Dissertation Wekerle Oct2013

Dissertation Wekerle Oct2013
Dynamics of the Canadian Arctic Archipelago
throughflow: A numerical study with a finite
element sea ice and ocean model
Claudia Wekerle
Universität Bremen, 2013
Dynamics of the Canadian Arctic Archipelago
throughflow: A numerical study with a finite
element sea ice and ocean model
Von der Fakultät für Physik und Elektrotechnik
der Universität Bremen
zur Erlangung des akademischen Grades
einer Doktorin der Naturwissenschaften (Dr. rer. nat.)
genehmigte Dissertation
von
Claudia Wekerle
Gutachter: Prof. Dr. Peter Lemke
Zweitgutachter: Prof. Dr. Thomas Jung
Eingereicht am 4. Juni 2013
Tag des Promotionskolloquiums: 13. September 2013
Declaration of Authorship
I certify that the work presented here is, to the best of my knowledge and
belief, original and the result of my own investigations, except as acknowledged, and has not been submitted, either in part or whole, for a degree at
this or any other University.
Bremerhaven, June 2013
Abstract
The Canadian Arctic Archipelago (CAA) connects the Arctic Ocean and
Baffin Bay through narrow channels and is one of the key gateways where
freshwater leaves the Arctic. It has therefore the potential to affect the deep
convection in the northern North Atlantic.
Representing the CAA in traditional global models still poses a challenge
due to the small scale nature of the narrow passages. In this study we
apply a global, multi-resolution sea ice ocean model (the Finite Element
Sea ice Ocean Model, FESOM) with refinement in the CAA up to ∼5 km
while keeping a coarse resolution setup otherwise. With this model setup, a
hindcast simulation for the period 1968-2007 was performed.
The first goal of this thesis is to assess the model behavior in the CAA
region and in the Arctic Ocean. The model assessment revealed good agreement with sea ice conditions in the Arctic Ocean and with fluxes through the
main gates of the Arctic Ocean. During the period 1968-2007 the mean volume transports through Lancaster Sound and Nares Strait amount to 0.86
Sv (1 Sv = 106 m3 /s) and 0.91 Sv, respectively. The monthly mean volume transport through western Lancaster Sound is highly correlated with
the observational estimate (r=0.81). A comparison of simulated sectionally
averaged velocities in Nares Strait with observational estimates reveals good
agreement (r=0.57). The simulated mean CAA freshwater export rate is 123
mSv, slightly higher than the observational estimate (101±10 mSv). The
local refinement of ∼5 km allows to investigate the freshwater contribution
of individual narrow straits to the Parry Channel.
In the second part of the thesis, the mechanisms driving the interannual
variability of freshwater transports through the CAA are analyzed. The
i
interannual variability is determined by sea surface height (SSH) gradients
between the Arctic Ocean and northern Baffin Bay. The variability of fluxes
through Lancaster Sound and Nares Strait is mainly determined by that of
the SSH on the shelf along the Beaufort Sea coast and in the northeastern
Baffin Bay, respectively. Sea level variations north of the CAA are explained
by changes in the wind regimes (cyclonic vs. anticyclonic) associated to
release or accumulation of freshwater from the Beaufort Gyre, whereas sea
level in the northeastern Baffin Bay can be attributed to ocean-atmosphere
heat fluxes over the Labrador Sea. Both processes are linked with the North
Atlantic Oscillation type of atmospheric variability.
In the last part of the thesis, the effect of mesh resolution in the CAA
area is evaluated by performing experiments with and without highly resolved
archipelago (∼5 km vs. ∼24 km resolution). Increased resolution in the CAA
leads to higher freshwater transports through the CAA; at the same time
transports on the eastern side of Greenland are reduced. The ‘redirection’
of Arctic freshwater affects convection in the Labrador Sea and thus the
Atlantic meridional overturning circulation.
We conclude that multi-resolution models like FESOM are promising
tools for global climate modeling, as they are able to present small scale
processes in a global setup.
ii
Zusammenfassung
Das Kanadisch-Arktische Archipel (CAA) verbindet den Arktischen Ozean
mit der Baffin-Bucht und ist eines der Hauptausgänge durch welches süßere
Arktische Wassermassen in den salzhaltigeren Nordatlantik fließen. Das
Süßwasser arktischen Ursprungs hat das Potenzial, die Tiefenwasserbildung
im Nordatlantik zu beeinflussen.
Aufgrund der engen Wasserstraßen ist es nicht möglich, das kanadische
Archipel in traditionellen globalen Ozeanmodellen adequat zu repräsentieren.
In dieser Arbeit wird das Finite Elemente Meereis Ozean Modell (FESOM)
verwendet, um die Ozean- und Meereisdynamik des Archipels zu simulieren.
Der Vorteil von FESOM liegt in der Finite-Elemente Diskretisierung, welche
die Verwendung von unstrukturierten Gittern ermöglicht. Dadurch kann das
kanadische Archipel hoch aufgelöst werden (in dieser Modellversion bis ∼5
km), während andere Teile der Weltozeane gröber aufgelöst werden können.
Das erste Ziel dieser Arbeit ist die Validierung des Modells anhand von
Beobachungen im Kanadischen Archipel und im Arktischen Ozean. Die
Modellergebnisse zeigen gute Übereinstimmung mit den Meereisbedingungen
im Arktischen Ozean und mit gemessenen Volumen- und Süßwassertransporten durch die Ausfallstore des Arktischen Ozeans (Beringstraße, Davisstraße, Framstraße und Barentssee). Während der Zeitperiode 1968-2007
beträgt der simulierte mittlere Volumentransport durch den Lancastersund
0.86 Sv (1 Sv = 106 m3 /s), während der Transport durch die Naresstraße
0.91 Sv beträgt. Der monatliche simulierte Volumentransport durch den
westlichen Lancastersund ist mit Beobachtungsdaten signifikant korreliert
(r=0.81). Ein Vergleich von simulierten mittleren Geschwindigkeiten in der
Naresstraße mit beobachteten Geschwindigkeiten zeigt gute Übereinstimmung
iii
iv
(r=0.57). Der mittlere simulierte Süßwasserexport des kanadischen Archipels
beträgt 123 mSv, und ist etwas höher als der Beobachtungswert (101±10
mSv). Durch die lokale Gitterverfeinerung von ∼5 km lässt sich der Süßwassereinfluss durch die einzelnen engen Passagen in den Parry-Kanal analysieren.
Im zweiten Teil dieser Arbeit werden die Kontrollmechanismen der zwischenjährlichen Variabilität der Süßwassertransporte durch das CAA untersucht. Die zwischenjährliche Variabilität wird durch den Meeresspiegelgradienten zwischen dem Arktischen Ozean und der nördlichen Baffin-Bucht
bestimmt. Die Variabilität der Transporte durch den Lancastersund wird
hauptsächlich durch den Meeresspiegel entlang der Beaufortsee bestimmt,
während der Süßwassertransport durch die Naresstraße durch den Meeresspiegel in der nordöstlichen Baffin-Bucht bestimmt wird. Meeresspiegeländerungen nördlich des kanadischen Archipels können duch Änderungen
im Windregime (zyklonisch vs. antizyklonisch) erklärt werden, welche mit
einer Akkumulation oder Freigabe von Süßwasser im Beaufortwirbel verbunden sind. Meeresspiegeländerungen in der nordöstlichen Baffin-Bucht
dagegen können mit Wärmeflüßen zwischen Ozean und Atmosphäre in der
Labradorsee erklärt werden. Beide Prozesse sind an die atmosphärische Variabilität der Nordatlantischen Oszillation gekoppelt.
Im letzten Teil dieser Arbeit wird der Effekt der Gitterauflösung im
Bereich des Kanadischen Archipels untersucht, indem Experimente mit und
ohne hoch aufgelöstes Archipel durchgeführt wurden (∼5 km vs. ∼24 km
Auflösung). Hohe Auflösung im CAA führt zu höheren Süßwassertransporten
durch das CAA; gleichzeitig sinken die Transporte östlich von Grönland.
Durch diese ’Umleitung’ des Arktichen Süßwassers ändert sich die Tiefenwasserbildung in der Labradorsee, und damit auch die Atlantische meridionale Umwälzzirkulation.
Zusammenfassend lässt sich sagen, dass multi-skalen Modelle wie FESOM im Vergleich zu tradionellen Ozeanmodellen Vorteile bieten: So können
etwa in einer globalen Konfiguration klein-skalige Prozesse modelliert werden, während eine bessere Darstellung der klein-skaligen Prozesse zu einer
besseren Darstellung der groß-skaligen Zirkulation führt.
Contents
1 Introduction
1
1.1
The Canadian Arctic Archipelago - Background . . . . . . . .
1
1.2
The climate relevance of the CAA . . . . . . . . . . . . . . . .
5
1.3
Modeling the CAA ocean dynamics . . . . . . . . . . . . . . .
7
1.4
Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
9
1.5
Contents of this thesis . . . . . . . . . . . . . . . . . . . . . . 10
2 Methods
2.1
The ocean model . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.1.1
2.2
11
Parameterization of subgrid-scale processes . . . . . . . 14
The sea ice model . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2.1
Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2.2
Thermodynamics . . . . . . . . . . . . . . . . . . . . . 19
2.3
Coupling of the ocean and sea ice components . . . . . . . . . 24
2.4
Finite element discretization . . . . . . . . . . . . . . . . . . . 26
2.5
Mesh generation . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.6
Model setup and atmospheric forcing . . . . . . . . . . . . . . 33
2.7
Computation of transports . . . . . . . . . . . . . . . . . . . . 35
3 Model assessment
37
3.1
Arctic Ocean sea ice . . . . . . . . . . . . . . . . . . . . . . . 37
3.2
Freshwater budget of the Arctic Ocean . . . . . . . . . . . . . 40
3.3
3.2.1
Transports across the main gates of the Arctic Ocean . 40
3.2.2
Arctic Ocean freshwater content . . . . . . . . . . . . . 42
The simulated CAA ocean dynamics . . . . . . . . . . . . . . 44
3.3.1
3.4
Mean sea surface height and circulation in the CAA
region . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.2 Parry Channel . . . . . . . . . . . . . . . . . . . . . .
3.3.3 Nares Strait . . . . . . . . . . . . . . . . . . . . . . .
3.3.4 Cardigan Strait and Hell Gate . . . . . . . . . . . . .
3.3.5 Constituents of the freshwater transport variability .
3.3.6 Overall transports through the CAA . . . . . . . . .
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4 Interannual variability
4.1 The role of sea surface height . . . . . . . . . . . . .
4.2 Large scale atmospheric forcing . . . . . . . . . . . .
4.3 Arctic Ocean forcing on Lancaster Sound throughflow
4.4 Sea level in Baffin Bay and Nares Strait throughflow
4.5 Summary . . . . . . . . . . . . . . . . . . . . . . . .
5 Impact of mesh resolution in the CAA
5.1 Motivation . . . . . . . . . . . . . . . .
5.2 Freshwater transport through the CAA
5.3 Large scale circulation . . . . . . . . .
5.4 Summary and conclusion . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
44
46
52
55
55
56
57
.
.
.
.
.
59
59
63
65
67
69
.
.
.
.
71
71
73
77
81
6 Conclusions and Outlook
83
A Vertical resolution
87
B A list of symbols
89
Bibliography
105
Acknowledgment
107
Chapter 1
Introduction
1.1
The Canadian Arctic Archipelago - Background
The Arctic Ocean, enclosed by American and Eurasian landmasses and Greenland, has only limited connections to the world oceans. The stratification of
water masses in the Arctic Ocean is strongly influenced by the presence of
fresh surface waters originating from high river runoff through Siberian and
North American rivers, precipitation in form of snow and low evaporation
due to the isolating ice cover, and inflow of less saline Pacific Water through
Bering Strait (Serreze et al., 2006; Dickson et al., 2007). These relatively
fresh waters exit mainly through two passages: Fram Strait and the Canadian Arctic Archipelago (CAA).
While Fram Strait is a deep (sill depth 2,600 m) and broad passway, the
CAA is a shallow shelf system with narrow straits, characterized by a complex
bathymetry. Water and sea ice exiting the CAA have to pass through one
of the four exits: Lancaster Sound, Smith Sound or Jones Sound to Baffin
Bay, or Fury and Hecla Strait to Foxe Basin (see Figure 1.1). Parry Channel,
which opens out at its eastern end into Lancaster Sound, is the largest strait
of the CAA and has a width of 52.3 km at its narrowest point. Nares Strait,
the CAA’s easternmost strait which ends in Smith Sound in the south, has
a minimum width of 27.7 km.
1
2
Chapter 1. Introduction
Figure 1.1: Model bathymetry of the Canadian Arctic Archipelago.
Table 1.1: Abbreviations used in Figure 1.1.
AG
BaI
BI
BMC
BS
DI
EI
GL
FHS
HC
JS
KC
KWI
Amundsen Gulf
Banks Island
Baffin Island
Byam Martin Channel
Barrow Strait
Devon Island
Ellesmere Island
Greenland
Fury and Hecla Strait
Hell Gate/Cardigan Strait
Jones Sound
Kennedy Channel
King William Island
LS
MCC
MCS
MI
PeS
PRI
PS
PWI
PWS
QEI
RC
SS
VI
Lancaster Sound
McClintock Channel
McClure Strait
Melville Island
Penny Strait
Prince Regent Inlet
Peel Sound
Prince of Wales Island
Prince of Wales Strait
Queen Elizabeth Islands
Robeson Channel
Smith Sound
Victoria Island
1.1. The Canadian Arctic Archipelago - Background
3
Hell Gate and Cardigan Strait open out into Jones Sound and have a
minimum total width of 12.4 km (see Melling (2000) for more details). Fury
and Hecla Strait is 2 km wide. The CAA is a shelf basin with sills ranging
from 125 m in western Lancaster Sound and Hell Gate to 220 m in Nares
Strait (Melling, 2000). The shallow sills retain salty Arctic Ocean waters
originating from the North Atlantic and permit fresh surface water to flow
southwards.
Despite the potential importance of the CAA region in the climate system,
in situ measurements of the ocean hydrography and velocity in this region
are still rather sparse in time and space due to its remote location and harsh
weather conditions. The circulation in the Canadian Arctic Archipelago is
generally considered to be directed from the Arctic Ocean towards the south
east (Walker , 1977). Based on data from the cruise of HMCS Labrador
in 19541 from Baffin Bay through Parry Channel, Prince of Wales Strait
to the Beaufort Sea, Bailey (1957) was one of the first to give a detailed
description of the density structure of the water masses along the North
West Passage (NWP), a sea route between the Atlantic and Pacific oceans
passing through the Canadian archipelago. He described that Arctic waters
enter Parry Channel, the northern most part of the North West Passage,
mainly through McClure Strait, Byam Martin Channel and Penny Strait.
The flow through Parry Channel bifurcates at around 102◦ W, where parts
of the water flows southwards through McClintock Channel and then enters
Parry Channel again through Peel Sound (Wang et al., 2012a). Mooring
measurements in Peel Sound in April 1981 revealed a northward transport of
0.17 Sv, which equals one-third of the eastward transport in Barrow Strait
(Prinsenberg and Bennet, 1989).
Long term oceanographic measurements were carried out primarily in
Barrow Strait (western Lancaster Sound) for the period of 1998-2011 (Prinsenberg et al., 2009; Peterson et al., 2012): Volume and freshwater transport
exhibits a strong seasonal variability, with peaks in summer and small fluxes
in fall and early winter. Apart from the strong eastward current which dom1
The first large vessel to circumnavigate North America through the North West Passage from east to west in a single year was the Canadian government icebreaker Labrador.
4
Chapter 1. Introduction
inates the central and southern part of the Lancaster Sound cross section,
observations indicated a weak westward counter current on the northern side
appearing particularly in the summer months. Peterson et al. (2012) found
a significant correlation of transport through this strait with northeastward
wind anomalies in the Beaufort Sea, indicating that a cyclonic wind pattern
over the Beaufort Sea would favor high transports through Lancaster Sound.
Mooring measurements in Cardigan Strait and Hell Gate were performed
from 1998 on (Melling et al., 2008), estimating volume fluxes of 0.2 Sv and
0.1 Sv, respectively (for the 1998-2002 time period). There is an indication
that the strongest Arctic outflow occurs from January through September,
with weaker and even reversing flows in autumn and early winter.
Moorings deployed from 2003-2006 indicated that the flow through Nares
Strait is mainly directed from north to south, with a small counter current
accounting for around 5% of the southward flow (Münchow and Melling,
2008). For this time period, the mean volume transport was estimated as
0.57±0.09 Sv (excluding the top 30 m).
The main oceanic feature in eastern Lancaster Sound and western Baffin
Bay is the Baffin Island Current, which receives water mass contributions
from Smith, Jones and Lancaster Sounds, as well as from the West Greenland
Current, and continues southwards along the western side of Baffin Bay.
Fissel et al. (1982), based on observations in the summers of 1978 and 1979,
observed the Baffin Island Current to penetrate westwards into Lancaster
Sound for 35 to 75 km along its northern side, before crossing to the southern
side and exiting to the east. In the mouth of Lancaster Sound, they observed
surface speeds up to 75 cm/s, and an average width of the flow of 10 to 30
km.
The CAA is covered by sea ice most of the year, with some regions in the
south and east opening up in late summer. September is generally the month
with the lowest sea ice concentration. Freezing starts again in October, and
the sea ice is landfast for around half a year (Melling, 2002). Thick multiyear ice persists along the northern boundary of the CAA, being in fact
the thickest sea ice that can be found in the Arctic Ocean. The sea ice
flux across the Arctic Ocean - CAA boundary was estimated by Agnew et al.
1.2. The climate relevance of the CAA
5
(2008) by evaluating AMSR-E imagery. Their study revealed an export of 77
km3 of sea ice volume each year through Amundsen Gulf and McClure Strait
into the Arctic Ocean (averaged over the period 2002-2007), while Lancaster
Sound exported around 102 km3 of ice volume each year into Baffin Bay.
This study also showed that sea ice around the Queen Elizabeth Islands is
mostly landfast. The ice drift through the narrow channels of the CAA is
partly controlled by the presence of ice arches, which form especially in the
northern part of Nares Strait (Kwok , 2005). Polynyas, fixed regions of open
water that are isolated within thicker pack ice, are widely distributed over
the area of the archipelago (Hannah et al., 2009). Several mechanisms lead
to their formation: Tides, as they mix the warmer bottom water with the
colder surface waters slowing or eliminating the formation of ice, winds and
currents as they drive away consolidated ice, and ice bridges (especially in
Nares Strait) as they block ice upstream of the polynya area.
Climate models predict a reduction of the Arctic sea ice in the 21st century, possibly enabling a transit of the NWP for common open-water ships
and moderately ice-strengthened ships during the warmest months (Smith
and Stephenson, 2013). However, as concluded by Wilson et al. (2004), these
projections have to be regarded with care. Increased summer air temperatures would lead to a reduction of first year ice in the region of the Queen
Elizabeth Islands, allowing more multi year ice to reach the NWP through
Penny Strait and Byam Martin Channel, along with a southward shift of the
Beaufort Sea pack ice. The increase in drifting multi year ice in the NWP
and the southward shift of pack ice present significant hazards for navigation. Additionally, the sea ice conditions in the CAA are highly variable, and
occasionally there will be still summers of heavy ice conditions.
1.2
The climate relevance of the CAA
Freshwater exported from the Arctic Ocean may have a significant impact on
the Atlantic Meridional Overturning Circulation (AMOC). Deep water formation (DWF) in the Labrador and Greenland Seas is initiated by a strong
surface heat loss to the atmosphere, and in the Greenland Sea it is addi-
6
Chapter 1. Introduction
tionally activated by brine rejection from sea ice formation. This happens
particularly in the winter months. Vertical mixing leads to rising of warmer
waters and sinking of colder and more saline surface waters (Kuhlbrodt et al.,
2007). However, strong pulses of freshwater, as exported from the Arctic
Ocean, stabilize the density stratification in the convection areas and reduce
DWF. Deep water formed in the interior of the basins is exported by a cyclonic boundary current and advected southwards, and thus contributes to
the global overturning circulation (Straneo, 2006). It is plausible, therefore,
that an adequate representation of Arctic freshwater export in climate models
is crucial when it comes to predicting climate variability and change.
Figure 1.2: Ocean circulation downstream of the CAA (LS: Lancaster Sound,
JS: Jones Sound, SS: Smith Sound, BIC: Baffin Island Current, WGC: West
Greenland Current, EGC: East Greenland Current).
Several studies were carried out to analyze the impact of freshwater exported west and east of Greenland, and came to different conclusions. Goosse
et al. (1997) performed experiments with a very coarse resolution sea ice-
1.3. Modeling the CAA ocean dynamics
7
ocean model (3◦ × 3◦ horizontal resolution), and revealed that opening the
CAA passages in the model resulted in decrease of surface salinity and density in the Labrador Sea, along with a reduction of the North Atlantic Deep
Water (NADW) outflow by 5%. Wadley and Bigg (2002) confirmed this
finding: When opening the CAA passages in their model (which has a horizontal mesh resolution ranging from 100 to 800 km), the Atlantic overturning
streamfunction decreased from 25 to 20 Sv. In contrast, Komuro and Hasumi
(2005) and Jahn et al. (2009) came to opposite conclusions. By resolving the
CAA throughflow instead of closing the CAA, both model studies showed an
increase of the AMOC. The authors explained this by the redirection of freshwater pathways east and west of Greenland: If the CAA is closed, the Arctic
freshwater exists mainly through Fram Strait, which is then transported via
the East and West Greenland Current into the central Labrador Sea where
it diminishes the convection activity. In agreement with this, Myers (2005)
reported that enhanced freshwater export through Davis Strait has almost
no impact on the Labrador Sea Water (LSW) formation since only very little
of the freshwater enters the interior of the Labrador Sea (see Figure 1.2). All
above studies are based on models with a rather coarse resolution.
1.3
Modeling the CAA ocean dynamics
As discussed in section 1.1, hydrographic observations are sparse in the CAA.
Hence, numerical simulations can be used to reproduce the ocean circulation
and to help understand the ocean dynamics in the CAA region. However,
very high horizontal grid resolution is required to adequately resolve the narrow CAA straits. Because of high requirement for computational resources,
numerical simulations using traditional models with high enough resolution
were limited to relatively short integration time, or confined to regional model
domains. Table 1.2 presents an overview of model simulations of the CAA
throughflow, revealing large differences especially in the horizontal resolution.
Note that all models presented in Table 1.2 are based on structured meshes.
Jahn et al. (2012) showed that the state-of-the-art ocean models, even the
regional ones, represent a large spread in the simulated CAA transport.
8
Chapter 1. Introduction
Table 1.2: Models simulating the CAA throughflow.
model
1
UVic ESCM
2
ORCA025
3
OCCAM
4
NEMOv1.9/LIM2
5
NAME
6
NEMOv3.1/LIM2
model
domain
horizontal
resolution
vertical
levels
analysis
period
global
global
global
AO/NA∗
NP/NA/AO∗
CAA
1.8◦ × 0.9◦
10-27.75 km
8.3 km
22-50 km
1/12◦ (∼9 km)
6.5 - 9.5 km
32
46
66
46
45
46
1950-2007
1965-2002
1989-2006
1958-2001
1979-2004
climatological
forcing
1
Jahn et al. (2009), 2 Lique et al. (2009), 3 Aksenov et al. (2010), 4 Houssais and Herbaut
(2011), 5 McGeehan and Maslowski (2012), 6 Wang et al. (2012a), ∗ NA - North Atlantic,
AO - Arctic Ocean, NP - North Pacific
The geometrical complexity of the CAA suggests that exploring the role of
the CAA may benefit from using models formulated on unstructured meshes.
Their resolution can be conveniently refined locally with other parts of the
ocean left relatively coarse. Kliem and Greenberg (2003) applied a regional
diagnostic finite element ocean model to calculate the velocity field in the
CAA based on observational data for salinity and temperature. Sea ice models using the finite element method (coupled to a slab ocean) have been used
to simulate the sea ice conditions in the CAA region, and demonstrated the
advantage of unstructured meshes in resolving small scale dynamics (Lietaer
et al., 2008; Terwisscha van Scheltinga et al., 2010).
In this work numerical simulations are used to study the CAA ocean and
sea ice dynamics. The simulations are performed with the Finite Element
Sea-ice Ocean Model (FESOM, Danilov et al., 2004; Wang et al., 2008; Timmermann et al., 2009), an ocean general circulation model using unstructured
triangular surface meshes. Examination of its behavior on long time scales
showed that FESOM performs well in simulating the past ocean variability and in representing particular climate change scenarios (Sidorenko et al.,
2011; Wang et al., 2012b). However, these simulations used relatively coarse
resolution in the CAA. Instead of refinement, the depth and width of the
1.4. Objectives
9
CAA straits were adjusted as in most coarse-resolution models on structured
meshes. In this work the model explicitly resolves the major CAA straits.
1.4
Objectives
This thesis focuses on the following questions:
• The finite element model applied in this work allows for a local refinement in areas of interest, while other parts of the world oceans can be
kept relatively coarse. Taking advantage of this model feature, a new
model configuration of FESOM with the CAA highly resolved within
a global setup has been developed. How does this approach improve
the simulation of the CAA sea ice and ocean dynamics, and does it
also improve the simulation of the Arctic Ocean dynamics? In order to
investigate this, we compare the model to available observations, and
also to previous modeling studies.
• There is evidence that freshwater transport through the CAA has an
impact on the deep water formation sites in the northern North Atlantic. Keeping this in mind, understanding the interannual variability
of CAA freshwater transport is of high importance for our knowledge
of the global climate. Therefore, what is the interannual variability of
these transports? Which mechanisms drive the interannual variability?
What is the role of atmospheric forcing in this context?
• A new high resolution modeling approach has been developed, and it is
thus important to investigate the impact of mesh resolution. When resolving the CAA accurately, does this have an impact on the freshwater
transports through the CAA? Does it have an impact on the convection activity in the northern North Atlantic and thus on the large scale
ocean circulation? Hence, is it relevant for climate models to resolve
the CAA adequately?
10
Chapter 1. Introduction
1.5
Contents of this thesis
The structure of this thesis is as follows. Chapter 2 describes the numerical
structure of FESOM by presenting the governing equations for the ocean and
sea ice component, their coupling, their numerical discretization based on the
finite element method, and the model setup and configuration used in this
thesis. Chapter 3 assesses the model performance in the Arctic Ocean and
in the CAA and focuses on Arctic Ocean sea ice conditions, its freshwater
budget and the CAA sea ice and ocean dynamics. The interannual variability of freshwater transport through the CAA and its driving mechanism is
presented in Chapter 42 . The sensitivity to mesh resolution and impacts on
the Atlantic Ocean circulation are discussed in Chapter 53 . The last chapter
summarizes and discusses the main findings of this thesis.
2
The major part of Chapters 3 and 4 is adopted for the following publication: Wekerle,
C., Q. Wang, S. Danilov, T. Jung and J. Schröter, Freshwater transport through the
Canadian Arctic Archipelago in a multi-resolution global model: Model assessment and
the driving mechanism of interannual variability, submitted manuscript to ”Journal of
Geophysical Research”, April 2013.
3
The major part of Chapter 5 is adopted for the following publication: Wekerle, C.,
Q. Wang, S. Danilov, T. Jung, J. Schröter, and P.G. Myers, The climate relevance of
the Canadian Arctic Archipelago: A multi-resolution study, submitted manuscript to the
journal ”Geophysical Research Letters”, May 2013.
Chapter 2
Methods
In this work a global version of the Finite-Element-Sea ice-Ocean-Model (FESOM) is applied. FESOM was developed in the Climate Dynamics section
of the Alfred Wegener Institute. It consists of a finite element ocean model
Danilov et al. (2004); Wang et al. (2008), coupled to a sea ice model (Timmermann et al., 2009). The advantage of this model is the following: The
finite element discretization of the governing equations allows us to employ
unstructured meshes, and therefore to locally refine the mesh in areas of
interest and keep it coarse in other parts of the global oceans. This multiresolution approach is very suitable for studying the CAA throughflow because the local fine resolution enables to resolve the complex bathymetry and
narrow straits, and the global configuration avoids the open boundary and
allows for studying the impact of the CAA throughflow on the global ocean
circulation.
2.1
The ocean model
The ocean component solves the standard set of hydrostatic primitive equations with the Boussinesq approximation. The system of governing equations
is split into two subproblems, the dynamical part and the thermodynamical
part, which are solved separately.
The dynamical part includes the momentum equations in three dimen11
12
Chapter 2. Methods
sions, the vertically integrated continuity equation and the hydrostatic pressure equation which are solved for horizontal velocity, sea surface height and
pressure:
1
∂t u + v · ∇3 u + f (k × u) + g∇η + ∇p = ∇ · Ah ∇u + ∂z Av ∂z u,
ρ
Z z=η 0
u dz = 0,
∂t η + ∇ ·
(2.1)
(2.2)
z=−H
∂z p = −g ρ,
(2.3)
where v ≡ (u, w) ≡ (u, v, v) is the velocity vector, f = f (θ) is the Coriolis
parameter dependent on the latitude θ, k is the vertical unit vector, g is
the gravitational acceleration, ρ0 and ρ are the mean sea water density and
R0
the deviation from it, respectively, η is the sea surface height, p = z g ρ dz
is the hydrostatic pressure anomaly obtained by integrating the hydrostatic
equation in the vertical from z = 0, Ah and Av are the lateral and vertical momentum diffusion coefficients, respectively. The upper limit in the
integration in equation 2.2 is set to zero, which implies a linear free-surface
approximation. ∇ and ∇3 stand for the 2-dimensional and 3-dimensional
gradient and divergence operators, respectively. In equation 2.2, the transport of freshwater into and out of the ocean through the surface is omitted
(transport into the ocean occurs through precipitation, river runoff and sea
ice melting, while transport out of the ocean occurs through evaporation and
sea ice freezing).
In the global configuration, the above equations are solved in the domain
S
Ω, which is limited by three different types of boundaries ∂Ω = 3i=1 Γi .
Γ1 : {z = 0} stands for the ocean surface, Γ2 : {z = −H(θ, λ)} stands for
the ocean bottom and Γ3 stands for the lateral vertical rigid walls. For the
dynamical part of the model, the boundary conditions at the surface and the
bottom are:
Av ∂z u = τ,
p = 0 on Γ1 ,
Av ∂z u + Ah ∇H · ∇u = −Cd u|u| on Γ2 ,
(2.4)
(2.5)
2.1. The ocean model
13
where τ is the wind stress and Cd is the bottom drag coefficient.
On the vertical rigid walls, no normal flow is allowed. In this configuration of FESOM, no-slip boundary conditions are applied, meaning that the
tangential velocity along the vertical wall is zero:
u · n = 0 and u = 0 on Γ3 ,
(2.6)
with n being the unit normal vector to the rigid wall. The vertical velocity
w is diagnosed from the continuity equation:
∂z w = −∇ · u,
(2.7)
which has the following kinematic boundary conditions at the surface and at
the bottom:
w = ∂t η on Γ1 ,
(2.8)
w = −∇H · u on Γ2 .
(2.9)
In the thermodynamical part of the ocean model we solve the tracer equations for potential temperature T and salinity S:
∂t C + v · ∇3 C = ∇ · Kh ∇C + ∂z Kv ∂z C,
(2.10)
with C standing for T or S and Kh and Kv standing for the lateral and vertical
diffusivity for the particular tracer, respectively. The following boundary
conditions have to be fulfilled for the tracer equations:
Kv ∂z C = −q, on Γ1
(Kh ∇C, Kv ∂z C) · n3 = 0 on Γ2 ∪ Γ3 ,
(2.11)
(2.12)
where q stands for the surface flux for T and S.
The density ρ is diagnosed via the equation of state according to Jackett and
Mcdougall (1995):
ρ = ρ(T, S, p).
(2.13)
14
Chapter 2. Methods
In the initial state, horizontal velocity is set to zero. The initial conditions
for salinity and potential temperature are taken from a climatological data
set, which will be discussed later in Section 2.6.
2.1.1
Parameterization of subgrid-scale processes
Mixing from mesoscale eddies in the ocean interior mainly occurs along directions tangent to the local density surface. In this model configuration we
employ z-levels, which do not correspond to the isopycnal surfaces. Therefore, the diffusion tensor is rotated to align the diffusive fluxes along the
neutral directions according to Redi (1982). As the neutral density slope is
generally small in the ocean interior, the small slope approximation is applied
in the Redi diffusion tensor. Gent and McWilliams (1990) also suggested a
scheme to parameterize mesoscale eddies by introducing an advective bolus
flux. Following these schemes, the right hand side of Equation 2.10 can be
rewritten as
∇ · (Kredi + KGM )∇3 C,
(2.14)
with Kredi and KGM being symmetric and antisymmetric tensors, respectively. The left term stands for the Redi diffusive flux, while the right term
stands for the Gent/McWilliams skew flux (see Griffies, 1998). The isopycnal diffusivity and the skew diffusivity are set to the same value of 0.006
ms−1 ∆, where ∆ is the square root of the surface triangle area. However,
the tensor Kredi + KGM is not bounded as the isoneutral slope increases (e.g.
in the mixed layer), which might lead to numerical instabilities. To avoid
these instabilities a tapering scheme is applied with a critical neutral slope
of 0.004.
Vertical mixing is strong in the surface layers of the ocean and relatively
weak below the thermocline; its parameterization in the model is provided
by the Pacanowski and Philander scheme (Pacanowski and Philander , 1981)
which ensures that vertical mixing increases for a weaker stratification. The
vertical momentum diffusion coefficient, Av , and the vertical tracer diffusion
coefficient, Kv , are expressed as:
2.2. The sea ice model
15
Av =
(
Kv =
(
v0
(1+α·Ri )n
+ Av0
v0 + Av0
v0
(1+α·Ri )n
if Ri > 0,
if Ri ≤ 0,
+ Av0
Av + Kv0
if Ri > 0,
if Ri ≤ 0.
(2.15)
(2.16)
where v0 , α and n are adjustable parameters set to 10−2 m2 s−1 , 5 and 2,
respectively. Av0 and Kv0 stand for the background vertical diffusion for
momentum and tracers, respectively, and are set to 10−4 m2 s−1 and 10−5
m2 s−1 . The above formulas depend on the Richardson number Ri , which
expresses the ratio of potential to kinetic energy:
Ri =
N2
,
(∂z u)2 + (∂z v)2
(2.17)
with N being the buoyancy frequency. The mixing scheme of Timmermann
et al. (2002) is introduced (a diffusivity of 0.01 m2 s−1 is applied over a depth
defined by the Monin-Obukhov length when it is positive) in order to avoid
unrealistically shallow mixed layers in summer.
To reduce dissipation of momentum, the harmonic horizontal viscosity of
2.1 is replaced by a biharmonic one with B∆3 , where B =0.027 ms−1 .
2.2
The sea ice model
Sea ice is an important part of the climate system. First, it affects the heat
and mass transfer between the ocean and the atmosphere due to its insulating
properties. Second, it influences the radiation balance due to its high albedo,
resulting in a high reflection of the incoming solar radiation. Third, the brine
release during sea ice formation affects the convection activity in deep water
formation areas. Therefore, the coupling of an ocean model with a sea ice
model is of high importance if we want to accurately simulate the ocean
dynamics.
Sea ice in the model is assumed to be a slab of ice on which snow can
16
Chapter 2. Methods
accumulate; on every element of the mesh it is described by the quantities ice
thickness, snow thickness, ice concentration and ice velocity. Ice thickness
and ice concentration can change due to freezing and melting (thermodynamic processes) and due to deformation, while the ice drift is affected by
wind and ocean drag, by the Coriolis force, by the sea surface height gradient and by internal forces within the ice (dynamic processes). In this section
the governing equations of the sea ice component of FESOM will be presented, i.e. the continuity equations and the momentum equation, along
with a description of the ice thermodynamics (which follow mainly the work
by Parkinson and Washington (1979) and Semtner (1976)) and the ice dynamics (which follow the approaches by Hunke and Dukowicz (2001)). A
detailed description of the sea ice model can be found in Timmermann et al.
(2009).
Sea ice consists of individual ice floats which drift freely in regions of low
ice concentration or is packed closely in regions of high ice concentration.
In order to model sea ice, we consider it as a 2-dimensional continuum.
Prognostic variables are:
• effective ice thickness h, defined as the ice volume per area averaged
over the ice covered and ice-free part of the element
• effective snow thickness hs defined in the same way as h
• ice concentration A, a dimensionless quantity ranging from zero to one
specifying the fraction of the ice-covered area of an element
• ice (and snow) drift velocity uice
The evolution in time of the quantities h, hs and A is described by the
continuity equations
∂t h + ∇ · (uice h) = Sh ,
(2.18)
∂t hs + ∇ · (uice hs ) = Ss n,
(2.19)
∂t A + ∇ · (uice A) = SA ,
0 ≤ A ≤ 1,
(2.20)
2.2. The sea ice model
17
where the first and second terms are the derivative in time of the prognostic
variable and advection terms, respectively. The terms on the right hand side
stand for sources and sinks including freezing and melting (thermodynamic
processes), snow fall and snow to ice transformation.
The ice velocity uice = (uice , vice ) is computed from the 2-dimensional
momentum equation:
m (∂t + f k×) uice = A (τai + τoi ) + F − mg∇η.
(2.21)
It includes the effect of the Coriolis force, the effect of stress due to wind
(τai ) and due to ocean velocity (τoi ), the internal forces (F) and the gravity
force on a tilted ocean surface. m = ρi h + ρsn hs denotes the ice mass per
area, where ρi = 910 kg/m3 stands for the density of ice. The sea surface
height η and the ocean velocity u, which is needed for the computation of
τoi , are taken from the ocean model as part of the coupling scheme.
The process to solve the above equations is the following: First, we solve
the momentum equation for ice velocity. Second, the continuity equation is
solved by applying a splitting technique: In the advection step preliminary
values for sea ice and snow thickness and ice concentration are computed by
setting the sources and sinks in equations 2.18-2.20 to zero. Then these variables are updated by accounting for the sources and sinks. In the following,
the dynamic and thermodynamic part of the sea ice model will be further
described.
2.2.1
Dynamics
The forces appearing in equation 2.21 acting on the sea ice will be further
described in this section.
Rheology The internal forces F within the sea ice are expressed as the divergence of the stress tensor σ: F = ∇·σ. According to Hunke and Dukowicz
(1997) we consider the sea ice as a nonlinear elastic viscous compressible fluid.
The EVP model consists of the following time-dependent equations, which
18
Chapter 2. Methods
have to be solved for σ1 := σ11 + σ22 , σ2 := σ11 − σ22 and σ12 :
P
1 ∂σ1 σ1
+
+
= DD ,
E ∂t
2ζ 2ζ
1 ∂σ2 σ2
+
= DT ,
E ∂t
2η
1 ∂σ12 σ21
1
+
= DS .
E ∂t
2η
2
(2.22)
(2.23)
(2.24)
DD , DT and DS are the divergence, the horizontal tension and the shearing
strain rates, respectively. They depend on the components of the strain rate
tensor, ǫ̇, and are defined as
DT = ǫ̇11 + ǫ̇22 ,
(2.25)
DD = ǫ̇11 − ǫ̇22 ,
(2.26)
DS = 2ǫ̇12 .
(2.27)
The shear viscosity η, the bulk viscosity ζ and ∆ are calculated from the
formulas
P
,
2∆
P
ζ= 2 ,
2e ∆
η=
2
∆ = [DD
+
(2.28)
(2.29)
1
1
(DT2 + DS2 )] 2 ,
2
e
(2.30)
where e is an empirical parameter set to 2. The ice strength P depends on
the ice concentration A and the ice thickness h:
P = P ∗ h exp(−C(1 − A)),
(2.31)
with P ∗ and C being empirical parameters set to 27500 Nm−2 and 20, respectively. This formulation ensures that the strength of the ice is high when the
ice is thick and when it has a high compactness. An implicit time stepping
scheme (backward Euler) is used to solve equations 2.22-2.24.
2.2. The sea ice model
19
Shear stresses The wind stress term in equation 2.21 depends on the
relative motion between the wind velocity at 10 m height, u10 , which is
prescribed by atmospheric forcing (see Section 2.6), and the ice velocity:
τai = Cd,ai ρa u10 |u10 |,
(2.32)
where Cd,ai stands for the ice-atmosphere drag coefficient and ρa = 1.3 kg/m3
stands for the air density. The ice velocity is small compared to the wind
velocity, and is thus ignored in the above equation.
The exchange of momentum between the ocean surface and the sea ice is
expressed by
τio = Cd,io ρ0 (uice − us )|uice − us |,
(2.33)
where Cd,io stands for the ice-ocean drag coefficient. The ocean surface velocity us is taken from the ocean component as part of the coupling scheme.
2.2.2
Thermodynamics
The ice thermodynamics, i.e. the melting and freezing of sea ice, are based
on energy balances for the ice-ocean and ice-atmosphere interface following
the work by Parkinson and Washington (1979). The heat conductivity of
sea ice follows the scheme described by Semtner (1976).
Surface energy budget The heat fluxes at the boundary between atmosphere and ocean, and in the ice-covered grid cell between atmosphere and
sea ice, are:
Qai,ao = Q↓SW + Q↑SW + Q↓LW + Q↑LW + Qs + Ql ,
(2.34)
where QSW and QLW stand for the short wave and long wave radiative fluxes,
respectively. The symbols ↓ and ↑ stand for downwelling and upwelling radiative fluxes, respectively. Qs and Ql denote the turbulent fluxes of sensible and
latent heat, respectively. In the model, the downwelling shortwave and longwave radiative fluxes are prescribed by the atmospheric forcing, described in
Section 2.6. The upwelling shortwave radiation is computed from the down-
20
Chapter 2. Methods
welling shortwave radiation, scaled with the albedo α, the reflecting power
of a surface:
(2.35)
Q↑SW = α · Q↓SW .
The albedo depends on the surface type and on the surface temperature.
Table 2.1 presents the albedo coefficients used in this model configuration.
Table 2.1: Values for the albedo α used in
FESOM.
albedo
surface
frozen snow
melting snow
frozen ice without snow layer
melting ice
open water
0.81
0.77
0.7
0.68
0.07
The upwelling longwave radiation is computed based on the Stefan-Boltzmann-Law, which describes the power radiated from a black body in terms
of its temperature:
(2.36)
Q↑LW = ǫσTs4 ,
with the emissivity ǫ set to 0.97 and being the same for ocean or sea ice
surface, and the Stefan-Boltzmann-coefficient being σ =5.67 10−8 Wm−2 K−4 .
Ts is the ocean or ice/snow surface temperature.
The turbulent fluxes of sensible and latent heat are calculated from the
bulk aerodynamic formulas according to Parkinson and Washington (1979):
Qs = ρa cp Ce,ao |u10 |(Ta − Ts ),
(2.37)
Ql = ρa L Ch,ao |u10 |(qa − qs ),
(2.38)
where cp stands for the specific heat capacity of water. L stands for the
latent heat of vaporization, and in case sea ice exists, it stands for the latent
heat of sublimation. L is set to 2.501·106 J kg−1 for evaporation, and to
2.835·106 J kg−1 for sublimation. The coefficients Ce,ao and Ch,ao are heat
2.2. The sea ice model
21
transfer coefficients for sensible and latent heat, respectively. Ta and qa are
air temperature and specific humidity at 10 m height, and are prescribed by
the atmospheric forcing. The air at the surface is assumed to be saturated,
and the saturated specific humidity at the surface, qs , is computed according
to Large and Yeager (2004) and depends on the the surface temperature Ts :
q1
exp
qs =
ρa
q2
Ts
,
(2.39)
with q1 =6.89 · 640380 kg/m3 and q2 =-5107.4 K.
Growth rates for the ice-free area Every element contains a part which
is covered by open water, specified by the ice concentration A. The net
heat flux from the atmosphere into the ocean is given by equation 2.34. In
equations 2.36 and 2.37, the surface temperature Ts is set to the temperature
of the ocean surface layer. In equation 2.35, the albedo is set to the value for
open water. If Qao is negative and the ocean surface temperature is above
the freezing temperature Tf 1 , the heat flux is used to cool down the ocean
surface layer; while a positive value for Qao would result in a warming of the
ocean surface layer. If the ocean surface temperature falls below the freezing
temperature Tf , a portion of the water freezes; the growth rate of sea ice
measured in meter ice per second is then:
(∂t h)ow = −
Qao
.
ρi L i
(2.40)
Growth rates for the snow- and ice-covered area When computing
the thermodynamic growth of sea ice in areas which are already ice covered,
contributions of ice growth at the ice-atmosphere interface and at the iceocean interface have to be summed up.
Considering the ice-atmosphere interface, we have to add to the surface
heat budget (equation 2.34) the conductive heat flux Qc . The conductive
heat flux is assumed to be vertically constant and to depend linearly on the
1
The freezing temperature Tf depends according to Gill
p (1982) on the the salinity of
the ocean surface layer: Tf = −0.0575Ss + 1.7105 · 10−3 Ss3 − 2.155 · 10−4 Ss2 .
22
Chapter 2. Methods
temperature difference between the bottom and the surface of the ice/snow.
It accounts for both, the snow layer and the ice layer:
Qc = κi
Tb − Ts
,
h∗i
with h∗i = hi + hsn
κi
κsn
(2.41)
with κi = 2.1656 W m−1 K−1 and κsn = 0.31 W m−1 K−1 being the heat
conductivity for ice and snow, respectively. Tb is the temperature at the
bottom of the ice and is set to the freezing temperature Tf . The variables
hi = h/A and hsn = hs /A denote the actual thickness of the ice and snow,
respectively. The only unknown in the above equation is the temperature at
the surface of the ice (or snow), Ts , which has to be determined from the the
surface energy budget:
Q↓SW + Q↑SW + Q↓LW + Q↑LW + Qs + Ql + Qc = 0.
(2.42)
Ts appears in terms 4, 5, 6, and 7. The Newton-Rhapson-Scheme is applied
to solve this equation iteratively for Ts . If Ts exceeds the freezing point of
freshwater, 0◦ C, Ts is set to 0◦ C. After Ts is solved, it is substituted into
equation 2.34 to calculate the net heat flux, which will be used to melt ice
and snow or form ice. The change of snow and ice thickness is calculated as:
Qai
,
ρsn Li
Qai
(∂t h)a = −
− ∂ t hs ,
ρi L i
∂ t hs = −
(2.43)
(2.44)
where ρsn = 290 kg/m3 is the density of snow. It is assumed that first the
total snow layer melts, followed by the ice layer.
On the bottom side of the ice, we assume that melting as well as freezing
can occur. The rate of melting or freezing is calculated from the heat flux at
the ocean-ice interface Qoi :
(∂t h)o = −
Qoi
.
ρi L i
(2.45)
The heat flux between ocean and ice is part of the coupling scheme and is
2.2. The sea ice model
23
described in Section 2.3.
The ice growth rate depends strongly on the ice thickness. In reality, sea
ice ranges from thin to thick ice floes in the area covered by one element in the
model. It is thus not reasonable to compute the ice growth rate depending
on the effective ice thickness, which presents a mean value over the element.
In order to incorporate the dependence of ice growth on ice thickness, seven
ice thickness classes ranging from 0 to 2 · h are introduced according to the
approach by Hibler (1984), instead of taking the effective thickness. For
every ice class, the ice growth is computed according to equations 2.43, 2.44
and 2.45 and weighted with 71 .
Mean growth rates The above calculated growth rates at the ice surface,
the ice bottom and in the open ocean are weighted according to the sea ice
concentration. The combined thermodynamic growth rate is then:
∂t h = A · ((∂t h)a + (∂t h)o ) + (1 − A) · (∂t h)ow .
(2.46)
Updating ice concentration After computing the thermodynamic growth
rates, the ice concentration A is updated according to Hibler (1979). In case
of freezing, SA , the term on the right hand side of equation 2.20, is proportional to the area of open water (1 − A):
SA =
1−A
{(∂t h)ow + (∂t h)o } ,
h0
(2.47)
with h0 being an empirical parameter, the lead closing parameter, which is
set to 0.5 m. In case of ice melting, SA is determined by the formula
SA =
A
{(∂t h)ow + (∂t h)a + (∂t h)o } ,
2h
The updated ice concentration A then equals SA · ∆t.
(2.48)
24
Chapter 2. Methods
2.3
Coupling of the ocean and sea ice components
The ocean and sea ice models are coupled via heat exchange, salt flux due
to melting and freezing of ice and snow, and momentum exchange.
The heat flux between ocean and sea ice is given by the formula
Qoi = ρw cp,w Ch,io u∗ (Ts − Tf ),
(2.49)
where Ts stands for the ocean surface temperature, cp,w stands for the specific
heat capacity of seawater at constant pressure, Ch,io stands for the transfer
coefficient of the exchange of sensible heat between the ocean and the bottom
side of the ice. The friction velocity u∗ is computed as
u∗ =
p
Cd,io |us − uice |,
(2.50)
where Cd,io denotes the oceanic drag coefficient and us denotes the ocean
velocity of the surface layer.
The salinity of the upper ocean layer is significantly affected by freezing
and melting of sea ice. Ice growth is associated with discharge of salt into the
ocean layer beneath the ice, which increases the density there. In contrast,
ice melting leads to an input of freshwater into the ocean surface layer. In the
model, melting and freezing result in a virtual salt flux, which is described
by
ρi
ρsn
salt
Fice
= (Ss − Sice ) (∂t h)th + Ss
(∂t hs )th .
(2.51)
ρw
ρw
The first term describes the salt flux due to thermodynamic changes in the
ice thickness, while the second term considers the snow thickness. The ice
salinity, Sice , is set to a constant value of 6 psu in the model, with the salinity
of snow assumed to be zero.
The difference of precipitation and evaporation (P − E) also provides
a source of freshwater. If the air temperature is above the freezing point
(Ta ≥ 0◦ C), the precipitation runs completely into the ocean. For Ta < 0◦ C,
2.3. Coupling of the ocean and sea ice components
25
the precipitation on the ice-covered area of the grid cell is accumulated as
snow. Therefore, the salinity flux due to P − E is given by
FPsalt
−E = Ss ·
(
P −E
(1 − A) · (P − E)
if Ta ≥ 0◦ C
if Ta < 0◦ C
(2.52)
The freshwater input due to river runoff is also considered as a virtual
salt
salt flux and denoted by Frunof
f.
The total virtual salt flux into the surface layer of the ocean model is
salt
salt
F salt = Fice
+ FPsalt
−E + Frunof f .
(2.53)
The exchange of momentum between ice and ocean is computed using
the formula
τio = Cd,io ρo (uice − us )|uice − us |,
(2.54)
where Cd,io stands for the ice-ocean drag coefficient. The stress between the
open ocean surface and the atmosphere is
τao = ρa Cd,ao u10 |u10 |,
(2.55)
with the atmosphere-ocean drag coefficient Cd,ao . The total ocean surface
stress is the weighted average of the two terms:
τo = A · τio + (1 − A) · τao .
(2.56)
Drag and transfer coefficients In this model configuration, the oceanatmosphere drag coefficient Cd,ao and the transfer coefficients for the exchange of sensible and latent heat Ce,io and Ch,io , respectively, are functions
of wind speed and the atmospheric stability ζ according to the formulas by
26
Chapter 2. Methods
Large and Yeager (2004):
2.7 · 10−3
u10
+ 0.142
,
u10
13090
p
= Cd,ao ,
(
p
18 Cd,ao
ζ>0
p
=
ζ ≤ 0.
32.7 Cd,ao
Cd,ao =
(2.57)
Ce,ao
(2.58)
Ch,ao
(2.59)
The exchange coefficients for the atmosphere - ice interaction are set to constant values:
2.4
Ch,ai = Ce,ai = 1.75 · 10−3 ,
(2.60)
Cd,ai = 1.32 · 10−3 .
(2.61)
Finite element discretization
The equations described above are solved by applying the finite element
method (FEM). We will now focus on the ocean component; the application
of the FEM to the sea ice equations works analogously. The finite element
discretization is explained in more detail in Danilov et al. (2004); Wang et al.
(2008); Timmermann et al. (2009).
First, a variational formulation of the governing equation is introduced.
We obtain this formulation by multiplying equation 2.1 with a test function
ũ and integrating it over the domain Ω. By applying integration by parts
to the right hand side term, only terms with derivatives up to first order are
left. An appropriate function space V for the variational formulation is the
Sobolev space H 1 (Ω) of square integrable functions with square integrable
weak derivatives of first order. If we apply boundary conditions 2.4-2.6 and
assume ũ = 0 on Γ3 , the problem defined by equation 2.1 can be reformulated
2.4. Finite element discretization
27
as: Find u ∈ H 1 (Ω), such that for each ũ ∈ H 1 (Ω) applies:
1
[∂t u + v · ∇3 u + f (k × u) + g∇η + ∇p] · ũ dΩ =
ρ
ZΩ
Z
Z 0
Cd u|u|ũ dΓ2 . (2.62)
τ ũ dΓ1 −
− [Ah ∇u · ∇ũ + Av ∂z u · ∂z ũ] dΩ +
Z
Ω
Γ1
Γ2
Similarly, the variational formulation of equation 2.2 is: Find η ∈ H 1 (Γ1 ),
such that for each η̃ ∈ H 1 (Γ1 ) applies:
Z
∂t η η̃ dΓ1 −
Γ1
Z
u · ∇η̃ dΩ = 0.
(2.63)
Ω
Variables u and η fulfilling equations 2.62 and 2.63 are called weak solutions
of the problem given by equations 2.1 and 2.2 and boundary conditions 2.4,
2.5 and 2.6. The formulations 2.62 and 2.63 have to be discretized in order to solve the problem numerically. The functional spaces V = H 1 (Ω)
and W = H 1 (Γ1 ) are replaced by finite-dimensional subspaces Vh ⊂ V and
Wh ⊂ W , respectively. The surface and 3D volume are discretized using
triangles and tetrahedra, respectively (see Figure 2.1). More details on mesh
generation will be given in Section 2.5. The prognostic variables u and η
Figure 2.1: Left: The surface triangulation and prisms below each triangle.
Right: Decomposition of a prism into three tetrahedrals. [Figures from S.
Harig]
28
Chapter 2. Methods
are approximated as polynomials using basis functions of the subspaces Vh
and Wh , respectively. We employ piecewise linear basis functions in two dimensions {Mj }j=1..M for η, while piecewise linear basis functions in three
dimensions {Nj }j=1..N are used for ocean velocity (the so-called P1 -P1 discretization), where M and N denote the number of 2D and 3D nodes of the
triangulation Th , respectively. Velocity and sea surface height can thus be
approximated as
uh =
N
X
(uj , vj )Nj ,
(2.64)
ηj M j ,
(2.65)
j=1
ηh =
M
X
j=1
where uj , vj and ηj denote the model values for velocity and sea surface
height in node j. Basis functions Mj and Nj equal one in node j and go
linearly to zero at neighboring nodes. The discrete expressions for velocity
and sea surface height, uh and ηh , are now inserted into equations 2.62 and
2.63. This approach is called Galerkin-method. As test functions we choose
the basis functions of the discrete subspaces Vh and Wh . This approach leads
to large systems of linear equations, which are solved iteratively for the nodal
values uj , vj and ηj . After obtaining discrete solutions for horizontal velocity
and sea surface height, we solve for the vertical velocity w. The equation
involves first order derivatives. A numerical solution of this equation leads
to difficulties, and in order to avoid these problems, we introduce a new
variable, the vertical velocity potential Φ, such that w = ∂z Φ. The variational
formulation of equation 2.7 is then: Find Φ ∈ H 1 (Ω), such that for each
Φ̃ ∈ H 1 (Ω) applies:
Z
∂z Φ∂z Φ̃ dΩ = −
Ω
Z
u · ∇Φ̃ dΩ
Ω
(2.66)
2.4. Finite element discretization
29
The potential Φ is approximated in the finite element sense as
Φh =
N
X
Φj Nj ,
(2.67)
j=1
with nodal values Φj . After computing Φh , the vertical velocity w is obtained
as an element-wise constant function.
The remaining equations for tracers (eq. 2.10) are solved in a similar
manner, i.e. setting up the variational formulation, discretizing this formulation with the Galerkin method, and solving the resulting large systems of
linear equations. With T and S obtained, the nodal values of the density are
computed via the equation of state (eq. 2.13). Then the hydrostatic pressure
is computed from equation 2.3 considered in the finite difference sense.
Time stepping scheme of the ocean component In the model, time
stepping of the momentum equation works in the following way: As explicit
Coriolis and vertical viscosity terms can limit the possible time steps in some
situations, we treat the Coriolis term semi-implicitly (by estimating this term
as a weighted sum over steps n + 1 and n) and the vertical viscosity term implicitly. The elevation term is also treated implicitly; this has the advantage
that it suppresses gravity waves.
Solving the dynamic part of the ocean model with the standard Galerkin
method leads to major numerical problems, which are associated with the
advection-dominated flow with high Reynolds numbers on the one side and
the possibility of pressure modes because of using the same horizontal test
functions for velocity and elevation on the other side. To overcome these
problems, the characteristic Galerkin and pressure projection methods are
introduced.
Before starting the iteration, potential temperature and salinity are initialized by setting them to climatological values; furthermore velocity is initialized by setting it to zero. Before each iteration step, the atmospheric
forcing data is read in. The dynamics and thermodynamics are staggered
in time by one half time step. First, density and hydrostatic pressure are
30
Chapter 2. Methods
computed from salinity and potential temperature. The solution of the horizontal momentum and continuity equation is performed in three steps with
the pressure-projection method, which is a predictor-corrector-method:
(i) in the predictor step, an auxiliary horizontal velocity u∗ is calculated,
(ii) then the sea surface height is determined,
(iii) in the correction step the correct velocity is determined.
Next, the vertical velocity is computed by determining the vertical velocity
potential. As a last step, tracer equations are solved for potential temperature and salinity. Tracer advection is solved with the flux-corrected-transport
(FTC) method. The time stepping scheme of the ocean component is described in more detail in Wang et al. (2008).
Time stepping scheme of the coupled ocean-ice model After the
initialization of ocean and ice variables, the atmospheric forcing data is read
in. Every iteration step of FESOM consists of the following steps:
(i) In the ocean to ice-coupling routine, the variables potential temperature, salinity and horizontal velocity of the ocean surface layer, Ts , Ss ,
and us , and the sea surface height η are handed over to the sea ice
component.
(ii) One time step of the sea ice model is performed, resulting in variables
ice concentration A, thickness h and snow thickness hs .
(iii) In the ice to ocean-coupling routine, the fluxes of heat, freshwater and
momentum are handed over to the ocean component.
(iv) One time step of the ocean model is performed, resulting in variables
horizontal ocean velocity u, vertical ocean velocity w, potential temperature T , salinity S, density ρ, pressure p and sea surface height
η.
2.5. Mesh generation
2.5
31
Mesh generation
In order to numerically solve the governing equations with the finite element
method, a triangulation of the model domain Ω is needed. Here, the triangulation consists of triangular elements on the ocean surface and of tetrahedral
elements in the 3D domain. First, a triangular unstructured 2D surface mesh
is generated. In this work the surface mesh is generated with the software
package Triangle (Shewchuk , 1996). The mesh quality, i.e. the extent in
which the triangle resembles an equilateral triangle, is further improved by
several relaxation methods. Also, there has to be a transition zone between
areas of high and coarse resolution. Then the 3-dimensional mesh is generated by cutting the volume into horizontal layers, the z-layers. The resulting
prisms are then divided into three tetrahedral elements. 3D nodes are always
aligned vertically below the 2D surface nodes.
In this work, the model domain covers the global oceans. For the surface
mesh, we use a nominal horizontal resolution of 1.5◦ throughout most of the
world’s ocean; resolution is doubled along the coastlines and further improved
to 24 km north of 50◦ N. The two major CAA straits are artificially widened
to allow for the presence of at least three grid points; this approach is similar
to the one used in traditional climate models. A second surface mesh was
produced, with a further refinement up to 5 km in the CAA (depicted in
Figure 2.2). The two surface meshes are denoted by LOW and HIGH, and
their resolution is given in Table 2.2. Fury and Hecla Strait connecting
the CAA and Foxe Basin (with a width of 2 km) is closed in our model.
In the vertical both meshes have 55 z-levels with thickness of 10 m in the
top 100; vertical resolution is gradually increasing downwards (see Table A1
in the appendix). The model bathymetry is taken from the IBCAO data
(Jakobsson et al., 2008) for the Arctic region and from the 1 min resolution
GEBCO data for other regions. A linear combination of the two datasets is
taken between 64◦ N and 69◦ N . To avoid the north pole singularity, rotated
coordinates with the north pole displaced to Greenland are used.
32
Chapter 2. Methods
Figure 2.2: Resolution in km of the global mesh in the CAA area (top) and
in the northern hemisphere (bottom).
2.6. Model setup and atmospheric forcing
33
Table 2.2: Resolution of the unstructured meshes.
HIGH
LOW
CAA
5 km
north of 50◦ N 24 km
globally
1.5◦ · 1.5◦ cos(lat)
2.6
24 km
24 km
1.5◦ · 1.5◦ cos(lat)
Model setup and atmospheric forcing
This configuration of FESOM is forced by atmospheric data taken from
the Common Ocean-Ice Reference Experiment version 2 data set (COREII, Large and Yeager , 2008). The forcing fields used in FESOM are listed
in Table 2.3. The CORE-II data is provided on the T62 grid (with a zonal
resolution of approximately 1.875 degree). Figure 2.3 shows the mean temperature and wind field in the CAA region in winter and summer from the
CORE-II dataset. In winter, temperatures are extremely low, up to -36◦ C,
while in summer temperature rises above the freezing point in the southern
part of the CAA.
Table 2.3: Atmospheric forcing fields taken from CORE-II
(Large and Yeager , 2008) used in this FESOM configuration.
variable
unit
resolution in time
10m zonal wind
10m meridional wind
10m air temperature
10m specific humidity
precipitation
downward shortwave radiation
downward longwave radiation
m/s
m/s
K
kg/kg
mm/s
W/m2
W/m2
6-hourly
6-hourly
6-hourly
6-hourly
monthly
daily
daily
34
Chapter 2. Methods
Figure 2.3: Mean air temperature and wind field at 10 m height in the CAA
region from the CORE-II data set for January (top) and July (bottom) 19582007. The temperature is shown by color patches, and the wind field is shown
by arrows.
2.7. Computation of transports
35
River runoff was taken from the dataset described in Dai et al. (2009),
which contains the monthly streamflow at the farthest downstream stations
of the world’s 925 largest ocean-reaching rivers.
Two hindcast simulations were carried out with the only difference being
the horizontal resolution in the CAA region. The simulations were initialized
with the mean temperature and salinity fields from the Polar Hydrography
Center global ocean climatology version 3 (PHC 3.0, Steele et al., 2001).
The initial sea ice concentration and thickness were taken from the long
term mean of a previous simulation. A sea surface salinity restoring to the
monthly PHC 3.0 climatology with a piston velocity of 10 m per 60 days
was applied. The model was integrated for the time period 1958-2007 and
the last 40 years are used in the analysis. Monthly mean and for some time
periods daily fields of the prognostic variables were saved for analysis.
2.7
Computation of transports
Volume transport is calculated by integrating normal velocity u · n over a
vertical cross section of area A:
Z
u · n dA.
(2.68)
Tvol =
A
Freshwater transport through a cross section of area A is calculated relative
to Sref = 34.8 psu, which is the mean salinity of the Arctic Ocean (Aargaard
and Carmack , 1989):
Tliqf w =
Z
A
Sref − S
u · n dA.
Sref
(2.69)
Sea ice volume transport is calculated as:
Tice =
Z
h
l
ρi
· uice · n dl,
ρw
(2.70)
with the ice density ρi = 920 kg m−3 and water density ρw = 1000 kg m−3 .
Freshwater transport in form of sea ice is computed by multiplying Tice with
36
Chapter 2. Methods
(Sref − Sice )/Sref , where Sice = 6 mSv.
The liquid freshwater content hf w at a particular grid point is defined as
the depth integrated salinity anomaly relative to Sref = 34.8 psu, integrated
from the surface to a depth of 500 m:
hf w =
Z
500m
z=0m
Sref − S
dz.
Sref
(2.71)
To obtain a time series of the liquid freshwater content in a region Ω we
spatially integrate hf w :
F W Cliq =
Z
hf w dΩ.
(2.72)
Ω
In this work the Arctic region is defined by Davis Strait, Fram Strait, Bering
Strait and Barents Sea Opening.
The sea ice freshwater content of the Arctic Ocean is calculated as:
F W Cice =
Sref − Sice
Vice ,
Sref
(2.73)
where Vice is the sea ice volume. Again we consider here the area of the
Arctic Ocean bounded by the straits mentioned above.
The streamfunction of meridional overturning circulation at a given latitude θ and time t is computed by zonally and vertically integrating the
meridional velocity v:
Tovert (θ, z, t) =
Z
0
z
Z
φeast
v(φ, θ, z ′ , t) dφ dz ′ .
φwest
(2.74)
Chapter 3
Model assessment in the Arctic
Ocean and CAA
We now assess the model performance, by focusing on sea ice conditions in
the Arctic Ocean (Section 3.1), on the freshwater (FW) budget of the Arctic
Ocean (Section 3.2) and on the ocean dynamics of the CAA (Section 3.3). In
this chapter, we only analyze model results from the simulation with highly
resolved CAA (simulation HIGH, see Table 2.2).
3.1
Arctic Ocean sea ice
Figure 3.1 shows the simulated sea ice extent (defined as the total area of
grid cells with at least 15% sea ice concentration) in the northern hemisphere
together with satellite passive microwave data from NSIDC (Fetterer et al.,
2009). The sea ice extent exhibits a strong seasonal cycle with peaks in
March and minimum in September. For the period 1979-2007, the mean
March (September) ice extent amounts to 15.76 ·106 km2 (7.58 ·106 km2 )
and 15.58 ·106 km2 (6.74 ·106 km2 ) for model and observation, respectively,
revealing that the summer ice extent is slightly overestimated by the model.
There is a strong decrease in the sea ice extent over the time period. For the
period 1990-2007, the modeled and observed March (September) ice extent
decreases by 0.42 ·106 km2 (0.89 ·106 km2 ) and 0.56 ·106 km2 (0.94 ·106 km2 )
37
38
Chapter 3. Model assessment
per decade, respectively. The 2007 summer minimum is well represented in
the model. The root mean square error of modeled ice extent is E=0.27 ·106
km2 , while the correlation coefficient between the modeled and observed sea
ice extent anomalies is r=0.87 (N∗ =9, significance 99%; the significance level
for the correlation is computed from the effective degrees of freedom based
on the formula by Chelton (1983)).
6
10 km
2
15
10
5
1980
1985
1990
1995
2000
2
FESOM
NSIDC
2
6
10 km
2005
0
−2
1980
1985
1990
1995
2000
2005
Figure 3.1: Mean monthly sea ice extent (top) and its anomaly (bottom) from
the model (blue) and satellite observation (red). The anomaly is calculated
by subtracting the mean seasonal cycle from the time series.
During the ten ICESat campaigns between 2003 and 2008 measurements
of sea surface elevation were conducted in the Arctic Ocean. Each of the
campaigns lasted ∼34 days and was conducted during different time periods
in spring and autumn. Based on these measurements, Kwok and Cunningham (2008) computed the ice freeboard and then estimated the ice thickness,
the mean of which is depicted in Figure 3.2 together with the model results.
As shown by both the observation and model results, the sea ice is the thickest along the northern CAA and Greenland coast and declines toward the
Siberian shelf. The model underestimates the ice thickness north of the CAA
3.1. Arctic Ocean sea ice
39
in both seasons. In spring, it slightly overestimates the ice thickness on the
Siberian shelf. In autumn, it underestimates the ice thickness in the Eurasian
Basin. The root mean square error between modeled ice thickness and observations amounts to E=0.54 m for spring and E=0.72 m for autumn. Considering large uncertainties in the observations (∼0.7 m, according to Kwok and
Cunningham (2008)), the model reasonably reproduced the observed spatial
and seasonal variation in sea ice thickness.
Figure 3.2: Mean sea ice thickness for spring (top) and autumn (bottom)
from ICESat measurements (Kwok and Cunningham, 2008) (left) and from
FESOM simulations (right). The spring and autumn means were calculated
by taking into account all available datasets for 2003-2007. The same time
periods are used for deriving the model results.
40
3.2
3.2.1
Chapter 3. Model assessment
Freshwater budget of the Arctic Ocean
Transports across the main gates of the Arctic
Ocean
Table 3.1 provides means of volume, liquid freshwater and sea ice transports
through the main gates of the Arctic Ocean derived from FESOM for the
time period 1968-2007. The table also shows observational estimates for
these straits. Freshwater transports were calculated with a reference salinity
of Sref = 34.8 psu, which is the mean salinity of the Arctic Ocean (Aargaard
and Carmack , 1989).
Davis Strait, located between Baffin Island and Greenland, is characterized by two currents with opposite directions: The northward flowing West
Greenland Current (WGC) at the eastern side of the strait is relatively warm
and salty, while the southward flowing Baffin Island Current (BIC) at the
western side is relatively cold and fresh. The BIC combines the CAA outflow of Arctic origin with the warmer WGC waters. The mean Davis Strait
liquid freshwater transport in the model is 110.6 mSv, well comparable to
observation. The mean volume transport in the model is 1.81 Sv, lower than
observed mean values, but still in the uncertainty range.
Fram Strait is the only deep connection between the Arctic Ocean and
the North Atlantic (Schauer et al., 2008). The East Greenland Current
(EGC) carries fresh and cold Arctic waters and sea ice southward, while
the West Spitzbergen Current (WSC) carries warm Atlantic waters into the
Arctic Ocean. Simulated net volume (1.34 Sv) and liquid freshwater (56.1
mSv) transports through Fram Strait agree well with the observations; the
modeled southward and northward transports are, however, much lower than
observations. Schauer et al. (2008) observed a 12 Sv northward flow and a
14 Sv southward flow, whereas FESOM simulates 3.3 Sv northwards and 4.6
Sv southwards at the observation section at 79◦ N. The underestimation of
the magnitude of transports through Fram Strait was shown in other models
as well (Lique et al., 2009; Jahn et al., 2009) and the reason might partly be
that the retroflection point of the Atlantic current is too far to the south in
3.2. Freshwater budget of the Arctic Ocean
41
simulations. Fram Strait is the main Arctic sea ice export route. Although
the model very well represents the sea ice extent and thickness, the volume
export is overestimated compared to available observations.
Table 3.1: Simulated volume and freshwater budget of the Arctic Ocean for
the time period 1968-2007 and observational estimates. Liquid freshwater
transports are given in mSv with a reference salinity of Sref =34.8 psu, volume
transports in Sv and ice transports in mSv. For observations, net transports
are given, unless otherwise noted. Standard deviations are calculated from
annual mean net fluxes. Fluxes into the Arctic Ocean have a negative sign,
and vice versa.
Budget term
Inflow
Vol. transports
Fram Strait
Davis Strait
Bering Strait
Barents Section
Liq. FW transports
Fram Strait
Davis Strait
Bering Strait
Barents Section
Ice Exports
Fram Strait
Davis Strait
Bering Strait
Barents Section
1
FESOM
Outflow Net
Obs.
Std
-3.28
-0.64
-0.65
-3.0
4.62
2.45
0.01
0.53
1.34
1.81
-0.64
-2.46
0.43
0.31
0.08
0.38
8.7
-13.5
-52.5
6.6
47.3
124.1
0.9
2.6
56.1
110.6
-51.6
9.2
9.2
17.2
6.6
3.8
111.0
24.0
-0.6
8.2
19.8
4.8
2.4
4.9
1
2
2±2.7
2.6±1, 3 2.3± 0.7
4
-0.8±0.2
5
-1.8
6
2
84±17
92±34, 3 116±41
6
-76±10
6
3±3
6
73±11
2
17
Schauer et al. 2008 (1997-2007), 2 Cuny et al. 2005 (09/1987-08/1990), 3 Curry et al. 2011
(08/2004-09/2005), 4 Woodgate et al. 2010 (1991-2007), 5 Skagseth et al. 2008 (1997-2006),
6
Serreze et al. 2006. Numbers in brackets denote the time period of observations.
42
Chapter 3. Model assessment
The Barents Sea receives salty water through the Barents Sea Opening,
contributing to a small net freshwater outflow as indicated in both model
simulations and observations.
Bering Strait is the only connection of the Arctic Ocean with the Pacific Ocean, and is ∼85 km wide and ∼50 m deep (Woodgate et al., 2010).
Pacific waters entering through Bering Strait present an important source of
freshwater to the Arctic Ocean. The Bering Strait freshwater inflow accounts
for 30 % of the total Arctic Ocean freshwater sources (Serreze et al., 2006).
The model slightly underestimates the Bering Strait volume and freshwater
inflow because of the coarse (24 km) resolution used in this region.
3.2.2
Arctic Ocean freshwater content
Compared to adjacent oceans the Arctic Ocean stores a large amount of
freshwater, including both liquid water and sea ice. The liquid part of the
freshwater is concentrated in the surface layers leading to a strong density
stratification defined mostly by salinity.
Figure 3.3 depicts the interannual variability of liquid freshwater content
(LFWC) and sea ice volume simulated by FESOM. The LFWC was calculated relative to 34.8 psu and was integrated from the surface to a depth of
500 m. The LFWC simulated by FESOM has local maxima in the years 1973,
1981, 1990 and 2000, which is comparable to those simulated by Köberle and
Gerdes (2007) (1970, 1982 ,1990 and 2000) and by Lique et al. (2009) (1966,
1981, 1988, 2001). The recent increase in LFWC, also shown by Rabe et al.
(2011), is represented in the model.
The variability in LFWC is strongly linked to freshwater exports of the
Arctic Ocean (shown in Figure 3.4). The changing rate of the LFWC and the
total export rate through the main gates of the Arctic Ocean are relatively
well correlated (r=-0.57, N∗ =12, significance 94%). A decrease in the Arctic
Ocean freshwater content is associated with increased exports of freshwater.
The correlation of the changing rate of LFWC with FW exports through
Davis Strait is the highest (r=-0.66, N∗ =12, significance 98%, not shown in
the figure), but not significant for Fram Strait FW export (r=0.18). This
3.2. Freshwater budget of the Arctic Ocean
43
points out the importance of liquid freshwater exports through the CAA in
determining the changes in the Arctic freshwater content.
The simulated sea ice volume (Figure 3.3) has local maxima in the late
1970’s and late 1980’s, and is decreasing from then on. This is similar to
the model results described by Lique et al. (2009) and Rothrock and Zhang
(2005). A change in sea ice volume must result from changes in sea ice export
and sea ice production (freezing minus melting). The study by Rothrock and
Zhang (2005) showed that sea ice production is most strongly controlled by
temperature over the entire Arctic basin, whereas the sea ice export through
Fram Strait (which dominates the overall Arctic Ocean sea ice export) is
controlled by local winds. In our simulation, there is no significant correlation
between changing sea ice volume and the Arctic Ocean sea ice export (r=0.18, time series not shown). Melting and freezing of sea ice in the interior of
the Arctic appears to be the most important in determining the ice volume
variability (r=0.52, N∗ =34, significance 99%, time series not shown).
4
1000 km
3
2
0
−2
−4
−6
1970 1975 1980 1985 1990 1995 2000 2005
Figure 3.3: Time series of simulated annual mean liquid freshwater content
anomaly of the Arctic Ocean from the surface to 500m depth (solid), sea ice
volume anomaly (dashed) and their sum (solid line with dots). The Arctic
Ocean is defined as the region bounded by Davis Strait, Fram Strait, Bering
Strait and Barents Section.
44
Chapter 3. Model assessment
60
40
mSv
20
0
−20
−40
−60
1970 1975 1980 1985 1990 1995 2000 2005
Figure 3.4: Time series of the simulated changing rate of annual mean liquid
freshwater content in the Arctic Ocean (solid) and the anomaly of the total
freshwater flux through all the Arctic Ocean gates (dashed).
3.3
The simulated CAA ocean dynamics
In this section we concentrate on the circulation pattern in the CAA region (Section 3.3.1), volume and freshwater transports through the CAA’s
main straits including a comparison with available observational data (Sections 3.3.2-3.3.4) and the constituents of the freshwater transport variability
(Section 3.3.5). Section 3.3.6 summarizes the main findings of the CAA
throughflow analysis.
3.3.1
Mean sea surface height and circulation in the
CAA region
The simulated sea level is higher in the Arctic Ocean than in Baffin Bay
(Figure 3.5b), due to lower salinity and thus lower density of the surface layer
in the Arctic Ocean. This results in a net outflow from the Arctic Ocean to
Baffin Bay. Along with the sea surface height (SSH) gradient between the
Arctic Ocean and Baffin Bay, there is a gradient of SSH across the main
straits of the CAA: In Parry Channel SSH is higher on the southern side of
the strait, in Nares Strait SSH is higher on the western side of the strait.
Over the CAA, the SSH is the highest in the west and decreases to the east.
3.3. The simulated CAA ocean dynamics
45
Figure 3.5: (a) Mean simulated horizontal velocity at 50 m depth and (b)
mean simulated sea surface height relative to the global mean for the time
period 1968-2007. Red boxes in (b) indicate areas over which SSH is averaged
(see Section 4.1).
46
Chapter 3. Model assessment
Figure 3.5a depicts the simulated mean horizontal velocity field at 50 m
depth. In the Canadian Basin of the Arctic Ocean there is an anticyclonic
circulation transporting Upper Halocline Waters (UHW) along the Canadian Arctic shelf westwards. This anticyclonic circulation is confined to the
top 200 m, and partly enters the CAA through Nares Strait, the channels
between the Queen Elizabeth Islands and McClure Strait. Below this anticyclonic current, a cyclonic undercurrent with its core at around 500 m (in
our simulation) transports Lower Halocline Waters (LHW) eastwards (not
shown; see Aksenov et al., 2010, for the definition of water masses). The
simulated mean velocity at 50 m depth is highest in the eastern Parry Channel and in the northern Nares Strait where the straits are narrow. Regarding
Parry Channel, the maximum velocity of 22 cm/s is reached north of Prince
of Wales Island. In Nares Strait a maximum velocity of 31 cm/s is reached in
its northern end, in Robeson Channel. At the same location Münchow and
Melling (2008) observed velocities exceeding 40 cm/s at around 100 m depth
in August 2003, with the simulated velocity in this particular month being
57 cm/s at 100 m depth. Arctic waters exiting the CAA through Lancaster,
Jones and Smith Sounds enter the North Atlantic through Baffin Bay. They
remain confined to the western side of Baffin Bay where they feed the Baffin
Island Current.
3.3.2
Parry Channel
Parry Channel is the main passage connecting the Beaufort Sea in the west
with Baffin Bay in the east. It comprises (from west to east) McClure Strait,
Viscount Melville Sound, Barrow Strait and Lancaster Sound. McClure
Strait is one of the major entrances to Parry Channel, apart from Byram
Martin and Penny Straits at its northern side and McClintock Channel, Peel
Sound and Prince Regent Inlet at its southern side (Figure 1.1). There is
a net flow from west to east, with the largest amount of water entering
through McClure Strait (see discussion later in this section). There is also a
cyclonic circulation around Prince of Wales Island with a southward flow in
McClintock Channel and northward flow reentering Parry Channel through
3.3. The simulated CAA ocean dynamics
47
Peel Sound (see Figure 3.5a). This cyclonic current loop results partly from
sills located south of Prince of Wales Island and in Barrow Strait which restrain further westward flow (Wang et al., 2012a). The northward transport
through Peel Sound was estimated as 0.17 Sv for April 1981 by Prinsenberg
and Bennet (1989). Our simulated value is 0.20 Sv for the same time period.
Table 3.2: Mean transports through the CAA. FESOM output is given for
the time period 1968-2007. FW transports are given in mSv with a reference
salinity of Sref =34.8 psu. Standard deviations are calculated from annual
mean net fluxes. HG & CS stands for Hell Gate and Cardigan Strait.
Budget term
FESOM
Net
Std
Vol. Transports (Sv)
Lancaster Sound
0.86
Nares Strait
0.91
HG & CS
0.04
Liq. FW Transports (mSv)
Lancaster Sound
71.36
Nares Strait
47.87
HG & CS
3.46
Ice Exports (mSv)
Lancaster Sound
4.16
Nares Strait
5.26
HG & CS
0.13
1
0.16
0.16
0.01
12.82
7.89
0.52
1.07
1.63
0.07
Obs.
1
0.46±0.09, 1∗ 0.53, 2 0.68
3
0.57±0.09
4
0.2/0.1 for CS/HG
1
32±6, 5 45 ±15
1
2.1, 6 2.7
Peterson et al. 2012 (1998-2011), 1∗ Peterson et al. 2012 (1998-2006), 2 Prinsenberg et al.
2009 (1998-2006), 3 Münchow and Melling 2008 (Aug 2003 to Aug 2006, excluding the top
30 m), 4 Melling et al. 2008, 5 Prinsenberg and Hamilton 2005 (1998-2001, excluding the
top 30 m), 6 Agnew et al. 2008 (2002-2007). Numbers in brackets denote the time period
of observations.
48
Chapter 3. Model assessment
FW Transport (mSv)
Vol. Transport (Sv)
Since 1998 the Bedford Institute of Oceanography operates an array of
moorings (Peterson et al., 2012) in western Lancaster Sound. At the location
of the mooring array, western Lancaster Sound has a maximum depth of 285
m and a width of 68 km. This geometry is well resolved in the model (max.
depth: 270 m, width: 71 km). Figure 3.6 depicts a comparison of monthly
mean net volume and freshwater transport of the observational data with
model output.
1.5
1
0.5
0
100
50
0
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
Figure 3.6: Volume (top) and freshwater (bottom) transports through western Lancaster Sound from observations (red, Peterson et al. 2012) and FESOM (blue).
The model result and the observation are highly correlated (r=0.81, N∗ =20,
significance 99% and r=0.82, N∗ =19, significance 99%, for volume and freshwater, respectively) and have a root mean square error of 0.35 Sv (volume
transport) and 37 mSv (freshwater transport). The fact that we find the
highest correlation at zero time lag indicates that the model simulates the
seasonal cycle quite well: Both the simulated volume and freshwater transport shows a maximum in August, as observed. In contrast, McGeehan and
Maslowski (2012) found the maximum volume transport through Lancaster
Sound to occur in March, and the maximum freshwater transport to occur
in August (based on their modeling study). The strength of the variability
3.3. The simulated CAA ocean dynamics
49
is captured very well by our model, as indicated by the standard deviation
(σ=0.35 Sv and σ=25 mSv for the observed volume and freshwater transport,
respectively, and σ=0.27 Sv and σ=23 mSv for the simulated time series).
The observational transport covers the whole depth of the cross section, although Icycler profilers captured the shallow depths only for some years.
Therefore, CTD data at 40 m depth had to be extrapolated to the surface
for approximating the freshwater transport when Icycler profilers were absent
(Peterson et al., 2012). Another source of uncertainty in the observations
is the sparseness of the mooring array (Peterson et al., 2012), consisting of
four moorings during the period 2001-2006 and only two moorings during the
periods 1998-2001 and 2006-2011. The observational transport depends on
the weighting of the individual moorings. The mean Lancaster Sound volume transport of 0.68 Sv for the period 1998-2006 estimated by Prinsenberg
et al. (2009) was revised to a new estimate of 0.53 Sv by using a different
weighting (Peterson et al., 2012). These uncertainties in observations make
the quantitative comparison with model results very difficult.
The time series of simulated annual mean volume, freshwater, and sea ice
transport through western Lancaster Sound are shown in Figure 3.7. The
simulated volume transport has local maxima in the years 1984, 1989, 1997
and 2002, and local minima in 1981, 1995 and 1999, which are comparable to
those simulated by McGeehan and Maslowski (2012) (maxima: 1984, 1990,
1997, minima: 1981, 1988, 1999). The mean volume transport through Lancaster Sound amounts to 0.86 Sv. Other modeling studies obtained volume
transports of 0.76 Sv (McGeehan and Maslowski (2012), for the years 19792004) and 1.02 Sv (Aksenov et al. (2010), for the years 1989-2006). The mean
liquid freshwater transport simulated by FESOM is 71 mSv, comparing to
48 mSv simulated by McGeehan and Maslowski (2012) and 64 mSv simulated by Aksenov et al. (2010). Both the simulated volume and freshwater
transports are higher than the individual observations (Table 3.2).
50
Chapter 3. Model assessment
Lancaster Sound
Vol. Transport (Sv)
1.2
1
0.8
0.6
0.4
0.2
0
1970 1975 1980 1985 1990 1995 2000 2005
FW Transport (mSv)
100
75
50
25
FW transport anomaly (mSv)
0
1970 1975 1980 1985 1990 1995 2000 2005
20
10
0
−10
−20
1970 1975 1980 1985 1990 1995 2000 2005
Figure 3.7: Annual mean transports through western Lancaster Sound. Top:
Net volume transport. Middle: Liquid (solid line) and solid phase (dashed
line) freshwater transports. Thin lines denote mean values for the period
1968-2007. Bottom: Freshwater transport anomaly (black line) and its
R
R
′
S
−S̄
various contribution: A ū SSref dA (green line), A u′ ref
dA (red line) and
Sref
R ′ S′
u Sref dA (blue line).
A
3.3. The simulated CAA ocean dynamics
51
Due to the local fine resolution, the present model simulation allows the
quantification of the relative contributions to the transport through the Parry
Channel. Time series and means of net volume and freshwater transports into
the channel are shown in Table 3.3 and Figure 3.8.
Table 3.3: Mean net transport exiting Lancaster Sound and transports through the
gates entering Parry Channel for the time period 1968-2007. Transports into the channel
have a positive sign.
Net volume Net FW
flux (Sv)
flux (mSv)
Lancaster
0.86
71
Byam Martin
Penny
McClure
Prince of Wales
McClintock
Peel
0.27
0.09
0.45
0.01
-0.27
0.29
22
7
40
2
-24
26
Net volume transport through McClure Strait accounts for 53% of the
total throughflow at western Lancaster Sound, whereas Byam Martin Strait
and Penny Strait contribute 32% and 11%, respectively. Volume transports
through McClintock Channel and Peel Sound balance each other, and the
transport through Prince of Wales Strait is negligible. All transports entering Parry Channel exhibit a similar interannual variability compared to the
Lancaster Sound outflow (r=0.88, r=0.94, r=0.96, r=-0.97 and r=0.98, N∗
ranging from 11 to 13, significance 99%, for Lancaster Sound transport with
transport at Byam Martin Channel, Penny Strait, McClure Strait, McClintock Channel and Peel Sound, respectively). This indicates that transports
through the individual channels are driven by the same forcing mechanism.
52
Chapter 3. Model assessment
Vol. transport (Sv)
1.5
Lanc
Penn
ByMa
MClu
MCli
Peel
1
0.5
0
−0.5
1970
1975
1980
1985
1990
1995
2000
2005
Figure 3.8: Simulated annual mean net volume transports into Parry Channel
Positive values mean transports into the channel (Penny Strait: green, Byam
Martin Strait: cyan, McClure Strait: blue, McClintock Channel: red, Peel
Sound: orange). The transport at the Lancaster Sound (black) is also shown.
3.3.3
Nares Strait
Nares Strait is bordered by landmasses from both sides with Greenland in the
east and Ellesmere Island in the west. Time series of annual mean transports
through Kennedy Channel (located in the central Nares Strait, with a width
of 27.7 km) are depicted in Figure 3.10. In our simulation, Arctic Ocean
waters are transported southwards, with almost no northward flow in Nares
Strait. The simulated mean volume and freshwater transports of 0.91 Sv and
47.87 mSv are slightly higher than simulated transports by McGeehan and
Maslowski (2012) (0.77 Sv and 10.38 mSv, respectively) and Aksenov et al.
(2010) (0.58 Sv and 20 mSv, respectively). The simulated interannual variability matches with the one simulated by McGeehan and Maslowski (2012),
both showing maxima in volume transport in the years 1984, 1990 and 1997,
and minima in the years 1981, 1988 and 1999.
From August 2003 until August 2006 a mooring array was operated in
Kennedy Channel as part of the Arctic Sub-Arctic Ocean Flux experiment
(Münchow and Melling, 2008). Four of the eight deployed moorings provided
data during the whole three years. A mean volume transport of 0.57±0.09 Sv
was obtained (excluding the top 30 m). The simulated volume transport for
3.3. The simulated CAA ocean dynamics
53
velocity (cm/s)
20
15
10
5
0
−5
Jan03
Jul03
Jan04
Jul04
Jan05
Jul05
Jan06
Jul06
Figure 3.9: Sectionally averaged along channel flow through Kennedy Channel excluding the top 30 m. The red line depicts observations (Münchow and
Melling, 2008) and the blue line shows the model output as 11-day running
means.
this time period amounts to 0.69 Sv, when excluding the top 30 m (0.81 Sv for
the whole cross section). Figure 3.9 compares sectionally averaged velocity
through the channel obtained from the mooring with simulated velocities
and reveals relatively good agreement in magnitude and correlation (r=0.57,
N∗ =65, significance 99%). The standard deviation of the observed velocity
is larger (σ=3.49 cm/s) than the standard deviation of simulated velocity
(σ=2.34 cm/s). The lower variability in the model simulation might be
linked to the coarse resolution of the atmospheric wind forcing; which should
be explored in future studies.
54
Chapter 3. Model assessment
Nares Strait
Vol. Transport (Sv)
1.2
1
0.8
0.6
0.4
0.2
0
1970 1975 1980 1985 1990 1995 2000 2005
FW Transport (mSv)
100
75
50
25
FW transport anomaly (mSv)
0
1970 1975 1980 1985 1990 1995 2000 2005
20
10
0
−10
−20
1970 1975 1980 1985 1990 1995 2000 2005
Figure 3.10: The same as Figure 3.7, but for Kennedy Channel, located in
the central Nares Strait.
3.3. The simulated CAA ocean dynamics
3.3.4
55
Cardigan Strait and Hell Gate
Waters exiting Jones Sound have to pass the narrow channels of Cardigan
Strait and Hell Gate with widths of 8 and 4 km, respectively, hence volume
exports through this gate are small compared to Lancaster Sound and Nares
Strait (see Table 3.2). With 5 km resolution we cannot resolve these narrow
channels, they are widened to have at least one grid point between the solid
side walls in order to allow water to pass through. However, the simulated
volume transport through Cardigan Strait and Hell Gate is still much lower
than the observation. Resolving and studying the flow through these channels
is beyond the scope of the current work.
3.3.5
Constituents of the freshwater transport variability
Freshwater transport depends on both the salinity and horizontal velocity
fields. In order to determine the constituents of the variability, we split up
velocity u and salinity S into its mean (denoted by bar) and its anomaly
(denoted by prime), u = ū + u′ and S = S̄ + S ′ . Freshwater transport
through a cross section A can thus be expressed as the sum of the following
R S −S̄
four terms: the mean advection of mean salinity, A ū ref
dA, the advecSref
R
′
tion of salinity anomalies by the mean currents, A ū SSref dA, the anomalous
R
S
−S̄
dA, and the anomalous advection of
advection of mean salinity, A u′ ref
Sref
R ′ S′
salinity anomalies, A u Sref dA. The bottom panels of Figures 3.7 and 3.10
indicate that the freshwater transport anomaly through Lancaster Sound and
Nares Strait is determined mainly by the volume transport anomaly. This
is in agreement with other modeling studies (Jahn et al., 2009; Lique et al.,
2010; Houssais and Herbaut, 2011). The correlation between the volume and
freshwater transport is r=0.99 (N∗ =11, significance 99%) and r=0.95 (N∗ =12,
significance 99%) for Lancaster Sound and Nares Strait, respectively.
56
3.3.6
Chapter 3. Model assessment
Overall transports through the CAA
In our simulation, volume transports through Lancaster Sound make up 47%
of the total fluxes through the CAA, with Nares Strait and Hell Gate/Cardigan
Strait contributing with 50% and 3%, respectively (see Table 3.2). The respective values for net liquid freshwater transports amount to 58%, 39%
and 3%. Volume transport through Nares Strait is the largest among the
three straits, while Lancaster Sound shows the highest freshwater transport.
Larger freshwater transport through Lancaster Sound can be explained by
the pathway of Arctic Ocean waters. The western entrance to Parry Channel is closer to the Beaufort Sea and thus to the Beaufort Gyre which is the
main freshwater storage of the Arctic Ocean. Nares Strait in contrast, with
its northern entrance located in the Lincoln Sea, is farther away from the
Beaufort Gyre. Also, due to the larger depth of Nares Strait compared to
Lancaster Sound, a higher fraction of saltier LHW enters this strait. Our results are consistent with those described in the work of Aksenov et al. (2010):
higher freshwater transports through Lancaster Sound than through Nares
Strait, because Lancaster Sound is dominated by less salty UHW, and Nares
Strait is dominated by the saltier LHW.
Observed net volume transports through the CAA sum up to 1.33 Sv
(0.46 Sv for Lancaster Sound, 0.57 Sv for Nares Strait and 0.3 Sv for Cardigan Strait/ Hell Gate, Table 3.2), while Davis Strait volume transport was
estimated as 2.6 Sv and 2.3 Sv for two different observational periods (Table
3.1). This discrepancy in the observations can be attributed to different time
periods, different instrumentation and the sparseness of instrument coverage
across the sections. The simulated net volume transport through the CAA
(and through Davis Strait) is 1.81 Sv, and thus lower than observational
estimates for Davis Strait throughflow and higher than those for the CAA
throughflow. The simulated mean freshwater export of the CAA amounts to
123 mSv, and is slightly higher than the climatological value of 101±10 mSv
estimated by Serreze et al. (2006).
Simulated freshwater exports in form of sea ice are small compared to
liquid freshwater exports. They make up 7% of the liquid freshwater exports
3.4. Summary
57
of the CAA in our model. Based on observations, Peterson et al. (2012)
obtained a similar value for Lancaster Sound: the observed mean solid freshwater transport through this section represents 6% of the observed mean
liquid freshwater transport over the period 2003-2007.
Transports through Lancaster Sound and Nares Strait are highly correlated (r=0.86, N∗ =13, significance 99%, for volume transports, and r=0.80,
N∗ =14, significance 99%, for liquid freshwater transports), indicating related
forcing mechanisms for both straits. This will be further discussed in the
next section.
3.4
Summary
The CAA is a remote area, covered by sea ice most of the year, and only
a limited number of hydrographic observations are available. Therefore,
high resolution modeling studies are required in order to better estimate
its oceanographic properties and understand the flow dynamics. The present
study assesses a new global finite element sea ice-ocean model (FESOM) with
high mesh resolution (∼5 km) in the CAA region and intermediate resolution
in the Arctic Ocean and northern North Atlantic (∼24 km). FESOM is an
unstructured-mesh model, which enables us to refine the mesh locally in areas
of interest. This method is particularly useful in the CAA area and allows us
to investigate the characteristics of the individual CAA channels. Our study
differs from previous coarse resolution Arctic freshwater circulation studies
(e.g. Jahn et al., 2009) in the treatment of the CAA bathymetry. While
modelers in traditional global ocean models manually widen the CAA straits
and sometimes keep only one strait, our resolution of 5 km more realistically
represents the geometry of the CAA main straits. The global configuration
with local mesh refinement also obviates the need for open boundaries in
regional models.
The focus of this study lies on volume and freshwater transports through
the CAA and its interannual variability. Mean net volume (liquid freshwater) transports through the CAA’s main straits Lancaster Sound and Nares
Strait for the time period 1968-2007, as simulated here, amount to 0.86 Sv (71
58
Chapter 3. Model assessment
mSv) and 0.91 Sv (48 mSv), respectively. The simulated variability of volume
transport through Lancaster Sound matches well with observations (r=0.81),
and also a comparison of sectionally averaged velocity in Nares Strait shows
reasonable agreement of the model and observations (r=0.57). The variability of freshwater transport through the CAA is mainly determined by the
variability of volume transport as shown by observations (Peterson et al.,
2012) and previous model studies (Jahn et al., 2009; Houssais and Herbaut,
2011). The total simulated CAA mean freshwater transport is 123 mSv,
slightly higher than the estimate (101±10 mSv) of Serreze et al. (2006).
The high mesh resolution allows us to quantify the individual contributions to the Parry Channel inflow. 53% of the total throughflow enters
the channel through McClure Strait, while 32% and 11% are contributed by
Byam Martin Strait and Penny Channel, respectively. Transports through
the straits feeding the Parry Channel throughflow show a similar interannual
variability, which is due to the same driving mechanism.
Chapter 4
Interannual variability of
freshwater transports through
the CAA and driving
mechanisms
In this section the mechanisms driving the freshwater transport variability
through the archipelago will be discussed, with focus on the interannual
variability. We concentrate on the liquid component of freshwater, which
dominates the total simulated export (93% of the CAA freshwater export is
liquid freshwater, see Section 3.3.6).
4.1
The role of sea surface height
The SSH difference between the Arctic Ocean and Baffin Bay (Figure 3.5b)
not only leads to a net outflow from the Arctic Ocean, its variability also
drives the variation of the CAA throughflow. The correlation of annual mean
freshwater transports though Lancaster Sound and Nares Strait with the
along strait SSH gradients is significant (Figure 4.1). At Lancaster Sound,
the upstream SSH variation plays the major role, and in particular SSH
variation on the western side of the strait. There is no significant correlation
59
60
Chapter 4. Interannual variability
of transport with downstream SSH. On the contrary, in Nares Strait it is
rather the downstream SSH and in particular the eastern side that correlates
with transports most significantly. Houssais and Herbaut (2011) have also
found that the transport through the CAA is highly correlated with the
along strait SSH gradient for both straits. However, in their case the volume
transport through Lancaster Sound is not so strongly correlated with the
upstream SSH as in our case. They only found high correlation with the along
strait SSH gradient rather than SSH at one single side. Regarding Nares
Strait they observed a high correlation of transport with the downstream
SSH, which is similar to our results.
Lancaster Sound
Nares Strait
1
1
0.5
0.5
0
0
−0.5
−0.5
a)
−1
−2
b)
−1
0
lag (years)
1
2
−1
−2
−1
0
lag (years)
1
2
Figure 4.1: Cross correlation of the anomalies of annual mean FW transport
through (a) western Lancaster Sound and (b) Nares Strait with annual mean
SSH upstream (dashed line), downstream (dashed-dotted line) and their difference (solid line). SSH was calculated as the average over the red boxes
shown in Figure 3.5b.
To quantify the amount of freshwater transported through the CAA explained by a certain SSH difference, we perform a linear regression analysis.
The time series of annual mean freshwater (and volume) transport T Stransport
is regressed on the along strait SSH difference T S∆SSH by solving the equation
T Stransport (i) = a · T S∆SSH (i) + b + ǫ(i), i = 1, .., N
(4.1)
in a least squares sense, with N being the size of the time series, pa-
4.1. The role of sea surface height
61
rameters a, b and residuum ǫ to be solved. Following the regression analysis
we found that an increase of 20 cm in the SSH difference between McClure
Strait and Lancaster Sound would result in an increase in freshwater (volume) transport of 138 mSv (1.6 Sv). In Nares Strait an increase of 20 cm
would lead to an increase in freshwater (volume) transport of 105 mSv (2.2
Sv).
The spatial maps of correlations between SSH and freshwater transports
through Lancaster Sound and Nares Strait reveal more details (Figure 4.2).
For both straits a positive correlation is apparent along the Beaufort Sea
coast. SSH and transports are negatively correlated in eastern Baffin Bay
and in the Labrador Sea. The correlation maps for both straits in Figure 4.2 are very similar, as expected from the fact that transports through
both straits are highly correlated. It is then expected that the same large
scale atmospheric circulation pattern is responsible for the transport variability through the two straits, which increases (decreases) the SSH along
the American coast in the Arctic Ocean and decreases (increases) the SSH
in the Labrador Sea and Baffin Bay at the same time. This will be further
discussed in the following sections.
62
Chapter 4. Interannual variability
Figure 4.2: Correlation of the simulated annual mean FW transport through
Lancaster Sound (left) and Nares Strait (right) with annual mean sea surface
height. Black contours indicate areas of 95% significance.
4.2. Large scale atmospheric forcing
4.2
63
Large scale atmospheric forcing
As seen in the previous section there is an indication that there could exist
a large scale forcing mechanism which simultaneously leads to a change of
SSH in the Arctic and in the Labrador Sea/Baffin Bay. By investigating the
relationship of CAA transports with sea level pressure, we observe that transports through both straits are negatively correlated with sea level pressure
in the Arctic Ocean and around Greenland, and are positively correlated in
the northern North Atlantic (Figure 4.3).
The correlation pattern are similar for both straits. This dipole structure
is related to the North Atlantic Oscillation (NAO), the dominant mode of
atmospheric variability in the North Atlantic and Arctic Ocean (Dickson
et al., 2000). A high NAO-index is characterized by a low pressure anomaly
in the Arctic and around Greenland, and by a high pressure anomaly in
the central North Atlantic. A low pressure anomaly in the Arctic drives
a cyclonic wind anomaly, whereas a strong pressure difference between the
two anomaly cells leads to strong winds over the Labrador Sea. The FW
transports through Lancaster Sound and Nares Strait are highly correlated
with each other, and both are correlated with the NAO index (Figure 4.4).
The correlation coefficients between the NAO index and the FW transports
are r=0.68 (N∗ =11, significance 97%) and r=0.52 (N∗ =12, significance 90%)
for Lancaster Sound and Nares Strait, respectively.
This is consistent with the study by Condron et al. (2009) who described
experiments with artificial forcing consisting of extremely negative and positive NAO phases, and observed an acceleration of Arctic Ocean freshwater
exports, also through the CAA, during the positive NAO phase. Apparently
the same large scale atmospheric pattern controls the transport through both
CAA straits. In section 4.1 we demonstrated that transport through Lancaster Sound can be explained by SSH changes in the Arctic Ocean, whereas
Nares strait transport is more linked to SSH changes in Baffin Bay. In the
following we will explain the changes in SSH upstream and downstream of
the CAA with the variability of the large scale atmospheric forcing.
64
Chapter 4. Interannual variability
Figure 4.3: Correlation of annual mean freshwater transport through western
Lancaster Sound (left) and Nares Strait (right) with annual mean sea level
pressure. Red contours indicate areas of 95% significance.
65
20
1
10
0.5
0
0
Lancaster Sound
Nares Strait
NAO
−10
−20
1970
1975
1980
1985
1990
1995
2000
NAO index
FW transport anomaly [mSv]
4.3. Arctic Ocean forcing on Lancaster Sound throughflow
−0.5
−1
2005
Figure 4.4: The index of the North Atlantic Oscillation (green) and anomalies
of net freshwater transports through Lancaster Sound (blue line) and Nares
Strait (dashed blue line). Time-series are 3 year-running means. The NAO
index was provided by NOAA Climate Prediction Center at their web site
http://www.cpc.ncep.noaa.gov/products/precip/CWlink/ENSO/verf/new.nao.shtml.
4.3
Arctic Ocean forcing on Lancaster Sound
throughflow
Figure 4.1 shows that Lancaster Sound transport is highly correlated with sea
level changes upstream. In the last section we pointed out that high transports through Lancaster Sound are connected with a low pressure anomaly in
the Arctic Ocean, which is associated with a cyclonic wind anomaly. In the
following the impact of winds will be further investigated by applying a multiple linear regression analysis. At every node of the surface mesh the time
series of the annual mean FW transport was regressed on the annual mean
surface wind for time period 1968-2007. We write the dependent variable,
the simulated FW transport Tf w , as a linear combination of the parameters
a, b and c:
Tf wi = a ui + b vi + c + ǫi , i = 1, ..., N,
(4.2)
with ui and vi being the independent variables, observed zonal and merid-
66
Chapter 4. Interannual variability
ional wind components from the CORE-II data set, respectively, ǫ being an
error term, and N denoting the number of data points in time. The coefficients a, b and c are found for each grid point by minimizing the residual
term in a least square sense. Figure 4.5 depicts the correlation coefficient of
modeled and predicted data values, i.e. of Tf w and a · u + b · v, and arrows
indicate the normalized optimal wind direction (a, b) · √a21+b2 . The Lancaster
Sound freshwater transport has the highest correlation with wind west of
Banks Island in the Beaufort Sea (the maximum correlation is r=0.73, at
-118.8◦ E, 72.8◦ N). This corresponds to northeastward winds. Our result is
consistent with Peterson et al. (2012) who performed a similar analysis with
Lancaster Sound volume transports derived from observations for the period
of 1998-2011 and had a similar finding. Figure 4.5 reveals that a positive (negative) freshwater transport anomaly through Lancaster Sound corresponds
to a cyclonic (anticyclonic) wind anomaly pattern in the Beaufort Sea.
Figure 4.5: Correlation of the simulated annual mean freshwater transport
through western Lancaster Sound with the annual mean wind field at 10
m height. Arrows denote the corresponding optimal wind direction. Red
contours indicate areas of 95% significance.
4.4. Sea level in Baffin Bay and Nares Strait throughflow
67
Proshutinsky et al. (2002) demonstrated that the Beaufort Gyre, the main
freshwater storage of the Arctic Ocean, accumulates freshwater during an
anticyclonic wind regime and releases freshwater during a cyclonic regime.
An anticyclonic wind regime in the Beaufort Sea leads to accumulation of
water masses in the gyre and thus raised SSH in the center of the gyre,
whereas a cyclonic regime leads to divergence of Beaufort Gyre surface water
and raised SSH along the American coast. Therefore, it is the wind regime
over the western Arctic Ocean that drives the variability of SSH along the
American coast, thus the variability of FW transport through the Lancaster
Sound. And the wind regime is strongly linked to the large scale atmospheric
forcing (Figure 4.3).
4.4
Sea level in Baffin Bay and Nares Strait
throughflow
In Section 4.1 we showed that the transport through Nares Strait is correlated
with sea level in eastern Baffin Bay (Figure 4.1). The correlation of Nares
Strait transport with sea level pressure (Figure 4.3) reveals that high transports are associated with a low pressure anomaly over Greenland and a high
pressure anomaly at around 30◦ latitude. This entails intensifying storms
over the northern North Atlantic, cooling in the Labrador Sea, a stronger cyclonic circulation south of Greenland (subpolar gyre) and a decrease of SSH
in this area (Lohmann et al., 2008). This is also apparent when correlating
the simulated annual mean freshwater transport through Nares Strait with
the surface heat flux (Figure 4.6), which reveals a high negative correlation
in the Labrador Sea, more precisely in the area which is ice free in the winter.
The analysis of simulated SSH time series in several locations in the eastern Baffin Bay (green boxes in Figure 4.6) shows that there is no time lag in
SSH between the three locations, neither on daily, monthly nor annual time
scales (Figure 4.7). This indicates that it is not the advection of low density
waters within the West Greenland Current, but fast wave propagation that
leads to low sea level along the West Greenland coast.
68
Chapter 4. Interannual variability
Figure 4.6: Correlation of annual mean freshwater transport through Nares
Strait with surface heat flux (downward positive). Green boxes are taken for
the computation of sea surface height indices. Red contours indicate areas
of 95% significance.
This explanation of SSH variability in the eastern Baffin Bay is similar to
the one by Houssais and Herbaut (2011). They also draw the conclusion that
during a NAO-positive phase the cooling in the Labrador Sea leads to lower
SSH there, which then propagates northwards into eastern Baffin Bay as
fast waves. Another hypothesis was proposed by McGeehan and Maslowski
(2012). They have only considered the seasonal signal and suggested that
it is the volume flux of the West Greenland Current that drives the SSH in
eastern Baffin Bay. However, the instantaneous response of the SSH at Smith
Sound to Labrador Sea SSH suggests that slow advection processes, while
undeniably present on the seasonal scale, cannot explain the SSH variability
at Smith Sound in our simulation.
4.5. Summary
m
−0.5
a)
−0.6
−0.7
−0.4
m
69
B1
1970
B2
1975
B3
1980
1985
1990
1995
2000
2005
b)
−0.6
−0.8
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
−0.3
m
c)
−0.6
−0.9
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Figure 4.7: Time series of (a) annual, (b) monthly and (c) daily (for 2006)
sea surface height averaged over the green boxes in Figure 4.6. B1 is the
northernmost box and B3 the southernmost box.
4.5
Summary
A statistical analysis has been performed to explain the interannual variability of liquid freshwater transports through Lancaster Sound and Nares
Strait. The main driver of transports is the along strait SSH difference, with
Lancaster Sound transport being mainly driven by the upstream SSH variation and Nares Strait mainly by the downstream SSH variation. The varying
sea level upstream of Lancaster Sound can be explained by the change in
the large scale wind regime in the Arctic Ocean, whereas the varying SSH in
northern Baffin Bay can be explained by ocean-atmosphere heat exchanges
over the Labrador Sea.
The variability of the CAA transports is related to the NAO, the dominant large scale atmospheric pressure pattern in the North Atlantic. Re-
70
Chapter 4. Interannual variability
sponses to the positive and negative phases of the NAO can be observed
both in the Beaufort Sea and in Baffin Bay. In the positive NAO phase,
the atmospheric circulation over the western Arctic Ocean is mainly in the
cyclonic phase (see e.g. Proshutinsky et al., 2002), leading to a loss in the
freshwater storage in the Beaufort Gyre and an increase in SSH along the
American coast, and thus increased FW transport through Lancaster Sound.
In the Labrador Sea the strong cooling during the NAO positive phase results in low SSH, which is propagated through fast waves northwards to
Smith Sound leading to higher transports through Nares Strait.
Chapter 5
Impact of mesh resolution in
the CAA on the Atlantic Ocean
circulation
5.1
Motivation
The Arctic Ocean is strongly stratified due to large surface freshwater input
from river runoff, excess of precipitation over evaporation and Bering Strait
inflow (Serreze et al., 2006). Arctic freshwater is exported through the narrow
straits of the Canadian Arctic Archipelago (CAA) and through the relatively
deep and wide Fram Strait into the North Atlantic. Freshwater exported from
the Arctic Ocean may have a significant impact on the Atlantic Meridional
Overturning Circulation (AMOC, see Goosse et al., 1997; Wadley and Bigg,
2002; Cheng and Rhines, 2004; Komuro and Hasumi , 2005). It is plausible,
therefore, that an adequate representation of Arctic freshwater export in
climate models is crucial when it comes to predicting climate variability and
change.
Despite the importance of the freshwater export through the Arctic gateways its monitoring and numerical simulation has been a challenge, especially
for the CAA. In situ measurements of the ocean hydrography and velocity in
the narrow straits of the CAA are still rather sparse in time and space due
71
72
Chapter 5. Impact of mesh resolution in the CAA
to their remote location and harsh weather conditions (Prinsenberg et al.,
2009; Münchow and Melling, 2008). Parry Channel and Nares Strait in the
CAA, for example, have a minimum width of about 52 km and 28 km, respectively. In traditional climate models these straits are too narrow to be
explicitly resolved, and in practice they are manually widened in order to
to allow for some freshwater export. Most of our understanding of oceanic
key processes in the CAA is based on high-resolution regional models (e.g.
Wang et al., 2012a; McGeehan and Maslowski , 2012) given past and present
computational limitations. Global high-resolution modeling studies are upcoming, e.g. by Aksenov et al. (2010), but cover shorter time periods so far.
Besides, the study by Jahn et al. (2012) shows a large spread of the stateof-the-art models in simulating the Arctic Ocean freshwater exports, with
better model-to-model agreement in the CAA than in Fram Strait.
A new generation of global models is emerging which employ multiresolution mesh methods, such as finite elements, and therefore allow local refinement without the necessity for nesting. The availability of such
models opens up new horizons for both dynamical downscaling and upscaling. The main motivation for developing multi-scale climate models is based
on the following two hypotheses: The representation of local dynamics can
be improved in a global set-up (downscaling); and a better representation
of small-scale processes leads to a better simulation of the large scale circulation (upscaling). To test these hypotheses, it is necessary to identify
dynamically important regions where meso-scale processes and hence local
mesh refinement are believed to be important. The geometrical properties of
the CAA (small-scale aspects) and its proximity to the deep convection sites
(with potential climate relevance) make it a very promising testbed.
In this work FESOM is used with and without mesh refinement in the
CAA in order to establish the role of resolution for representing the freshwater
transport through the CAA and to understand possible impacts on the largescale ocean circulation. This is one of the first studies to test the abovementioned hypothesis for one special case —the CAA.
Two experiments (denoted by HIGH and LOW) were performed, with the
only difference in the configuration being the mesh resolution in the CAA
5.2. Freshwater transport through the CAA
73
area. In experiments HIGH and LOW, the mesh resolution in the CAA is 5
km and 24 km, respectively. North of 50◦ N the resolution in both experiments
is 24 km, and in the global oceans it is 1.5◦ (see Section 2.5, Figure 2.2 and
Table 2.2). Figure 5.1 depicts the global mesh used in experiment HIGH.
Figure 5.1: Northern hemisphere view of the global mesh used for experiment
HIGH. The color patch shows the mesh resolution. The CAA region has a
resolution of 5 km. Green lines indicate Lancaster Sound (A), Nares Strait
(B), a section near 53◦ N (C) and the Greenland-Iceland-Norway section (D).
5.2
Freshwater transport through the CAA
Time series of liquid freshwater transports (relative to a salinity of 34.8 psu,
the mean salinity of the Arctic Ocean) through the two main straits of the
CAA are shown in Figure 5.2a,b. The freshwater export rates are higher
for both straits in HIGH than they are in LOW. In Lancaster Sound the
mean liquid freshwater transport amounts to 33 mSv and 71 mSv in LOW
and HIGH, respectively, while in Nares Strait it is 26 mSv and 48 mSv. Importantly, the total liquid freshwater transport through the CAA in HIGH
(119 mSv) is much closer to the climatological liquid freshwater transport
74
Chapter 5. Impact of mesh resolution in the CAA
(101 mSv) estimated by Serreze et al. (2006). Freshwater transports in form
of sea ice through the CAA account only for 7% of the total freshwater
transports (9 mSv and 2 mSv in cases HIGH and LOW, respectively).
FW Transport (mSv)
80
a)
60
40
20
0
1970
150
Nares Strait
1980
1990
100
2000
c)
100
50
0
1998 2000 2002 2004 2006
velocity (cm/s)
FW Transport (mSv)
FW Transport (mSv)
Lancaster Sound
100
80
b)
60
40
20
0
1970
20
15
1980
1990
2000
Obs.
HIGH
d)
10
5
0
−5
2003
2004
2005
LOW
2006
Figure 5.2: (a) Annual mean net liquid freshwater transport through Lancaster Sound. Horizontal lines denote the mean value for 1968-2007. (b)
Same as (a) but for Nares Strait. (c) Monthly mean net liquid freshwater transport through Lancaster Sound for experiments HIGH, LOW and
observation (red, Peterson et al., 2012). (d) 11-day running mean of sectional velocity in Nares Strait excluding the top 30 m. The running mean
is shown to better illustrate the low frequency variability which is the focus
of the current work. The red line depicts the observation (Münchow and
Melling, 2008). In all four plots, results from experiments HIGH and LOW
are depicted by blue and cyan lines, respectively. Freshwater transports are
calculated relative to 34.8 psu and southward transports have a positive sign.
Mooring arrays have been deployed in the western Lancaster Sound since
1998 to monitor the volume and freshwater transport (Prinsenberg et al.,
2009; Peterson et al., 2012). The observed and simulated monthly mean
freshwater transports in Lancaster Sound are shown in Figure 5.2c. Evidently, there is a good correlation between the observed time series with the
5.2. Freshwater transport through the CAA
75
one obtained by HIGH (r=0.82); considerably lower correlations are found for
LOW (r=0.62). Furthermore, HIGH also shows improvements on the level
of freshwater export variability as indicated by the standard deviation of the
monthly mean data (observations: σ = 25 mSv, HIGH: σ = 23 mSv, LOW:
σ = 10 mSv). On the other hand, its mean value for this period (68 mSv) is
too high compared to the estimates obtained from the observations (34 mSv).
A pronounced overestimation of the Lancaster Sound transport compared to
the same observational dataset was also reported in a regional high resolution
model (Lu et al., 2010). It is possible that this overestimation is due to more
general model deficiencies which are not directly related to model resolution.
It is also conceivable that model validation is hampered by large observational uncertainties given that the results presented here were derived from
only two (some periods four) moorings which generally did not observe the
uppermost 30 m, which contains most of the freshwater. This might leave
the observational estimate sensitive to the methods used for interpolation
in the horizontal and extrapolation in the vertical (Peterson et al., 2012),
respectively.
From August 2003 to August 2006 a mooring array was operated in
Kennedy Channel in Nares Strait (Münchow and Melling, 2008). Considering the fact that the upper 30 m depth was missing in the observation,
Münchow and Melling (2008) only provided the total volume transport below 30 m depth for Nares Strait (0.57 ± 0.09 Sv). Corresponding volume
transports simulated by FESOM for the same time period amount to 0.69 Sv
and 0.25 Sv in runs HIGH and LOW, respectively. Figure 5.2d compares the
sectionally averaged velocity in Kennedy Channel from the mooring array
and simulations. The experiments HIGH and LOW have similar correlations
with the observation (r = 0.57 and 0.49, respectively); the mean value of
HIGH, however, is much closer to the observed value.
There is a large discrepancy in the observations of CAA and Davis Strait
transports. The observed Davis Strait volume transport is 2.3 - 2.6 Sv (Curry
et al., 2011; Cuny et al., 2005), while the observed CAA volume transport is
only 1.33 Sv (0.46 Sv for Lancaster Sound estimated by Peterson et al. (2012),
0.57 Sv for Nares Strait estimated by Münchow and Melling (2008) and 0.3
76
Chapter 5. Impact of mesh resolution in the CAA
Sv for Hell Gate and Cardigan Strait estimated by Melling et al. (2008)).
This discrepancy is also found in the freshwater transports. The simulated
Davis Strait freshwater transport of HIGH (111 mSv) is much closer to the
observational estimate (ranging between 92 and 116 mSv according to Curry
et al. (2011) and Cuny et al. (2005)) than the value of LOW (65 mSv).
Despite noticeable differences, the freshwater transports in the two simulations are highly correlated (r = 0.85 for Lancaster Sound and r = 0.79 for
Nares Strait). The interannual variability of freshwater transport through
the CAA is correlated with the variability of the along strait sea surface
height difference, which is in turn driven by local forcing associated with the
large scale atmospheric circulation (Houssais and Herbaut, 2011). This may
explain why the correlation for LOW is relatively high, despite its relatively
poor representation of the CAA.
As discussed above, the liquid freshwater export through the CAA is enhanced by 64 mSv when horizontal resolution is increased from about 24 to 5
km. This increase is accompanied by a decrease of freshwater export through
Fram Strait by 20 mSv; this indicates that the near-surface Arctic freshwater
content becomes lower in the case HIGH (most pronounced in the Lincoln
Sea and north of Greenland, see Figure 5.3). An observational estimate of
the Fram Strait freshwater export for 1998-2008 is 66 mSv (relative to a reference salinity of 34.9 psu, contributions to the transport over the continental
shelf are estimated by simulations, de Steur et al., 2009). Considering the
same reference salinity, the value of HIGH for 1998-2007 (60 mSv) matches
better with observations than the value of LOW (81 mSv). From idealized
model simulations employing the Island Rule for the flow around Greenland
it follows that the circulation east and west of Greenland balances each other
and that the individual contributions depend on friction and winds (Joyce
and Proshutinsky, 2007). In the next section we will show that the changed
distribution and amount of freshwater export on the two sides of Greenland
leads to changes in the large-scale circulation of the North Atlantic.
5.3. Large scale circulation
77
Figure 5.3: Mean liquid freshwater content integrated from 500 m depth to
the surface relative to 34.8 psu for 1968-2007 for experiment LOW (left) and
its difference from experiment HIGH (HIGH minus LOW, right).
5.3
Large scale circulation
Figures 5.4a and b show the mean AMOC streamfunction in experiment
LOW and the difference between HIGH and LOW, respectively. The strength
of the intermediate overturning cell increases over the whole Atlantic when
the CAA region is better resolved. The largest increase in the North Atlantic
amounts to about 2 Sv. The sizable change in the large-scale ocean circulation
highlights the importance of adequately resolving the details of the freshwater
export from the Arctic in climate simulations.
The North Atlantic Deep Water (NADW), which feeds the intermediate
cell of the AMOC, is exported from the Labrador Sea by the Deep Western
Boundary Current (DWBC). The DWBC has primarily a thermohaline origin
through the processes of deep convection and water-mass ventilation in the
GIN (Greenland-Iceland-Norwegian) Seas and in the Labrador Sea (Böning
et al., 2006; Schweckendiek and Willebrand , 2005). Previous studies indicate
that both regions have significant impact on the overturning circulation (Latif
et al., 2006; Schweckendiek and Willebrand , 2005; Danabasoglu et al., 2012;
Eden and Jung, 2001).
78
Chapter 5. Impact of mesh resolution in the CAA
Sv
a)
Depth (m)
1000
12
2000
3000
4000
12
3
4
7
5 6
0
2
10
8
4
5
−2
−1
0
−1
−20
0
20
40
Latitude
−5
60
Sv
b)
Depth (m)
1000
2000
0.6
4000
1
0.8
1.2
1.4
0.4
3000
15
10
6
0
5000
6000
11
11
10
9
8
1.6 1.8 1.4
0.6
0.8
1
1.2
1
0
0.2
0.4
2
0.2
−1
5000
6000
−20
0
20
Latitude
40
60
−2
Figure 5.4: (a) Mean AMOC streamfunction for the time period 1968-2007
for experiment LOW and (b) its difference from experiment HIGH (HIGH
minus LOW). (c) Mean winter (January-March) mixed layer depth for 19682007 for experiment LOW and (d) its difference from experiment HIGH
(HIGH minus LOW). (e) Mean salinity from the surface to 50 m depth for
1968-2007 for experiment LOW and (f) its difference from experiment HIGH
(HIGH minus LOW).
5.3. Large scale circulation
79
During wintertime mixed layer depth (MLD) represents a measure of the
convection intensity. Figures 5.4c and d show the winter mean MLD in
run LOW along with the difference between HIGH and LOW. Deep MLD
are observed in the Labrador Sea, south of the Greenland-Iceland-Scotland
Ridge, with intermediate depths in the GIN Seas; this is in agreement with
observations (de Boyer Montégut et al., 2004). When the CAA region is
better resolved, the MLD increases significantly in the northern Labrador
Sea (from 1, 480 m in run LOW to 1, 670 m in HIGH for the 40 years mean
maximum); on the other hand, it reaches to lower depths south of Greenland.
Moderate increases in MLD are found in the northern GIN seas.
The changes in the MLD can be explained by changes in the Arctic freshwater pathways (see Figure 1.2 for the circulation pattern in Baffin Bay and
the Labrador Sea). The freshwater exported through Fram Strait flows southward along the east Greenland coast as part of the East Greenland Current
(EGC). When the CAA region is better resolved, allowing for more export,
the freshwater transport in the EGC is reduced. This slightly weakens the
stratification along the ice edge in the Greenland Sea, thus intensifying the
convection slightly as shown by the increased MLD. As shown in Figure 5.5a,
overflow export from the GIN Seas is only marginally affected.
The EGC feeds the West Greenland Current (WGC) after passing the
southern tip of Greenland. A reduction in the freshwater transport in the
EGC decreased the WGC and hence the freshwater input to the Labrador
Sea (changes in the salinity of the top 50 m are shown in Figure 3f). This
explains the enhanced convection in the northern Labrador Sea in HIGH
(Figure 5.4d). Although the freshwater export through the CAA increases in
case HIGH, it propagates southward along the shelf and does not penetrate
the northern Labrador Sea as that from the WGC does (Myers, 2005; Myers
et al., 2009). Part of the anomalous freshwater from the CAA in case HIGH
recirculates in the North Atlantic subpolar gyre (SPG) and increases the
stratification south of Greenland, leading to weaker convection there.
Chapter 5. Impact of mesh resolution in the CAA
Transport (Sv)
Transport (Sv)
80
6
HIGH
LOW
2
0
30
b)
20
10
0
2500
Depth (m)
a)
4
2000
c)
1500
1000
500
1970 1975 1980 1985 1990 1995 2000 2005
Figure 5.5: (a) Overflow (defined by σ0 ≥ 27.8) export rate from the GIN
Seas (section D in Figure 5.1). (b) Dense portion (σ0 ≥ 27.7) of the DWBC
transport near 53◦ N (section C in Figure 5.1). (c) Maximum winter (JanuaryMarch) mixed layer depth in the box 60◦ -50◦ W and 55◦ -62◦ N. All time series
are 3-year filtered. Horizontal lines denote the mean value for 1968-2007.
As the convection and water-mass ventilation in this region is less significant than in the Labrador Sea’s interior, the resulting DWBC transport at
the southwestern Labrador Sea (near 53◦ N) is higher in HIGH (Figure 5.5b).
This explains the enhanced strength of the AMOC (Figure 5.4b).
Figure 5.5c shows the time series of the maximum winter mean MLD
in the Labrador Sea. As expected, this proxy for open ocean convection
correlates well with the DWBC transport (r = 0.56 (0.55) for time lags of 0
(1) year in case of HIGH, with convection leading for positive lags). Both
simulations capture the reduced convection in the early 1970s associated
with the Great Salinity Anomaly (Lazier , 1980) and the strong convection
events in the mid-1970s, mid-1980s, and early 1990s (Lazier et al., 2002;
Böning et al., 2006). The fact that LOW reproduces the decadal variability
of DWBC and AMOC is consistent with the notion that it is primarily linked
to the atmospheric circulation associated with the North Atlantic Oscillation
5.4. Summary and conclusion
81
(e.g., Curry et al., 1998; Lazier et al., 2002; Eden and Jung, 2001).
5.4
Summary and conclusion
In this work the multi-resolution model FESOM is used to study the freshwater export through the CAA and its impact on the large scale ocean circulation. By using a multi-resolution approach it becomes possible to realistically
represent the narrow straits of the CAA in a global model. A comparison
with observations reveals that the numerical representation of the freshwater
transport through the CAA benefits from the use of a locally refined mesh.
Increased mesh resolution in the CAA region not only allows for more freshwater to exit the Arctic through the CAA, it also improves the representation
of interannual freshwater export variability.
Better resolving the CAA region leads to a redirection of Arctic Ocean
freshwater transports, with more export through the CAA and less export
through Fram Strait. This redirection has significant effects on the stratification of the near surface waters in the Labrador Sea and thus open ocean
convective activity. One possible explanation for this finding is that freshwater from the EGC and WGC more vigorously penetrates into the interior
of the Labrador Sea than that from the Baffin Island and Labrador Currents
(Myers, 2005; Komuro and Hasumi, 2005). The total Arctic freshwater export to the North Atlantic is higher when the CAA region is more adequately
resolved. Part of the increased freshwater enters the subpolar gyre and increases the stratification south of Greenland. Overall, however, ventilation
and DWBC is intensified, leading to a stronger AMOC.
The high-resolution regional modeling study by McGeehan and Maslowski
(2011) shows that mesoscale eddies may play an important role in transporting freshwater from the western Labrador Sea into its interior. In this paper
we only increased the model resolution locally in the CAA, leaving horizontal resolution in the Labrador Sea region relatively coarse (24 km). Future
work aims to increase model resolution simultaneously in both the CAA and
the deep convection regions to more comprehensively assess the role of Arctic freshwater export in the climate system. The work of McGeehan and
82
Chapter 5. Impact of mesh resolution in the CAA
Maslowski (2011) also shows that the major shelf to interior liquid freshwater flux on annual time scales occurs from Hamilton Bank southward, as
found from the coarse resolution model study by Myers (2005). Therefore,
the results in our work are expected to be sensible in general, which needs
to be verified in future work.
Using the CAA as a case study we demonstrate that multi-resolution
models allow to simultaneously improve the representation of small-scale dynamical processes in a global setup (downscaling) and explicitly account for
the influence of meso-scale processes in dynamically active key-regions on the
large-scale circulation (upscaling). Our results therefore indicate that multiresolution models might prove extremely useful when it comes to significantly
advancing the field of climate modeling.
Chapter 6
Conclusions and Outlook
The Canadian Archipelago connects the Arctic Ocean with Baffin Bay and
is one of the main gateways for Arctic Ocean freshwater to enter the North
Atlantic. It is a remote area with harsh weather conditions, and thus only
a limited number of hydrographic observations are available. The area is
characterized by narrow straits, which are not accurately represented in traditional ocean general circulation models. In this work, we apply a global
configuration of FESOM with the Canadian archipelago explicitly highly resolved (∼5 km). The local high resolution allows for studying small-scale
processes, while the global set-up enables us to investigate the impact of
these small-scale processes on the large-scale circulation. Hindcast simulations were performed for the period 1958-2007, forced by the CORE-II
atmospheric reanalysis data. Two model set-ups (high and coarse resolution
in the CAA) allow us to investigate the impact of locally refined meshes. The
main findings of this work are:
• Assessment of the model in the CAA region revealed a high correlation
of modeled and observed transport through Lancaster Sound and averaged velocity in Nares Strait (r=0.81 and r=0.57, respectively). The
modeled seasonal cycle of the Lancaster Sound transport matches well
with the observed seasonal cycle, indicating that the model realistically
represents the CAA ocean dynamics. In fact, as hydrographic observations mostly do not cover the oceanic surface layer (e.g. Münchow and
83
84
Chapter 6. Conclusions and Outlook
Melling, 2008), our results provide useful information for this missing
region in observations. As the model resolves also the narrow straits of
the CAA, the relative contributions to the transport through the Parry
Channel was quantified.
• The validation of the model in the Arctic Ocean demonstrated that the
model is capable to realistically simulate ocean and sea ice conditions
in this region. Sea ice concentration in the northern hemisphere can be
realistically simulated, although FESOM slightly overestimates sea ice
concentration in the summer months.
• The interannual variability of freshwater transport through the CAA
is driven mainly by the sea surface height gradient between the Arctic
Ocean and Baffin Bay, with Lancaster Sound transport being driven
by the upstream sea level variation and Nares Strait transport by the
sea level variation in northeastern Baffin Bay. The variability of fluxes
through Lancaster Sound and Nares Strait is mainly determined by
that of the SSH on the shelf along the Beaufort Sea coast and in the
northeastern Baffin Bay, respectively. Sea level variations north of
the CAA are explained with changes in the wind regimes (cyclonic vs.
anticyclonic) associated to release or accumulation of freshwater from
the Beaufort Gyre, whereas sea level in the northeastern Baffin Bay
can be attributed to ocean-atmosphere heat fluxes over the Labrador
Sea. Both processes are linked to the North Atlantic Oscillation type
of variability.
• Mesh refinement in the CAA region leads to a ’redirection’ of freshwater pathways, with more freshwater export through the CAA and less
through Fram Strait. These changed freshwater pathways affect the
stratification of the near surface layers in the Labrador Sea, and thus
the convection activity in this region, leading to a stronger AMOC.
Future perspectives As a next step it should be considered to not only
highly resolve the CAA, but also other important regions of the world oceans.
85
Especially when studying the impact of Arctic Ocean freshwater exports on
the deep water formation in the North Atlantic, areas with strong convection
activity like the Labrador Sea and Greenland Seas as well as overflow areas
like the Denmark Strait should be better resolved simultaneously.
The ice conditions in Nares Strait are characterized by ice bridges, which
form frequently in its northern part. The FESOM ice component is not able
to simulate these ice bridges accurately. Further developing the sea ice model
remains a task in the future work.
Tides play an important role in the Canadian archipelago, especially in
mixing colder surface waters with warmer bottom waters. This has an impact
on the formation of polynyas which form frequently in the archipelago region
(Hannah et al., 2009). Hence, it should be considered to incorporate tides
into the model.
The resolution of the atmospheric forcing is rather coarse. Especially in
Nares Strait, which is bounded by landmasses from both sides, orographic
effects can have an important impact on the surface wind field. Preliminary
results of FESOM-simulations with high resolution forcing data from Environment Canada (CMC GDPS Reforecast, 33 km resolution, Smith et al.
(2013)) indicate that differences in the simulated ocean dynamics of the CAA
compared to simulations with CORE-II forcing are rather small. Regarding
sea ice conditions though, the high resolution forcing leads to higher sea ice
transport through Nares Strait and improves the simulated ice concentration in the CAA region. An additional experiment with this high resolution
forcing interpolated in time and space onto the CORE-II grid also shows
improvements in the simulation of sea ice. This indicates that improvement
is not just due to the resolution of the Environment Canada atmospheric
forcing, but also due to its quality. The aspect of atmospheric forcing should
be further investigated.
86
Chapter 6. Conclusions and Outlook
Appendix A
Vertical resolution
The vertical resolution is high in the surface layers and decreases towards
the bottom (see Table A1).
Table A.1: Model z-layers with depths in m.
no.
depth
1
0
2
10
3
20
4
30
5
40
6
50
7
60
8
70
9
80
10
90
11
100
12
120
13
140
14
160
15
180
16
200
17
230
18
270
19
320
20
370
21
420
22
470
23
520
24
570
25
620
26
670
27
730
28
800
29
890
30
1000
31
1120
32
1260
33
1400
34
35
1540 1680
36
1820
37
1960
38
2120
39
2300
40
2500
41
2750
42
3000
43
3250
44
45
3500 3750
46
4000
47
4250
48
4500
49
4750
50
5000
51
5250
52
5500
53
5750
54
55
6000
87
88
Appendix A. Vertical resolution
Appendix B
A list of symbols
Table B.1: Symbols.
symbol
meaning
A
Ah
Av
Av0
C
Cd
Cd,ao
Cd,io
Ce,ao
ice concentration
lateral momentum diffusion coefficient
vertical momentum diffusion coefficient
background vertical momentum diffusion coefficient
tracer
bottom drag coefficient
atmosphere-ocean drag coefficient
ice-ocean drag coefficient
atmosphere-ocean transfer coefficient for the
exchange of sensible heat
ice-ocean transfer coefficient for the exchange
of sensible heat
atmosphere-ocean transfer coefficients for the
exchange of latent heat
ice-ocean transfer coefficients for the
exchange of latent heat
empirical parameter controlling the ice strength
specific heat capacity of water
Ce,io
Ch,ao
Ch,io
c
cp
89
90
Appendix B. A list of symbols
Table B.1: Symbols.
symbol
meaning
F
F salt
FPsalt
−E
internal force within the ice
virtual salt fux into the ocean
virtual salt fux into the ocean due to
precipitation minus evaporation
virtual salt fux into the ocean due to ice melting
virtual salt fux into the ocean due to river runoff
Coriolis parameter
gravitational acceleration
ocean bottom depth
effective ice thickness
effective snow thickness
lateral diffusivity
vertical diffusivity
background vertical diffusivity
Gent/McWilliams diffusion tensor
Redi diffusion tensor
latent heat of vaporization or sublimation
vertical unit vector
ice mass per area
buoyancy frequency
2D unit normal vector
3D unit normal vector
ice strength
empirical parameter controlling the ice strength
hydrostatic pressure
long wave radiative flux
short wave radiative flux
conductive heat flux
turbulent flux of latent heat
turbulent flux of sensible heat
surface flux
specific humidity of air
salt
Fice
salt
Frunof
f
f
g
H(x, y)
h
hs
Kh
Kv
Kv0
KGM
Kredi
L
k
m
N
n
n3
P
P∗
p
QLW
QSW
Qc
Ql
Qs
q
qa
91
Table B.1: Symbols.
symbol
meaning
Ri
S
Sice
Ss
Sref
Sh
Ssn
SA
T
Ta
Tf
Ts
t
u
us
u10
uice
u∗
v
w
α
ǫ
ǫ̇
η
λ
Φ
ρ
ρ0
ρa
ρi
ρsn
ρw
Richardson number
salinity
salinity of sea ice
salinity of the oceanic surface layer
reference salinity
sources and sinks in the ice thickness continuity equation
sources and sinks in the snow thickness continuity equation
sources and sinks in the ice concentration continuity equation
potential temperature
air temperature
freezing point of sea water
potential temperature of the oceanic or ice surface layer
time
horizontal ocean velocity
horizontal ocean velocity of the surface layer
horizontal wind field at 10 m height
sea ice velocity
auxiliary horizontal velocity
3D ocean velocity
vertical velocity
albedo
emissivity
tensor of deformation rates
sea surface height
longitude
vertical velocity potential
mean density of sea water
deviation from the mean density of sea water
air density
density of sea ice
density of snow
density of freshwater
92
Appendix B. A list of symbols
Table B.1: Symbols.
symbol
meaning
σ
θ
τo
τai
τao
τio
Ω
Γ1 , Γ2 , Γ3
∇
∇3
∆
stress tensor
latitude
ocean surface stress
stress between atmosphere and ice
stress between atmosphere and ocean
stress between ice and ocean
model domain
boundaries of the domain Ω
2D gradient and divergence operator
3D gradient and divergence operator
square root of the surface triangle area
93
abbreviation
Table B.2: Abbreviations.
meaning
AMOC
AO
BIC
CAA
CTD
DWBC
DWF
EGC
FEM
FESOM
FW
GIN Sea
LFWC
LHW
LSW
MLD
NA
NADW
NAO
NP
NWP
SSH
UHW
WGC
WSC
Atlantic Meridional Overturning Circulation
Arctic Ocean
Baffin Island Current
Canadian Arctic Archipelago
conductivity temperature depth
deep western boundary current
deep water formation
East Greenland Current
finite element method
Finite Element Sea ice Ocean Model
fresh water
Greenland-Iceland-Norwegian Sea
liquid freshwater content
lower halocline waters
Labrador Sea Water
mixed layer depth
North Atlantic
North Atlantic Deep Water
North Atlantic Oscillation
North Pacific
North West Passage
sea surface height
upper halocline waters
West Greenland Current
West Spitzbergen Current
94
Appendix B. A list of symbols
Bibliography
Aargaard, K., and E. C. Carmack (1989), The Role of Sea Ice and Other Fresh
Water in the Arctic Circulation, J. Geophys. Res., 94, 14,485–14,498.
Agnew, T., A. Lambe, and D. Long (2008), Estimating sea ice area flux across
the Canadian Arctic Archipelago using enhanced AMSR-E, J. Geophys.
Res., 113, doi:10.1029/2007JC004582.
Aksenov, Y., S. Bacon, A. C. Coward, and N. P. Holliday (2010), Polar
outflow from the Arctic Ocean: A high resolution model study, Journal of
Marine Systems, 83 (1-2), 14 – 37, doi:10.1016/j.jmarsys.2010.06.007.
Bailey, W. (1957), Oceanographic Features of the Canadian Archipelago,
Journal of the Fisheries Research Board of Canada, 14, 731–769, doi:
10.1139/f57-03.
Böning, C., M. Scheinert, Dengg, Biastoch, and A. Funk (2006), Decadal
variability of subpolar gyre transport and its reverberation in the North
Atlantic overturning, Geophys. Res. Lett., 33, doi:10.1029/2006GL026906.
Chelton, D. (1983), Effects of sampling errors in statistical estimation, DeepSea Research, 30, 1083–1103.
Cheng, W., and P. Rhines (2004), Response of the overturning circulation
to high-latitude fresh-water perturbations in the North Atlantic, Climate
Dynamics, 22, 359–372, doi:10.1007/s00382-003-0385-6.
Condron, A., P. Winsor, C. Hill, and D. Menemenlis (2009), Simulated Response of the Arctic Freshwater Budget to Extreme NAO Wind Forcing,
J. Climate, 22, 2422–2437, doi:10.1175/2008JCLI2626.1.
95
96
Bibliography
Cuny, J., P. Rhines, and R. Kwok (2005), Davis Strait volume, freshwater and
heat fluxes, Deep-Sea Res. I, 52, 519–542, doi:10.1016/j.dsr.2004.10.006.
Curry, B., C. M. Lee, and B. Petrie (2011), Volume, Freshwater, and Heat
Fluxes through Davis Strait, 2004-05, Journal of Physical Oceanography,
41, 429–436, doi:10.1175/2010JPO4536.1.
Curry, R., M. McCartney, and T. Joyce (1998), Oceanic transport of subpolar
climate signals to mid-depth subtropical waters, Nature, 391, 575–577, doi:
10.1038/35356.
Dai, A., T. Qian, K. E. Trenberth, and J. D. Milliman (2009), Changes
in Continental Freshwater Discharge from 1948 to 2004, J. Climate, 22,
2773–2792, doi:10.1175/2008JCLI2592.1.
Danabasoglu, G., S. Yeager, Y.-O. Kwon, J. Tribbia, A. S. Phillips, and
J. Hurrell (2012), Variability of the Atlantic Meridional Overturning Circulation in CCSM4, J. Climate, 25, 5153–5172, doi:10.1175/JCLI-D-1100463.1.
Danilov, S., G. Kivman, and J. Schröter (2004), A finite-element ocean
model: principles and evaluation, Ocean Modelling, 6 (2), 125–150, doi:
10.1016/S1463-5003(02)00063-X.
de Boyer Montégut, C., G. Madec, A. S. Fischer, A. Lazar, and D. Iudicone (2004), Mixed layer depth over the global ocean: An examination of
profile data and a profile-based climatology, J. Geophys. Res., 109, doi:
10.1029/2004JC002378.
de Steur, L., E. Hansen, R. Gerdes, M. Karcher, E. Fahrbach, and J. Holfort
(2009), Freshwater fluxes in the East Greenland Current: A decade of
observations, Geophys. Res. Lett., 36, doi:10.1029/2009GL041278.
Dickson, R., T. Osborn, J. Hurrell, J. Meincke, J. Blindheim, B. Adlandsvik,
T. Vinje, G. Alekseev, and W. Maslowski (2000), The Arctic Ocean Response to the North Atlantic Oscillation, J. Clim., 13, 2671–2696.
Bibliography
97
Dickson, R., B. Rudels, S. Dye, M. Karcher, J. Meincke, and I. Yashayaev
(2007), Current estimates of freshwater flux through Arctic and subarctic
seas, Progress in Oceanography, 73, 210–230.
Eden, C., and T. Jung (2001), North Atlantic Interdecadal Variability:
Oceanic Response to the North Atlantic Oscillation, J. Climate, 14, 676–
691, doi:10.1175/1520-0442(2001)014¡0676:NAIVOR¿2.0.CO;2.
Fetterer, F., K. Knowles, W. Meier, and M. Savoie (2009), Sea Ice Index,
Digital media, boulder, CO: National Snow and Ice Data Center.
Fissel, D., D. Lemon, and J. Birch (1982), Major Features of the Summer
Near-Surface Circulation of Western Baffin Bay, 1978 and 1979, Arctic, 35,
180–200.
Gent, P., and J. McWilliams (1990), Isopycnal Mixing in Ocean Circulation
Models, J. Phys. Oceanogr., 20, 150–155.
Gill, A. (1982), Atmosphere-Ocean Dynamics, Academic Press, New York.
662 pp.
Goosse, H., T. Fichefet, and J.-M. Campin (1997), The effects of the water flow through the Canadian Archipelago in a global ice-ocean model,
Geophysical Research Letters, 24, 1507–1510, doi:10.1029/97GL01352.
Griffies, S. (1998), The Gent-McWilliams Skew Flux, J. Phys. Oceanogr., 28,
831–841, doi:10.1175/1520-0485(1998)028¡0831:TGMSF¿2.0.CO;2.
Hannah, C., F. Dupont, and M. Dunphy (2009), Polynyas and Tidal Currents
in the Canadian Arctic Archipelago, Arctic, 62, 83–95.
Hibler, W. (1979), A Dynamic Thermodynamic Sea Ice Model, J. Phys.
Oceanogr., 9, 815–846.
Hibler, W. (1984), The role of sea ice dynamics in modeling CO2 increases, in
Climate Processes and Climate Sensitivity, edited by J. E. Hansen and T.
Takahashi, pp. 238–253, Geophys. Monogr. Ser., AGU, Washington D.C.,
doi:10.1029/GM029p0238.
98
Bibliography
Houssais, M.-N., and C. Herbaut (2011), Atmospheric forcing on
the Canadian Arctic Archipelago freshwater outflow and implications for the Labrador Sea variability, J. Geophys. Res., 116, doi:
10.1029/2010JC006323.
Hunke, E., and J. Dukowicz (1997), An Elastic-Viscous-Plastic Model for
Sea Ice Dynamics, Journal of Physical Oceanography, 27, 1849–1867.
Hunke, E., and J. Dukowicz (2001), The Elastic-Viscous-Plastic Sea Ice Dynamics Model in General Orthogonal Curvilinear Coordinates on a SphereIncorporation of Metric Term, Monthly Weather Review, 130, 1848–1865.
Jackett, D., and T. Mcdougall (1995), Minimal Adjustment of Hydrographic
Profiles to Achieve Static Stability, J. Atmos. Oceanic Technol., 12, 381–
389, doi:10.1175/1520-0426(1995)012¡0381:MAOHPT¿2.0.CO;2.
Jahn, A., B. Tremblay, L. Mysak, and R. Newton (2009), Effect of the largescale atmospheric circulation on the variability of the Arctic Ocean freshwater export, Clim. Dyn., 34, 201–222, doi:10.1007/s00382-009-0558-z.
Jahn, A., Y. Aksenov, B. A. de Cuevas, L. de Steur, S. Häkkinen, E. Hansen,
C. Herbaut, M.-N. Houssais, M. Karcher, F. Kauker, C. Lique, A. Nguyen,
P. Pemberton, D. Worthen, and J. Zhang (2012), Arctic Ocean freshwater: How robust are model simulations?, J. Geophys. Res., 117, doi:
10.1029/2012JC007907.
Jakobsson, M., R. Macnab, L. Mayer, R. Anderson, M. Edwards, J. Hatzky,
H.-W. Schenke, and P. Johnson (2008), An improved bathymetric portrayal of the arctic ocean: Implications for ocean modeling and geological, geophysical and oceanographic analyses, Geophysical Research Letters,
35 (L07602), doi:10.1029/2008GL033520.
Joyce, T., and A. Proshutinsky (2007), Greenland’s Island Rule and the
Arctic Ocean circulation, Journal of Marine Research, 65, 639–653.
Bibliography
99
Kliem, N., and D. A. Greenberg (2003), Diagnostic Simulations of the Summer Circulation in the Canadian Arctic Archipelago, Atmosphere-Ocean,
41, 273–289.
Köberle, C., and R. Gerdes (2007), Simulated Variability of the Arctic Ocean
Freshwater Balance 1948-2001, J. Phys. Oceanogr., 37, 1628–1644, doi:
10.1175/JPO3063.1.
Komuro, Y., and H. Hasumi (2005), Intensification of the Atlantic Deep
Circulation by the Canadian Archipelago Throughflow, J. Phys. Oceanogr.,
35, 775–789.
Kuhlbrodt, T., A. Griesel, M. Montoya, A. Levermann, M. Hofmann, and
S. Rahmstorf (2007), On the driving processes of the Atlantic meridional
overturning circulation, Rev. Geophys., 45, doi:10.1029/2004RG000166.
Kwok, R. (2005), Variability of Nares Strait ice flux, Geophys. Res. Lett., 32,
doi:10.1029/2005GL024768.
Kwok, R., and G. F. Cunningham (2008), ICESat over Arctic sea ice: Estimation of snow depth and ice thickness, J. Geophys. Res., 113, doi:
10.1029/2008JC004753.
Large, W., and S. Yeager (2004), Diurnal to decadal global forcing for ocean
and sea-ice models: The data sets and flux climatologies, Tech. rep., Technical Report TN-460+STR, NCAR.
Large, W., and S. Yeager (2008), The global climatology of an interannually varying air-sea flux data set, Climate Dynamics, 33, 341–364, doi:
10.1007/s00382-008-0441-3.
Latif, M., C. Böning, J. Willebrand, A. Biastoch, J. Dengg, N. Keenlyside,
U. Schweckendiek, and G. Madec (2006), Is the thermohaline circulation
changing?, Journal of Climate, 19, 4631–4637, doi:10.1175/JCLI3876.1.
Lazier, J. (1980), Oceanographic conditions at Ocean Weather Ship Bravo,
Atmosphere-Ocean, 18, 227–238, doi:10.1080/07055900.1980.9649089.
100
Bibliography
Lazier, J., R. Hendry, A. Clarke, I. Yashayaev, and P. Rhines (2002), Convection and restratification in the Labrador Sea, 1990-2000, Deep-Sea Research I, 49, 1819–1835.
Lietaer, O., T. Fichefet, and V. Legat (2008), The effects of resolving the
Canadian Arctic Archipelago in a finite element sea ice model, Ocean Modelling, 24 (3–4), 140 – 152, doi:10.1016/j.ocemod.2008.06.002.
Lique, C., A. M. Treguier, M. Scheinert, and T. Penduff (2009), A modelbased study of ice and freshwater transport variability along both sides of
Greenland, Climate Dynamics, 33, 685–705, doi:10.1007/s00382-008-05107.
Lique, C., A. M. Treguier, B. Blanke, and N. Grima (2010), On the origins
of water masses exported along both sides of Greenland: A Lagrangian
model analysis, J. Geophys. Res., 115, doi:10.1029/2009JC005316.
Lohmann, K., H. Drange, and M. Bentsen (2008), Response of the North
Atlantic subpolar gyre to persistent North Atlantic oscillation like forcing,
Clim. Dyn., 32, 273–285, doi:10.1007/s00382-008-0467-6.
Lu, Y., S. Nudds, F. Dupont, M. Dunphy, C. Hannah, and S. Prinsenberg
(2010), High-resolution Modelling of Ocean and Sea-ice Conditions in the
Canadian Arctic Coastal Waters, in Proceedings of the 20th International
Offshore and Polar Engineering Conference, beijing, China. 20-26 June
2010.
McGeehan, T., and W. Maslowski (2011), Impact of shelf-basin freshwater transport on deep convection in the western Labrador Sea, J. Phys.
Oceanogr., 41, 2187–2210, doi:10.1175/JPO-D-11-01.1.
McGeehan, T., and W. Maslowski (2012), Evaluation and control mechanisms of volume and freshwater export through the Canadian Arctic
Archipelago in a high-resolution pan-Arctic ice-ocean model, J. Geophys.
Res., 117, doi:10.1029/2011JC007261.
Bibliography
101
Melling, H. (2000), Exchanges of freshwater through the shallow straits of
the North American Arctic, in The Freshwater Budget of the Arctic Ocean,
edited by E. L. Lewis et al., pp. 479–502, Springer, New York.
Melling, H. (2002), Sea ice of the northern Canadian Arctic Archipelago, J.
Geophys. Res., 107, doi:10.1029/2001JC001102.
Melling, H., T. A. Agnew, K. K. Falkner, D. A. Greenberg, C. M. Lee,
A. Münchow, B. Petrie, S. J. Prinsenberg, R. M. Samelson, and R. A.
Woodgate (2008), Fresh-water fluxes via Pacific and Arctic outflows across
the Canadian polar shelf, in Arctic-Subarctic Ocean Fluxes: Defining the
Role of the Northern Seas in Climate, edited by R. R. Dickson et al., pp.
193–247, Springer.
Münchow, A., and H. Melling (2008), Ocean current observations from Nares
Strait to the west of Greenland: Interannual to tidal variability and forcing,
J. Mar. Res., 66, 801–833.
Myers, P., C. Donnelly, and M. Ribergaard (2009), Structure and variability of the West Greenland Current in Summer derived from 6 repeat standard sections, Progress in Oceanography, 80, 93–112, doi:
10.1016/j.pocean.2008.12.003.
Myers, P. G. (2005), Impact of freshwater from the Canadian Arctic
Archipelago on Labrador Sea Water formation, Geophys. Res. Lett., 32,
doi:10.1029/2004GL022082.
Pacanowski, R., and S. Philander (1981), Parameterization of Vertical Mixing
in Numerical Models of Tropical Oceans, J. Geophys. Res., 11, 1443–1451.
Parkinson, C., and W. Washington (1979), A Large-Scale Numerical Model
of Sea Ice, J. Geophys. Res., 84, 311–337.
Peterson, I., J. Hamilton, S. Prinsenberg, and R. Pettipas (2012), Wind
Forcing of Volume Transport through Lancaster Sound, J. Geophys. Res.,
in press., doi:10.1029/2012JC008140.
102
Bibliography
Prinsenberg, S., and E. Bennet (1989), Transport between Peel Sound and
Barrow Strait in the Canadian Arctic, Continental Shelf Research, 9, 427–
444.
Prinsenberg, S., and J. Hamilton (2005), Monitoring the Volume, Freshwater
and Heat Fluxes Passing through Lancaster Sound in the Canadian Arctic
Archipelago, Atmosphere-Ocean, 43, 1–22.
Prinsenberg, S., J. Hamilton, I. Peterson, and R. Pettipas (2009), Observing
and interpreting the seasonal variability of the oceanographic fluxes passing
through Lancaster Sound of the Canadian Arctic Archipelago, in Influence
of Climate Change on the Changing Arctic and Sub-Arctic Conditions,
edited by J. Nihoul, pp. 125–143, Springer.
Proshutinsky, A., R. H. Bourke, and F. A. McLaughlin (2002), The role of the
Beaufort Gyre in Arctic climate variability: Seasonal to decadal climate
scales, Geophys. Res. Lett., 29, doi:10.1029/2002GL015847.
Rabe, B., M. Karcher, U. Schauer, J. Toole, R. Krishfield, S. Pisarev,
F. Kauker, R. Gerdes, and T. Kikuchi (2011), An assessment of Arctic
Ocean freshwater content changes from the 1990s to the 2006-2008 period,
Deep-Sea Research I, 58, 173–185, doi:10.1016/j.dsr.2010.12.002.
Redi, M. (1982), Oceanic Isopycnal Mixing by Coordinate Rotation, J. Phys.
Oceanogr., 12, 1154–1158.
Rothrock, D., and J. Zhang (2005), Arctic Ocean sea ice volume:
What explains its recent depletion?, J. Geophys. Res., 110, doi:
10.1029/2004JC002282.
Schauer, U., A. Beszczynska-Moeller, W. Walczowski, E. Fahrbach,
J. Piechura, and E. Hansen (2008), Variation of Measured Heat Flow
Through the Fram Strait Between 1997 and 2006, in Arctic-Subarctic
Ocean Fluxes: Defining the Role of the Northern Seas in Climate, edited
by R. D. et al., pp. 65–85, Springer.
Bibliography
103
Schweckendiek, U., and J. Willebrand (2005), Mechanisms Affecting the
Overturning Response in Global Warming Simulations, Journal of Climate, 18, 4925–4936, doi:10.1175/JCLI3550.1.
Semtner, A. (1976), A Model for the Thermodynamic Growth of Sea Ice
in Numerical Investigations of Climate, J. Phys. Oceanogr., 6, 379–389,
doi:10.1175/1520-0485(1976)006¡0379:AMFTTG¿2.0.CO;2.
Serreze, M. C., A. P. Barrett, A. G. Slater, R. A. Woodgate, K. Aagaard,
R. B. Lammers, M. Steele, R. Moritz, M. Meredith, and C. M. Lee (2006),
The large-scale freshwater cycle of the Arctic, J. Geophys. Res., 111, doi:
10.1029/2005JC003424.
Shewchuk, J. (1996), Triangle: Engineering a 2D Quality Mesh Generator
and Delaunay Triangulator, first Workshop on Applied Computational Geometry (Philadelphia, PA), 124-133, ACM, May 1996.
Sidorenko, D., Q. Wang, S. Danilov, and J. Schröter (2011), FESOM under
Coordinated Ocean-ice Reference Experiment forcing, Ocean Dynamics,
61, 881–890, doi:10.1007/s10236-011-0406-7.
Skagseth, O., T. Furevik, R. Ingvaldsen, H. L. K. Mork, K. Orvik, and
V. Ozhigin (2008), Volume and Heat Transports to the Arctic Ocean Via
the Norwegian and Barents Seas, in Arctic-Subarctic Ocean Fluxes: Defining the Role of the Northern Seas in Climate, edited by R. R. Dickson
et al., pp. 45–64, Springer.
Smith, G., F. Roy, P. Mann, F. Dupont, B. Brasnett, J.-F. Lemieux,
S. Laroche, and S. Bélair (2013), A New Atmospheric Data Set for Forcing
Ice-Ocean Models: Evaluation of Canadian GDPS Reforecasts, Q. J. R.
Met. S., submitted manusscript.
Smith, L., and S. Stephenson (2013), New Trans-Arctic shipping routes navigable by midcentury, PNAS, 110, 4855–4856, doi:10.1073/iti1313110.
104
Bibliography
Steele, M., R. Morley, and W. Ermold (2001), PHC: a global ocean hydrography with a high-quality Arctic Ocean, J. Climate, 14, 2079–2087,
doi:10.1175/1520-0442(2001)014¡2079:PAGOHW¿2.0.CO;2.
Straneo, F. (2006), On the Connection between Dense Water Formation,
Overturning, and Poleward Heat Transport in a Convective Basin, J. Phys.
Oceanogr., 36, 1822–1840, doi:10.1175/JPO2932.1.
Terwisscha van Scheltinga, A., P. G. Myers, and J. D. Pietrzak (2010), A
finite element sea ice model of the Canadian Arctic Archipelago, Ocean
Dynamics, 60 (6), 1539–1558, doi:10.1007/s10236-010-0356-5.
Timmermann, R., A. Beckmann, and H. Hellmer (2002), Simulations of iceocean dynamics in the Weddell Sea. 1. Model configuration and validation,
J. Geophys. Res, 107 (C33024).
Timmermann, R., S. Danilov, J. Schröter, C. Böning, D. Sidorenko, and
K. Rollenhagen (2009), Ocean circulation and sea ice distribution in a
finite element global sea ice-ocean model, Ocean Modelling, 27 (3-4), 114 –
129, doi:10.1016/j.ocemod.2008.10.009.
Wadley, M., and G. Bigg (2002), Impact of flow through the canadian
archipelago and bering strait on the north atlantic and arctic circulation:
An ocean modelling study, Q.J.R. Meteorol. Soc., 128, 2187–2203,, doi:
10.1256/qj.00.35.
Walker, E. (1977), Aspects of Oceanography in the Archipelago, Sidney, B.C.:
Institute of Ocean Sciences, Patricia Bay.
Wang, Q., S. Danilov, and J. Schröter (2008), Finite element ocean circulation model based on triangular prismatic elements, with application
in studying the effect of topography representation, J. Geophys. Res.,
113 (C05015), doi:10.1029/2007JC004482.
Wang, Q., P. Myers, X. Hu, and A. Bush (2012a), Flow constraints
on pathways through the Canadian Arctic Archipelago, Atm.-Oce., doi:
10.1080/07055900.2012.704348.
Bibliography
105
Wang, X., Q. Wang, D. Sidorenko, S. Danilov, J. Schröter, and T. Jung
(2012b), Long term ocean simulations in FESOM: Validation and application in studying the impact of Greenland Ice Sheet melting, Ocean Dynamics, doi:10.1007/s10236-012-0572-2.
Wilson, K., J. Falkingham, H. Melling, and R. D. Abreu (2004), Shipping in
the Canadian Arctic: Other Possible Climate Change Scenarios, in Geoscience and Remote Sensing Symposium, 2004. IGARSS ’04. Proceedings.,
vol. 3, pp. 1853–1856.
Woodgate, R. A., T. J. Weingartner, and R. Lindsay (2010), The 2007 Bering
Strait oceanic heat flux and anomalous Arctic sea-ice retreat, Geophys.
Res. Lett., 37 (L01602), doi:10.1029/2009GL041621.
106
Bibliography
Acknowledgment
First of all, I would like to thank my supervisor Dr. Qiang Wang for his
invaluable help during my PhD studies. Thanks also to Dr. Sergey Danilov
for his great support. I would like to thank Prof. Thomas Jung and Dr.
Jens Schröter for their great support and for giving me the opportunity to
work on my PhD thesis in their research department. Thanks to Prof. Peter
Lemke for his encouragement and for being a reviewer of my thesis. Thanks
a lot to Dr. Dmitry Sidorenko, Dr. Sven Harig and Dr. Martin Losch for
their help and support.
I would like to thank Xuezhu Wang for her suggestions and help. I would
also like to thank Verena Haid, Vibe Schourup-Kristensen, Madlen Gebler,
Konstantin Korchuk, Madlen Kimmritz, Leander Wachsmuth and my other
colleagues from the Climate Dynamics group for making it a such a pleasure
working at AWI.
Great thanks goes to Prof. Paul Myers (University of Alberta) for giving
me the opportunity to study in Edmonton for two months. It was a great experience, and helpful discussions with him and Xianmin Hu greatly improved
my work.
I gratefully acknowledge Dr. Simon Prinsenberg (Bedford Institute of
Oceanography) and Dr. Andreas Münchow (University of Delaware) for providing me with observational data.
I am very grateful to Dr. Agnieszka Beszczynska-Möller for giving me
the chance to participate in the Polarstern expedition ARK XXVI/1 to Fram
Strait through which I gained knowledge in operational oceanography.
I am very grateful to my family, and especially Özgür, for their support,
patience and care.
107
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement