawanuck and Fort Severn.

5.4.2 Climate projections

Downscaled climate projections were produced for Big Trout Lake and Lansdowne

House using two climate models which performed well in comparison to the observed temperature for the climatic baseline period of 1961-1990 (Table 3). The difference between modeled MAAT results and observations was 0.2oC or less for both locations. The modeled precipitation tends to be scattered on both sides of the observed values for both of the locations with CGCM2 producing higher values and HadCM3 producing lower values for the annual precipitation. Winter precipitation was over simulated for both models and summer precipitation was under-simulated. As a result of this lack of consistency in reproducing the observed climate, we will focus on the temperature thresholds for palsa formation and occurrence, the MAATs, and number of days below -10oC per year.

144

Table 5-3. Observed climate and downscaled climate projections for Big Trout Lake and Lansdowne House. Model MAAT Number of Winter Summer Annual July January Continentality T (oC) days < - Prec. Prec. Prec. T (oC) T (oC) T (oC) 10oC/yr (mm) (mm) (mm) Big Trout Lake Observed -2.7*,** 120A 72.4A 253.0 609.1 16.2 -23.7 39.9 CGCM2 -2.5*,** 122A 164 281.0 782.0 15.4 -22.0 37.4 HadCM3 -2.8*,** 121A 95A 211.0 528.0A 15.7 -23.6 39.3 Lansdowne House Observed -1.3** 112 80.5A 291.0 699.5 17.2 -22.3 39.5 CGCM2 -1.1** 109 164 281.0 782.0 16.6 -20.1 36.7 HadCM3 -1.4** 112 95A 211.0 528.0A 16.6 -21.8 38.4

*,** indicates MAAT criteria at the -2oC and 0oC thresholds, respectively, have been met; A indicates the palsa threshold has been met, in reference to Table 5-1.

5.4.3 MAAT threshold

Projected MAAT values are reported in Table 5-4 for the four weather stations. For Big

Trout Lake, the -2oC threshold derived from Fennoscandian palsas, is attained for all four model simulations for the 2020s (2010-2039), only for one in the 2050s (2040-2069) and for none in the

2080s (2070-2099). The net warming by the end of the 2080s is approximately 2oC with a slightly warmer response with the CGCM2 than with the HadCM3. For Lansdowne House the threshold was not met during the baseline period, and consistent with Big Trout Lake, a gradual warming of approximately 2oC was projected by the end of the 2080s with all four simulations indicating a MAAT of greater than 0oC. Peawanuck shows a baseline temperature 0.6oC warmer than the observations from Peawanuck (Table 5-4) for CGCM2 but 0.9oC cooler for HadCM3.

Both models indicate a stronger warming for these two near coastal sites compared to the in-land

145 sites of Big Trout Lake and Lansdowne House with a warming by the end of the 2080s of over

6oC for CGCM2 and 5oC for HadCM3 using the A2 scenario. The magnitude of warming was substantially lower using the B2 scenario.

Table 5-4. Projected MAAT values for the four weather stations. Baseline T (oC) 2020s T (oC) 2050s T (oC) 2080s T (oC) Fort Severn Observed -4.6*,** ------CGCM2, A2 -4.8*,** -2.5*,** -0.6** 3.0 CGCM2, B2 -4.4*,** -2.2*,** -1.1** 0.1 HadCM3, A2 -4.8*,** -3.6*,** -2.7*,** 0.3 HadCM3, B2 -4.9*,** -3.7*,** -2.7*,** -1.3** Peawanuck Observed -3.7*,** ------CGCM2, A2 -3.1*,** -1.1** 0.6 3.4 CGCM2, B2 -3.1*,** -0.7** 0.2 1.2 HadCM3, A2 -4.8*,** -3.6*,** -2.7*,** 0.3 HadCM3, B2 -4.9*,** -3.7*,** -2.7*,** -1.3** Big Trout Lake Observed -2.7*,** ------CGCM2, A2 -2.5*,** -2.1*,** -1.31** -0.1** CGCM2, B2 -2.5*,** -2.1*,** -1.1** -0.3** HadCM3, A2 -2.8*,** -2.7*,** -2.0*,** -1.1** HadCM3, B2 -2.8*,** -2.7*,** -1.7** -1.3** Lansdowne House Observed -1.3** ------CGCM2, A2 -1.1** -0.67** 0.10 1.28 CGCM2, B2 -1.1** -0.72** 0.30 1.03 HadCM3, A2 -1.39** -1.22** -0.62** 0.33 HadCM3, B2 -1.40** -1.29** -0.31** 0.06

*,** indicates MAAT thresholds at the -2oC and 0oC, respectively, have been met.

146

For both models, the -2oC threshold is not attained by the 2080s, and for CGCM2 this threshold was also not attained by the 2020s and 2050s. At the 2050s horizon, the two models differ greatly. Fort Severn has identical results for HadCM3 as Peawanuck which is likely a reflection of the coarse resolution of the data. For HadCM3 the two locations share the same grid box whereas for CGCM2 their data come from neighboring grid boxes. The latter might explain the warmer response under CGCM2 scenarios. The CGCM2 results indicate a colder temperature for Fort Severn than Peawanuck which is consistent with its more northerly location, and in this case, identical to the HadCM3 result with only a difference of 0.2oC from the observations

(Table 5-4). By the 2080s the CGCM2 results for these locations indicate a warming of close to

8oC for the A2 scenario, the largest of the four locations. This CGCM2 model also shows the location failing the -2oC threshold by the 2050s for both the A2 and B2 scenarios; however, the

HadCM3 projection did not fail the threshold for the 2050s, although it did by the 2080s.

Assessment using the 0oC threshold derived from the adjacent Quebec region was applied to all four weather stations; there was a transformational trend shift towards the latter years of the projected emission scenarios. For Big Trout Lake, the condition for palsa occurrence is met well into the 2080s for the two models under the A2 and B2 emission scenarios. For Lansdowne

House, the threshold is met until the 2020s based on the CGCM2 model and until the 2050s for the HadCM2 model. At Peawanuck, again under coarse resolution, the palsa condition is met until the 2020s under the CGCM2 model; until the 2050s with HadCM3 under A2 emission scenario, and until the 2080s with HadCM3 under the B2 emission scenario. At Fort Severn, the conditions for palsa occurrence were met into the 2050s, however, by the 2080s only HadCM3 under the B2 emission scenario met the conditions at the 0oC threshold.

147

5.4.4 Number of Days below -10oC per year

The projected number of days below -10oC per year was reported in Table 5-5 for Big

Trout Lake and Lansdowne House. This analysis was not done for Peawanuck and Fort Severn as the coarse resolution climate model data were available on a monthly and not on a daily basis as is required for this metric. Although Big Trout Lake meets this criterion for its baseline, it is at the cusp of this threshold and falls just below the threshold for the 2020s and onwards for both climate models. The models differ only in the magnitude of reduction of days below the threshold. By the 2080s, the range of days for these models and scenarios is 56 to 84 representing a reduction of 30% to over 50%. However, it should be noted that most of this reduction occurs after the 2050s and before then the reduction ranged from 8 to 22%.

Table 5-5. Projected number of days below -10oC per year for Big Trout Lake and Lansdowne House. Number of Days Climate Models Baseline T (oC) 2020s T (oC) 2050s T (oC) 2080s T (oC) Big Trout Lake CGCM2, A2 122* 115 94 56 CGCM2, B2 121* 114 93 60 HadCM3, A2 121* 119 110 84 HadCM3, B2 121* 119 103 83 Lansdowne House CGCM2, A2 108 101 78 42 CGCM2, B2 108 101 78 48 HadCM3, A2 112 109 100 70 HadCM3, B2 112 110 91 77

* indicates the number of days per year that are below the -10oC threshold at 120 days have been met.

148

5.5 Discussion

This study provides a first time climate change assessment for the HBL with a particular focus on climate variables related to the formation and occurrence of palsas. To determine climate variables that influence palsa formation and occurrence in the HBL, climate data from weather and climate stations at the northern and southern palsa limits were analyzed. Criteria used to determine the formation and occurrence of palsa formation in the HBL were imported for comparison from other areas of the world which host palsas, in particular Fennoscandia and

Quebec. First, these criteria were evaluated locally for the climate conditions of the HBL. The region met the thermal requirements thresholds derived from Fennoscandian palsas on MAATs and number of days below -10oC per year. For the optimum ranges, the July temperatures in the

HBL were warmer and January temperatures cooler, leading to greater continentality. In addition, the moisture regime differed with more total precipitation and more summer time precipitation.

We note that due to the limited availability of climate data, and lack of continuous data, we were only able to frame the HBL palsa region with two climate stations (Big Trout Lake,

Lansdowne House) just south of the southern extent of HBL palsas and two stations (Peawanuck,

Fort Severn) at the northern edge, along the Hudson Bay coast. For the more complete climate data sets (Big Trout Lake and Lansdowne House) we were able to generate downscaled results to examine climate model projections with more temporal resolution (needed for the days below -

10oC metric). For Peawanuck and Fort Severn, the historical data were too sparse to conduct

149 statistical downscaling and, therefore, we relied on coarse resolution climate model projections at these two stations. In addition, the precipitation record for these two stations was too sparse; this prevented a reliable characterization of the precipitation regime. It should be noted that the roles of precipitation and soil hydrology are key factors in the formation, preservation, and degradation of the palsa life cycle (Seppälä, 1986); however, without reliable precipitation and soil moisture data in this region, the influence of these factors on HBL palsa formation and occurrence could not be fully determined.

This study further examines the climate change impacts on palsas by applying climate change projections, under A2 and B2 emission scenarios, using the identified climate thresholds for palsas in the HBL. We recognize that the projection data analysis focuses on climate conditions based on thermal requirements that are fundamental for palsa genesis, growth and sustenance. Due to both the lack of reliable precipitation and continuous temperature data, we focused on the changes using the -2oC (Fennoscandian) and 0oC (Quebec) MAAT climate thresholds, and the 120 days below -10oC per year thresholds in examining the climate projections. Although we used two models (CGCM2 and HadCM3) under two emissions scenarios (A2 and B2), this does not necessarily mean that palsas will rapidly disappear in the

HBL region. It is important to consider that the existing structures have a considerable thermal inertia that will require a succession of years of an energy imbalance to completely thaw the existing palsas. Estimation of when palsas may disappear from the HBL is beyond the framework of this analysis and would require the use of further palsa modeling (eg. An and

Allard, 1995).

150

5.6 Conclusion

Our study identified thermal thresholds for the formation and occurrence of HBL palsas and examined how these thresholds are met using climate projections for the remainder of the century. Our results indicate that 1) the climate conditions for palsa genesis and growth in the

HBL is already close to the threshold for viability under the current climate conditions, and 2) the climate conditions by the end of the century will not be able to support the genesis of palsas and will likely contribute to palsa degradation. The Big Trout Lake and Lansdowne House results showed that current conditions are near the edge of the -2oC MAAT thresholds and that throughout the rest of this century, the thresholds will increasingly not be met due to a warming of approximately 2oC, which can result in a substantial reduction (exceeding 50% in some scenarios) of the days below -10oC. Under the 0oC MAAT threshold, Big Trout Lake can maintain favourable conditions for palsa occurrence until the end of the century; however, the

Lansdowne House region will experience a different fate as the climate may become unfavorable earlier, surpassing 0oC by mid-century. The projected temperature changes at coastal Peawanuck and Fort Severn climate stations were considerably larger by at least a factor of two, and in one instance by a factor of four, more than the warming of the two in-land weather stations. This attribute reflects concurrent changes in the sea ice distribution of the Hudson Bay with an effective elimination of the wintertime ice platform by the 2050s (Gagnon and Gough, 2005).

Although we were not able to address the 120 days below -10oC threshold metric for Peawanuck and Fort Severn, due to incomplete and unavailable climate data, this metric is not independent of the MAAT. Thus, it is reasonable to apply the behavior of this metric observed at Big Trout

Lake and Lansdowne House and conclude that there will be substantial reductions of this metric

151 at Peawanuck and Fort Severn as well. Finally, we note that these projections pertain to the thermal regime necessary for the presence of palsas and these projections may be significantly modified if the hydrological regime is substantially altered in the future.

5.7 Acknowledgements

The authors would like thank Joyce Zhang for the statistical downscaling contribution for this study. Portions of this research were funded and supported by the Wildlife Research and

Development Section of the Ontario Ministry of Natural Resources, the Ontario Ministry of

Environment, and the Department of Physical and Environmental Sciences at the University of

Toronto Scarborough.

5.8 References

An, W., and Allard, M. 1995. A mathematical approach to modeling palsa formation: Insights on processes and growth conditions. Cold Region Science and Technology 23, 231-244.

Collins, M., Tett, SFB., and Cooper, C. 2001. The internal climate variability of HadCM3, a version of the Hadley Centre coupled model without flux adjustments. Climate Dynamics 17, 61–81.

Cyr, S., and Payette, S. 2010. The origin and structure of wooded permafrost mounds at the arctic treeline in eastern Canada. Plant Ecology & Diversity 3, 35-46.

Environment Canada. 2012. "Climate Data Online - Canada’s National Climate Archive." Government of Canada. Accessed Online: 01 September 2012 from: http://climate.weather.gc.ca/index_e.html

152

Flato, GM., and Boer, GJ. 2001. Warming Asymmetry in Climate Change Simulations. Geophysical. Research Letters 28, 195-198.

Fronzek, S., Luoto, M., and Carter, TR. 2006. Potential effect of climate change on the distribution of palsa mires in Subarctic Fennoscandia. Climate Research 32(1), 1-12.

Gagnon, AS., and Gough, WA. 2002. Hydro-climatic trends in the Hudson Bay region, Canada. Canadian Journal of Water Resources 27, 245-262.

Gagnon, AS., and Gough, WA. 2005. Trend and Variability in the Dates of Ice Freeze-up and Break-up over Hudson Bay and James Bay. Arctic 58(4), 370-382.

Gough, WA., and Leung, A. 2002. Nature and fate of Hudson Bay permafrost. Regional Environmental Change 2, 177-184.

Gurney, SD. 2001. Aspects of the genesis, geomorphology and terminology of palsas: perennial cryogenic mounds. Progress in Physical Geography 25, 249-260.

Hinkel, K., Paetzoid, F., Nelson, FE. and Bockheim, JG. 2001. Patterns of soil temperature and moisture in the active layer and upper Permafrost at Barrow, Alaska: 1993-1999. Global and Planetary Change 29, 293-309.

Kim, SJ., Flato, GM., Boer,GJ., and McFarlane, NA. 2002. A coupled climate model simulation of the Last Glacial Maximum, Part 1: transient multi-decadal response. Climate Dynamics 19, 515-537.

Kim, SJ., Flato, GM. and Boer, GJ. 2003. A coupled climate model simulation of the Last Glacial Maximum, Part 2: approach to equilibrium. Climate Dynamics 20, 635-661.

Kujala, K., Seppälä, M., and Holappa, T. 2008. Physical properties of peat and palsa formation. Cold Regions Science and Technology 52, 408-414.

Lewis, T., and Lamoureux, SF. 2010. Twenty-first century discharge and sediment yield predictions in a small high Arctic watershed. Global and Planetary Change 71, 27-41.

Lewkowicz, AG., and Coultish, TL. 2004. Beaver damming and palsa dynamics in a subarctic mountainous environment, Wolf Creek, Yukon Territory, Canada. Arctic, Antarctic, and Alpine Research 36(2), 208-218.

Luoto, M., Fronzek, S. and Zuidhoff, FS. 2004. Spatial modeling of palsa mires in relation to climate in Northern Europe. Earth Surface Processes and Landforms 29, 1373-1387.

153

Nakićenović, N. 2000. Greenhouse gas emissions scenarios. Technological Forecasting and Social Change 65(2), 149-166.

Nakićenović, N., and Swart, R. ed. 2000. Special Report on Emissions Scenarios: A special report of Working Group III of the Intergovernmental Panel on Climate Change, Cambridge University Press, ISBN 0-521-80081-1, 978-052180081-5 (pb: 0-521-80493- 0, 978-052180493-6).

Parviainen, M., and Luoto, M. 2007. Climate envelopes of mire complex types in Fennoscandia. Geografiska Annaler, Series A: Physical Geography 89(2), 137-151.

Pissart, A. 2002. Palsas, lithalsas and remnants of periglacial mounds. A progress report. Progress in Physical Geography 26 (4), 605-621.

Seppälä, M. 1986. The Origin of Palsas. Physical Geography 68, 141-147.

Tam. A. 2009. Permafrost in Canada’s Subarctic Region of Northern Ontario. Master’s Thesis. Department of Geography, University of Toronto, Toronto.

Thibault, S., and Payette, S. 2009. Recent permafrost degradation in bogs of the James Bay area, Northern Quebec. Permafrost and Periglacial Processes 20, 383-389.

Vallée, S., and Payette, S. 2007. Collapse of permafrost mounds along a subarctic river over the last 100 Years (Northern Quebec). Geomorphology 90, 162-170. van Everdingen, Robert, ed. 1998 revised May 2005. Multi-language glossary of permafrost and related ground-ice terms. Boulder, CO: National Snow and Ice Data Center.

Wilby, RL. and Dawson, C.W. 2012. The Statistical Downscaling Model: Insights from one decade of application. International Journal of Climatology (in press).

Zuidhoff, FS., and Kolstrup, E. 2000. Changes in palsa distribution in relation to climate change in Laivadalen, Northern Sweden, especially 1960-1997. Permafrost and Periglacial Processes 11, 55-69.

154

Chapter 6 Discussion and Conclusion 6 Chapter 6

6.1 Discussion and conclusion

The aim of this thesis is to assess the climatic potential for Canadian permafrost under changing climate conditions along a geographical south-to-north transect. The main research questions of this thesis were addressed individually in Chapters 2 to 5 as standalone chapters.

The objectives of this research were achieved by identifying site-specific characteristics and by determining climate trends for each study locations using available climate data. Climate change projections were determined using GCMs to assess future changes to the potential permafrost distributions along the geographical south-to-north transect.

The contributions of this research further supports an attempt to address a data gap issue in permafrost research along this study transect where historical and continuous active layer and permafrost measurements are either limited or non-existent. Readily accessible data in this research, such as climatological data (collected from Environment Canada weather stations) and soil physical data (collected from field visits), permitted the use of simple permafrost tools to calculate local potential permafrost, primarily the application of the Stefan equation and the

Stefan Frost Number index. More comprehensive tools have been developed and are available, but these tools require significantly more data inputs and field measurements. For example, this research had considered the application of the temperature at the top of permafrost (TTOP) model (Smith and Riseborough, 1996; Henry and Smith, 2001; Smith and Riseborough, 2002)

155 and the Kudryavtsev equation (Anisimov et al., 1997). The Kudryavtsev equation is widely applied for greater accuracy of active later thickness calculations; however this formulation requires greater data inputs that include air and ground temperature measurements, temperature and amplitude measurements of the snow cover and vegetation, height of vegetation, snow thickness, soil moisture content, and soil thermal properties. This comprehensive tool is most preferential and would enhance the accuracy of active layer thickness calculations by considering the inputs listed above. However, the greater data requirements may reduce the tool’s feasibility, the spatial and temporal application, for locations where historical and present day data are limited or when specific inputs are not available. As a result, the simpler tools, such as those related to the Stefan equation, are applicable for locations with limited data, such as the case in

Chapter 2 for Peawanuck, Ontario.

Chapter 2 focuses on Peawanuck, Ontario, located within the subarctic region of northern

Ontario in the Hudson Bay Lowlands, and provides for the first time, calculations of the Stefan

Frost (Fs) Numbers for the period from 1959 to 2011. The Fs is an extension of the Frost

Number (F) that was previously investigated in this region by Gough and Leung (2002). The Fs formulation includes the FDD and TDD indices with the addition of a soil thermal conductivity ratio parameter modifying the FDD, the Stefan equation. The dimensionless Fs results provide an indication of the potential for permafrost, with Fs ≥ 0.67 for continuous permafrost distributions;

<0.67 for discontinuous; <0.6 for sporadic and <0.5 for permafrost absent. This chapter demonstrated that these permafrost tools, F and Fs, can be applied to assess the climatic potential for permafrost, which further supports the linkage between the climate conditions and potential permafrost distributions for Peawanuck. The coupling of the difference between the degree-days with the soil thermal conductivity ratio provides a rationalization of the asymmetric freezing and

156 thawing process associated with the inherent asymmetric frozen and unfrozen soil thermal conductivities. The frozen soil thermal conductivity is typically 1.5 to 2.0 times greater than the unfrozen soil thermal conductivity due to the presence of soil moisture providing an explanation for the presence of permafrost when climate conditions appear not to be favourable (Nixon and

McRoberts, 1973; Shur and Jorgenson, 2007; Tam, 2009; Waller et al., 2012). Six Kf/Ku ratios were considered where three were based on observed soil conditions for peat, silt and clay, and the remaining three were additional theoretical scenarios. The three theoretical Kf/Ku ratio scenarios investigated in this Chapter were to explore the impacts of thermal offsetting effects on

Fs results. For comparison, no thermal offset is represented in the first scenario by a Kf/Ku ratio =

1, where Fs = F. The second scenario of a Kf/Ku ratio = 1.5 represents conditions close to the individual Peawanuck field values for peat, silt, and clay that were based on field observations, and the final scenario of a Kf/Ku ratio = 2.0 represents a strong thermal offset effect on the soil thermal conductivities. The results demonstrated the importance of the peat soil composition at

Peawanuck, and the continuing support for permafrost presence. As soil moisture content affects the soil thermal conductivity, these scenarios of Kf/Ku further provide insight on the variation of the soil moisture content, which can vary spatially and temporally. Climate conditions were analyzed using available and historical weather observation station data and soil properties. Due to incomplete and missing climate data, the records for Peawanuck, Ontario, were combined with its former northern location prior to the 1986 relocation at Winisk, Ontario. Since 1959, indication of climate warming at Peawanuck was observed, with statistical significance, by a decreasing trend in the FDD accumulation. For the TDD, a very weak cooling trend was observed, however this was not statistically significant. In parallel to the Fs calculations, Stefan depths calculations using the Stefan equation provided an indication of the active layer thicknesses by producing the Stefan freezing and thawing depths. From 1959, the Stefan depths

157 of freezing showed statistically significant decreasing trends while the Stefan depths of thawing showed increasing trends, however these were not statistically significant. With the demonstration of the Fs and the Stefan depths, further work is explored in the next chapter by using a more sophisticated method for calculating the Stefan depth and by expanding this methodology northwards along the study transect.

In Chapter 3, research was undertaken to refine the methodology in simulating active layer thickness profiles using the Stefan equation and to provide insight on the permafrost potential at the five locations within the study transect. Application of the Stefan equation requires the assumption that the soil material and corresponding soil thermal conductivity within a layer are homogenous. Xie and Gough (2013) proposed a modified Stefan equation algorithm to calculate multilayered soil profiles of various soil types and thermal conductivities. As

Chapter 2 established the linkage between the climate conditions and potential permafrost distributions for Peawanuck, Chapter 3 applied this concept to a broader geographic context. The study locations in Chapter 3 includes the subarctic, Peawanuck from Chapter 2, and the low and high arctic regions of Canada with the intent of comparing active layer profiles and identifying the various soil types and soil thermal conductivity values. A common period from 2004 to 2011 was selected for this analysis to allow comparison and to validate the simulated active layer thicknesses with field observations taken from 2006 to 2011. Moving north along the transect, the active layer thickness profiles became thinner as was anticipated by the respective decrease in mean annual air temperatures. When the XG-Algorithm active layer simulations results were compared to field observations at two study locations with measurements at the same date, the

XG-Algorithm systematic error was calculated to be was +0.5% for the Peawanuck, Ontario

(RMSE of 2.4 cm); and +1.2% for Alert, Nunavut (RMSE of 4.7 cm). To assess the climatic

158 potential for permafrost, the Stefan Frost Number from Nelson and Outcalt (1987) was applied to the climate data using the known soil thermal conductivity values. Fs results in Chapter 3, for all study locations, demonstrated that permafrost is present and the distribution extent remains continuous, except for the HBL where climate conditions appear to favour discontinuous permafrost, a result consistent with Chapter 2. Chapter 2 and field observations in Tam (2009) indicate the presence of continuous permafrost within the HBL; this suggests that the HBL region may be in a transition from continuous to discontinuous permafrost, and that the field observations made could be that of “relict” continuous permafrost. Chapter 3 further investigates the possibility of continuous permafrost in the HBL by varying the soil thermal conductivity ratio in a sensitivity analysis, as previously demonstrated in Chapter 2. For the HBL under Fs condition 1, the thermal offset effect was removed to demonstrate no thermal offset effect. This resulted in a classification of discontinuous permafrost, below the Fs ≥ 0.67 continuous extent threshold. Condition 2 applied observed field conditions of heterogeneous soil types, in contrast to the homogenous soil composition assumption in Chapter 2, where the soil thermal conductivity ratio was set at 1.6; the results for three of the years resulted in the possible classification of continuous permafrost, which is consistent with Gough and Leung (2002). The remainder of the years indicated discontinuous permafrost potential. Condition 3 applied a soil thermal conductivity ratio of 2.0, a scenario reflective of enhanced soil moisture content resulting in Fs values exceeding 0.67 for the HBL indicating continuous permafrost for all years.

The remaining four locations within the study (Rankin Inlet, Resolute Bay, Eureka, and Alert) were greater than 0.67 supporting continuous permafrost. This Chapter provides support for the notions put forward by Gough and Leung (2002) to explain why the HBL has had continuous permafrost in the past due to the thermal offset, the effect of asymmetric frozen and unfrozen thermal conductivities. In addition to this, the more recent data (2004 to 2011) coupled with the

159 results of Chapter 2 suggest that the region is on the cusp of continuous/discontinuous permafrost climate conditions.

Chapter 4 combines the concepts of the previous chapters to project future climate projections using climate model projections from GCMs to assess permafrost potential at the five study locations within the study transect. This assessment was conducted under IPCC 4th

Assessment Report emission scenarios, A1B, A2, and B1 from 2011 to 2100, which are built into the Localizer Tool, provided by Climate Labs at the University of Toronto Scarborough and the

University of Prince Edward Island, with climate data obtained from Environment Canada

(UTSC, 2013). Validation of the multi-model ensemble MAAT showed differences ranging from

-0.2 to -0.1°C between the modeled values compared with the observed local baseline climate record. The projected rates of climate change for the periods of 2011-2040, 2041-2070, and

2071-2100 were applied to the study locations’ 1971-2000 baseline temperature to simulate future temperatures, with the inherent assumption that the temperature variability of the baseline period is preserved in the projections. The future projected temperatures were ‘bias-corrected’ for locations with limited data availability (<70% data availability), such as at Peawanuck,

Ontario, where the local rate of climate change from the Localizer Tool required correction by the application of a better quality climate data set from Moosonee, Ontario (Fenech, 2009). All locations were projected to experience climate-warming changes with rates ranging from +1.5 to

+7.1°C until the end of the century. Fs was calculated for the projection periods with results projecting climate conditions suitable for maintaining continuous permafrost for Resolute Bay,

Eureka and Alert, Nunavut. Changes in the climate conditions for permafrost are expected in the

Subarctic at Peawanuck for the near future from discontinuous to sporadic permafrost; and, in the Low Arctic at Rankin Inlet by mid-century from continuous to discontinuous including

160 potential sporadic permafrost distributions. Although these calculations were site specific using both localized climate and soil moisture conditions, the results can provide an indication of what may be expected regionally as the continuous permafrost “front” slowly propagates northward over the remaining years of this century.

Chapter 4 applied the emission scenarios of IPCC AR4 that were built into the Localizer

Tool, it should be discussed that at the time of this research, the final draft Report of the Working

Group I contribution to the IPCC 5th Assessment (AR5) was recently released in fall 2013, and the final report expected to be release later in 2014. A difference between the IPCC AR4 and

AR5 is the methodology for projecting future climate change scenarios. The AR4 considers various emission scenarios of atmospheric [CO2] versus the scenarios of radiative forcing targets in the AR5, which also accounts for future emitted greenhouse concentrations (Slater and

Lawrence, 2013). The extent of warming for the Canadian Arctic is expected to range between

+2.0 to +5.0 °C by 2100 (Smith et al., 2005; Zhang et al., 2008; Throop et al., 2012). The projected change in mean global climate based on the AR4 emission scenarios range within +1.1 to +6.4 °C by 2100 (Solomon et al., 2007; Zhou et al., 2009). Considering the radiative forcing targets using RCP2.6, RCP4.5, RCP6.0 and RCP8.5 scenarios, the projected mean temperature changes for the arctic by 2099 are +2.2, +3.8, +4.5 and +7.8 °C, respectively (Slater and

Lawrence, 2013). As both assessment reports have indicated mutual agreement in the future warming extent of the arctic, the projections produced and applied in this study, although based on the AR4 with the Localizer Tool, remain a rational and relevant approach for assessing climate change impacts.

161

Chapter 5 focuses on the climate change impacts specifically on palsa formations and occurrences within the HBL at four locations. As demonstrated in the previous chapters, HBL climate conditions are expected to shift towards favouring discontinuous and sporadic permafrost by the end of the century. The fate of palsas in the HBL may serve as surface features and indicators for underlying permafrost, which may provide insight into climate change impacts that has not been assessed for this region of northern Ontario. Based on the results of this chapter, climate conditions in the 2080s, for the HBL, may no longer be favourable for the formation of palsas based on the -2oC threshold derived from Fennoscandian palsas; however, the continued presence of palsas may be expected when considering the 0oC threshold derived from neighbouring Quebec palsas (Seppälä, 1986; Parviainen and Luoto, 2007; Kujala et al., 2008;

Cyr and Payette, 2010). This suggests that although climate warming may be shifting towards permafrost unfavourable conditions, the existence of palsas may provide refuge for permafrost within its central frozen core, consistent with the projected sporadic permafrost distributions of

Chapter 4. It is also important to note that there may be a lag-time response in climate change impacts to palsas. As suggested in earlier chapters, permafrost may not disappear rapidly in response to climate warming, and that the thermal inertia requires additional time to compensate in order to thaw the frozen ground material, allowing for a temporary disequilibrium between permafrost conditions and concurrent climate conditions. Furthermore, the small differences between the projected MAAT for Peawanuck in this chapter and the preceding chapter may be addressed by the different projection methodologies applied. Chapter 5 applied two individual climate models, the CGCM2 and the HadCM3, for projections using two IPCC emission scenarios and then statistically downscaled the data using SDSM, while Chapter 4 utilized multi- model ensemble means produced from 20-24 model means using three IPCC emission scenarios.

162

In conclusion, the combination of the four research studies provides a climate change impact assessment on potential permafrost distributions in Canada, along the south-to-north study transect. Chapters 2 and 3 identified the main factors affecting permafrost, the climate conditions and the soil thermal conductivity, which can be incorporated into various permafrost tools such as the Stefan equation, the Frost number, and the Stefan Frost Number. Chapters 2 and

3 served to illustrate the transferability of the use of these permafrost tools to other locations where climate and soil physical properties are available; and in the case of Chapter 2, fragmented climate data can be combined in the application of these tools. The results from Chapters 2 and 3 further demonstrated that although climate conditions have changed towards being unfavourable to continuous permafrost, the importance of the soil thermal conductivity can provide support in maintaining the observed continuous permafrost distributions within the HBL at Peawanuck.

Chapter 4 provided climate change projections for the study transect based on multiple GCMs under three future emission scenarios, which projected that climate warming will likely have an effect on changing the permafrost distributions, through a lag-time response. The increase in mean annual air temperatures will result in an increase in the thawing degree-days and a decrease in the freezing degree-days accumulations. Chapter 4 indicated that the area at risk of permafrost degradation will most likely range from the subarctic to the low arctic regions near Peawanuck to

Rankin Inlet, where sporadic and discontinuous permafrost may be expected by the end of the century. Chapter 5 focuses at the southernmost end of the study transect within the HBL at four locations by exploring the fate of palsas, an indicator of permafrost, with the climate projections from two GCMs, and statistical downscaling models for two locations. These results indicated that warming climate conditions by the end of the century will not support the formation of new palsas; however, the presence of existing palsas could be supported which may also become refuges for permafrost. Chapter 5 provides further support to the conclusion of Chapter 4 for the

163

HBL, where the climate projections by the end of this century indicates change in the permafrost potential towards a sporadic distribution, by the presence of palsas. Finally, with these changes to the permafrost distributions, this may also lead to other profound impacts for the local landscape, such as slope failures and mass movements, and to the communities located in the

North, especially on building infrastructure, transportation and food security (French, 2007;

Etzelmüller et al., 2011; Lewkowicz, 2011).

6.2 Recommendations for further research

Permafrost research encompasses many disciplines such as geology, soil science, ecology, hydrology, and climatology (Muller, 2008); as such, there are many complexities to consider when examining the impact of climate change. Recommendations for further research include 1) determining the time lag associated with climate conditions on permafrost distributions by using permafrost modelling; 2) examining the impact of future hydrological conditions on the soil moisture content and soil thermal conductivity within the Stefan Frost

Number calculations; 3) gaining a better understanding of the implications of positive feedback effects associated with the release of greenhouse gasses and carbon storages of soil organic carbon contained within the permafrost layers; and finally, 4) furthering permafrost field studies and the installing of weather observation stations in remote locations to enhance spatial resolution.

Further research into the time lag associated with climate conditions on permafrost distributions can help to establish time frames in which permafrost distributions may change

(Anisimov et al., 1997). This could be accomplished by the inclusion of a time dependency

164 factor to an iterative approach in calculating active layer development, such as a modified XG-

Algorithm.

The impacts on future hydrological conditions should be investigated as the increase in moisture by precipitation and soil moisture content can increase the soil thermal conductivities within ground materials. An increase in the soil thermal conductivity can influence the results of the Stefan equation that forms the Stefan Frost Number, and the XG-Algorithm. For example, the Stefan Frost Number contains the Kf/Ku ratio that influences the freezing degree-days variable, an increase in the Kf may enhance freezing or an increase in Ku may reduce the freezing potential, both could affect the permafrost potential. Further research is also required on developing nf values based on snow depths for these study locations.

A further understanding of the positive feedback effects associated with the release of greenhouse gasses and carbon storages of soil organic carbon contained within the permafrost layers is of great interest (French, 2007; Grosse et al., 2011; Watanabe et al., 2011). The scope of this thesis did not examine the amount of carbon stored within the permafrost, nor account for any amplified greenhouse effects from the release of the greenhouse gasses. For climate change scenarios, this thesis relied upon the available emissions scenarios provided by the IPCC

(TGICA, 2007). At the time of this thesis, the final draft Report of the Working Group I contribution to the IPCC 5th Assessment Report "Climate Change 2013: The Physical Science

Basis" was not available for citation, quotation or distribution. Studies into the amount of stored carbon in permafrost regions can provide comprehensive understanding on future atmospheric greenhouse gas concentrations and its effects on global climate change, especially if carbon sinks become sources (Reyes et al., 2010).

165

Finally, given the vast geography of northern Canada, a further research recommendation for permafrost research is to expand permafrost field studies and weather observation stations where possible, especially in remote locations where data is currently limited or missing. By expanding monitoring and field stations, the availability of climate and environmental data can satisfy the data requirements of more comprehensive permafrost tools, such as TTOP and the

Kudryavtsev equation. Further data availability can assist in validation of GCMs and permafrost model performances with observations prior to assessing site-specific climate change impacts

(Anisimov et al., 1997; Bonnaventure and Lewkowicz, 2011). Long-term climate and permafrost monitoring programs should be expanded in the Canadian North, especially within the Canadian

Subarctic and Low Arctic regions as support in this research where permafrost change is expected within this century (Smith et al., 2005).

166

6.3 References

Anisimov, O., Shiklomanov, N., and Nelson, F. 1997. Global warming and active-layer thickness: results from transient general circulation models. Global and Planetary Change 15, 61-77.

Bonnaventure, PP., and Lewkowicz, AG. 2011. Modelling climate change effects on the spatial distribution of mountain permafrost at three sites in northwest Canada. Climate Change 105, 293-315.

Cyr, S., and Payette, S. 2010. The origin and structure of wooded permafrost mounds at the arctic treeline in eastern Canada. Plant Ecology & Diversity 3, 35-46.

Dobinski, W. 2011. Permafrost. Earth-Science Reviews 108, 158-169.

French, M. 2007. The Periglacial Environment. 3rd Ed. West Sussex: John Wiley & Sons.

Gough, WA., and Leung, A. 2002. Nature and fate of Hudson Bay permafrost. Regional Environmental Change 2, 177-184.

Grosse, G., Harden, J., Turetsky, M., McGuire, AD., Camill, P., Tarnocai, C., Frolking, S., Schuur, EAG., Jorgenson, T., Marchenko, S., Romanovsky, V., Wickland, KP., French, N., Waldrop, M., Bourgeau-Chavez, L., and Striegl, RG. 2011. Vulnerability of high- latitude soil organic carbon in North America to disturbance. Journal of Geophysical Research 116, G00K06, doi: 10.1029/2010JG001507.

Henry, K., and Smith, M. 2001. A Model-based Map of Ground Temperatures for the Permafrost Regions of Canada. Permafrost and Periglacial Processes 12, 389-398.

Kujala, K., Seppälä, M., and Holappa, T. 2008. Physical properties of peat and palsa formation. Cold Regions Science and Technology 52, 408-414.

167

Lewkowicz, AG. 2011. Slope hummock development, Fosheim Peninsula, Ellesmere Island, Nunavut, Canada. Quaternary Research 75, 334-346.

Muller, S. 2008. Frozen in Time: Permafrost and Engineering Problems, HM French and FE Nelson (eds). Reston, VA: American Society of Civil Engineers.

Nelson, F., and Outcalt, S. 1987. A Computational Method for Prediction and Regionalization of Permafrost. Arctic and Alpine Research 19, 279-288.

Nixon, J., and McRoberts, E. 197: A study of factors affecting the thawing of frozen soils. Canadian Geotechnical Journal 20, 439-452.

Parviainen, M., and Luoto, M. 2007. Climate envelopes of mire complex types in Fennoscandia. Geografiska Annaler, Series A: Physical Geography, 89(2), 137-151.

Reyes, AV., Froese, DG., and Jensen, BJL. 2010. Permafrost response to last interglacial warming: field evidence from non-glaciated Yukon and Alaska. Quaternary Science Reviews 29, 3256-3274.

Seppälä, M., 1986: The Origin of Palsas. Physical Geography 68, 141-147.

Shur, Y., and Jorgenson, M. 2007. Patterns of Permafrost Formation and Degradation in Relation to Climate and Ecosystems. Permafrost and Periglacial Processes 18, 7–19.

Slater, AG., and Lawrence, DM. 2013. Diagnosing Present and Future Permafrost from Climate Models. Journal of Climate 26, 5608-5623.

Smith, MW., and Riseborough, DW. 1996. Permafrost monitoring and detection of climate change. Permafrost and Periglacial Processes 7, 301-310.

Smith, MW., and Riseborough, D. 2002. Climate and the Limits of Permafrost: A Zonal Analysis. Permafrost and Periglacial Process 13, 1-15.

Smith, SL., and Burgess, MM., Riseborough, D., and Nixon, FM. 2005. Recent Trends from Canadian Permafrost Thermal Monitoring Network Sites. Permafrost and Periglacial Process 16, 19-30.

168

Tam, A. 2009. Permafrost in Canada's Subarctic Region of Northern Ontario. Master’s Thesis. Department of Geography, University of Toronto, Toronto.

TGICA. 2007. “General Guidelines on the use of Scenario Data for Climate Change Adaptation Assessment, Version 2.” Intergovernmental Panel on Climate Change. Accessed online: 01 May 2013 from: http://www.ipcc-data.org/guidelines

UTSC. 2013. Localizer. University of Toronto Scarborough. Accessed online: 22 May 2013 from: http://climatechange.utsc.utoronto.ca/

Waller, RI., Murton, JB., and Kristensen, L. 2012. Glacier-permafrost interactions: Processes, product and glaciological implications. Sedimentary Geology 255-256, 1-28.

Watanabe, S., Laurion, I., Chokmani, K., Pienitz, R., and Vincent, WF. 2011. Optical diversity of thaw ponds in discontinuous permafrost: A model system for water color analysis. Journal of Geophysical Research 116, G02003, doi: 10.1029/2010JG001380.

Xie, C., and Gough, WA. 2013. A simple thaw-freeze algorithm for multilayered soils using the Stefan equation. Permafrost and Periglacial Processes, DOI: 10.1002/ppp.1770.