Is realistic Antarctic sea ice extent in climate models the result of excessive ice drift?

34 Pages·2014·3.83 MB·English

by  P. Uotila

Accepted Manuscript
Is realistic Antarctic sea ice extent in climate models the result of excessive ice
drift?
P. Uotila, P.R. Holland, T. Vihma, S.J. Marsland, N. Kimura
PII: S1463-5003(14)00052-3
DOI: http://dx.doi.org/10.1016/j.ocemod.2014.04.004
Reference: OCEMOD 892
To appear in: Ocean Modelling
Received Date: 28 October 2013
Revised Date: 10 April 2014
Accepted Date: 21 April 2014
Please cite this article as: Uotila, P., Holland, P.R., Vihma, T., Marsland, S.J., Kimura, N., Is realistic Antarctic sea
ice extent in climate models the result of excessive ice drift?, Ocean Modelling (2014), doi: http://dx.doi.org/
10.1016/j.ocemod.2014.04.004
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Is realistic Antarctic sea ice extent in climate models
the result of excessive ice drift?
P. Uotilaa,b,, P.R. Hollandc, T. Vihmab, S.J. Marslanda, N. Kimurad
aCSIRO–Marine & Atmospheric Research, Aspendale, Australia
bFinnish Meteorological Institute, Helsinki, Finland
cBritish Antarctic Survey, Cambridge, the UK
dNational Institute of Polar Research, Tachikawa, Japan

Abstract
Forthefirsttime,wecomputethesea-iceconcentrationbudgetofafullycoupled
climate model, the Australian ACCESS model, in order to assess its realism in
simulating the autumn–winter evolution of Antarctic sea-ice. The sea-ice con-
centration budget consists of the local change, advection and divergence, and
the residual component which represents the net effect of thermodynamics and
ridging. Although the model simulates the evolution of sea-ice area reasonably
well, its sea-ice concentration budget significantly deviates from the observed
one. Themodelledsea-icebudgetcomponentsdeviatefromobservedclosetothe
Antarcticcoast,wherethemodelledicemotionismoreconvergent,andnearthe
ice edge, where the modelled ice is advected faster than observed due to incon-
sistencies between ice velocities. In the central ice pack the agreement between
the model and observations is better. Based on this, we propose that efforts to
simulate the observed Antarctic sea-ice trends should focus on improving the
realism of modelled ice drift.
Keywords: thermodynamics, divergence, advection
∗Correspondingauthor.
Email address: [email protected](P.Uotila)
Preprint submitted to Ocean Modelling April 26, 2014
1. Introduction
1
2 The Antarctic sea ice is expanding and climate models have difficulties
3 in simulating this trend (Turner et al., 2013a), for yet unknown reasons. A
4 small number of climate model simulations, however, show a similar increase of
5 Antarctic sea ice extent to the observed one which may indicate that the inter-
6 nal variability of the climate system, rather than forcing due to greenhouse gas
7 concentrations, plays a significant role (Zunz et al. , 2013). This hypothesis is
8 supportedbyMahlsteinetal.(2013),whostudiedAntarcticsea-iceareaderived

9 from a large ensemble of 23 climate models and found that the internal sea-ice
10 variabilityislargeintheAntarcticregionindicatingthatboththeobservedand
11 modelled trends can represent natural variations along with external forcings.
12 Moreover, Polvani and Smith (2013) analysed forced and preindustrial control
13 model simulations of four climate models to see whether their Antarctic sea-ice
14 trends are due to the internal variability or not. They found that the observed
15 Antarctic trend falls within the distribution of trends arising naturally from
16 the coupled atmosphere–ocean–sea ice system and concluded that it is difficult
17 to attribute the observed trends to anthropogenic forcings. Consistent with
18 Polvani and Smith (2013), Swart and Fyfe (2013) show that when accounting
19 for internal variability, an average multi-model sea-iceareatrendisstatistically
20 compatible with the observed trend.
21 However, the validity of the hypothesis that the Antarctic sea-ice increase
22 is due to the internal variability of the climate system remains uncertain be-
23 cause the models used to test the hypothesis show biases in the mean state and
24 regional patterns, and overestimate the interannual variance of sea-ice extent,
25 particularly in winter (Zunz et al. , 2013). To confirm the argument of natural
26 variability,amodelwouldhavetoexplaintheobservedsea-iceincreasewhilesi-
27 multaneously responding to anthropogenic forcings. Hence, it appears that the
28 modelscannotbeusedtotestpreciselywhethertheobservedseaiceexpansion
29 is due to the internal variability of the climate system or not.
30 In addition to the above mentioned model based studies, a recent observa-
2
31 tionalstudysupportstosomeextenttheargumentofinternalvariability. Meier
32 et al. (2013) analysed satellite data and showed that the Antarctic sea-ice ex-
33 tent in 1964 was larger than anytime during 1979–2012. This is a robust result,
34 because within the wide range of uncertainty in the 1964 satellite estimate, the
35 1964iceextentishigherthanthemonthlySeptemberaverageofanyoftheyears
36 of the satellite record from 1979–2012 and remains on the highest end of the
37 estimates even when taking into consideration the variation within the month.
38 AccordingtoMeieretal.(2013),theicecovermaycurrentlyberecoveringfrom
39 a relatively low level back to higher conditions seen in the 1960s. Hence, this
40 result suggests that the current 33 year increase in the sea-ice extent is due to
41 the long-term variability of the climate system. Whether this long-term vari-
42 ability is only due to the internal variability or due to the combined effects of
43 forcings and the internal variability remains unclear.
44 Observations can also be used to show that the Antarctic sea ice concentra-
45 tion trends are closely associated with trends in ice drift or with trends related
46 to thermodynamics (Holland and Kwok, 2012). The observed Antarctic sea-ice
47 drift trends can be explained by changes in local winds and the aspects of local
48 winds can be attributed to large-scale atmospheric circulation modes (Uotila et
49 al. , 2013b), which have experienced significant changes in the last thirty years
50 (Solomonetal.,2007;Turneretal.,2013b). Moreover,HollandandKwok(2012)
51 show where the evolution of Antarctic sea ice is controlled either by thermo-
52 dynamic or dynamic processes during its autumnal expansion and in winter.
53 This is particularly valuable because the relatively weak overall Antarctic sea
54 ice trend consists of strong regional but opposing trends (Turner et al., 2009).
55 HollandandKwok(2012)suggestthat,bycomparingtheirobservationalresults
56 with similarly processed climate model output, one can diagnose faults in a cli-
57 mate model due to thermodynamic or dynamic processes when simulating the
58 Antarctic sea ice. This is the motivation of our study — to investigate whether
59 a fully coupled climate model produces realistic contributions from thermody-
60 namic and dynamic sea-ice evolution. In this way we should be able to address
61 whichprocessesinthemodelaretoopoorlyrepresentedtorealisticallysimulate
3
62 the currently observed sea-ice state, its variability and its trends. Results from
63 such an analysis have not yet been published.
64 Related to this, recent studies have shown that coupled ocean–ice models,
65 whereatmosphericstatesareprescribed, canreproduceobservedAntarcticsea-
66 icetrendsunderrealisticatmosphericforcingand/orwhentheyareconstrained
67 withobservations. Massonnetetal.(2013)assimilatedseaiceconcentrationinto
68 anocean–icemodeltogenerateAntarcticsea-icevolumetimeseriesfrom1980–
69 2008. Additionally,Zhang(2013)showsbyanocean–icemodelthatintensifying
70 winds result in increase in sea ice speed, convergence and sea-ice deformation .
71 Thesea-icedeformationincreasesthevolumeofthickiceintheocean–icemodel
72 along with a significant sea-ice concentration increase in the Southern Weddell
73 Sea. Importantly, Holland et al. (2014, submitted) show that a free-running
74 ocean–icemodelforcedbyatmosphericre-analysescanreproduceAntarcticsea-
75 ice concentration and drift trends as observed. Hence, atmospheric states of a
76 fully coupled climate model seem crucial for the modelled sea-ice trends. Ac-
77 cordingly, an assessment of the thermodynamic and dynamic processes related
78 to the evolution of sea-ice concentration in a fully coupled climate model is an
79 important next step to understand why climate models have not been able to
80 simulate Antarctic sea ice realistically.
81 We hypothesise that climate models simulate the seasonal evolution of inte-
82 grated Antarctic sea-ice area, and integrated extent, reasonably well, even with
83 relatively unrealistic dynamic and thermodynamic components of the sea-ice
84 concentration budget, partly due to the balancing of biases of these compo-
85 nents. For example, during its autumnal expansion sea ice is advected over a
86 larger area when its speed is higher, but at the same time it melts more at
87 the northernmost ice edge where the ocean and atmosphere are warm and the
88 thermodynamics limits the dynamical expansion of sea ice. In order to produce
89 observed regional sea-ice concentration trends in decadal time scales, and the
90 overallsea-iceareaorextenttrendsfortherightreasons,andthereforewiththe
91 correct mass, energy and momentum fluxes, climate models need to simulate
92 regional dynamical and thermodynamical processes correctly.
4
93 To test the success of our hypothesis, we compare modelled dynamic and
94 thermodynamic components of the Antarctic April–October sea-ice concentra-
95 tionbudgetasderivedfromtheoutputofawellperformingstate-of-the-science
96 climate model with the observed budget of Holland and Kwok (2012). The ob-
97 served sea-ice concentration budget data of Holland and Kwok (2012) is only
98 available from April to October which limits our analysis to these months. We
99 presentthemodels,methodsanddatausedforthisanalysisinthenextsection.
100 In the Results and Discussion section, we compare modelled sea-ice concentra-
101 tion budgets with observed ones and discuss how their differences affect the
102 sea-ice evolution. Finally, in the last section we present the main conclusions of
103 this study along with their implications.
2. Methods and data
104
Table1: Modelexperimentsusedinthisstudy.
Name Years Short description and reference.
historical 1850–2005 Historicalsimulationsthatuseevolvingforcingsuchasvolcanoes,
aerosols, greenhouse gas concentrations and land use changes
(Taylor et al., 2012).
rcp85 2005–2100 A future projection simulation forced with specified concentra-
tions(RCPs),consistentwithahighemissionsscenario(Tayloret
al., 2012).
CORE-II IAF 1948–2007 The second phase of The Coordinated Ocean-ice Reference Ex-
periments (COREs) that uses inter-annually varying prescribed
atmospheric forcing (IAF) of Large and Yeager (2009) under the
experimental protocols introduced in Danabasoglu et al. (2014).
105 Weanalysedatafromfourhistorical andonercp85 realisationsimulatedby
106 theAustralianCommunityClimateandEarth-SystemSimulatorcoupledmodel
107 version 1.0 (ACCESS1.0) and 1.3 (ACCESS1.3) as submitted to the phase five
108 of the Coupled Model Inter-comparison project (CMIP5) database (Table 1,
109 Figure 1 and Dix et al., 2013). ACCESS1.0 and ACCESS1.3 differ in two
110 important aspects: their sea-ice albedos are different and their atmospheric
111 cloudmicrophysicsschemesaredifferent. Boththesedifferencescanbeexpected
112 to affect the sea-ice performance. Therefore we wanted to see how much their
5
Figure 1: Horizontal bars illustrate total time extent of model simulations and observations
usedinthisstudy. Timeperiodsselectedfortheanalysisarehighlightedwithnon-transparent
colours with the start and end years written, while time periods excluded from the analysis
areshownwithtransparent,faintercolours.
113 sea-ice concentration budgets differ. The ACCESS configurations are one of
114 the better performing CMIP5 models in terms of global sea-ice extent with a
115 climatology relatively close to the observed one (Uotila et al., 2013a; Liu et al.,
116 2013), thus justifying its selection for this study.
117 Moreover, similar analysis as for the ACCESS coupled model (ACCESS-
118 CM; Bi et al., 2013a) output, are carried out for the output from an ACCESS
119 ocean–sea ice model (ACCESS-OM; Bi et al., 2013b) simulation forced with
120 prescribedatmosphericconditionsandbulkformulaeofLargeandYeager(2009)
121 followingtheCoordinatedOcean-iceReferenceExperimentphase2Inter-annual
122 Forcing (CORE-II IAF) protocols as described in Griffies et al. (2012) (Table
123 1). Following Danabasoglu et al. (2014), we use the fifth cycle of a CORE-II
124 IAF simulation for the analysis of ACCESS-OM presented here. Note that the
6
125 ACCESS-OM simulation ends in 2007 which is the last year of CORE-II IAF.
126 The ACCESS-CM and ACCESS-OM configurations share the ocean and
127 sea-ice models and by analysing their differences we can assess the role of the
128 prescribed atmospheric forcing in driving changes in the Antarctic sea-ice con-
129 centration. The sea-ice model of ACCESS is the LANL Community Ice CodE
130 version 4.1 Hunke and Lipscomb (2010), which uses the elastic-viscous-plastic
131 rheology, and the ocean model is an implementation of the 2009 public release
132 of the NOAA/GFDL MOM4p1 community code (Griffies et al., 2009). Both
133 ACCESS-CMandACCESS-OMuseanidenticalhorizontaldiscretisationonan
134 orthogonal curvilinear tripolar grid with a nominal one degree resolution hav-
135 ing additional refinements in the Arctic, in the Southern Ocean, and near the
136 Equator. The ACCESS-CM atmospheric model has a horizontal resolution of
◦ ◦
137 1.25 latitude by 1.875 longitude. ACCESS-OM is forced by CORE forcing
◦
138 withsphericalT62resolution(approximately1.9 ),althoughmanymeteorolog-
139 ical variables, such as winds, are based on the NCEP/NCAR reanalysis with a
140 coarser horizontal resolution of 2.5◦ latitude × 2.5◦ longitude.
141 There is a significant difference in the computation of sea-ice surface en-
142 ergy balance between ACCESS-CM and ACCESS-OM. As described in Bi et
143 al. (2013a) ACCESS-CM has a semi-implicit atmospheric boundary layer that
144 requires determination of the surface heat flux using a zero-layer thermody-
145 namic calculation following Semtner (1976). In contrast, ACCESS-OM uses a
146 4-layer sea-ice thermodynamic discretisation that allows for a more realistic in-
147 ternal sea-ice temperature profile. In the multi-layer thermodynamic approach
148 (ACCESS-OM), the sea-ice temperatures and net top and basal surface heat
149 fluxes are together calculated iteratively, with a heat capacity that depends on
150 internal material properties. The simpler zero-layer approach (ACCESS-CM)
151 only accounts for top and basal sea-ice temperatures and assumes a linear in-
152 ternal sea-ice temperature profile with no heat capacity. As shown by Cheng
153 et al. (2008), an increased number of sea ice layers results in more realistic sea-
154 icethermodynamics. Despitethisdifference,havingbothACCESS-CMCMIP5
155 and ACCESS-OM CORE-II simulations available is clearly an asset for our
7
156 evaluation that is not available for many climate models.
Following Holland and Kwok (2012), we compute April–October (from 1
April to 31 October) daily sea-ice concentration budgets for ACCESS-CM real-
isations and for the ACCESS-OM experiment as,
∂A
+u·∇A+A∇·u=f −r, (1)
∂t
157 based on daily sea-ice concentration (A) and velocity (u). The concentration
158 change from freezing minus melting (f), and the concentration change from

159 mechanical ice redistribution processes (r), such as ridging and rafting, are
160 resolved as a residual component (f −r). In general, and in the Antarctic in
161 particular,wherethesea-icedrifttendstobedivergent,themagnitudeoff can
162 be expected to be much larger than that of r.
Next, daily sea-ice concentration budgets are integrated over the April–
October period for each year. The integral of the first term from the left in
(1) provides the net change in the sea-ice concentration from the beginning to
end of the period. The integral of the second term in (1) is the contribution
to the sea-ice concentration change by the advection, the integral of the third
term is the contribution by the divergence and the integral on the right hand
side is the net contribution by the thermodynamic and ridging processes. After
reorganising, the integrated ice concentration budget can be represented as,
t2 t2 t2 t2
∂A
dt=− u·∇Adt− A∇·udt+ (f −r)dt, (2)
(cid:2) ∂t (cid:2) (cid:2) (cid:2)
t1 t1 t1 t1
where we denote the term on the left hand side of (2) as difference or dadt;
the first term on the right hand side as advection or adv; the second term
as divergence or div; and the third term as residual or res. Accordingly, the
integrated budget and its components can be expressed compactly as
dadt=adv+div+res. (3)
8
163 It is important to understand that the three components on the right hand
164 sideof(3)areinterdependentand,forexample,regionsexperiencinglargerates
165 of divergence are likely to experience ice growth under cold atmospheric condi-
166 tions. Anotherexamplewouldbeacasewheretheicemeltdecreasesthesea-ice
167 concentrationandthickness,andconsequentlyresultsinafastermovingseaice,
168 which in turn affects the divergence and advection.
169 Finally, integrated components of sea-ice concentration budget are used to
170 computetheiraveragevaluesover19-yearperiodsof1992–2010(ACCESS-CM)
171 and 1989–2007 (ACCESS-OM). These periods were selected because they are
172 ascloseaspossibletotheobservationalresultscovering1992–2010,whichisthe
173 longest period with reliable sea-ice concentration budget observations available
174 (Holland and Kwok, 2012). The observed sea-ice concentration budget was cal-
175 culated on a 100×100 km2 grid, which has a resolution close to the ACCESS
176 model grid (nominally 1◦latitude × 1◦longitude). Following Holland and Kwok
177 (2012), we apply a low pass filter, where every grid point is replaced by the
178 mean value of a 9-cell square centred on that point, on adv, div, and res in (3)
179 to ensure the comparability of the model output with the observations. Model
180 based results are robust and rather similar with or without the smoothing, but
181 Holland and Kwok (2012) observation based results require smoothing to re-
182 ducegrid-scalenoiseinthederivatives. Notethattocoverthewhole1992–2010
183 period we joined four ACCESS-CM historical simulations, which end in 2005,
184 withthercp85 simulationfrom2006–2010resultinginfourcombinationsoftime
185 series – one combination for ACCESS1.0 and three for ACCESS1.3 (Figure 1).
186 To quantify the similarity between the observed and modelled sea-ice, the nor-
187 malised root-mean-square-error (NRMSE) was computed between the observed
188 andmodelledsea-iceconcentration. Wealsocomparethemodelledsea-icearea,
189 computed as the area integral of ice concentration, with the sea-ice area based
190 on observational HadISST data (Rayner et al., 2003), and we assess the agree-
191 ment of modelled ice drift with a 2003–2010 ice velocity climatology computed
192 fromobservationbaseddata(Kimuraetal.,2013). Kimuraetal.(2013)havere-
193 centlypublishedadaily icevelocity productona37.5kmresolution gridwhich
9