ATMOSPHERIC DRIVERS of SNOWFALL and SNOW COVER ABLATION VARIABILITY WITHIN the GREAT LAKES BASIN of NORTH AMERICA by Zachary J

ATMOSPHERIC DRIVERS OF SNOWFALL

AND SNOW COVER ABLATION VARIABILITY WITHIN

Zachary J. Suriano

A dissertation submitted to the Faculty of the University of Delaware in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Climatology

Spring 2018

ATMOSPHERIC DRIVERS OF SNOWFALL

AND SNOW COVER ABLATION VARIABILITY WITHIN

THE GREAT LAKES BASIN OF NORTH AMERICA

Zachary J. Suriano

Approved: ______Delphis F. Levia, Ph.D. Chair of the Department of Geography

Approved: ______Estella A. Atekwana, Ph.D. Dean of the College of Earth, Ocean, & Environment

Approved: ______Ann L. Ardis, Ph.D. Senior Vice Provost for Graduate and Professional Education

I certify that I have read this dissertation and that in my opinion it meets the academic and professional standard required by the University as a dissertation for the degree of Doctor of Philosophy.

Signed: ______Daniel J. Leathers, Ph.D. Professor in charge of dissertation

Signed: ______David A. Robinson, Ph.D. Member of dissertation committee

Signed: ______Sara A. Rauscher, Ph.D. Member of dissertation committee

Signed: ______Tracy L. DeLiberty, Ph.D. Member of dissertation committee

Signed: ______Gina R. Henderson, Ph.D. Member of dissertation committee ACKNOWLEDGMENTS

This research was partially supported by the United States National Oceanic and Atmospheric Administration Climate Program Office (grant NA14OAR4310207), the 2016 and 2017 Dr. John R. Mather Graduate Research Grant, and the University of Delaware. Any opinions, findings, conclusions, or recommendations expressed in this study are those of the authors and do not necessarily reflect the views of the funding sources. I wish to acknowledge the support, insight, and guidance provided by my advisor, Daniel Leathers, and my committee members David Robinson, Gina Henderson, Sara Rauscher, and Tracy DeLiberty. Their technical assistance and advice was critical to the success of this study and my graduate education. I also acknowledge the members of our NOAA grant research team at Rutgers University, the University of Georgia, and the United States Naval Academy. Gratitude is also extended to the faculty of the Department of Geography for their formal and informal mentoring that furthered my professional development. Finally, thank you to Lindsey, Madeleine, and my entire family for their continuous love and support.

iv TABLE OF CONTENTS

LIST OF TABLES ...... viii LIST OF FIGURES ...... ix ABSTRACT ...... xiii

Chapter

1 INTRODUCTION ...... 1

1.1 Lake-effect Snowfall ...... 2 1.2 Snow Cover Ablation ...... 4 1.3 Dissertation Organization ...... 9

REFERENCES ...... 10

2 SYNOPTICALLY CLASSIFIED LAKE-EFFECT SNOWFALL TRENDS TO THE LEE OF LAKES ERIE AND ONTARIO ...... 16

2.1 Abstract ...... 16 2.2 Introduction ...... 17 2.3 Data and Methodology ...... 19

2.3.1 Datasets ...... 19 2.3.2 Synoptic Classification ...... 21 2.3.3 Lake-effect Classification ...... 24

2.4 Results and Discussion ...... 26

2.4.1 Seasonal Snowfall Distribution ...... 26 2.4.2 Temporal Snowfall Trends ...... 28 2.4.3 Causes of Snowfall Variability and Trends ...... 29 2.4.4 Limitations ...... 33

2.5 Summary and Conclusions ...... 34 2.6 Acknowledgements ...... 37

REFERENCES ...... 38

v 3 SPATIOTEMPORAL VARIABILITY OF GREAT LAKES BASIN SNOW COVER ABLATION EVENTS ...... 51

3.1 Abstract ...... 51 3.2 Introduction ...... 52 3.3 Data and Methodology ...... 53

3.3.1 Snow Depth Data ...... 53 3.3.2 Basin Definition ...... 55 3.3.3 Ablation Definition and Calculation ...... 56

3.4 Results ...... 58

3.4.1 Basin-Scale Ablation Events ...... 58 3.4.2 Sub-Basin Scale Spatial Patterns and Trends in Ablation Events ...... 60

3.5 Discussion and Conclusions ...... 62 3.6 Acknowledgements ...... 65

REFERENCES ...... 66

4 GREAT LAKES BASIN SNOW COVER ABLATION AND SYNOPTIC- SCALE CIRCULATION ...... 77

4.1 Abstract ...... 77 4.2 Introduction ...... 78 4.3 Methodology and Datasets ...... 80

4.3.1 Temporal Synoptic Index ...... 80 4.3.2 Snow Dataset and Great Lakes Basin Definition ...... 80 4.3.3 Ablation Definition ...... 81

4.4 Results ...... 83

4.4.1 Synoptic Analysis ...... 83 4.4.2 Ablation Case Studies ...... 85 4.4.3 Inter-annual Variability of Synoptic Types ...... 88

4.5 Discussion and Conclusions ...... 89 4.6 Acknowledgements ...... 92

REFERENCES ...... 93

vi 5 CONCLUSIONS ...... 105

Appendix

A PERMISSIONS ...... 110

vii LIST OF TABLES

Table 2.1.: Average surface meteorological characteristics for the seven lake- effect synoptic weather types at 0900, 1500, 2100, 0300 UTC...... 42

Table 2.2.: Average monthly lake-effect snow (cm), and ratio of monthly to seasonal lake-effect snowfall total (%) for Lake Erie and Lake Ontario from 1950-2009. These defined regions are shown in Figure 2.3...... 43

Table 4.1.: Meteorological characteristics in Flint, MI at 0900, 1500, 2100, and 0300 UTC and statistical summary for the ten most common synoptic weather types leading to ablation in the Great Lakes basin...... 96

viii LIST OF FIGURES

Figure 2.1.: Adapted from Suriano and Leathers (2017); composite sea-level pressure fields (hPa) for the seven lake-effect synoptic types a) WNW-1, b) W-1, c) SW-1, d) WSW-1, e) W-2, f) WSW-2, and g) NW-1. Red shades correspond to higher SLP while blue shades correspond to lower SLP...... 44

Figure 2.2.: Average seasonal snowfall (cm) associated with lake-effect synoptic types (a), and (b) ratio of snowfall associated with lake-effect synoptic types to snowfall from all synoptic types in the eastern Great Lakes region fro 1950-2009 winter seasons. Darker blues represent higher values while lighter blues to white represent lower values...... 45

Figure 2.3.: Map of the study region. The grid cells used for defining Lake Ontario and Lake Erie lake-effect snowfall are shown in blue. Buffalo, NY is labeled...... 46

Figure 2.4.: Lake-effect snowfall (cm) and snowfall trends (cm year-1) associated with Lakes Ontario and Erie for (a) 1950-2009 using the initial snowfall amounts (black) and snowfall only from stations identified in Kunkel et al. (2009) (grey), (b) initial snowfall amounts during two 30-year periods corresponding to 1950-1979 (black) and 1980-2009 (grey), and (c) a 21-year moving trend in initial lake-effect snowfall from 1960-1999 at an interval of 1 year...... 47

Figure 2.5.: Lake-effect synoptic type frequency (days) and trend (days year-1) during two 30-year periods of 1950-1979 (black) and 1980-2009 (grey) (a), and (b) scatterplot of lake-effect synoptic type frequency (days) and lake-effect snowfall (cm) associated with Lakes Ontario and Erie from 1950-2009...... 48

ix Figure 2.6.: Linearly extrapolated change in snowfall (cm) during 1950-1979 seasons based on (a) changes in the frequency of lake-effect synoptic types, (b) changes in snowfall rates (snowfall intensity), (c) total changes in snowfall due to frequency and intensity changes (panel a plus panel b), and (d) observed snowfall data. Colors are consistent across all four panels and with Fig. 2.7. Blues represent increases in snowfall while yellows represent decreases in snowfall. Darker shades correspond to larger changes with lighter shades corresponding to smaller changes...... 49

Figure 2.7.: As in Figure 2.6 except for 1980-2009 seasons...... 50

Figure 3.1.: Monthly 1960-2009 averaged runoff into the basin (black-solid), over lake evaporation (grey-solid), and over lake precipitation (black-short dash) on the left y-axis in mm over the lake surface, and lake-water levels (grey-long dash) on the right y-axis in meters. Runoff, evaporation, and precipitation data come from the Great Lakes Environmental Research Laboratory (https://www.glerl.noaa.gov/) while lake-water levels are provided by the U.S. Army Corps of Engineers (http://www.lre.usace.army.mil/)...... 69

Figure 3.2.: Map depicting the 57 1-degree grid cells (dark edges) representing the Great Lakes Basin. Portions of two sub-regions are highlighted based on significant trends in the annual number of ablation events: the northeastern and northwestern Lake Superior drainage basins (black), and the eastern Lake Huron and eastern Georgian Bay drainage basins (dark grey)...... 70

Figure 3.3.: Average monthly ablation event frequency across the Great Lakes basin from 1960-2009 (a), and (b) standardized average monthly ablation event frequency for the four defined threshold levels from 1960-2009 ...... 71

Figure 3.4.: Total number of ablation events greater than 2.54 cm across the Great Lakes basin from 1960-2009. Darker shades represent progressively more events (days)...... 72

Figure 3.5.: Probability of an ablation event greater than 2.5 cm from 1960-2009 across the Great Lakes basin for October through April (fraction). Lighter shades represent a lower probability while darker shades represent higher probability...... 73

x Figure 3.6.: Month of peak ablation event frequency for all ablation events, and for events at the defined threshold levels across the Great Lakes basin from 1960-2009. Colors indicate different months...... 74

Figure 3.7.: Trend (a) and statistical significant (b) in the annual frequency of ablation events across the Great Lakes basin from 1960-2009. For panel (a), in days year-1, brown shades represent negative trends while blue represent positive trends. In panel (b), reported in 1 – p-value, darker red shades correspond to lower p-values and thus higher statistical significance...... 75

Figure 3.8.: Annual frequency of snow ablation events (solid) and associated trend line (dashed) in the (a) northwest and northeast Lake Superior drainage basins, and (b) eastern Georgian Bay and eastern Lake Huron drainage basins...... 76

Figure 4.1.: Composite sea-level pressure fields in hPa for the ten most common synoptic weather types leading to ablation. Grouped by category: southerly flow (S1-5), rain-on-snow (R1-3), high overhead (H1-2)...... 97

Figure 4.2.: Sea-level pressure in hPa (a), surface air temperature in °C (b), surface dew point temperature in °C (c), surface wind speed in ms-1 (d), and percent cloud cover (e) at 0000 UTC (1800 CST) during case study 1 on February 27, 1974...... 98

Figure 4.3.: Snow depth preceding the event (a) and spatial distribution of ablation across the Great Lakes basin for case study 1 on February 27, 1974. Darker shades represent deeper snow depths and greater ablation in cm...... 99

Figure 4.4.: Sea-level pressure in hPa (a), surface air temperature in °C (b), surface dew point temperature in °C (c), surface wind speed in ms-1 (d), and percent cloud cover (e) at 0000 UTC (1800 CST), and daily precipitation total in cm (f) during case study 2 on February 13, 1984. 100

Figure 4.5.: Snow depth preceding the event (a) and spatial distribution of ablation across the Great Lakes basin for case study 2 on February 13, 1984. Darker shades represent deeper snow depths and greater ablation in cm...... 101

Figure 4.6.: Sea-level pressure in hPa (a), surface air temperature in °C (b), surface dew point temperature in °C (c), surface wind speed in ms-1 (d), and percent cloud cover (e) at 0000 UTC (1800 CST) during case study 3 on March 18, 2001...... 102

xi Figure 4.7.: Snow depth preceding the event (a) and spatial distribution of ablation across the Great Lakes basin for case study 3. Darker shades represent deeper snow depths and greater ablation in cm on March 18, 2001. .... 103

Figure 4.8.: Inter-annual frequency from 1960-2009 of the general ablation- inducing synoptic categories: southerly flow (black), rain-on-snow (gray), and high-pressure overhead (dashed). The frequencies of the rain-on-snow and high overhead synoptic categories are significantly decreasing and increasing respectively...... 104

xii ABSTRACT

This dissertation examines the relationships between snow and synoptic-scale atmospheric circulation in the Great Lakes region of North America in a series of three journal articles. The first assesses the variability and long-term trends of lake-effect snowfall along the eastern shores of Lakes Erie and Ontario, and determines the particular synoptic-scale weather types that drive the variability in snowfall. These weather type frequencies explain over 68% of inter-annual lake-effect snowfall variability, and between 89-95% of the observed linear changes in snowfall can be explained by long-term changes in the frequency and snowfall rates of these synoptic patterns. The second article builds a climatology of snow ablation events within the Great Lakes basin by isolating ablation from a daily gridded snow depth product. Ablation events are latitudinally-dependent, with peak probability of an event shifting northwards during the spring months in conjunction with enhanced incoming solar radiation, surface air temperatures, and atmospheric moisture. No long-term changes in the seasonal timing of ablation events are detected within the basin, however two spatially coherent regions corresponding to the northern Lake Superior and the eastern Lake Huron/Georgian Bay drainage basins did experience significant decreases and increases in inter-annual ablation event frequency from 1960-2009, respectively. Such changes are hypothesized to be driven by changes in the frequency of particular mid- latitude cyclones influencing the region and long-term trends in lake-effect snowfall.

xiii The third article employs a synoptic-classification procedure that identifies and analyzes the atmospheric conditions that lead to snow ablation events across the Great Lakes basin. Three primary categories of synoptic weather types lead to ablation, corresponding to ‘southerly flow’, ‘rain-on-snow’, and ‘high-pressure overhead’ patterns. Each pattern influences the meteorological conditions forcing ablation at the surface, and exhibits substantial inter-annual variability. The second and third most common ablation-inducing synoptic weather type categorizes, ‘high-pressure overhead’ and ‘rain-on-snow’, are respectively increasing and decreasing in inter- annual frequency from 1960-2009. Together, these three articles showcase the variable forcings of snow in the Great Lakes basin, and highlight the importance of understanding the links between atmospheric circulation and cryospheric water resources.

xi v Chapter 1

INTRODUCTION

The Great Lakes region of North America encompasses a portion of the United States and Canada, including the eight states of Illinois, Indiana, Michigan, Minnesota, New York, Ohio, Pennsylvania, and Wisconsin, and the province of Ontario. Collectively, the region is the largest surface freshwater system on Earth, containing some 21% of the world’s surface water supply across the five primary Great Lakes (Superior, Michigan, Huron, Erie, and Ontario), Lake St. Clair, and the Georgian Bay. Climatologically, the region undergoes large intra-seasonal variations in temperature, and is located at the cusp of ephemeral snow cover over the North American continent. As such, the Great Lakes region experiences a high degree of snow variability. In some regions, average winter season snow depths range from over 70 cm to less than 5 cm inter-annually. Such differences in snow cover can have a large impact on society, the regional climate, and the hydrologic cycle. Additionally, large and rapid accumulation and ablation events can be particularly detrimental in the form of hazardous travel conditions, power outages, snowmelt induced flooding, transport of pollutants, and/or water stress. For communities living in the region, considerable resources are invested into preventative and reactive measures to ensure human safety, including road treatments, snow removal, dam water release, and ice jam removal. It is critical in efficiently utilizing these resources to have a robust understanding in how and why snow varies in the region. This study utilizes a synoptic weather classification procedure to quantify what specific weather patterns

1 lead to snowfall and snow cover ablation events in the Great Lakes region, and is used to attribute observed variability in snow to atmospheric circulation.

1.1 Lake-effect Snowfall Lake-effect snow is the enhancement of snowfall downwind of lakes occurring during the late fall and winter months, as the relatively warm waters of the lakes destabilize the boundary layer of overrunning cold air masses (Eichenlaub 1970). The strength and vertical height of the destabilized layer is proportional to the development of low-level convective cloud bands and snowfall; heightened temperature gradients between the air mass and the lake surface yield deeper and less stable layers, stronger convection, and more lake-effect snowfall (Niziol et al. 1995). This lake-effect process can produce twice as much snow in locations downwind of the lakes relative to locations further inland (Norton and Bolsenga 1993). Boundary layer instability is arguably the most critical process in the formation of lake-effect snow (Kristovich et al. 2003), where a temperature difference between the lake-water-surface and 850 hPa of at least 13°C is necessary for lake-effect snow (Holrody 1971; Niziol et al. 1995). This temperature difference is very important in the development of cloud bands, where minor differences of only 2°C can alter snowfall totals by 30-40%.

In addition to boundary layer instability, suitable winds and a lifting mechanism are necessary for snowfall. While uplift can be achieved through instability alone, it is often enhanced via frictional and thermal convergence over the lake surface, or via orographic features inland. The 850 hPa steering wind speed, direction, and shear dictate the uplift strength and the potential for band formation. Wind speeds should ideally fall between 15-20 ms-1 for optimal air parcel residence

2 time over the lake to receive the appropriate heat and moisture fluxes from the water surface, and to ensure land breezes do not dominate flow (Niziol et al. 1995). Wind direction is equally important to the development of lake-effect snow bands as it determines the fetch, or distance an air parcel travels over the lake surface. Minimum fetch is 80 km, where generally the longer the fetch, the greater the potential for stronger convergence, convection, and snowfall. Lake-effect snow can cause significant financial and physical hardship to the surrounding region, with impacts on transportation, agriculture, and socioeconomics (Norton and Bolsenga 1993; Schmidlin 1993; Kunkel et al. 2002; Changnon et al. 2006). Snow removal, road treatment, and power restoration budgets for individual cities in the region often exceed one million (U.S.) dollars annually, and can cripple the infrastructure for days, or up to several weeks in extreme cases. The additional snowfall can benefit some sectors of the economy however, including recreation and winter-product sales (Schmidlin 1993; Kunkel et al. 2002). The long-term behavior of snowfall in the region is well documented, where multiple studies have examined trends in snowfall within the Great Lakes region. However, depending on the specific region and time period examined, different trends have been observed (Leathers et al. 1993; Norton and Bolsenga 1993; Leathers and Ellis 1996; Grover and Sousounis 2002; Burnett et al. 2003; Ellis and Johnson 2004; Kunkel et al 2009; Bard and Kristovich 2012; Hartnett et al. 2014; Loveless et al. 2014). While a majority have found increasing snowfall trends with time, Bard and Kristovich (2012) and Hartnett et al. (2014) respectively found a trend reversal such that snowfall began to decrease in the 1980s downwind of Lake Michigan and after the 1970s in central New York. Both of these regions regularly experience lake-effect

3 snowfall. Linear trend analysis within this manuscript is subject to additional bootstrapping techniques for robustness. Examining lake-effect snow not by the commonly defined 80-100km lake belt mentioned above (Eichenlaub 1970; Dewey 1979; Norton and Bolsenga 1993; Scott and Huff 1996), but based on snow falling during lake effect synoptic situations, will allow for investigation of actual lake-effect snowfall amounts, not just snowfall within the lake belts (Suriano and Leathers 2017a). The robust investigation of snowfall forced by lake-effect synoptic condition over the eastern Great Lakes region conducted in this study supplements existing knowledge and specifically explains the physical mechanisms behind the trends noted in the literature. Additionally, while 21st century model projections support decreases in lake-induced snowfall in the region (Suriano and Leathers 2016), understanding the forcing mechanisms behind current trends in snowfall caused by lake effect synoptic situations may increase the understanding on what factors will drive projected changes into the future.

1.2 Snow Cover Ablation According to the American Meteorological Society’s Glossary of Meteorology, ablation is considered “all processes that remove snow, ice, or water from a glacier, snowfield, etc; in the sense of the opposite of accumulation.” Snow ablation is a major contributor to soil moisture, stream flow, and groundwater, with over a billion people of the world’s population living in snowmelt-dominated regions (Barnett et al. 2005). In regions with ephemeral snow cover, there are challenges in predicting and preparing for the magnitude and timing of multiple snowmelts each year. A lack of understanding can lead to negative societal consequences including both flooding and drought conditions along snowmelt fed rivers, and the transport of excess nutrients,

4 chemicals, or pollution. Changnon (2008) found that from 1972-2006, snowmelt flooding resulted in $3.365 billion (2006 dollars) in damages and that snowmelt related floods had the largest spatial extent of any type of flood with an average of eight states affected per event. Snowmelt flooding was also found to represent approximately 7% of the 815 flood related fatalities in the United States from 1996- 2005 (Ashley and Ashley 2008). Ashley and Ashley (2008) also noted that of all floods in their study period, the 6th most deadly flood was a snowmelt event in 1996 where 22 people died in Pennsylvania, Virginia, Vermont, and West Virginia (also see Leathers et al. 1998). In the Great Lakes region, runoff is dominated by snowmelt (Barnett et al. 2005) with snowmelt being the primary driver behind the seasonal cycle of Great Lakes water levels in spring and summer (Quinn 2002). Changes in snowfall could therefore substantially change the annual runoff into the basin, potentially changing lake levels and soil moisture availability, along with other impacts. Variable lake level has been shown to impact a variety of environmental and ecological factors including fish habitats, sediment-water nutrients, aquatic vegetation, and marsh bird breeding abundance (Barry et al. 2004; Chow-Fraser 2005; Timmermans et al. 2008; Steinman et al. 2012; Fracz and Chow-Fraser 2013). Processes of snow ablation include melt, sublimation, wind erosion, and avalanche, all of which decrease the mass of snow/water in the pack. In the Great Lakes region, where large differences in elevation are uncommon, avalanches are exceedingly rare and likely have little to no impact on daily ablation. Similarly, snowpack ablation caused by surface and wind-aided sublimation and wind erosion are relatively minor on daily scales. Sublimation and blowing snow transport are

5 found to remove only an average of 7% of local annual precipitation, or 28 mm yr-1 of liquid over portions of Canada, where the effect is minimal over a single storm total (Dery and Yau 2002). Of this snow pack loss, blowing snow transport corresponds to only a minute fraction. Thus, daily ablation in the Great Lake region is primarily forced by snowmelt. With daily snow data from the National Weather Service Cooperative Observer Program (see www.nws.noaa.gov.om/coop/) having no information on water equivalent, or snow mass, it can be difficult to differentiate decreases in snow depth due to melt and other physical factors such as snowpack compaction. Compaction or compression can substantially decrease snow depth without modifying the mass of the snow through destructive metamorphism of fallen snow crystals. As falling snow reaches the surface and collects, destructive metamorphism begins resulting in the loss of sharp crystalline facets of snow and the development of rounded branches, eventually leading to well-rounded crystals (Colbeck 1983a,b). This process occurs due to mechanical stress, sintering, and gradients of vapor pressure surrounding individual snow crystals (Colbeck 1983a,b; Colbeck 1998). Well-rounded crystals can pack together more closely than their highly faceted counterparts, producing a denser, thinner snowpack with time. Under high-wind conditions, deposited snow can have densities as high as 0.4 Mg m-3 (1.0 Mg m-3 = 1.0 g cm-3) compared to more typical densities of 0.1 Mg m-3 (Mellor 1977). Similarly, with increasing deposition of snow, more overburden force is placed on the snowpack, increasing the compression or densification rate (Colbeck 1973; Mellor 1977). While very initial densification rates of low-density snow (ρ < 0.1 Mg m-3) that is buried quickly during heavy snowfall conditions can exceed 10-4 s-1

6 (Mellor 1977), a more typical value of initial fractional settling rate (compression rate) is 0.01 h-1 (1% per hour) caused by destructive metamorphism (Gunn 1965; Barry et al. 1990). To overcome the potential of snowpack compaction artificially enhancing ablation frequency, prior studies have utilized snow depth decreases as a proxy for snowmelt and ablation only when certain conditions are met including temperature and depth change thresholds (Grundstein and Leathers 1998; Grundstein and Leathers 1999; Leathers et al. 2004; Dyer and Mote 2007). Using snow depth changes as a proxy for ablation over short periods has been shown to introduce error (e.g., Hubley 1954), however by accounting for temperature and fresh snow accumulations, this error is limited (Dyer and Mote 2007). Trends in snow ablation were investigated by Dyer and Mote (2007) over North America from 1960-2000. They found peak ablation occurring in April with increasing frequency of March ablation events and decreasing frequency of May ablation events, indicating a shift towards earlier ablation of the snowpack. This is likely the result of high sensible heat fluxes in March, resulting from increased frequency of Dry Moderate air masses over central Canada, coupled with a decrease in Dry Polar air masses (Dyer and Mote 2007). Across the middle latitude regions of the Northern Hemisphere, a number of different synoptic conditions lead to snowmelt (Leathers et al. 1998; Leathers et al. 2004; McCabe et al. 2007; Ashley and Ashley 2008; Mazurkiewicz et al. 2008; Bednorz 2009; Bednorz and Widig 2015). In the North American Great Plains, Grundstein and Leathers (1998, 1999) identified three weather patterns associated with snowmelt that all involve a mid-latitude cyclone tracking through the region and the

7 advection of warm air from the Gulf of Mexico. Similarly, ablation events in the Polish-German lowlands most frequently occur when warm maritime air is advected into the region via pressure patterns of anomalously strong low-pressure systems in the northern Atlantic, and high-pressure over the Mediterranean (Bednorz 2009). These patterns often result in anomalously high precipitation, generally classified as rain-on-snow events (McCabe et al. 2007; Mazurkiewicz et al. 2008; Levia and Leathers 2011). Ablation in the West Siberian Plain is associated with different circulation types that include westerly, southwesterly, and southerly flow types (Bednorz and Wibig 2015). In the central Appalachians, a majority of ablation events occur during ‘moist’ air masses with cloudy, windy conditions and high dew point temperatures, however ‘dry’ air masses with clear skies also cause events (Leathers et al. 2004). As previously indicated, the frequency and magnitude of an ablation event is dependent, in part, on the occurrence of particular atmospheric patterns that provide sufficient meteorological conditions for melt. As such, synoptic classification techniques can be employed to assist in the evaluation of the atmosphere’s influence on the snowpack where days with similar atmospheric conditions are represented as a single synoptic weather type that influences the regional snowpack in an unique manner (Yarnal 1993). Examining the synoptic-scale atmospheric patterns associated with snow cover ablation events in the Great Lakes region will expand the knowledge needed in forecasting major melt events and can help the regional communities be better prepared, limiting losses.

8 1.3 Dissertation Organization This dissertation addresses three primary research objectives regarding the variability of snow in the Great Lakes region and its atmospheric forcing in a series of published and/or submitted manuscripts. Chapters 2-4 are written as self-contained studies with the structure, scope, and detail of a peer-reviewed scientific journal article. Chapter 2 focuses on the attribution of lake-effect snowfall trends to variability in the frequency and snowfall rates of particular synoptic-scale weather patterns. This chapter was published in the journal Climate Research (Suriano and Leathers 2017b). Chapter 3 develops a climatology of snow cover ablation events across the entire Great Lakes basin, detailing the seasonal cycle of ablation spatially and temporally, over a variety of ablation magnitudes. Chapter 3 was published in the journal Hydrological Processes (Suriano and Leathers 2017c). Chapter 4 addresses the synoptic-scale forcings of snow cover ablation events in the Great Lakes basin, identifying and analyzing the patterns that lead to ablation in the basin. This chapter is currently in review at the Journal of Applied Meteorology and Climatology, and was submitted in October 2017 (Suriano and Leathers, in review). The final chapter, Chapter 5, ties the results of the three studies together, speaks to the implications of the entire work, and offers avenues for further research.

REFERENCES

Ashley, S.T., and W.S. Ashley (2008) The storm morphology of deadly flooding events in the United States. International Journal of Climatology, 28: 493-503. doi:10.1002/joc.1554.

Bard, L., and D. Kristovich (2012) Trend Reversal in Lake Michigan Contribution to Snowfall. Journal of Applied Meteorology and Climatology, 51: 2038-2046. DOI: 10.1175/JAMC-D-12-064.1.

Barnett, T.P., Adam, J.C., and D.P. Lettenmaier (2005) Potential impacts of a warming climate on water availability in snow-dominated regions. Nature, 438: 303-309, doi:10.1038/nature04141.

Barry, M.J., Bowers, R. and F.A. DeSzalay (2004). Effects of hydrology, herbivory and sediment disturbance on plant recruitment in a Lake Erie coastal wetland. American Midland Naturalist, 151: 217-232, doi:10.1674/0003- 0031(2004)151[0217:EOHHAS]2.0.CO;2.

Barry, R., Prevost, M., Stein, J., and A.P. Plamondon (1990) Application of a Snow Cover Energy and Mass Balance Model in a Balsam Fir Forest. Water Resources Research, 26: 1079-1092.

Bednorz, E. (2009) Synoptic conditions for rapid snowmelt in the Polish-German lowlands. Theoretical and Applied Climatology, 97: 279–286, doi:10.1007/s00704-008-0063-z.

Bednorz, E., and J. Wibig (2015) Spatial distribution and synoptic conditions of snow accumulation and snow ablation in the West Siberian Plain. Quastiones Geographicae. doi:10.1515/quageo-2015-0029.

Burnett, A., Kirby, M., Mullins, H., and W. Patterson (2003) Increasing Great Lake- effect snowfall during the Twentieth Century: a regional response to global warming? Journal of Climate, 16: 3535-3541. DOI: 10.1175/1520- 0442(2003)016<3535:IGLSDT>2.0.CO;2.

Changnon, S.A., Changnon, D., and T. Karl (2006) Temporal and spatial characteristics of snowstorms in the contiguous United States. Journal of Applied Meteorology and Climatology, 45: 1141-1155. DOI: 10.1175/JAM2395.1.

10 Changnon, S.A. (2008) Assessment of Flood Losses in the United States. Journal of Contemporary Water Research and Education, 138: 38–44, doi:10.1111/j.1936-704X.2008.00007.x.

Chow-Fraser, P. (2005) Ecosystem response to changes in water level of Lake Ontario marshes: lessons from the restoration of Cootes Paradise Marsh. Hydrobiologia, 539: 189-204. DOI: 10.1007/s10750-004-4868-1.

Colbeck, S.C. (1973) Theory of metamorphism of wet snow. USA Cold Regions Research Engineering Laboratory, Research Report 313.

Colbeck, S.C. (1983a) Theory of Metamorphism of Dry Snow. Journal of Geophysical Research 88: 5475-5482.

Colbeck, S.C. (1983b) Snow Particle Morphology in the Seasonal Snow Cover. Bulletin of the American Meteorological Society 64: 602-609.

Colbeck, S.C. (1998) Sintering in a dry snow cover. Journal of Applied Physics 84: 4585-4589. DOI: 0.1063/1.368684.

Déry, S.J., and M.K. Yau (2002) Large-scale mass balance effects of blowing snow and surface sublimation. Journal of Geophysical Research: Atmospheres, 107: ACL 8-17, doi:10.1029/2001JD001251

Dewey, K.F. (1979) An Objective Forecast Method Developed for Lake Ontario Induced Snowfall Systems. Journal of Applied Meteorology and Climatology, 18: 787-793. DOI: 10.1175/1520-0450(1979)018<0787:AOFMDF>2.0.CO;2

Dyer, J.L., and T.L. Mote (2007) Trends in snow ablation over North America. International Journal of Climatology, 27: 739–748, doi:10.1002/joc.1426.

Eichenlaub, V. (1970) Lake effect snowfall to the lee of the Great Lakes: its role in Michigan. Bulletin of the American Meteorological Society, 51: 403-412. DOI: 10.1175/1520-0477(1970)051<0403:LESTTL>2.0.CO;2.

Eichenlaub, V. (1979). Weather and Climate of the Great Lakes Region. University of Notre Dame Press. Notre Dame, IN.

Ellis, A.W., and J. Johnson (2004) Hydroclimatic analysis of snowfall trends associated with the North American Great Lakes. Journal of Hydrometeorology, 5: 471-486. DOI: 10.1175/1525- 7541(2004)005<0471:HAOSTA>2.0.CO;2.

11 Fracz, A., and P. Chow-Fraser (2013) Impacts of declining water levels on the quantity of fish habitat in coastal wetlands of eastern Georgian Bay, Lake Huron. Hydrobiologia, 702: 151-169, doi:10.1007/s10750-012-1318-3.

Grover, E.K. and P.J. Sousounis (2002) The influence of large-scale flow on fall precipitation systems in the Great Lakes basin. Journal of Climate. 15: 1943- 1956.

Grundstein, A.J., and D.J. Leathers (1998) A case study of the synoptic patterns influencing midwinter snowmelt across the northern Great Plains. Hydrological Processes, 12: 2293–2305, doi:10.1002/(SICI)1099- 1085(199812)12:15<2293::AID-HYP797>3.0.CO;2-9.

Grundstein, A.J., and D.J. Leathers (1999) A spatial analysis of snow-surface energy exchanges over the northern Great Plains of the United States in relation to synoptic scale forcing mechanisms. International Journal of Climatology, 19: 489–511, doi:10.1002/(SICI)1097-0088(199904)19:5<489::AID- JOC373>3.0.CO;2-J.

Gunn, K.L.S. (1965) Measurements on new-fall snow. McGill University, Montreal, Stormy Weather Group Scientific Report MW-44.

Hartnett, J., Collins, J., Baxter, M., and D. Chambers (2014) Spatiotemporal Snowfall Trends in Central New York. Journal of Applied Meteorology and Climatology, 53:2685-2697. DOI: 10.1175/JAMC-D-14-0084.1.

Holroyd, E. (1971). Lake-effect cloud banks as seen from weather satellites. Journal of the Atmospheric Sciences. 28: 1165-

Hubley, R. (1954) The problem of short period measurements of snow ablation. Journal of Glaciology 16: 437-440. DOI: 10.3189/002214354793702542.

Kristovich, D., Laird, N., and M. Hjelmfelt (2003) Convective evolution across Lake Michigan during a widespread Lake-Effect snow event. Monthly Weather Review, 131: 643-655. DOI: 10.1175/1520- 0493(2003)131<0643:CEALMD>2.0.CO;2

Kunkel, K., Westcott, N., and D. Kristovich (2002) Assessment of potential effects of climate change on heavy lake-effect snowstorms near Lake Erie. Journal of Great Lakes Research, 28: 521-536. DOI: 10.1016/S0380-1330(02)70603-5.

12 Kunkel, K., Ensor, L., Palecki, M., Easterling, D., Robinson, D., Hubbard, K., and K. Redmond (2009) A new look at lake-effect snowfall trends in the Laurentian Great Lakes using a temporally homogeneous data set. Journal of Great Lakes Research, 35: 23-29. DOI: 10.1016/j.jglr.2008.11.003.

Leathers, D.J., Mote, T.L., Kluck, D., Kuivinen, K., and S. McFeeters (1993) Temporal characteristics of USA snowfall 1945-46 through 1984-85. International Journal of Climatology, 13: 65-76. DOI: 10.1002/joc.3370130105.

Leathers, D.J, and A.W. Ellis (1996) Synoptic mechanisms associated with snowfall increases to the Lee of Lakes Erie and Ontario. International Journal of Climatology, 10: 1117-1135. DOI: 10.1002/(SICI)1097- 0088(199610)16:10<1117::AID-JOC80>3.0.CO;2-4.

Leathers, D.J., Kluck, T., and S. Kroczynski (1998) The severe flooding event of 1996 across north-central Pennsylvania. Bulletin of the American Meteorological Society, 79: 785-797.

Leathers, D.J., Graybeal, D., Mote, T.L., Grundstein, A., and D.A. Robinson (2004) The Role of Airmass Types and Surface Energy Fluxes in Snow Cover Ablation in the Central Appalachians. Journal of Applied Meteorology, 43: 1887–1899, doi:10.1175/JAM2172.1.

Levia, D.F., and D.J. Leathers, 2011: Rain-induced snowmelt. In: Singh, V.P., P. Singh and U.K. Haritashya (Eds.), Encyclopedia of Snow, Ice, and Glaciers. Springer-Verlag, Heidelberg, Germany: 915-917, doi:10.1007/978-90-481- 2642_433.

Mazurkiewicz, A.B., Callery, D.G., and J.J. McDonnell (2008) Assessing the controls of the snow energy balance and water available for runoff in a rain-on-snow environment. Journal of Hydrology, 354: 1-14, doi:10.1016/j.jhydrol.2007.12.027.

McCabe, G., Clark, M., and L. Hay (2007). Rain-on-snow events in the western United States. Bulletin of American Meteorological Society, 88: 319-328, doi:10.1175/BAMS-88-3-31920.

Mellor, M. (1977). Engineering properties of snow. Journal of Glaciology, 19: 15-66.

Niziol, T., Snyder, W., and J. Waldstreicher (1995) Winter Weather Forecasting throughout the Eastern United States. Part IV: Lake Effect Snow. Weather Forecasting, 10:61-77. DOI: 10.1175/1520- 0434(1995)010<0061:WWFTTE>2.0.CO;2.

13 Norton, D., and S. Bolsenga (1993) Spatiotemporal trends in Lake Effect and continental snowfall in the Laurentian Great Lakes, 1951-1980. Journal of Climate, 6: 1943-1956. DOI: 10.1175/1520- 0442(1993)006<1943:STILEA>2.0.CO;2.

Quinn, F. (2002) Secular changes in Great Lakes water level seasonal cycles. Journal of Great Lakes Research, 28: 451–465.

Schmidlin, T. (1993) Impacts of severe winter weather during December 1989 in the Lake Erie snowbelt. Journal of Climate, 6: 759-767. DOI: 10.1175/1520- 0442(1993)006<0759:IOSWWD>2.0.CO;2.

Scott, R., and F. Huff (1996) Impacts of the Great Lakes on regional climate conditions. Journal of Great Lakes Research, 22: 845-863. DOI: 10.1016/S0380-1330(96)71006-7.

Steinman, A.D., Ogdahl, M.E., Weinert, M., Thompson, K., Cooper, M.J., and D.G. Uzarski (2012) Water level fluctuation and sediment-water nutrient exchange in Great Lakes coastal wetlands. Journal of Great Lakes Research, 38: 766- 775, doi:10.1016/j.jglr.2012.09.020.

Suriano, Z.J., and D.J. Leathers (2016) Twenty-first century snowfall projections within the Eastern Great Lakes region: Detecting the presence of a lake- induced snowfall signal in GCMs. International Journal of Climatology, 36: 2200-2209. DOI: 10.1002/joc.4488.

Suriano, Z.J., and D.J. Leathers (2017a) Synoptically classified lake-effect snowfall trends to the lee of Lakes Erie and Ontario. Climate Research, 74: 1-13, doi:10.3354/cr01480.

Suriano, Z.J., and D.J. Leathers (2017b) Synoptic climatology of lake-effect snowfall conditions in the eastern Great Lakes region. International Journal Climatology, 37: 4377-4389, doi:10.1002/joc.5093.

Suriano, Z.J., and D.J. Leathers (2017c) Spatiotemporal Variability of Great Lakes Basin Snow Cover Ablation Events. Hydrological Processes., 31: 4229-4237. doi:10.1002/hyp.11364.

Suriano, Z.J., and D.J. Leathers. Great Lakes basin snow cover ablation and synoptic- scale circulation. Submitted to the Journal of Applied Meteorology and Climatology. (October 2017).

14 Timmermans, S.T.A., Badzinski, S.S., and J.W. Ingram (2008) Associations between breeding marsh bird abundances and Great Lake hydrology. Journal of Great Lakes Research, 34: 351-364, doi:10.3394/0380- 1330(2008)34[351:ABBMBA]2.0.CO;2

Yarnal, B. (1993) Synoptic Climatology in Environmental Analysis: a primer. Belhaven Press (London), 195 pp.

15 Chapter 2

SYNOPTICALLY CLASSIFIED LAKE-EFFECT SNOWFALL TRENDS TO THE LEE OF LAKES ERIE AND ONTARIO

Suriano, Z.J. and D.J. Leathers (2017) Synoptically classified lake-effect snowfall trends to the lee of Lakes Erie and Ontario. Climate Research, 74: 1-13. DOI: 10.3354/cr01480.

2.1 Abstract

Recent research has indicated that snowfall in portions of the Great Lakes region subject to lake-effect snow has undergone a trend reversal, with snowfall declining in recent decades. This study examines the seasonal variability and trends specifically in synoptically classified lake-effect snow across the eastern Great Lakes region, and investigates the mechanisms responsible for observed changes. Using a synoptic climatological approach, days are identified where the synoptic-scale conditions are conducive to lake-effect snowfall and the associated snowfall is analyzed. Seven synoptic types over the November – March snowfall season are identified with characteristics of lake-effect conditions. Snowfall from these seven lake-effect synoptic types represents between 45-53% of the seasonal snowfall total along the eastern shores of Lakes Erie and Ontario with snowfall totals being highest during January. Lake-effect snowfall exhibits a 60-year increasing trend downwind of Lakes Erie and Ontario however through examination over shorter 30-year periods, a change in the trend of snowfall is occurring around 1980. While a true trend reversal is not

16 detected, lake-effect snowfall significantly increases from 1950-1979 before exhibiting no significant trend from 1980-2009. The inter-annual variability of seasonal lake-effect snowfall is highly related to the frequency of lake-effect synoptic types where an increase (decrease) in synoptic type occurrence leads to enhanced (diminished) lake-effect snowfall totals. Depending on the period examined, long-term changes of lake-effect synoptic types’ frequency and snowfall rates represent between 89-95% of the observed changes in lake-effect snow. Keywords: Great Lakes, synoptic classification, snowfall variability, climate change, lake-effect

2.2 Introduction Lake-effect snow (LES) is the enhancement of snowfall downwind of lakes occurring during the late fall and winter months. This is the result of increased convection due, in part, to a heightened temperature gradient between the relatively warm lake surface and overlying cold air masses (Eichenlaub 1979, Niziol et al. 1995, Kristovich et al. 2003). Lake-effect processes can produce twice as much snow in locations downwind of the lakes relative to locations further inland (Norton and Bolsenga 1993); this excessive snow can have substantial negative impacts on the surrounding region including on transportation, agriculture, economics, and natural habitats (Norton and Bolsenga 1993, Schmidlin 1993, Kunkel et al. 2002, Changnon et al. 2006). The additional snowfall can benefit some sectors of the economy such as recreation and winter-product sales (Schmidlin 1993, Kunkel et al. 2002). Numerous studies have examined trends in snowfall within the Great Lakes region (Leathers et al. 1993, Norton and Bolsenga 1993, Leathers and Ellis 1996, Grover and Sousounis 2002, Burnett et al. 2003, Ellis and Johnson 2004, Kunkel et al

17 2009, Bard and Kristovich 2012, Hartnett et al. 2014, Loveless et al. 2014), with some directly investigating LES. While a majority of these studies find increasing snowfall trends with time, Bard and Kristovich (2012) and Hartnett et al. (2014) respectively found a trend reversal with decreasing trends after the 1970-80’s downwind of Lake Michigan and in central New York State. Both studies noted increased air temperature as a possible forcing of declining snowfall. Typically, LES is defined as snowfall within an 80-100 km lake belt (Eichenlaub 1970, Dewey 1979, Norton and Bolsenga 1993, Scott and Huff 1996). Instead, here LES is defined based on snow falling during lake-effect synoptic patterns (Ellis and Leathers 1996, Leathers and Ellis 1996, Ellis and Johnson 2004; Suriano and Leathers 2017). This allows for the isolation of synoptic-driven LES amounts, not just snowfall that occurs within the lake belts, which could be derived from different systems. Loveless et al. (2014) did in part examine trends in snowfall accumulation from different storm types for Oneonta, NY, utilizing low-pressure storm tracks for differentiation of storms, but only examined one observation station. Synoptic classifications allow for daily weather events to be distinctly categorized, facilitating the evaluation of the atmosphere’s influence on the underlying land surface. Techniques include regionalization, circulation pattern classifications, and weather typing (Yarnal 1993, Sheridan and Lee 2014). Synoptic weather typing creates individual synoptic types that represent multiple days with similar atmospheric conditions, permitting researchers to relate synoptic-scale weather patterns to smaller- scale processes and track the frequency of their occurrence. This synoptic weather typing technique has proven effective in separating snowfall into lake-effect and non-

18 lake-effect types in previous research (Leathers and Ellis 1996, Ellis and Leathers 1996, Karmosky 2007, Suriano and Leathers 2017). This study utilizes an eigenvector-based synoptic weather typing classification technique (Kalkstein and Corrigan 1986) to generate a daily synoptic calendar for the eastern Great Lakes region. Snowfall events during the November through March snow season are isolated and analyzed based on the synoptic conditions that produced them (Suriano and Leathers 2017). A robust investigation of snowfall forced by lake- effect synoptic conditions, over the entire eastern Great Lakes region will help explain the changing snowfall trends unique to LES. Specifically, this study addresses the role of changing frequency of synoptic types and changing snowfall rates on varying snowfall trends during the 1949-50 through 2008-09 snowfall seasons. Additionally, while 21st century model projections suggest initial increases in lake-induced snowfall in the region followed by rapid decreases (Suriano and Leathers 2016), exploring the current trends in lake-effect snowfall will increase the understanding of potential factors that could drive changes in LES in the future. This information supplements and expands upon the existing literature on lake-effect snowfall, details mechanisms responsible for its change, and quantifies the relationship between the synoptic type and snowfall.

2.3 Data and Methodology

2.3.1 Datasets

Snow data come from a daily North American snowfall, snow depth, temperature, and liquid precipitation dataset interpolated onto a 1-degree grid for the period 1950-2009 (Dyer and Mote 2006, Kluver et al. 2016). The dataset for this study

19 has been updated from Dyer and Mote (2006). In the current version grid cell values are interpolated directly to a 1-degree lattice, whereas Dyer and Mote (2006) generated ¼ degree grids before interpolating to a 1-degree grid. Data originates as cooperative station observations in the United States from the TD3200 data from the National Centers for Environmental Information (NCEI, formerly NCDC, U.S. Department of Commerce 2003), and from observations archived in the Meteorological Service of Canada National Climate Data Center and Information Archive. A quality control method outlined in Robinson (1989) was applied to the station data, which omits unreasonable values and tests the internal consistency of the data. Data that passes the quality control are interpolated onto a 1-by-1 degree latitude-longitude grid using the Spheremap spatial interpolation procedure from the University of Delaware (Willmott et al. 1984, Willmott et al. 1985). Spheremap uses a modified version of Shepard’s inverse-distance algorithm of interpolation onto a two-dimensional Cartesian plane before projecting onto a spherical lattice. As station density varies with space and time, a variable search radius is used for each grid box (Kluver et al. 2016). Data are presently stored at Rutgers University (http://climate.rutgers.edu/snowcover/noaamelt/). For this study, snowfall observations are clipped to the eastern Great Lakes region bounded approximately by 40 to 46° north latitude and 73 to 85° west longitude, and to 1950 – 2009. Kluver et al. (2016) validated the interpolated snowfall dataset to station point observations by comparing it to the Community Collaborative Rain, Hail & Snow Network (CoCoRaHS; Cifelli et al. 2005) stations. Across North America, the interpolated daily snowfall dataset is approximately 2.0 cm lower than the CoCoRaHS

20 observations per event. In some regions within the Great Lakes basin, the interpolated dataset has slightly more negative biases approaching -5.0 cm. The underestimation found by Kluver et al. (2016) is partially attributed to the interpolation method, which smooths higher frequency signals through averaging of multiple station observations. The study also highlights that differences are additionally due to the inconsistent station locations of the cooperative and CoCoRaHS data. For the lake-effect criteria (see section 2.3.3), 850 hPa temperature and wind data are from the NCEP/NCAR Reanalysis Project and are acquired from NOAA- ERSL Physical Sciences Division (http://www.esrl.noaa.gov/psd/; Kalnay et al. 1996). The specific reanalysis product was selected over other (i.e. NARR, ERA) due primarily to its length of record. Lake-water temperature data are obtained from NOAA’s Great Lakes Environmental Research Laboratory (GLERL) (http://www.glerl.noaa.gov/). The GLERL data are selected over the NCEP/NCAR Reanalysis due to the surface temperature of the reanalysis likely not being indicative of lake-water temperature at its current spatial resolution (2.5°). At that resolution, in the regions corresponding to the lakes in the reanalysis product, both land surface and lake-water temperatures are factored into the final surface temperatures, likely artificially limiting the water-air temperature difference. Lake surface temperatures modeled by GLERL’s Large Lake Thermodynamics model are available as monthly averages on a per-year basis by lake. GLERL lake surface temperatures are found to agree well with monitored water surface temperatures (Croley II and Hunter 1994).

2.3.2 Synoptic Classification Daily synoptic weather types are developed for Buffalo, New York (WBAN#14733, 42.941° N, 78.732° W) from 1950-2009 using an eigenvector-based

21 approach similar to that of Kalkstein and Corrigan’s (1986) Temporal Synoptic Index (TSI). The TSI procedure has been employed in multiple studies, successfully classifying synoptic-scale weather types for a variety of applications (Davis 1991, Kalkstein et al. 1990, Ellis and Leathers 1996, Leathers and Ellis 1996, Siegert et al. 2016, among others). Four times-daily observations (0900, 1500, 2100, 0300 UTC) of meteorological variables are obtained from Kent State University (http://sheridan.geog.kent.edu/ssc.html), with data originating from the National Centers for Environmental Information (NCEI, Surface Data Hourly Global (DS3505) http://www.ncdc.noaa.gov). The specific four times-daily observation periods are selected to represent the standard 24-hour calendar day in the Eastern Time zone as effective as possible. Variables include temperature, dew point, atmospheric pressure, meridional and zonal wind components, and cloud cover. An unrotated principal components analysis (PCA) is conducted on the meteorological observations to reduce the original 24 variables (6 variables, 4-times daily) into a set of components that are linearly independent and ordered by explained variance (Kalkstein and Corrigan 1986). The PCA is conducted at the seasonal level for winter (DJF), spring (MAM), and autumn (SON). These seasonal classifications are chosen to limit the influence of the annual cycle on the synoptic patterns present within the generally temperate climatic region of the Great Lakes basin. Without the seasonal level analysis, much of the explained variance from the PCA would represent the annual cycle, limiting the effectiveness of the procedure. Seasonal PCA loadings of all components with eigenvalues greater than 1.0 are retained for further analysis. Five PCs where retained during each of the seasons. Multiplying the eigenvector of each component by the original data generates

22 component scores for each day, indicating the relative importance of each component for a given day. These daily component scores are clustered with an initial 20-cluster solution using within group-average linkage clustering to group days with similar component scores into individual clusters, or synoptic types. Within group-average linkage clustering is generally considered the most appropriate for synoptic weather typing due to its differentiation of extreme and more normal weather days into appropriate clusters (Kalkstein et al. 1987). This clustering method is typically found to minimize within-cluster variance while maximizing between-cluster variance. This results in the development of a calendar where each day is categorized as a particular synoptic type. All days with the same synoptic classifications are composited to produce maps depicting sea level pressure, surface air temperature, and 500 hPa geopotential height for that given synoptic type (NOAA-ESRL PSD, http://www.esrl.noaa.gov/psd/, Kalnay et al. 1996). If the characteristics of the synoptic types are similar, they are qualitatively combined using knowledge of local weather patterns to fine-tune the clustering’s autonomous nature (Siegert et al. 2016). It should be noted that snowfall or other forms of precipitation are not used to define the synoptic type. Furthermore, while synoptic types are generated for three meteorological seasons to reduce the impact of the annual cycle, analysis will be further constrained to the November-March snowfall season. Thus synoptic types from the winter (DJF) months are combined with autumn types occurring only in November, and spring types occurring only in March to produce the November-March season. The final result of the TSI is a 60-year, daily synoptic calendar for the November-March season. Daily snowfall from the 1x1 degree interpolated dataset is

23 combined with the daily synoptic calendar allowing for snowfall to be matched with the synoptic type occurring on the same calendar day. This permits analysis of the spatial relationships between snowfall and each synoptic type. It should be noted that the TSI is intended to produce classifications that have similar synoptic-scale features such that snowfall at the seasonal level and over larger spatial scales can be analyzed. The procedure is not designed to define micro-scale phenomena such as frictional convergence along the lakeshore or vorticities within the cloud bands associated with lake-effect snow. These smaller scale features can influence snowfall at a localized level, however this study is not focused on these meso- and micro-scales. Due to the focus on the synoptic-scale, conditions at Buffalo are sufficient to determine the synoptic scale situation for the lake-effect regions of the eastern Great Lakes despite being located within Lake Erie’s basin, as opposed to Lake Ontario’s.

2.3.3 Lake-effect Classification Specific emphasis is placed on the snowfall associated with lake-effect synoptic types. A synoptic type is considered lake-effect based on criteria developed in Suriano and Leathers (2017). Lake-effect synoptic conditions are considered as:

1) wind flow at 850 hPa provides favorable fetch over the lakes (ranging

NNW-SSW flow),

2) 850 hPa winds surpass 5 m s-1 nor exceed 20 m s-1,

3) directional wind shear between the surface (2-m) and 850 hPa is less than

30°, and

24 4) the temperature difference, or lapse rate, between the lake-water and 850

hPa is at least 13°C.

Average conditions for the lakes and synoptic types are used to calculate the lake-water to 850 hPa temperature difference. The 850 hPa temperatures within the grid cells directly above the lakes are the reanalysis composite of all snowfall producing occurrences of each synoptic type. As lake-water temperature data are only available monthly, the lake-water temperatures are the average conditions for the months and years corresponding to when each individual synoptic type actually occurred. The calculation was initially conducted individually for both Lakes Erie and

Ontario, however in all cases when one lake had sufficient instability (> 13°C) they both did.

Lake ice has been shown to influence the formation of lake-effect snow

(Burnett et al. 2003, Gerbush et al. 2008, Wang et al. 2011, Vavrus et al. 2013). In this study, the impact of lake ice is indirectly accounted for. Days with substantial lake ice coverage are included in the analysis of snowfall associated with individual synoptic types just as days without lake ice are. Thus total seasonal snowfall and seasonal snowfall rates include days where no, or reduced, lake-effect snow was observed due to the presence of lake ice.

25 2.4 Results and Discussion

2.4.1 Seasonal Snowfall Distribution

The TSI procedure for Buffalo, NY results in the identification of 43 synoptic types during the three seasons (autumn, winter, spring) spanning the November-March snowfall season over 1950-2009. Of these types, seven met the criteria for LES development and their sea-level pressure fields are depicted in Figure 2.1. Further information on the surface meteorological characteristics of the seven synoptic weather types, including temperature, dewpoint, winds, sea-level pressure, and cloud cover, can be viewed in Table 2.1. The frequencies of the seven lake-effect synoptic types are combined into a single grouping that contains all days during November- March when any of the seven individual synoptic types occurred, corresponding to 2374 total days. On average, lake-effect synoptic types occur approximately 40 days each season. Of the 2374 lake-effect synoptic type days, 2307 (97%) produced snowfall. For the purposes of this study, the snow produced by these lake-effect synoptic types is considered to be LES, however the limitations discussed in section 3.4 should be noted. The average seasonal LES distribution is shown in Figure 2.2a. Higher 1950- 2009 seasonally averaged LES totals in excess of 185 and 120 cm year-1 exist downwind of Lake Ontario and Lake Erie, respectively (σ = 73.2, σ = 45.6). The spatial distribution of snowfall per event is similar, with peak snowfall rates downwind of the lakes (not shown). Comparing snowfall totals of the lake-effect types to snowfall from all types, LES comprises 45-53% of the seasonal snowfall total in the grid boxes immediately downwind of the lakes (Figure 2.2b). This is similar to Norton

26 and Bolsenga (1993) who found that LES can double the amount of snow received downwind of the lakes over a typical season. Intra-seasonal LES is additionally inspected. For all five months, average LES is greater downwind of Lake Ontario than downwind of Lake Erie, likely due to the orientation of the lakes and to other physical differences highlighted below. By examining the ratio of monthly LES to total seasonal LES (Table 2.2), more information can be gained. For both Lakes Ontario and Erie, a majority of LES downwind of the lakes occurs during the month of January, respectively contributing 36% and 33% of the seasonal LES. Also for both lakes, January is followed by December, February, November, and March in decreasing order of respective amounts of LES produced. During November and December, LES downwind of Lake Erie makes up a higher percentage of the seasonal LES total than for LES associated with Lake Ontario. During January and February, the opposite occurs such that LES downwind of Lake Erie makes up a smaller percentage of the seasonal LES total than the monthly percentages associated with Lake Ontario. This switch over is likely, in part, due to the temperature of the Lakes and the likelihood of lake ice development. During November, Lake Erie is substantially warmer than Lake Ontario due to its shallower depth and lower latitude. This can result in a larger lake-water to 850 hPa temperature difference and is indicative of stronger convective instability and relatively increased LES. By January and February, Lake Erie is commonly colder than Lake Ontario due to its shallower depth and typically has a much higher percentage of ice cover (Assel et al. 2003). This increased ice coverage on Lake Erie relative to Lake Ontario lessens the convective instability and can greatly reduce LES (Cordeira and Laird 2008, Gerbush et al. 2008).

27 2.4.2 Temporal Snowfall Trends To assess 1950-2009 trends in LES associated with Lake Ontario and Lake Erie collectively, LES within the grid cells that most closely align with the lake belts defined in the literature are isolated (Eichenlaub 1970, Dewey 1979, Norton and Bolsenga 1993, Scott and Huff 1996). Grid cells used for this analysis are: 41.5°N 80.5°W, 42.5°N 79.5°W, 42.5°N 78.5°W, 43.5°N 76.5°W, and 43.5°N 75.5°W (Figure 2.3). The collective LES from both lakes exhibits a long-term increasing trend of 0.81 cm year-1 (p < 0.05) (Figure 2.4a). Leathers and Ellis (1996) found seasonal snowfall increases from November-March of approximately 0.8-2.0 cm year-1 downwind of Lakes Erie and Ontario from 1931-1990 using individual station data. Burnett et al. (2003) noted a 1.5 cm increase in LES stations across the Great Lakes region compared to non-lake-effect stations from 1931-2001; however, this did include lake-effect sites associated with Lakes Michigan and Superior. Kunkel et al. (2009) analyzed snowfall associated with stations deemed homogeneous in the Great Lakes region, finding an increase of 0.6 cm year-1, however similar to Burnett et al. (2003), trends were examined across all of the lakes. Hartnett et al. (2014) found a 1.16 + 0.31 cm year-1 increase in snowfall from 1931-2012 in central New York State however, their study did not distinguish lake-effect from non-lake-effect snow, utilizing stations outside those generally considered lake-effect impacted regions.

While trends in this study are broadly similar to those in the literature, it should be noted that the other studies utilized different periods of record and different spatial scales than this study. Using only the stations identified by Kunkel et al. (2009) within the interpolated dataset, the 0.81 cm year-1 increase in LES is reduced to 0.36 cm year- 1 (p = 0.06). Only three stations identified in Kunkel et al. (2009) fall within the grid

28 cells analyzed in this research and the smaller trend could be a result of the station locations. Recent research has indicated snowfall in portions of the Great Lakes region has undergone a trend reversal where snowfall steadily increased until the 1970-80s before declining thereafter (Bard and Kristovich 2012, Hartnett et al. 2014). In response to these conclusions, trends in LES are additionally examined over two equal 30-year periods (1950-1979, 1980-2009), and by computing a 21-year moving average of the trend with a 1-year window. Examining the trend in Lake Ontario and Lake Erie collective LES as two distinct periods (grid cells discussed previously, Figure 2.3), there is a stop to significant trends around 1980 (Figure 2.4b). From 1950-1979, LES increases by 3.24 cm year-1 (p < 0.01). However from 1980-2009, LES does not exhibit a significant trend. This is further supported by the 21-year moving average LES trends calculated for the 1960-1999 seasons (Figure 2.4c). LES increases during the first 20 seasons through 1979 (x = 1.83, σ = 1.17). In 1980, the 21-year moving average LES trend becomes negative and stays negative until 1993 (x = -0.81, σ = 0.63). During the final seven seasons (1993-1999), the LES trends hover around zero but are variable (x = 0.18, σ = 0.77). While a true trend reversal is not detected in LES downwind of Lakes Ontario and Erie, LES does appear to be behaving non-linearly over the 60-year period with a halt to increasing snowfall trends around 1980.

2.4.3 Causes of Snowfall Variability and Trends Air temperatures are certainly related to snowfall trends (Bard and Kristovich 2012, Hartnett et al. 2014, among others). However, we hypothesize that the frequency of lake-effect synoptic types and the rate of snowfall per day (snowfall intensity) play dominant roles in explaining snowfall variability (Leathers and Ellis 1996, Ellis and

29 Johnson 2004), particularly in explaining the apparent change in LES trend after 1980 (Figure 2.3). Examining the time series of lake-effect synoptic types’ seasonal frequency, no long-term trend exists. However similar to LES, a change in trend is apparent when the long-term trend is examined over two equal 30-year periods (Figure 2.5a). From 1950-1979, the frequency of lake-effect synoptic types increased by approximately 0.43 days year-1 (p < 0.05). From 1980 onwards, no significant trend in lake-effect synoptic types’ frequency is detected. To determine the effect of lake-effect synoptic types’ frequency on LES variability, simple linear regression analysis is conducted. The seasonal frequency of lake-effect synoptic types is significantly correlated to the average LES from Lakes Ontario and Erie (0.802, p < 0.01) (Figure 2.5b). A similar relationship is noted when the same analysis is conducted on the two time series after they are de-trended (0.834, p < 0.01). De-trending is conducted by calculating the differences in the original data from the linear regression line. This suggests that the number of lake-effect synoptic types occurring each season can explain a large percentage of the inter-annual variability of LES associated with Lakes Ontario and Erie, independent of the long- term trend. Changes in the frequency of lake-effect synoptic types are likely the dominant force behind the apparent change in trend of LES. To understand the magnitude of LES changes caused by lake-effect synoptic types’ frequency changes, a snowfall term is linearly extrapolated (Equation 1).

(1) SF_Freq = freq_trend x SF x years

The ‘freq_trend’ term is the trend in lake-effect synoptic types’ frequency in days year-1, ‘SF’ is the average daily LES in cm day-1 by grid cell, and ‘years’ are the

30 number of years in the analyzed period. This calculation is conducted for each cell in the study region. If there is no trend in lake-effect synoptic types’ frequency, the resulting calculated snowfall value is zero. Figure 2.6(a) shows the calculated snowfall changes due to lake-effect synoptic type frequency changes from 1950-1979. Across the entire region, lake-effect synoptic types’ frequency changes result in an increase in

LES.

The magnitude of LES changes due to the impact of a changing rate of snowfall (snowfall intensity) is also assessed through a similar linearly extrapolated calculated snowfall term (Equation 2).

(2) SF_Int = int_trend x freq x years

The ‘int_trend’ term is the trend in snowfall intensity in cm day-1 year-1 for each grid cell, determined by regressing the average LES per day against time. ‘Freq’ is the average number of days of a lake-effect synoptic type, and ‘years’ are the number of years in the analyzed period. Similar to equation one, if there is no trend in the snowfall intensity, the calculated snowfall value for that grid cell is zero. Figure 2.6(b) depicts the predominately positive calculated snowfall changes due to changes in snowfall intensity during 1950-1979.

The addition of the snowfall changes due to lake-effect synoptic type frequency and snowfall intensity changes, account for large snowfall increases of approximately 150 cm to the east of Lake Ontario over the 1950-1979 period (Figure

2.6c). In addition, substantial snowfall increases of 60-75 cm are also found along the

31 northeastern shores of Lake Erie over this period. Figure 2.6(d) depicts the linearly extrapolated LES changes observed in the region. The combined frequency and intensity snowfall changes account for 94.5% of the observed snowfall changes downwind of the lakes.

The same process is conducted for the second time period, 1980-2009. Figure

2.7(a-d) is the same as Figure 2.6(a-d), but for this later 30-year period. During this time period, lake-effect synoptic types exhibit a decreasing trend, thus the calculated snowfall changes due to frequency are negative (Figure 2.7a). Calculated snowfall changes due to snowfall intensity changes vary by grid cell, with positive and negative changes dispersed across the region (Figure 2.7b). Figure 2.7(c) shows the combined calculated snowfall change from lake-effect synoptic type frequency and snowfall intensity changes. Compared to 1950-1979, these two factors collectively result in relatively small increases or decreases in snowfall in 1980-2009. Comparing these changes to the linearly extrapolated observed snowfall change in the region (Figure

2.7d), the combined frequency and intensity driven snowfall changes account for

89.4% of the observed snowfall change downwind of the lakes.

During both time periods, calculated snowfall changes are lower than what is observed. This suggests there could be other factors influencing LES trends in the region, with their influence being more apparent during 1980-2009. While synoptic type frequency and snowfall intensity influences appear dominant, other factors such as the effect of air temperature changes (Kunkel et al. 2009, Bard and Kristovich 2012,

Hartnett et al. 2014), changes in ice cover (Assel et al. 2003), or intra-synoptic type

32 changes including humidity, winds, and other characteristics may also be contributing to changes in LES totals.

2.4.4 Limitations Beyond the negative bias found in the interpolated snowfall observations (Kluver et al. 2016), two other limitations exist that should be considered when examining the results and conclusions of this study. First, lake-effect snow events are not necessarily confined to a standard 24-hour day. It is possible for a single lake- effect snowfall event to occur during parts of two consecutive days. Second, it is important to note that the interpolated data originate from Cooperative Observer Program (COOP) observations in the United States. It is documented in the literature that time of observation by COOP observers has varied during the 20th century and can vary by location (Karl et al. 1986, Kunkel et al. 2007). Stations in close geographic proximity but with different time of observations may cause differences in recorded snowfall. This potential bias could result in a portion of the “daily” snowfall for certain stations within a grid cell of the interpolated dataset to be assigned to the synoptic type occurring the following day. Both limitations may cause a dampened lake-effect snowfall signal directly downwind of the lakes, and an enhanced lake-effect snowfall signal in regions that typically receive small amounts of lake-effect snowfall such as those further away from the lakes. This should be considered when examining the spatial distribution of lake-effect snowfall. However despite these limitations, the snowfall from the identified lake-effect synoptic types represents between 45-53% of the total seasonal snowfall directly downwind of the lakes, which is in line with the values seen in the literature (Braham and Dungey, 1984, Kelly, 1986, Norton and Bolsenga, 1993).

33 2.5 Summary and Conclusions This study examined the November-March seasonal snowfall associated with lake-effect synoptic types in the eastern Great Lakes region, and the mechanisms responsible for their change during the period 1950-2009. A synoptic climatological approach was utilized to identify and isolate synoptic-scale weather types consistent with lake-effect snowfall patterns that regularly occur in the region (Ellis and Leathers 1996, Leathers and Ellis 1996, Ellis and Johnson 2004, Suriano and Leathers 2017). Of the synoptic types classified, seven were identified as lake-effect synoptic patterns by their 850 hPa winds, directional wind shear with height, and large 850 hPa to lake- surface temperature lapse rates. Snowfall occurring on days with lake-effect synoptic types is considered lake-effect snow (LES) and was examined at both seasonal and monthly time scales across the region. As expected, the spatial distribution of average LES revealed higher totals downwind of both Lakes Ontario and Erie, with lesser totals further inland. Snowfall is most prevalent during the month of January for regions downwind of both Lakes, however differences between the Lakes in the relative monthly contributions to the seasonal LES total are detected. During the second half of the winter season, the ratio of monthly to total LES is smaller for Lake Erie compared to Lake Ontario, indicating the potential influence of lake ice on snowfall (Cordeira and Laird 2008, Gerbush et al. 2008). A linear trend analysis for winter seasons 1949-50 through 2008-09 revealed that LES increased by 0.81 cm year-1 collectively downwind of Lakes Ontario and Erie. However, breaking the long-term trend into two 30-year periods reveals a more interesting history. A change in trend of LES is detected downwind of the Lakes around 1980. While a true trend reversal in not detected in the region as seen in snowfall in other portions of the Great Lakes region (Bard and Kristovich 2012,

34 Hartnett et al. 2014), LES significantly increased by 3.24 cm year-1 from 1950-1979, before no longer exhibiting a significant trend from 1980-2009. Trends remain robust after a bootstrapping analysis. Changes in the seasonal frequency of lake-effect synoptic types and the rate of snowfall (snowfall intensity) were hypothesized to be the primary drivers for changes in LES (Leathers and Ellis 1996). Similar to LES, the seasonal frequency of lake- effect synoptic types also exhibits a change in trend around 1980. After de-trending both variables, lake-effect synoptic type frequency is strongly correlated to seasonal LES downwind of Lakes Ontario and Erie, explaining over 68% of the variance in LES. This suggests the frequency of lake-effect synoptic types are likely the dominant force behind the apparent change in trend of LES. The magnitude of the snowfall changes based on synoptic type frequency and snowfall intensity changes are also examined. Over both periods examined (1950-1979, 1980-2009) this calculated snowfall change due to frequency and intensity changes represented 94.5% and 89.4% of the observed snowfall changes, respectively. This leads to the conclusion that changes in LES over these periods are predominately caused by changes in the frequency of lake-effect synoptic types and snowfall intensity within these types. However for both periods, calculated snowfall changes are less than the observed changes, with the underrepresentation of observed changes being more pronounced during 1980-2009. This suggests there could be additional factors influencing LES trends in the region, with their influence being more pronounced in more recent times. Additional influences could include intra-synoptic type variability, changes in lake ice cover, increasing air temperatures, or other unknown factors.

35 Comparing some of this study’s results to Leathers and Ellis (1996) and Ellis and Leathers (1996), a number of similarities exist. In their studies, five synoptic types for Syracuse, NY were identified as LES producers compared to the seven found in this study. This difference is likely the result of a larger study period used in this study, and with the meteorological season-based Temporal Synoptic Index (TSI) methodology used here, as opposed to the TSI being conducted over a single five- month period in Leathers and Ellis (1996) and Ellis and Leathers (1996). Furthermore, during the 1950/51 – 1981/82 period, Leathers and Ellis (1996) reported a majority of snowfall changes were due to synoptic type frequency and snowfall intensity changes. Just as in this study, changes in synoptic type frequency and snowfall intensity also accounted for a majority (94.5%) of observed snowfall changes from 1950-1979. However, here we find this relationship between changing frequency and intensity, and LES persists during times without significant snowfall increases (1980-2009), furthering the novel aspects of the study. Snowfall is an important component of the hydrologic cycle within the eastern Great Lakes region, influencing water resources, the economy, transportation, winter recreation, and natural habitats. Particularly, LES plays a pivotal role, resulting in 45- 53% of the seasonal snowfall totals downwind of the lakes. The results of this study further the understanding of LES seasonal variability and trends, and what predominately influences LES. Despite LES no longer exhibiting a strong increasing trend in the region, the frequency of lake-effect synoptic types and the rate, or intensity, of snowfall, appear to be the major drivers of LES changes. Future work will expand this analysis, particularly investigating the driving force(s) of lake-effect snowfall intensity changes, with an emphasis on intra-synoptic type variability (the

36 changing character of synoptic types), and changing lake surface and 850 hPa air temperature differences.

2.6 Acknowledgements The authors would like to acknowledge the financial support received from the National Oceanic and Atmospheric Administration Climate Program Office (NA14OAR4310207), and the Dr. John R. Mather Graduate Research Award from the University of Delaware. The authors would like to thank Thomas Estilow from Rutgers University and Scott Sheridan from Kent State University for their assistance in data acquisition. Gratitude is also extended to David Robinson from Rutgers University, Gina Henderson from the United States Naval Academy, and Tracy DeLiberty and Sara Rauscher from the University of Delaware for their helpful comments and suggestions. The authors would also like to thank the anonymous reviewers for their feedback.

REFERENCES

Assel R, Cronk K, Norton D (2003) Recent Trends in Laurentian Great Lakes Ice Cover. Clim Change 57:185-204.

Bard L, Kristovich D (2012) Trend Reversal in Lake Michigan Contribution to Snowfall. J Appl Meteorol Clim 51:2038-2046. DOI: 10.1175/JAMC-D-12- 064.1.

Burnett A, Kirby M, Mullins H, Patterson W (2003) Increasing Great Lake-effect snowfall during the Twentieth Century: a regional response to global warming? J Climate 16:3535-3541. DOI: 10.1175/1520- 0442(2003)016<3535:IGLSDT>2.0.CO;2.

Changnon S, Changnon D, Karl T (2006) Temporal and spatial characteristics of snowstorms in the contiguous United States. J Appl Meteorol Clim 45:1141- 1155. DOI: 10.1175/JAM2395.1.

Cifelli R, Doesken N, Kennedy P, Carey LD, Rutledge SA, Gimmestad C, Depue T (2005) The Community Collaborative Rain, Hail, and Snow Network: Informal Education for Scientists and Citizens. B Am Meteorol Soc 86:1069- 1077.

Cordeira JM, Laird NF (2008) The Influence of Ice Cover on Two Lake-Effect Snow Events over Lake Erie. Mon Weather Rev 136:2747–2763. DOI: 10.1175/2007MWR2310.1.

Croley II TE, Hunter TS (1994) Great Lakes Monthly Hydrologic Data. NOAA Data Report ERL GLERL, National Technical Information Service, Springfield, Virginia, 22161.

Davis R (1991) A Synoptic Climatological Analysis of Winter Visibility Trends in the Mideastern United-States. Atmos Environ, B Urban Atmos 25:165–175. DOI: 10.1016/0957-1272(91)90052-G.

Dewey KF (1979) An Objective Forecast Method Developed for Lake Ontario Induced Snowfall Systems. J Appl Meteorol 18:787-793. DOI: 10.1175/1520- 0450(1979)018<0787:AOFMDF>2.0.CO;2

Dyer JL, Mote T (2006) Spatial variability and trends in observed snow depth over North America. Geophys Res Lett 33:L16503. DOI: 10.1029/2006GL027258.

38 Eichenlaub V (1970) Lake effect snowfall to the lee of the Great Lakes: its role in Michigan. B Am Meteorol Soc 51:403-412. DOI: 10.1175/1520- 0477(1970)051<0403:LESTTL>2.0.CO;2.

Ellis AW, Leathers DJ (1996) A synoptic climatological approach to the analysis of lake-effect snowfall: Potential forecasting applications. Weather Forecast 11:216-22. DOI: 10.1175/1520-0434(1996)011<0216:ASCATT>2.0.CO;2

Ellis AW, Johnson J (2004) Hydroclimatic analysis of snowfall trends associated with the North American Great Lakes. J Hydrometeorol 5:471-486. DOI: 10.1175/1525-7541(2004)005<0471:HAOSTA>2.0.CO;2.

Gerbush MR, Kristovich DAR, Laird NF (2008) Mesoscale Boundary Layer and Heat Flux Variations over Pack Ice–Covered Lake Erie. J Appl Meteorol and Clim 47:668–682. DOI: 10.1175/2007JAMC1479.1.

Grover E, Sousounis P (2002) The Influence of Large-Scale Flow on Fall Precipitation Systems in the Great Lakes Basin. J Clim 15:1943-1956. 10.1175/1520- 0442(2002)015<1943:TIOLSF>2.0.CO;2

Hartnett J, Collins J, Baxter M, Chambers D (2014) Spatiotemporal Snowfall Trends in Central New York. J Appl Meteorol Clim 53:2685-2697. DOI: 10.1175/JAMC-D-14-0084.1.

Kalkstein LS, Corrigan P (1986) A synoptic climatological approach for geographical analysis: assessment of sulfur dioxide concentrations. Ann Assoc Am Geogr 76:381-395. DOI: 10.1111/j.1467-8306.1986.tb00126.x

Kalkstein LS, Tan G, Skindlov JA (1987) An Evaluation of Three Clustering Procedures for Use in Synoptic Climatological Classification. J Clim Appl Meteorol 26: 717-730

Kalkstein L, Dunne P, Vose R (1990) Detection of Climatic-Change in the Western North-American Arctic Using a Synoptic Climatological Approach. J Clim 3:1153–1167. DOI: 10.1175/1520-0442(1990)003<1153:DOCCIT>2.0.CO;2.

Kalnay E, Kanamitsu M, Kistler R et al. (1996) The NCEP/NCAR 40-year reanalysis project. B Am Meteorol Soc 77:437-471.

Karl TR, Williams CN, Young PJ, Wendland WM (1986) A Model to Estimate the Time of Observation Bias Associated with Monthly Mean Maximum, Minimum, and Mean Temperatures for the United States. J Appl Meteorol Clim 25:145-160. DOI: 10.1175/1520- 450(1986)025<0145:AMTETT>2.0.CO;2

39 Karmosky C (2007) Synoptic Climatology of Snowfall in the Northeastern United States: An Analysis of Snowfall Amounts from Diverse Synoptic Weather Types. Master’s Thesis, University of Delaware, Newark, DE.

Kluver D, Mote T, Leathers D, Hendersen GR, Chan W, Robinson DA (2016) Creation and validation of a comprehensive 1 degree by 1 degree daily gridded North American dataset for 1900 to 2009: Snowfall. J Atmos Ocean Tech 33:857-871. DOI: 10.1175/JTECH-D-15-0027.1

Kristovich D, Laird, N., Hjelmfelt, M (2003) Convective evolution across Lake Michigan during a widespread Lake-Effect snow event. Mon Weather Rev 131:643-655. DOI: 10.1175/1520-0493(2003)131<0643:CEALMD>2.0.CO;2

Kunkel K, Westcott N, Kristovich D (2002) Assessment of potential effects of climate change on heavy lake-effect snowstorms near Lake Erie. J Great Lakes Res 28:521-536. DOI: 10.1016/S0380-1330(02)70603-5.

Kunkel K, Ensor L, Palecki M, Easterling D, Robinson D, Hubbard K, Redmond K (2009) A new look at lake-effect snowfall trends in the Laurentian Great Lakes using a temporally homogeneous data set. J Great Lakes Res 35:23-29. DOI: 10.1016/j.jglr.2008.11.003.

Leathers D, Mote T, Kluck D, Kuivinen K, McFeeters S (1993) Temporal characteristics of USA snowfall 1945-46 through 1984-85. Int J Climatol 13:65-76. DOI: 10.1002/joc.3370130105.

Leathers D, Ellis AW (1996) Synoptic mechanisms associated with snowfall increases to the Lee of Lakes Erie and Ontario. Int J Climatol 10:1117-1135. DOI: 10.1002/(SICI)1097-0088(199610)16:10<1117::AID-JOC80>3.0.CO;2-4.

Loveless D, Godek M, Bleckman J (2014) Developing a Climatology of snowfall events in Oneonta, New York. Northeastern Geoscience 32:44-55.

Niziol T, Snyder W, Waldstreicher J (1995) Winter Weather Forecasting throughout the Eastern United States. Part IV: Lake Effect Snow. Weather Forecast 10:61- 77. DOI: 10.1175/1520-0434(1995)010<0061:WWFTTE>2.0.CO;2.

Norton D, Bolsenga S (1993) Spatiotemporal trends in Lake Effect and continental snowfall in the Laurentian Great Lakes, 1951-1980. J Climate 6:1943-1956. DOI: 10.1175/1520-0442(1993)006<1943:STILEA>2.0.CO;2.

Robinson, D (1989) Construction of a United States historical snow database. Proceedings of the 45th Eastern Snow Conference, 50-59.

40 Sheridan SC, Lee CC (2014) Synoptic Climatology. Oxford Bibliographies in geography. New York: Oxford University Press. DOI: 10.1093/OBO/9780199874002-0088.

Siegert CM, Leathers DJ, Levia DF (2016) Synoptic typing: interdisciplinary application methods with three practical hydroclimatological examples. Theor Appl Climatol DOI: 10.1007/s00704-015-1700-y.

Schmidlin T (1993) Impacts of severe winter weather during December 1989 in the Lake Erie snowbelt. J Clim 6:759-767. DOI: 10.1175/1520- 0442(1993)006<0759:IOSWWD>2.0.CO;2.

Scott R, Huff F (1996) Impacts of the Great Lakes on regional climate conditions. J Great Lakes Res 22:845-863. DOI: 10.1016/S0380-1330(96)71006-7.

Suriano ZJ, Leathers DJ (2016) Twenty-first century snowfall projections within the Eastern Great Lakes region: Detecting the presence of a lake-induced snowfall signal in GCMs. Int J Climatol 36:2200-2209. DOI: 10.1002/joc.4488.

Suriano ZJ, Leathers DJ (2017) Synoptic Climatology of Lake Effect Snowfall Conditions in the Eastern Great Lakes Region. Int J Climatol 37: 4377-4389. DOI: 10.1002/joc.5093.

U.S. Department of Commerce (2003) TD-3200 Surface Summary of the Day, National Climatic Data Center, NOAA. [Available online: http://www.ncdc.noaa.gov].

Vavrus S, Notaro M, Zarrin A (2013) The Role of Ice Cover in Heavy Lake-Effect Snowstorms over the Great Lakes Basin as Simulated by RegCM4. Mon Weather Rev 141:148–165. DOI: 10.1175/MWR-D-12-00107.1.

Wang J, Bai X, Hu H, Clites A, Colton M, Lofgren B (2011) Temporal and Spatial Variability of Great Lakes Ice Cover, 1973–2010. J Clim 25:1318–1329. DOI: 10.1175/2011JCLI4066.1.

Willmott C, Rowe C, Philpot W (1984) Spheremap. Center for Climatic Research, Department of Geography, University of Delaware, Newark DE.

Willmott C, Rowe C, Philpot W (1985) Small-Scale Climate Maps: A Sensitivity Analysis of Some Common Assumptions Associated with Grid-point Interpolation and Contouring. American Cartographer. 12:5-16.

Yarnal B (1993) Synoptic Climatology in Environmental Analysis: a primer. Belhaven Press (London). 195 pg.

41 Table 2.1.: Average surface meteorological characteristics for the seven lake-effect synoptic weather types at 0900, 1500, 2100, 0300 UTC.

Synoptic Temperature Dewpoint Sea-level Wind Wind Cloud Type (°C) (°C) pressure speed direction cover (hPa) (ms-1) (°) (/10) 0900 UTC - WNW-1 1.4 -1.8 1010 4.6 284 8.7 W-1 2.3 -1.2 1020 3.1 266 8.4 SW-1 -9.8 -13.4 1014 3.9 220 7.6 WSW-1 -5.9 -8.9 1008 8.7 246 9.2 W-2 -6.7 -9.8 1011 4.4 271 8.6 WSW-2 -9.1 -12.7 1025 4.0 254 6.8 NW-1 -7.0 -10.7 1016 3.1 329 6.5

1500 UTC WNW-1 1.4 -2.7 1013 5.6 293 8.6 W-1 2.8 -1.7 1024 3.9 285 8.2 SW-1 -7.6 -11.6 1012 5.9 213 9.1 WSW-1 -6.8 -10.0 1011 9.2 253 9.2 W-2 -7.1 -10.8 1014 5.2 278 8.2 WSW-2 -7.6 -11.9 1027 4.3 247 6.7 NW-1 -5.7 -11.5 1019 4.2 333 6.3

2100 UTC WNW-1 1.8 -3.5 1014 5.6 292 8.7 W-1 3.6 -2.1 1024 4.1 280 7.8 SW-1 -4.8 -8.8 1009 7.1 224 9.7 WSW-1 -6.6 -10.6 1013 9.8 258 9.1 W-2 -6.6 -11.5 1015 5.9 276 7.7 WSW-2 -5.0 -10.6 1026 5.0 242 7.2 NW-1 -3.5 -11.4 1019 4.9 308 5.7

0300 UTC WNW-1 0.0 -4.2 1016 3.8 290 7.5 W-1 1.2 -2.6 1025 1.2 279 6.8 SW-1 -5.2 -8.4 1009 6.6 232 9.6 WSW-1 -8.3 -12.1 1016 7.7 263 8.8 W-2 -8.5 -12.6 1018 4.6 276 7.0 WSW-2 -6.6 -10.8 1026 3.3 226 6.9 NW-1 -6.5 -11.5 1021 2.8 296 3.5

42 Table 2.2.: Average monthly lake-effect snow (cm), and ratio of monthly to seasonal lake-effect snowfall total (%) for Lake Erie and Lake Ontario from 1950- 2009. These defined regions are shown in Figure 2.3.

Lake Erie Lake Ontario Snowfall Percent of Snowfall Percent of (cm) Total (%) (cm) Total (%)

Nov 13.1 12.9 16.5 9.4 Dec 28.0 27.5 43.5 24.8 Jan 34.2 33.6 64.3 36.7 Feb 21.9 21.5 44.9 25.6 Mar 4.5 4.4 6.0 3.4

Season 101.7 175.2

Figure 2.3.: Map of the study region. The grid cells used for defining Lake Ontario and Lake Erie lake-effect snowfall are shown in blue. Buffalo, NY is labeled.

Figure 2.4.: Lake-effect snowfall (cm) and snowfall trends (cm year-1) associated with Lakes Ontario and Erie for (a) 1950-2009 using the initial snowfall amounts (black) and snowfall only from stations identified in Kunkel et al. (2009) (grey), (b) initial snowfall amounts during two 30-year periods corresponding to 1950-1979 (black) and 1980-2009 (grey), and (c) a 21- year moving trend in initial lake-effect snowfall from 1960-1999 at an interval of 1 year.

Figure 2.5.: Lake-effect synoptic type frequency (days) and trend (days year-1) during two 30-year periods of 1950-1979 (black) and 1980-2009 (grey) (a), and (b) scatterplot of lake-effect synoptic type frequency (days) and lake- effect snowfall (cm) associated with Lakes Ontario and Erie from 1950- 2009.

Figure 2.6.: Linearly extrapolated change in snowfall (cm) during 1950-1979 seasons based on (a) changes in the frequency of lake-effect synoptic types, (b) changes in snowfall rates (snowfall intensity), (c) total changes in snowfall due to frequency and intensity changes (panel a plus panel b), and (d) observed snowfall data. Colors are consistent across all four panels and with Fig. 2.7. Blues represent increases in snowfall while yellows represent decreases in snowfall. Darker shades correspond to larger changes with lighter shades corresponding to smaller changes.

Figure 2.7.: As in Figure 2.6 except for 1980-2009 seasons.

50 Chapter 3

SPATIOTEMPORAL VARIABILITY OF GREAT LAKES BASIN SNOW COVER ABLATION EVENTS

Suriano, Z.J., and D.J. Leathers. 2017. Spatiotemporal variability of Great Lakes basin snow cover ablation events. Hydrological Processes, 31: 4229-4237, doi: 10.1002/hyp.11364.

3.1 Abstract

In the Great Lakes basin of North America, annual runoff is dominated by snowmelt. This snowmelt-induced runoff plays an important role within the hydrologic cycle of the basin, influencing soil moisture availability and driving the seasonal cycle of spring and summer Lake levels. Despite this, relatively little is understood about the patterns and trends of snow ablation event frequency and magnitude within the Great Lakes basin. This study uses a gridded dataset of Canadian and United States surface snow depth observations to develop a regional climatology of snow ablation events from 1960-2009. An ablation event is defined as an inter- diurnal snow depth decrease within an individual grid cell. A clear seasonal cycle in ablation event frequency exists within the basin and peak ablation event probability is latitudinally dependent. Most of the basin experiences peak ablation frequency in March, while the northern and southern regions of the basin experience respective peaks in April and February. An investigation into the inter-annual frequency of ablation events reveals ablation events significantly decrease within the northeastern and northwestern Lake Superior

51 drainage basins and significantly increase within the eastern Lake Huron and Georgian Bay drainage basins. In the eastern Lake Huron and Georgian Bay drainage basins, larger ablation events are occurring more frequently, and a larger impact on hydrology can be expected. Trends in ablation events are attributed primarily to changes in snowfall and snow depth across the region.

3.2 Introduction In snow-dominated regions, snow ablation is a critical hydrologic process, influencing soil moisture, stream flow, and groundwater. While important to regional hydrology, snow ablation also represents a societal and environmental hazard through snowmelt-induced flooding and excess nutrient and pollution transport (Changnon 2008; Ashley and Ashley 2008). In the Great Lakes basin of North America, snowmelt is the primary driver behind the seasonal cycle of Great Lakes water levels in the spring and summer, and dominates the annual runoff into the basin (Quinn 2002; Barnett et al. 2005; Figure 3.1). This suggests that changes to the frequency and timing of snowmelt events could substantially change the timing of runoff into the basin and seasonal lake-water levels. Such water-level changes impact a variety of environmental and ecological factors in the Great Lakes basin including fish habitats, sediment-water nutrients, aquatic vegetation, and marsh bird breeding abundance

(Barry et al. 2004; Timmermans et al. 2008; Steinman et al. 2012; Chow-Fraser 2013). While the physical processes behind snow ablation have been extensively studied in a variety of North American regions (e.g. Grundstein and Leathers 1998; Leathers et al. 2004; McCabe et al. 2007; Mazurkiewicz et al. 2008), there has been relatively limited research into the spatiotemporal variability of North American snow cover ablation. Dyer and Mote (2007) did examine the seasonal timing of collective

52 North American ablation events, detecting a shift towards an earlier onset of ablation during the 1960-2000 period. This supports decreasing trends of snow cover in the spring across the continent (Frei et al. 1999; Dyer and Mote 2006). In light of this conclusion and understanding the importance of snowmelt to the hydrology of the Great Lakes basin, it is critical to understand when and where ablation events are occurring specifically in the Great Lakes basin, and if there are significant changes to their seasonal timing. This research examines the spatial patterns and trends of snow ablation frequency and magnitude from 1960-2009 within the Great Lakes basin to determine the seasonal distribution of ablation events and ascertain if changes to the frequency or seasonal timing of ablation events are occurring. Additionally, as there is substantial spatial variability in seasonal snowfall across basin (Suriano and Leathers 2017a,b); this research investigates the spatial and temporal variability of ablation events and trends at a sub-basin scale seeking to identify particular regions within the basin with a higher susceptibility to changing climatic conditions.

3.3 Data and Methodology

3.3.1 Snow Depth Data

Snow depth data spanning 1960-2009 are obtained from a quality-controlled daily North American dataset, interpolated onto a 1-degree grid (Dyer and Mote 2006; Kluver et al. 2016), presently stored at Rutgers University (http://climate.rutgers.edu/snowcover/noaamelt/). This dataset is selected over other snow depth products such as the National Weather Services’ National Operational Hydrologic Remote Sensing Center (NOHRSC) Snow Data Assimilation System

53 (SNODAS) primarily due to its length of record and use in similar studies (Dyer and Mote 2007). In creation of the dataset (Dyer and Mote 2006) snow depth data are interpolated from stations within the United States’ cooperative observer network (U.S. Department of Commerce 2003), and the Meteorological Service of Canada (Braaten 1996). Furthermore, data underwent a quality control procedure as described in Robinson (1989). However, likely due to the well-documented nature of snow observations (Robinson 1989; Doesken and Robinson 2009) and the interpolation scheme, snow depth values within the dataset occasionally increase or decrease between successive days, then rebound without meteorological conditions being favorable for such depth changes. This is in part attributed to a lack of consistency in station reporting. A station may provide data one day, and then not provide it the next. While the interpolation method uses a variable search radius, in regions with a limited number of stations or in regions where moderate-to-large differences in snow depth values exist between stations within the same grid cell, a station not reporting for a day then reporting the next could have a large impact on an interpolated snow depth change. The differences in snow depth values between stations within the same grid cell could result from variations in elevation, a high prominence of micro/meso-scale meteorological events, or measurement errors/biases. In light of this, further quality control on the snow depth dataset was deemed necessary prior to performing an ablation calculation. A routine is developed to test the consistency of the daily change in snow depth where days and grid cells are flagged when: 1) an increase in depth occurs that is greater than 125% of new snow accumulations, and 2) snow depth decreases despite the day’s maximum temperature

54 remaining below -3°C and a large (10 cm) snowfall not occurring the day before (potential for compaction of fresh snow). The additional 25% of freedom in snow accumulations from the first criteria is the value found to maximize the efficiency of the flagging routine, such that when depth changes are meteorologically plausible, the day is not typically flagged. Based on this quality control procedure, data believed to be erroneous are flagged as missing within the dataset. No effort is made to generate replacement values using physically- or statistically-based algorithms. Within the Great Lakes basin, 2.7% of the potential gridded snow depth measurements are flagged as missing. With this further quality control, there is greater confidence that changes in snow depths are grounded in physical changes to the snowpack and not a result of stations’ observational inconsistencies.

3.3.2 Basin Definition To determine which of the 1-degree grid cells of the snow depth dataset constitute the Great Lakes basin, a centroid method is applied. If a grid cell’s centroid falls within the spatial boundary of the Great Lakes basin, that grid cell is considered within the basin at the 1-degree spatial resolution of the snow depth dataset. The spatial boundary of the basin is based on the Hydrological Units from the United States Geological Survey’s “Watershed Boundary Dataset”

(http://nhd.usgs.gov/wbd.html), and from the “Drainage Areas Dataset” by Natural Resources Canada (http://geogratis.gc.ca/). A single grid cell (42.5°N, 77.5°W) was added to this basin definition as a vast majority of the cell fell within the basin with the exception of a relatively narrow band that included the cell’s centroid. Fifty-seven 1-degree grid cells constitute the Great Lakes region, bounded approximately by 41- 51° North latitude and 75-93° West longitude (Figure 3.2).

55 3.3.3 Ablation Definition and Calculation With snow data from the cooperative network having no information on water equivalent, snow depth changes due to snowmelt have been defined as ablation (Grundstein and Leathers 1998; Leathers et al. 2004; Dyer and Mote 2007). As observed snow depth may also decrease due to sublimation, measurement errors, and compaction, these factors must be acknowledged before a snow depth decrease may be considered ablation. Sublimation can often be considerable over large scales in regions where air masses with low water vapor content are common. However, this effect is minimal over a single storm total (Dery and Yau 2002) and air masses over the Great Lakes region generally contain relatively high water vapor content. As such, the loss of daily snow depth due to sublimation is considered negligible, and not accounted for in this analysis. Effects of drifting snow are not accounted for in this study, as it is standard snow measurement practice to measure in areas where wind effects and drifting are minimized (U.S. Department of Commerce 2013). In addition to sublimation and drifting, changes in snow depth not associated with melt can be attributed to non-physical processes related to the measurement process. Variation in the time of observation by Cooperative Observers Program observers can influence snow depth values (Kunkel et al. 2007). Depending on the timing of a sub-diurnal scale snowmelt or accumulation event, stations in close geographic proximity, but with different time of observations, may record different daily snow depth values. This may result in erroneous daily snow depth changes in the interpolated value not grounded in the meteorological conditions; however this impact should be minimized by the additional quality control measures taken to flag and remove unrealistic snow depth changes.

56 Compaction can substantially decrease snow depth without modifying the mass of the snow through destructive metamorphism of fallen snow crystals. This process results in a denser, thinner snowpack over time (Colbeck 1983a,b). With increasing deposition of snow, more overburden force is placed on the snowpack, increasing the compression or densification rate (Mellor 1977). This may increase the number of detected ablation events attributed to snowmelt. However, by incorporating temperature criteria into the ablation calculation, this influence of compaction can be minimized (Dyer and Mote 2007). As such, an event of decreasing snow depth is only considered ablation if it occurs when the maximum daily temperature on the second day of the associated event is above 0°C. Under these conditions, the snowpack can be assumed to be relatively isothermal and mature, removing the effect of snowpack compaction as effectively as possible (Dyer and Mote 2007). In this study, an ablation event is considered an inter-diurnal decrease in snow depth greater than 2.5 cm within an individual grid cell, only during instances where the maximum daily temperature on the second day of the associated event is above 0°C. The 2.5 cm threshold is used to isolate events that could be considered hydrologically significant, and represents a measurable quantity for observers. Snow accumulations during defined ablation events are not uncommon and additionally must be accounted for in the ablation calculation. As such, the calculated depth decrease of a given ablation event is added to the recorded snowfall on the second day of the event. Ablation events are examined monthly and annually in each grid cell within the Great Lakes basin from 1960-2009 during the September-August snow season. To determine the seasonal cycle of ablation event frequency and magnitude of the entire

57 basin, events from all 57-cells are summed by month into whole basin values and monthly averages over the 50-year period are calculated. In establishing if there is a shift in the frequency and/or seasonal timing of ablation events, monthly and annual trends are calculated using simple linear regression. Trend analysis is conducted on all ablation events within the basin collectively, and for each of the 57 cells individually due to the variability of snow conditions within the basin. Autocorrelation tests are performed to examine a 1-year lag’s impact on the significance of monthly and annual trends, yielding a mean monthly value of 0.09. As no strong (> 0.3) or significant (p < 0.05) correlations existed for any month, no action is taken to address autocorrelation in the analysis.

3.4 Results

3.4.1 Basin-Scale Ablation Events

Over the Great Lakes basin, snow ablation event frequency exhibits a clear seasonal pattern during 1960-2009 (Figure 3.3). An ablation event is defined as an inter-diurnal snow depth decrease greater than 2.5 cm day-1 for an individual grid cell. In examining the seasonal cycle for the entire basin, the total ablation events across all grid cells are summed by month, and averaged over the period 1960-2009. The average frequency of ablation events begins to increase during mid autumn and reaches maximum frequency during March, with approximately 300 events year-1 (σ = 79.1) across the 57 grid cells within the basin. After March, the frequency declines quickly to less than one event year-1 in June, yielding over 900 annual events year-1 (σ = 148.6) (Figure 3.3a). The seasonal cycle in event frequency is also examined for ablation events at four different threshold levels. Ablation threshold levels are used to

58 investigate the magnitude of ablation events. Thresholds are defined for the Great Lakes basin as minor, moderate, major, and extreme events, respectively, corresponding to between 2.5 and 5 cm, 5 and 10 cm, 10 and 20 cm, and greater than 20 cm change in snow depth between successive days. The threshold values were chosen based on the size distribution of events within the basin. For all threshold levels, ablation event frequency increases during the autumn and winter months, reaches a maximum in March, and then rapidly declines to a summer-time minimum (Figure 3.3b). The monthly means of event frequency in Figure 3.3b are standardized by the annual mean of each threshold level, due to the higher frequency of smaller ablation events and the successively smaller number of events with each threshold level. The pattern in Figure 3.3b indicates that ablation events across all threshold levels are relatively consistent in their seasonal occurrences. The non-standardized average ablation event frequencies are depicted in Figure 3.3c. Trends in ablation frequency are calculated for all events greater than 2.5 cm and no significant monthly or annual trends are detected, suggesting the seasonal cycle is not significantly changing over time. Ablation events at the moderate, major, and extreme thresholds similarly exhibit non-significant trends; however, significant trends in minor ablation events (2.5 – 5 cm) are detected. The number of minor ablation events in April significantly decreased by 1.1 events year-1 from 1960-2009 (p < 0.05) across the basin. Additionally, the annual number of minor ablation events exhibits a significant decreasing trend (-1.9 events year-1, p < 0.05), indicating the decrease in annual minor ablation events could be the result of the decrease in April event frequency. This is supported by a strong statistical correlation (0.677, p < 0.01) between the number of annual and April ablation events. Minor ablation events in

59 April represent approximately 63% of the total April events. Despite no significant trends in minor events for any other months, the decrease in minor April ablation events suggests a trend towards an earlier end to the snow melt season across the entire Great Lakes basin.

3.4.2 Sub-Basin Scale Spatial Patterns and Trends in Ablation Events The Great Lakes basin covers a relatively large geographic area and snow is highly variable within the region. As such, patterns in ablation frequency are investigated at the sub-basin scale, revealing substantial spatial variability during 1960-2009. Generally, there are more ablation events at higher latitudes than at lower ones, but what is most distinctive is the enhanced frequency of events within close proximity to the leeward shores of the Lakes (Figure 3.4). These regions closely correspond to the lake-effect snow belts (Scott and Huff 1996), implying a relationship between ablation frequency and lake-effect snow. This is not unexpected as lake-effect snow greatly increases the depth of the snowpack in these regions (Scott and Huff 1996, Suriano and Leathers 2017a, Suriano and Leathers 2017b) and a larger and more persistent snowpack would increase the likelihood of ablation events. There is a distinct seasonal pattern of snow ablation event frequency within the Great Lakes basin (Figure 3.5). For most of the region, higher ablation event probability occurs during February-April, with the region of peak probability shifting northward in later spring months. Conversely during autumn, increased event probability moves southward over the basin. This is emphasized by spatially examining the month of peak ablation event frequency within the basin (Figure 3.6; top panel). A majority of the basin experiences peak ablation event frequency in March, however regions in the southerly and the most northerly portions of the basin

60 experience their peaks in February and April, respectively. While this is the case for ablation events of all sizes, the pattern is not consistent when examined for ablation events at different magnitude thresholds (Figure 3.6). As the threshold level for ablation events increases, the month of peak event frequency, particularly in the southern regions of the basin, tends to shift towards the earlier winter months. This is attributed to the decreasing likelihood of a progressively larger snowpack existing later into the winter season for the different threshold levels. The larger the ablation event threshold, the larger the snowpack must be to allow for such an event to occur; often by March or April, there is not enough snow on the ground to warrant large events in the southern regions of the basin. Thus the larger events can only occur when there is enough snow for them to occur, typically earlier in the winter season. While no long-term trend in annual ablation event frequency is detected for the Great Lakes basin as a whole, there are significant long-term trends in annual frequency at the sub-basin scale indicating certain regions may be more susceptible to changing climatic conditions (Figure 3.7). The most spatially coherent region exhibiting significant trends in annual ablation event frequency is in the north of the basin, consisting of portions of the northwestern and northeastern Lake Superior drainage basins (for region, see Figure 3.2). In this region, trends range from -0.11 events year-1 (p < 0.05) to -0.22 events year-1 (p < 0.01) indicating significant decreases in the annual number of ablation events over time. This is particularly apparent in the trends of minor ablation events (not shown). Significant decreases in annual event frequency for the minor threshold level range from approximately -0.05 (p < 0.05) to -0.17 events year-1 (p < 0.01), indicating relatively small ablation events are becoming less common.

61 As regions north of Lake Superior experience decreases in the frequency of annual ablation events, parts of the eastern Georgian Bay and eastern Lake Huron drainage basins (Figure 3.2) experience significant increases in ablation event frequency (Figure 3.7). These sub-basins have respective increasing trends of 0.21 events year-1 (p < 0.01) and 0.15 events year-1 (p < 0.01) that indicate significant increases in the frequency of ablation events during 1960-2009. In the eastern Lake Huron region, annual ablation event frequency is significantly increasing (p < 0.05) for the three largest ablation threshold levels: moderate, major, and extreme (not shown). This indicates that ablation events, particularly larger events, are becoming more common within these sub-basins.

3.5 Discussion and Conclusions A regional climatology of snow cover ablation event frequency is developed for the Great Lakes basin where an ablation event is defined as an inter-diurnal snow depth decrease exceeding a critical value within an individual grid cell. Ablation event frequency is examined collectively, and at different threshold levels (minor: 2.5-5 cm, moderate: 5-10 cm, major: 10-20 cm, extreme: 20+ cm) to allow for an investigation into the magnitude of ablation events. Analysis identifies a clear seasonal cycle in ablation event frequency, with March representing the peak month of event frequency for the basin at all threshold levels. Examining the spatial variability of the seasonal cycle within the basin, most of the basin experiences peak ablation frequency in March. However regions generally in the northern and southern portions of the basin experience respective peak ablation event frequencies in April and February. Dyer and Mote (2007) found peak ablation event frequency for North American occurs in April. This earlier peak in the Great Lakes basin is expected as the basin represents the

62 southerly edge of typical North American snow cover. The latitude-dependent seasonal pattern in ablation event probability follows the seasonal cycles of incoming solar radiation, temperature, and vapor pressure that provide the energy necessary for snow cover ablation. This increased energy impacts the Great Lakes basin prior to locations further north, aiding in explaining the differing peak detected in this study compared to Dyer and Mote (2007). Multiple studies have noted a change in the seasonal cycle of snow cover and ablation in diverse regions over North America, with snow cover exhibiting decreasing trends during the spring months, tending towards earlier snowmelt (Frei and Robinson 1999; Frei et al. 1999; Brown 2000; Dyer and Mote 2006; Dyer and Mote 2007). In this study, no significant shift in the seasonal cycle of all monthly snow ablation events is detected for the Great Lakes basin. However, the annual frequency of specifically minor ablation events (2.5 - 5 cm) within the basin significantly decreases in conjunction with declining events in April. This decline in minor ablation event frequency suggests a trend towards an earlier end to the snow melt season across the entire Great Lakes basin. This is likely associated with changes in basin snow cover during the spring driven by changing temperatures, and will be further examined in future work. At the sub-basin scale, significant trends in the annual frequency of ablation events are detected in two spatially coherent regions. In the northwestern and northeastern Lake Superior drainage basins, annual ablation event frequency significantly decreases by as much as than -0.22 events year-1 (p < 0.01). Linearly extrapolated, this reduces the number of events per year from approximately 23.4 in 1960, to less than 12.3 events per year in 2009; a reduction of nearly 50% (Figure

63 3.8a). Minor ablation events additionally exhibit significant decreasing trends in these sub-basins. These trends indicate ablation events, particularly smaller ablation events, are becoming less frequent. The decreasing trend in this region is attributed to significantly decreasing trends in snow depth in this region stretching northwest into central Canada (Dyer and Mote 2006). Dyer and Mote (2006) attribute this change in snow depth, in part, to extra-tropical cyclones that track over the region changing in frequency and/or intensity (Isard et al. 2000), resulting in changing temperature and precipitation patterns in the region. With a shallower snowpack, fewer ablation events can occur. The second region exhibiting significant trends in annual ablation event frequency is in the eastern Georgian Bay and eastern Lake Huron drainage basins. Regions in these sub-basins exhibit significant increases in ablation events by as much as 0.21 events year-1 (p < 0.01). Over the 50-year period, this represents a linearly extrapolated increase in the annual number of events by 74% from approximately 14.5 in 1960, to over 25 annual events in 2009 (Figure 3.8b). The frequencies of the largest ablation threshold levels (moderate, major, and extreme) are also significantly increasing. This is an indication that larger ablation events are becoming more common in these sub-basins, and a larger impact to the hydrology can be expected. Larger daily ablation events may increase snowmelt-related runoff and increase flooding event risk. These increasing trends in ablation frequency are attributed to significant increases in lake-effect snowfall in the region (Suriano and Leathers 2017b), where more snowfall creates a larger snowpack and increases the associated potential for more or larger ablation events to occur.

64 The results of this study have highlighted the nature of snow cover ablation events for the Great Lakes basin, and emphasized the spatial variability at the sub- basin scale. Changes to the frequency of ablation events will influence the occurrence and magnitude of snowmelt-induced runoff into the Lakes that may impact water resources, ecological habitats, and/or terrestrial flooding events. Understanding the regional complexities to snow ablation is critical for communities preparing for the impacts of a changing climate. Further research is currently ongoing, investigating the atmospheric conditions associated with ablation events in the Great Lakes basin. This future work will explore the relationships between synoptic-scale weather patterns and ablation frequencies, seeking to explain some of the changes in ablation frequency detected in this study.

3.6 Acknowledgements The authors would like to acknowledge the financial support received from the National Oceanic and Atmospheric Administration Climate Program Office (NA14OAR4310207). Gratitude is extended to Thomas Estilow at Rutgers University for his assistance in snow data acquisition, and to Sara Rauscher and Tracy DeLiberty from the University of Delaware, David Robinson from Rutgers University, and Gina Henderson from the United States Naval Academy for their helpful suggestions and comments. Gratitude is also extended to the comments provided by two anonymous reviewers.

REFERENCES

Ashley, S.T., & Ashley, W.S. (2008). The storm morphology of deadly flooding events in the United States. International Journal of Climatology, 28, 493-503. doi:10.1002/joc.1554.

Barnett T.P., J.C. Adam, & Lettenmaier, D.P. (2005). Potential impacts of a warming climate on water availability in snow-dominated regions. Nature, 438, 303- 309, doi:10.1038/nature04141.

Barry M.J., R. Bowers, & De Szalay, F.A. (2004). Effects of hydrology, herbivory and sediment disturbance on plant recruitment in a Lake Erie coastal wetland. American Midland Naturalist, 151, 217-232, doi:10.1674/0003- 0031(2004)151[0217:EOHHAS]2.0.CO;2.

Braaten, R.O. (1996). An updated Canadian snow depth database. Contract report prepared for Atmospheric Environmental Service, September, 1996, 24.

Brown, R.D. (2000). Northern Hemisphere snow cover variability and change, 1915- 97. Journal of Climate, 13, 2339-2355.

Changnon, S. (2008). Assessment of flood losses in the United States. Journal of Contemporary Water Resources Education, 138, 38-44, doi:10.1111/j.1936- 704X.2008.00007.x.

Colbeck, S.C. (1983a). Theory of Metamorphism of Dry Snow. J. Geophy. Res. 88, 5475-5482.

Colbeck, S.C. (1983b). Snow Particle Morphology in the Seasonal Snow Cover. Bulletin of the American Meteorological Society, 64, 602-609.

Dery, S., & Yau, M. (2002). Large-scale mass balance effects of blowing snow and surface sublimation. Journal of Geophysical Research, 107, 4679, doi:10.1029/2001JD001251.

Doesken, N.J. & Robinson, D.A. (2009). The Challenges of Snow Measurement. Historical Climate Variability and Impacts in North America, doi:10.1007/978-90-481-2828-0_15.

Dyer, J.L., & Mote, T.L. (2006). Spatial variability and trends in observed snow depth over North America. Geophysical Research Letters, 33, L16503, doi:10.1029/2006GL027258.

66 Dyer, J.L., & Mote, T.L. (2007). Trends in snow ablation over North America. International Journal of Climatology, 27, 739-748, doi:10.1002/joc.1426.

Fracz, A., & Chow-Fraser, P. (2013). Impacts of declining water levels on the quantity of fish habitat in coastal wetlands of eastern Georgian Bay, Lake Huron. Hydrobiologia, 702, 151-169, doi:10.1007/s10750-012-1318-3.

Frei, A., & Robinson, D.A. (1999). Northern Hemisphere snow extent: regional variability 1972-1994. International Journal of Climatology, 19, 1535-1560.

Frei, A., D.A. Robinson, & Hughes, M.G. (1999). North American snow extent: 1900- 1994. International Journal of Climatology, 7, 1517-1534.

Grundstein, A.J, & Leathers, D.J. (1998). A case study of the synoptic patterns influencing midwinter snowmelt across the Northern Great Plains. Hydrological Processes, 12, 2293-2305.

Kluver, D., T.L. Mote, D.J. Leathers, G.R. Henderson, W. Chan, & Robinson, D.A. (2016). Creation and Validation of a Comprehensive 1° by 1° Daily Gridded North American Dataset for 1900–2009: Snowfall. Journal of Atmospheric and Oceanic Technology, 33, 857–871, doi:10.1175/JTECH-D-15-0027.1.

Kunkel, K.E., M.A. Palecki, K.G. Hubbard, D.A. Robinson, K.T. Redmond, & Easterling, D.R. (2007). Trend Identification in Twentieth-Century U.S. Snowfall: The Challenges. Journal of Atmospheric and Oceanic Technology, 24, 64-73, doi:10.1175/JTECH2017.1

Leathers, D.J., D. Graybeal, T.L. Mote, A.J. Grundstein, & Robinson, D.A. (2004). The role of air mass types and surface energy fluxes in snow cover ablation in the central Appalachians. Journal of Applied Meteorology, 43, 1887–1898.

Mazurkiewicz, A.B., D.G. Callery, & McDonnell, J.J. (2008). Assessing the controls of the snow energy balance and water available for runoff in a rain-on-snow environment. Journal of Hydrology, 354, 1-14, doi:10.1016/j.jhydrol.2007.12.027.

McCabe, G., M. Clark, & Hay, L. (2007). Rain-on-snow events in the western United States. Bulletin of American Meteorological Society, 88, 319-328, doi:10.1175/BAMS-88-3-31920.

Mellor, M. (1977). Engineering properties of snow. Journal of Glaciology, 19, 15-66.

Quinn, F.H. (2002). Secular changes in Great Lakes water level seasonal cycles. Journal of Great Lakes Research, 28, 451-465.

67 Robinson, D.A. (1989). Construction of a United States historical snow database. Proceedings of the 45th Eastern Snow Conference, 50-59.

Scott, R., & Huff, F. (1996). Impacts of the Great Lakes on regional climate conditions. Journal of Great Lakes Research, 22, 845-863, doi:10.1016/S0380- 1330(96)71006-7.

Steinman, A.D., M.E. Ogdahl, M. Weinert, K. Thompson, M.J. Cooper, & Uzarski, D.G. (2012). Water level fluctuation and sediment-water nutrient exchange in Great Lakes coastal wetlands. Journal of Great Lakes Research, 38, 766-775, doi:10.1016/j.jglr.2012.09.020.

Suriano, Z.J., & Leathers, D.J. (2017a): Synoptic Climatology of Lake-Effect Snowfall Conditions in the Eastern Great Lakes Region. International Journal of Climatology, 37: 4377-4389, doi:10.1002/joc.5093.

Suriano, Z.J., & Leathers, D.J. (2017b): Synoptically Classified Lake-Effect Snowfall Trends to the Lee of Lakes Erie and Ontario. Climate Research, 74: 1-13. doi: 10.3354/cr01480.

Timmermans, S.T.A., S.S. Badzinski, & Ingram, J.W. (2008). Associations between breeding marsh bird abundances and Great Lake hydrology. Journal of Great Lakes Research, 34, 351-364, doi:10.3394/0380- 1330(2008)34[351:ABBMBA]2.0.CO;2

U.S. Department of Commerce (2003). TD-3200 Surface Summary of the Day, National Climatic Data Center, NOAA. [Available online: http://www.ncdc.noaa.gov].

U.S. Department of Commerce (2013). Snow Measurement Guidelines for National Weather Service Surface Observing Programs. NOAA Office of Climate, Water and Weather Services. September 2013.

Figure 3.3.: Average monthly ablation event frequency across the Great Lakes basin from 1960-2009 (a), (b) standardized average monthly ablation event frequency for the four defined threshold levels from 1960-2009, and (c) as in (b) but for non-standardized frequencies

Figure 3.4.: Total number of ablation events greater than 2.54 cm across the Great Lakes basin from 1960-2009. Darker shades represent progressively more events (days).

Figure 3.6.: Month of peak ablation event frequency for all ablation events, and for events at the defined threshold levels across the Great Lakes basin from 1960-2009. Colors indicate different months.

Figure 3.7.: Trend (a) and statistical significance (b) in the annual frequency of ablation events across the Great Lakes basin from 1960-2009. For panel (a), in days year-1, brown shades represent negative trends while blue represent positive trends. In panel (b), reported in 1 – p-value, darker red shades correspond to lower p-values and thus higher statistical significance.

76 Chapter 4

GREAT LAKES BASIN SNOW COVER ABLATION AND SYNOPTIC-SCALE CIRCULATION

Suriano, Z.J., and D.J. Leathers. Great Lakes basin snow cover ablation and synoptic- scale circulation. Journal of Applied Meteorology and Climatology. (Submitted October 2017).

4.1 Abstract

Synoptic-scale atmospheric conditions play a critical role in determining the frequency and intensity of snow cover ablation events. Using a synoptic weather classification technique, distinct regional circulation patterns influencing the Great Lakes basin of North America are identified and examined in conjunction with daily snow ablation events from 1960-2009. An ablation event is considered in this study as an inter-diurnal decrease in areal-weighted average snow depth greater than 2.54 cm in magnitude over the entire Great Lakes basin. General meteorological characteristics associated with ablation-causing synoptic types are examined and three individual case studies from prominent synoptic types are presented to understand the diversity of meteorological influences on regional snow ablation. Results indicate a variety of synoptic weather conditions lead to snow ablation in the Great Lakes basin. The ten most common synoptic types result in 65% of the 392 ablation events detected from 1960-2009. Collectively, snow ablation in the Great Lakes basin most commonly occurs when there is advection of warm and moist air into the region providing the sensible and latent heat fluxes needed for melt, but

77 ablation frequently occurs during rain-on-snow events and in instances of high- pressure overhead. Ablation magnitude is highest during rain-on-snow synoptic types and the inter-annual frequency of these types is significantly decreasing by over 37%. Conversely, the frequency of high-pressure overhead synoptic types is significantly increasing by over 30% from 1960-2009. Such changes may influence the hydrologic impact of these synoptic types on ablation over time. Keywords: synoptic classification, snowmelt, warm air advection, rain-on-snow

4.2 Introduction Snow cover ablation plays a key role in the hydrologic cycle in many regions across the globe, contributing to soil moisture availability, stream flow, and groundwater supplies (Barnett et al. 2005). The lack of consistency in the volume and/or timing of snowmelt events associated with ephemeral snow cover, can however lead to harmful societal and environmental consequences. The variable release of water from the snowpack may result in snowmelt induced flooding, pollutant or excess nutrient transport, and water stress along snowmelt-fed rivers. During 1972-2006, 48 major snowmelt floods occurred in the United States, resulting in an average of $69 million in losses per event (Changnon 2008). In the Great Lakes basin of North America, annual runoff is dominated by snowmelt (Barnett et al. 2005), and snowmelt is considered the primary driver behind the seasonal cycle of the Great Lakes’ water levels in spring and summer (Quinn 2002). A number of negative ecological and environmental consequences are linked to variable lake levels including those impacting wildlife habitats (Fracz and Chow- Fraser 2013). This suggests changes to snow ablation events within the basin could

78 substantially alter the timing of runoff, influencing soil moisture availability, lake levels, and ecological environments. Snow ablation event variability is in part a function of the synoptic-scale atmospheric conditions that influence the heat and moisture transfers over a region (Grundstein and Leathers 1999). As such, synoptic classification techniques can be employed to assist in the evaluation of the atmosphere’s influence on the snowpack. These techniques represent days with similar atmospheric conditions as a single synoptic weather type (Yarnal 1993), where each type influences the region and snowpack in a unique manner. Synoptic classification techniques have been utilized in a number of studies examining ablation in mid-latitude regions globally (Grundstein and Leathers 1999; Leathers et al. 2004; Bednorz 2009; Bednorz and Wibig 2015). In many cases, snow ablation is favored during synoptic types that advect warm and moist air into a region and/or result in rain-on-snow precipitation. To determine the synoptic-scale weather conditions that result in snow ablation events in the Great Lakes basin, this study utilizes an eigenvector-based synoptic weather typing classification technique similar to that of Kalkstein and Corrigan (1986) to create a daily synoptic calendar from 1960-2009. Snow ablation events exceeding 2.54 cm day-1 averaged over the entire basin are identified, and the associated synoptic-scale weather patterns are analyzed. By using a synoptic classification approach, insights are gained into how each synoptic type modifies the snowpack individually and the frequency of each type can be tracked over time. Detailed case studies from three of the most prominent synoptic types are presented to highlight the diversity of meteorological conditions that lead to snow ablation.

79 4.3 Methodology and Datasets

4.3.1 Temporal Synoptic Index

Daily synoptic weather types are developed from 1960-2009 based on station data from Flint, Michigan (WBAN# 14826, 42.97 °N, 83.75 °W) using a procedure similar to the Temporal Synoptic Index (TSI; Kalkstein and Corrigan 1986). The procedure classifies synoptic-scale weather types by relying on a principal components analysis (PCA) and subsequent cluster analysis of meteorological observations (e.g. Suriano and Leathers 2017a, 2017b). The Flint, MI station is chosen based on its relatively centralized location within the Great Lakes basin and its sufficiently long and complete period of record. The TSI is conducted for the four climatological seasons individually (SON, DJF, MAM, and JJA) as opposed to a single annual analysis which can limit the effectiveness of the procedure (Suriano and Leathers 2017a, 2017b). The final result of the procedure is a daily synoptic calendar from 1960-2009 where each day is classified as a particular synoptic weather type that can be analyzed in conjunction with daily snow cover ablation events.

4.3.2 Snow Dataset and Great Lakes Basin Definition Snow data come from a quality controlled, daily North American dataset of snowfall, snow depth, temperature and precipitation interpolated onto a 1-degree grid, archived at Rutgers University (http://climate.rutgers.edu/snowcover/noaamelt/; Dyer and Mote 2006; Kluver et al. 2016). Data are interpolated from observation stations within the United States’ Cooperative Observer network and the Meteorological Service of Canada. The extensive quality control conducted on the data is outlined in Robinson (1989) and Suriano and Leathers (2017c).

80 For this study, the Great Lakes basin is defined at a 1-degree resolution, and is based on the grid cell centroids of the snow dataset being contained within the basin’s geographical boundary. The boundary is defined by the Hydrological Units from the United States Geological Survey’s “Watershed Boundary Dataset” (http://nhd.usgs.gov/wbd.html), and the “Drainage Areas Dataset” by Natural Resources Canada (http://geogratis.gc.ca/). Based on an individual case where the centroid method was deemed inappropriate, a single grid cell (42.5°N, 77.5°W) was added to this definition, resulting in 57 1-degree grid cells constituting the Great Lakes basin.

4.3.3 Ablation Definition As snow data from the cooperative network contains no information on water equivalent, changes in daily snow depth have been defined as ablation in prior work (Grundstein and Leathers 1998; Leathers et al. 2004; Dyer and Mote 2007; Suriano and Leathers 2017c). In this study, an ablation event is considered an inter-diurnal decrease in basin-wide average snow depth exceeding 2.54 cm. Further criteria defining an ablation event include using only those instances where the maximum daily temperature on the second day of the associated ablation event is above 0 °C (Dyer and Mote 2007; Suriano and Leathers 2017c). Daily snow depth for the Great

Lakes basin is determined by calculating a daily areal weighted average depth based on the 57 grid cells contained within the basin. While the Great Lakes basin covers a relatively large geographic area and snow depths can be spatially variable, average seasonal snow cover very often covers the entire basin, permitting analysis for the basin on the whole. Using a 2.54 cm threshold for a basin-wide ablation event does not mean this much snow is ablated

81 across all portions of the basin; an ablation event for a given day may be spatially inhomogeneous depending on the current snow depth and meteorological conditions. Furthermore, an ablation event of at least 2.54 cm can only occur if the average basin- wide snow depth is at least that depth. It is possible for an event greater than the threshold to occur only in a portion of the overall basin, but not be included in this analysis due to the basin-wide ablation value not meeting the 2.54 cm level. This is specifically the case for events that cause melt in the southern and eastern portion of the basin while snow accumulates in the western and northern portions, potentially resulting in a dampened ablation signal across the entire basin. Despite this, valuable information regarding the atmospheric controls on snow ablation dynamics for the basin on the whole can be gathered. It is important to note that beyond snowmelt, snow depth decreases may also occur due to physical factors such as sublimation and compaction of the snowpack, and/or due to non-physical factors related to the measurement process. Sublimation can considerably reduce the mass of a snowpack over large spatial scales in regions with a high occurrence of dry air masses. This effect however is minimal for a single storm (Déry and Yau 2002) and air masses over the Great Lakes region generally contain relatively high water vapor content. Similarly, compaction of the snowpack due to destructive metamorphism of snow crystals can substantially decrease the depth of the snowpack without modifying its mass (Colbeck 1983). By utilizing the temperature requirement in defining ablation events, the effect of snowpack compaction is removed as effectively as possible. Under conditions where the temperature exceeds 0 °C, the portion of depth change attributable to compaction is minimized as the snowpack can be assumed to be relatively mature and isothermal

82 (Dyer and Mote 2007). The impact of depth changes due to errors or biases in the measurement processes should be minimized by the quality control measures taken to flag and remove snow depth changes inconsistent with the meteorological conditions (Suriano and Leathers, 2017c).

4.4 Results An ablation event for the Great Lakes basin is defined as an inter-diurnal decrease of areal-weighted basin-wide average snow depth in excess of 2.54 cm. In the Great Lakes basin, 7.8 events occur per year (σ =3.9) on average, with a range of 2 to 21 events year-1. When an event occurs, an average of 3.7 cm of snow ablates across the entire Great Lakes basin (σ =1.2). The seasonal cycle of ablation indicates events predominantly occur between the months of December and April, however on average, only two months have more than one ablation event per year: February (1.5 events year-1) and March (3.5 events year-1). Ablation magnitude per event is relatively consistent during the melt season. No statistically significant trends in ablation frequency or magnitude are detected from 1960-2009.

4.4.1 Synoptic Analysis

With no ablation events within the basin occurring during the summer months

(JJA), those months are excluded from analysis. The PCA of the TSI procedure yields five principal components with eigenvalues greater than 1.0 for each of the remaining climatological seasons. In autumn (SON), winter (DJF), and spring (MAM) the five components respectively explain 80.5%, 79.3%, and 77.7% of their seasonal variance. Clustering the seasonal PCA loadings using within group average linkage clustering with an initial 20-cluster solution (Kalkstein et al. 1987; Suriano and Leathers 2017a,

83 2017b), yields the synoptic weather types across the September through May snow season and a daily synoptic calendar from 1960-2009 is generated. The daily synoptic calendar is then examined in conjunction with the identified daily ablation events to determine the synoptic weather types responsible for ablating snow within the Great Lakes basin. While multiple synoptic weather types result in at least one basin-wide snow ablation event in excess of 2.54 cm from 1960-2009, only synoptic weather types that lead to at least one ablation event every four years are analyzed. Ten synoptic weather types meet this criterion, producing 65% of the 392 total ablation events, and are examined in greater detail (Figure 4.1, Table 4.1). Table 4.1 presents the composited meteorological conditions in Flint, MI for each of these ten synoptic weather types at four daily observation periods, along with general statistics. Composite sea-level pressure fields for each of the ten synoptic weather types indicate the dominant pattern for ablation in the Great Lakes basin (61%) is that of low pressure to the west and high pressure to the east (Figure 4.1 – S1-5, Table 1). This general pressure pattern is depicted in five of the synoptic weather types (S1-5; S for southerly flow component), resulting in south to south-southwest flow into the basin and advection of warm and potentially moist air north from the Gulf of Mexico. This advection would provide sufficient sensible and latent heat fluxes into the snowpack necessary for melt. Despite having similar synoptic-scale atmospheric patterns, the meteorological conditions for the individual synoptic weather types with this southerly flow pattern do vary, in part due to their timing within the seasonal cycle (Table 4.1). Two other dominant sea-level pressure patterns are present among the most common ablation-causing synoptic weather types: rain-on-snow and overhead high

84 patterns. Weather types R1-3 (R for rain-on-snow) are classified as rain-on-snow synoptic types due to a cyclonic system tracking over portions of the basin resulting in advection of warm and moist air and liquid precipitation. For instance, type R1 produces approximate precipitation rates between 3-6 mm day-1 per associated ablation event averaged across the basin. In addition to the warm air and dew point temperatures that result in relatively high sensible and latent heat fluxes into the snowpack, the rain-on-snow precipitation should contribute to snowmelt (Leathers et al. 1998; Levia and Leathers 2011). For weather types H1-2 (H for high pressure overhead), there is a high- pressure center directly over the Great Lakes basin, resulting in afternoon air temperatures above 5 °C, average dew point temperatures below -3°C, and light and variable winds in Flint, MI (Table 4.1 – H1-2). For weather type H1, there is almost no cloud cover in Flint and based on these average conditions, the bulk of energy needed for ablation likely results from relatively warm air and high levels of incoming solar radiation reaching the snowpack (not shown). While not examined in detail, the synoptic types that cause the remaining 35% of ablation can still generally be grouped into one of these three main categories.

4.4.2 Ablation Case Studies

The magnitude of ablation during a particular synoptic weather type is determined by the availability of snow to ablate and the synoptic weather type’s specific atmospheric conditions. To examine the diversity of meteorological conditions associated with differing types, three case studies are examined for the most prominent patterns leading to ablation in the Great Lakes basin: southerly flow, rain-on-snow, and overhead high patterns. Surface pressure, temperature, dew point

85 temperature, wind speed, and cloud cover maps are generated from the National Center for Environmental Prediction (NCEP) Reanalysis data to understand the particular conditions during select ablation events over the Great Lakes basin (Kalnay et al. 1996). (a) Type S1 – Southerly flow pattern – February 27, 1974 The weather pattern on February 27, 1974 provides an example of synoptic weather type S1, a southerly flow pattern (Figure 4.2a). A low-pressure center over North Dakota and high pressure along the Atlantic coast results in relatively warm and moist air being drawn north from the Gulf of Mexico. Examining meteorological variables over the region at 0000 UTC (1800 CST), surface air and dew point temperatures greater than 0°C advance into a majority of the basin from the southwest (Figure 4.2b, 4.2c). Wind speeds range from 3 m s-1 north of Lake Superior to over 9 m s-1 east of Lake Erie (Figure 4.2d). Cloud cover exceeds 50% across the region with maximum cloud cover of over 90% existing in northern Michigan (Figure 4.2e). Snow cover is present across almost the entire basin on February 27th, 1974 and is at depths in excess of 50 cm in the northern portions (Figure 4.3a). During the event, average basin-wide ablation is 5.12 cm with much of the snowmelt occurring in the central portions of the basin and east of Lake Erie where higher wind speeds are observed (Figure 4.3b). In southern Michigan, the snowpack is completely melted during the event. Due to air and dew point temperatures exceeding 0°C with moderate wind speeds over much of the basin, a majority of melt is due to relatively large turbulent fluxes providing energy to the snowpack (not shown). (b) Type R1 – Rain-on-Snow pattern – February 13, 1984

86 A representative example of synoptic weather type R1 occurred on February 13, 1984 where the rain-on-snow event ablated an average of 4.58 cm of snow across the basin. The 1000 hPa low-pressure system northwest of the basin and the strong high-pressure center over the Northern Atlantic Ocean contributes to south- southeasterly flow (Figure 4.4a). This flow results in the advection of air and dew point temperatures exceeding 10-12°C into the eastern portion of the basin stretching as far north as 46°N at 0000 UTC (Figures 4.4b, 4.4c). Wind speeds at this time are a moderate 6-8 m s-1 over much of the basin (Figure 4.4d). Peak cloud cover of over 80% corresponds closely to the tongue of warm and moist air, however a majority of the basin exhibits less than 50% cloud cover (Figure 4.4e). Additionally, there is substantial liquid precipitation during the event, with regions in southern Ontario receiving in excess of 3.0 cm over the calendar day (Figure 4.4f). Examining the snow depth across the basin prior to the event, there is a strong latitudinal gradient with the northern regions having a much deeper snowpack (Figure 4.5a). Regions that exhibited the greatest loss of snow during the event correspond closely to those that experienced the warmest air and dew point temperatures, and the highest liquid precipitation totals (Figure 4.5b). The advection of warm and moist air results in relatively high latent and sensible heat transfer into the snowpack, greatly contributing to melt. The liquid precipitation additionally contributes to snowmelt by giving up sensible heat upon contact with the relatively colder snowpack, and/or by eventually freezing onto the snowpack and releasing latent heat. (c) Type H1– Overhead high pattern – March 18, 2001 March 18, 2001 is an example of synoptic weather type H1, where a 1028 hPa High is directly over the Midwest and Great Lakes regions (Figure 4.6a). Surface air

87 temperatures are above freezing at 0000 UTC over almost the entire basin with a maximum around 4°C in southern Michigan (Figure 4.6b), and dew point temperatures below freezing (Figure 4.6c). Wind speeds are generally weak over much of the region with the exception of the eastern most reaches of the basin in New York State where speeds in excess of 6 m s-1 are recorded (Figure 4.6d). With high-pressure overhead, cloud cover in the basin is minimal (< 20%) allowing for high levels of solar radiation to reach the surface (Figure 4.6e). Similar to the other case studies, the deepest snow depths on March 18, 2001 are in the northern and eastern portions of the Great Lakes basin with much of southern Michigan having less than 3 cm of snow (Figure 4.7a). During the basin-wide average 4.55 cm ablation event, snow depth is ablated over much of the basin (Figure 4.7b). Relative maxima of ablation exist east of each of the five lakes including a number of grid cells south of Lake Superior. A majority of the energy needed for snowmelt during the event came from two sources: the above freezing air temperatures, and the clear skies and relatively high mid-March sun angle allowing for a higher solar radiative flux into the snowpack.

4.4.3 Inter-annual Variability of Synoptic Types While snow cover is necessary for an ablation event to occur, examining the frequency of the previously identified synoptic types indicates how atmospheric circulation responsible for ablation varies and may be changing over time. Examining the inter-annual frequency of synoptic types that lead to ablation by the general categories presented previously (southerly flow, rain-on-snow, and high pressure overhead), substantial variability is present from 1960-2009 (Figure 4.8). The southerly flow synoptic types are most prevalent, occurring approximately 26.3 days

88 year-1, followed by high-pressure overhead and rain-on-snow types occurring 12.0 and 11.0 days year-1, respectively. Trends in the inter-annual frequency of each synoptic type category are examined using simple linear regression. The frequency of rain-on- snow synoptic types is significantly declining by 0.10 days year-1, or 5.1 days over the 50-year period (p < 0.01). Such a decline represents an approximate 37% decrease. Conversely, the frequency of high-pressure overhead types is significantly increasing by 0.07 days year-1, or 3.5 days over 1960-2009 (p < 0.05). Southerly flow synoptic types do not exhibit statistically significant linear trends (p > 0.05).

4.5 Discussion and Conclusions In the Great Lakes basin of North America, three general synoptic weather types can explain the majority of basin-wide snowmelt events greater than 2.54 cm during the 1960-2009 period: southerly flow, rain-on-snow, and high-pressure overhead patterns. The most common type is that of a mid-latitude cyclone to the west and an anticyclone to the east and south of the basin resulting in advection of unseasonably warm and potentially moist air from the Gulf of Mexico. Similarly, a second synoptic type leading to ablation also has a mid-latitude cyclone in the vicinity of the basin leading to advection of warm and moist air, however for these types liquid precipitation aids in snowmelt. The third type is that of high-pressure overhead resulting in warm temperatures and low cloud cover basin-wide, thus increasing the potential for enhanced levels of incoming solar radiation reaching the snowpack and inducing melt. Three case studies, one representing each of the general synoptic weather types, are used to demonstrate the meteorological variability in snowmelt events between the types. In synoptic type S1, the southerly flow type, air and dew point

89 temperatures greater than 0 °C across much of the basin and relatively high wind speeds approaching 10 ms-1 provide the turbulent heat exchange necessary for melt. Case study two depicts the liquid precipitation and advection of unseasonably warm and moist air associated with type R1 into the central and eastern portions of the basin, leading to melt. In the third case study, for type H1, the overhead high pressure leads to minimal cloud cover (< 20%) over much of the basin allowing for high levels of solar radiation to reach the surface. Additionally, surface air temperatures do reach 4°C at 0000 UTC in portions of the basin and this contributes to the warming and melting of the snow pack. To ascertain if long-term changes to the frequency of these patterns are occurring, simple linear regression is performed. The inter-annual frequency of high- pressure overhead synoptic types are significantly increasing by over 34% from 1960- 2009. These types result in approximately 3.8 cm of snow ablation per event and contribute substantially to the total amount of snow ablated across the basin. The frequency of rain-on-snow synoptic types are significantly declining over the 50-year period, with a linearly extrapolated decrease from 13.6 to 8.5 days year-1 from 1960 to 2009. On average, rain-on-snow synoptic types ablate the highest amount of basin- wide average snow cover of the general synoptic categories, at approximately 4.0 cm event-1. Rain-on-snow conditions are often associated with the largest snowmelt- induced flooding events in mid-latitude regions due to the addition of liquid precipitation to the snow-generated runoff (e.g. Leathers et al. 1998). Detecting the significant decrease in the frequency of these synoptic conditions is important as fewer instances of rain-on-snow synoptic conditions suggest a decrease to the hydrologic

90 impact of these types and their probability of producing potentially hazardous flooding conditions. The synoptic weather conditions leading to ablation in the Great Lakes basin are broadly similar to conditions shown to cause snow ablation in other mid-latitude regions. In the North American Great Plains, the three synoptic patterns associated with snowmelt involve a mid-latitude cyclone tracking through the region and advecting warm air from the south (Grundstein and Leathers 1998; Grundstein and Leathers 1999). Ablation in the West Siberian Plain is associated with four different circulation types that include westerly, southwesterly, and southerly flow types (Bednorz and Wibig 2015). Melt events in the Polish-German lowlands are favored when anomalously strong low-pressure systems are in the northern Atlantic and high- pressure is over the Mediterranean (Bednorz 2009). Such a pattern brings warm maritime air and often higher-than-usual precipitation into the region, inducing snowmelt. In the central Appalachians, a majority of ablation events occur during ‘moist’ air masses with cloudy, windy conditions and high dew point temperatures, however ‘dry’ air masses with clear skies also cause events (Leathers et al. 2004). This manuscript focused on the atmospheric forcings of snow cover ablation in the Great Lakes basin, highlighting the diversity of atmospheric and meteorological conditions that lead to snow ablation events. The variability of snow cover ablation however is dependent on both the overlying atmospheric conditions that induce ablation and the availability of sufficient snow depth to be ablated. While snow cover exists across the entire basin seasonally, there is substantial temporal and spatial variability in the depth of snow. Preliminary findings suggest the average seasonal basin-wide snow depth is significantly declining by approximately 25% over 1960-

91 2009, while particular sub-regions have differing trends potentially due to changing snowfall totals. Such changes in average snow depth could affect the frequency or magnitude of ablation events, based on the availability of snow to ablate. A thorough investigation into the role of spatial and temporal snow depth variability on snow ablation is currently underway. Additionally, while this study briefly inspected and makes reference to energy fluxes associated with melt events, a more exhaustive examination into these snow-surface fluxes for each of the primary ablation-causing synoptic weather types will be conducted in future studies. Continued research into interactions between snow ablation and atmospheric circulation is useful for operational hydrological forecasting and regional water management in general.

4.6 Acknowledgements Partial funding for this research was provided by the National Oceanic and Atmospheric Association Climate Program Office grant NA14OAR4310207.

REFERENCES

Barnett, T.P., J.C. Adam, and D.P. Lettenmaier, 2005: Potential impacts of a warming climate on water availability in snow-dominated regions. Nature, 438,303-309, doi:10.1038/nature04141.

Bednorz, E., 2009: Synoptic conditions for rapid snowmelt in the Polish-German lowlands. Theor. Appl. Climatol., 97, 279–286, doi:10.1007/s00704-008-0063- z.

Bednorz, E. and J. Wibig, 2015: Spatial distribution and synoptic conditions of snow accumulation and snow ablation in the West Siberian Plain. Quastiones Geographicae. doi:10.1515/quageo-2015-0029.

Changnon, S.A., 2008: Assessment of Flood Losses in the United States. J. Contemp. Water Res. Educ., 138, 38–44, doi:10.1111/j.1936-704X.2008.00007.x.

Colbeck, S., 1983: Theory of Metamorphism of Dry Snow. J. Geophys. Res.-Oceans Atmospheres, 88, 5475–5482, doi:10.1029/JC088iC09p05475.

Déry, S.J., and M.K. Yau, 2002: Large-scale mass balance effects of blowing snow and surface sublimation. J Geophys Res Atmospheres, 107, ACL 8-17, doi:10.1029/2001JD001251

Dyer, J.L., and T.L. Mote, 2006: Spatial variability and trends in observed snow depth over North America. Geophys. Res. Lett., 33, doi:10.1029/2006GL027258.

——, and ——, 2007: Trends in snow ablation over North America. Int. J. Climatol., 27, 739–748, doi:10.1002/joc.1426.

Fracz, A., and P. Chow-Fraser, 2013: Impacts of declining water levels on the quantity of fish habitat in coastal wetlands of eastern Georgian Bay, Lake Huron. Hydrobiologia, 702, 151–169, doi:10.1007/s10750-012-1318-3.

Grundstein, A.J., and D.J. Leathers, 1998: A case study of the synoptic patterns influencing midwinter snowmelt across the northern Great Plains. Hydrol. Process., 12, 2293–2305, doi:10.1002/(SICI)1099- 1085(199812)12:15<2293::AID-HYP797>3.0.CO;2-9.

93 ——, and ——, 1999: A spatial analysis of snow-surface energy exchanges over the northern Great Plains of the United States in relation to synoptic scale forcing mechanisms. Int. J. Climatol., 19, 489–511, doi:10.1002/(SICI)1097- 0088(199904)19:5<489::AID-JOC373>3.0.CO;2-J.

Kalkstein, L., and P. Corrigan, 1986: A Synoptic Climatological Approach for Geographical Analysis - Assessment of Sulfur-Dioxide Concentrations. Ann. Assoc. Am. Geogr., 76, 381–395, doi:10.1111/j.1467-8306.1986.tb00126.x.

——, G. Tan, and J. Skindlov, 1987: An Evaluation of 3 Clustering Procedures for Use in Synoptic Climatological Classification. J. Clim. Appl. Meteorol., 26, 717–730,

Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-year reanalysis project. Bull. Am. Meteorol. Soc., 77, 437–471, doi:10.1175/1520- 0477(1996)077<0437:TNYRP>2.0.CO;2.

Kluver, D., T. Mote, D. Leathers, G. R. Henderson, W. Chan, and D. A. Robinson, 2016: Creation and Validation of a Comprehensive 1° by 1° Daily Gridded North American Dataset for 1900–2009: Snowfall. J Atmospheric Ocean Technol, 33, 857–871,doi:10.1175/JTECH-D-15-0027.1

Leathers, D.J., T. Kluck, and S. Kroczynski, 1998: The severe flooding event of 1996 across north-central Pennsylvania. Bull. Am. Meteorol. Soc., 79, 785-797.

Leathers, D.J., D. Graybeal, T.L. Mote, A. Grundstein, and D.A. Robinson, 2004: The Role of Airmass Types and Surface Energy Fluxes in Snow Cover Ablation in the Central Appalachians. J. Appl. Meteorol., 43, 1887–1899, doi:10.1175/JAM2172.1.

Quinn, F., 2002: Secular changes Great Lakes level seasonal cycles. J. Gt Lakes Res, 28, 451–465.

Suriano, Z.J, and D.J. Leathers, 2017a: Synoptically classified lake-effect snowfall trends to the lee of Lakes Erie and Ontario. Clim. Res., 74, 1-13, doi:10.3354/cr01480.

94 ——, and ——, 2017b: Synoptic climatology of lake-effect snowfall conditions in the eastern Great Lakes region. Int. J. Climatol., 37, 4377-4389, doi:10.1002/joc.5093.

——, and ——, 2017c: Spatiotemporal Variability of Great Lakes Basin Snow Cover Ablation Events. Hydro. Proc., 31: 4229-4237, doi:10.1002/hyp.11364.

Yarnal, B., 1993: Synoptic Climatology in Environmental Analysis: a primer. Belhaven Press (London)., 195 pp.

95 Table 4.1.: Meteorological characteristics in Flint, MI at 0900, 1500, 2100, and 0300 UTC and statistical summary for the ten most common synoptic weather types leading to ablation in the Great Lakes basin.

Southerly Flow Rain-on-Snow High Variable S1 S2 S3 S4 S5 R1 R2 R3 H1 H2

0900 UTC Temperature (°C) -0.2 0.8 -0.1 2.0 -5.6 7.0 1.1 6.3 -4.5 -1.1 Dew point (°C) -3.8 -4.4 -3.0 -2.0 -8.2 4.3 -2.6 1.7 -7.9 -3.5 Sea-level pressure (hPa) 1019 1022 1024 1018 1020 1015 1021 1009 1028 1019 Wind speed (ms-1) 4.6 3.8 1.8 2.6 2.4 4.5 3.4 5.2 1.3 1.9 Wind direction (°) 200 180 178 207 139 197 136 192 325 3 Cloud cover (1/10s) 7.0 3.1 6.5 0.9 3.7 8.9 7.2 6.1 1.0 7.9

1500 UTC Temperature 2.1 7.5 4.3 9.1 -0.9 8.5 5.5 10.7 2.1 0.6 Dew point -2.0 -1.9 -1.4 0.1 -4.4 6.1 -0.2 3.8 -6.8 -3.3 Sea-level pressure 1018 1021 1024 1018 1017 1015 1020 1007 1030 1022 Wind speed 5.4 6.1 3.1 4.3 4.8 5.4 5.5 7.4 0.5 3.1 Wind direction 202 198 177 214 149 198 154 217 25 20 Cloud cover 7.5 5.4 9.6 1.0 7.5 9.6 9.1 9.3 0.5 8.5

2100 UTC Temperature 6.3 13.5 8.9 15.2 4.7 11.6 11.1 13.6 7.2 5.2 Dew point 0.1 1.0 -0.8 0.9 -1.0 8.0 2.8 4.2 -7.1 -3.3 Sea-level pressure 1015 1016 1021 1015 1012 1013 1015 1004 1028 1022 Wind speed 2.3 6.5 2.7 4.6 5.6 5.2 5.6 8.1 0.3 2.8 Wind direction 204 200 180 211 164 202 153 230 244 18 Cloud cover 7.3 8.2 8.9 2.2 9.4 8.9 8.4 8.8 0.7 7.4

0300 UTC Temperature 4.4 9.0 4.7 8.5 4.3 9.4 7.5 7.9 0.3 0.0 Dew point 0.4 2.2 -0.5 1.7 1.4 7.1 3.0 2.9 -6.3 -3.0 Sea-level pressure 1016 1016 1021 1014 1007 1013 1012 1008 1028 1023 Wind speed 4.8 4.7 2.9 3.4 6.8 4.4 5.2 4.8 1.8 2.5 Wind direction 201 184 150 182 168 208 150 243 137 75 Cloud cover 7.0 6.3 8.2 1.6 9.2 8.8 8.6 5.5 1.5 3.7

Total Number of 57 29 28 27 13 30 20 16 17 17 Ablation Events Mean Ablation 4.2 3.6 3.5 3.4 3.5 4.4 3.5 3.6 4.0 3.8 Magnitude (cm) Standard Deviation of 1.8 1.0 0.9 0.8 0.9 1.2 1.1 1.2 1.8 1.1 Mean (cm) Maximum Ablation 12.4 7.1 5.8 5.2 5.3 8.2 7.2 6.9 10.5 6.3 Magnitude (cm)

100

101

102

Figure 4.7.: Snow depth preceding the event (a) and spatial distribution of ablation across the Great Lakes basin for case study 3. Darker shades represent deeper snow depths and greater ablation in cm on March 18, 2001.

103

Figure 4.8.: Inter-annual frequency from 1960-2009 of the general ablation-inducing synoptic categories: southerly flow (top), rain-on-snow (middle), and high-pressure overhead (bottom). The frequencies of the rain-on-snow and high overhead synoptic categories are significantly (p < 0.05) decreasing and increasing respectively.

104 Chapter 5

CONCLUSIONS

The three studies conducted for this dissertation address several interrelated questions regarding the variability of snow in the Great Lakes region and the atmospheric conditions associated with this variability. The dissertation as a whole effectively traces snow through the winter hydrologic cycle, beginning with the atmospheric mechanisms leading to the formation of lake-effect snow, and ending with the processes forcing snowmelt. As evident from the studies, there are many unique synoptic-scale atmospheric patterns that directly influence the accumulation and ablation of snow across the Great Lakes region. The first study used a synoptic weather type classification approach to examine seasonal lake-effect snowfall in the eastern Great Lakes region, and the mechanisms responsible for their change during the period 1950-2009. Prior research indicated snowfall trends in this region underwent a trend reversal, with snowfall declining since 1970-1980. This study supports the non-linear behavior of long-term snowfall trends, specifically for lake-effect snow, which makes up approximately 50% of the region’s total snowfall. While a true trend reversal is not detected in this study, lake-effect snowfall significantly increases from 1950-1979, before exhibiting no significant trends from 1980-2009. In attributing the atmospheric and climatic forcings of snowfall variability and trends, results indicate the variability in the frequency of particular synoptic weather types are primarily responsible for the variability in snowfall, where more (less) frequent types result in higher (lesser) lake-effect snowfall

105 totals. When the impact of long-term changes in the frequency of synoptic weather types on lake-effect snowfall totals is coupled with the impact of long-term changes in the synoptic weather types’ respective snowfall rates, between 89-95% of the observed changes in lake-effect snowfall can be explained. Thus, this study finds the frequency of lake-effect synoptic types and the rate, or intensity, of snowfall associated with them, appear to be the major drivers of lake-effect snowfall changes. The second study developed further quality control for a daily 1° gridded snow dataset, and built a climatology of snow cover ablation events across the entire Great Lakes basin. The quality control procedure tests the consistency of daily snow depth change within each grid cell against meteorologically plausible conditions suitable for melt, using snowfall and temperature thresholds. The procedure brings greater confidence that detected changes in snow depths are grounded in physical changes to the snowpack, and not a result of station observational inconsistencies or measurement errors. Analysis of the basin’s snow cover ablation climatology identifies a clear seasonal cycle in ablation frequency, with March representing the peak month of event frequency regardless of ablation magnitude. Spatially, ablation is latitude dependent, with the zone of peak ablation probability shifting northward while progressing from January to May. This is due to (1) the northern ‘march’ of increasing surface air temperatures, atmospheric moisture, and incoming solar radiation during the spring months providing the necessary energy for melt, and (2) the inability of an ablation event to occur after the snowpack has been completed ablated. While multiple studies have noted a change in the seasonal cycle of snow cover and ablation in regions over North America, in this study, no significant shift in the seasonal cycle of all monthly snow ablation events is detected for the Great Lakes basin. However, long-term

106 changes in the inter-annual frequency of ablation events are detected in two spatially homogeneous regions corresponding to the northern Lake Superior (decreasing) and eastern Lake Huron/Georgian Bay (increasing) drainage basins. The significant increases in ablation events in the Lake Huron/Georgian Bay basins are particularly important as the frequency of large ablation events (5-10cm, 10-20cm, 20+cm) are increasing, suggesting a greater hydrologic impact from such events is likely. The third study builds upon the conclusions of study 2, while utilizing a similar synoptic weather classification methodological procedure as study 1 to quantify the atmospheric conditions that lead to snow ablation across the Great Lakes basin. While a large number of synoptic weather types are found to cause at least one 2.54 cm or greater basin-wide average snow ablation event, the most common and impactful weather types can be categorized into three primary groups: southerly flow patterns, rain-on-snow patterns, and high-pressure overhead patterns. Southerly flow patterns are the most common, representing over 61% of the ablation events studied. These patterns result in the strong advection of warm, and sometimes moist, air northward into the basin from the Gulf of Mexico, providing sufficient latent and sensible heat fluxes into the snowpack to result in melt. Rain-on-snow patterns similarly are driven by warm and moist air advection and are accompanied by liquid precipitation that may also contribute to melt. Arguably, rain-on-snow patterns pose a greater risk for snowmelt-induced flooding. High-pressure overhead patterns lead to snow ablation via clear skies allowing for high levels of incoming solar radiation to impact the snowpack. Both rain-on-snow and high-pressure overhead patterns each cause approximately 20% of the ablation events examined. The three primary ablation- inducing synoptic weather patterns influence the snowpack in a unique manner and

107 exhibit differing inter-annual variability. Over the 1960-2009 study period, long-term linear trend analysis reveals the relative frequencies of high-pressure overhead and rain-on-snow patterns are respectively increasing and decreasing significantly. With rain-on-snow patterns causing the largest amount of ablation per event, a decline in the frequency of rain-on-snow patterns indicate a decreased potential for ablation events by these patterns and thus, a decreased risk for hazardous flooding conditions that often result. Taken together, these studies have highlighted the variability of snowfall and snow cover ablation events within the Great Lakes region of North American, and described the synoptic-scale atmospheric conditions that explain the variability. A major snowmelt event results in over $69 million (2007 U.S. Dollars) in losses on average, while major lake-effect storms can cause millions in environmental and societal disruption. By understanding the connections between atmospheric circulation and snow dynamics, this research established a pathway of the physical conditions that influence snow in the Great Lakes region, and can be used to aid forecasting and budgeting for the potentially harmful consequences of melt and accumulation events in a changing climate. The following represent areas of suggested further research, most of which are currently under investigation by the author. 1. Examine the driving force(s) of lake-effect snowfall intensity changes, with an emphasis on intra-synoptic type variability (the changing character of synoptic types), and changing lake surface and 850 hPa air temperature differences.

108 2. Investigate the long-term changes in basin snow depths, seeking to understand how such changes have contributed to the variability and trends in snow ablation events within the Great Lakes basin. 3. Quantify the snow-surface atmospheric energy fluxes during each of the primary ablation-causing synoptic weather types to fully comprehend the environmental and meteorological conditions the control snowmelt.

109 Appendix

PERMISSIONS

Chapter 2 © Inter Research, Climate Research. Used with permission. Suriano, Z.J. and D.J. Leathers (2017) Synoptically classified lake-effect snowfall trends to the lee of Lakes Erie and Ontario. Climate Research, 74: 1-13. DOI: 10.3354/cr01480.

Chapter 3 © John Wiley & Sons, Hydrological Processes. Used with permission. Suriano, Z.J., and D.J. Leathers. 2017. Spatiotemporal Variability of Great Lakes Basin Snow Cover Ablation Events. Hydrological Processes, 31: 4229-4237, doi: 10.1002/hyp.11364.

110