Introduction

Antarctic ice shelves are prominent, constituent parts of ice sheets due to their ice–ocean–atmosphere interface and their vulnerability to regional and global changes in atmospheric and oceanic temperatures (Reference MercerMercer, 1978; Reference Vaughan and DoakeVaughan and Doake, 1996; Reference Shepherd, Wingham and RignotShepherd and others, 2004). The majority of mass loss from the Antarctic ice sheet is from ice shelves via either iceberg calving (the production of icebergs that break off from the ice shelf) or basal melting (the melting of ice at the bottom of the ice shelf due to higher ocean temperatures or increased pressure) (Reference Jacobs, Hellmer, Doake, Jenkins and FrolichJacobs and others, 1992; Reference Williams, Warner and BuddWilliams and others, 2002). Iceberg calving events are sporadic, but occur in a natural cycle of advance and retreat of the ice-shelf front over periods typically lasting several decades (Reference BuddBudd, 1966; Reference Scambos, Hulbe, Fahnestock, Domack, Burnett, Leventer, Conley, Kirby and BindschadlerScambos and others, 2003).

The retreat and disintegration of several ice shelves during the past few decades (Reference Rott, Rack, Skvarca and De AngelisRott and others, 2002; Reference De Angelis and SkvarcaDe Angelis and Skvarca, 2003; Reference Braun, Humbert and MollBraun and others, 2009) has raised significant concerns that these occurrences could become more frequent in a warming climate (Reference JanssenJanssen, 2009). As a result, large amounts of mass could be removed nearly instantaneously with calving events (Reference Bassis, Fricker, Coleman and MinsterBassis and others, 2008). The collapse of the Larsen B ice shelf (Reference Rignot, Casassa, Gogineni, Krabill, Rivera and ThomasRignot and others, 2004; Reference Scambos, Bohlander, Shuman and SkvarcaScambos and others, 2004) has proven that ice shelves are dynamically coupled with the flow of inland ice and can modulate ice flow far upstream of the grounding line. This phenomenon affects the discharge of grounded ice and has a direct influence on sea-level rise (Reference Bassis, Fricker, Coleman and MinsterBassis and others, 2008). To fully understand the processes behind changes in ice-shelf rifting systems and the interactions between the atmosphere, ocean and ice, it is necessary to monitor such systems over an extended period of time.

The Amery Ice Shelf is the largest ice shelf in East Antarctica (Fig. 1a). It drains continental ice from an area of >10⁶ km² (Reference AllisonAllison, 1991) through a section of coastline that represents ∼2% of the total continental circumference (Reference Budd, Smith, Wishart and OuraBudd and others, 1967). The last major calving event occurred in late 1963 or early 1964, when a ∼10 000 km² iceberg detached itself from the Amery Ice Shelf, removing nearly one-seventh of the shelf’s total surface area (Reference BuddBudd, 1966; Reference Fricker, Young, Allison and ColemanFricker and others, 2002a,Reference Frickerb). A precursor to calving is the initiation and subsequent propagation of rifts, which penetrate the entire ice shelf (Reference Fricker, Young, Coleman, Bassis and MinsterFricker and others, 2005). Rifts propagate and widen until they eventually become detachment boundaries for tabular icebergs. The Loose Tooth (LT) rift system (Fig. 1c) is an active rift system located at the front of the Amery Ice Shelf, which is likely to calve and produce a large iceberg in the near future. Additionally, an analysis of historical data indicates that a similar Loose Tooth-sized calving event preceded the last major calving event of 1963–64 (Reference Fricker, Young, Allison and ColemanFricker and others, 2002a).

Fig. 1. The Loose Tooth rift system. (a) MODIS Mosaic of Antarctica (MOA), 29 February 2004 (Reference Haran, Bohlander, Scambos and FahnestockHaran and others, 2005), showing the location of the Amery Ice Shelf in East Antarctica. (b) Map showing the formation of suture zone where two ice streams merged (gray curve). (c) Envisat ASAR image acquired on 11 February 2011 showing an enlarged view of the Loose Tooth rift system.

The system consists of two longitudinal-to-flow rifts ∼30 km apart that were shaped over the 1980s (denoted L1 and L2), and two transverse-to-flow rifts (denoted T1 and T2) that were initiated at the tip of L1 in 1995, when L1 was 23 km long (Reference Fricker, Young, Allison and ColemanFricker and others, 2002a, Reference Fricker, Young, Coleman, Bassis and Minster2005; Reference Bassis, Coleman, Fricker and MinsterBassis and others, 2005). Meanwhile, the longitudinal-to-flow strain rates began to exceed transverse-to-flow strain rates (Reference Young and HylandYoung and Hyland, 2002), and L2 stopped propagating (Reference Fricker, Young, Coleman, Bassis and MinsterFricker and others, 2005). There are two other longitudinal-to-flow rifts west of the LT rift system (denoted L3 and L4). The T2 tip propagated at ∼4 m d⁻¹ from late 1999 to early 2004, exhibiting a seasonal dependence with significantly higher rates during the austral summer period (Reference Fricker, Young, Coleman, Bassis and MinsterFricker and others, 2005); however, a slowing trend of rift T2 has recently been reported (Reference BassisBassis and others, 2007).

Data and Method

To quantify the rift changes, we measured the rift length over 9 years (February 2004 to February 2012) using Envisat ASAR data. Antarctic sea ice peaks in September (at the end of the Southern Hemisphere winter) and retreats to a minimum in February–March. From February 2004 to February 2012 we gathered data from Envisat ASAR, which operates in the C-band (5.6 cm wavelength), on the LT rift system in wide swath mode.

All Envisat ASAR data were processed with precise orbits and geometric corrections. We obtained images at a 75 m resolution with registration accuracy within one pixel. The histogram of each image was linearly stretched to enhance the rifts and then filtered according to an efficient imagedenoising scheme using principal component analysis with local pixel grouping (Reference Zhang, Dong, Zhang and ShiZhang and others, 2010). We then extracted the rift edges semi-automatically by combining an artificial drawing method with an improved watershed algorithm (Reference Wang, Gong, Ying, Yang, Cheng and RanWang and others, 2010). The rift lengths were estimated diagonally along the upstream rift wall according to the extracted edges from point ‘V’ in the triple junction (Fig. 1c) to the rift tips.

Results and Discussion

There has been an increase in the length and width of both rifts T1 and T2 over the past 9 years (Fig. 2). From 2004 to 2012, T1 lengthened from 5.3 to 18.3 km and T2 propagated from 17.0 to 24.3 km (Fig. 3a). T1 is propagating at ∼4.49 m d⁻¹, while T2 is growing at ∼2.53 m d⁻¹. Assuming that rift T2 will propagate at an average rate of ∼2.53 m d⁻¹ from 2013 to 2015, an iceberg between T2 and L1 will not calve during 2012–15, as projected previously by Reference Fricker, Young, Coleman, Bassis and MinsterFricker and others (2005). The exception is, of course, the occurrence of unforeseen events, including significant seismic activities or tsunami, as was the case with the iceberg break-off from the Sulzberger and Ross Ice Shelves (Reference MacAyeal, Okal, Aster and BassisMacAyeal and others, 2009; Reference Brunt, Okal and MacAyealBrunt and others, 2011). Rift T1 started to grow southward in February 2009 and then westward in February 2010 (Fig. 2).

Fig. 2. Loose Tooth edges extracted from Envisat ASAR (2004–12) with the background image of Envisat ASAR on 17 February 2012. The gray curve is the suture zone formed where two ice streams merged.

Fig. 3. Rift length and propagation rate. (a) Measured rift lengths derived from Envisat ASAR from February 2004 to February 2012. The slopes of the solid line and dotted line show that rift T1 has had a higher propagation rate than rift T2 since 2004. (b) Propagation rates for every 2 years from 2000 to 2012. The data from 2000 to 2012 are from Reference Fricker, Young, Coleman, Bassis and MinsterFricker and others (2005). The solid line shows the slowing of rift T2. The solid and dotted lines show a similar trend for both rifts since 2004.

Combining the measurement results from Reference Fricker, Young, Coleman, Bassis and MinsterFricker and others (2005) with the data in Figure 3a, Figure 3b indicates that rift T1 has had a higher propagation rate than rift T2 since 2004. However, the propagation rates of both rifts show a similar trend over the past 9 years (Fig. 3b), increasing from 2004 to 2005, 2008 to 2009, and 2010 to 2011, and decreasing from 2006 to 2007, 2009 to 2010, and 2011 to 2012. Furthermore, propagation rates of rift T1 have decreased more significantly than those of rift T2 since 2009, when rift T1 started to grow southward and then westward, almost parallel to the ice flow. We surmise that the changing distributions of the strain field in the rift tip might stop the rift forward. On the whole, both rifts show a retarding trend since 2005; Reference BassisBassis and others (2007) also found an overall decrease in propagation rates for T2 between 2000 and 2006.

Evidence has been presented that rift propagation occurs in episodic bursts, primarily driven by the internal stress in the ice shelf caused by the gravitational spreading of the ice. This evidence opposes the idea that short-term environmental forces (e.g. winds or ocean tides) can initiate propagation (Reference Bassis, Coleman, Fricker and MinsterBassis and others, 2005, Reference Bassis, Fricker, Coleman and Minster2008; Reference Joughin and MacAyealJoughin and MacAyeal, 2005; Reference JanssenJanssen, 2009). Internal stress is also a dominant force in rift widening (Reference Joughin and MacAyealJoughin and MacAyeal, 2005). Rift widths increase steadily at a rate similar to that of ice-shelf spreading, whereas rift lengths grow episodically by the coalescence of several discrete rift-tip propagation bursts, occurring as a series of micro- and mesoscale cracks ahead of the rift tips (Reference SchulsonSchulson, 2001; Reference Bassis, Coleman, Fricker and MinsterBassis and others, 2005, Reference Bassis2007; Reference Joughin and MacAyealJoughin and MacAyeal, 2005). According to Reference WeertmanWeertman (1957), the stress, τ_i, within the ice shelf is related to ice thickness and the densities of ice and water:

(1)

where ρ _i is the density of the ice, ρ _w is the density of the sea water and H _i is the ice thickness. The Amery Ice Shelf elevation change rate during 2003–08 is −0.06 ± 0.008 m a⁻¹ (Reference Pritchard, Ligtenberg, Fricker, Vaughan, Van den Broeke and PadmanPritchard and others, 2012), and the thickness change rate during 1994–2008 is 0.92 ma⁻¹ (Reference Shepherd, Wingham, Wallis, Giles, Laxon and SundalShepherd and others, 2010). Following Eqn (1), large-scale stress is relatively constant during 1994– 2008, as the ice thickness change is relatively small in proportion to the ice thickness (∼270 m near the LT rift system (Reference Craven, Allison, Fricker and WarnerCraven and others, 2009)).

However, melange, which fills the rifts and is composed of sea ice, accumulated and wind-blown snow, ice fragments broken off the ice shelf, and marine ice may modulate the stress field (Reference Rignot and MacAyealRignot and MacAyeal, 1998; Reference Khazendar and JenkinsKhazendar and Jenkins, 2003; Reference Fricker, Young, Coleman, Bassis and MinsterFricker and others, 2005). It has been suggested that melange may impede rift propagation and iceberg calving (Reference Hulbe, Rignot and MacAyealHulbe and others, 1998; Reference Rignot and MacAyealRignot and MacAyeal, 1998) by viscous coupling, and thick melange may likewise slow propagation (Reference Larour, Rignot and AubryLarour and others, 2004; Reference Fricker, Young, Coleman, Bassis and MinsterFricker and others, 2005). With its seasonally variable characteristics and side-walls that are thinner than the rift, melange may melt more easily than the surrounding ice shelf due to ocean warming (Reference Rignot and MacAyealRignot and MacAyeal, 1998; Reference Larour, Rignot and AubryLarour and others, 2004; Reference Fricker, Young, Coleman, Bassis and MinsterFricker and others, 2005). The composition as well as the accumulation and attrition process of melange ice may be a key factor in rift propagation.

Reference Oza, Singh, Vyas and SarkarOza and others (2011) studied the interannual variations in surface melting over the Amery Ice Shelf from 2000 to 2009, showing a decreasing trend in surface melting since 2005 (melting index (MI) and cumulative melting surface (CMS) in Fig. 4). Reference Craven, Allison, Fricker and WarnerCraven and others (2009) estimated a snow accumulation rate of 0.71 ± 0.08 m a⁻¹ near the LT. In addition, the basal melting rate of the Amery Ice Shelf dramatically decreases from the southern grounding zone to the downstream ice shelf; the basal melting rate at the front is −9.5 ± 5.6 Gt a⁻¹, with a net basal refreezing rate of 0.2 ± 0.1 m a⁻¹(Reference Yu, Liu, Jezek, Warner and WenYu and others, 2010). Reference Wen, Wang, Wang, Jezek, Liu and AllisonWen and others (2010) also calculated the spatial distribution of the basal melting/ freezing rates (m a⁻¹)beneath the Amery Ice Shelf, and found that the tips of both rifts are located within the freeze-on area. There remains a significant amount of marine ice formed in the calving front through the refreezing process, originally found by Reference Fricker, Popov, Allison and YoungFricker and others (2001), which could bind the ice blocks together, give the melange more mechanical strength, and slow rift propagation (Reference BassisBassis and others, 2007). The average accretion rate of marine ice near the calving front is 1.1 ± 0.2 ma⁻¹. We speculate that the melange thickness in LT might be increasing with the decrease of surface and basal melting, the increase of surface accumulation and basal refreezing. This would slow rift propagation.

Fig. 4. Interannual variations of the melting index (MI), cumulative melting surface (CMS), cumulative monthly mean air temperature (CMAT) and the number of positive degree-days with temperatures greater than 4°C (PDD4) for the month of January from 1998 to 2010 over the Amery Ice Shelf (Reference Oza, Singh, Vyas and SarkarOza and others, 2011).

As an external forcing for rifts, wind forcing could influence the amount of wind-transported snow, as well as the surface and basal melting rates (Reference Leonard, Tremblay, MacAyeal and JacobsLeonard and others, 2008; Reference Pritchard, Ligtenberg, Fricker, Vaughan, Van den Broeke and PadmanPritchard and others, 2012). Such influences could then affect the rift motion over a long-term scale. Moreover, the changing propagation rates of both rifts (Fig. 3b) share a similar trend with air temperature (CMAT in Fig. 4) and the number of positive degree-days with temperatures greater than 4°C (PDD4 in Fig. 4) for January 2004–09 (Reference Oza, Singh, Vyas and SarkarOza and others, 2011), indicating that air temperature is an important environmental force. Both wind forcing and air temperatures may influence rift propagation by affecting the melange ice thickness and other properties, which further demonstrates that melange ice is a key influencer in rift motion. Additionally, tips of both rifts encountered suture zones (Figs 1b and 2), which could reduce stress concentrations ahead of the rift tips (Reference BassisBassis and others, 2007), potentially hindering them.

Conclusion

Time-series observations of the LT rift system from 2004 demonstrate the obvious propagation and widening of two rifts, primarily driven by internal stress. With an average propagating speed of 4.49 m d⁻¹ for T1 and 2.53 m d⁻¹ for T2 and barring the occurrence of any unforeseen events, the rift will not break during 2012–15 as previously projected by Reference Fricker, Young, Coleman, Bassis and MinsterFricker and others (2005). Rift T1 has changed its propagation direction since 2009. Moreover, rift T1 has had a higher propagation rate than rift T2 since 2004; however, the propagation rates of both rifts have shown a similar downturn since 2005. Thickening melange ice in the rifts or an encounter with the suture zone could potentially affect propagation rates. Wind-forcing and air-temperature changes may lead to changes in the melange ice, consequently impacting rift development. As to the undetermined factors, additional field studies and autonomous observations are necessary to further enhance our understanding of rift propagation mechanisms and to predict calving times.