Arctic marginal ice zone trending wider in summer and narrower in winter



[1] Declines in Arctic sea ice extent and thickness suggest scientifically important changes in the spatial pattern of the marginal ice zone (MIZ)—a dynamic and biologically active band of sea ice cover close to open ocean. Arctic MIZ widths were measured in satellite era sea ice concentrations over the years 1979–2011 using a recently published MIZ width analysis method. Over the record, the warm season (July–September) MIZ width increased by 13 km decade−1, amounting to a 39% widening. This widening trend resulted from a marked poleward advancement of the MIZ poleward edge into regions where sea ice was increasingly younger and thinner at the beginning of annual melt, while the MIZ's equatorward edge moved comparatively little. The warm season MIZ widening contrasted in sign and strength with a cold season (February–April) MIZ narrowing trend (4 km decade−1, amounting to a 15% narrowing).

1 Introduction

[2] The decline of Arctic sea ice extent [Cavalieri and Parkinson, 2012] and the replacement of thick, multiyear ice by thin, first-year ice [Maslanik et al., 2011] suggest scientifically important, yet largely unstudied, changes in the position, width, and area of the marginal ice zone (MIZ)—a dynamic and biologically active band of sea ice cover close to open ocean [Squire, 1998]. MIZ width is a fundamental length scale for polar ecosystem dynamics and climate dynamics [Wadhams, 2000]. The width of the MIZ is a buffer zone that protects the stable morphology of the inner ice from wave penetration [Squire, 2007], represents the distance over which the atmospheric boundary layer converts to its stable polar form [Shaw et al., 1991], establishes an important spatial dimension for marine habitat selection [Ribic et al., 1991], and impacts human accessibility to the Arctic [Rogers et al., 2012].

[3] The importance of the MIZ has motivated some estimates of its width based on the early satellite record [Comiso and Zwally, 1984], field campaigns [Muench, 1983], and ship-based observations [Xie et al., 2011]. The National Ice Center provides a daily MIZ product based on manual inspection of various passive and active imagery since July 2006 [NIC, 2011]. Here we apply our recently published MIZ analysis method [Strong, 2012] to satellite data to uncover a marked widening trend in the warm season MIZ and contrast it with comparatively modest cold season MIZ narrowing.

2 Data and Methods

2.1 Marginal Ice Zone Identification and Width Analysis

[4] We analyzed the warm season (July–September) MIZ and cold season (February–April) MIZ over the domain and subdomains shown in Figure 1a. MIZ identification and width analysis followed algorithms detailed in [Strong, 2012], and the method is briefly summarized here using 29 August 2010 as an example (Figures 1b and 1c). For MIZ identification, sea ice concentrations included values interpolated over islands smaller than Greenland (gray shading, Figure 1a; black outlines, Figure 1b) and the MIZ was defined as a body of marginal ice (0.15≤c≤0.80) that adjoined both pack ice (c>0.80) and sparse ice (c<0.15). The 0.80 threshold was chosen to match the maximum concentration considered to be “close ice” by the World Meteorological Organization [WMO, 1985] and to match the upper concentration limit used in the National Ice Center (NIC) daily MIZ product.

Figure 1.

(a) The full analysis domain is shown and was subdivided into the East Siberian-North American (ES-NA) sector and the Atlantic sector (separated by 90°E and 90°W). As in [Strong, 2012], no analyses were performed for semienclosed water bodies (orange shading) and sea ice concentrations were interpolated over islands smaller than Greenland and not adjoined with semienclosed water bodies (gray shading). (b) From 29 August 2010 sea ice concentrations retrieved by the Bootstrap algorithm [Comiso, 1986]: pack ice (gray; c>0.80), marginal ice (white; 0.15≤c≤0.80), and sparse ice (blue; c<0.15). Orange shading shows semienclosed water bodies, and black curves show small islands over which sea ice concentrations were interpolated. (c) Shading shows solution to Laplace's equation within the MIZ (ψ), and width measurements following the gradient of ψ are black curves (only subset shown for clarity).

[5] Once the MIZ boundaries and area were identified (e.g., white shading, Figure 1b), we solved for the idealized (smooth) sea ice concentration field (ψ; shading Figure 1c) satisfying Laplace's equation ∇2ψ=0 within the MIZ with boundary conditions ψ=0.15 at the sparse ice edge, ψ=0.80 at the pack ice edge, and ψ=0.48 where the MIZ met land (0.48 is the average of 0.80 and 0.15). MIZ width () was then the length of a curve following the gradient of ψacross the MIZ (black curves, Figure 1c). For each day, a hemispheric summary measure of MIZ width (w) was defined by averaging with respect to distance along the MIZ perimeter, and we time averaged the daily w values for 3 month blocks during each year (denoted math formula for February–April (the “cold season”) and math formula for July–September (the “warm season”)). We calculated an area-weighted MIZ latitude for July–September of each year (math formula). The probability of MIZ (PMIZ) at a location for some period of time was the number of days that MIZ was observed at that location divided by the number of days in the period (0≤PMIZ≤1).

2.2 Sea Ice Data

[6] We identified and measured MIZ width in three different sets of passive microwave sea ice concentration data from the National Snow and Ice Data Center (NSIDC): concentrations based on the Bootstrap algorithm [Comiso, 1986], concentrations based on the National Aeronautics and Space Administration (NASA) Team algorithm [Cavalieri et al., 1984], and concentrations from the “Climate Data Record of Passive Microwave Sea Ice Concentration” (CDR) which is a blending of different algorithms intended to produce a consistent record over time [Meier et al., 2011]. We also measured MIZ width in the National Ice Center (NIC) daily MIZ product after projecting NIC daily shape files [NIC, 2011] onto the Bootstrap stereographic grid. In all cases, the polar data gap and scattered pixels of missing data were filled using a thin plate spline radial basis function [Duchon, 1977] on the NSIDC stereographic projection. We found that, for periods of data set overlap, warm season MIZs measured in NASA Team data were approximately three times wider than Bootstrap-based MIZs (Figure S1 in supporting information), a result consistent with possible low bias in the NASA Team algorithm [Notz, 2013]. In contrast, seasonal mean Bootstrap-based MIZ widths agreed with CDR-based widths to within 11 km and agreed with NIC-based widths to within 30 km (Figure 2a). CDR and NIC data were available only back to 1987 and 2006, respectively, whereas Bootstrap concentrations extended back to November 1978. The results presented in section 3 are based on the Bootstrap concentrations extending back to 1979 (the first full year of data).

Figure 2.

(a) Time series of MIZ width for two seasons and three data sets. Trend lines for Bootstrap results are shown in black. Maps at right show examples of pack ice (gray; c>0.80), MIZ (white; 0.15≤c≤0.80), and sparse ice (blue; c<0.15) for (b) 29 August 1983 and (c) 29 August 2010.

[7] We used an April mean sea ice age (math formula) derived by combining sea ice concentration data with sea ice motion analyzed from tracks of drifting buoys and manned stations maintained by the International Arctic Buoy Programme [Rigor et al., 2002; Rigor and Wallace, 2004]. We used an April mean sea ice thickness (math formula) calculated from the observationally forced Pan-Arctic Ice-Ocean Modeling and Assimilation System [Zhang and Rothrock, 2003; Schweiger et al., 2011].

3 Results

[8] Cold season MIZ width (math formula; blue curve, Figure 2a) averaged 71 km over 1979–2011, and the downward trend line (4 km decade−1; Figure 2a) indicates a 15% narrowing over the record. The cold season narrowing trend was largely determined by the Atlantic sector (where most of the Arctic MIZ perimeter resided) and resulted from poleward contraction of the MIZ equatorward edge (c=0.15) with comparatively little movement of the MIZ poleward edge (c=0.80). As detailed in [Strong, 2012], the atmosphere lacks sufficient circulation or wind stress trends to account for the cold season narrowing, suggesting an important role for thermodynamic forcing. Oceanic heat content is a principal regulator of the sea ice edge position [e.g., Bitz et al., 2005], and Atlantic sector warming over the satellite era (related in part to the upward trending Atlantic Multidecadal Oscillation [e.g., Deser et al., 2009]) is consistent with a poleward contraction of the MIZ equatorward edge. The remainder of this section focuses on warm season results.

[9] Warm season MIZ width (math formula; orange curve, Figure 2a) averaged 124 km over 1979–2011, and the upward trend line (13 km decade−1; Figure 2a) indicates a 39% widening over the record. Early in the record, large regions of pack ice covered much of the Arctic basin in August, confining the MIZ to a narrow band around its periphery (e.g., Figure 2b). Later in the record, pack ice contracted poleward away from the coasts of Alaska and eastern Siberia, leaving behind a broad MIZ (e.g., Figure 2c). Comparison of Figures 2b and 2c illustrates that the warm season MIZ widening trend was the result of the MIZ's poleward edge advancing markedly poleward with comparatively small contraction of the MIZ's equatorward edge. Full domain trends in MIZ latitude and width were driven largely by changes in the East Siberian-North American (ES-NA) sector, where the warm season MIZ widened by 22.5 km decade−1 while moving poleward by more than 0.8° of latitude per decade (math formula and math formula; Table 1). In the Atlantic sector, poleward motion of the MIZ was approximately 30% slower (0.62° decade−1), and MIZ width exhibited a slight narrowing trend (Table 1).

Table 1. Trend in Sea Ice Variables for 1979–2011 in Regions Defined in Figure 1aa
 East Siberian- Full
 North AmericanAtlanticDomain
  1. a

    All values are statistically significant at the 95% confidence level.

July–September MIZ width (math formula; km decade−1)23−413
July–September MIZ latitude (math formula; ° decade−1)0.840.620.59
April sea ice thickness (math formula; m decade−1)−0.26−0.11−0.15
April sea ice age (math formula; years decade−1)−4.0−1.7−2.8

[10] An increase in the portion of the ice cover that is young and thin at the onset of melt could lead to an increase in the spatial coverage of marginal ice (0.15≤c≤0.80), and hence wider MIZs. During the first half of the record, the probability of July–September MIZ (PMIZ) was highest around the periphery of the Arctic basin (red shading, Figure 3a) where mean sea ice thickness at the beginning of the melt season ranged from approximately 2 to 3 m (Figure 3c shows mean April sea ice thickness (math formula) for the first half of the record with stippling where PMIZ≥0.25). The high PMIZ periphery during the first half of the record (Figure 3a) also coincided spatially with regions where mean sea ice age was generally less than 10 years (Figure 3e shows mean April sea ice age (math formula) for the first half of the record with stippling where PMIZ≥0.25).

Figure 3.

(Top) The first half of the record (1979–1995) and (bottom) the second half of the record (1996–2011). Mapped variables are (a, b) probability of July–September MIZ (PMIZ), (c, d) April mean sea ice thickness (math formula; meters), and (e, f) April mean sea ice age (math formula; years). Red meridians separate the ES-NA and Atlantic sectors defined in Figure 1a, and stippling is shown on Figures 3c–3f where PMIZ≥0.25.

[11] Comparing the first and second halves of the record, the warm season MIZ widened poleward, particularly in the ES-NA sector (Figure 3b). The poleward MIZ widening was collocated with sea ice thinning that produced math formula in the 2 to 3 m range over much of the ES-NA sector (Figure 3d shows math formula for the second half of the record with stippling where PMIZ≥0.25). Likewise, sea ice age decreased over much of the ES-NA sector, expanding the young ice regions that were favored for MIZ formation (Figure 3f shows math formula for the second half of the record with stippling where PMIZ≥0.25). Similar to the MIZ width results, full domain trends in sea ice age and thickness were largely driven by changes in the ES-NA sector (Table 1).

4 Sensitivity Analysis and Discussion

[12] The warm season MIZ widening in the ES-NA sector was collocated with other important changes in the Arctic sea ice system that could affect passive microwave retrievals of sea ice concentration, including emergence of exceptional melt pond coverage [Rosel and Kaleschke, 2012] and ice losses associated with record low extents [Parkinson and Comiso, 2013]. Sea ice concentration variability between the Bootstrap and NASA Team algorithms tended to be larger within the MIZ (Figure S2 in supporting information), and passive microwave retrievals of sea ice concentration may be biased low, particularly during summer when melt ponds may cover large fractions of first-year ice [Meier and Notz, 2010]. To test the sensitivity of our reported warm season widening trend to sea ice concentration measurement bias, we first assumed a concentration measurement error of −20% distributed uniformly over space and time (i.e., true concentration underreported by 20%) and then replaced the Bootstrap concentrations (c) with the lesser of {1.0;c/0.8}. The adjusted concentrations yielded a smaller mean warm season MIZ width (90 km versus 124 km) and a smaller but still statistically significant (α=0.05) warm season trend over the record (7.9 km decade−1 versus 13 km decade−1). The widening trend remained statistically significant when measurement errors of −10% and −30% were assumed (Figure S3 in supporting information). The widening trend also remained statistically significant when measurement error adjustments were applied only at lower concentrations (c<0.5) or only at higher concentrations (c≥0.5) (not shown). Finally, the widening trend vanished when measurement errors were assumed to strengthen linearly over the record from −20% to −30% (Figure S3 in supporting information).

[13] MIZ width trends are associated with changes in sea ice concentration gradients (our MIZ is defined by concentration contours), and some insights into MIZ width variability are available from studies of sea ice seasonality, compactness, and deformation dynamics [e.g., Comiso and Zwally, 1984; Stern and Lindsay, 2009]. The equatorward edge of the MIZ coincides with sea ice extent by definition (both use the 0.15 concentration contour), and the dynamics governing MIZ width variability are consequently related to the suite of factors that control the sea ice edge [e.g., Bitz et al., 2005]. Comparatively little is known about controls on the MIZ poleward (high concentration) edge, and poleward advance of the MIZ poleward edge has far outpaced poleward contraction of the MIZ equatorward edge. Further understanding of the linkages between MIZ width and environmental controls will require more extensive observational and modeling analysis.


[14] C. Strong was supported by National Science Foundation Arctic Sciences Division grant 1022485. I. Rigor is supported by the US Interagency Arctic Buoy Program. The authors thank the editor and two anonymous reviewers for insightful comments.

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