: processes, observaons and models

Cryosphere and Change CIE 4602 2015 – 2016 February 10th OUTLINE

1) lifecycle: formaon, growth and melt 2) Observaons: – Microwave and opcal signatures – Arcc sea ice – Antarcc sea ice 3) Models: dynamics and thermodynamics 4) Sea ice model performance References:

• IPCC Fih Assessment report (AR5) • hps://nsidc.org/cryosphere/seaice • hp://earthobservatory.nasa.gov/ • Canadian Ice Service (www.ec.gc.ca/glaces-ice) • JPL Polar Oceanography Group • hp://www.arcc.noaa.gov • Sea Ice Physics and (Shokr & Sinha, Wiley, 2015) Introducon to Sea Ice

• A major component of the polar ecosystem: – Habitat for plants and animals at all trophic levels (plankton, algae, fish, birds, seals, penguins, bears and whales) • Also a component of the : – changes the surface (incoming SW) – insulates the ocean from heat loss (outgoing LW)

– barrier to gas (H2O, CO2) and momentum exchanges – alters ocean density à thermohaline circulaon àearly indicator of and amplifier of climate perturbaons

Ice-albedo feedback: melng ice has a lower albedo (absorbs more sunlight, thus melts more) Excursus – Thermohaline circulaon

• Also MOC for Meridional [Morrison, Frolicher & Sarmiento, Phys. Today, 2015] Overturning Circulaon (turns every ~1000 years) or Great Ocean Conveyor Belt.

• Slow density-driven ocean circulaon (as opposed to the fast wind-driven circulaon that dominates the ocean upper few hundred meters). • Downwelling: salty cold from new sea ice formaon (e.g. Weddell Sea, Barents Sea) gets this circulaon started.

• Upwelling: thought to occur mostly along density surfaces in the by à Important because of heat uptake (of excess energy in the climate system), carbon sink and nutrient supply (from wind-driven . deep water enriched by the biological pump). Figure I.12, [WMO,1970]). Since its formation from suspensions as young ice, the

process of growth into first-year and multi-year ice is one of steady , thickness increase,1 - Sea ice lifecycle and surface erosion. The thickness of the overlying layer on top of the sea ice increases in general as the sea ice grows. In this section, we will review the

• Sea ice may be roughly divided into five different age classes (new, young, development stages of sea ice and the basic terminology required for its classification. thin first year, first year and mulyear ice) used as a proxy for ice thickness.

small pressure ridge lead large pressure ridge rafting

Sea Snow level

Salt rejecon desalinaon

New Young Thin First Year Ice Multiyear Ice Ice Ice First Year

Thickness Since its formaon from crystal suspensions as new and young ice, the <10 cm <30 cm <70 cm <200 cm ~400 cm process of growth into first year and mulyear ice is one of steady Salinity 25 ‰ 15 ‰ 4-15 ‰ 4-5 ‰ 2 ‰ thickness increase (by thermodynamic growth and deformaon), surface erosion and desalinaon, with an increasing layer of snow depth. Snow ~10 cm ~10 cm ~30 cm – – Figure I.12 – Sea ice terminology ([WMO, 1970])

Formation

Sea ice formation into new ice [see Figure I.13(a)] begins at the sea surface with a

suspension of ice known as frazil [Tucker et al., 1992]. When ice forms in calm

, the frazil rises to form an unconsolidated layer of crystals known as grease ice,

which stabilizes the sea surface and suppresses the formation of capillary waves in the

presence of wind. Continued results in a smooth, thin, elastic ice known as dark

nilas. Consolidation progresses by water crystallization, with a resulting increase in

salinity of the remaining liquid. Some of the saline brine is forced out of the ice mass to

the sea beneath and to the surface. The remainder of brine is trapped within the ice in

21 Figure I.12, [WMO,1970]). Since its formation from crystal suspensions as young ice, the process of growth into first-year and multi-year ice is one of steady desalination, thickness increase,1 - Sea ice lifecycle and surface erosion. The thickness of the overlying snow layer on top of the sea ice increases in general as the sea ice grows. In this section, we will review the development stages of sea ice and the basic terminology required for its classification.

small pressure ridge lead large pressure ridge rafting

Sea Snow level

New Young Thin First Year Ice Multiyear Ice Ice Ice First Year

Thickness <10 cm <30 cm <70 cm <200 cm ~400 cm

Salinity 25 ‰ 15 ‰ 4-15 ‰ 4-5 ‰ 2 ‰ Snow ~10 cm ~10 cm ~30 cm – – Figure I.12 – Sea ice terminology ([WMO, 1970])

Formation

Sea ice formation into new ice [see Figure I.13(a)] begins at the sea surface with a suspension of known as frazil [Tucker et al., 1992]. When ice forms in calm seas, the frazil rises to form an unconsolidated layer of crystals known as grease ice, which stabilizes the sea surface and suppresses the formation of capillary waves in the presence of wind. Continued freezing results in a smooth, thin, elastic ice known as dark nilas. Consolidation progresses by water crystallization, with a resulting increase in salinity of the remaining liquid. Some of the saline brine is forced out of the ice mass to the sea beneath and to the surface. The remainder of brine is trapped within the ice in

21 Sea ice formaon

• Formaon of new ice begins at the sea surface with a random suspension of ice crystals known as frazil (aer salt rejecon). • When ice forms in calm seas, the frazil connues to form an unconsolidated layer of crystals known as grease ice, which stabilizes the sea surface and suppresses the formaon of capillary waves in the presence of wind. Salt rejecon eliminates about 80% of the inial salt contents. Ice producon and salinity enhancements go hand-in-hand.

Sea ice development

• Connued freezing results in a smooth elasc thin ice known as dark nilas, becoming brighter as it thickens.

• Currents or winds oen push the nilas around so that they slide over each other, a process known as raing.

• Under calm condions, sea ice growth progresses by steady crystallizaon and brine drainage (air filling), some of which will remain trapped in vercally elongated brine pockets. Connued growth takes place at the boom of the slab (basal freezing as congelaon ice).

Mixture of brine and air pockets give old ice its (electromagnecally) bright appearance Sea ice development

• Wave acon (parcularly in the SH) causes the ice to lump and form small rounded floes called pancakes.

Super- • On thin ice, snow deposion may weigh down the ice enough to cause flooding imposed ice mainly controlled by the snow cover. Bare sea disintegrates more quickly than snow- covered sea ice, and this pattern of differential melting will result in the appearance of melt ponds, hummocks, drainage channels and weathered ridges [see Figure I.13(e)]

[Tucker et al., 1992]. Some first-year ice survives the summer melt, becoming thick multiyear ice, with a top layer transformed into a porous, low salinity cover, and a surface relief that becomes increasingly modulated by snow deposition and wind erosion [see

Figure I.13(c)].

Sea ice development

• Growth into first-year ice is a process of connued brine drainage (desalinaon with formaon of air pockets), with increases in ice thickness and snow load. Meanwhile, the ice surface will undergo connual deformaon under forcing from wind, waves

and ocean currents, forming leads, pressure ridges and rubble fields.

(a) New Ice Leads (b) First-year ice (c) Multi-year ice Rubble fields

(d) Ridge and Lead e) Melt Divergence causes lower Convergence sea ice concentraons Figure I.13 – Development stages of sea ice produces thicker ice (Photos courtesyand enhanced sea ice of the Canadian Ice Service and the JPL Polar Oceanography Group) by deformaon. producon. à Dynamical effects on sea ice state Dielectric properties of sea ice

The electrical properties of sea ice are relevant, along with structural features such as surface roughness and the distribution of inhomogeneities, in understanding how it interacts with the incident electromagnetic radiation [Onstott, 1979]. At frequencies higher than about 1 GHz, the propagation of electromagnetic energy into sea ice is best

23 mainly controlled by the snow cover. Bare sea disintegrates more quickly than snow- covered sea ice, and this pattern of differential melting will result in the appearance of melt ponds, hummocks, drainageSea ice development channels and weathered ridges [see Figure I.13(e)]

[Tucker et • al.,As the water and air temperatures rise each summer, the sea ice starts to melt. 1992]. Some first-year ice survives the summer melt, becoming thick • Bare sea ice disintegrates faster than snow- multiyear ice, with a top layer transformed into a porous, low salinity cover, and a surface covered sea ice (by differenal melng) resulng in the appearance of melt ponds, drainage relief that becomeschannels, weathered ridges and increasingly modulatedhummock fields by snow deposition. and wind erosion [see

… snow and sea ice have different (solar input) and Figure I.13(c)].thermal conducvies (flux of sensible heat from the atmosphere)

• Ice that survives the summer melt season is called (perennial) mul-year ice: characterized by a porous, low salinity ice with a smooth relief modulated by melt and snow deposion. (a) New Ice (b) First-year ice (c) Multi-year ice

(d) Ridge and Lead e) Melt

Figure I.13 – Development stages of sea ice (Photos courtesy of the Canadian Ice Service and the JPL Polar Oceanography Group)

Dielectric properties of sea ice

The electrical properties of sea ice are relevant, along with structural features such as surface roughness and the distribution of inhomogeneities, in understanding how it interacts with the incident electromagnetic radiation [Onstott, 1979]. At frequencies higher than about 1 GHz, the propagation of electromagnetic energy into sea ice is best

23 2 - Observaons Satellite sensors: recording of reflected, scaered and emied radiaon from the surface.

à Microwaves: - Sea ice concentraon • passive microwaves: Longest - Sea ice moon - radiometers (SSMI, AMSR) record - Snow depth… • acve microwaves: - scaerometers (Quikscat, ASCAT) - Sea ice extent - MY ice fracon - almeters (Cryosat) - SAR à Opcal: • visible imagers (MODIS) - Sea ice temperature Limited - Sea ice/snow albedo • Infrared imagers (AVHRR, VIIRS) by • LIDAR (ICESat) - Sea ice thickness

Higher resoluon opcal, infrared and SAR data are key to visual idenficaon of surface features and ice types (for operaonal ice monitoring), but supplementary in climate studies because of their limited coverage. Opcal signatures VIS1 (0.6 µm) IR4 (10 µm)

- Reflecon of solar radiaon, a direconal funcon:

(depends on sunlight, are pervasive and difficult to detect) (retrieving albedo requires BRDF and atmospheric correcons)

à Sea water albedo (0.1) << à Sea ice albedo (0.5) <<

à Fresh snow albedo (0.9) … very variable Figure I.19 – AVHRR visible and infrared images of Barrow, AK in March of 2003 Opcal reflectance (AVHRR, VIS) 4.2.4 Multisensor approach VIS1 (0.6 µm) IR4 (10 µm)

- IR radiaon depends on surface temperature and emissivity In an attempt to increase the level of information that can be exploited in segmenting the - IR emissivies of ice, snow and sea water are similar (~0.96) data, present trends point at multisensor approaches to sea ice classification using leaving radiaon a strong funcon of physical (skin) simultaneous microwave and optical satellite temperatureobservations, ([Gray et: al., 1992], [Steffen

& Schweiger, 1990], [Drinkwater et al., 1991], [Remund,2000]). (requires cloud masking and atmospheric correcons)

à Sea ice temperature << sea water temperature Table I.4 – Summary of sea ice emission/backscatter signatures Figure I.19 – AVHRR visible and infrared images of Barrow, AK in March of 2003 Opcal radiance (GR = gradient ratio, PR = polarization ration, TB = brightness (~ temperature)Tair) (~ Tfreezing = -1.8°C)

(AVHRR, IR) 4.2.4 Multisensor approach Surface Type Microwave Optical Backscatter Emission Infrared Albedo

In an attempt to increase the level of information that can be exploited in segmenting the Open water Wind Positive GR High TB Low dependent Strong PR data, present trends point at multisensor approaches to sea ice classification using New ice Low Medium simultaneous microwave and optical satellite observations, ([Gray et al., 1992], [Steffen & Schweiger, 1990], [Drinkwater et al., 1991], [RemFirstund,2000]). Year Medium Zero GR Low TB High Ice Zero PR

Multiyear High Negative GR Low TB High Ice Zero PR Table I.4 – Summary of sea ice emission/backscatter signatures

(GR = gradient ratio, PR = polarization ration, TB = brightness temperature)

34 Surface Type Microwave Optical Backscatter Emission Infrared Albedo

Open water Wind Positive GR High TB Low dependent Strong PR New ice Low Medium

First Year Medium Zero GR Low TB High Ice Zero PR Multiyear High Negative GR Low TB High Ice Zero PR

34 backscatter with measurements of brightness temperatures, the ability to classify sea ice improves further [Beaven, 1995]. The currently operational polar observing satellites allow the determination of the extent of the sea ice cover, a classification of sea ice, and delimiting the snow cover over large areas, albeit with a fairly large error for some of the parameters, depending on factors such as time of the year and atmospheric conditions.

temperature T and its directionaltemperature emissivity ∈T (andθ), theits directionallatter intemperature turn emissivity related T to and∈ the(θ its) ,bistatic thedirectional latter in emissivityturn related ∈ to(θ )the, the bistatic latter in turn related to the bistatic 4.2.1 Microwave backscatter radar cross-section (see Section 4.4radar in cross-sectionChapter II). The (see radiance Sectionradar 4.4is usuallycross-section in Chapter expressed II) (see. The asSection radiancea 4.4 isin usually Chapter expressed II). The radianceas a is usually expressed as a

brightness temperature TB, modeledbrightness as (see temperatureAppendix A): T B, modeledbrightness as (see temperature Appendix T A):B, modeled as (see Appendix A): Radar sensors, such as scatterometers or the higher resolutionAs shownsynthetic on aperture the left radar panel in Figure I.18, open water (and new ice) is characterized by a P 1 P 1 P 1 (SAR), use electromagnetic pulses at low microwave frequencies (L, TTGd C ==and Xr bands) to ∈Ω()θθφ TTGd (,)==r ∈Ω ()θθφTTGd== (,)r ∈Ω()θθφ (,) strong polarization,B whereas! [] the differenceB between vertically! [] andB horizontally polarized! [] k ∆νπ4 k ∆νπ4 kB∆νπ4 Ω excite the Earth’s surface. The backscatter cross-section is calculated as: B Ωsurface B Ωsurface surface radiances for other sea ice types remains small. On the basis of this difference, currently P 2 σWhere0 = Pr r is the radiometric power,Where kPBr theis theBoltzmann radiometric constant,Where power, P∆ν r kis B the thethe frequency Boltzmannradiometric constant,power, kB∆ν the the Boltzmann frequency constant, ∆ν the frequency PR22/(4π ) t operational algorithms parameterize the extent of polar sea ice in terms of the bandwidth ofMicrowave signatures the radiometric channel,bandwidth and G( ofθ the,φ) theradiometric receiver bandwidthchannel,antenna gainand of G(pattern.theθ ,radiometricφ) the receiver channel, antenna and gain G(θ pattern.,φ) the receiver antenna gain pattern. Where Pr and Pt are the received and transmitted power and R is the distance to the polarization ratio PR at 19 GHz (surface temperature and emissivity) defined by: PASSIVE surface. ACTIVE Brightness Temperature T 19( V ) −T 19( H ) 19 GHz V 19 GHz HV PR = B 1937 19GHzB GHz V H 1937 GHz HV 37 GHz V T B 19( V ) +TB 19( H )

Observe too that, while the emissivity of first year ice remains quite flat throughout the

microwave spectrum, that of multiyear ice increases steadily, mainly due to stronger

volume scattering contributions from inhomogeneities and snow.

Figure I.16 – RADARSAT (5.3 GHz HH) SAR image of Barrow, AK in March of 2003 Microwave backscaer (Backscatter angles range from 20Figure to 50 degrees) I.17 – SSM/IMicrowave radiaon images of Barrow,Figure I.17AK in– SSM/IMarch imagesof 2003 of Figure Barrow, I.17 AK – SSM/I in March images of 2003 of Barrow, AK in March of 2003

(SAR, C-band HH) (SMMI) The main geophysical parameters that shape the backscatter cross-section, σ0, are a) the As with the radar backscatter, theAs major with thefeatures radar i nbackscatter, microwaveAs the emissivities with major the featuresradar from backscatter, isean microwave ice the majoremissivities features from in microwavesea ice emissivities from sea ice surface dielectric permittivity, Reflecvitywhich acts on the absolute backscatter level, b) theEmissivity surface à 1- Reflecvity FY roughness, which influences- Dielectric permivity the angulargenerally distributio involven of backscatter, strong and versus c) thegenerally weakpresence (surface of involve and strong volume) versusgenerally scattering weak involve (surface and the strong and large v versusolume) weak scattering (surface and and the vlargeolume) scattering and the large difference in dielectric permittivity between sea ice and open water (see Figure I.17, the - Surface roughness & volume difference28 in dielectric permittivitydifferenceinhomogeneies between in sea dielectric ice and permittivity open water between(see Figure sea I.17,ice and the open water (see Figure I.17, the sea ice cover north of Point Barrowsea ice iscover invariably north of w armer Pointsea than Barrow ice the cover is water invariably north opening of Point w armer Barrow than PR = polarizaon rao the is invariably water opening warmer than the water opening • Water permivity >> sea ice permivity GR = gradient rao • Water emissivity increases with frequency (Debye model) extending along the western coast).extending The along algorithms the western that extending characterize coast). The along sea algorithms icethe using western that coast). characterizeMY The algorithms sea ice using that characterize sea ice using • Sea ice emissivity decreases with frequency (volume scaering) microwave radiances are based onmicrowave surface emissivityradiances aredifferences basedmicrowave on at surface separate radiances emissivity frequencies are differencesbased on surface at separate emissivity frequencies differences at separate frequencies • Polarizaon difference is stronger for surface scaering (water) than volume scaering (sea ice) and polarizations (i.e. spectral gradientand polarizations and polar ization(i.e. spectral ratios)and gradientand polarizations they and rely polar on (i.e. aization smallspectral ratios) gradient and theyand polarrely onization a small ratios) and they rely on a small Figure I.18 – Measured microwave emissivities of sea ice [EpplerIce concentraon from et al., 1992] and set of empirically determined distributionset radiances of empirically that of sea corres ice determinedpond species toset speciesin radiances ofPR-GR empirically representative space that corres [Com determined pond ofiso et topassive microwaves al., radiances species 1997] representative that correspond of to species representative of (MY is multiyear ice, FY is first year ice, NI is new ice and OW is water) “pure” multiyear ice, first year “pure” ice and multiyear open water ice, ([Cavalieri first year“pure” iceet al., andmultiyear 1984], open waterice,[Thomas, first ([Cavalieri year ice et and al., open 1984], water [Thomas, ([Cavalieri et al., 1984], [Thomas,

1993]). 1993]). 1993]). The principal basis thus for distinguishing among first and multiyear ice lies in their

differing spectral emissivity gradients, a feature best parameterized by the spectral

gradient ratio GR defined by: 31 31 31 T 37( V ) −T 19( V ) GR = B B TB 37( V ) +TB 19( V )

32 2 - Observaons: Arc sea ice

Sea of Arcc sea ice follows a dominantly Okhotsk ancyclonic wind-driven circulaon (polar subsidence), constrained by land:

• Ice trapped in the Beaufort Siberia Gyre may circulate around the Beaufort Arcc for several years (more me to bump around and Gyre Transpolar grow) Dri

Stream North Water • The Transpolar Dri Stream Hudson pushes some ice against Bay Greenland and the Canadian Baffin Strait Bay Barents Archipelago (thickness Sea increase by compression) and some ice out of the Arcc basin through the Labrador (quickly melted) Sea 0° Arc sea ice Aug 2007 - Dec 2008 From Quikscat scaerometer

Things to note: • During the melng season: - quick sea ice retreat - passing fronts change the appearance of the sea ice cover • Minimum extent: only thick ice survives at the end of the summer. • During the growth season: perennial (MY) and seasonal (FY) ice are disnguished based on their disncve backscaer signatures (bright MY, darker FY) • and TDS export Observations: Cryosphere Chapter 4

4.2.2.2 Longer Records of Ice Extent terrestrial proxies (e.g., Macias Fauria et al., 2010; Kinnard et al., 2011). The records constructed by Kinnard et al. (2011) and Macias Fauria et For climate analysis, the variability of the sea ice cover prior to the al. (2010) suggest that the decline of sea ice over the last few decades commencement of the satellite record in 1979 is also of interest. There has been unprecedented over the past 1450 years (see Section 5.5.2). are a number of pre-satellite records, some based on regionalObservations: obser- In Cryospherea study of the marginal seas near the Russian coastline using ice Chapter 4 vations taken from ships or aerial reconnaissanceArc sea ice extent (e.g., Walsh and extent data from 1900 to 2000, Polyakov et al. (2003) found a low Trends Chapman, 2001; Polyakov et al., 2003) while others were based on frequency multi-decadal oscillation near the that shifted to a Ice concentration dominanta) decadal Annual oscillationice extent in the Chukchi Sea. -10.0 (% per decade) 10.0 18 A more comprehensive1.0 basin-wide record, compiled by Walsh and

) -3.8±0.3 (% per decade) a) Daily ice extent Chapman (2001)2 0.5, showed very little interannual variability until the km 16 last three to6 four decades. For the period 1901 to 1998, their results 0.0 show a summer mode that includes an anomaly of the same sign over ) 2 14 nearly the entire-0.5 Arctic and that captures the sea-ice trend from recent Ice extent

km satellite data. -1.0FigureSummer minimum 4.3 shows an updated record of the Walsh and

anomaly (10 6 Chapman data set with longer time coverage (1870 to 1978) that is 0 12 -1.5 extent record in 2012 more robust because it includes additional historical sea ice observa- tions (e.g., from Danish1980 meteorological1985 1990 stations).1995 2000 A comparison2005 2010of this 10 updated data set with that originally reported by Walsh and Chapman Multiyear ice concentration b) Multiyear ice coverage (Jan-1) (2001) shows similar interannual variability that is dominated by a 5.0 -50.0 (% per decade) 50.0

8 ) Ice extent (1

1979-1988 nearly constant2 extent of the (January–February–March) and 4.5

1989-1998 autumn (October–November–December)km ice cover from 1870 to the 6 6 1999-2008 1950s. The absence4.0 of interannual variability during that period is due 2009-2012 to the use of climatology3.5 to fill gaps, potentially masking the natural signal. Sea ice data from 1900–2011 as compiled6 2 by Met Office Hadley 4 3.0 -0.80±0.2 x10 (km per decade) Centre are also plotted for comparison. In this data set, the 1979–2011 JFMAMJJASOND values were MYice area (10 derived2.5 From from variousQuikscat sources, including satellite data, as described by Rayner2.0 et al. (2003). Since the 1950s, more in situ data are b) WinterSeasonal cycle (DJF) c) Spring (MAM) available and have1980 been homogenized1985 1990 with1995 the satellite2000 2005record (Meier2010 50oN et al., 2012). These data show a consistent decline in the sea ice cover Thickness 60oN c) Ice thickness 90oE that is relatively4.0 moderate during the winter but more dramatic during -1.0 (m per decade) 0.0 the summer months. Satellite data from other sources are also plotted 4 3.5 Winter • in Figure 4.3, including Scanning Multichannel Microwave-0.62 (m per Radiome decade) - Long climate record from passive microwave records ter (SMMR) and3.0 Special Sensor Microwave/Imagerfrom 1979 to present (35 (SSM/I) data using yrs): the Bootstrap 2.5 Algorithm (SBA) as described by Comiso and Nishio (2008) and National Aeronautics and Space Administration (NASA) à 2.0 ICESat Decline in Arcc sea ice extent, with reducons more pronounced in summer. o Team Algorithm (NT1) as described by Cavalieri et al. (1984) (see Sup- 90 W Ice thickness (m) 1.5 Regression of submarine observations plementary Material). DataEM su rvfromeys (Northe thAdvanced Pole) Microwave Scanning à Decline in extent of older and thicker mulyear ice. Radiometer - Earth1.0 Observing System (AMSR-E) using the Bootstrap 1980 1985 1990 1995 2000 2005 2010 d) Summer (JJA) e) Autumn (SON) Algorithm (ABA) and the NASA Team Algorithm Version 2 (NT2) are Drift speed also presented.d) Sea iceThe drif errort speed bars represent one standard deviation of the 4 interannual variability during the satellite period. Because of the use of -1 - Decrease in sea ice extent has been aributed to 2 Buoy drift:thermodynamic effects0.55±0.04 (km day-1 per decade) -2.5 (like increases (km day per decade) 2.5

to) fill data gaps from 1870 to 1953, the error bars in the -1 Walsh and Chapman1 data were set to twice that of the satellite period in surface air temperature and ocean heat storage promong earlier melt) and and 1.5 times higher for 1954 to 1978. The apparent reduction of the dynamical sea ice extent from0 1978 to 1979 is in part due to the change from effects (increased export of older ice and deformaon by wind forcing). surface observations to satellite data. Generally, the temporal distri- butions from the-1 various sources are consistent with some exceptions Sea ice drift speed that may be attributed anomaly (km day to possibleSatellite errors ice in drift the (Oct-May) data (e.g.,: Screen, 2011 0.94±0.3 (km day-1 per decade) and Supplementary-2 Material). Taking this into account, the various sources provide similar1980 basic1985 information1990 1995 and conclusions2000 2005 about2010 the Length of melt season changinge) extent Average and length variability of melt of theseason Arctic sea ice cover. -2.4 -1.6 -0.8 0.0 0.8 1.6 2.4 (day per decade) Trend (% IC yr-1) 130 -30.0 30.0 4.2.2.3 Multi-year/Seasonal Ice Coverage Figure 4.2 | (a) Plots of decadal averages of daily sea ice extent in the Arctic (1979 to 120 1988 in red, 1989 to 1998 in blue, 1999 to 2008 in gold) and a 4-year average daily The winter extent and area of the perennial and multi-year ice cover ice extent from 2009 to 2012 in black. Maps indicate ice concentration trends (1979– in the Central 110Arctic (i.e., excluding Greenland Sea multi-year ice) for 2012) in (b) winter, (c) spring, (d) summer and (e) autumn (updated from Comiso, 2010). 5.7±0.9 (day per decade) 100 325 Length of melt (day) 90 1980 1985 1990 1995 2000 2005 2010

Figure 4.6 | Summary of linear decadal trends (red lines) and pattern of changes in the following: (a) Anomalies in Arctic sea ice extent from satellite passive microwave observa- tions (Comiso and Nishio, 2008, updated to include 2012). Uncertainties are discussed in the text. (b) Multi-year sea ice coverage on January 1st from analysis of the QuikSCAT time series (Kwok, 2009); grey band shows uncertainty in the retrieval. (c) Sea ice thickness from submarine (blue), satellites (black) (Kwok and Rothrock, 2009), and in situ/electro- magnetic (EM) surveys (circles) (Haas et al., 2008); trend in submarine ice thickness is from multiple regression of available observations within the data release area (Rothrock et al., 2008). Error bars show uncertainties in observations. (d) Anomalies in buoy (Rampal et al., 2009) and satellite-derived sea ice drift speed (Spreen et al., 2011). (e) Length of melt season (updated from Markus et al., 2009); grey band shows the basin-wide variability.

331 Arc sea ice moon à Sea ice moon, largely driven by surface winds, affects sea ice concentraon and thickness by convergence/divergence (closing or opening ice) and advecon (export to melng latudes).

• Evoluon into a younger, 1987-2012 thinner pack with narrowing band of old ice along the northern coast of Canada.

• Loss of MY ice is aributed to 1) export through Fram Strait and 2) advecon into the , where it melts in the summer.

• Large variability associated with atmospheric circulaon: Northern Annular Mode NOAA www.climate.gov and [Maslanik, 2011] (NAM) Sea ice moon vectors derived by maximum cross-correlaon of sequenal daily (opcal/microwave) images and buoys… Excursus: Northern Annular Mode A recurrent atmospheric circulaon paern also known as Arcc Oscillaon (AO) or North Atlanc Oscillaon (NAO): – Posive phase: strong zonal winds (jet stream) shied poleward, lowering surface pressure and temperatures in the pole. – Negave phase: weaker zonal winds (jet stream) with greater movement of cold polar air into middle 2650 latudes JOURNAL OF CLIMATE VOLUME 15

à More posive NAM NAM+ induces a small cyclonic trend in a dominant ancyclonic circulaon, thereby producing: - Weakening of the Beaufort Gyre - Shi in TDS towards Fram Strait

Sea ice moon (buoys) and surface FIG.2.AnalyzedfieldsofSLPandSIMforDec1993.Dotsmark positionspressure contours (NWP) [Rigor, 2002] of IABP buoys, and arrows show buoy velocities. Contours NAM+ à increased advecon of sea ice towards the Fram Strait are shown every 2 hPa. NAM- à increased advecon of sea ice towards the Beaufort Sea at Boulder. The ice chart data [sea-ice data on a digital grid (SIGRID)] were produced by the Arctic and Ant- arctic Research Institute in St. Petersburg, and were also obtained from NSIDC. Figure 2 shows the analyzed fields of SLP and SIM for December 1993. For comparison, the buoy positions during that month are marked with dots, and the buoy velocities are shown as black arrows. (Analogous plots for other periods can be obtained from the IABP Web server.) This figure shows the strong correspondence between SLP and SIM noted by Thorndike and Colony (1982), who found that geostrophic winds account for more than 70% of the variance in daily SIM.

3. Climatology The mean (1979–98) fields of SLP and SIM for winter (January–March) and summer (July–September) shown FIG.3.SeasonalmeanfieldsofSLPandSIMfor1979–98:(a) in Fig. 3 exhibit some well-known features of Arctic winter and (b) summer. climate such as the Beaufort high in SLP, which drives the anticyclonic Beaufort gyre in SIM, and the Trans- polar Drift Stream (the zone of high ice velocity across The drift of sea ice across the isobars in these long- the toward Fram Strait). The winter field term means (Fig. 3) reflects the influence of the ocean of SIM exhibits an export of ice into the East Siberian currents upon SIM. On timescales longer than a year Sea from the Arctic Ocean, and import of ice from the the contributions from the winds and ocean currents in Laptev and Kara Seas into the Arctic Ocean. The main driving SIM are roughly equal, but as shown in Fig. 2, sink of ice from the Arctic Ocean is through Fram Strait. the drift of sea ice on shorter timescales (￿1yr)follows In the annual mean, 900 000 km2 of ice flow through the wind (Thorndike and Colony 1989). On short time- Fram Strait to the North Atlantic (Colony and Thorndike scales SIM can be approximated by the simple rule of 1984). The SIM map for summer exhibits a cyclonic thumb that the ice drifts with a speed of about 1% of gyre in the eastern Arctic associated with a low in the and 5￿ to the right of the geostrophic winds (e.g., Thorn- SLP field. Although the Beaufort high is much weaker dike and Colony 1982; Zubov 1943). And as noted by during the summer and has retreated onto land over Aagaard (1989), the drift of sea ice and the surface ocean Alaska, the mean field of SIM in the Beaufort Sea re- currents are decoupled from the deeper currents of the mains anticyclonic. The fields for the transition seasons Arctic Ocean, which tend to flow counterclockwise, op- (not shown) resemble the winter climatology. posite to the flow at the surface. Arc sea ice thickness à Submarine and satellite records suggest that the thickness of Arcc sea ice (hence total volume) is also decreasing.

Satellite LIDAR (ICESat) From ice freeboard Assuming an average density of ice and snow

Submarine From ice dra (1975-2000) L15501 KWOK AND ROTHROCK: ARCTIC SEA ICE THICKNESS L15501 … declassified military data

[Kwok, 2009]

Figure 2. (a) Data points from U.S. Navy cruises used by RPW08, and the data release area (irregular polygon). (b) Interannual changes in winter and summer ice thickness from RPW08 and K09 centered on the ICESat campaigns. Blue error bars show residuals in the regression and quality of ICESat data. (c, d) Spatial patterns of ice thickness in winter (Feb–Mar) and fall (Oct–Dec) of 1988. (e) Mean at summer minimum (1978–2000). (f, g) Spatial patterns of mean winter (Feb–Mar) and fall (Oct–Dec) ice thickness from ICESat (2003–2008). (h) Mean sea ice concentration at summer minimum (2003–2008). Quantities in Figures 2c, 2d, 2f, and 2g are mean and standard deviation of ice thickness within the DRA.

Table 1. Mean Ice Thickness at the End of Melt Season in the Six Regions of the Arctic Ocean From Submarine Cruises in 1958–1976, 1993–1997, and ICESat Acquisitions in 2003–2007a Period Change Period 1, Period 2, Period 3, (2)–(1) (3)–(1) (3)–(2) Region 58–76 93–97 03–07 Thickness Percent Thickness Percent Thickness Percent Chukchi Cap 1.95 0.98 0.70 0.97 50 1.25 64 0.28 29 Beaufort Sea 1.95 0.98 0.97 À0.97 À50 À0.97 À50À 0.00À 0 Canada Basin 3.45 2.05 1.70 À1.40 À40 À1.75 À51 0.35 17 3.77 2.27 1.89 À1.51 À40 À1.89 À50 À0.38 À17 Nansen Basin 3.88 2.05 2.11 À1.83 À47 À1.77 À46À 0.06À 3 Eastern Arctic 3.24 1.30 1.24 À1.94 À60 À2.00 À62 0.06 5 All Regions 3.02 1.62 1.43 À1.40 À46 À1.59 À53 À0.19 À12 À À À À À À aMean ice thickness is shown in meters, and changes in thickness are shown in meters and percent.

3of5 Observations: Cryosphere Chapter 4

Ice concentration a) Annual ice extent -10.0 (% per decade) 10.0 1.0

) -3.8±0.3 (% per decade) 2 0.5 km 6 0.0 -0.5 Ice extent -1.0 anomaly (10 -1.5 1980 1985 1990 1995 2000 2005 2010 Multiyear ice concentration b) Multiyear ice coverage (Jan-1) 5.0 -50.0 (% per decade) 50.0 ) 2 4.5 km 6 4.0 3.5 3.0 -0.80±0.2 x106 (km 2 per decade)

MYice area (10 2.5 2.0 Observations: Cryosphere 1980 1985 1990 1995 2000 2005 2010 Chapter 4 Thickness c) Ice thickness 4.0 -1.0 (m per decade) 0.0 Ice concentration a) Annual3.5 ice exWitentnter -0.62 (m per decade) -10.0 (% per decade) 10.0 3.0 1.0

) 2.5 -3.8±0.3 (% per decade) 2 0.5 km

6 2.0 ICESat

0.0Ice thickness (m) 1.5 Regression of submarine observations EM surveys (North Pole) -0.5 1.0 Ice extent -1.0 1980 1985 1990 1995 2000 2005 2010 anomaly (10 Drift speed 4 d)-1.5 Sea ice drift speed -1 2 Buoy drift: 0.55±0.04 (km day-1 per decade) -2.5 (km day per decade) 2.5

) 1980 1985 1990 1995 2000 2005 2010 -1 1 Multiyear ice concentration b) Multiyear ice coverage (Jan-1) Arc sea ice melt and moon -50.0 (% per decade) 50.0 5.0 0 ) 2 4.5 km

6 -1 Sea ice drift speed 4.0 anomaly (km day Satellite ice drift (Oct-May): 0.94±0.3 (km day-1 per decade) • Trends towards extended duraon of melt season 3.5 -2 1980 1985 1990 1995 2000 2005 2010 6 2 3.0 -0.80±0.2 x10 (km per decade) Length of melt season e) Average length of melt season

MYice area (10 2.5130 -30.0 (day per decade) 30.0 2.0 1201980 1985 1990 1995 2000 2005 2010 From passive microwaves: Thickness c) Ice thick110ness counng from first day of 4.0 -1.0 (m per decade) 0.0 100 5.7±0.9 (day per decade) advance to last day of retreat

3.5Length of melt (day) Winter -0.62 (m per decade) 3.0 90 1980 1985 1990 1995 2000 2005 2010 2.5 Figure 4.6 | Summary of linear decadal trends (red lines) and pattern of changes in the following: (a) Anomalies in Arctic sea ice extent from satellite passive microwave observa- 2.0 ICESat tions (Comiso and Nishio, 2008, updated to include 2012). Uncertainties are discussed in the text. (b) Multi-year sea ice coverage on January 1st from analysis of the QuikSCAT à Ice thickness (m) Regression of submarine observations time series (Kwok, 2009); Most affected areas are the East Siberian and Beaufort Seas: up to grey 1.5band shows uncertainty in the retrieval. (c) Sea ice thickness from submarine (blue), satellites (black) (Kwok and Rothrock, 2009), and in situ/electro- 3 month EM surveys (North Pole) magnetic (EM) surveys (circles) (Haas et al., 2008); trend in submarine ice thickness is from multiple regression of available observations within the data release area (Rothrock et 1.0 al., 2008). Error barslengthening of the ice free summer season with earlier melt and later freeze-up show uncertainties in observations. (d) Anomalies in buoy (Rampal et al., 2009) and satellite-derived sea ice drift speed (Spreen et al., 2011). (e) Length of melt 1980 1985 1990 1995 2000 2005 2010 season (updated from Markus et al., 2009); grey band shows the basin-wide variability. Drift speed d) Sea ice drift speed 4 -1 331 2 Buoy drift: 0.55±0.04 (km day-1 per decade) -2.5 (km day per decade) 2.5 ) -1 1

0

-1 Sea ice drift speed anomaly (km day Satellite ice drift (Oct-May): 0.94±0.3 (km day-1 per decade) -2 1980 1985 1990 1995 2000 2005 2010 Length of melt season e) Average length of melt season à posive trend in dri speed is aributed to a weaker and thinner sea ice cover 130 -30.0 (day per decade) 30.0

(rather than changes in wind speed, in the absence of a definite NAM trend) 120

110

100 5.7±0.9 (day per decade) Length of melt (day) 90 1980 1985 1990 1995 2000 2005 2010

Figure 4.6 | Summary of linear decadal trends (red lines) and pattern of changes in the following: (a) Anomalies in Arctic sea ice extent from satellite passive microwave observa- tions (Comiso and Nishio, 2008, updated to include 2012). Uncertainties are discussed in the text. (b) Multi-year sea ice coverage on January 1st from analysis of the QuikSCAT time series (Kwok, 2009); grey band shows uncertainty in the retrieval. (c) Sea ice thickness from submarine (blue), satellites (black) (Kwok and Rothrock, 2009), and in situ/electro- magnetic (EM) surveys (circles) (Haas et al., 2008); trend in submarine ice thickness is from multiple regression of available observations within the data release area (Rothrock et al., 2008). Error bars show uncertainties in observations. (d) Anomalies in buoy (Rampal et al., 2009) and satellite-derived sea ice drift speed (Spreen et al., 2011). (e) Length of melt season (updated from Markus et al., 2009); grey band shows the basin-wide variability.

331 2 - Observaons: Antarcc sea ice 0° Antarcc • The wind driven ACC is the Circumpolar strongest King Hakon Current system on Earth: sea ice Sea East Wind moon is more intense in Drake Wedell Dri Passage Gyre SH due to higher winds and lower land constraints Antarcc Amery Cooperaon Peninsula Shelf Sea Ronne • Bellings- Shelf East Sea ice moves in a hausen Antarcca clockwise direcon, with a Sea net northward component: Transantarcc freezing in the inner pack is Amundsen Ross mountains Sea Shelf aided by divergence, then Ross advected to the margins Polynya Ross where it melts (*) Gyre • Gyres in the Weddell and Ross Seas form regions of maximum ice export.

(*) “Antarcc sea ice freshwater pump” 2 - Observaons: Antarcc sea ice Aug 2007 - Dec 2008 From Quikscat scaerometer • Lile perennial or MY ice survives (except in the Weddell Sea and in small patches around the coast) resulng in ice that is younger, thinner, warmer, saler, and more mobile than Arcc sea ice.

• Most new ice is formed along the coasts, as the northward- moving ice leaves areas of open water which rapidly refreeze: Ross and Ronne Shelf

• Roaming in summer Antarcc sea ice extent Chapter 4 Observations: Cryosphere Seasonal cycle development but progress is limited by knowledge of snow thickness 20 and the paucity of suitable validation data sets. A recent analysis of the a) Daily ice extent ICESat record by Kurtz and Markus (2012), assuming zero ice freeboard, found• negligibleDecadal monthly averages overlap each other, trends in ice thickness over the 5-year record. ) 2 15 with a slightly overall posive trend overall km

4.2.3.3 Sea Ice Drift 6 0 Using a 19-year data set (1992–2010) of satellite-tracked sea ice 10 motion,• HollandSignificant regional differences: decreases in and Kwok (2012) found large and statistically sig- nificant decadal trends in Antarctic ice drift that in sectors are 1979-1988 caused by changes in local winds. These trends suggest acceleration of Ice extent (1 1989-1998 Amundsen and Bellingshausen Seas and 5 1999-2008 the wind-driven and deceleration of the . The 2009-2012 changes inincreases in the Ross and King Haakon Seas. meridional ice transport affect the freshwater budget near the Antarctic coast. This is consistent with the increase of 30,000 km2 yr–1 in the net area export of sea ice from the Ross Sea shelf coastal 0 polynya region between 1992 and 2008 (Comiso et al., 2011). Assum- JFMAMJJASOND ing an annual average thickness of 0.6 m, Comiso et al. (2011) estimat- b) Winter (JJA) c) Spring (SON) ed an increase in volume export of 20 km3 yr–1 whichTrends is similar to the • Temperatures in the atmosphere and 45oE rate of production in the Ross Sea coastal polynya region for the same Southern Ocean have increased, but period discussed in Section 4.2.3.5. 4.2.3.4 Timing of Sea Ice Advance, Retreat and Ice Season sea ice connues to grow faster than Duration it melts, the causes not yet fully 60oS In the Antarctic there are regionally different patterns of strong change 135oW in ice duration (>2 days yr–1). In the northeast and west Antarctic understood: 50oS Peninsula and southern Bellingshausen Sea region, later ice advance d) Summer (DJF)e) Autumn (MAM) (+61 ±15 days), earlier retreat (–39±13 days) and shorter duration - Snow deposion? (+100 ±31 days, a trend of –3.1 ± 1.0 days yr–1) occurred over the period 1979/1980–2010/2011 (Stammerjohn et al., 2012). These - Ocean fluxes? changes have strong impacts on the marine ecosystem (Montes- 4 - Meltwater from the shelves - stabilizing and cooling the upper ocean surface? Hugo et al., 2009; Ducklow et al., 2011). The opposite is true in the adjacent western Ross Sea, where substantial lengthening of the - Changes in atmospheric circulaon – condioned to ozone depleon and GHG increase? ice season of 79 ±12 days has occurred (+2.5 ± 0.4 days yr–1) due to earlier advance (+42 ±8 days) and later retreat (–37 ±8 days). Patterns of change in the relatively narrow East Antarctic sector are generally of a lower magnitude and zonally complex, but in certain regions involve changes in the timing of sea ice advance and retreat -2.4 -1.6 -0.8 0.0 0.8 1.6 2.4 –1 Trend (% IC yr-1) of the order of ±1 to 2 days yr (for the period 1979–2009) (Massom Figure 4.7 | (a) Plots of decadal averages of daily sea ice extent in the Antarctic et al., 2013). (1979–1988 in red, 1989–1998 in blue, 1999– 2008 in gold) and a 4-year average daily ice extent from 2009 to 2012 in black. Maps indicate ice concentration trends 4.2.3.5 Antarctic Polynyas (1979–2012) in (b) winter, (c) spring, (d) summer and (e) autumn (updated from Comiso, 2010). Polynyas are commonly found along the coast of . There are two different processes that cause a polynya. Warm water upwelling which routine observations of sea ice and snow properties were made. keeps the surface water near the freezing point and reduces ice pro- Their compilation included a gridded data set that reflects the regional duction (sensible heat polynya), and wind or ocean currents move ice differences in sea ice thickness. A subset of these ship observations, away and increase further ice production (latent heat polynya). and ice charts, was used by DeLiberty et al. (2011) to estimate the annual cycle of sea ice thickness and volume in the Ross Sea, and to An increase in the extent of coastal polynyas in the Ross Sea caused investigate the relationship between ice thickness and extent. They increased ice production (latent heat effect) that is primarily respon- found that maximum sea ice volume was reached later than maxi- sible for the positive trend in ice extent in the Antarctic (Comiso et mum extent. While ice is advected to the northern edge and melts, the al., 2011). Drucker et al. (2011) show that in the Ross Sea, the net interior of the sea ice zone is supplied with ice from higher latitudes ice export equals the annual ice production in the Ross Sea polynya and continues to thicken by thermodynamic growth and deformation. (approximately 400 km3 in 1992), and that ice production increased by Satellite retrievals of sea ice freeboard and thickness in the Antarctic 20 km3 yr–1 from 1992 to 2008. However, the ice production in the Wed- (Mahoney et al., 2007; Zwally et al., 2008; Xie et al., 2011) are under dell Sea, which is three times less, has had no statistically significant

332 Antarcc sea ice dri and thickness

Trends in sea ice moon (arrows) Mean sea ice thickness over vs. trends in sea ice concentraon (colors) LETTERS NATURE GEOSCIENCE2003-2008 from DOI: 10.1038/NGEO1627ICESat

a in the ice-melting zone. In areas of mean northerly winds King Haakon Sea (Bellingshausen, Cosmonaut and Dumont D’Urville seas; Fig. 1a) southward advection opposes thermodynamic growth of the ice cover, and freezing extends closer to the ice edge. The mean concentration difference over autumn is dominated by freezing, with advection and divergence being minor contrib- utors during this period (Supplementary Fig. S2). However, in 0.02 the Pacific sector and Weddell Sea, trends in the autumn con- centration difference seem to be strongly influenced by dynamics 0.0 yr¬1 (Supplementary Fig. S3). In contrast, trends in the King Håkon Sea are controlled by thermodynamics. Supplementary Fig. S4 ¬0.02 shows the proportion of the autumn concentration difference 0.005 m s¬1 yr¬1 trend that is explained by trends in dynamical processes. The ratio is noisy, but after heavy smoothing, it confirms that dynamic trends dominate in the Pacific sector. Trends in freezing in this sector can actually oppose the ice-concentration changes, because dynamical processes are progressively replacing thermodynamics. Ice-concentration losses in the Weddell Sea also seem to be caused by decreased northward advection, but the concentration increase in King Håkon Sea and other changes around East Antarctica Ross Sea contain a strong thermodynamic[Kurtz, 2012] component. The wind trends in these regions suggest that changes in cold- and warm-air advection b [Holland, 2012] explain the thermodynamic trends. The ultimate cause of the wind and ice changes lies in the large- scale climate variability of the Southern Hemisphere. Antarctic sea Wind driven changes: ice can contain 3–5-year cyclic anomalies that might be partly aliased into our calculations1,16,21, but our trends cover several such cycles and• are consistentLimited informaon on ice thickness with longer-term studies18. Aspects • Dri à acceleraon of the Ross gyre and of the wind trends (and therefore ice-motion trends) can be attributed to large-scalechange (satellite retrievals, under modes such as the Southern Annular deceleraon of the Weddell gyre 0.6 Mode and El Niño/Southern Oscillation3,19,22. Modern trends in 0.0 hPa yr¬1 these modes could arisedevelopment, are limited by knowledge of through natural variability, but some • Concentraon à increased ice producon in evidence suggests that they are forced by the Southern Hemisphere 4,23 ¬0.6 ozone hole and increasedsnow thickness, and the record is short) greenhouse gases . Our conclusions the Ross Shelf and increased cold air that ice-motion trends are dominated by winds, and that winds 0.1 m s¬1 yr¬1 contribute significantly to ice concentration trends through both advecon over King Haakon Sea dynamic and thermodynamic effects, reinforce the need for a better understanding of both the wind changes and the anthropogenic forcing of relevant climate modes. Our conclusions are of fundamental importance in rectifying the failure of present climate models to hindcast the recent increase in Antarctic sea ice11. In particular, they suggest that surface winds and ice dynamics and thermodynamics must be accurately represented. Our data set provides an observational map of changes against which models can be compared, and any faults Figure 3 Autumn (April–June) 1992–2010 ice motion and concentration can be diagnosed using our decomposition of the ice-concentration trends and| their relation to wind forcing. Wind-driven changes in ice budget into dynamic and thermodynamic components. When motion are clearly linked to changes in ice concentration. a, Ice-motion climate models can hindcast ice-concentration increases we will trend vectors overlaid on ice-concentration trends. b, ERA-Interim 10-m have good reason to believe their forecasted ice loss under the wind trend vectors overlaid on trend in sea-level pressure. White, grey and effects of climate change. black contours show underlay field trends significant at 90%, 95% and Our data offer a new view of surface change relevant to all 99% respectively; black vectors have meridional trends significant at components of the Antarctic climate. The good fit between ice >90%; magenta contour in b shows extent of concentration trends. motion and reanalysis wind trends, in an area of extremely sparse in situ data, is testament to the power of satellite sounder data Before examining trends, it is instructive to first decompose the assimilation into ERA-Interim. It also implies confidence that these mean April–October ice-concentration budget20 (Supplementary winds can be used to force models of Antarctic ice and ocean Fig. S1). Freezing in the inner pack is maintained by divergence, trends over recent decades. The large and widespread changes in which supports greater ocean–atmosphere heat exchange than is ice motion imply considerable changes in sea-surface forcing, both otherwise possible in consolidated ice. In the main export regions, directly through observed changes in ice stress and indirectly by ice is advected to the margins and then melted, because it is validating trends in reanalysis wind stress. The ice-motion trends thermodynamically unsustainable there even in winter; the ice suggest that increased cyclonic forcing has accelerated the Ross cover in these regions extends hundreds of kilometres further Gyre, supporting its possible involvement in ice-sheet melting and equatorward than it would in the absence of northward advection. Ross Sea freshening24. The decrease in Weddell Sea ice cyclonicity This illustrates the Antarctic sea-ice freshwater pump, which suggests a Weddell Gyre deceleration, implying that gyre changes contributes brine to Antarctic Bottom Water close to the continent alone cannot explain the warming of Antarctic Bottom Water and to Antarctic Intermediate Water and mode exported to the abyssal Atlantic25. Changes in meridional ice

874 NATURE GEOSCIENCE VOL 5 DECEMBER 2012 www.nature.com/naturegeoscience | | | [Thompson, 2002] Excursus: Southern Annular Mode 1979-2000 Geopotenal Height at 500 hPa à ice-moon trends in Antarcca are dominated by winds, which contribute significantly to ice concentraon trends through both dynamics (advecon, convergence) and thermodynamic effects (cold/warm air advecon)

The atmospheric circulaon over the southern ocean has definitely moved towards a posive SAM phase (stronger surface westerlies, aributed in large part to ozone depleon in the stratosphere), Surface Air with associated decreases in mean temperature and Temperature pressure in the pole:

“Ozone depleon à stratospheric cooling à strengthening of vortex à intensificaon of surface westerly winds à increase in zonal winds stress à drives an ekman response (intensified northward velocity) à increased ice advecon and producon”

Also linked to warming of Antarcc Peninsula Summary observaons

• Over the satellite period (1979 to present): – Decrease in annual mean Arcc sea ice extent, parcularly strong in summer, with accompanying reducon in average winter sea ice thickness and an increase in sea ice dri speeds. – Lengthened annual period of Arcc surface melt over the satellite period. – Mild but significant increase in Antarcc sea ice extent, with acceleraon of the Ross Gyre and deceleraon of the Weddell Gyre. 3 – Sea ice models: dynamics and thermodynamics • Examples: – CICE, Los Alamos Sea Ice Model (Hunke & Lipscomb, 2013) in CESM – LIM, Lovain-La-Neuve Sea Ice Model (Vancoppenolle et al., 2009) in NEMO

• All dynamic-thermodynamic models: – Vercal thermodynamics: growth/melt by heat exchange (Maykut, 1971) – Horizontal dynamics: dri and deformaon through ice moon (Hibler, 1979)

• With a subgrid-scale ice thickness distribuon g(x,h,t) : !g ! " " Determines ice fracons in = " (hg! ) " u #$g " g$ #u #+% each thickness category !t !h Each category prescribes surface dynamics and interior ice characteriscs dh/dt: rate of 1) advecon (roughness, albedo, salinity, …) thermodynamic 2) convergence/divergence of the pack ice growth/melt 3) mechanical redistribuon Ψ (ridging/ raing) Solved in terms of ice velocity field Vercal thermodynamics

à Sea ice growth is controlled by the balance of fluxes at the upper (air-ice) and lower (ice-ocean) interfaces:

Shortwave solar radiaon (Fsw) c f α = reflected fracon (albedo, cf) F i = absorbed fracon F LWé 0 LWê α FSW Fsh+Flh Longwave infrared radiaon: T air FLWê = incoming = FLWê(Tair, albedo, cf) F = outgoing = ε σ T 4 T0 snow LWé 0 Fc ice Turbulent mixing of sensible and latent heat: i0 T(h) Fsh = sensible heat flux =Fsh(uair, Tair-T0, z0) Flh = latent heat flux = Flh (uair, qair-q0, z0) F * F *= ocean heat flux = F *(u , T -T ) c T sh sh ice h ocean h uair = wind speed Tocean Conducon: uice = ice dri q = air specific humidity Fc = top = k dT/dz|0 air F * z0 = surface roughness length sh Fc*= boom = k dT/dz|h cf = cloud fracon ε = ice emissivity

Excursus: turbulent mixing

Determines the exchange of heat and momentum between the air-ice and ice-ocean components. Accomplished by small-scale turbulence or eddies. Usually parameterized using bulk formulas: Most used is the Classical aerodynamic method Monin-Obukhov Similarity Theory Sensible heat (air/ice) Fsh ! CT uair (Tair " T0 ) F ! C u q " q Ablaon of surface Latent heat (air/ice) lh q air ( air 0 ) ice by sublimaon

* * Sensible heat (ice/ocean) F sh ! CT uwi (Tocean " Th ) ! ! C u u Stress components (air/ice) ! air " D air air ! ! C* u u Stress components (ice/ocean) ! ocean " D wi wi

Bulk transfer coefficients Determined Driving gradient empirically (depend on stability, surface roughness) background mean speed Vercal thermodynamics Energy fluxes dT F = (1!" )#(1! i )# F + F ! F ! F ! F ! k on top surface TOP 0 SW LW $ LW % sh lh dz 0

Thermal balance: dh F F (T ) 0 F (T = 0! C) > 0? = ! TOP Surface TOP 0 = TOP 0 melt dt qmelt

FTOP Heat conducon T0 (ADJUSTED) (Determines the temperature profile) snow ice "T " "T !c = k + QSW (z) h "t "z "z T(h) QSW = absorbed solar heat capacity (c) and thermal conducvity (k) depend on ice temperature and salinity Th = -1.8 C (FIXED) FBOTTOM dh FBOTTOM Boom Energy fluxes on * dT = ! F = F ! k growth/melt boom surface BOTTOM sh dz dt qmelt Simplified model of thermodynamic growth

Suppose that the sea ice growth rate is determined by the Tair difference between Tair and Tocean:

Fc h Fc = !keff (Tair ! Tocean ) / h Fourier’s Law

dh dh keff F = q = ! (Tair ! Tocean ) / qmelt Tocean c melt dt dt h

Where keff is the effecve sea ice thermal conducvity and qmelt = ρice x Lfusion is the energy required to melt a unit volume of sea ice, implying 1) Thin ice grows the fastest Latent heat 2) Growth rate decreases with ice thickness of fusion 3) Snow delays ice growth (insulang effect by lower conducvity)

à A thermodynamic equilibrium thickness is reached when ice becomes thick

enough that no heat can be conducted through it: this is about 3 meters in the Arcc (ridging can boost this up to 10 m), and about 1-2 meters in the Antarcc. … snow depths and ocean heat fluxes are larger in the Antarcc A = ice concentraon Horizontal dynamics f = parameters η = sea surface height Wind is the primary force responsible for sea ice moon: σ = internal stress tensor ! conservaon !u ! ! ! ! ! ! ! of linear m = A(" air +" ocean ) # mf k $ u # mg%& + % '( momentum !t Drag force Coriolis Ocean lt Internal stress

Changes in ice momentum due to air and ocean stresses -Tension/compression (τair, τocean) go against the and ocean surface -Shear lts, and are lost to internal stress. ! ! ! = ! ("";P) 2 σshear Rheology: establishes the relaon P 1! e σ between strain rate (spaal σc σnormal derivaves of ice moon) and P e = 2 internal stress. !! isotropic plasc yield curve σc(P) 1 # "u "u & - weak under tension viscous-plasc i j !!ij = + - strong in shear connuum % ( 2 $ "x j "xi ' - strongest under compression. à P (ice strength) depends on thickness and concentraon… Largest deformaon rates are localized along linear features… (ridges/cracks)

Ellesmere Nares Island strait (Canada) Fram Strait Greenland

hp://ocean.dmi.dk from MODIS

Chapter 2 Sea ice environment

equilibrium. The central part of the ridge along the water line is refrozen forming the consolidated layer of the ridge. A typical sketch of the cross-section of a first-year ridge together with the basic geometrical parameters that describe the ridge shape is schematically shown in Fig. 2.1. Fig. 2.2 shows photo of first-year ice ridges from the North-Western Barents Mechanical redistribuon Sea area.

Solved aer horizontal dynamics (strain rate)

Redistribuon funcon ψ:

(h,g, !) ( !) (h) ( !) w (h,g) ! " = # 0 " $% +# r " $ r [Shafrova, 2007]

1) α0, αr are the lead opening and closing rates. Fig. 2.1. Principal cross section sketch of an ice ridge.

• Divergence and shear create open water … from the spaal derivaves of ice moon • Convergence and shear produce thicker ice by raing and ridging ! = Mechanical redistribution 2) wr(h,g) describes mechanical redistribuon onto different thickness categories g(h)dh • Thinnest 15% is lost to deformaon • Different redistribuon funcons: – uniform up to certain liming height Fig. 2.2. Photo of first-year ridges in the Barents Sea, May 2005. g(h)dh [Bailey, 2010] – exponenal distribuon – remains an open queson

h

6 Summary models

• In addion to core dynamic and thermodynamic components, sea ice models include a large number of parameterizaons. • Major sea ice model uncertaines affect: – Mechanical redistribuon under deformaon – Formulaon of internal stress (anisotropy) – Surface processes like snow depth and melt ponds • Also connected to improvements in modeling of: – Atmosphere (winds, cloud cover, carbon deposion) – Ocean (heat advecon and mixing) • Newer developments: – Sea ice microstructure evoluon – Biochemistry

Evaluation of Climate Models Chapter 9

A further persistent problem is insufficient marine stratocumulus cloud ocean heat uptake, sea level rise, and coupled modes of variability. in the eastern tropical Pacific, caused presumably by weak coastal There is little evidence that CMIP5 models differ significantly from upwelling off South America leading to a warm SST bias (Lin, 2007). CMIP3, although there is some evidence of modest improvement. Many Although the problem persists, improvements are being made (Achuta- improvements are seen in individual CMIP5 ocean components (some Rao and Sperber, 2006). now including interactive ocean biogeochemistry) and the number of relatively poor-performing models has been reduced (thereby reducing 9.4.2.5.2 Tropical inter-model spread). New since the AR4, process-based model evalua- tion is now helping identify the cause of some specific biases, helping CMIP3 and CMIP5 models exhibit severe biases in the tropical Atlantic to overcome the limits set by the short observational records available. Ocean, so severe that some of the most fundamental features—the east–west SST gradient and the eastward shoaling thermocline along 9.4.3 Sea Ice the equator—cannot be reproduced (Figure 9.14; (Chang et al., 2007; Chang et al., 2008; Richter and Xie, 2008; Richter et al., 2013). In many Evaluation of sea ice performance requires accurate information on ice models, the warm SST bias along the Benguela coast is in excess of concentration, thickness, velocity, salinity, snow cover and other fac- 5°C and the Atlantic warm pool in the western basin is grossly under- tors. The most reliably measured characteristic of sea ice remains sea estimated (Liu et al., 2013a). As in the Pacific, CMIP3 models suffer the ice extent (usually understood as the area covered by ice with a con- 9 double ITCZ error in the Atlantic. Hypotheses for the complex Atlantic centration above 15%). Caveats, however, exist related to the uneven bias problem tend to draw on the fact that the Atlantic Ocean has a far reliability of different sources of sea ice extent estimates (e.g., satellite smaller basin, and thus encourages a tighter and more complex land– vs. pre-satellite observations; see Chapter 4), as well as to limitations of atmosphere–ocean interaction. A recent study using a high-resolution this characteristic4 – Sea ice model performance as a metric of model performance (Notz et al., 2013). coupled model suggests that the warm eastern equatorial Atlantic SST bias is more sensitive to the local rather than basin-wide trade wind bias and to a wet Congo basin instead of a dry Amazon—a finding that differs from previous studies (Patricola et al., 2012). Recent ocean ) model studies show that a warm subsurface temperature bias in the 2 km eastern equatorial Atlantic is common to virtually all ocean models 6 • Evaluaon of sea ice model performance forced with ‘best estimated’ surface momentum and heat fluxes, owing to problems in parameterization of vertical mixing (Hazeleger and requires informaon on ice concentraon, Haarsma, 2005). Toniazzo and Woolnough (2013) show that among a tent (10 variety of causes for the initial bias development, ocean–atmosphere velocity, thickness, salinity, snow cover and coupling is key for their maintenance.

Sea ice ex other factors… 9.4.2.5.3 Tropical

CMIP3 and CMIP5 models simulate equatorial Indian Ocean climate à Thickness from satellites (ICEsat, Cryosat): reasonably well (e.g., Figure 9.14), though most models produce weak westerly winds and a flat thermocline on the equator. The models show uncertaines are significant, spaal coverage a large spread in the modelled depth of the 20°C isotherm in the east- limited and the length of the me series is too ) ern equatorial Indian Ocean (Saji et al., 2006). The reasons are unclear 2 km but may be related to differences in the various parameterizations of short to derive trends. However, thickness is in 6 vertical mixing as well as the wind structure (Schott et al., 2009). general not well reproduced by models…

CMIP3 models generally simulate the Seychelles Chagos thermocline tent (10 ridge in the Southwest Indian Ocean, a feature important for the Indian monsoon and tropical cyclone activity in this basin (Xie et al., 2002). The à Regarding sea ice extent: mulmodel mean models, however, have significant problems in accurately representing Sea ice ex its seasonal cycle because of the difficulty in capturing the asymmetric errors do not exceed 10-15% (CMIP5) nature of the monsoonal winds over the basin, resulting in too weak a semi-annual harmonic in the local Ekman pumping over the ridge region compared to observations (Yokoi et al., 2009b). In about half of the models, the thermocline ridge is displaced eastward associated Figure 9.22 | Mean (1980–1999) seasonal cycle of sea ice extent (the ocean area with with the easterly wind biases on the equator (Nagura et al., 2013). a sea ice concentration of at least 15%) in the Northern Hemisphere (upper) and the Southern Hemisphere (lower) as simulated by 42 CMIP5 and 17 CMIP3 models. Each model is represented with a single simulation. The observed seasonal cycles (1980– 9.4.2.6 Summary 1999) are based on the Hadley Centre Sea Ice and Sea Surface Temperature (HadISST; Rayner et al., 2003), National Aeronautics and Space Administration (NASA; Comiso There is high confidence that the CMIP3 and CMIP5 models simulate and Nishio, 2008) and the National Snow and Ice Data Center (NSIDC; Fetterer et al., the main physical and dynamical processes at play during transient 2002) data sets. The shaded areas show the inter-model standard deviation for each ensemble. (Adapted from Pavlova et al., 2011.)

787 Evaluation of Climate Models Chapter 9 ) !"#$%&'(")*"+*!,-./*%)0*!,-.1*2 * km

Sea ice model trends 6

September Ice Extenttent (10

Clear response in !"#$%&'!!( Sea ice ex Arcc trends to !"#$%&$'()(à Regarding Arcc ice: models tend *+,-.-,/-01'$2( GHG loading 3456,7(((()(to underesmate the observed *6,/.-,8-01'$2(( 93:9;<=)(*6,>.-,>701'$2( decrease in summer minimum 9 extent. ) 2

km !")"à%&'!! ( CMIP5 Sll, a number of models already 6 CMIP3 !"#$%&$'()(predict ice free Arcc summer *<8,-.<,6+01'$2(

tent (10 3456,7(((()(condions by 2050. *?,>.-,8@01'$2(( 93:9;<=)(*/,/.-,6>01'$2( Sea ice ex CMIP5 Arcc (le) and Antarcc (right) summer ice extent trends (1979-2010) From 1953-2011 RCP4.5 multi-model(c) CMIP5 meanArctic ice extent S eptembersuggests trends (1979 - 2010) 51%(d) of CMIP5 Antarctic ice extent February trends (1979 - 2010) à Regarding Antarcc summer: the observed trend is externally forced. From 1979-2011 it is the majority of models exhibit 63%. decreasing trends, in contrast to the observed mild increase. Are discrepancies an imprint of natural variability or model shortcomings?

(106 km2 per decade) (106 km2 per decade)

Figure 9.24 | (Top and middle rows) Time series of sea ice extent from 1900 to 2012 for (a) the Arctic in September and (b) the Antarctic in February, as modelled in CMIP5 (coloured lines) and observations-based (NASA; Comiso and Nishio, 2008) and NSIDC; (Fetterer et al., 2002), and dashed thick black lines, respectively). The CMIP5 multi- model ensemble mean (thick red line) is based on 37 CMIP5 models (historical simulations extended after 2005 with RCP4.5 projections). Each model is represented with a single simulation. The dotted black line for the Arctic in (a) relates to the pre-satellite period of observation-based time series (Stroeve et al., 2012). In (a) and (b) the panels on the right are based on the corresponding 37-member ensemble means from CMIP5 (thick red lines) and 12-model ensemble means from CMIP3 (thick blue lines). The CMIP3 12-model means are based on CMIP3 historical simulations extended after 1999 with Special Report on Emission Scenarios (SRES) A2 projections. The pink and light blue shadings denote the 5 to 95 percentile range for the corresponding ensembles. Note that these are monthly means, not yearly minima. (Adapted from Pavlova et al., 2011.) (Bottom row) CMIP5 sea ice extent trend distributions over the period 1979–2010 for (c) the Arctic in September and (d) the Antarctic in February. Altogether 66 realizations are shown from 26 different models (historical simulations extended after 2005 with RCP4.5 projections). They are compared against the observations-based estimates of the trends (green vertical lines in (c) and (d) from Comiso and Nishio (2008); blue vertical line in (d) from Parkinson and Cavalieri (2012)). In (c), the observations-based estimates (Cavalieri and Parkinson, 2012; Comiso and Nishio, 2008) coincide.

789 Summary model performance • Large-scale sea ice processes like basic thermodynamics are well understood and represented: – Current climate models reproduce seasonal cycle of Arcc/ Antarcc sea ice extent to within 10-15%, although regional distribuons of sea ice concentraon are poorly simulated. – Underesmaon of trend in Arcc summer extent – Wrong trend in Antarcc summer extent • Details of sea ice dynamics and deformaon not well captured, spaal thickness distribuons in the ballpark but poorly simulated. • Simulaon of sea ice is affected by errors in the representaon of the atmosphere and ocean, some feedbacks sll poorly understood. • Model vs. data comparisons connue to be a rich area for research. Quesons?