Landfast ice thickness in the Canadian Arctic Archipelago from observations and models

13 Pages·2016·2.81 MB·English

TheCryosphere,10,1463–1475,2016
www.the-cryosphere.net/10/1463/2016/
doi:10.5194/tc-10-1463-2016
©Author(s)2016.CCAttribution3.0License.
Landfast ice thickness in the Canadian Arctic Archipelago from
observations and models
StephenE.L.Howell1,FrédéricLaliberté1,RonKwok2,ChrisDerksen1,andJoshuaKing1
1ClimateResearchDivision,EnvironmentandClimateChangeCanada,Toronto,Canada
2JetPropulsionLaboratory,CaliforniaInstituteofTechnology,Pasadena,California,USA
Correspondenceto:StephenE.L.Howell([email protected])
Received:23March2016–PublishedinTheCryosphereDiscuss.:30March2016
Revised:15June2016–Accepted:21June2016–Published:12July2016
Abstract. Observed and modelled landfast ice thickness 1 Introduction
variability and trends spanning more than 5 decades
within the Canadian Arctic Archipelago (CAA) are The World Meteorological Organization (WMO, 1970) de-
summarized. The observed sites (Cambridge Bay, fines landfast sea ice as “sea ice which remains fast along
Resolute, Eureka and Alert) represent some of the thecoast,whereitisattachedtotheshore,toanicewall,to
Arctic’s longest records of landfast ice thickness. Ob- anicefront,orovershoals,orbetweengroundedicebergs”.
served end-of-winter (maximum) trends of landfast In the Arctic, this ice typically extends to the 20–30m iso-
ice thickness (1957–2014) were statistically signif- baths (Mahoney et al., 2007, 2014). It melts each summer
icant at Cambridge Bay (−4.31±1.4cmdecade−1), andre-formsinthefallbutthereareregionsalongthenorth-
Eureka (−4.65±1.7cmdecade−1) and Alert erncoastoftheCanadianArcticArchipelago(CAA)where
(−4.44±1.6cmdecade−1) but not at Resolute. Over multi-yearlandfastice(alsotermedan“iceplug”)ispresent.
the 50+-year record, the ice thinned by ∼0.24–0.26m at The two most prominent regions of multi-year landfast sea
Cambridge Bay, Eureka and Alert with essentially negli- ice in the CAA are located in Nansen Sound and Sverdrup
gible change at Resolute. Although statistically significant Channel (Serson, 1972, 1974) (Fig. 1). It has been docu-
warming in spring and fall was present at all sites, only mentedthaticeremainedintactfrom1963to1998inNansen
low correlations between temperature and maximum ice Sound and from 1978 to 1998 in Sverdrup Channel (Jeffers
thickness were present; snow depth was found to be more et al., 2001; Melling, 2002; Alt et al., 2006). The extreme
strongly associated with the negative ice thickness trends. warmyearof1998disintegratedtheiceinbothregionsand
Comparison with multi-model simulations from Coupled theirsurvivalduringthesummermeltseasoninrecentyears
Model Intercomparison project phase 5 (CMIP5), Ocean hasoccurredlessfrequently(Altetal.,2006).Overtheentire
Reanalysis Intercomparison (ORA-IP) and Pan-Arctic Arctic,landfasticeextentisdecliningat7%decade−1 since
Ice-Ocean Modeling and Assimilation System (PIOMAS) themid-1970s(Yuetal.,2014).
show that although a subset of current generation models Recordsoflandfasticethicknessprovideannualmeasures
havea“reasonable”climatologicalrepresentationoflandfast of ice growth that can also almost entirely be attributed to
ice thickness and distribution within the CAA, trends are atmospheric forcing with negligible deep ocean influence
unrealistic and far exceed observations by up to 2 orders on local ice formation. While the key forcings on landfast
of magnitude. ORA-IP models were found to have positive ice and offshore ice are different, the seasonal behaviour of
correlationsbetweentemperatureandicethicknessoverthe landfast ice can nevertheless provide useful information for
CAA, a feature that is inconsistent with both observations understanding the interannual variability of ice thickness in
andcoupledmodelsfromCMIP5. both regimes. Presently, there is no pan-Arctic network for
monitoring changes in landfast ice but available measure-
ments suggest thinning in recent years. Thickness measure-
ments near Hopen, Svalbard, revealed thinning of landfast
PublishedbyCopernicusPublicationsonbehalfoftheEuropeanGeosciencesUnion.
1464 S.E.L.Howelletal.:LandfasticethicknessintheCAAfromobservationsandmodels
data assimilation systems have become of sufficient spatial
resolution to assess potential landfast ice thickness changes
withintheCAA.
Thisanalysisexaminesthetrendsofmeasuredlandfastice
thickness, snow depth and air temperature over a 50+-year
periodbetween1957and2014andcomparestheresultswith
theearlieranalysisbyBC92.Wethenusethisobservational
foundation to evaluate the representativeness of landfast ice
in state-of-the-art global climate models, assimilation sys-
temsandreanalysisproducts.
Figure1.MapofthecentralCanadianArcticArchipelagoshowing
2 Datadescription
thelocationofthelandfastsnowandthicknessobservations.
2.1 Observations
ice in the Barents Sea region by 11cmdecade−1 between
Landfast ice thickness and corresponding snow depth mea-
1966and2007(Gerlandetal.,2008).Fromacompositetime
surementhavebeenmaderegularlyatmanycoastalstations
series of landfast ice thickness from 15 stations along the
throughout Canada since about 1950. These data are qual-
Siberian coast, Polyakov et al. (2010) estimate an average
itycontrolledandarchivedattheCanadianIceService(CIS)
rate of thinning of 3.3cmdecade−1 between the mid-1960s
and represent one of the few available sources of continu-
andearly2000s.RelativelyrecentobservationsbyMahoney
ous ice thickness measurements in the Arctic. In general,
et al. (2007) and Druckenmiller et al. (2009) found longer
thickness measurements are taken once per week, starting
ice-freeseasonsandthinnerlandfasticecomparedtoearlier
after freeze-up when the ice is safe to walk on and contin-
records.
uing until breakup or when the ice becomes unsafe. Com-
At four sites in the CAA, Brown and Cote (1992) (here-
plete details of this data set are provided by Brown and
inafter, BC92) provided the first examination of the inter-
Cote (1992). The data set is available on the CIS web site
annual variability of end-of-winter (maximum) landfast ice
(http://www.ec.gc.ca/glaces-ice/, see Archive followed by
thickness and associated snow depth over the period 1957–
IceThicknessData).FoursitesintheCAAwereselectedfor
1989. Their results highlighted the insulating role of snow
study:Alert,Eureka,ResoluteandCambridgeBay(Fig.1).
cover in explaining 30–60% of the variance in maximum
Althoughthereareothersitesinthedatabase,thesesitesare
icethickness.SimilarresultswerealsoreportedbyFlatoand
the only ones than span the same 55-year period between
Brown(1996)andGoughetal.(2004).Intherecordexam-
1960and2014.TherecordatMouldBay,usedinBC92,ter-
ined by BC92, no evidence for systematic thinning of land- minatedintheearly1990s.Togetherthesesitescover∼20◦
fasticeintheCAAwasfound.Landfasticethicknessrecords
inlatitude(Fig.1)adjacenttoanareaofthickArcticseaice
atseveraloftheseCAAsitesarenowover50yearsinlength,
thatexperiencedthehighestthinninginrecentyears(Kwok
whichrepresentsanadditionofmorethan2decadesofmea-
andRothrock,2009;Laxonetal.,2013).Valuesofmaximum
surementssinceBC92duringaperiodthatsawdramaticre-
orend-of-wintericethicknessandcorrespondingsnowdepth
ductions in the extent and thickness of Arctic sea ice (e.g.
duringtheicegrowthseasonwereextractedfromtheweekly
KwokandRothrock,2009;Stroeveetal.,2012).
iceandsnowthicknessdataattheselectedsites(seeSupple-
Thesparsenetworkoflong-termobservationsofsnowand
ment). As this study is concerned with annual variability in
icethicknessintheArctic(clearlyexhibitedbyonlyfouron-
maximumicethickness,themainperiodofinterestextends
goingmeasurementssitesoperatedbyEnvironmentCanada
fromSeptembertolateMay.
intheCAA)hasmadetheuseofmodelsimperativetopro-
The other source of observed data used in this study
vide a broader regional scale perspective of sea ice trends
was Environment Canada’s monthly mean air temperature
inawarmingclimate.Giventhecoarsespatialresolutionof
records at Alert, Eureka, Resolute and Cambridge Bay for
globalclimatemodels,previousstudiesfocusingontheCAA
which a complete description is provided by Vincent et
havereliedoneitheraone-dimensionalthermodynamicdy-
al.(2012).
namic model (Flato and Brown, 1996; Dumas et al., 2006)
or a regional three-dimensional ice–ocean coupled model 2.2 Models
(e.g.SouandFlato,2009).Specifically,Dumasetal.(2006)
found projected maximum ice thickness decreases of 30cm The representation of CAA landfast sea ice thickness
by 2041–2060 and 50cm by 2081–2100 and Sou and Flato within the Coupled Model Intercomparison project phase 5
(2009)reportedapotential17%decreaseinoverallicethick- (CMIP5)isanalysedusingthe1850–2005Historicalexperi-
nessthroughouttheCAAby2041–2060.However,inrecent mentfollowedbythe2006–2099RepresentativeConcentra-
years some global climate models, reanalysis products and tion Pathway 8.5 (RCP85) experiment (Taylor et al., 2012)
TheCryosphere,10,1463–1475,2016 www.the-cryosphere.net/10/1463/2016/
S.E.L.Howelletal.:LandfasticethicknessintheCAAfromobservationsandmodels 1465
Table1.CMIP5modelsusedinthisstudy,thenumberofrealiza-
tionswithicedataandthenumberofrealizationswithseaicetrans-
portdata. aily
d
#ofsimulations #ofsimulations Arctic al-time
bbBCccNaccn--UEcc-ssSEmmMS11M2--11-m 1115 MHHMaaIIddRRGGOOEECCMM-5E22S--CEMSC-CHEM 1314 PIOMAS APL/PSC∼22kminthe POPTED 1958–2015NCEP/NCAR NSIDCnear-re
CMCC-CESM 1 MPI-ESM-LR 3
CCMMCCCC--CCMMS 11 MMPRII--ECSGMC-MM3R 11 eading SI-SAF
CNRM-CM5 5 CCSM4 6 R O
AACFISCCOICCR-EEEOSSS-MSSM11k..033.6.0 11110 GNNGFooFrrDDEELLSS--MMCESM11M--MM32EG 1111 UR025.4 Universityof◦ORCA0.25 NEMO3.2LIM2 1993–2010ERA-Interim EUMETSAT
EC-EARTH 6 GFDL-ESM2M 1 n
o
inmcm4 1 CESM1(BGC) 1 ati
n
FGOALS-g2 1 CESM1(CAM5) 3 bi
m
MIROC-ESM 1 CESM1(WACCM) 3 o
c
A
◦ m TI
ORAP5.0 ECMWFORCA0.25 NEMO3.4LIM2 1985–2013ERA-Interi NOAA/OS
y)
g
o
ol
e T
h A
(cawdTonaeandtrabcebsleenanrsoet1erwt)ar.iateidnMovednepodatn(hcvtfchar(eolvrysimaasrebsidealtehabtelhseiicrBcoes)nur,itdgth2i)hiscmhwtkhenAetreeetsmCmsuepos(neesvtrpdraaeh.rtiueTafrrobheirclee(EvDCsnaiaMvrtt)iiaa,rIobPsCnl5eemeadnteiaatcnrsteea)- GLORYS2V3 MercatorOcéan◦ORCA0.25 NEMO3.1LIM2(withEVPr 1993–2013ERA-Interim IFREMER/CERS
tal Data Analysis (www.ceda.ac.uk). Ensemble r6i1p1 and
y
rc7oir1rupp1tefdromdatma.oWdeel oEbCta-EinAtRhTeHmuwlteir-emoredmelovmeedanbeocfautrseendosf AERctic monthl
and their statistical significance at each grid point by creat- T-Ar ap
bSitniuwoginladt.rhWteaeetdnuiaosslit.esr(eiab2um0tt1dio5oids)netraloinbftdouttrLaieocanncldiofbsouetrnrhtttréhofeeoutirgnahitlne.atr(ea2Mrnn0noa1unl6at)vel.avWCraiaareirbaloioblbiitstlyaiitmiynausatlnhiadne- ECCO-v4 JPL-NASA-MI∼40kminthe MITgcmMITgcm 1991–2011ERA-Interim NSIDCBootstr
s.
mcaullstii-gmniofidcealnmceeabnyofifrPsteapresrofnorcmoirnreglaatiFonisshearndtrathnesifrosrmtatiasntid- eristic daily
thenapplyingthesamemethodasforthetrends.Theinverse aract Team
Fisher transform is applied after obtaining the multi-model h A
omdfeesWtachnreeipaahtnliisgdoohnietiosnsftvs-terihegsetsniogimfilauecttetaihonionccdeem..tShoiedceeklntshe(esSsatopvrpatloeuneedstixafrl.fo,om2r0aa1c1soe;mlFeocptrligeoetnet A-IPmodelc CGLORS CMCC◦ORCA0.25 NEMO3.2.1LIM2 1982–2012ERA-Interim NSIDCNAS
R
et al., 2015; Haines et al., 2014; Zuo et al., 2015; Masina O ed
e((a2EOTtn0Rad0RaAb3lAAl.)-e,.I-snI2Ss2Pti0uem))1rpi5i(pamlB)aon,tard(itlfDoimrnnfoegrameoSs2meeydttmshataetelhtm.ee,eOtm2(ac0PPple1.aeIa,1Onrna2)-M.tA0uR1rAre5ceaSt;idn)cCaa(tlhZaIyecshwveiaas-enOlrlgIiecneeatroeanbrendcttaoRiaMmnlo.eop,tdhda2rref0oirls1iocon6mkng), Table2.Summaryof Modelname InstituteResolution OceanmodelSeaicemodel TimeperiodconsideredAtmosphericforcing Seaiceproductassimilat
www.the-cryosphere.net/10/1463/2016/ TheCryosphere,10,1463–1475,2016
1466 S.E.L.Howelletal.:LandfasticethicknessintheCAAfromobservationsandmodels
Table 3. Observed maximum ice thickness, snow depth and surface air temperature at four landfast ice sites in the Canadian Arctic
Archipelago.Theboldtextindicatesstatisticalsignificanceofthelineartrendat95%orgreater.
CambridgeBay Resolute Eureka Alert
Period 1960–2014 1957–2014 1957–2014 1957–2014
IceThickness,hice
Meanofmaxhice(m) 2.11±0.19 2.02±0.19 2.27±0.23 1.98±0.22
Trendofmaxhice(cmdecade−1) −4.31±1.4 −0.5±1.6 −4.65±1.7 −4.44±1.6
Dayofmaxhice 24May±17 25May±21 26May±12 27May±16
Trendofdayofmaxhice(daysdecade−1) −0.87±1.5 −6.2±1.5 −2.0±0.1 −3.0±1.2
Snowdepth(hsnow)
MeanOct–Mayhsnow(cm) 8.4±4.2 22.6±10 17.6±5.8 18.4±6.2
TrendofOct–Mayhsnow(cmdecade−1) −0.8±0.4 −0.75±0.8 0.54±0.5 0.26±0.5
Temperature
Winter(Dec—Feb)mean(◦C) −31.3±2.0 −30.8±1.9 −36.0±2.0 −31.2±1.6
Winter(Dec—Feb)(◦Cdecade−1) 0.59±0.2 0.35±0.1 0.23±0.2 0.38±0.1
Spring(Mar—May)mean(◦C) −20.0±1.8 −21.1±1.8 −24.9±2.0 −22.8±1.8
Spring(Mar—May)(◦Cdecade−1) 0.47±0.1 0.57±0.1 0.44±0.1 0.32±0.1
Summer(Jun—Aug)mean(◦C) 5.9±1.4 2.3±1.3 3.9±1.2 1.3±0.8
Summer(Jun—Aug)(◦Cdecade−1) 0.30±0.1 0.17±0.2 0.21±0.1 0.1±0.1
Fall(Sep—Nov)mean(◦C) −11.1±2.0 −13.8±2.0 −19.6±2.2 −18.0±1.7
Fall(Sep–Nov)(◦Cdecade−1) 0.60±0.2 0.67±0.1 0.68±0.2 0.56±0.1
3 Resultsanddiscussion:observations
3.1 Climatology
The average behaviour of landfast ice at the four sites over
the 50+-year record is summarized in Table 3. Ice growth,
approximately linear through most of the season, slows af-
terMarch(Fig.2).Icethicknessreachesamaximumof∼2–
2.3mbylateMayatallsites.Valuesareconsistentwiththose
reportedbyBC92andwithrecentobservationsofMellinget
al.(2015)andHaasandHowell(2015).Thestandarddevia-
tionsarenearlyuniform(at∼0.2m)acrossallsites,givinga
relativelylowcoefficientofvariation(CV;ameasureofrel-
ativedispersiondefinedastheratioofthestandarddeviation
to the mean) of ∼0.1. The thickest ice is found in Eureka
witha1957–2014meanof2.27m,whichislikelyduetocli-
matologically lower air temperatures in the fall and winter
(Table3). Figure2.Seasonalcycleofobservedmeanicethicknessatthefour
sites(1960–2014).
Snowdepthalsoappearstogrowlinearlythroughthesea-
son, peaking in May, but unlike ice thickness the monthly
variabilityishigh(CV∼0.4)(Fig.3).MeanOctobertoMay
snowdepthsatResolute,EurekaandAlertrangefrom∼18 fication throughout the ice growth season (BC92; Woo and
to 23cm compared to only ∼8cm at Cambridge Bay (Ta- Heron,1989).
ble 3). The rapid buildup of the snow cover due to storms
in the fall and early winter, which is evident over the Arc- 3.2 Trends
ticOceanmulti-yearicecover(Warrenetal.,1999;Webster
etal.,2014),isnotseeninthesesnowdepthrecordswithin ThetimeseriesofmaximumicethicknessatCambridgeBay,
theCAA.Thelinearbehaviourinsnowdepthislikelymain- Resolute,EurekaandAlertareillustratedinFig.4andsum-
tained by continuous wind-driven redistribution and densi- marizedinTable1.Statisticallysignificant(95%orgreater
TheCryosphere,10,1463–1475,2016 www.the-cryosphere.net/10/1463/2016/
S.E.L.Howelletal.:LandfasticethicknessintheCAAfromobservationsandmodels 1467
Figure3.Seasonalcycleofobservedmeansnowdepthatthefour Figure5.Timeseriesandtrendofobservedmeansnowdepthatthe
sites(1960–2014). foursitesfromOctoberthroughMay.
(−7.1cmdecade−1), no evidence of a trend at Eureka and
a positive trend at Resolute (10cmdecade−1), but only the
positive trend at Resolute was statistically significant at the
95%orgreaterconfidencelevel.Ourresultsfromthepresent
50+-yearrecordsuggestthatthenegativetrendatAlertisro-
bustandthetrendatEurekaisnownegativeandsignificant.
The trend at Resolute is now slightly negative, but it is not
statisticallysignificant.
Typically, ice thickness reaches its maximum in late
May with trends toward earlier dates of maximum
ice thickness present at all sites (significant at Reso-
lute, Eureka and Alert; Table 3). The significant trends
are between −2.0±0.1daysdecade−1 at Eureka and
−6.2±1.5daysdecade−1 at Resolute. At Resolute, the date
of maximum ice thickness is now on average more than a
month earlier than the early 1960s, although this is not re-
Figure4.Timeseriesandtrendofobservedmaximumicethickness flectedinthetrendinicethickness.Freezeonsetatthesesites
atthefoursites. also increases at ∼3–6daysdecade−1 (Howell et al., 2009)
anddemonstratesashortenedgrowthseasonatResolute,Eu-
rekaandAlert.Together,thetrendsoficethicknessandtheir
confidencelevel)negativemaximumicethicknesstrendsare recorded dates suggest a systematic thinning of landfast ice
present at Cambridge Bay (−4.31±1.4cmdecade−1), atCambridgeBay,EurekaandAlert.
Eureka (−4.65±1.7cmdecade−1) and Alert
(−4.44±1.6cmdecade−1) (Table 1). A slight negative 3.3 Icethicknesslinkageswithsnowdepthand
trend is present at Resolute but not statistically significant temperature
at the 95% confidence level (Table 1). Over the 50+-year
record, the ice thinned by ∼0.24–0.26m at Cambridge ThevariabilityoflandfastthicknessattheseArcticsiteswas
Bay, Eureka and Alert with essentially negligible change previously found to be largely driven by interannual varia-
at Resolute. These trends in the CAA are similar to trends tions in snow depth and air temperature (BC92; Flato and
on the Siberian coast (−3.3cmdecade−1) (Polyakov et Brown,1996).Withthe50+-yearrecordatthefoursites,we
al., 2010) but lower in magnitude compared to the Barents can examine the corresponding linkages to snow depth and
Sea(−11cmdecade−1)(Gerlandetal.,2008). temperature.
For the shorter record (late 1950s–1989, ∼30 years) in- For snow depth, the only trend that is statistically sig-
vestigated by BC92 there was a negative trend at Alert nificant at the 95% confidence is Cambridge Bay at
www.the-cryosphere.net/10/1463/2016/ TheCryosphere,10,1463–1475,2016
1468 S.E.L.Howelletal.:LandfasticethicknessintheCAAfromobservationsandmodels
−0.8±0.4cmdecade−1 (Table3,Fig.5).Incontrast,BC92
foundasignificantpositivetrendatAlert(4cmdecade−1),a
trendoflowsignificanceinEurekaandanegativeandsignif-
icant trend at Resolute (−3.3cmdecade−1). Looking at the
detrendedcorrelations(r)betweensnowdepthandicethick-
nessrevealsthestrongestcorrelationatResolute(r =−0.71)
followed by Eureka (r =−0.66), Alert (r =−0.47) and
Cambridge Bay (r =−0.31). Figure 6 provides evidence
fromextremeyearsoftheroleofdeepersnowinhibitingice
growth compared to thinner snow, but the positive trends in
snow thickness are not significant at Resolute, Eureka and
Alert.Thismayinpartbeduetothesinglepoint-wisesnow
depthandicethicknessmeasurementsmadeateachpointin
time,whichfailtocapturespatialheterogeneityinthesnow
depth–icethicknessrelationship.
Withrespecttoobservedtemperature,wefindsignificant
warmingtrendsinthespringandfallatallsitesoverthe50+-
year record (Table 3; Fig. 7). Significant warming is also
presentat allsitesinthe summerexceptResoluteand atall
sitesduringthewinterexceptEureka(Table3).Warmingis
highestduringthefall,at∼0.6◦Cdecade−1 atallsites(Ta-
ble3).Thedetrendedcorrelationbetweentemperature(win-
ter, spring, summer and fall) and maximum ice thickness is
weak at all sites. For example, the strongest detrended cor-
relation between maximum ice thickness and temperature
(winter and spring) is found at Cambridge Bay during the
winterandspringbutisonly∼0.4.
Also of interest is that the observed temperature trends
over this period differ considerably from the earlier period
Figure 6. Weekly time series of ice thickness and snow depth at
investigated in BC92, in which they reported cooling at all
EurekaandAlertfor(a)lowsnowyearsand(b)highsnowyears.
the sites, with a significant cooling trend at Eureka. It was
noted that the general cooling over their record coincided
with the 1946–1986 cooling trend over much of the east-
ern Arctic and northwestern Atlantic reported by Jones et
al.(1987).Thiscoolingtrendhaltedduringthe1980sandthe
warming,seeninthecurrentandlongerrecord,hasresumed
(Jones et al., 1999). Arctic land areas have experienced an
overall warming of about ∼2◦C since the mid-1960s, with
area-widepositivetemperatureanomaliesthatshowsystem-
atic changes since the end of the 20th century, which con-
tinuedthrough2014(JeffriesandRichter-Menge,2015).Re-
cently,warminginCanadianArcticregionswasfoundtobe
greater than the pan-Arctic trend by up to 0.2◦Cdecade−1
(Tivyetal.,2011).
4 Resultsanddiscussion:models
4.1 Climatology
In order to compare seasonal cycles and trends in land-
Figure7.TimeseriesobservedmeanairtemperaturebyEnviron-
fast ice thickness and snow depth between models and
mentCanadaduringwinter(DFJ),spring,(MAM),summer(JJA)
observations, we limit our comparison to models with
andfall(SON)attheCambridgeBay,Resolute,EurekaandAlert.
a reasonable representation of the CAA, i.e. those with
an open Parry Channel (i.e. bcc-csm-1-1, bcc-csm-1-1m,
TheCryosphere,10,1463–1475,2016 www.the-cryosphere.net/10/1463/2016/
S.E.L.Howelletal.:LandfasticethicknessintheCAAfromobservationsandmodels 1469
CNRM-CM5, ACCESS1-0, ACCESS1-3, FIO-ESM, EC-
EARTH, inmcm4, MIROC5, MPI-ESM-LR, MPI-ESM-
MR,MRI-CGCM3,CCSM4,NorESM1-M,NorESM1-ME,
GFDL-CM3,GFDL-ESM2G,GFL-ESM2M,CESM1-BCG,
CESM1-CAM5, CESM-WACCM). In these models, suffi-
cient spatial resolution allows us to find sample points that
are almost collocated to in situ observation locations. The
samplepointsweredeterminedbyfindingtheclosestocean
gridpointwheretheseaiceispackedforagoodportionof
theyear,butnotallyear.Gridpointswiththischaracteristic
thereforesharethemostimportantfeatureofthelandfastice
atourobservationslocations:itisnotperennial.Mathemat-
ically, we sought sample points where the sea ice concen-
trationisonaverageabove85%formorethan1monthbut
lessthan11monthsoverthe1955–2014period.TheEureka Figure8.CMIP5medianseaicethicknessseasonalcycle(1955–
2014) at stations (grey). Observations from 2 (black). Median
siteis,however,particularlychallengingformodelsbecause
of ORA-IP models CGLORS, ORAP5.0, GLORYS2V3 (blue),
itliesdeepinaverynarrowchannel,whichisonlyresolved
ECCO-v4(green)andUR025.4(red).Whiskersindicatethe5thand
by the MPI-ESM-MR in the CMIP5. As a result, for most
95thpercentiles.
models,thesamplepointforEurekaislocatedonthewest-
ernshoreofEllesmereIsland.Thisisaconsequenceofusing
samplesbecausesomemodelseitherdonotresolvesomeof
the channels in the CAA or have too perennial packed ice
cover (e.g.CESM1-CAM5), and then the samplepoints are
furtherfromtheobservationalsitethanwouldbedesired.We
chose to use sample points in our comparison to observa-
tionsinsteadofusingregionalaveragesfortwomainreasons.
The first reason is that using regional averages would have
lumped together different ice dynamics regimes that should
notnecessarilybeexpectedtocomparewelltopointobserva-
tionsonlandfastice.Thesecondreasonisthatweareofthe
opinion that the resolution in many of these models is suf-
ficiently high to warrant such a direct comparison and pro-
videsabetterbenchmarkthanregionalaveragesforlandfast
icemodellingintheCAA.
The seasonal cycle (1955–2014) of median ice thickness Figure9.SameasFig.8forsnowdepthandonlyforCMIP5models
from CMIP5 (black), ORA-IP models CGLORS, ORAP5.0 (grey)andobservations(black).
and GLORYS2V3 (blue), ECCO-v4 (green) and UR025.4
(red)isshowninFig.8.ORA-IPmodelshavebeensplitinto
threegroupsbased,respectively,ontheirhigh,mediumand thicknessdoesnot(Fig.8),unlikeinobservations.Thislikely
low ice thicknesses at Alert. Ice thickness from CMIP5 is reflects the fact that the grid cell thickness in sea ice mod-
comparable to observations (Fig. 2) at Cambridge Bay and elswiththicknessclassesarepresentstheaveragethickness
Resolutewithmaximumicethicknessreaching200cm.The over these classes. In August the thinner ice classes might
ORA-IPmodelsarelessconsistent.ECCO-v4tendstohave have melted but thicker ice classes can still be found, re-
thicker sea ice than observations at Cambridge Bay, Reso- sulting in a substantial average ice thickness over the grid
lute and Eureka but thinner at Alert. CGLORS, ORAP5.0 cell.Theseasonalcycleoverpackediceinthesemodelsthus
andGLORYS2V3,however,arecomparabletoobservations gives a reasonable representation of the seasonal cycle over
atCambridgeBay,ResoluteandEurekabuthaveextremely landfast ice in the CAA, especially in the southern region
thickandperennialiceclosetoAlert. oftheCAA.Overall,thiscomparisonshowshowrecentim-
The seasonal cycle (1955–2014) of median snow depth provements in sea ice model resolution allows comparisons
from CMIP5 is shown in Fig. 9. CMIP5 models indicate a withobservationsthatrequireddynamicaldownscalingtech-
linearincreasesimilartoobservationsreachingamaximum niquesinthepreviousgenerationofseaicemodels(i.e.Du-
of∼20cminAprilorMay.Thisislowerthantheobserved masetal.,2005;SouandFlato,2009).
maximumatResolute,EurekaandAlertbutisabouttwiceas Despite relatively high spatial resolution, PIOMAS does
muchasatCambridgeBay.Whilethesnowdepthreaches0 notresolveseasonalicethicknessalongthecoastsandwithin
duringthesummeratEurekaandAlertinmodels,theseaice the very narrow channels within the CAA (not shown). As
www.the-cryosphere.net/10/1463/2016/ TheCryosphere,10,1463–1475,2016
1470 S.E.L.Howelletal.:LandfasticethicknessintheCAAfromobservationsandmodels
Figure11.ComparisonofPIOMASicethicknesswithicethickness
observationsfromEnvironmentCanada’sicethicknessmonitoring
sites at Cambridge Bay and Resolute. The data cover the period
1979–2014.
consistentwiththedifferentinsituobservations(Fig.4)but
Figure10.Seasonalcycleofobservedmeanicethickness(left)and
with overestimated negative thickness trends. Although for
snowdepth(right)fromPIOMASatCambridgeBayandResolute
individual models this pattern is far from uniform, the gen-
(1979–2014).
eral pattern and magnitude of thickness trends tend to be
roughlyinaccordancewithtemperaturetrends(notshown).
a result, Cambridge Bay and Resolute Bay sites represent AsimilarbehaviourisobservedintheORA-IPmodels,with
the only long-term monitoring sites within the CAA suit- thenotableexceptionofCGLORS,wherepositivethickness
ableforcomparisonsincePIOMAS.Themonthlytimeseries trendsarefoundalmosteverywhere(Fig.12a).Thisisrobust
ofPIOMASiceandsnowthicknessestimatesatCambridge and it appears that the model is not completely equilibrated
BayandResoluteisshowninFig.10.Theseasonalcycleof intheCAAandexhibitslargemonth-to-monthadjustments.
icegrowthatCambridgeBayandResoluteisrepresentative Model ORAP5.0 also is not completely equilibrated in the
comparedtoobservations(Fig.2)butPIOMASestimatesre- region for years 1979–1984. During those years, it exhibits
tainmoreiceinAugustandSeptember,particularlyatRes- large interannual changes in thickness. For this reason, we
olute.IcegrowthreachesamaximuminAprilatCambridge areonlyconsideringyears1985–2013forthismodel.
and in May at Resolute which is 1 month earlier compared ForPIOMAS,thenorth–southoverestimatedtrendisalso
toobservations.Snowdepthfollowsalinearincreasesimilar present (not shown) as with CMIP5 and ORA-IP. Looking
toobservations(Fig.3),withgoodagreementatCambridge specifically at trends computed from 1979 to 2014 near the
Bay,butconsiderablyunderestimatessnowdepthatResolute observedsitesshowsthemeanmaximumicethicknesslinear
(Fig.10).Schweigeretal.(2011)performedadetailedcom- trendfromatCambridgeBaytobe−13.4+3.4cmdecade−1,
parison of PIOMAS ice thicknessvalues against in situ and which is almost double the observational trend of 6.2±
Ice, Cloud, and land Elevation Satellite (ICESat) ice thick- 2.4cmdecade−1. At Resolute, the PIOMAS linear trend is
ness observations and found strong correlations. They de- 24.0±4.1cmdecade−1,whichisconsiderablystrongerthan
terminedarootmeansquareerror(RMSE)of∼0.76mand theobservationaltrendof−4.9±3.51cmdecade−1.
notedthatPIOMASgenerallyoverestimatesthinnericeand
underestimates thicker ice. At both sites within the CAA, 4.3 Icethicknesslinkageswithsnowdepthand
PIOMAS ice thickness data are in reasonably good agree- temperature
ment with in situ observations with RMSEs of 0.29cm at
CambridgeBayand0.68cmatResolute(Fig.11).Thesys- Even though ORA-IP models have unrealistically large
tematicoverestimateofthinnericereportedbySchweigeret thickness trends, the pattern of interannual correlation (de-
al.(2011)ismoreapparentatResolutethanCambridgeBay trended) between winter temperatures and thicknesses is
(Fig.11).Thehigher-latituderegionsoftheCAAwherethere roughly consistent across models (Fig. 13). Some ORA-IP
isanintricatemixofseasonalfirst-yeariceandmulti-yearice modelsalsoexperiencepositivecorrelations(e.g.CGLORS,
isaproblemforPIOMASandthuscontributestothelarger ORAP5.0, GLORYS2V3 and UR025.4) that are mostly lo-
discrepancyatResolutecomparedtoCambridgeBay. catednorthoftheCAAorwithintheCAAinregionswhere
multi-year ice is known to be present. It is possible that
4.2 Trends warmer temperatures are associated with an increased flux
of thicker multi-year ice into the CAA, which is known to
Thespatialdistributionofmaximumseaicethicknesstrends occur(e.g.Howelletal.,2013),butthedrivingprocessesre-
from ORA-IP and CMIP5 is illustrated in Fig. 12. The sponsibleforthesepositivecorrelationsrequiremoreinves-
CMIP5 model mean exhibits a fairly uniform trend pattern, tigation.InCMIP5models,nomodelexhibitspositivecorre-
TheCryosphere,10,1463–1475,2016 www.the-cryosphere.net/10/1463/2016/
S.E.L.Howelletal.:LandfasticethicknessintheCAAfromobservationsandmodels 1471
Figure 12. (a–e) Maximum sea ice thickness trends in ORA-IP simulations. (f) Same for CMIP5 model-mean. From south to north, o’s
indicateCambridgeBay(green),Resolute(blue),Eureka(white)andAlert(black)andx’sindicatethecorrespondingmeasurementstations.
In(f),oneopermodelisshown.Thestipplingindicatespvalueslessthan0.05,correctedusingthefalsediscoveryratemethodwithaglobal
pFDRvaluelessthan0.10(Wilks,2006).Thecolourbarislinearfrom−10to10cmdecade−1 andsymmetriclogarithmicbeyondthese
values.
lationswithtemperaturethatresemblesORA-IPmodelsover 5 Conclusions
theCAA.AlthoughthetimeseriesfortheORA-IPmodelsis
short and the positive correlations are only statistically sig-
Over the 50+-year in situ observational record, statistically
nificantatafewgridpointsinCGLORSandUR025.4,this
significant negative trends in maximum (end-of-winter) ice
behaviourissufficientlyproblematictorecommendthatcare
thickness are present at Cambridge Bay, Eureka and Alert.
should be taken when using these ORA-IP models to study
Significantnegativetrendsinthedayofmaximumicethick-
theinterannualvariabilityintheCanadianArctic.
nessarealsopresentatResolute,EurekaandAlert.Together,
IntheCMIP5models,significantwintersnowdepthtrends
these trends suggest thinning of landfast ice in the CAA,
are more strongly negative in the north than in the south
where little evidence was found in the shorter record anal-
(Fig. 14). This is in disagreement with point observations
ysed in an earlier study (BC92). The interannual variability
presented in the previous sections that showed no signifi-
ofairtemperatureisonlyweaklycorrelatedtomaximumice
cant trendssnow depthtrends at Alertbut negativeand sig-
thickness (i.e. maximum correlation is ∼0.4). Snow thick-
nificant trends at Cambridge Bay. Although only based on
ness plays the dominant role in controlling maximum ice
limitedpointinsituobservations,thissuggeststhatoverthe
thickness variability given the high correlations at Resolute
last decades changes in winter precipitation at Alert must
and Eureka and reasonably high correlations at Alert and
have compensated the increased melting driven by increas-
CambridgeBay.
ingtemperatures,acompensationthatisclearlynotcaptured
Comparison of CMIP5, ORA-IP and PIOMAS simula-
inCMIP5models.
tions with observations indicate a reasonable representation
ofthelandfasticethicknessmonthlyclimatologywithinthe
CAA. This is particularly apparent when seasonal first-year
icedominatestheicescape(i.e.CambridgeBay).Despiteim-
www.the-cryosphere.net/10/1463/2016/ TheCryosphere,10,1463–1475,2016
1472 S.E.L.Howelletal.:LandfasticethicknessintheCAAfromobservationsandmodels
Figure13.(a–e)PearsoncorrelationofdetrendedmaximumseaicethicknessinORA-IPwithdetrendedONDJFMAMERA-INTERIM2m
temperature.(f)SamebutforCMIP5MODEL-MEAN.Thestipplingindicatespvalueslessthan0.05,correctedusingthefalsediscovery
ratemethodwithaglobalpFDRvaluelessthan0.10(Wilks,2006).
The overall thickness of ice within the CAA in the cur-
rent generation of models is too high. As a result, trends
are unrealistic and far exceed observations (by upwards of
−50cmdecade−1)inpartbecausetheinitialicethicknessis
too large. The problem is particularly acute in the ORA-IP
models where large and unrealistic interannual changes in
thicknesssuggestthatthemodelsarenotfullyequilibrated.
While the impact of the snow cover on ice thickness is
wellknown,thesignificantcorrelationsatResolute,Eureka
and Alert suggest that the higher sensitivity to changes in
snow depth could potentially mask the warming signal on
bothfastandoffshoreice.Thus,eveninthislimiteddataset,
wecanseethedominantroleplayedbysnowdepthindeter-
mining the interannual variability of the maximum landfast
icethickness.Thisagainhighlightsthattheprimaryfactoris
theamountandtimingofsnowaccumulationratherthanair
Figure 14. Same as Fig. 12f but for snow depth trends (ONDJF-
temperature.However,itisworthnotingthatfewofthecur-
MAM).
rentgenerationmodelsshowcoherentrelationshipsbetween
ice thickness, snow depth and temperature over the longer-
termrecord.
provements in spatial resolution, mixed ice types (i.e. sea-
sonal and multi-year) present at the sub-grid cell resolution
are likely problems for model estimates within the CAA.
TheCryosphere,10,1463–1475,2016 www.the-cryosphere.net/10/1463/2016/