East Antarctic Landfast Sea Ice Distribution and Variability, 2000–08

15 FEBRUARY 2012 F R A S E R E T A L . 1137

ALEXANDER D. FRASER Antarctic Climate & Ecosystems Cooperative Research Centre, and Institute for Marine and Antarctic Studies, University of Tasmania, Hobart, Tasmania, Australia

ROBERT A. MASSOM Antarctic Climate & Ecosystems Cooperative Research Centre, University of Tasmania, Hobart, and Australian Antarctic Division, Kingston, Tasmania, Australia

KELVIN J. MICHAEL Antarctic Climate & Ecosystems Cooperative Research Centre, and Institute for Marine and Antarctic Studies, University of Tasmania, Hobart, Tasmania, Australia

BENJAMIN K. GALTON-FENZI AND JAN L. LIESER Antarctic Climate & Ecosystems Cooperative Research Centre, University of Tasmania, Hobart, Tasmania, Australia

(Manuscript received 21 December 2010, in ﬁnal form 20 July 2011)

ABSTRACT

This study presents the first continuous, high spatiotemporal resolution time series of landfast sea ice extent along the East Antarctic coast for the period March 2000–December 2008. The time series was derived from consecutive 20-day cloud-free Moderate Resolution Imaging Spectroradiometer (MODIS) composite images. Fast ice extent across the East Antarctic coast shows a statistically significant (1.43% 60.30% yr21) increase. Regionally, there is a strong increase in the Indian Ocean sector (208–908E, 4.07% 60.42% yr21), and a nonsignificant decrease in the western Pacific Ocean sector (908–1608E, 20.40% 60.37% yr21). An apparent shift from a negative to a positive extent trend is observed in the Indian Ocean sector from 2004. This shift also coincides with a greater amount of interannual variability. No such shift in apparent trend is observed in the western Pacific Ocean sector, where fast ice extent is typically higher and variability lower than the Indian Ocean sector. The limit to the maximum fast ice areal extent imposed by the location of grounded icebergs modulates the shape of the mean annual fast ice extent cycle to give a broad maximum and an abrupt, relatively transient minimum. Ten distinct fast ice regimes are identified, related to variations in bathymetry and coastal configuration. Fast ice is observed to form in bays, on the windward side of large grounded icebergs, between groups of smaller grounded icebergs, between promontories, and upwind of coastal features (e.g., glacier tongues). Analysis of the timing of fast ice maxima and minima is also presented and compared with overall sea ice maxima/minima timing.

1. Introduction 2001; World Meteorological Organization 1970). It is a preeminent feature of the Antarctic coastal zone and an Landfast sea ice (fast ice) is sea ice that is held sta- important interface between the ice sheet and pack ice/ tionary (fast) by being attached to coastal features (e.g., ocean. The reliance of fast ice upon these coastal fea- the shoreline, glacier tongues, and ice shelves), groun- tures as anchor points means that it tends to form in ded icebergs, or grounded over shoals (Massom et al. narrow bands of widely varying widths but rarely ex- ceeding 150 km around East Antarctica (Giles et al. 2008). There are strong hemispheric contrasts in fast ice Corresponding author address: Alexander D. Fraser, Antarctic Climate and Ecosystems Cooperative Research Centre, University extent and persistence. In the Arctic, a lack of grounded of Tasmania, Private Bag 80, Hobart, TAS 7001, Australia. icebergs means fast ice typically grounds itself in shallow E-mail: [email protected] waters, with the seaward fast ice edge often located around

DOI: 10.1175/JCLI-D-10-05032.1

Ó 2012 American Meteorological Society Unauthenticated | Downloaded 10/02/21 01:08 PM UTC 1138 JOURNAL OF CLIMATE VOLUME 25 the 20–30-m isobath (Mahoney et al. 2007a; Lieser 2004). techniques applied to Synthetic Aperture Radar (SAR) By contrast, Antarctic icebergs often ground in water image pairs. Massom et al. (2009) created a time series of depths of 400–500 m (Massom et al. 2001) and act as fast fast ice extent off the Ade´lie Land coast from 1348– ice anchor points, giving much larger fast ice extents. 1438E using cloud-free National Oceanic and Atmo- Such shoals also occur some distance offshore. spheric Administration (NOAA) Advanced Very High Variability in fast ice extent is important for a number Resolution (AVHRR) data from 1992 to 1998. Several of reasons. It is likely a sensitive indicator of climate other studies have used remote sensing techniques to change (Heil et al. 2006; Mahoney et al. 2007a; Murphy obtain regional-scale information on fast ice formation et al. 1995) and is also closely associated with coastal and breakout, primarily in and around Lu¨ tzow-Holm polynyas. Coastal polynyas have far-reaching conse- Bay (Enomoto et al. 2002; Mae et al. 1987; Ushio 2006, quences in terms of Antarctic Bottom Water formation 2008) and the western Ross Sea (Brunt et al. 2006). While and hence global thermohaline circulation (Massom detailed studies of Arctic fast ice formation, breakup, and et al. 1998; Rintoul 1998; Tamura et al. 2008; Williams extent have taken place on regional scales (e.g., Mahoney et al. 2008). A recent study (Massom et al. 2010) has et al. 2007a), there has been no similar Antarctic study shown that fast ice acts to mechanically stabilize fragile providing large-scale information on fast ice distribution glacier tongues and ice shelves, delaying calving and and its spatiotemporal variability. ultimately affecting ice sheet mass balance, as well as The East Antarctic coast has persistent cloud cover, prolonging the residence times of ungrounded icebergs averaging 93% in October compared with a global av- (Massom 2003). Despite the physical signiﬁcance of fast erage of 70% (Spinhirne et al. 2005). Thus, a major ice, it is currently not represented in global climate cir- challenge when using visible–thermal infrared (TIR) culation models or coupled ice–ocean–atmosphere models. time series (which are heavily affected by cloud)— such Fast ice also has several important biological functions as the National Aeronautics and Space Administration at various trophic levels. It forms a habitat for micro- (NASA) Aqua and Terra Moderate Resolution Imaging organisms (e.g., McMinn et al. 2000) and plays a crucial Spectroradiometer (MODIS) instruments—is to derive role in the life cycle of Emperor penguins and Weddell high-quality imagery of the surface (i.e., the sea ice zone) seals (Massom et al. 2009; Kooyman and Burns 1999). by appropriate treatment of cloud cover. Fraser et al. Antarctic fast ice also affects logistical operations, acting (2009) presented an algorithm to effectively produce to both facilitate and impede navigation and base resupply. cloud-free composite imagery of the East Antarctic sea To date, research has focused on overall sea ice extent ice zone on a temporal scale suitable for studying large- without discriminating between pack ice and fast ice scale fast ice formation–breakout events. Fraser et al. (e.g., Comiso 2009; Comiso and Nishio 2008; Cavalieri (2010) used these MODIS composite images, along with and Parkinson 2008; Lemke et al. 2007; Zwally et al. 6.25-km resolution NASA Advanced Microwave Scan- 2002). Relatively little is known about larger-scale as- ning Radiometer for Earth Observing System (EOS) pects of Antarctic fast ice distribution and its spatio- (AMSR-E) passive microwave sea ice concentration temporal variability. In particular, little is known about composite images, to develop methods for retrieval of the atmospheric and oceanic controls on the growth and fast ice extent along the East Antarctic coast. The work breakout of fast ice around the Antarctic coast. Much of presented in this paper directly builds upon these earlier the research has largely focused on the acquisition and works (Fraser et al. 2009, 2010), to retrieve the ﬁrst high analysis of measurements of physical or biological as- spatiotemporal resolution (2 km, 20 day) maps of East pects of fast ice localized close to bases—for example, Antarctic fast ice, during the period from March 2000 to Heil (2006), Kawamura et al. (1997), Lei et al. (2010), December 2008, providing information on the season- Ohshima et al. (2000), Purdie et al. (2006), Smith et al. ality of fast ice occurrence. A further major aim is to (2001), Tang et al. (2007), and Uto et al. (2006). Several provide large-scale fast ice information in support of regional-scale studies of Antarctic fast ice have been detailed but localized data acquired within the newly- conducted using remote sensing techniques, but none established Antarctic Fast Ice Network (AFIN) (Heil have combined large-scale coverage, long time series, and et al. 2011). The MODIS dataset was chosen because high spatiotemporal resolution. Kozlovsky et al. (1977) it combines excellent polar synoptic coverage (a 2330- studied East Antarctic fast ice on a broad spatial scale km-wide swath) with moderate resolution (1 km in (08–1608E), but sampling was sparse and sporadic and did TIR bands) and is readily available. This dataset also not resolve fast ice formation or decay. has broad and regular coverage of the narrow but Giles et al. (2008) presented two snapshots of East zonally extensive fast ice zone—a zone that cannot Antarctic (888–1708E) fast ice extent from November easily be covered and monitored by satellite SAR data 1997 and 1999, derived using image cross-correlation and remains largely unresolved in lower-resolution

Unauthenticated | Downloaded 10/02/21 01:08 PM UTC 15 FEBRUARY 2012 F R A S E R E T A L . 1139

composite images were produced during the summer and early autumn–late spring, and TIR composites were produced when solar illumination was lacking.

AMSR-E sea ice concentration composite images (6.25-km resolution, Spreen et al. 2008) were also con- structed over identical 20-day windows. The MODIS composite images were then analyzed (in conjunction with the AMSR-E composites during times of persistent cloud cover or otherwise low-quality portions in the MODIS composite image) to determine pixels with fast ice cover. These fast ice maps form the basis of the time series presented here. Fraser et al. (2010) conducted an error analysis of this method of fast ice detection and identified two error regimes: (i) error regime 1 represents a higher level of confi- FIG. 1. Map showing regions of the Southern Ocean used in this dence (1s error of 61.5%), where $90% of the fast study (after Zwally et al. 1983). ‘‘B & A Seas Sector’’ (608–1308W) ice in the image was classified from a single MODIS is the Bellingshausen and Amundsen Seas sector. The dashed line composite; and shows the longitudinal extent of the fast ice study region (108W– (ii) error regime 2 (1s error of 64.38%) where the 1728E). Fast ice regimes, as defined in section 3a, are identified with roman numerals (i)–(x). equivalent AMSR-E or previous–next MODIS composite was used to classify .10% of the fast ice.

(i.e., 12.5–25.0 km) passive microwave sea ice concen- The Antarctic continent was masked using the Mosaic tration data. of Antarctica (MOA) coastline product (Scambos et al. 2007). European Centre for Medium-Range Weather 2. Datasets and methods Forecasts (ECMWF) Interim Reanalysis data (Berrisford et al. 2009) from 1989 to 2008 [Mean Sea Level Pressure We create a series of 159 consecutive 2-km, 20-day (MSLP), 10-m wind vectors, and 2-m surface tempera- resolution cloud-free MODIS composite images, from tures on a 1.58grid] were formed into 20-day and annual March 2000 to December 2008, of the East Antarctic climatologies to assist in interpretation of fast ice vari- fast ice zone (63.58–728S, 108W–1728E, see dashed box in ability. Moreover, 20-day sea ice concentration compos- Fig. 1), using the methods outlined in Fraser et al. (2009) ite images were also created from the combined Scanning and Fraser et al. (2010). The method is summarized as Multichannel Microwave Radiometer (SMMR) and Spe- follows. cial Sensor Microwave Imager (SSM/I) dataset (Comiso (i) Assess the cloud content of each granule prior to 1999), to compare overall sea ice extent/area with fast ice downloading MODIS imagery, using a modified extent. This product uses the Bootstrap algorithm, and MOD35 cloud mask product (Ackerman et al. a threshold of 15% was used to compute sea ice area. 2006; Fraser et al. 2009), and discard the cloudiest These data are available from http://nsidc.org/data/ half to reduce processing requirements. Over the nsidc-0079.html. Bathymetric information was obtained 8.8-yr study period, around 150 000 MODIS gran- from the Smith and Sandwell (1997) dataset (version ules were used to produce the 159 composite 11.1, updated in 2008; available at http://topex.ucsd.edu/ images, totaling over 12 terabytes of data (reduced WWW_html/mar_topo.html). from a possible ;300 000 granules and ;25 tera- The evaluation of trend significance for the fast ice bytes). extent time series closely follows that used by Cavalieri (ii) Acquire 20 days of MODIS granules of the East and Parkinson (2008), with the metric of a continuous Antarctic coast. R value (the ratio of the linear trend to the standard (iii) Perform cloud masking using a modified MODIS deviation) used to indicate the significance of a particu- cloud mask. lar trend. This R value was converted to a confidence (iv) Reproject masked granules to a common grid. interval by assuming a two-tailed Student’s t distribution (v) Finally, average the reprojected images to form and using a lookup table with the appropriate number cloud-free composite images of the surface. Visible of degrees of freedom (DOF). Following Cavalieri and

Unauthenticated | Downloaded 10/02/21 01:08 PM UTC 1140 JOURNAL OF CLIMATE VOLUME 25

Parkinson (2008), the number of DOFs is set to the exceptions are at 118 and 158E where the shallow number of years of data—two (i.e., the number of DOFs continental shelf region is wider (i.e., ;50 km), is reduced by two when fitting a linear trend, so here, allowing a number of small icebergs to ground and with approximately 9 years of data, we use DOF 5 7). It provide anchors for fast ice. A thin (up to 30 km) is important to note that the use of null hypothesis sig- strip of seasonally recurring fast ice is also found at nificance testing has been criticized because of the ar- 278–358E, again where icebergs ground in shallow bitrary nature of significance levels, the effect of sample regions. size on significance levels, and the difficulty of correctly (ii) 358–408E (Lu¨tzow–Holm Bay—A region of ex- interpreting the result when rejecting or accepting the tensive multiyear fast ice surrounds Syowa base. null hypothesis (e.g., Nicholls 2001). These criticisms are Despite relatively deep bathymetry in the center overcome to some extent by using and reporting the of the bay (up to ;900 m), groups of icebergs R value (thus providing a continuum of significance). grounded to the west of Riiser–Larsenhalvøya Here, we perform trend analyses on the previously de- (;698S, 348E) and to the north of Syowa station fined (Zwally et al. 1983) Indian Ocean (IO; 208–908E) (at ;698S, 39.58E) anchor fast ice formations. and western Pacific Ocean (WPO; 908–1608E) sectors. These icebergs naturally reinforce the sheltering effect of Lu¨ tzow–Holm Bay, maintaining the frequently extensive fast ice. Additionally, relatively 3. Results and discussion calm atmospheric conditions are often encountered This section first presents a discussion of regional dif- in the bay during December and January; that is, ferences in fast ice distribution and morphology around during the fast ice breakout season, according to the East Antarctic coast, and then presents the 8.8-year ECMWF Interim Reanalysis data (see Fig. 3). time series of fast ice extent. This includes a regional This may reduce the occurrence of wind-driven fast analysis, and comparison with pack-ice area/extent ice breakout. Fast ice formed under such quiescent within those regions. conditions is mainly thermodynamically formed and can attain considerable thickness (Ushio a. Location and persistence of fast ice features 2006). (iii) 408–508E (Enderby Land Coast)—A seasonally Figure 2 shows a map of the average percentage of recurring, approximately 50-km-wide strip of fast time with fast ice cover throughout the 8.8-yr study pe- ice is found along this coastline, of which little is riod (e.g., a value of 50% likely indicates that fast ice is multiyear fast ice. Several thousand small grounded present for half of each year on average) and the ba- icebergs, clustered along several NW–SE-aligned thymetry for the study region (Smith and Sandwell bathymetric ridges, act to anchor the fast ice be- 1997). Ten distinct regions of fast ice cover were iden- tween the coast and the 400–500-m isobath. tified from visual inspection of Fig. 2, with the assistance (iv) 508–578E (Amundsen Bay to Cape Boothby)— of a SAR image mosaic of Antarctica, the RADARSAT-1 Fast ice rarely forms extensively in this region, Antarctic Mapping Project (RAMP) (Jezek 2002). These despite the fact that the shallow bathymetry per- regions were identified to both document and distinguish mits a line of grounded icebergs about 40 km from regional differences in fast ice formation regimes around the coast. Few grounded icebergs exist between the East Antarctic coast. The SAR mosaic was compared this line and the coast. It is possible that this dis- with the bathymetric map to determine the approximate tance is too wide for fast ice to span in the absence location of zones of grounded icebergs, that is, in waters of suitable onshore atmospheric/oceanic forcing. shallower that 400 to 500 m. In SAR imagery, icebergs Also, the continental shelf is very narrow here, appear as consistently bright (i.e., high backscatter) tar- possibly allowing warmer waters from the diver- gets under freezing conditions, compared with the typi- gence of the Weddell Gyre and the Antarctic Cir- cally lower backscatter values from sea ice (Williams et al. cumpolar Current to penetrate more easily onto 1999; Gladstone and Bigg 2002). From west to east, these the continental shelf (Meijers et al. 2010), which regions are classified as follows. could explain the reduced fast ice extent here. (i) 108W–358E (Haakon VII Sea Coast)—This region (v) 578–688E (Mawson Coast)—Small grounded ice- is characterized by low fast ice extents, with little bergs closely follow the contour of undersea ridges, to no multiyear fast ice. This is a possible conse- leading to recurring and distinctively shaped fast quence of the continental shelf break being very ice features extending ;50 km from the coast in close (;20 km) to the coast, leading to few grounded this area. Fast ice is present for much of the year, icebergs to act as anchor points for fast ice. Localized particularly in the 578–628E section. Multiyear fast

Unauthenticated | Downloaded 10/02/21 01:08 PM UTC 15 FEBRUARY 2012 F R A S E R E T A L . 1141

FIG. 2. (a),(c) Fast ice coverage map averaged over the 8.8-yr study period (from March 2000 to December 2008). A value of 100% is given to fast ice that covers the pixel for the entire time series of 159 images spanning the 8.8-yr study period. (b),(d) Corresponding bathymetry maps for (a),(c). Bathymetry is from Smith and Sandwell (1997). Coastline is from Scambos et al. (2007). See the text for a full discussion of fast ice formation regimes. The maximum width of the fast ice zone varies widely across the East Antarctic coast, reaching a maximum width of ;225 km at around 1508E (immediately east of the iceberg B-9B).

ice is found only in the most sheltered bays (e.g., precludes iceberg grounding, leading to a lack of Edward VIII Gulf, ;578E). fast ice anchor points. (vi) 688–718E(CapeDarnley)—A line of small (viii) 748–818E (Ingrid Christensen Coast)—A narrow grounded icebergs extends northeast from Cape strip of fast ice often covers this coast, forming Darnley, leading to extensive and frequent (though along the eastern margin of the Amery Ice Shelf, not typically multiyear) fast ice coverage. This fast the Polar Record Glacier, and hundreds of small ice feature frequently extends to the western edge grounded icebergs. Multiyear fast ice is found at of the Amery Ice Shelf. the eastern edge of the Amery Ice Shelf. Further (vii) 718–748E (north face of Amery Ice Shelf terminus)— offshore (;100 km), a large, seasonally-recurring No signiﬁcant fast ice forms in this dynamic region of offshore fast ice exists, of which the polynya region, likely due to the deep bathyme- main body is present for around half of each try (600–700 m) in Prydz Bay. The depth here season. This is anchored around a group of small

Unauthenticated | Downloaded 10/02/21 01:08 PM UTC 1142 JOURNAL OF CLIMATE VOLUME 25

FIG. 3. Prevailing wind diagrams generated from ECMWF ERA-Interim reanalysis data for the grid point at 698S, 398E, that is, near Syowa Station, in Lu¨ tzow–Holm Bay. The y axis represents binned wind strength, the x axis represents binned wind direction, and frequency in each bin is indicated by shading–contours. (a) Prevailing wind plot averaged over all months in the entire ERA-Interim period (1989–2010). The most frequently observed wind is an easterly, with a strength of 5–10 m s21. (b) Climatology (1989–2010) of prevailing wind for between DOY 341 and 20 (i.e., early December to late January in the following year). This period corresponds to the time of maximum rate of fast ice breakout across the Indian Ocean sector (see Fig. 6b). This plot shows that the modal wind speed is generally lower in Lu¨ tzow–Holm Bay during this period (the most frequently observed wind speed is 2–5 m s21, also from the east) than the annual modal wind speed.

grounded icebergs (centered on approximately (;50 km) strip of fast ice which is oriented parallel 678S, 78.58E). to the coast, similar to the fast ice morphology (ix) 818–1528E (West Ice Shelf to Cook Ice Shelf)— found in the aforementioned Enderby Land Coast This zonally extensive region is characterized by and Mawson Coast regions (regions iii and v). The several coastal features that are north–south aligned, strip of fast ice is more extensive in the western for example, iceberg tongues, glacier tongues, ice half of this region. Multiyear fast ice is found shelves, coastal promontories, and groups of smaller between Lauritzen and Slava Bays (1548–1568E) grounded icebergs. Irregular-shaped fast ice re- and also off the coast of the closed Russian gions are located on the eastern (windward) side Leningradskaya Station (;159.58E). Very few of these coastal features (as opposed to the major- grounded icebergs are present in this region. The ity of fast ice features found from 108Wto818E, recurring fast ice feature to the east of this region which typically run parallel to the coast). Latent likely forms in the oceanic lee of Cape Adare. heat polynyas (driven by katabatic winds) are typi- Wind speeds in the region, as shown in ECMWF cally found on the western (lee) side of the features ERA-Interim reanalyses (Berrisford et al. 2009, (Barber and Massom 2007). Several extensive (a not shown here), are typically relatively low, which few 1000 km2 in area) regions of multiyear fast ice precludes the formation of a signiﬁcant latent heat are encountered in this region. Every major north– polynya from the Cape. There is no evidence to south protrusion (i.e., the West Ice Shelf, Shackle- support the presence of the ocean ridge, reportedly ton Ice Shelf, line of grounded icebergs north of centered at 68.58S, 1578E (Smith and Sandwell Vincennes Bay, Dalton Iceberg Tongue, Dibble 1997, see Fig. 2d), as there appear to be no Iceberg Tongue, and Mertz Glacier Tongue) has grounded icebergs at this location. The ridge is a latent heat polynya on its western side (Massom also not present in other more recent bathymetry et al. 1998; Tamura et al. 2008) and a fast ice products, e.g., Timmermann et al. (2010). feature on its eastern side. (x) 1528–1728E (Cook Ice Shelf to Cape Adare)—This It is suggested that these regimes can be combined ﬁnal region is characterized by a relatively narrow into two larger regions with broadly similar fast ice

Unauthenticated | Downloaded 10/02/21 01:08 PM UTC 15 FEBRUARY 2012 F R A S E R E T A L . 1143

FIG. 4. Conceptual schematic diagrams of two East Antarctic fast ice formation regimes, with a similar color scale to Fig. 2. The continent and coastal protrusions are shown in white, and the ocean–pack ice–polynyas are shown in black. (top) Fast ice formation regime in the vicinity of a protrusion into the Antarctic Coastal Current. Multiyear fast ice (shown here in red) is frequently observed attached to the upstream side of the coastal protrusion. We hypothesize that fast ice growth is achieved via dynamic interception of pack ice (shown here in yellow and green) for coastal conﬁgurations such as this, leading to the characteristic sawtooth shape of fast ice features. (bottom) Fast ice formation regime in the vicinity of several small grounded icebergs some distance offshore. Note the lack of multiyear fast ice. A strip of ﬁrst-year fast ice, often ;50 km wide, forms when fast ice extends to the farthest grounded icebergs.

characteristics: 1) between 818 and 1558E, a region char- (Beaman and Harris 2005). Initial fast ice growth can acterized by a large number of coastal protrusions; and occur abruptly between the coast and the closest groun- 2) west of 818E, plus the small region from 1558;1728E, ded icebergs, or alternatively, fast ice can form between where relatively few coastal protrusions exist. These two closely spaced grounded icebergs some distance offshore. formation regimes are shown schematically in Fig. 4. Subsequent growth then occurs between existing fast ice Regime 1, representative of region (vi) and the exten- features and other grounded icebergs. Little multiyear sive region (ix), occurs when a N–S-aligned protrusion fast ice is observed under this formation regime. The into the (westward) Antarctic Coastal Current (ACoC) proportion of thermodynamically formed fast ice may exists. This protrusion can be in the form of a coastal be higher under this regime than for regime 1 because promontory, one or more large tabular grounded ice- this regime does not rely on pack ice advection for for- bergs, or a near-contiguous group of smaller grounded mation to occur. The distance from the shore to the fast icebergs. This regime is characterized by extensive fast ice edge during maximum extent is typically shorter ice formation on the upstream side of the protrusion under this formation regime than regime 2. Under this (multiyear fast ice in many cases) and a coastal polynya regime, the shape of the fast ice feature at maximum in the oceanic lee of the protrusion. The ACoC generally extent broadly resembles the 400–500-m bathymetry advects pack ice into the protrusion, resulting in dynami- contour. cally formed fast ice frequently extending over 100 km b. Fast ice time series trend analysis and from the shore (e.g., region ix in particular). The loca- overall trend analysis tion of the fast ice edge thus progresses toward the east during the growth season as more advected pack ice is The time series of fast ice extent for the entire region held fast against pre-existing fast ice. (108W–1728E) is shown in Fig. 5, with the mean annual Regime 2, representative of regions (iii), (v), (viii), cycle given in Fig. 5b. The 8.8-yr mean annual cycle be- and (x), typically occurs where small icebergs ground gins each year by steadily declining to the fast ice min- some distance from the coast, at a depth of 400–500 m imum of ;120 000 km2 at around Day of Year (DOY)

Unauthenticated | Downloaded 10/02/21 01:08 PM UTC 1144 JOURNAL OF CLIMATE VOLUME 25

FIG. 5. (a) Fast ice time series (thick line) for the East Antarctic coast (108W–1728E), showing a statistically sig- niﬁcant (99% conﬁdence level) increase of 4012 6830 km2 yr21. Note the major impact of extraordinarily extensive fast ice along the Enderby Land and Mawson coasts from 2006 to 2008. (b) The shape of the annual fast ice cycle, produced from the 8.8-yr dataset. This cycle is also shown as a thin line in (a). (c) Solid line: Fast ice extent anomaly (differences between the observed fast ice extent and the 8.8-yr mean for that time period). Dashed line: Linear trend for the 8.8-yr period.

61–80 (early mid-March). This is followed by a period the shape of the overall sea ice extent and area cycles is of rapid fast ice growth to a relatively broad maximum more sinusoidal, with fairly slow seasonal formation (from of ;388 000 km2 (persisting from ; DOY 141 to 300, or mid-March to late September) being followed by relatively mid-May to late October). Following DOY 300, the fast rapid retreat, typically from mid-October to early ice extent declines until the end of the year. In contrast, February (Gloersen et al. 1992).

FIG. 6. As in Fig. 5, but for the Indian Ocean sector (thick line), showing a statistically signiﬁcant increase of 4444 6457 km2 yr21. (b) Note the smooth annual cycle, possibly reﬂecting the relatively high portion of thermodynamically formed fast ice in this region.

Unauthenticated | Downloaded 10/02/21 01:08 PM UTC 15 FEBRUARY 2012 F R A S E R E T A L . 1145

FIG. 7. (a) As in Fig 6, but for the western Paciﬁc Ocean sector, showing a slight but nonsigniﬁcant decrease (579 6525 km2 yr21).

The annual cycle is more sinusoidal in the IO sector This result showing the more transient nature of dy- (Fig. 6b) than the WPO sector (Fig. 7b), though it is still namically formed fast ice in Antarctica, characterized substantially broader around the maximum than the by series of breakouts and reformations, has also been minimum. This possibly reflects the different formation observed in the Arctic by Mahoney et al. (2007b). regimes in each region (see Fig. 4). The proportion of There is a pronounced peak in the mean annual fast thermodynamically rather than dynamically formed fast ice extent cycle in the WPO sector at DOY 261–280. ice in the IO sector is likely higher than that in the WPO Initially, it was suspected that an anomalously high fast sector because of the greater number of north–south ice maximum extent in 2006 was solely contributing to coastal promontories in the latter sector. These features this peak. This anomalous maximum fast ice extent was act to intercept pack ice that is drifting westward around caused by advection of pack ice against the coast, driven the coast within the ACoC, leading to more dynamically by strong easterly winds (see section 3c and Fig. 8 for formed fast ice (Giles et al. 2008; Jezek 2002). Maps of a detailed description of this event). The contribution of fast ice in Giles et al. (2008) show higher radar back- this event to the peak in the annual cycle was evaluated scatter (indicating rougher ice on the scale of the radar by calculating a truncated mean (whereby the maximum wavelength, or ;5 cm in this case) for fast ice on the and minimum values were removed from the mean cal- upstream side of these coastal promontories. Similarly, culation). The truncated mean annual cycle also exhibited the SAR mosaic of Antarctica (Jezek 2002) shows low a peak at DOY 261–280, indicating that the timing of the backscatter for fast ice along the Mawson and Enderby fast ice maximum extent occurs within this period. land coasts, both of which are regions with no significant N–S-aligned promontories of protrusions. It is hypoth- esized here that the dynamic fast ice formation leads to TABLE 1. Table of fast ice extent trend results by region. The the jagged shape of the mean annual cycle in regions R value represents the ratio of trend slope to its standard deviation. containing a high proportion of dynamically formed fast An R value greater than 3.499 indicates a statistically significant ice. Significant fast ice breakout also typically begins to trend with greater than 99% confidence and is shown here in a bold font. occur earlier in the WPO sector than the IO sector (see Figs. 6b and 7b) and continues at a steady rate throughout Sector km2 yr21 %yr21 R value the summer breakout season. This suggests that dynam- East Antarctica 4012 6830 1.43 60.30 4.84 ically formed fast ice may be mechanically weaker than (108W–1728E) thermodynamically formed fast ice, leading to episodic Indian Ocean (208–908E) 4444 6457 4.07 60.42 9.73 breakout, which may occur at the interfaces between Western Pacific Ocean 2579 6525 20.40 60.37 21.10 existing fast ice and newer dynamically formed fast ice. (908–1608E)

Unauthenticated | Downloaded 10/02/21 01:08 PM UTC 1146 JOURNAL OF CLIMATE VOLUME 25

2 TABLE 2. Table of fast ice (FI) maximum and minimum extents (km ) for East Antarctic coast, Indian Ocean, and western Paciﬁc Ocean sectors.

East Antarctica Indian Ocean sector Western Paciﬁc Ocean (108W–1728E) (208–908E) (908–1608E) Minimum Maximum Minimum Maximum Minimum Maximum 2000 392 000 144 000 206 000 2001 133 000 375 000 34 000 140 000 87 000 199 000 2002 96 000 400 000 41 000 148 000 51 000 226 000 2003 110 000 362 000 24 000 146 000 79 000 211 000 2004 83 000 341 000 8 900 137 000 69 000 184 000 2005 115 000 374 000 46 000 150 000 68 000 192 000 2006 104 000 447 000 26 000 179 000 72 000 238 000 2007 127 000 395 000 66 000 162 000 54 000 195 000 2008 191 000 407 000 92 000 173 000 76 000 199 000 Mean 120 000 388 000 42 000 153 000 69 000 205 000

We also performed a regional analysis on the fast ice a longer-term positive trend. The observed increase is time series, comparing the IO and WPO sectors. Table 1 concentrated mainly in the IO sector (see Fig. 6), with summarizes the results of the trend analyses, and Table 2 a trend of 4444 6457 km2 yr21. Interannual variability shows the value of the fast ice minimum/maximum areal in areal extent in this region is relatively large, especially extent in each year. in fast ice minima. As with the entire East Antarctic For the entire East Antarctic coast and from 2000 to coast, annual minima in the IO sector seem to have little 2008, a positive trend (increase) in fast ice extent of relation to previous or subsequent maxima. Fast ice max- 4012 6830 km2 yr21 is observed. Though this trend is ima in this region appear to follow a bimodal distribution, statistically signiﬁcant (at the 99% conﬁdence level), the with the 2000–05 maxima falling near the 8.8-yr mean time series is too short to determine whether it is part of cycle value, while the 2006–08 maxima are considerably

FIG. 8. (a) Prevailing wind diagram generated from 1989 to 2010 ECMWF ERA-Interim reanalysis data for the grid point at 67.58S, 1508E, that is, east of the 2000–08 location of the large tabular iceberg B-9B (location shown in Fig. 2). Axes are as in Fig. 3. The most frequently observed wind is from the southeast, with a strength of 5–10 m s21. (b) Wind difference plot showing the difference between the prevailing wind (a) and the wind during the interval DOY 261 to 280, in 2006. Red shading indicates lower frequency than the prevailing wind, and blue indicates stronger. During this interval, strong winds (.25 m s21) are observed from the east and southeast. In this case these winds may have advected pack ice against the coast, B-9B, and/ or preexisting fast ice, temporarily forming fast ice via dynamic advection.

Unauthenticated | Downloaded 10/02/21 01:08 PM UTC 15 FEBRUARY 2012 F R A S E R E T A L . 1147 higher than that of the previous 6 years. In contrast, there to ground, leading to dynamic fast ice growth via in- is no significant trend in fast ice extent for the WPO terception of pack ice (e.g., to the east of iceberg B-9B, sector (2579 6525 km2 yr21). at ;1508E). An apparent change in fast ice extent trend is ob- Annually averaged (annual minimum to subsequent served in the IO sector from ;2004 onward (Fig. 6). annual minimum) fast ice conditions are shown in Fig. 11 Prior to 2004, a slightly negative trend is observed, and (except for the image labeled ‘‘2008,’’ where no data interannual variability is relatively small. From 2004 from 2009 were analyzed and hence the 2009 minimum onward, the trend becomes strongly positive and vari- was unavailable, making this image more biased toward ability increases. This change in trend in the IO sector greater fast ice coverage). This figure clearly shows the contributes strongly to the trend observed for the entire origin of the positive trend in the IO sector (208–908E, East Antarctic coast. In the IO sector, minima range from see Fig. 6), especially over the period 2006–08. In par- ;9000 km2 (in 2004) to ;92 000 km2 (2008), while max- ticular, much more of the fast ice along the Enderby ima range from ;137 000 km2 (2004) to ;179 000 km2 Land and Mawson coasts (408;508E and 578;688E, (2006). No such change is observed in the WPO sector respectively) survives the summertime melt during 2007 (Fig. 7), where there is relatively little interannual and 2008, contributing to the progressively higher min- variability (but values still span quite a large range), imum fast ice extents during these years. Additionally, and the negative trend continues throughout the 8.8-yr more fast ice forms north of the region from Amundsen record. In this sector, minima range from ;51 000 km2 Bay to Cape Boothby (508–578E) during the maxima of (2002) to ;87 000 km2 (2001), while maxima range from 2006, 2007, and 2008, contributing to the higher max- ;184 000 km2 (2004) to ;238 000 km2 (2006). imum extent observed in the IO sector during these The relationship between each year’s fast ice maximum years. and the subsequent minimum was also analyzed. Anom- In the WPO sector, the main contribution to the origin alously low minima are observed in 2002 (;96 000 km2), of the anomalously high maximum extent in 2006 (DOY 2004 (;83,000 km2), and 2006 (;104 000 km2,compared 261–280) can be traced to an extensive fast ice feature to with the mean of ;120 000 km2). Of these years, only the the east of iceberg B-9B [until recently (Young et al. 2004 maximum was anomalously low (;341 000 km2 com- 2010), centered at approximately 678209S, 1488239E]. pared with the mean maximum extent of ;388 000 km2). Here, the westward-flowing ACoC advects pack ice into In fact, the 2006 maximum was anomalously high the region between B-9B and the coast, often forming (;447 000 km2). The only anomalously high minimum heavily consolidated pack ice which can temporarily form (;191 000 km2) was encountered in 2008, which was fol- fast ice (Massom et al. 2001; Barber and Massom 2007). lowed by an anomalously high maximum (;407 000 km2). Dense clusters of small grounded icebergs, that is, those It appears that there is little correlation between max- to the north of B-9B, act in a similar fashion to individual ima and subsequent minima, except from 2006 to 2008 in large grounded icebergs (Massom et al. 2001; Barber the IO sector, which recorded strongly positive extent and Massom 2007). In this way, fast ice can extend across anomalies for almost the entire 3-yr period. waters deeper than the maximum depth of iceberg grounding (;450 m) and be present more than 200 km c. Fast ice extent climatology and annual offshore. The 2006 maximum is possibly an extreme mani- minimum-to-minimum averages festation of this phenomenon, and an in-depth investiga- The fast ice extent by time of year, averaged over the tion into this event is planned. Examination of ECMWF 8.8-yr time series, is shown in Figs. 9 and 10. These fig- Interim reanalysis data for this period shows anomalously ures represent a fast ice climatology. Comparison be- strong easterly winds (see Fig. 8). This provides evidence tween these figures and Fig. 2 reaffirms the close links for pack ice advection being an important contributor to between bathymetry, iceberg grounding, and fast ice ex- fast ice growth in this region. Ocean currents are also tent discussed in previous sections. In many regions, and likely to be an important factor in this region, but data are particularly where minimal coastal protrusions are pres- lacking. ent, fast ice reaches its maximum extent early in the Analysis of the fast ice climatology (Figs. 9 and 10) season and is unable to grow past this because of a lack reiterates the important differences in the nature of fast of grounded icebergs in waters deeper than approxi- ice formation and breakout in each sector, as described mately 400–500 m. It is suggested that these regions in section 3a. Fast ice growth and breakout along the (e.g., the Mawson and Enderby Land coasts) may have Enderby Land and Mawson coasts, and also the region a higher fraction of thermodynamically formed fast ice. to the east of the Cook Ice Shelf (;1528E), both occur This is in contrast to regions where the presence of first between the coast and the closest grounded ice- shallow bathymetry allows large numbers of icebergs bergs, before progressing to nearby grounded icebergs,

Unauthenticated | Downloaded 10/02/21 01:08 PM UTC 1148 JOURNAL OF CLIMATE VOLUME 25

FIG. 9. Fast ice climatology map for the 8.8-yr period, in 20-day increments, from DOY 1–180 (see Fig. 10 for DOY 181–365). Each panel shows the fraction of observations during that DOY interval with fast ice cover. The color scale is the same as that used in Fig. 2; that is, it represents the proportion of time over which fast ice coverage occurs. eventually forming a strip (typically ,50 km wide) of scale of this dataset, detailed analysis of fast ice sea- fast ice across the coast. This is in contrast to much of the sonality is outside of the scope of this work. fast ice in the WPO sector. As previously mentioned, in d. Comparison between fast ice extent and overall this region, several coastal protrusions allow fast ice to regional sea ice extent and area form windward (upstream) of these features. As the season progresses, the fast ice grows more in an eastward The relationship between fast ice extent and overall direction as more pack ice is intercepted by the preex- sea ice extent was examined using SSM/I passive mi- isting fast ice. Fast ice retreat in this sector then occurs crowave sea ice concentration 20-day composite images by recession of the fast ice largely from east to west in generated over the same time period. Comparisons for spring–summer. the whole coast, the IO sector, and the WPO sector are The length of the fast ice coverage per year at a given shown in Figs. 12, 13, and 14, respectively. Note the location, that is, the seasonality or fast ice season dura- difference in overall sea ice extent and area between the tion, is an important parameter, responding to both IO and WPO sectors, a consequence of the greater pack oceanic and atmospheric forcing (Heil 2006; Heil et al. ice extent in the IO sector, relating to the location of 2006; Mahoney et al. 2005). Because of the large spatial the eastern part of the Weddell Gyre and the southern

Unauthenticated | Downloaded 10/02/21 01:08 PM UTC 15 FEBRUARY 2012 F R A S E R E T A L . 1149

FIG. 10. As in Fig. 9, but from DOY 181–365 (see Fig. 9 for DOY 1–180). boundary of the Antarctic Circumpolar Current (Gloersen ice minimum extents during years when more extensive et al. 1992). Across the East Antarctic coast and both pack ice is present. subregions, pack ice area/extent maxima are uncorrelated This fast ice time series mirrors the longer-term trends with fast ice maxima, with correlation coefficients (R)of in overall sea ice extent/area in the region (dating back to 0.22, 0.14, and 0.2 for the entire coast, IO, and WPO 1978). For example, both Comiso (2009) and Cavalieri sectors, respectively. However, the minima are strongly and Parkinson (2008) show a larger increase in sea ice correlated (correlation coefficients of 0.93, 0.94, and extent in the IO sector (;1.9 61.4% decade21) than the 0.81, respectively). This is not necessarily an indication WPO sector (;1.4 61.9% decade21), though neither that fast ice extent and overall sea ice extent share a trend is significant at the 95% confidence level (Cavalieri common forcing; rather the relative fraction of fast ice and Parkinson 2008). comprising overall sea ice increases (i.e., the ratio of e. Variability in timing of fast ice maxima seaicetofasticedecreasestoaminimum)duringthe and minima summertime sea ice minimum (seeTable2,Fig.15). Additionally, fast ice is highly vulnerable to ocean waves Fast ice extent is thought to respond to atmospheric (Crocker and Wadhams 1988, 1989; Langhorne et al. and oceanic forcing in a complex manner (e.g., Heil 2001), and it may be that pack ice acts as a protective 2006; Mahoney et al. 2007a; Massom et al. 2009). It is, buffer to their destructive effect, leading to larger fast however, outside the scope of this paper to carry out a

Unauthenticated | Downloaded 10/02/21 01:08 PM UTC 1150 JOURNAL OF CLIMATE VOLUME 25

FIG. 11. Fast ice conditions by season (from one minimum to the subsequent minimum), using the same color scale as Fig. 2; that is, it represents the proportion of time over which fast ice coverage occurs. Since no 2009 data were analyzed, the 2008 season is incomplete, and biased toward higher fast ice coverage. detailed analysis of the effects of atmospheric and oce- and maximum overall sea ice extents over the same time anic variability and change on the observed variability in period (2000–08) of ;770 000 and ;9 300 000 km2, timing of fast ice maxima/minima and spatiotemporal respectively (Comiso 1999), that is, a ratio of ;12 to 1 for behavior. Such work is planned for the future, using the overall sea ice to fast ice. The fast ice minimum almost time series presented here. always occurs later than the overall sea ice minimum, Timing of maximum and minimum fast ice extent are a likely consequence of the pack ice protecting the fast ice important fast ice parameters that are sensitive to changes from swell-induced breakout. in climate (Heil et al. 2006; Mahoney et al. 2005). The The timing of fast ice minimum extent is observed to minimum fast ice extent, averaged across the entire region, occur progressively earlier in the IO sector (on the order typically occurs during DOY 61–80 (early–mid-March), of 5 days yr21) throughout the 8.8-yr time series (see with a mean value of ;120 000 km2. Maximum typically Table 3). No obvious trend is observed in the timing of occurs during DOY 261–280 (mid–late September), with the overall sea ice extent in the IO sector. In general, a mean value of ;388 000 km2 (see Table 3)—that is, the the timing of maximum/minimum fast ice extent dis- ratio of maximum to minimum fast ice extent is typically plays higher variability than the corresponding timing ;3.2 to 1. This compares with mean regional minimum of overall sea ice maximum/minimum extent, reﬂecting

Unauthenticated | Downloaded 10/02/21 01:08 PM UTC 15 FEBRUARY 2012 F R A S E R E T A L . 1151

FIG. 12. Time series comparison between overall sea ice extent– FIG. 13. As in Fig. 12, but for the Indian Ocean sector (208–908E). area and fast ice extent along the entire East Antarctic coast (108W–1728E). The scale for overall sea ice extent and area is on During winter, the variability of fast ice extent is much the left. Fast ice extent has been scaled by a factor of 10 for inter- comparison purposes, and uses the scale on the right. lower than the summer, a consequence of the limit placed upon maximum fast ice extent due to the location of grounded icebergs (see section 3e). both the lower fast ice extent compared to overall sea ice, and the complex controls that limit fast ice extent 4. Summary and further work (e.g., grounded icebergs acting as anchor points). The new time series presented in this paper entails the f. Percentage of fast ice comprising overall sea ice most comprehensive and detailed information on vari- extent/area ability in the distribution of Antarctic fast ice to date and The percentage of fast ice area comprising overall sea represents an important new climatic baseline against ice area throughout the 8.8-yr time series for the East which to gauge future change and variability along the Antarctic coast is shown in Fig. 15. The percentage of entire coastal and near-coastal East Antarctic envi- fast ice extent comprising overall sea ice extent is also ronment in a warming climate. This is by virtue of the shown in this figure. Minimum, maximum, and mean sensitivity of fast ice to change/variability in atmo- values for each year are shown in Table 4. We assume spheric and oceanic circulation and temperature, ocean here that fast ice concentration is 100%; hence, fast ice extent is equivalent to fast ice area. This is a reasonable assumption based on field observations. A strong seasonal cycle is observed in the fraction of overall sea ice area/extent, which is fast ice. Following the maximum fast ice percentage early in the season (around DOY 41–60, late February), the fast ice extent (area) percentage slowly decreases to a broad minimum, with a mean value of 3.77% (4.51%). A rapid increase in the fast ice percentage is observed from around DOY 341–365 (mid-December), with maximum percentage oc- curring during DOY 41–60 in most years, except 2003 and 2005, when the maximum fast ice extent percentage occurred during DOY 21–40. The value of the maximum (minimum) extent percentage varies between 16.0% and 21.0% (3.2% to 4.2%), with the maximum (minimum) area percentage showing similar variability. FIG. 14. As in Fig. 12, but for the western Pacific Ocean sector Note that the largest variability in this quantity occurs (908–1608E). Note that fast ice extent in this sector comprises during the summer sea ice/fast ice minimum, where the a much greater portion of overall sea ice extent compared to the relative variability of both sea ice and fast ice is greater. Indian Ocean sector (Fig. 13).

Unauthenticated | Downloaded 10/02/21 01:08 PM UTC 1152 JOURNAL OF CLIMATE VOLUME 25

FIG. 15. (a) Percentage of overall sea ice that is fast ice, throughout the 8.8-yr time series. Minimum, maximum, and mean values for each year are given in Table 2. Maximum fast ice percentage typically occurs at approximately the same time as the sea ice minimum. Timing of minimum fast ice percentage has larger variability, reflecting the large variability in timing of the fast ice maximum. (b) Climatology of the data presented in (a), using the same vertical scale. waves/storms, and the larger-scale pack ice realm. The part of the record, which both determines and domi- new dataset also provides an important means of spa- nates the upward 9-yr trend, is attributed to the anom- tially extending detailed (though geographically limited) alous persistence of extensive fast ice off the Mawson point measurements of fast ice formation and break-up and Enderby Land coasts, both in summer and winter processes at sites close to Antarctic bases within the (and to a lesser extent during the shoulder seasons of AFIN program (Heil et al. 2011). spring and autumn). Atmospheric factors influencing Although the time series is currently short, that is, ;9 this variability are currently unknown but are under years, certain spatiotemporal patterns are apparent that investigation. highlight broad regional differences in factors relating to By comparison, the concomitant lower variability and fast ice formation, persistence, and seasonality. In terms lack of a similar abrupt change around 2004/05 in the WPO of overall extent across the entire East Antarctic region sector is attributed to differences in the physical setting (from 108W to 1728E), there has been a statistically sig- (given that different atmospheric forcing patterns may nificant increase (at the 99% confidence level) of 1.43 also play a key role). In this broad sector, extensive north– 60.3% yr21. This pattern is in fact dominated by the south-oriented coastal promontories and grounded ice- signal from the IO sector, where an increase in fast ice berg assemblages are more of a factor than they are in the extent of 4.07 60.42% yr21 is observed. In contrast, the IO sector. These lead to generally greater persistence in 9-yr trend in the adjacent WPO sector is of a (non- the WPO sector of more dynamically formed fast ice significant) decrease of 20.40 60.37% yr21. Nested masses that also extend further offshore. More work is within these linear trends are major contrasts in the de- necessary, however, to tie down the processes and regional gree of interannual variability, however. In the IO sector, mechanisms involved (in both sectors). With this in the period 2000–03 is characterized by low variability mind, research is underway to investigate links between and a slight decreasing trend, while an abrupt upward coastal configuration and spatiotemporal characteristics trend combined with much higher interannual variabil- of fast ice coverage. Additional information on fast ice ity occurs in the subsequent period (2004–08). This latter type (and process of formation) as a function of region

Unauthenticated | Downloaded 10/02/21 01:08 PM UTC 15 FEBRUARY 2012 F R A S E R E T A L . 1153

TABLE 3. Table of the timing of fast ice and overall sea ice minimum and maximum extent, shown in DOY range format for the period March 2000–December 2008. Bold entries indicate which event occurred earlier (overall sea ice minimum extent or fast ice minimum extent).

East Antarctica Indian Ocean Western Pacific (108W–1728E) (208–908E) (908–1608E) Minimum Maximum Minimum Maximum Minimum Maximum Fast ice Sea ice Fast ice Sea ice Fast ice Sea ice Fast ice Sea ice Fast ice Sea ice Fast ice Sea ice 2000 261–280 261–280 221–240 261–280 261–280 221–240 2001 061–080 041–060 241–260 261–280 081–100 041–060 181–200 261–280 061–080 041–060 241–260 261–280 2002 061–080 041–060 201–220 261–280 081–100 041–060 281–300 261–280 061–080 041–060 201–220 261–280 2003 061–080 041–060 261–280 261–280 081–100 061–080 221–240 281–300 061–080 041–060 161–180 261–280 2004 061–080 041–060 281–300 261–280 061–080 041–060 201–220 261–280 081–100 041–060 161–180 261–280 2005 061–080 041–060 241–260 261–280 061–080 041–060 241–260 261–280 061–080 041–060 241–260 241–260 2006 061–080 041–060 261–280 281–300 061–080 041–060 241–260 261–280 041–060 041–060 261–280 281–300 2007 061–080 041–060 221–240 261–280 061–080 041–060 221–240 261–280 061–080 041–060 221–240 281–300 2008 041–060 041–060 281–300 281–300 041–060 041–060 241–260 301–320 061–080 041–060 281–300 281–300 can be derived from backscatter characteristics related breakout frequency and extent in the Ade´lie Land coastal to fast ice roughness/deformation using satellite SAR sector of East Antarctica and the annual breeding success data (Giles et al. 2008). In addition, a SAR-based al- of the Emperor penguin population at Dumont d’Urville gorithm has been developed that will provide higher for the period 1992–99 (based on AVHRR data analysis). spatial and temporal resolution maps of fast ice coverage In particular, the new fast ice data will be compared with in regions-of-interest (Giles et al. 2011), to complement improved penguin demographic information to investi- the broader-scale coverage of the MODIS product and gate possible relationships between fast ice coverage better resolve breakout/formation events related to the and penguin mortality, for example. Moreover, spatio- passage of storms. Combination of the two datasets temporal variability in fast ice coverage has recently been should enable improved understanding of processes implicated in the breeding success and ecology of Ade´lie and mechanisms responsible for the strong inter- and penguins at Be´ chervaise Island in East Antarctica intraregional differences noted in this study, particularly (Emmerson and Southwell 2008), and the new dataset when merged with meteorological analysis. will allow a more detailed analysis and assessment of the The broader physical and biological implications of linkages. In addition, a linkage has recently been pro- the observed change/variability in East Antarctic fast posed between fast ice and the stability of floating ice ice coverage from 2000 to 2008 are currently unknown sheet margins, that is, the Mertz Glacier Tongue prior but may be significant and are under investigation. Plans to its calving in 2010 (Massom et al. 2010). This study are in place to apply the new dataset to a temporal ex- suggests that changes in fast ice coverage may have im- tension of the regional study of Massom et al. (2009), portant ramifications for ice loss from the vulnerable ice which showed an apparent linkage between fast ice ex- sheet margins that are under increasing threat from tent and wind-driven variability in intraseasonal fast ice oceanic warming; the new dataset will enable wider

TABLE 4. Table of maximum and minimum percentage of overall sea ice extent/area which is fast ice (see Fig. 15).

East Antarctica (108W–1728E) Maximum fast ice percentage compared to: Minimum fast ice percentage compared to: Overall sea ice extent (%) Overall sea ice area (%) Overall sea ice extent (%) Overall sea ice area (%) 2000 3.9 4.8 2001 21.0 34.0 4.0 4.8 2002 19.4 34.1 4.2 4.9 2003 17.9 25.9 3.7 4.4 2004 20.7 33.9 3.2 3.7 2005 19.6 35.1 3.3 4.1 2006 20.6 32.2 3.6 4.3 2007 16.4 26.6 4.1 4.7 2008 16.0 24.1 4.0 4.9 Mean 19.0 30.7 3.8 4.5

Unauthenticated | Downloaded 10/02/21 01:08 PM UTC 1154 JOURNAL OF CLIMATE VOLUME 25 assessment of this potentially important role of fast ice icebergs ground (Beaman and Harris 2005). In most re- in a changing climate. In all cases, a better under- gions, this forms an upper limit on maximum extent of standing is required of the role of change/variability in fast ice coverage offshore, given that grounded iceberg large-scale patterns (modes) of atmospheric circulation assemblages for anchor points for fast ice growth by both in driving the observed regional change/variability in fast thermodynamic and dynamic processes, a finding that is ice coverage across East Antarctica, and this is once in line with other studies, for example, Giles et al. (2008) again the focus of current research. and Massom et al. (2009). Further fast ice growth past Although based on a temporally limited dataset, the this physical boundary is achieved only in the most shel- current study confirms an intimate though complex as- tered regions (e.g., Lu¨tzow–Holm Bay) or regions where sociation between fast ice variability and that of the pack ice is continually advected into other protruding surrounding regional pack ice. This implies that pro- coastal features, for example, near 1508E (Massom et al. jected changes in Antarctic pack ice (e.g., Bracegirdle 2001). et al. 2008) may significantly impact fast ice distribution. Given the wide-ranging importance of improved es- On the broadest scale, the regional short-term trends in timation of fast ice coverage, we plan to use the same fast ice extent shown in this study broadly agree with technique to extend the time series both forward and longer-term (1979–2008) trend in overall sea ice (com- back in time, the latter using NOAA AVHRR data from prising both pack ice and fast ice) extent/area (e.g., pre-2000 (where and when available). Future fast ice Cavalieri and Parkinson 2008; Comiso 2009). However, extent derivations will exploit imagery from MODIS East Antarctic fast ice extent displays greater variability and the Ocean Land Color Instrument (OCLI) onboard than overall sea ice extent/area on monthly through sea- the European Space Agency Sentinel-3 satellite (due sonal to interannual time scales. This is likely due to the for launch in 2013, http://www.eoportal.org/directory/ often abrupt transitions that typically occur between fast presGMESSentinel3Mission.html). Further plans are ice breakout and growth. In general, fast ice maxima are to extend Antarctic coverage to circumpolar and to apply uncorrelated with sea ice extent (and area) maxima, the technique to mapping regional Arctic fast ice. Work is while the opposite is true for the respective minima. This underway to semi-automate the algorithm, an important relationship is thought to be indicative of the effect of step given the large volume of data involved and the need pack ice as a protective buffer against ocean swell/wave- for routine coverage, to create a gap-free climatic dataset induced fast ice breakout (Langhorne et al. 2001). This that is equivalent to the important satellite passive linkage is further suggested by the fact that the fast ice microwave–derived pack ice time series that dates back minimum typically occurs after the overall sea ice min- to 1979 (Comiso 2009). imum. However, fast ice minima (maxima) are generally uncorrelated with subsequent maxima (minima). Signif- Acknowledgments. This work was carried out with the icant interannual variability is also observed in the timing support of the Australian Government’s Cooperative of fast ice minima and maxima, although the latter is Research Centres Program through the Antarctic Cli- difficult to determine accurately given the broad nature mate and Ecosystems Cooperative Research Centre (ACE of the peak in the annual cycle. The minimum extent of CRC) and contributes to Australian Antarctic Science fast ice is observed to occur progressively earlier in the Project 3024 and CPC Project 18. MODIS data were IO sector, despite there being no corresponding trend in obtained from the NASA Level 1 Atmosphere Archive the timing of minimum overall sea ice extent. Further and Distribution System (LAADS) (http://ladsweb. analysis is necessary to determine possible causes. nascom.nasa.gov/). Smith and Sandwell bathymetry data The mean annual growth and decay cycle differs for were obtained from the Scripps Institution of Oceanog- fast ice compared to overall sea ice in that it is charac- raphy (http://topex.ucsd.edu/marine_topo/mar_topo.html). terized by a temporally broad maximum and an abrupt, AMSR-E ASI sea ice concentration data were obtained relatively short minimum. By comparison, the seasonal from the University of Hamburg (ftp://ftp-projects.zmaw. cycle of Antarctic pack ice is characterized by a rela- de/seaice/AMSR-E/). SSM/I sea ice concentration data tively long expansion phase (typically from February (dataset reference NSIDC-0079) were obtained from through August–October) followed by a rapid decay the NASA Earth Observing System Distributed Active from November through December (Gloersen et al. 1992). Archive Center (DAAC) at the U.S. National Snow and The shape of the annual fast ice cycle itself (in particular Ice Data Center (NSIDC), University of Colorado, the broad maximum, and the relatively short growing Boulder (http://www.nsidc.org). ECMWF interim re- season compared to that for overall sea ice) is thought to analysis data were obtained from the ECMWF Research reflect the link between fast ice presence and shallow Data Server (http://data.ecmwf.int/data/). Thanks to bathymetry: 400–500 m is the maximum depth at which anonymous reviewers for their helpful comments and

Unauthenticated | Downloaded 10/02/21 01:08 PM UTC 15 FEBRUARY 2012 F R A S E R E T A L . 1155 discussion. The authors wish to thank P. Reid and ——, ——, P. Heil, and G. Hyland, 2011: Semi-automated feature- T. Tamura for their helpful review comments, A. Worby tracking of East Antarctic sea ice from Envisat ASAR imag- for discussions regarding the data, and D. FitzGerald ery. Remote Sens. Environ., 115, 2267–2276. Gladstone, R., and G. R. Bigg, 2002: Satellite tracking of icebergs and R. Warner for discussions about statistics. in the Weddell Sea. Antarct. Sci., 14, 278–287. Gloersen, P., W. J. Campbell, D. J. Cavalieri, J. C. Comiso, C. L. Parkinson, and H. J. Zwally, 1992: Arctic and Antarctic sea REFERENCES ice, 1978–1987: Satellite passive microwave observations and analysis. NASA Special Publication SP-511, 290 pp. Ackerman, S. A., and Coauthors, 2006: Discriminating clear-sky Heil, P., 2006: Atmospheric conditions and fast ice at Davis, East from cloud with MODIS. NASA Goddard Space Flight Center Antarctica: A case study. J. Geophys. Res., 111, C05009, Algorithm Theoretical Basis Doc. ATBD-MOD-06, 125 pp. Barber, D. G., and R. A. Massom, 2007: The role of sea ice in Arctic doi:10.1029/2005JC002904. and Antarctic Polynyas. Polynyas: Windows to the World, Vol. ——, C. W. Fowler, and S. Lake, 2006: Antarctic sea-ice velocity as 74, W. Smith and D. G. Barber, Eds., Elsevier, 1–43. derived from SSM/I imagery. Ann. Glaciol., 44, 361–366. Beaman, R. J., and P. T. Harris, 2005: Bioregionalisation of the ——, S. Gerland, and M. A. Granskog, 2011: An Antarctic moni- George V Shelf, East Antarctica. Cont. Shelf Res., 25, 1657– toring initiative for fast ice and comparison with the Arctic. 1691. Cryosph. Discuss., 5, 2437–2463, doi:10.5194/tcd-5-2437-2011. Berrisford, P., D. Dee, K. Fielding, M. Fuentes, P. Kallberg, Jezek, K., cited 2002: RAMP AMM-1 SAR image mosaic of S. Kobayashi, and S. Uppala, 2009: The ERA-interim archive. Antarctica. [Available online at http://nsidc.org/data/nsidc- ECMWF ERA Rep. Series No. 1, 16 pp. 0103.html.] Bracegirdle, T. J., W. M. Connolley, and J. Turner, 2008: Antarctic Kawamura, T., K. I. Ohshima, T. Takizawa, and S. Ushio, 1997: climate change over the twenty first century. J. Geophys. Res., Physical, structural, and isotopic characteristics and growth 113, D03103, doi:10.1029/2007JD008933. processes of fast sea ice in Lu¨ tzow-Holm Bay, Antarctica. Brunt, K. M., O. Sergienko, and D. R. MacAyeal, 2006: Observa- J. Geophys. Res., 102 (C2), 3345–3355. tions of unusual fast-ice conditions in the southwest Ross Sea, Kooyman, G. L., and J. Burns, 1999: Weddell seal versus Emperor Antarctica: Preliminary analysis of iceberg and storminess penguin: Boss of the Ross Sea. Amer. Zool., 39, 9–19. effects. Ann. Glaciol., 44, 183–187. Kozlovsky, A. M., Y. L. Nazintsev, V. I. Fedotov, and N. V. Cavalieri, D. J., and C. L. Parkinson, 2008: Antarctic sea ice vari- Cherepanov, 1977: Fast ice of the Eastern Antarctic (in Russian). ability and trends, 1979–2006. J. Geophys. Res., 113, C07004, Proc. Sov. Antarc. Exped., 63, 1–129. doi:10.1029/2007JC004564. Langhorne, P. J., V. A. Squire, C. Fox, and T. G. Haskell, 2001: Comiso, J., cited 1999: Bootstrap Sea Ice Concentrations from Lifetime estimation for a land-fast ice sheet subjected to ocean NIMBUS-7 SMMR and DMSP SSM/I, 2000–2008. National swell. Ann. Glaciol., 33, 333–338. Snow and Ice Data Center. [Available online at http://nsidc. Lei, R., Z. Li, B. Cheng, Z. Zhang, and P. Heil, 2010: Annual cycle org/data/nsidc-0079.html.] of landfast sea ice in Prydz Bay, east Antarctica. J. Geophys. ——, 2009: Variability and trends of the global sea ice cover. Sea Res., 115, C02006, doi:10.1029/2008JC005223. Ice, 2nd ed. D. Thomas and G. Dieckmann, Eds., Wiley- Lemke, P., and Coauthors, 2007: Observations: Changes in snow, Blackwell, 205–246. ice and frozen ground. Climate Change 2007: The Physical ——, and F. Nishio, 2008: Trends in the sea ice cover using en- Science Basis, S. Solomon et al., Eds, Cambridge University hanced and compatible AMSR-E, SSM/I, and SMMR data. J. Press, 337–383. Geophys. Res., 113, C02S07, doi:10.1029/2007JC004257. Lieser, J. L., 2004: A numerical model for short-term sea ice Crocker, G. B., and P. Wadhams, 1988: Observations of wind- forecasting in the Arctic. Alfred Wegener Institute for Polar generated waves in Antarctic fast ice. J. Phys. Oceanogr., 18, and Marine Research Rep. on Polar and Marine Research 485, 1292–1299. 93 pp. ——, and ——, 1989: Breakup of Antarctic fast ice. Cold Reg. Sci. Mae, S., T. Yamanouchi, and Y. Fujii, 1987: Remote sensing of fast Technol., 17, 61–76. ice in Lu¨ tzowholmbukta, East Antarctica, using satellite Emmerson, L., and C. Southwell, 2008: Sea-ice cover and its in- NOAA-7, 8 and aircraft. Ann. Glaciol., 9, 251–251. fluence on Ade´lie penguin reproductive performance. Ecol- Mahoney, A., H. Eicken, L. Shapiro, and A. Graves, 2005: Defining ogy, 89, 2096–2102. and locating the seaward landfast ice edge in northern Alaska. Enomoto, H., F. Nishio, H. Warashina, and S. Ushio, 2002: Satellite Proceedings of the International Conference on Port and Ocean observation of melting and break-up of fast ice in Lu¨ tzow- Engineering under Arctic Conditions, Vol. 3, POAC, 991–1001. Holm Bay, East Antarctica. Polar Meteor. Glaciol., 16, 1–14. ——, ——, A. G. Gaylord, and L. Shapiro, 2007a: Alaska landfast Fraser, A. D., R. A. Massom, and K. J. Michael, 2009: A method for sea ice: Links with bathymetry and atmospheric circulation. compositing polar MODIS satellite images to remove cloud J. Geophys. Res., 112, C02001, doi:10.1029/2006JC003559. cover for landfast sea-ice detection. IEEE Trans. Geosci. Re- ——, ——, and L. Shapiro, 2007b: How fast is landfast sea ice? A mote Sens., 47, 3272–3282, doi:10.1109/TGRS.2009.2019726. study of the attachment and detachment of nearshore ice at ——, ——, and ——, 2010: Generation of high-resolution East Barrow, Alaska. Cold Reg. Sci. Technol., 47, 233–255. Antarctic landfast sea-ice maps from cloud-free MODIS sat- Massom, R. A., 2003: Recent iceberg calving events in the Ninnis ellite composite imagery. Remote Sens. Environ., 114, 2888– Glacier region, East Antarctica. Antarct. Sci., 15, 303–313. 2896, doi:10.1016/j.rse.2010.07.006. ——, P. T. Harris, K. J. Michael, and M. J. Potter, 1998: The dis- Giles, A. B., R. A. Massom, and V. I. Lytle, 2008: Fast-ice distri- tribution and formative processes of latent-heat polynyas in bution in East Antarctica during 1997 and 1999 determined East Antarctica. Ann. Glaciol., 27, 420–426. using RADARSAT data. J. Geophys. Res., 113, C02S14, ——, K. L. Hill, V. I. Lytle, A. P. Worby, M. J. Paget, and I. Allison, doi:10.1029/2007JC004139. 2001: Effects of regional fast-ice and iceberg distributions on

Unauthenticated | Downloaded 10/02/21 01:08 PM UTC 1156 JOURNAL OF CLIMATE VOLUME 25

the behaviour of the Mertz Glacier polynya, East Antarctica. Spinhirne, J. D., S. P. Palm, and W. D. Hart, 2005: Antarctica Ann. Glaciol., 33, 391–398. cloud cover for October 2003 from GLAS satellite lidar ——, K. Hill, C. Barbraud, N. Adams, A. Ancel, L. Emmerson, and profiling. Geophys. Res. Lett., 32, L22S05, doi:10.1029/ M. Pook, 2009: Fast ice distribution in Ade´lie Land, East 2005GL023782. Antarctica: Interannual variability and implications for Em- Spreen, G., L. Kaleschke, and G. Heygster, 2008: Sea ice remote peror penguins (Aptenodytes forsteri). Mar. Ecol. Prog. Ser., sensing using AMSR-E 89-GHz channels. J. Geophys. Res., 374, 243–257. 113, C02S03, doi:10.1029/2005JC003384. ——, A. B. Giles, H. A. Fricker, R. Warner, B. Legre´sy, G. Hyland, Tamura, T., K. I. Ohshima, and S. Nihashi, 2008: Mapping of sea ice N. Young, and A. D. Fraser, 2010: Examining the interaction production for Antarctic coastal polynyas. Geophys. Res. between multi-year landfast sea ice and the Mertz Glacier Lett., 35, L07606, doi:10.1029/2007GL032903. Tongue, East Antarctica; Another factor in ice sheet stability? Tang, S., D. Qin, J. Ren, J. Kang, and Z. Li, 2007: Structure, salinity J. Geophys. Res., 115, C12027, doi:10.1029/2009JC006083. and isotopic composition of multi-year landfast sea ice in Nella McMinn, A., C. Ashworth, and K. Ryan, 2000: In situ net primary Fjord, Antarctica. Cold Reg. Sci. Technol., 49, 170–177. productivity of an Antarctic fast ice bottom algal community. Timmermann, R., and Coauthors, 2010: A consistent dataset of Aquat. Microb. Ecol., 21, 177–185. Antarctic ice shelf topography, cavity geometry, and global Meijers, A. J. S., A. Klocker, N. L. Bindoff, G. D. Williams, and ocean bathymetry. Earth Syst. Sci. Data Discuss., 3, 231–257, S. J. Marsland, 2010: The circulation and water masses of the doi:10.5194/essdd-3-231-2010. Antarctic shelf and continental slope between 308 and 808E. Ushio, S., 2006: Factors affecting fast-ice break-up frequency in Deep-Sea Res. II, 57, 723–737. Lu¨ tzow-Holm Bay, Antarctica. Ann. Glaciol., 44, 177–182. Murphy, E. J., A. Clarke, C. Symon, and J. J. Priddle, 1995: Tem- ——, 2008: Land-fast ice variation during the past five decades– poral variation in Antarctic sea-ice: Analysis of a long term Lu¨ tzow-Holm Bay, Antarctica. Ice Climate News, 11, 13–14. fast-ice record from the South Orkney Islands. Deep-Sea Res. Uto, S., H. Shimoda, and S. Ushio, 2006: Characteristics of sea-ice I, 42, 1045–1062. thickness and snow-depth distributions of the summer land- Nicholls, N., 2001: Commentary and analysis: The insignificance of fast ice in Lu¨ tzow-Holm Bay, East Antarctica. Ann. Glaciol., significance testing. Bull. Amer. Meteor. Soc., 82, 981–986. 44, 281–287. Ohshima, K. I., T. Kawamura, T. Takizawa, S. Ushio, and Williams, G. D., N. L. Bindoff, S. J. Marsland, and S. R. Rintoul, T. Miyakawa, 2000: Current variability under landfast sea ice in 2008: Formation and export of dense shelf water from the Lu¨tzow-Holm Bay, Antarctica. J. Geophys. Res., 105, 17 121– Ade´lie Depression, East Antarctica. J. Geophys. Res., 113, 17 132. 4039, doi:10.1029/2007JC004346. Purdie, C. R., P. J. Langhorne, G. H. Leonard, and T. G. Haskell, Williams, R. N., W. G. Rees, and N. W. Young, 1999: A technique 2006: Growth of first-year landfast Antarctic sea ice de- for the identification and analysis of icebergs in synthetic ap- termined from winter temperature measurements. Ann. Gla- erture radar images of Antarctica. Int. J. Remote Sens., 20, ciol., 44, 170–176. 3183–3199, doi:10.1080/014311699211697. Rintoul, S. R., 1998: On the origin and influence of Ade´lie Land World Meteorological Organization, 1970: WMO sea-ice nomen- Bottom Water. Ocean, Ice and the Atmosphere: Interactions at clature. Terminology, codes and illustrated glossary. Geneva the Antarctic Continental Margin. Antarctic Research Series, Secretariat of the World Meterological Organization Tech. Geophys Monogr., Vol. 75, Amer. Geophys. Union, 151–171. Rep. 259, 147 pp. Scambos, T., T. Haran, M. Fahnestock, T. Painter, and J. Bohlander, Young, N., B. Legre´sy, R. Coleman, and R. A. Massom, 2010: 2007: MODIS-based Mosaic of Antarctica (MOA) data sets: Mertz Glacier Tongue unhinged by giant iceberg. Austr. An- Continent-wide surface morphology and snow grain size. Re- tarc. Mag., 18, 19. mote Sens. Environ., 111, 242–257. Zwally, H. J., J. C. Comiso, C. L. Parkinson, W. J. Campbell, F. D. Smith, I. J., P. J. Langhorne, T. G. Haskell, H. J. Trodahl, R. Frew, Carsey, and P. Gloersen, 1983: Antarctic Sea Ice, 1973–1976: and M. R. Vennell, 2001: Platelet ice and the land-fast sea ice Satellite passive-microwave observations. NASA Special of McMurdo Sound, Antarctica. Ann. Glaciol., 33, 21–27. Publication SP-459. 459 pp. Smith, W. H. F., and D. Sandwell, 1997: Global seafloor topogra- ——, ——, ——, D. J. Cavalieri, and P. Gloersen, 2002: Variability phy from satellite altimetry and ship depth soundings. Science, of Antarctic sea ice 1979–1998. J. Geophys. Res., 107, 3041, 277, 1956–1962. doi:10.1029/2000JC000733.

Unauthenticated | Downloaded 10/02/21 01:08 PM UTC