Volume 42, Issue 18 p. 7598-7605
Research Letter
Free Access

Sea ice convergence along the Arctic coasts of Greenland and the Canadian Arctic Archipelago: Variability and extremes (1992–2014)

Ron Kwok

Corresponding Author

Ron Kwok

Jet Propulsion Laboratory, California Institute of Technology, Pasadena, California, USA

Correspondence to: R. Kwok,

[email protected]

Search for more papers by this author
First published: 11 September 2015
Citations: 55


After the summer of 2013, a convergence-induced tail in the thickness distribution of the ice cover is found along the Arctic coasts of Greenland and Canadian Arctic Archipelago. Prompted by this, a normalized ice convergence index (ICI) is introduced to examine the variability and extremes in convergence in a 23 year record (1992–2014) of monthly ice drift. Large-scale composites of circulation patterns, characteristic of regional convergence and divergence, are examined. Indeed, the ICI shows the June 2013 convergence event to be an extreme (i.e., ICI > 2). Furthermore, there is a cluster of 9 months over a 17 month period with positive ICIs (i.e., >1) following the record summer minimum ice extent (SMIE) in 2012; the imprint of ice dynamics from this cluster of positive ICIs likely contributed to higher SMIEs in 2013 and 2014. The impact of convergence on SMIE is discussed, and the increase in Arctic ice volume in 2013 is underscored.

Key Points

  • Impact of regional ice convergence on thickness distribution and sea ice roughness
  • Record of extremes and variability of regional sea ice convergence since 1992–2014
  • Potential impact of ice dynamics on observed variability of summer ice extent in 2013 and 2014

1 Introduction

Beyond thermodynamic ice growth, it is ice dynamics that maintains the characteristic tails of the thickness distributions, g(h) (see examples in Figure 1a) of Arctic sea ice [Thorndike et al., 1975]. Deformed and thicker ice (rafts and pressure ridges) is formed by convergence and shear between ice floes and, in particular, convergence at coastal boundaries of the Arctic Basin. The thicker ice is more likely to survive the summer to form the “perennial” ice cover of the Arctic Ocean and slow the decline of sea ice coverage. Deformation-induced redistribution also increases local ice strength and roughness and alters the mechanical response of the ice cover to atmosphere and ocean forcing [e.g., Martin et al., 2014]. In the melt season, there are close linkages between melt pond coverage and surface relief created by ice deformation—meltwater tends to collect on topographically low ice, and topography controls pond depth and limits the horizontal distance water can travel [e.g., Eicken et al., 2004].

Details are in the caption following the image
Ice convergence near the Arctic coasts of the Canadian Arctic Archipelago in summer of 2013. (a) Two-month ice thickness distributions within the multiyear ice cover (>70% coverage) during the growth season before (2012–2013) and after (2013–2014) the summer convergence in 2013 (quantities show mean/standard deviations of each distribution). (b, top) Mean June–August ice drift and (b, bottom) convergence of ice cover from change in polygon areas between October (dashed) and May (colored); colored polygon is obtained by back propagation of dashed polygon boundaries from October to May with observed ice drift. (c) Contrasts in thickness changes due to convergence and melt (within defined polygon) in 2013 compared to 2011 and 2012. Ice thickness estimates are from CryoSat-2 (KC15).

Even though ice deformation is of geophysical interest in a number of contexts [e.g., Kwok and Sulsky, 2010], progress in improving and verifying ridging schemes in sea ice models has been slow. Capturing and resolving changes in g(h) at small spatial scales (e.g., in the field) have been proven to be difficult—especially the mechanical redistributions associated with the nonuniform ice motion between ice floes in the open Arctic Ocean [Lipscomb et al., 2007]. At larger spatial scales, detecting changes in g(h) at the appropriate temporal scales was not available until the near-monthly coverage of the Arctic provided by CryoSat-2 (CS-2). The regional convergence of sea ice from ice drift and the availability of observed g(h) before and after the summer of 2013, reported by Kwok and Cunningham [2015] (henceforth, KC15), offered a first look at the potential impact of large-scale ice convergence on the redistribution of thick ice along the Arctic coasts of Greenland and the Canadian Arctic Archipelago (CAA). At the end of summer 2013, KC15 and Tilling et al. [2015] found thicker ice and a longer tail in the ice thickness distribution in the multiyear ice (MYI) cover northwest of the CAA. In addition to a cooler summer, the mean circulation pattern (noted in KC15) also showed that the thicker ice is attributable to on-shore ice drift that contributed to significant regional convergence (up to 23%, as sampled by the red ice parcel, see Figure 1b, right) and deformation of the ice cover.

In this note, the drift pattern that led to the ice convergence in KC15 is examined more closely, and this occurrence is placed within the context of monthly drift patterns between 1992 and 2014. The following questions are addressed: (1) What is the contribution of the cooler summer and convergence in 2013 to the thicker ice cover? (2) Is the observed regional convergence in summer of 2013 an atypical occurrence in the 23 year record of variability and extremes?, and (3) What is the potential impact of ice convergence in this region on summer ice extent? The paper is organized as follows. The next section describes the data set used. Section 3 reviews briefly the observations from the summer of 2013, addresses the first question, and describes an ICI index used as an indicator of the regional ice convergence along the Arctic coasts of Greenland and the CAA. The 2013 summer is examined within the context of the variability of the ICI over the 23 year record. The potential impact of convergence events is reviewed, and the last section concludes the paper.

2 Data Description

2.1 Satellite Ice Motion, Multiyear Ice Fraction, and Ice Thickness

Gridded fields of ice drift are blended ice motion (100 km grid spacing) from two Special Sensor Microwave Imager (SSMI) radiometer channels (37 and 85 GHz). Ice motion estimates are derived from sequential fields of daily passive microwave brightness temperature using the procedures described by Kwok et al. [1998] and Kwok [2008]. Gridded fields (25 km spacing) of MYI fraction are derived from radar backscatter acquired by the Advanced Scatterometer (ASCAT) and processed using the approach in Kwok [2004]. ASCAT is a moderate resolution wide-swath C-band scatterometer that provides daily mapping of the Arctic Ocean. Gridded ice thickness fields from CS-2 are those discussed in KC15.

2.2 IceBridge ATM Elevations

The Level-1B lidar Elevation and Return Strength data set from Airborne Topographic Mapper (ATM) data [Krabill, 2014] from the Operation IceBridge campaigns in 2013 and 2014 is used to provide measures of ice surface roughness. The data contain ATM spot elevation measurements (~1 m footprint) over sea ice. The ATM scanning geometry provides an across-track scan swath of 250 m with typical elevation accuracy better than 10 cm. Ice surface roughness is calculated as the detrended standard deviation of ATM lidar elevations (over 25 km segments) from Operation IceBridge. Detrending removes potential biases in local height variability due to geoid residuals in the elevation data.

3 Data Analysis

In this section, the ice convergence near Greenland and the CAA in summer of 2013 is reviewed, the relative contribution of reduced melt/convergence to the observed October ice thickness is examined, an index for measuring ice convergence and for examining the variability and extremes of ice convergence in a 23 year record is introduced, and the potential impact of ice convergence on summer ice extent is discussed.

3.1 Convergence and Reduced Melt in Summer 2013

Here the focus is on ice convergence that occurred during the summer months of 2013—details regarding the 4 years (2011–2014) of Arctic ice thickness and volume from CS-2 referenced here can be found in KC15. Compared to the preceding winter of 2012–2013 (Figure 1a), the g(h) of the MYI cover (with >70% coverage) in 2013–2014 is thicker ice and has a longer tail in all months. Examination of ice drift suggests that the observed g(h) is attributable to convergent ice motion seen in the mean summer (June–August) drift pattern (Figure 1b, right). The strong wind-driven onshore ice drift is forced by the relative location of the high- and low-pressure centers over the Arctic Ocean. The high is centered just north of Mackenzie Bay with the low centered near the North Pole. Regionally, the convergence sampled by ice parcel (initial area in red: Figure 1b, right) shows an area reduction of 23% (final area: polygon with dashed boundaries) between May and October. This is equivalent to an increase in thickness of ~30% if ice volume was conserved. The ice thickness in October of 2013 ( urn:x-wiley:00948276:media:grl53483:grl53483-math-0001), at 2.43 m, is more than a meter thicker that the October thicknesses in the preceding 2 years (see Figure 1c). Since the summer of 2013 is also cooler [KC15; Tilling et al., 2015], one question is whether urn:x-wiley:00948276:media:grl53483:grl53483-math-0002 is a consequence of change in thickness due to reduced melt ( urn:x-wiley:00948276:media:grl53483:grl53483-math-0003) or to convergence ( urn:x-wiley:00948276:media:grl53483:grl53483-math-0004).

Given the observed ice thickness in May (hmay) and October (hoct), urn:x-wiley:00948276:media:grl53483:grl53483-math-0005 and urn:x-wiley:00948276:media:grl53483:grl53483-math-0006 can be estimated by first calculating the thickness of the May ice parcel ( urn:x-wiley:00948276:media:grl53483:grl53483-math-0007) due to the observed convergence (α)—see Figure 1b. Within the ice parcel, the difference between urn:x-wiley:00948276:media:grl53483:grl53483-math-0008 and hmay then gives urn:x-wiley:00948276:media:grl53483:grl53483-math-0009 due to convergence. Finally, differencing urn:x-wiley:00948276:media:grl53483:grl53483-math-0010 and hoct gives the change urn:x-wiley:00948276:media:grl53483:grl53483-math-0011 due to melt. Consistent with a cooler summer, the results (in Figure 1c) show that ice melt in 2013 ( urn:x-wiley:00948276:media:grl53483:grl53483-math-0012 = 1.13 m) is lower by ~0.4 m when compared to 2011 and 2012 (~1.53 m). In contrast, urn:x-wiley:00948276:media:grl53483:grl53483-math-0013 due to convergence, at 0.84 m, is twice as large. Ice convergence was negligible in 2012, and the ice cover was in fact divergent in 2011. Thus, the contribution of ice convergence to urn:x-wiley:00948276:media:grl53483:grl53483-math-0014 is more significant in 2013 summer even though the resultant thickness is a consequence of both convergence and, to a less extent, a cooler summer.

In a different analysis, Ricker et al. [2015] considered the higher MYI freeboard in November 2013 compared to March 2013 to be unlikely after the melt season and surmised that higher ice freeboards may be due to retrieval issues. Even though the retrieval issues merit attention, they neglected to consider the potential of ice convergence and deformation (discussed above) in creating such inconsistencies. Broadly, instead of attributing the ice thickness of this region solely to thermodynamics, the results here highlight the role of dynamics as a source of variability in Arctic Ocean ice thickness that should not be discounted when interpreting retrieval results.

3.2 Observed Convergence in Summer 2013: Is This an Extreme Event?

To place the summer 2013 convergence into a broader context, the drift patterns in a 23 year record (1992–2014) of monthly ice drift is examined. As a measure of the regional convergence in the vicinity of CAA and Greenland, a fluxgate is positioned across the Arctic (dashed red line in Figure 2b) that is nearly parallel to the coast of the CAA and Greenland. The monthly ice area flux (in km2) across this gate is then used as a measure of the strength of onshore drift. A time series of normalized ice convergence index (ICI, see Figure 2a) is then constructed with the monthly area flux, F(t), where
Details are in the caption following the image
Normalized ice convergence index (ICI). (a) Monthly variability of ICI between 1992 and 2014 (middle). Top/bottom graphs show five positive/negative extremes (i.e., |ICI| > 2.0) for the period. (b) Composites of mean ice drift over the Arctic Ocean for positive (ICI > 1.0) and negative (ICI < 1.0) deviations and average (−1.0 ≤ ICI ≤ 1.0) values of the index. Total and number of winter (W: October–May)/summer (S: June–September) months in each composite are indicated. Contours are isobars of sea level pressure from National Centers for Environmental Prediction/National Center for Atmospheric Research reanalysis (interval: 4 hPa in Figure 2a and 2 hPa in Figure 2b). Dashed line (red) in Figure 2b (left) is the gate used for area flux calculations.

urn:x-wiley:00948276:media:grl53483:grl53483-math-0016 and σF are the mean and standard deviation of the monthly area flux over the period (1992–2014). Positive deviations from the mean (i.e., ICI > 0), advection toward the CAA and Greenland coasts, are indicative of convergence. For the 23 years, urn:x-wiley:00948276:media:grl53483:grl53483-math-0017 is 38.0 × 103 km2 and σF is 64.0 × 103 km2. In effect, there is always a mean flux through the gate and the circulation is, in the mean, slightly convergent. Based on the analysis of errors in monthly ice flux described by Kwok and Rothrock [1999], uncertainty in ICI is ~0.2. There is no statistically significant trend in ICI for this period.

To characterize the drift and large-scale atmospheric patterns that are conducive to local convergence and divergence, composite patterns of mean monthly sea level pressure and ice drift based on the positive (ICI+: ICI > +1.0) and negative (ICI: ICI < −1.0) deviations and neutral values (ICIo: −1.0 ≤ ICI ≤ 1.0) of the index are constructed. The ICI+ composite (Figure 2b, left) shows the positioning of mean high- (in the central Beaufort) and low- (in Iceland) pressure centers act to drive the onshore flow. The axis of the Transpolar Drift Stream (TDS) is rotated westward resulting in reduced advection of ice from the Beaufort Sea and Siberian Coast toward the Fram Strait. During ICI, the high-pressure center is situated in the center of the Arctic Basin driving coastal divergence along the CAA coast with a TDS that delivers ice from the Beaufort Sea to the Fram Strait. As expected, patterns selected by the ICI have near-opposing effects that impact the behavior of the ice cover (discussed in following sections). The mean circulation pattern of the Arctic, with a characteristic Beaufort Gyre and TDS, is more evident in ICIo. The seasonal loading of the composite patterns (i.e., number of winter (October–May) and summer (June–September) months) is uneven. The more summer months in ICI+ and less winter months in ICI is likely associated with the different mean drift/atmospheric patterns in winter and summer [see Kwok et al., 2013]. Examples of five monthly drift fields with |ICI| > 2.0, 2 standard deviations from the mean, shown in Figure 2a (top and bottom) can be compared to the ICI+ and ICI composites with strongly convergent and divergent drift patterns, respectively.

Does the summer of 2013 represent an extreme on record? Indeed, the month of June (with ICI = 2.1—see Figure 2a) can clearly be considered an extreme when ICI > 2.0 (or 2 standard deviations from the mean) in the 23 year record. Combined with positive ICIs in July and August (0.9 and 1.7, respectively), there is a net ice convergence over the summer. The mean June-August ice drift pattern (Figure 1b) can be compared to the ICI+ composite.

3.3 Variability and Extremes in ICI (1992–2014)

Since ice drift is largely wind driven, the ICI can be related to larger-scale atmospheric oscillations. Correlations of the monthly ICI time series to the Arctic Oscillation (AO) and the North Atlantic Oscillation (NAO) indices (0.42 and 0.33, respectively) are moderate. Thus, the drift patterns due to the juxtaposition of the high- and low-pressure centers in the two ICI extremes are related to the hemispheric-scale AO and NAO.

There are two periods of note in the 23 year ICI record. Based on the local density of extreme ICIs, two clusters stand out (circled in Figure 2): one between November 1997 and May 1999 and the second—more recent—is between November 2012 and March 2014. In the first cluster, there were 8 months (out of the 19 month period) with ICI < −1.0, and in the second cluster, there were 9 months (out of 17 month period) with ICI > 1.0 and four of which with ICI > 2.0. The period of the first cluster coincides with the year of the Surface Heat Budget of the Arctic Ocean (SHEBA) field campaign [Uttal et al., 2002]. Unfortunately, the westward drifting station was established in October 1997, and thus, the SHEBA records do not contain observations of ice condition upstream in the CAA region for large-scale assessment of ice thickness anomalies due to the sustained negative ICI extremes. Here the focus is on the second cluster that appeared after the record summer minimum ice extent (SMIE) in summer of 2012 and spans the summer of 2013, which motivated the present study. In fact, there were five other ICI+ months after the summer of 2013 that created more thick ice in the region. Within this period, the SMIE of 2013 and 2014 are both higher than the SMIE of 2012 by over ~1.5 × 106 km2, and the Arctic Ocean ice volume had increased by ~2500 km3 (KC15). Could this clustering of extreme ICI+s be related to the larger SMIEs following the record of 2012? This is discussed below.

3.4 Potential Impact of Ice Convergence on SMIEs

Direct attribution of SMIE to ice deformation may be difficult since regional convergence or divergence in the vicinity of the CAA and Greenland coasts affects the response of the ice cover to subsequent atmospheric and ocean forcing both directly and indirectly and at varying time and space scales. The effects of convergence depend on the location and extent of thick/deformed ice during summer melt (therefore, advection) coupled with the time-varying summer melt processes (e.g., melt pond evolution) that are less well understood at this time [e.g., Kwok and Untersteiner, 2011]. Below, only potential impacts of ice convergence on SMIE are discussed because comprehensive observations necessary to trace and understand the evolution of the needed sea ice parameters (ice thickness, surface roughness, and melt pond coverage) are not available. Observational issues are addressed briefly in the last section.

3.4.1 Thicker Ice

The mean annual ice regional convergence near the CAA and Greenland coasts serves to maintain the tails in the thickness distribution and to produce the thickest ice found in the Arctic Ocean. To first order, the most direct and immediate impact of regional ice convergence on SMIE, as discussed here, is the advection of that thicker ice into the Beaufort Sea. Since thicker ice is more likely to survive the summer, the advection of this ice into this region of high melt rates in the summer would reduce overall area loss.

3.4.2 Surface Roughness and Melt Ponds

Perhaps just as important, but less direct, is the impact of roughness of the deformed ice surface on melt pond coverage. During summer melt, water from snow and ice melt accumulates in ponds with low albedo. This further enhances the melt rate, creating a strong positive feedback between the absorbed downwelling shortwave radiation and the resulting melt pond coverage of the ice surface [Kwok and Untersteiner, 2011]. Available meltwater covers a larger area over smoother seasonal ice compared to that of rougher, deformed ice [Fetterer and Untersteiner, 1998; Polashenski et al., 2012]. A recent study by Schröder et al. [2014], using a melt pond model with more realistic processes (rather than based on parameterization) that includes surface topography [Scott and Feltham, 2010], suggests strong correlations between spring pond fraction and SMIE.

Since the deformed ice discussed here was created during a 17 month period, it could have a profound impact on the SMIEs in 2013 and 2014. Figure 3 shows snapshots of surface roughness from two spring Operation IceBridge campaigns contrasting the smoother ice cover in 2013 with the rougher ice that covers a larger area in 2014; these snapshots show a rougher ice cover (see inset of Figure 3b) created during the cluster of ICI+ months. This suggests that increased ice roughness prior to the 2013 and 2014 melt seasons may have lowered the melt pond area fraction, and hence the albedo during summer of 2013/2014, potentially contributing to higher SMIE. This lends credence to substantial influence of the combined impact of surface topography and meltwater on ice loss during the melt season.

Details are in the caption following the image
Comparison of surface roughness in (a) March–April 2013 and (b) March–April 2014. MYI ice regions with >50/70% coverage are shown in dark/light gray. Inset in Figure 3b shows distributions of surface roughness in areas with >70% MYI (quantities are means and standard deviations of distributions).

3.4.3 Timing of Convergence Events

It is also important to note that mechanical redistribution (convergence) of thinner seasonal ice into deformed ice (from larger into smaller areas) could have different consequences; redistribution creates thicker ice while reducing thickness of thinner parts of the g(h); i.e., for a given area, average volume remains unchanged. The impact of convergence in the summer is different from that in the winter. During the winter or growth season, the ice area in the Arctic Ocean depleted by ice convergence is replaced by seasonal ice. Depending on the winter conditions, water in the opened areas (convergence leads to openings elsewhere) has an opportunity to grow and replenish the ice area consumed in deformation, contributing to increased ice production.

This is not true of deformation during summer melt. Since there is no freezing of the opened areas, deformation contributes directly to open water production or reduction of ice concentration during the summer. Also, convergence/divergence changes absorbed downwelling shortwave radiation by decreasing/increasing low albedo water surfaces and surface air temperature. To effect a higher SMIE, it would be better to have higher ice deformation during the early winter and spring—coupled with favorable drift patterns—rather than the summer. The higher 2013 SMIE is likely the response of the ice cover to a combination of four ICI+ months (in the 17 month cluster between November 2012 and March 2014) prior to the summer of 2013, together with the cooler summer temperatures in the Arctic, rather than just the extreme ICI in June. The effects of ice convergence on SMIE deserve attention when considering the predictability of Arctic ice coverage.

4 Conclusions

The present note examines regional ice convergence, near the coasts of Greenland and the Canadian Arctic Archipelago (CAA), during the summer of 2013 and its significance within the context of a 23 year record (1992–2014). Compared to the preceding years, the contribution of ice convergence to the thicker multiyear ice cover in October 2013 is higher than that of reduced melt due to a cooler summer.

A normalized ice convergence index (ICI), which measures the magnitude of monthly onshore ice drift in the region, is introduced to assess the variability and extremes in ice convergence. The record indicates that the ICI in June of 2013 was indeed an extreme (ICI > 2.0, 2 standard deviations from the mean) and August was also high (ICI = 1.7). Moreover, both months belonged to a 17 months period (cluster), between November 2012 and March 2014, that was dominated by positive ICIs: there were 9 months with ICI > 1.0, in which four had ICI > 2.0. This 17 month cluster of ice convergence extremes follows the record summer minimum ice extent (SMIE) of 2012: the period includes the summer of 2013 and the higher SMIEs in 2013 and 2014—both of which are >1.5 × 106 km2 higher than the SMIE of 2012. In addition, there was an increase in Arctic Ocean ice volume by ~2500 km3 and longer tails are seen in the ice thickness distributions (KC15).

Ice convergence near the Arctic coasts of Greenland and the Canadian Arctic Archipelago (CAA) is a source of some of the thickest ice in the Arctic and alters the response of the ice cover to atmospheric and oceanic forcing at different time and space scales. Certainly, the thicker, more deformed ice adds to the variability and challenges the predictability of the ice cover in response to summer melt. In addition to thickness, the expected linkages between surface roughness, meltwater coverage, and September ice extent are recognized—increases in roughness reduce melt pond coverage and reduce ice loss associated with insolation. The timing of the convergence events, whether during the growth or melt season, is also an important consideration. Further, the difficulty in direct attribution of the SMIE to ice convergence due to its longer-range effects (in time and space) and to the variety of factors that contributes to ice extent at the end of the summer is recognized: it preconditions the ice cover but could be overwhelmed by summer forcing. For the 17 month cluster of positive extremes in ICI, the signature of ice convergence on the increased SMIE in 2013/2014 and ice volume is strongly suggested.

Here observations suggest a linkage between ice convergence and the variability of summer ice coverage. However, a detailed understanding of the processes has yet to be elucidated. The needed observations for a more detailed analysis of the impact of ice deformation on SMIE are lacking. A useful time series of surface observations would capture the regional ice convergence and trace the trajectory and the evolution of surface parameters (ice thickness, surface roughness, and melt pond) on a fixed collection of ice floes. This highlights the need for close coupling of observations of the surface, typically acquired independently, required to resolve and separate the contributions of the thermodynamics and dynamics processes that are necessary for quantifying ice thickness, surface roughness, and pond coverage. Thus, even though near-simultaneous observations of these parameters are difficult with current satellite and field technologies, the importance of the coordination between these observations deserves special attention for future planning.


I thank the two anonymous reviewers for their valuable comments and suggestions and Shirley Pang for software support during the course of this work. Data from Operation IceBridge and the Special Sensor Microwave Imager (SSMI) are available through the National Snow and Ice Data Center (NSIDC—http://nsidc.org); data from the EUMETSAT Advanced Scatterometer (ASCAT) are available through the Physical Oceanography Distributed Active Archive Center (PO.DAAC—http://podaac.jpl.nasa.gov) at the Jet Propulsion Laboratory. R.K. performed this work at the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration.

The Editor thanks two anonymous reviewers for their assistance in evaluating this paper.