MODELING THE CLIMATIC SENSITIVITY OF THE INCEPTION OF THE

FENNOSCANDIAN ICE SHEET

BY

RACHEL PASSIG OIEN

THESIS

Submitted in partial fulfillment of the requirements

for the degree of Master in Science in Geology

in the Graduate College of the

University of Illinois at Urbana-Champaign, 2016

Urbana, Illinois

Adviser:

Associate Professor Alison Anders

ii

ABSTRACT

Understanding the sensitivity of ice sheets to climate forcing is crucial for predicting

future behavior of ice sheets. Multiple studies have analyzed the inception of the Quaternary ice

sheets using time-dependent global climate models. The results have indicated that although

carbon dioxide levels, summer insolation, and sea level are important, ice sheet growth is most

dependent on the distribution and quantity of precipitation. I test this conclusion by analyzing

how precipitation and temperature effect the growth of an ice sheet using two different sets of

climate data. I use the Parallel Ice Sheet Model (PISM), a coupled ice sheet and ice shelf model,

to simulate the mass balance and dynamics of the Fennoscandian Ice Sheet (FIS) at the time of

its inception. One climate scenario is constructed using modern temperature and precipitation

records scaled to represent conditions at the last glacial maximum (LGM) with a -7ᵒC

temperature offset and 50% decrease in precipitation. I compare this scenario with the LGM

modeled climate from the PMIP3 palaeo-GCM. When forced with the scaled modern climate,

PISM first initiates glaciation the southern Scandinavian mountains and an ice sheet expands

north and south over 1000 years. The thickest ice remains in the south near the present day

Jotunheimen ice cap. In contrast, the evolution of the ice sheet using the PMIP3 climate features

unrealistically large ice thicknesses along the southern coast of Norway and Sweden, extending

across the Norwegian Channel into Denmark. Glaciers do not propagate along the Scandinavian

mountains to the north despite the presence of very cold temperatures. The difference between

these two models of the inception of the FIS is largely controlled by differences in the spatial

patterns of precipitation between the two climate data sets. A sensitivity test examining the

influence of incremental changes in precipitation and temperature ranging from the present

modern condition to the LGM scaling of modern patterns shows that while temperature limits

growth of the ice sheet to some extent, it does not change the general pattern of ice

iii

accumulation. Precipitation has a much larger influence on ice sheet evolution. I conclude that

the inception of the actual FIS is likely to have been sensitive to patterns of precipitation and

assert that future work on modeling ice sheets must consider the potential impact of small

changes in precipitation on ice sheet mass balance and movement.

iv

ACKNOWLEDGEMENTS

I would like to extend a special thanks to my adviser, Dr. Alison Anders for the

collaboration and expert knowledge regarding the choice to use PISM and guidance on the

project. Thank you to Dr. Ola Fredin (Norwegian Geological Survey) for the advice,

development of the project topic, and the constant questioning in the field which helped

developed the thesis topic. Thanks to David Wojtowicz and Ken Pattern for the availability and

assistance of setting up space on the SESE supercluster. Thank you to Derek for assisting in the

project set up. Thank you to my family whom gave me constant support to keep pushing

forward. I would like to extent a very special thank you to Dr. Andy Aschwanden, for a visit to

Fairbanks for collaboration along with continued support throughout the year. Finally, my

deepest gratitude to Dr. Anna Nesbitt, without her guidance and assistance in the last phase of

this project I do not believe I would‟ve been able to finish my thesis. Her expertise in Python

gave me crucial insight into modeling and figure design.

v

TABLE OF CONTENTS

INTRODUCTION………………………………………………………………………………...1

BACKGROUND………………………………………………………………………………….3

MODEL………………………………………………………………………….………………10

MODEL PARAMETERS AND FIXED BOUNDARY CONDITIONS………………………..12

PALAEOCLIMATE FORCINGS……………………………………………………………….14

EXPERIMENT DESIGN………………………………………………………………………..17

RESULTS………………………………………………………………………………………..18

DISCUSSION……………………………………………………………………………………22

CONCLUSIONS………………………………………………………………………………...26

REFERENCES…………………………………………………………………………………..27

COPYRIGHT PERMISSIONS………………………………………………………………….33

TABLES………………………………………………………………………………………....34

FIGURES………………………………………………………………………………………...35

1

INTRODUCTION

The global significance of potential sea level rise associated with the melting of the

Greenland and Antarctic Ice Sheets is undeniable, yet the response of these ice sheets to future

climate change remains uncertain. The physics of glacier flow, particularly the transitions of

slow moving ice sheets to rapid ice streams and grounded ice to floating ice in ice shelves, is

complex and non-linear, with feedbacks between calving rates, flow speed, strain heating and

basal conditions (e.g. Gunzberger, 2007; Bueler, 2009; Zhang et al., 2011; and Golledge et al.,

2014). The complex behavior of ice sheets and ice shelves makes numerical models

computationally expensive. Furthermore, the sensitivity of glaciers to basal conditions, which are

variable in space and time and impractical to measure directly, increases uncertainty (Zhang et

al., 2011). Most importantly, the fact that both temperatures and precipitation rates are changing

makes glacial mass balance change estimation complicated (Shepherd et al., 2012). The impact

of climate on ice sheets is not well understood, therefore it is crucial to understand how small

changes in temperature and precipitation impact ice sheet mass balance and dynamics.

Here, I examine the sensitivity of Fennoscandian Ice Sheet (FIS) to palaeoclimate

forcing at its inception to gain a better understanding of on controls ice sheet growth. The

Parallel Ice Sheet Model (PISM) solves for mass balance, ice dynamics and geophysical

interactions. I focus on assessing the influence of spatial patterns in temperature and

precipitation on the location and growth of the FIS. The influence of climate on the initial stages

of glacier formation is important because as ice sheets grow they strongly influence the local

climate and initiate feedbacks that favor continued growth of ice (Fredin, 2002). Therefore, they

become less sensitive to external climate forcing as they get larger. Thus, the greatest sensitivity

of glaciers to climate is expected as they are first forming.

2

Specifically, I measure changes in the location, thickness and growth of modeled ice as a

function of prescribed temperature and precipitation patterns. These analyses provide insight into

how the true palaeoclimate drove the growth of the FIS. Two sets of climate patterns are used,

one based on the current precipitation and temperature patterns, scaled to reflect the cooler, drier

conditions at the inception of the FIS, and the last glacial maximum (LGM) conditions from a

global paleoclimate model, PMIP3 (Pollard et al., 2000). These two representations of climate

have distinct spatial patterns in precipitation and temperature which allows for assessment of the

factors that control location of the earliest glaciation and the first phases of ice sheet expansion.

3

BACKGROUND

The Fennoscandian Ice Sheet (FIS), one of the extensive Quaternary ice sheets, is widely

studied due to its striking impact on the topography of Scandinavia. Norway is known for the

dramatic fjords along the western coast and the Scandinavian mountain range which spans the

length of Norway and northern Sweden and has peaks up to 2500m in elevation. The FIS also

produced hummocky topography throughout Sweden, extended into the modern day North Sea

where it carved the Norwegian Channel, and crossed the Baltic Sea into Finland (Figure 1). On

the continental shelf of the northern Atlantic Ocean extensive sedimentary plateaus were

deposited during the Quaternary which provide records of both climatic conditions and extent of

the FIS.

I am going to briefly summarize what is known to date about the glaciological inception

of the FIS, which most likely occurred post-Eemian interglacial during the early Weichselian,

between 115ka-100ka (Mangerud et al., 1981b). The last interglacial period in northern Europe

occurred during Marine Isotope Stage-5e (MIS-5e) which occurred about 125-115ka (Colleoni et

al., 2014). Palaeoclimate records from this period, sedimentological records, and numerical

models provide insight the inception. The timing is constrained by four sedimentological cores

along the southwestern coast of Norway near JT to SV (Figure 1). The presence of glaciomarine

silt indicates the existence of mountain glaciers at approximately 115ka (Mangerud 1981a;

Mangerud et al., 1981b; Fredin, 2002; and Mangerud, 2011). In a review paper, Fredin (2002)

concludes that sedimentological records and some early models support a southern Scandinavian

mountain nucleation area. Several numeric models have simulated the inception of the northern

hemisphere ice sheets during the last interglacial period, MIS-5e (e.g. Siegert et al., 2001;

Kageyama, 2004; Kubatzki, 2006; Colleoni et al., 2014).

4

The Eemian-Weichselian transition is a very difficult period for chronological constraints

because of the destructive nature of ice sheets and that it is older than radiocarbon capabilities.

While there have been analyses of multiple sediment cores and speleothems, the constraints on

the spatial of extension during Marine Isotope Stage-5d (MIS-5d) to Marine Isotope Sage- 4

(MIS-4) is extremely spotty across Scandinavia (Mangerud, 2004; Helmens, 2013). From

multiple studies of sediment cores, pollen data, and cave speleothems using U/Th dating, I have

constructed a map showing the known pattern of ice coverage during the Early Weichselian

(Figure 2) (Helle et al., 1981; Mangerud et al., 1981a; Mangerud, 1991; Baumann et al., 1995;

Lauritzen, 1995; Fronval et al, 1997; Larsen et al., 2000; Olsen et al., 2001; Mangerud, 2004;

Raunholm et al., 2004; Lekens et al., 2009; Skoglund et al., 2010; Mangerud, 2011; Olsen et al.,

2013; Helmens, 2013).

Okshola in Skjerstadfjorden and Stordalsgrotta cave systems are the two points near 76-

78◦N (Figure 2) (Lauritzen, 1995; Skoglun et al., 2010). The oxygen isotopic record in the cave

indicates vadose conditions which mean that the cave is dry and the speleothems are growing

(Lauritzen, 1995; Skoglun et al., 2010). The growth of speleothems implies that the cave was not

topped by a mountain ice cap or and ice sheet until 71-78ka (Lauritzen, 1995). At this time there

is a distinct change in the cave conditions to indicate an ice sheet (Lauritzen, 1995; Skoglun et

al., 2010). While there are cooler conditions throughout the period 130-71ka, which may indicate

alpine glaciers, the resolution of the timing and climatic signals in the speleothems cannot be

further constrained (Lauritzen, 1995; Mangerud, 2004; Skoglun et al., 2010; Mangerud, 2011).

Additionally, the Solki sediment record in northern Finland (point not shown) identifies that

there was not glaciers covering the area until 80-90ka (Helmens, 2013). The point in the

Norwegian Sea is a sediment core that records ice rafted debris (IRD). Due to the ocean currents,

5

the glacial debris is most likely from ice reaching the ocean on the southwestern coast (Fronval

et a., 1997). The three points along the southwestern coast are sedimentological cores (top

(Godøya), middle (Fjøsanger), and bottom (Jæren)) that have MIS-5d tills on top of Eemian

sediments (Mangerud et al., 1981a; Mangerud, 1991; Baumann et al., 1995; Larsen et al., 2000;

Olsen et al., 2001; Mangerud, 2004; Lekens et al., 2009; Mangerud, 2011; Olsen et al., 2013).

Finally, the most eastern point (Figure 2) is a record with Eemian sediments followed by two tills

interspaced with a peat layer (Helle, 1981). The first till corresponds to MIS-5d (Helle, 1981).

Collectively, these records give a general timing and location of the FIS inception.

There is substantial evidence concluding that the presence of warm Nordic seas caused a

delay in the inception of the FIS during MIS-5e (Born et al., 2010). With few reconstructions

continuing throughout last glaciological cycle, most northern hemisphere ice sheet modeling

jumps from the last interglacial (115ka) to the last glacial maximum at approximately 22ka

(Mangerud, 2011). The inception of the FIS has been placed in the early Weichselian

(Mangerud, 2011) which is generally overlooked in ice sheet numerical modeling. While there

has been previous work on the timing of the inception (Figure 2), I will be focusing on the

questions regarding the location and cause of the development of the FIS.

The transition from the previous interglacial, the Eemian, into the last glaciological cycle,

known as the Weichselian in Scandinavia, has been studied through numerical modeling,

paleoclimate reconstruction, and geomorphic data (Kubatzki, 2006, Mangerud, 2011, and

Colleoni et al., 2014). The glaciological inception of the FIS is defined as the change in

atmospheric and oceanographic conditions to a climate that can produce perennial ice on

continents which does not necessarily coincide with the earliest identifiable landforms due to ice

6

sheet growth (Stroeven et al., 2002). The MIS- 5e marks the largest change in summer insolation

at the higher latitudes of the northern hemisphere (Colleoni et al., 2014).

Sea level during this period is not well constrained. Colleoni et al. (2014) and Kubatzki

(2006) find that at 125ka sea level was four to ten meters above modern mean sea level;

Welbroeck (2002) places sea level at present-day levels up to 2m above mean sea level (a.m.s.l)

at 125ka. Lambeck and Chappell (2001) conclude that that at 125ka sea level was near present

day levels but it fell by five to six meters between 125-115ka due to the growth of the Laurentide

ice sheet. It can be assumed that sea level was approximately near modern height at the end of

the last interglacial because there was a similar amount of ice in the northern hemisphere as at

present (Coyne et al., 2007).

The Vøring plateau in the Northern Atlantic Ocean has provided a well-dated

sedimentological record of the past glaciations (Mangerud et al., 1979; Mangerud et al., 1981a;

Mangerud, 1981b; Fredin, 2002). Specifically, the sedimentary facies reflect a well-defined

transition from near-shore marine silts into basal tills (Mangerud et al., 1979; Fredin, 2002). The

sedimentary records also provide palaeoclimate information based on the types of foraminifera

present. Species such as Neogloboquadrina pachyderma, Globigerina pachyderma and

Globigerina quinqueloba are found in abundance when sea surface temperatures (SSTs) are

between 0 and 5◦C (Ehrenberg, 1861, Mangerud et al., 1979; Siegert et al., 2001; Risebrobakken,

2007). Siegert et al. (2001) suggests that the profusion of these particular species on the Vøring

plateau indicates a high-productivity zone in the North Atlantic Ocean during MIS-5e. The

occurrence and concentration of these foraminifera species also imply warm northern waters and

a potential precipitation increase across the western Eurasian Artic (Siegert et al., 2001). An

abundance of benthic foraminifera and pollen found within the Fjøsanger sediments, located just

7

off the coast, west of JT (Figure 1) represent a low-boreal habitat, a prime indication of a warm

climate (Mangerud et al., 1979; Mangerud, 1981a, Mangerud et al., 1981b). The palaeoclimate

data previously discussed correlates with a period of warm Nordic seas which is widely agreed to

have delayed the glaciological inception of the Fennoscandian ice sheet from 119ka until 114ka

(Born et al., 2010).

Numerical models can simulate the entire evolution of a whole ice sheet through time

whereas the sedimentary record is always limited in time and space by the available samples.

Recent numerical models have a few key features which are crucial for more accurately

portraying ice sheet behavior than was possible prior to the first decade of the 21st century.

Inception studies of northern hemisphere ice sheets use 3D thermomechanical ice sheet. Models,

(e.g. GREMLINS, CESM, GLIMMER, and SICOPOLIS), all of which allow grounded based ice

to flow through internal deformation (e.g. Kageyama, 2004; Kubatzi, 2006; Charbit et al., 2007;

Risebrobakken, 2007; Bonelli et al., 2009; Born et al., 2010; Gregory et al., 2012). Only one of

the models previously used for Northern hemisphere ice inception studies CESM with GRISLI

simulates ice streams, ice shelves, and steeply sloped areas such as fjords and U-shaped valleys

which are extremely prevalent in the Scandinavian mountains more realistically through a

shallow shelf approximation (Colleoni et al., 2014). The climate inputs used in previous northern

hemisphere inception models are global climate models (GCMs) such as IPSL CM4, CLIMER -

2, PMIP1, and FAMOUS, which at the highest resolution are modelled on a 40km x 40km grid

(e.g. Kageyama, 2004; Kubatzi, 2006; Risebrobakken, 2007; Charbit et al., 2007; Bonelli et al.,

2009; Born et al., 2010; and Gregory et al., 2012). Inception studies including Colleoni et al.,

(2014), Zweck and Huybrechts (2002), Born et al., (2010) and Gregory et al., (2012) use the

positive-degree-day (PDD) scheme from Reeh (1991) to model mass balance. The PDD feature

8

is important for ice sheet modeling because it includes in natural variability. In summary,

previous models have included representations of internal deformation, sliding in steeply sloped

areas, a variety of climate inputs, and a minimum resolution of 40km.

Prior models examined the roles that sea-level changes, vegetation, summer insolation

and carbon dioxide levels have played in the inception of the FIS (e.g., Siegert et al., 2001;

Kageyama, 2004; and Kubatzki, 2006). Colleoni et al. (2014) studied the effect of scaling the

winter and summer temperature vertical lapse rates, along with prohibiting the temperatures from

changing with altitude to see the effect on inception of the ice sheets in Eurasian. While their

study did not have ice nucleation within the Scandinavian mountains as the FIS is through to

have formed, their Eurasian ice was located northern Finland, eastern Russian and extending

across the Barents Sea to Novaya Zemlya. Their study concluded that the Eurasian ice was more

sensitive to the lack of precipitation in comparison to the North American ice sheet (Colleoni et

al., 2014). The models discussed in the review paper by Fredin (2002) concurred that the

southern Scandinavia mountains are the geographic location of the inception of the FIS (Fredin,

2002). The results of Gregory et al., (2012) show the FIS is thickest in the southern mountainous

region of Norway due to large precipitation rates in this region. The main commonalities of

Fredin, 2002; Zweck and Huybrecths, 2005; Charbit et al., 2007; Gregory et al., 2012, is that the

inception should occur in the highest peaks of the Scandinavian Mountains.

The impact of coarse resolution of GCMS on ice sheet evolution has not previously been

studied. The key unknowns include addressing the use of coarse global climate models for

modeling. Scandinavia‟s complex topography influences precipitation patterns which can

reasonably be expected to impact the inception of a mountain-based ice sheet. A climate dataset

with a resolution of 40km x 40km cannot capture important spatial variability in precipitation

9

expected in mountains. Additionally, a model can better represent an ice sheet forming in steeply

sloped mountainous areas using the SSA. A caveat is that when a model uses a low resolution

atmospheric GCM paired with low resolution topography, it will be unable to capture the

important climatic information due to changes in local topography. When a model uses measured

temperature and precipitation the information includes essential data, even when the

measurements are smoothed to a coarser gridding, there is still some influence of the real

topography within the climate. I use a modern climate precipitation dataset to capture the

influence of the modern precipitation due to the Scandinavian mountains. This was compared to

a GCM constructed for the same climatic conditions that the modern climate was scaled at in

order to compare the influence of grid resolution.

10

MODEL

A numerical model with specific characteristics was needed to answer the research

questions described above. The domain size of the model needed to be able to encompass the

entire FIS ice sheet at a minimum resolution of 20km. This spatial resolution is necessary due to

the complex topography, particularly the fjords along the west coast of Norway, for the where

the development of fast shear zones is anticipated. Simulations need to be conducted over

timescales of hundreds to a few thousand years for multiple climate conditions, necessitating a

computationally efficient model. In addition, the model needs to be appropriate for both shallow

sloping terrestrial areas and steeply sloping mountains. In some areas of an ice sheet, internal

deformation is the dominant force. In these areas the model needed to simulate frozen-based ice

through the shallow ice approximation (SIA) (Figure 3a). Where basal sliding is significant, such

as ice streams, ice shelves and steeply sloped areas, the shallow shelf approximation (SSA)

contributes significantly to calculated velocities (Figure 3b). Finally, the model must be able to

incorporate various climatic data sets and couple them into the growth of an ice sheet.

The Parallel Ice Sheet Model (PISM) can achieve all of the necessary functions. PISM is

a free, community source, large scale, ice-sheet ice-shelf coupled model created by Bueler and

Brown (2009) that can be run at a high spatial resolution using a desktop computer and scaled to

run in parallel on a supercomputer. I used PISM version 0.7 accessed from http://www.pism-

docs.org/wiki/doku.php?id=stable_version, which is the same as the Potsdam Parallel Ice Sheet

Model (PISM-PIK) as cited in Winklemann et al. (2011) and Martin et al. (2011). PISM

approximates ice sheet flow properties with the SIA and SSA, greatly reducing computational

expense as compared to full Stokes models, allowing the investigation of large ice sheets.

11

Furthermore, PISM has previously been applied to fully terrestrial ice sheets, at high spatial

resolution and varying timescales (Golledge et al., 2012, Golledge et al., 2014).

12

MODEL PARAMETERS AND FIXED BOUNDARY CONDITIONS

PISM model computations were carried out using the Keeling supercluster at the School

of Earth, Society, and the Environment, at the University of Illinois at Urbana-Champaign.

Model runs utilized up to 56 CPUS and took between 3 and 36 hours to complete. Model output

data were plotted using Python‟s Matplotlib Basemap toolkit.

PISM has numerous options, I used the basic setup for Greenland (Adalgeirsdottir et al.,

2014) because of the similarity of the FIS and the Greenland ice sheet in terms of size and

importance of steeply sloped topography. Some deviation from the Greenland model parameters

was necessary to maintain consistency with current understanding of the FIS. Specifically, I

changed calving parameters to allow for the formation of ice shelves, which are not significant in

Greenland, but were prominent in the FIS. A calving thickness threshold of 50m and an Eigen

calving K value of 1e17 were set. The calving rate is the rate of mass transport of the ice to the

shelf terminus. Another deviation from the Greenland setup was the pseudo plastic value for q,

which was changed to 0.25. This exponent controls the basal sliding velocity in the sliding law

equation (4) from Adalgeirsdottir et al., (2014). This is the recommended default value for PISM

but it not the choice for Greenland.

The topographic elevation is fixed for all simulations at modern values. The spatial

resolution of the model is 5km with an approximate latitude range of 0°E-55°E and longitude

range of 50°N-80°N. Bedrock topography and bathymetry were extracted from the ETOPO1

Global Relief Model, a data compilation provided by NOAA at 1 arc-minute resolution (Fig. 1;

Amante and Eakins, 2009). Reference locations for all simulations are given in Figure 1. Sea

level was fixed at modern height (0m), reflecting the slow rate of sea level change at the time of

inception. Between125ka to 115ka sea level was estimated to change by only 5-6m (Lambeck

13

and Chappell, 2001), therefore, the sea level offset over the 1000 years of the simulations would

be negligible.

14

PALAEOCLIMATE FORCINGS

The goal of this study is to examine how climate influenced the inception and early

evolution of the FIS. I assume that spatial variability in precipitation and temperature are likely

to be strong controls on the location of glacial inception. Two separate experiments were

conducted. The first experiment used modern climate data (SMC) and the second used a global

climate model based on LGM conditions (PMIP3). The modern data was scaled to represent the

climate cooling which would be necessary to trigger glaciation. All input data were projected

from WGS 1984 geographic coordinates onto UTM zone 32N using Python module PyProj, after

which they were interpolated onto a 5-km resolution Cartesian grid using SciPy and output as a

PISM-ready netCDF4 file.

In all of these experiments, mass balance is determined with a positive-degree day (PPD)

scheme, which calculates snowfall, snow melt, and snow refreeze based on air temperature,

precipitation, elevation, and a linear temperature threshold (Fischer et al, 2014). Random,

normally-distributed “white-noise” is added to the temperature cycle to simulate daily variability.

The “positive-degree days” are those that contribute to melting. The modeled climates have a

single averaged precipitation and temperature for each month and each 20km grid cell.

As the ice sheet grows, I account for expected changes in temperature due to increased

elevation. A standard temperature lapse rate of -6◦C/km was imposed (Adalgeirsdottir et al.,

2014). Precipitation rates did not change with changes in elevation of the ice sheet surface.

Modern climate is justified as a starting point for this experiment because of the

similarities between the Eemian and present-day interglacial periods in terms of environment

(Kukla et al., 2002) and sea-level (Lambeck and Chappell, 2001; Kubatzki, 2006; Coyne et al.,

2007; Born et al., 2010; and Colleoni et al., 2014). Obviously, considerably more and better

15

quality data are available for the modern climate than for the Eemain interglacial climate of

Scandinavia. Importantly, the modern measurements of climate are influenced by the complex

topography at all scales. The impact of complex topography on precipitation rates, in particular,

is notoriously hard to reproduce in large-scale climate models, and makes GCMs far from ideal

for examining the importance of spatial patterns in precipitation (e.g. Yorgun and Rood, 2014).

The first experiment was conducted using modern climate data, with monthly air

temperatures and precipitation rates across terrestrial and oceanic parts of the domain. This

modern climate data was acquired from multiple sources. The modern climate, land temperature

and precipitation observations were from the E-OBS dataset from the EU-FP6 project

ENSEMBLES (http://ensembles-eu.metoffice.com) by the data providers in the ECA&D project

(http://www.ecad.eu). This dataset came at gridded in 1degree latitude and longitude intervals.

There was not any missing data from the daily raw temperatures. The daily land temperature data

from 1950 to 2015 was averaged for each day of the year and then averaged into monthly values.

The 60-80◦N ocean surface temperature dataset was accessed from the National

Oceanographic Data Center Arctic Regional Climatology. This dataset was monthly averaged

temperatures at 1degree latitude and longitude intervals. The dataset for 50-60◦N was monthly

averages of ocean surface temperature data accessed from the GIN Seas Regional Climatology in

1degree latitude and longitude intervals. There is no apparent offset between these data sets at

the boundary between them. A nearest neighbor approximation was used to interpolate across the

land ocean boundary for each of the monthly averages to get a spatially continuous map of

monthly temperatures. Land and ocean temperatures were merged to produce continuous

monthly average temperature maps (Figure 4). Two time slices, January and July temperatures

are shown in Figure 3 to show the seasonal influence.

16

Precipitation data were similarly used to produce maps of monthly precipitation totals.

The average precipitation for each month was calculated for the period 1950-2015. Over land,

the E-OBS dataset was used. The land precipitation data was complete and did not require any

interpolation. Precipitation data over the ocean is sparse, making estimation of spatial and

temporal patterns of monthly precipitation totals impossible at the model grid resolution.

Therefore, a spatially and temporally uniform precipitation rate of 10mm/month was assumed

over the ocean based a coarse global climate model (Jones et al., 1998). The consequence of

using such a low precipitation is the ice would be „starved‟ once it reaches ocean cells.

Precipitation over the oceans is assumed to be unimportant for the inception of the FIS because it

began on land. Precipitation is displayed in Figure 5 and to highlight the influence of significant

autumn precipitation, August and January were displayed

The modern-based climate experiment explores how a modern spatial variability in

precipitation and temperature would have influenced ice sheet inception. Using a palaeo-global

climate model allows us to look at spatial patterns in precipitation and temperature that are

independent of modern conditions. The global climate model used for the second experiment was

acquired from the Paleoclimate Modelling Intercomparison Project (PMIP3) database via the

Earth System Grid Federation (ESGF) web portal (Peterschmitt et al., 2013). These data were

provided by NCAR‟s Community Climate System Model 4 (CCSM4) at a resolution of 288 x

192 x L26 for the globe (approximately 1.3ᵒ in the domain). Specifically, I used the monthly

averaged near surface air temperature (tas) (Figure 4, PMIP3) and precipitation (pr) (Figure 5,

PIMP3) from the 21ka PMIP3 simulation. This particular time slice represents the climatic

conditions during the Weichselian LGM. The difference of the two climates was taken by

subtracting the PMIP3 from the SMC to show the variance in the two palaeo forcings.

17

EXPERIMENT DESIGN

The experiments are designed to identify the conditions which would initiate the FIS, the

location where the FIS first forms and, the early pattern of growth. I chose to hold the climate

conditions constant throughout the simulations. Therefore, I chose a reference run using the

modern climate scaled (SMC) to LGM conditions, which was a -7◦C offset in each grid cell and

50% decrease in local precipitation (Figure 6). I examined the first 1000 years of evolution. The

sensitivity of the inception of the FIS to climate was explored by examining perturbations of

precipitation and temperature relative to the reference run. Specifically, I changed temperature in

two degree intervals from a -1◦C offset to -9

◦C offset while holding precipitation fixed at 50% of

modern (Figure 7). Then I changed precipitation by 25% increments for a fixed temperature at -

7◦C offset (Figure 8). I also compare the reference run to the LGM conditions of the PMIP3

GCM (PMIP3) (Figure 9). Again, the sensitivity of the PMIP3 climate was studied through one

degree intervals from a -5◦C offset to -9

◦C offset while holding precipitation fixed at 100%. Then

I changed precipitation by 50% increments for a fixed temperature. The difference in the spatial

patterns of precipitation and temperature between the modern climate and the PMIP3 climate

provides insight into how spatial patterns in climate (SMC vs. PMIP3) may have influenced the

inception and early evolution of the FIS.

18

RESULTS

In the literature, there are not any current numerical models analyzing the FIS at the

inception during the last glaciological cycle or during the last glacial maximum (Weichselian).

The lack in modeling of the FIS was the motivation for these experiments. A comprehensive list

of the completed simulations is found in Table 1. To test the viability of the modern climate, a

simulation was completed to test the location of ice accumulation to the present day location of

ice in Scandinavia. At the 20km resolution, the modeled ice accumulation is centered on the

current location of the present day Jotunheimen ice cap in Norway. Additionally, to highlight

the dissimilarities of the SMC and PMIP3, the difference (SMC minus PMIP3), is shown in the

palaeoclimate forcing figures 4 and 5. In the temperature difference figure, when the cell is red,

this indicates that SMC was warmer than the PMIP3 climate. In some locations during the month

of January, SMC is up to 40◦C warmer than the PMIP3 GCM. The blue in the precipitation

difference diagram indicates that the PMIP3 climate was wetter than the SMC. During the

autumn rains, the PMIP3 was up to 150mm more than the SMC. Finally, to test the accuracy of

the 20km grid resolution, a high resolution (5km by 5km) simulation of the reference run was

completed (Figure 10). The location of the inception and thickness reached for the high and

lower resolution experiments was consistent. The higher resolution run does show the details of

ice accumulation within the western fjord systems.

In the reference run, the FIS initiates in the southern Scandinavian mountains and not in

the northern latitudes. Figure 1 has various locations in Norway to use as reference points for ice

growth: Stavanger (SV), Trondheim (TR), Andøya (AY) and Jotunheimen (JT). The ice cap

grows to a thickness of 500m within 200 years in the same location as the present day

Jotunheimen ice field (Figure 6a). Elsewhere in the domain, the ice is significantly thinner and

19

nonexistent in the north. Thick ice is present in the high mountains and where precipitation is

heavy. The ice continues to thicken over time to the south and expand to the north toward AY.

Within the first 600 years of the experiment, the ice primarily accumulates around JT until it has

reached a thickness of 500m. Then the ice cap expands north to TR and south to SV along the

mountain divide. The zone of maximum ice thickness is initially near the coast but the ice divide

moves eastward by 800 years. Ice thickens rapidly from 0-600 years and the rate of thickening

slows from 600-1000 years. Finally, at 1000 years (Figure 6e) the most western fjords from SV

to JT have experienced ice thinning to 400m. The ice divide continues to move eastward with the

maximum ice thickness remaining at about 900m. There are only a few places that reach the

maximum ice thickness of 1400m. There is a continuous ice field that extends from TR to AY

with thicknesses up to 900m. North of AY, the ice field thins to 300m.

A sensitivity experiment was conducted to analyze temperature and precipitation

independently from one another in order to decipher the effect each parameter had in ice sheet

inception. The first set of runs was conducted with a constant 50% of precipitation of the MD

input while varying temperature at 2 degree offsets. Simulations were conducted for -1, -3, -5, -7,

and -9ᵒC offsets (Figure 7). The spatial pattern of ice thickness was similar for all temperature

offsets. When the temperature offset was -1ᵒC or -3

ᵒC the maximum ice thickness reached in the

high mountains was about 1200m. When the temperature offset is greater than -3ᵒC the

maximum ice thickness reached after 1000 years of evolution is 1400m. When temperature is

offset by at least -5ᵒC, the spatial pattern of ice accumulation has a noticeable change. At five

degrees of cooling relative to today, the ice sheet has limited thickness near AY and the primary

ice field remains along the western edge of Norway near JT. With offsets of -7ᵒC and -9

ᵒC, the

ice divide shifts inland from the west coast toward the known position of the FIS ice divide at the

20

LGM. At offsets less than -9ᵒC the ice remains thin between TR and AY, but when equal to -9

ᵒC

the ice is continuously thick along the high topography between these regions.

The second set of sensitivity experiments was conducted with a constant temperature

offset of -7ᵒC from the modern data. Precipitation rates of 25%, 50%, 75%, and 100% of modern

precipitation were examined. Precipitation is a stronger control on ice thickness at 1000 years

than temperature (Figure 8). The spatial distribution of ice strongly differs with precipitation. At

25% of modern precipitation (approximately half of the estimated LGM precipitation) ice is

significantly thinner than in the reference run. For precipitation greater than and equal to 75% of

modern values there is a significant buildup of ice along the Scandinavian mountains along with

the movement of the ice divide to the east towards Sweden There is also significant ice growth in

Denmark and on the continental shelf of the Baltic and North Seas with ice thicknesses up to

1200m.

The PMIP3 climate is based on a global climate model that includes the topographic

effects of the FIS. The precipitation and air temperature from the 21ka PMIP3 simulation,

therefore, show the influence of a large ice sheet. As shown in Figure 4, the cooler air

temperatures extend to significantly lower latitudes near present day southern Sweden and

Denmark than the SMC (Figure 9). Precipitation and temperature followed the same geographic

pattern. The first PMIP3 GCM experiment was held at the same LGM conditions as the

temperature and precipitation offsets of the SMC climate. The maximum ice thickness in the

GCM model occurred near SV and to the south across the North Sea into Denmark. Thick ice did

propagate to the north but not to the extent seen in the reference run. Moreover, the ice divide did

not shift inland as in the reference run. During the PMIP3 GCM sensitivity experiment (not

shown), temperature controlled ice thickness to a similar degree as the SMC experiment, but the

21

spatial pattern of ice accumulation was different. As in the SMC experiments, the effect of

precipitation was larger than the effect of temperature. In particular, at 25% of GCM-modeled

precipitation, the ice sheet is very limited at this scaled precipitation.

22

DISCUSSION

The goal of this project was to understand how temperature and precipitation affected the

formation and early growth of the Fennoscandian Ice Sheet. The key findings were that changes

in precipitation had a more significant impact on ice sheet growth than a similar magnitude of

temperature changes. A change of two degrees Celsius only effected a minimal change of 200m

of thickness in ice sheet growth. The results of the SMC sensitivity test confirmed that warmer

temperature offsets decreased the total ice thickness but did not change the location of ice

accumulation within the thousand-year time period. In contrast, changes in the quantity of

precipitation both limited the thickness and changed the location of ice accumulation.

Jotunheimen (JT) is a present day ice field, thus it is expected to accumulate ice under

conditions similar to modern. In my experiments, JT is an important area of ice accumulation.

Additionally, AY is well known for the record high-velocity ice streams during the last

glaciological cycle and was the prominent location of ice accumulation in the northern region of

Norway (Manguerd et al., 2011). SMC set at precipitation of 75% with a -7ᵒC offset supports ice

accumulation at the location of the present day JT (Figure 8c). The SMC-based model is in

agreement with geologic records (Mangerud et al., 1981a; Mangerud, 1981b; Fredin, 2002). In

contrast, the PMIP3-based simulation does not follow the known geologic ice positions and

instead produces a significant ice sheet in southern Sweden extending into Denmark across the

Norwegian Channel (Figure 9).

I compare the simulations based on the SMC with those using a palaeoclimate GCM

developed at LGM conditions. The FIS, at the LGM extent, itself is an important factor in

controlling the climatic pattern of the PMIP3 reconstruction. The presence of the full FIS

23

impacts the distribution of temperature and precipitation in PMIP3 (Figure 9) and, therefore,

created a significant different in the pattern of ice thickness along the western and southern coast

of Norway as compared with a scaled modern climate. This demonstrates the potential impact of

spatial variability in climate for growth of the ice sheet. The temperature gradient across Sweden

and Finland is reasonably uniform because this GCM was built with a 2-3km ice sheet in this

region. The PMIP3 climate results in an ice sheet growing along the southern coast of Norway,

southern Sweden, and across the English Channel. This result is not consistent with the

geological records (Fredin, 2002).

The clear differences between the SMC and GCM illustrate the impact of spatial patterns

in precipitation and temperature on modeled FIS initiation. It is worth noting that there is a

significant range in temperatures and precipitation between the two palaeoclimate forcings. The

PMIP3 GCM is up to 40◦C colder during the winter than the SMC reference run. In addition, the

PMIP3 GCM during the autumn rains is up to 150mm/month wetter than the SMC. The extreme

cold and wetness allow for a thicker ice sheet to develop. Even though the exaggerated climate

conditions allow for thicker ice, the geographic pattern is not consistent with any geologic

evidence (Helle et al., 1981; Mangerud et al., 1981a; Mangerud, 1991; Baumann et al., 1995;

Lauritzen, 1995; Fronval et al, 1997; Larsen et al., 2000; Olsen et al., 2001; Mangerud, 2004;

Raunholm et al., 2004; Lekens et al., 2009; Skoglund et al., 2010; Mangerud, 2011; Olsen et al.,

2013; Helmens, 2013). While the PMIP3 GCM has clear climatic differences to the SMC it is

good to keep in mind the coarse resolution of the modeled climate. Global climate models are

typically created at a very coarse scale resolution. In the case of the FIS, orographic precipitation

would likely have been important for ice sheet growth. Therefore, a mesoscale climate model

may be necessary to reproduce precipitation patterns relevant to the inception of the ice sheet.

24

Previous studies suggest how different factors control the growth of the FIS (e.g.,

Gregory et al., 2002; Charbit et al, 2007; Born et al., 2010; Colleni et la., 2014). For example,

Charbit et al., (2007) shows how six different atmospheric general circulation models predict

significant differences in air temperature and precipitation that create differences in the spatial

patterns of accumulation of ice in the northern hemisphere. My conclusions confirm that the

palaeoclimate input strongly controls spatial patterns in ice accumulation. The autumn rains

show the importance of a monthly climatic scheme and its influence on ice sheet accumulation

instead of an annually averaged climate. The importance of precipitation in controlling ice

thickness is further evidenced by the similarity in the spatial patterns of precipitation (Figure 5)

and ice thickness (Figures 6 & 9).

My experiments did not include a coupled ocean model. However, it has been suggested

that warm Nordic waters delayed the inception of the FIS in the northern latitudes of Scandinavia

from 119ka to 115ka (Born et al., 2010). My simulations include a southern starting point for the

FIS with gradual expansion to the north, in absence of a differential north-south near-surface

temperature gradient relative to the present. This result suggests that no special oceanic

conditions are required to initiate the FIS earlier in southern Norway than in areas farther north.

The inclusion of a variable ocean circulation model within an ice sheet model may be able to test

how sensitive the northern region is to changes in ocean surface temperatures in the future.

Another key future refinement of the model would be the use of precipitation and

temperature fields derived from a paleoclimate model of a more appropriate timeframe for the

inception than the 21ka LGM simulation used herein. Specifically, a time-dependent simulation

of the Eemian climate would better represent the true history of the region and may help better

constrain the influence of climate on the inception of the FIS. Currently, groups from Bristol

25

University and Kiel, Germany are working on palaeo-climate simulations of the Eemian in

Scandinavia. Contact and updates for this future project can be found on:

https://wiki.lsce.ipsl.fr/pmip3/doku.php/pmip3:design:li:final. With this model it will be

important to note the scale at which the model was created at to see if the orographic effect plays

a role in the precipitation pattern throughout Scandinavia. If the GCM for the Eemian is still at a

coarse resolution, even if it is the correct time period, it may not accurately represent how the

local precipitation patterns that are affected by the mountain range.

The question that remains is how much do we know about the Eemian climate? What is

the best way to record palaeo-precipitation? What resolution is needed to capture the orographic

effect of the Scandinavian mountains on precipitation? These are important questions for future

researchers investigating the inception of an ice sheet, particularly in areas of complex

topography.

Precipitation quantities and spatial distributions had stronger impacts on the location and

growth of ice sheets than temperature for both sets of palaeoclimate conditions examined in this

study. This finding suggests that it is important to consider how minor changes in precipitation

may affect ice sheet growth. Unfortunately, precipitation is not well constrained for many palaeo

records (Baker, 2009; Jones et al., 2010). Better constraints on palaeo precipitation would be

necessary for a higher resolution investigation Eemian climatic conditions which induced the

inception of the Fennoscandian Ice Sheet.

26

CONCLUSIONS

Modeling results emphasize the importance of precipitation patterns and amounts for the

inception of the FIS. Precipitation was the determining factor in where and how much ice

accumulated during the thousand-year simulation of inception of the FIS. Changes in

precipitation dramatically changed the location and growth of an incipient ice sheet.

Understanding the importance of a climatic model resolution is important to understand what

features will be prominent in each palaeoclimate simulation. A scaled modern climate leads to

early ice accumulation in the southern Scandinavian mountains that is consistent with geologic

records. Since modeled ice accumulation coincides with the present day JT ice cap, it suggests

that my assumption of a similarity between the Eemian interglacial and modern climate is

reasonable. The results of ice accumulation from the PMIP3 climate do not correlate with

geologic evidence which suggests that it was not the appropriate choice for the inception. Up to

two degree changes in temperature do not affect the pattern of modeled ice accumulation.

Temperature only impacts the ice thickness obtained within the thousand-year duration of the

experiment. This decreases the importance of higher resolution palaeoclimate records of

temperature and highlights the importance of obtaining the best possible record of palaeo

precipitation patterns.

27

REFERENCES

Abe-Ouchi, Ayako. Heinz Blatter, and Atsumu Ohmura. 1994. How does the Greenland ice sheet

geometry remember the ice age?. Global and Planetary Change. Vol 9. 133-142.

Adalgeirsdottir, Tolly, Antje Fitzner, Nick Golledge, Tore Hattermann, Moritz H• utten, Thomas

Kleiner, Marianne Madsen, Malou Maris, Tim Morey, Mirena Olaizola, Christian

Rodehacke, Nathan Shemonski, Sebastian Simonsen, Anne Solgaard, Ben Sperisen,

Synne H_yer Svendsen, Martin Tru_er, Shuting Yang, Ryan Woodard. 2014. PISM, a

Parallel Ice Sheet Model: User‟s Manual. Accessed from: http://www.pism-

docs.org/wiki/doku.php?id=overview

Amante, C. and B.W. Eakins. 2009. ETOPO1 1 Arc-Minute Global Relief Model: Procedures,

Data Sources and Analysis. NOAA Technical Memorandum NESDIS NGDC-24.

National Geophysical Data Center, NOAA. doi:10.7289/V5C8276M.

Baker, Paul A. 2009. Paleo-Precipitation Indicators. Encyclopedia of Paleoclimatology and

Ancient Environments. Encyclopedia of Earth Sciences. 746-748.

Baumann, Karl-Heinz, Klas S. Lackschewitz, Jan Mangerud, Robert F. Spielhagen, Thomas C.

W. Wolf-Welling. Rudiger Henrich and Heidemarie Kassens. 1995. Reflection of

Scandinavian Ice Sheet Fluctuations in Norwegian Sea Sediments during the past

150,000 years. Quaternary Research. Vol 43. 185-197.

Bonelli, S., S. Charbit, M. Kageyama. M.-N. Woillez, G. Ramstein, C. Dumas, and A. Quiquet.

2009. Investigating the evolution of major Northern Hemisphere ice sheets during the last

glacial-interglacial cycle. Climate of the Past. Vol 5. 329-345.

Born, A, M. Kageyama, and K. H. Nisancioglu. 2010. Warm Nordic Seas delayed glacial

inception in Scandinavia. Climate of the Past. Vol 6. 817-826.

Born, A., K.H. Nisancioglu, and B. Risebrobakken. 2011. Late Eemian warming in the Nordic

Seas as seen in proxy data and climate models. Paleoceanography. Vol 26. PA2207,

doi:10.1029/2010PA002027.

Bueler, Ed, Constantine Khroulev, Andy Aschwanden, Jeb Brown, and Nathan Schemonski.

2009. PISM, a Parallel Ice Sheet Model: User‟s Manual.

Bueler, Ed and Jed Brown. 2009. Shallow shelf approximation as a “sliding law” in a

thermomechanically coupled ice sheet model. Journal of Geophysical Research. Vol 114.

F03008. doi: 10.1029/2008JF001179.

Boulton, G.S., Payne, A., 1993. Simulation of the European ice sheet through the last glacial

cycle and prediction of future glaciation. SKB, Report 93-14, Swedishnuclear fuel and

waste management. Stockholm.

28

Bueler, Ed. 2009. Numerical Modeling of ice sheets and ice shelves. Lectures at Karthaus,

University of Alaska-Fairbanks.

Casty, Carol, Christoph C. Raible, Thomas F. Stocker, Heinz Wanner, and Jurg Luterbacher.

2007. A European pattern climatology 1766-2000. Clim Dyn 29. 791-805.

Charbit, S., C. Ritz, G. Philippon, V. Peyaud, M. Kageyama. 2007. Numerical reconstructions of

the Northern Hemisphere ice sheets through the last glacial-interglacial cycle. Climate of

the Past, European Geosciences Union. Vol 3. 15-37.

Colleoni, E., S. Masina, A. Cherchi, A. Navarra, C. Ritz, V. Peyaud, and B. Otto-Bliesner. 2014.

Modeling Northern Hemisphere ice-sheet distribution during MIS 5 and MIS 7 glacial

inceptions. Climate of the Past. Vol 10. 269-291.

Coyne, M. K., B. Jones, and D. Ford. 2007. Highstands during Marine Isotope Stage 5: evidence

from the Ironshore Formation of Grand Cayman, British West Indies. Quaternary Science

Reviews. Vol 26. 536-559.

Erhenberg, C. G., 1861. Uber die Tiefgrund-Verhaltnisse des Oceans am Eingange der

Davisstrasse une bei Island. K. preussiches Akademie von Wissenschaft der Berlin,

Monatsberichte. 275-31.

Fischer, R., S. Nowicki, M. Kelley, G. A. Schmidt. 2014. A system of conservative regridding

for ice-atmosphere coupling in a General Circulation Model (GCM). Geoscientific Model

Development 7. Vol 3. 883–907.

Fronval, Torben and Eystein Jansen. 1997. Eemian and early Weichselian (140-60ka)

paleoceanography and paleoclimate in the Nordic seas with comparison to Holocene

conditions. Paleoceanography. Vol 12. 443-462.

Fredin, Ola. 2002. Glacial inception and Quaternary mountain glaciations in Fennoscandia.

Quaternary International. Vol 95-96. 99-112.

Golledge, Nicholas et al, 2012. Last Glacial Maximum climate in New Zealand inferred from a

modeled Southern Alps icefield. Quaternary Science Reviews. Vol. 46. 30-45.

Golledge, N.R. et al. 2014. Antarctic contribution to meltwater pulse 1A from reduced Southern

Ocean overturning. Nature communications. DOI: 10.1038.

Gregory, J. M., O. J. H. Browne, A. J. Payne, J. K. Ridley, and I. C. Rutt. 2012. Modelling large-

scale ice-sheet – climate interactions following glacial inception. Climate of the Past. Vol

8. 1565-1580.

Gunzberger, Max. 2007. The science of ice sheets: The mathematical modeling and

computations simulations of ice flows. Isaac Newton Institute for Mathematical Sciences.

Seminar.

29

Haylock, M.R., N. Hofstra, A.M.G. Klein Tank, E.J. Klok, P.D. Jones, M. New. 2008. A

European daily high-resolution gridded dataset of surface temperature and precipitation.

J. Geophys. Res (Atmospheres). Vol 113. D20119, doi:10.1029/2008JD10201.

Helle, H., Snstegaard, E., Coope, G., Rye, N., 1981. Early Weichselian peat at Brumunddal,

southeastern Norway. Boreas. Vol 10. 369–379.

Helmens, Karin F. 2013. From The Last Interglacial-Glacial cycle (MIS 5-2) re-examined based

on long proxy records from central and northern Europe. Department of Physical

Geography and Quaternary Geology, Stockholm University. ISSN 1404-0344.

Holmlund, P. and J. Fastookt. 1995. A time dependent glaciological model of the Weichselian

Ice Sheet. Quaternary International. Vol. 27. 53-58.

Huybrechts, Philippe. 1994. The present evolution of the Greenland ice sheet: an assessment by

modelling. Global and Planetary Change. Vol 9. 39-51.

Huybrechts, Philippe. 2002. Sea-Level changes at the LGM from ice-dynamic reconstructions of

the Greenland and Antarctic ice sheet during the glacial cycles. Quaternary Science

Reviews. Vol 21. 203-231.

Jet Propulsion Laboratory. 2014. Ice Sheet System Model. University of California-Irvine and

NASA. Accessed from https://issm.jpl.nasa.gov/.

Johnsen, S. J., D. Dahl-Jensen, W. Dansgaard, and N. Gundestrup, 1995. Greenland paleotem-

peratures derived from GRIP bore hole temperature and ice core isotope profiles. Tellus.

Vol 47B. 624–629.

Jones, P.D., K.R. Briffa, T.P. Barnett, and S.F.B. Tett, 1998. High-resolution palaeoclimatic

records for the last millennium: interpretation, integration and comparison with General

Circulation Model control-run temperatures. The Holocene. Vol 8.4. 455-471.

Kageyama, M., S. Charbit, C. Ritz, M. Khodri, and G. Ramstein. 2004. Quantifying ice-sheet

feedbacks during the last glacial inception. Geophysical Research Letters. Vol 31.

L24203.

Kellogg, Thomas B. 1977. Paleoclimatology and Paleo-Oceanography of the Norwegian and

Greenland Seas: The last 450,000 years. Marine Micropaleontology. Vol 2. 235-249.

Kubatzki, Claudia, Martin Claussen, Reinhard Calov, and Andrey Ganopolski. 2006. Sensitivity

of the last glacial inception to initial and surface conditions. Climate Dynamics. Vol 27.

333-344.

Kukla, J. George, Michael L. Bender, Jacques-Louis de Beaulieu, Gerard Bond, Wallace S.

Broecker, Piet Cleveringa, Joyce E. Gavin, Timothy D. Herbert, John Imbrie, Jean

Jouzel, Lloyd D. Keigwin, Karen-Luise Knudsen, Jerry F. McManus, Josef Merkt,

Daniel R. Muhs, Helmut Müller, Richard Z. Poore, Stephen C. Porter, Guy Seret,

30

Nicholas J. Shackleton, Charles Turner, Polychronis C. Tzedakis, Isaac J. Winograd.

2002. Last Interglacial Climates. Quaternary Research. Vol 58. 2-13.

Lambeck, K. and J. Chappell. 2001. Sea level change through the Last Glacial Cycle. Science.

Vol 292. 679– 686.

Larsen, E., Sejrup, H. P., Janocko, J., Landvik, J. Y., Stalsberg, K. and Steinsund, P. I. 2000.

Recurrent interaction between the Norwegian Channel Ice stream and terrestrial-based ice

across southwest Norway. Boreas. Vol. 29. 185-203.

Lauritzen, Stein-Erik. 1995. High-Resolution Paleotemperature Proxy Record for the Last

Interglaciation Based on Norwegian Speleothems. Quaternary Research. Vol 43. 133-

146.

Lehman, S.J., et al., 1991.Initiation of the Fennoscandian ice-sheet retreat during the last

deglaciation. Nature. Vol. 349: 513-516.

Lekens, W.A.H., et al. 2009. Sedimentation history of the northern North Sea Margin during the

last 150 ka. Quaternary Science Reviews. doi:10.1016/j.quascirev.2008.11.010

Martin, M. A., Winkelmann, R., Haseloff, M., Albrecht, T., Bueler, E., Khroulev, C., and

A. Levermann. 2011. The Potsdam Parallel Ice Sheet Model (PISM-PIK) – Part 2:

Dynamic equilibrium simulation of the Antarctic ice sheet.The Cryosphere. Vol 5. 727-

740.

Mangerud, Jan. 1991. The Scandinavia Ice Sheet through the last interglacial/glacial cycle.

Reprint from: B. Frenzel (ed.) l99l: Klimageschichtliche Probleme der letzten 130 000

Jahre. 451 pp. G. Fischer, Stuttgart, New York. (Paleoklimaforschung Volume l).

Mangerud, Jan. 2004. Ice Sheet Limits in Norway and on the Norwegian continental shelf.

Quaternary Glaciations- Extent and Chronology. Developments in Quaternary Science.

DOI: 10.1016/S1571-0866(04)80078-2.

Mangerud, Jan et al. 2011. Glacial History of Norway. Developments of Quaternary Science.

Vol. 15. 279-298.

Mangerud, J. 1981b. The Early and Middle Weichselian in Norway: a review. Boreas. Vol 10.

381-393.

Mangerud, J., Sønstegaard, E., Sejrup, H.-P. & Haldorsen, S. 1981a. A continuous Eemian-Early

Weichselian sequence containing pollen and marine fossils at Fjøsanger, western

Norway. Boreas. Vol 10. 137-208.

Mangerud, J., Sønstegaard, E. & Sejrup, H.-P. 1979. Correlation of the Eemian (interglacial)

Stage and the deep-sea oxygen-isotope stratigraphy. Nature. Vol 277. 189-192.

31

Olsen, L., H. Sveian, and B. Bergstrøm. 2001. Rapid adjustments of the western part of the

Scandinavian ice sheet during the Mid- and Late Weichseliana new model. Norsk

Geologisk Tidsskrift Vol. 81. 93-118.

Olsen, L., H. Sveian, D. Ottesen, and L. Rise. 2013. Quaternary glacial, interglacial and

interstadial deposits of Norway and adjacent onshore and offshore areas. In Olsen, L.,

Fredin, O. and Olesen, O. (eds.) Quaternary Geology of Norway. Geological Survey of

Norway Special Publication. Vol 13. 79–144.

Ottesen, Dag et al.2005. The Vestfjorden-Trænadjupet paleo-ice stream drainage system, mid-

Norwegian continental shelf. Marine Geology. Vol. 218. 178-189.

Pedersen, Vivi Kathrine and David Lundbek Egholm. 2013. Glaciations in response to climate

variations preconditioned by evolving topography. Nature. Vol 493. 206-210.

Peterschmitt, J. Y., S. Denvil, and M. Morgan. 2013. The PMIP 3 Database: Latest News. In

EGU General Assembly Conference Abstracts. Vol 15. 9107.

Pollard, D. and PMIP participating groups. 2000. Comparison of ice-sheet surface mass budgets

from Paleoclimate Modeling Intercomparison Project (PMIP) simulations. Global and

Planetary Change. Vol 24. 79-106.

Raunholm, S., Larsen, E. & Sejrup, H.P. 2004. Weichselian interstadial sediments on Jæren (SW

Norway) – paleoenvironments and implications for ice sheet configuration. Norwegian

Journal of Geology. Vol. 84. 91-106.

Reeh, Niels. 1991. Parameterization of Melt Rate and Surface Temperature on the Greenland Ice

Sheet. Polarforschung. Vol 5913. 113-128.

Risebrobakken, Bjørg, Trond Dokken, Odd Helge Ottera, Eystein Jansen, Yongqi Gao, and

Helge Drange. 2007. Inception of the Northern European ice sheet due to constrating

ocean and insolation forcing. Quaternary Research. Vol 67. 128-135.

Sejrup, Hans Petter et al. 2003. Configuration, history and impact of the Norwegian Channel Ice

Stream. Boreas. DOI: 10.1080/03009480310001029.

Shackleton, Nicholas J., Maria Fernanda Sanchez-Goni, Delphine Paillerc, and Yves Lancelot.

2003. Marine Isotope Substage 5e and the Eemian Interglacial. Global and Planetary

Change. Vol 36. 151-155.

Shepherd, Anderw, et al., 2012. A reconciled Estimate of Ice-Sheet Mass Balance. Science. Vol.

338: 1183-1189.

Siegert, Martin J., Julian A. Dowdeswell, Morten Hald, and John-Inge Svendsen. 2001.

Modelling the Eurasian Ice Sheet through a full (Weichselian) glacial cycle. Global and

Planetary Change. Vol 31. 367-385.

32

Skoglund, R.Ø. and S. E. Lauritzen. 2010. Morphology and speleogenesis of Okshola (Fauske,

northern Norway): example of a multi-stage network cave in a glacial landscape.

Norwegian Journal of Geology. Vol. 90. 123-139.

Stroeven, Arjen P., Derek Fabel, and Shawn Marshall. 2002. Editorial: Quaternary International.

Vol. 95-96. 1-4.

Tucker, G. B. 1960. Precipitation over the North Atlantic Ocean. Meterological Office.

551.501.42 : 551.577.2 (261).

Waelbroeck C, L. Labeyrie L, E. Michel, JC Duplessy, JF McManu, K. Lambeck, E. Balbon, and

M. Labracherie. 2002. Sea-level and deep water temperature changes derived from

benthic foraminifera isotopic records. Quaternary Science Reviews.Vol 21. 295-305.

Winklemann, R. et al. 2011. The Potsdam Parallel Ice Sheet Model (PISM-PIK)-Part 1: Model

description. The Cryosphere. Vol. 5:715-726.

Yorgun, M. Soner and Richard B. Rood. 2014. An Object-Based Approach for Quantification of

GCM Biases of the Simulation of Orographic Precipitation. Part I: Idealized Simulations.

Journal of Climate. Vol 27. 9139–9154. doi: 10.1175/JCLI-D-14-00051.1.

Zhang, Huai, et al. 2011. Coupled Models and Parallel Simulations for Three-Dimensional Full-

Stokes Ice Sheet Modeling. Numer. Math. Theor. Meth. Appl. Vol 4. No.3: 359-381.

Zweck, Chris and Philippe Huybrechts. 2005. Modeling of the northern hemisphere ice sheets

during the last glacial cycle and glaciological sensitivity. Journal of Geophysical

Research. Vol 110. D07103

33

Copyright Permissions

PISM is jointly developed at the University of Alaska, Fairbanks (UAF) and the Potsdam

Institute for Climate Impact Research (PIK). UAF developers, who are in the Glaciers Group at

the GI, are supported by NASA's Modeling, Analysis, and Prediction and Cryospheric Sciences

Programs (grants NAG5-11371, NNX09AJ38C, NNX13AM16G, NNX13AK27G) and by the

Arctic Region Supercomputing Center. Accessed from http://www.pism-docs.org/wiki/doku.php.

34

TABLES

RUN CLIMATE TEMP

OFFSET(◦C)

PRECIP

OFFSET

(%)

RESOLUTION

(km)

Reference Run SMC -7 50 20

Sensitivity 1 SMC -9 50 20

Sensitivity 2 SMC -5 50 20

Sensitivity 3 SMC -3 50 20

Sensitivity 4 SMC -1 50 20

Sensitivity 5 SMC -7 75 20

Sensitivity 6 SMC -7 100 20

Sensitivity 7 SMC -7 25 20

High Resolution SMC -7 50 5

Modern Climate

(not shown) SMC 0 0 20

Experiment 2 PMIP3 0 100 20

Sensitivity 8

(not shown) PMIP3 -2 100 20

Sensitivity 9

(not shown) PMIP3 -1 100 20

Sensitivity 10

(not shown) PMIP3 +1 100 20

Sensitivity 11

(not shown) PMIP3 +2 100 20

Sensitivity 12

(not shown) PMIP3 0 50 20

Sensitivity 13

(not shown) PMIP3 0 150 20

Sensitivity 14

(not shown) PMIP3 0 200 20

Table 1. This is a complete list is of all of the experiments for this study. It includes the

experiment, climate scheme used, temperature offset, precipitation fraction, and model

resolution. Additionally, it includes the sensitivity experiment using the PMIP3 GCM which is

not shown in the figures below.

35

FIGURES

Figure 1: Topography of the model domain from ETOPO1. The points mark the cities of

Stavanger (SV), Trondheim (TR), and Andøya (AY). The Jotunheimen ice field (JT) is a modern

location of ice accumulation. The pink line represents present day sea level.

36

Figure 2: This is a schematic of the timing of ice coverage throughout Scandinavia. The records

are a combination of ice rafted debris off the continental shelf, sediment core in terrestrial and

marine environments and cave speleothems (Helle et al., 1981; Mangerud et al., 1981a;

Mangerud, 1991; Baumann et al., 1995; Lauritzen, 1995; Fronval et al, 1997; Larsen et al., 2000;

Olsen et al., 2001; Mangerud, 2004; Raunholm et al., 2004; Lekens et al., 2009; Skoglund et al.,

2010; Mangerud, 2011; Olsen et al., 2013; Helmens, 2013). From the known geological record,

FIS inception is located in the southern Scandinavian mountains. The tan represents land, the

pink line indicates present day sea level, and the blue indicates water.

37

Figure 3: A schematic showing the combination of the flow calculated by SIA and SSA within

the hybrid PISM model. In interior regions (a) SIA is large as internal deformation dominates. In

steeply sloping areas and an ice stream setting (b) both the flow by SIA and SSA is

approximated equally. SSA dominates in ice shelves (c).

b

a

c

& Steep Slopes

38

Figure 4: Temperature in the SMC (left column) and PMIP3 (middle column) are shown for

January, representing minimum temperatures, and July, representing summer temperatures. The

right column shows the difference (SMC-PMIP3) between these climate data. The black line

represents the outline of land at present sea level. Red indicates where the SMC was warmer than

the PMIP3 and blue indicates where the PMIP3 was colder than the SMC.

39

Figure 5: Monthly precipitation totals for SMC (left column) and PMIP3 (center column) are

shown for January and August. August is chosen to represent autumn precipitation maxima. The

right column shows the SMC minus the PMIP precipitation for each month. The black line

represents the outline of land at present sea level. Red indicates where the SMC was wetter than

the PMIP and blue indicates where the PMIP was wetter than the SMC.

40

Fig

ure

6:

Evolu

tion o

f ic

e th

icknes

s duri

ng t

he

refe

rence

run f

or

SM

C a

t -7

ᵒ C a

nd 5

0%

pre

cipit

atio

n.

Topogra

ph

y,

(a)

200 y

rs (

b)

400

yrs

(c)

600 y

rs (

d)

800 y

rs a

nd (

e) 1

000

yrs

. N

ote

the

south

ern l

oca

tion o

f n

ucl

eati

on.

The

pin

k l

ine

indic

ates

pre

sent

day s

ea l

evel

s.

41

Fig

ure

7:

The

sensi

tivit

y o

f 1000

-yea

r ic

e th

ickn

esse

s to

tem

per

ature

chan

ge.

The

refe

rence

run i

s pan

el d

and o

utl

ined

in t

he

bla

ck

box

. P

reci

pit

atio

n i

s hel

d f

ixed

at

50%

. T

hic

ker

ice

when

tem

per

ature

s ar

e at

cold

er o

ffse

ts y

et t

her

e is

lit

tle

chan

ge

in l

oca

tion o

f

accu

mula

tion.

The

pin

k l

ine

indic

ates

pre

sent

day s

ea l

evel

s.

42

Fig

ure

8:

Sen

siti

vit

y t

o 1

000

-yea

r ic

e th

icknes

ses

to p

reci

pit

atio

n. R

efer

ence

run i

s pan

el b

and o

utl

ined

in t

he

bla

ck b

ox

. T

emper

ature

is h

eld c

onst

ant

at a

-7

ᵒ C o

ffse

t. A

lim

itat

ion i

n p

reci

pit

atio

n a

ffec

ts t

he

loca

tion a

nd t

hic

knes

s in

ice

acc

um

ula

tion. S

ubst

anti

al

pre

cipit

atio

n l

eads

to t

he

dev

elopm

ent

of

a co

nti

nuous

ice

shee

t ac

ross

the

Sca

ndin

avia

n M

ounta

ins.

The

pin

k l

ine

indic

ates

pre

sent

day

sea

lev

els.

43

Fig

ure

9:

Evolu

tion o

f ic

e th

icknes

s fr

om

PM

IP3 p

alae

ocl

imat

e G

CM

. T

op

ogra

ph

y,

(a)

200 y

rs (

b)

400 y

rs (

c) 6

00 y

rs (

d)

80

0 y

rs a

nd

(e)

1000 y

rs. T

he

pin

k l

ine

indic

ates

pre

sent

day s

ea l

evel

s.

44

Figure 10: High Resolution simulation of ice thickness during the reference run for SMC at -7ᵒC

and 50% precipitation. The resolution of this run is on a 5km by 5km grid. Note the southern

location of nucleation particularly in the fjords near JT. The pink line indicates present day sea

levels. The resolution shows ice development in the same location as the coarser runs at 20km.