1. Introduction
A plumelike structure of sea ice often appears in the marginal ice zone (MIZ) in polar oceans (Fig. 1). This organized structure of sea ice is called an ice band. Formation of ice bands causes increases in the area of open water in MIZ, and rapid melting may be promoted. Therefore, it has been suggested that the seasonal evolution of the MIZ is likely influenced by ice-band pattern formation (e.g., Martin et al. 1983).
Satellite images of ice-band patterns from Moderate Resolution Imaging Spectroradiometer (MODIS) images by the National Aeronautics and Space Administration (NASA; available online at http://lance-modis.eosdis.nasa.gov/imagery/subsets/?mosaic=Arctic). (a) Greenland Sea (21 Mar 2010) and (b) Southern Ocean (12 Dec 2010); 10-m wind vectors (m s−1) of ERA-Interim (Dee et al. 2011) on 20 Mar 2010 and 11 Dec 2010, respectively. Wind direction of (a) coincides with the cloud streets direction.
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Satellite images of ice-band patterns from Moderate Resolution Imaging Spectroradiometer (MODIS) images by the National Aeronautics and Space Administration (NASA; available online at http://lance-modis.eosdis.nasa.gov/imagery/subsets/?mosaic=Arctic). (a) Greenland Sea (21 Mar 2010) and (b) Southern Ocean (12 Dec 2010); 10-m wind vectors (m s−1) of ERA-Interim (Dee et al. 2011) on 20 Mar 2010 and 11 Dec 2010, respectively. Wind direction of (a) coincides with the cloud streets direction.
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Satellite images of ice-band patterns from Moderate Resolution Imaging Spectroradiometer (MODIS) images by the National Aeronautics and Space Administration (NASA; available online at http://lance-modis.eosdis.nasa.gov/imagery/subsets/?mosaic=Arctic). (a) Greenland Sea (21 Mar 2010) and (b) Southern Ocean (12 Dec 2010); 10-m wind vectors (m s−1) of ERA-Interim (Dee et al. 2011) on 20 Mar 2010 and 11 Dec 2010, respectively. Wind direction of (a) coincides with the cloud streets direction.
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Ice bands show a variety of band spacing from O(1) km to O(10) km, where the distance from the center of one ice band to the adjacent one is defined as the band spacing. In this paper, we will discuss ice bands with relatively large band spacing (about 10 km), which have been observed from satellite, as shown in Fig. 1. Muench and Charnell (1977) presented basic characteristics of such large-scale ice bands, where the band spacing ranges between 6 and 12 km and the spacing becomes larger as the wind speed increases. The band pattern often spreads over the MIZ extending more than 100 km. Further, the long axis of an ice band tends to lie to the left (right) of the wind direction at 40°–90° in the Northern (Southern) Hemisphere (Muench and Charnell 1977; Wadhams 1983, 2000; Ishida and Ohshima 2009), where the long axis is defined as the band orientation that is perpendicular to the band spacing (see Fig. 2). The satellite images in Fig. 1 also exhibit this feature.
Momentum balance on the sea ice.
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Momentum balance on the sea ice.
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Momentum balance on the sea ice.
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Mechanisms of ice-band formation have been discussed since the 1980s. McPhee (1979, 1982, 1983) suggested that an ice edge band may be separated from the top of the ice pack by the reduction of drag between sea ice and ocean owing to intense buoyancy flux by melting when the ice edge moves into warm water. Wadhams (1983) suggested that inhomogeneity of the wave radiation stress in the fetch-limited open waters of various spacing would produce ice bands. The wave radiation stress is concentrated on floes at the downwind end of the open water, and it accelerates them toward the neighboring floes farther downwind. These cumulated ice floes may form the band structure. Wadhams (1983) applied this theory to ice bands with a band spacing of ~1 km, although this mechanism may work for larger-scale ice bands.
Mechanisms of relatively large-scale ice-band formation with a band spacing of about 10 km, which is our main subject, have been discussed from a point of view of a coupling with internal inertial–gravity waves. A steady wind may generate ice bands by generating internal lee waves behind an ice edge, where the vertical velocity is caused by the convergence/divergence of the Ekman transport as a result of differential stress between the open water and ice-covered water (Sjørberg and Mork 1985; Fujisaki and Oey 2011, hereinafter referred to as FO11). The internal waves are generated when the ice edge speed U is larger than the baroclinic phase speed cI; here, cI = 
In this paper, based on the internal inertia–gravity wave model, we will discuss basic ice-band characteristics such as
what determines the ice-band spacing; and
is there any favorable wind direction for ice-band formation?
This paper is organized as follows: In section 2, an ice–ocean coupled model is formulated using a hydrostatic, 1.5-layer model. In section 3, the above two questions are discussed by obtaining a plane wave solution in the vicinity of resonance. We present the discussion in section 4, where ice-band observations from satellites, nonhydrostatic effects, and effects of earth rotation are discussed. Further considerations on sea ice convergence by variations in ice concentration and by wind forcings are also included. Finally, we summarize the results in section 5.
2. Formulation
a. Sea ice model
b. Ocean model
Next, we considered a 1.5-layer ocean model driven by the surface stress. The 1.5-layer model is a layer model with the hydrostatic approximation, in which the lower layer is assumed to be so deep that the lower-layer motion is at rest. Since the polar oceans typically have a well-mixed upper layer underlain by a strong pycnocline, the 1.5-layer model would be suited for modeling the upper polar oceans (e.g., Muench 1983).
Ice–ocean coupled model consisting of 1.5 layers.
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Ice–ocean coupled model consisting of 1.5 layers.
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Ice–ocean coupled model consisting of 1.5 layers.
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Schematics of vertical flow generation: (a) Off-ice oceanic velocity perturbations and associated vertical velocity due to periodic ice–ocean stress divergence caused by off-ice wind. Mean ice–ocean stress causes along-ice Ekman transport that does not contribute to the divergence/convergence of the flow in the ocean. (b) Vertical velocity generation due to stress difference Δτ associated with the Ekman transport by the along-ice wind. (c) Vertical velocity generation due to the turning angle δθ associated with the Ekman transport by the along-ice wind. See appendix A for details.
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Schematics of vertical flow generation: (a) Off-ice oceanic velocity perturbations and associated vertical velocity due to periodic ice–ocean stress divergence caused by off-ice wind. Mean ice–ocean stress causes along-ice Ekman transport that does not contribute to the divergence/convergence of the flow in the ocean. (b) Vertical velocity generation due to stress difference Δτ associated with the Ekman transport by the along-ice wind. (c) Vertical velocity generation due to the turning angle δθ associated with the Ekman transport by the along-ice wind. See appendix A for details.
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Schematics of vertical flow generation: (a) Off-ice oceanic velocity perturbations and associated vertical velocity due to periodic ice–ocean stress divergence caused by off-ice wind. Mean ice–ocean stress causes along-ice Ekman transport that does not contribute to the divergence/convergence of the flow in the ocean. (b) Vertical velocity generation due to stress difference Δτ associated with the Ekman transport by the along-ice wind. (c) Vertical velocity generation due to the turning angle δθ associated with the Ekman transport by the along-ice wind. See appendix A for details.
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Consequently, the two equations with respect to ζ′ and A′ are derived as in (14) and (15). In the next section, an eigenvalue problem will be discussed using these equations.
3. Theory and results
a. Scaling
b. Plane wave solution
Dispersion relation of the internal inertia–gravity wave 
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Dispersion relation of the internal inertia–gravity wave
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Dispersion relation of the internal inertia–gravity wave
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
c. Perturbation expansion in the vicinity of the resonance condition
1) O(ϵ0) solution
Relationship between the band propagation speed 
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Relationship between the band propagation speed
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Relationship between the band propagation speed
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Equation (23) indicates that if the sea ice drift speed becomes large, the band scale tends to become large. This is consistent with results by Muench and Charnell (1977). Once 
Figure 7 displays nondimensional band spacings in terms of the nondimensional sea ice drift speed. The dashed line in this figure represents the nondimensional inertial frequency 
Relationship between the band propagation speed 
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Relationship between the band propagation speed
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Relationship between the band propagation speed
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
2) O(ϵ1/2) solution
Now, we will discuss the question of whether or not there are any favorable wind directions for ice-band formation. If δθ = 0, both G and sin(ϕ/2) in ν are maximum when θa = 0 because ω0 > 1 for the internal inertia–gravity waves. In other words, the growth rate is maximum when the band propagates in the direction that is parallel to the wind or when the long axis of the band lies in the right angle with respect to the wind. In this case, the convergence/divergence in the ocean transport is not owing to the Ekman transport but owing to the stress divergence term, that is, ∂/∂t∇ ⋅ τ in (10). This term is not necessarily small relative to the Ekman term because ω0 > 1. Figure 8 depicts the relationship between the ice band and the internal wave when mutual enhancement occurs. An important aspect in Fig. 8 is that the stress divergence term causes upwelling where ice concentration is small. The upwelling amplifies the internal waves that propagate along with the ice band. Then, the velocity field associated with the internal wave in turn aggregates ice and enhances the ice-band formation as in Fig. 8. This is similar to a kinematic explanation of hydrodynamic instability such as the Kelvin–Helmholtz instability (e.g., Baines and Mitsudera 1994). Note that the convergence/divergence in the Ekman transport, caused by the y-component wind, produces downwelling in the front side of the ice band. This downwelling does not induce mutual enhancement (see appendix B for more discussion), although it produces a lee-wave train behind the ice edge as discussed in Sjørberg and Mork (1985) and FO11.
Schematic diagram of the coupling between fluctuations of internal waves and the evolution of sea ice concentration. Broad arrows represent the flow field induced by the internal wave. Broken arrows denote the vertical flow generated by stress divergence, which is the same as Fig. 4a. Since the broken arrows lag behind the broad arrows by ~π/4, ridges and troughs of the internal waves are forced to grow owing to the vertical flow generated by the stress divergence. Further, since downwelling motions induced by internal waves are located below ice bands where ice concentration is high, the horizontal flow is convergent there, and the ice concentration increases further. Therefore, this coupling yields a positive feedback.
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Schematic diagram of the coupling between fluctuations of internal waves and the evolution of sea ice concentration. Broad arrows represent the flow field induced by the internal wave. Broken arrows denote the vertical flow generated by stress divergence, which is the same as Fig. 4a. Since the broken arrows lag behind the broad arrows by ~π/4, ridges and troughs of the internal waves are forced to grow owing to the vertical flow generated by the stress divergence. Further, since downwelling motions induced by internal waves are located below ice bands where ice concentration is high, the horizontal flow is convergent there, and the ice concentration increases further. Therefore, this coupling yields a positive feedback.
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Schematic diagram of the coupling between fluctuations of internal waves and the evolution of sea ice concentration. Broad arrows represent the flow field induced by the internal wave. Broken arrows denote the vertical flow generated by stress divergence, which is the same as Fig. 4a. Since the broken arrows lag behind the broad arrows by ~π/4, ridges and troughs of the internal waves are forced to grow owing to the vertical flow generated by the stress divergence. Further, since downwelling motions induced by internal waves are located below ice bands where ice concentration is high, the horizontal flow is convergent there, and the ice concentration increases further. Therefore, this coupling yields a positive feedback.
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Finally, the growth rate ν in (26) is plotted in Fig. 9 in terms of wind directions θa, where the parameter values are (ω0, k0) = (1.516, 8.33 × 10−1), |τai| = 1.0, and δd = Δτ/|τai| = 0.3. As for δθ = 0, ν becomes maximum when the ice band travels parallel to the wind direction. Nonzero but small δθ modifies θa in a consistent manner to the observation.
Relationship between the wind direction and nondimensional growth rate. Solid line denotes δθ = 0.1 case, dashed line denotes δθ = 0.0 case, and dotted line denotes δθ = 0.2 case. Schematic diagrams depict wind direction.
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Relationship between the wind direction and nondimensional growth rate. Solid line denotes δθ = 0.1 case, dashed line denotes δθ = 0.0 case, and dotted line denotes δθ = 0.2 case. Schematic diagrams depict wind direction.
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Relationship between the wind direction and nondimensional growth rate. Solid line denotes δθ = 0.1 case, dashed line denotes δθ = 0.0 case, and dotted line denotes δθ = 0.2 case. Schematic diagrams depict wind direction.
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
d. Numerical results of coupled development
In this subsection, we try to verify the ice-band pattern formation using the ice–ocean coupled numerical model with simple settings. The present theory implies that if there is a random sea ice concentration field initially, the band patterns may emerge gradually, having the most favorable band-axis direction with respect to the imposed wind.
We use an ice–ocean coupled model that is the same as that of FO11. The ocean model is based on the Princeton Ocean Model, which employs the primitive equations with hydrostatic as well as Boussinesq approximations (Mellor et al. 2002). The ice model uses the elastic–viscous–plastic rheology (Hunke and Dukowicz 1997) with ice collision parameterization (Sagawa 2007). The model domain is 160 km × 220 km, with a horizontal resolution of 250 m. The depth of the ocean is set to be 150 m, having the Bering Sea shelf break in mind, with a vertical resolution of 5 m. Figure 10 shows the initial density stratification, which approximates a two-layer model ocean. The stratification after 7 days is also depicted in Fig. 10, which indicates that the upper mixed layer further develops during the evolution of ice bands. This configuration is dynamically equivalent to the 1.5-layer model for the first baroclinic mode. In this experiment, the water temperature is set at −1.8°C, and all thermal fluxes are removed. That is, thermodynamic effects are eliminated because we want to investigate whether the ice bands are formed by dynamical effects only. The initial sea ice field is set with a random sea ice concentration (Fig. 11a). Mean sea ice concentration is 0.5 because we consider a free drift condition without internal stress. The ice thickness d = 0.5 m. The lateral boundary condition of this domain is periodic.
Stratification in the numerical experiments: Dashed line denotes the initial stratification, where a potential density profile is given by 0.15 × tanh[0.03(z − 50)] + 1026.3 (kg m−3), with z being the depth. Solid line denotes the stratification on day 7 after the wind starts blowing. Ocean temperature is constant and set at −1.8°C so that melting does not occur.
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Stratification in the numerical experiments: Dashed line denotes the initial stratification, where a potential density profile is given by 0.15 × tanh[0.03(z − 50)] + 1026.3 (kg m−3), with z being the depth. Solid line denotes the stratification on day 7 after the wind starts blowing. Ocean temperature is constant and set at −1.8°C so that melting does not occur.
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Stratification in the numerical experiments: Dashed line denotes the initial stratification, where a potential density profile is given by 0.15 × tanh[0.03(z − 50)] + 1026.3 (kg m−3), with z being the depth. Solid line denotes the stratification on day 7 after the wind starts blowing. Ocean temperature is constant and set at −1.8°C so that melting does not occur.
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Ice-band pattern formation: (a) Initial random sea ice concentration field on horizontal plane, where mean sea ice concentration is set at 0.5. Red vector denotes the wind vector. Wind velocity is Ua = (Ua, Va) = (15.0, 7.5) m s−1. (b) Ice-band patterns on day 3.5 on horizontal plane. (c) Ice-band patterns on day 7 on horizontal plane. Black line denotes the location representing the vertical sections of Figs. 12a and 12b. Red vector denotes τai and blue vector denotes τiw, where |τai| = 10.5 × 10−1 kg m−1 s−2 and |τiw| = 9.63 × 10−1 kg m−1 s−2, respectively.
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Ice-band pattern formation: (a) Initial random sea ice concentration field on horizontal plane, where mean sea ice concentration is set at 0.5. Red vector denotes the wind vector. Wind velocity is Ua = (Ua, Va) = (15.0, 7.5) m s−1. (b) Ice-band patterns on day 3.5 on horizontal plane. (c) Ice-band patterns on day 7 on horizontal plane. Black line denotes the location representing the vertical sections of Figs. 12a and 12b. Red vector denotes τai and blue vector denotes τiw, where |τai| = 10.5 × 10−1 kg m−1 s−2 and |τiw| = 9.63 × 10−1 kg m−1 s−2, respectively.
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Ice-band pattern formation: (a) Initial random sea ice concentration field on horizontal plane, where mean sea ice concentration is set at 0.5. Red vector denotes the wind vector. Wind velocity is Ua = (Ua, Va) = (15.0, 7.5) m s−1. (b) Ice-band patterns on day 3.5 on horizontal plane. (c) Ice-band patterns on day 7 on horizontal plane. Black line denotes the location representing the vertical sections of Figs. 12a and 12b. Red vector denotes τai and blue vector denotes τiw, where |τai| = 10.5 × 10−1 kg m−1 s−2 and |τiw| = 9.63 × 10−1 kg m−1 s−2, respectively.
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
The homogeneous wind starts blowing over the initial sea ice field. As a result, the ice-band patterns emerge all over the domain as in Fig. 11b. The ice bands in this simulation exhibit a regular, 10-km-scale band spacing (Fig. 11c), consistent with those of ice bands seen in the satellite images (e.g., Figs. 1, 13). Further, wind leans to the left with respect to the band propagation direction.
Figure 12b shows the vertical section of the vertical velocity along the black line in Fig. 11c. Mean amplitude of vertical velocity in a unit volume (1 m3) grows exponentially (Fig. 12c), indicating mutual intensification between the ice bands and the internal waves. After 120 h, the growth of the vertical velocity slows down a little, indicating that the ice band grows sufficiently so that ice concentration is close to one in the ice band. Sea ice concentration on day 7 along this section is also drawn in Fig. 12a. Comparing Fig. 12a with Fig. 12b, we found that the phase relationship between sea ice and internal waves is the same as that in the theoretical model (Fig. 8), in such a way that the upwelling vertical velocity is seen in front of each ice-band promoting convergence in ice motion. Therefore, the numerical results in Figs. 11 and 12 represent the ice-band pattern formation by resonant interaction between sea ice and internal waves. This implies the importance of the stress divergence term in (10) for the mutual enhancement. As for the role of δθ, Fig. 11c shows that δθ ~ 2°. This yields θa ≃ 3.5° for δd = cosδθ − (CDaw/CDai) ≃ 0.5, where the air–water drag coefficient CDaw = 1.5 × 10−3 and the air–ice drag coefficient CDai = 3.0 × 10−3 in the model. This θa is too small to explain the simulated angle fully, although the turning direction is consistent.
Development of vertical flows: (a) Sea ice concentration along the black line of Fig. 11c. (b) Vertical section of vertical flow along the black line of Fig. 11c. (c) Evolution of the vertical velocity amplitude squared with time in a unit volume (1 m3).
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Development of vertical flows: (a) Sea ice concentration along the black line of Fig. 11c. (b) Vertical section of vertical flow along the black line of Fig. 11c. (c) Evolution of the vertical velocity amplitude squared with time in a unit volume (1 m3).
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Development of vertical flows: (a) Sea ice concentration along the black line of Fig. 11c. (b) Vertical section of vertical flow along the black line of Fig. 11c. (c) Evolution of the vertical velocity amplitude squared with time in a unit volume (1 m3).
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
The numerical results are apparently different from those of FO11 although the numerical model used is the same. In particular, FO11 suggested that an ice band and a lee-wave train hardly occur for the wind with θa = 0. This difference probably comes from the initial configuration of the sea ice. In FO11 they configured the ice-band modeling with a Heaviside function–type initial ice concentration, where A = 0 in front of the ice edge, and A = 0.8 behind the ice edge, in contrast to our case in which the mean of A is 0.5 in order for sea ice to drift freely at least initially. Because A = 0.8 in FO11 is large enough for internal stress to become effective, sea ice does not react directly to the sinusoidal flow field of internal waves but acts as a low-pass filter. As a result, the vertical velocity caused by the stress divergence becomes much smaller than that of our case. Indeed, a similar experiment to FO11 with the initial ice edge of the Heaviside function form, but for A = 0.5 behind the ice edge, shows the band formation for θa = 0 (figure not shown).
e. Brief summary
Based on the theoretical and numerical modeling results, we have found a new unstable mode caused by resonant interaction between the internal inertia–gravity waves and sea ice drift. Results are summarized briefly as follows:
The ice-band scale is selected by the resonance condition in which the ice-band speed coincides with the phase speed of internal inertia–gravity waves.
The ice bands tend to develop favorably when the wind direction and the band propagation direction are nearly parallel (or the long axis of the band is almost perpendicular to the wind). The velocity acceleration caused by the wind parallel to the band propagation direction is important. The wind direction may turn to the left slightly in the Northern Hemisphere for the maximum growth rate, as seen in the numerical results. The Coriolis force acting on ice may yield the angle, consistent with the numerical results, although it is small compared with the simulated angle.
4. Discussion
a. Ice-band pattern formation in real ocean
The discussion so far implies that once instability occurs as a result of the positive feedback between the ice band and internal waves, the ice-band pattern emerges from the random initial ice concentration field. In this subsection, we present an example of the ice-band pattern formation seen in satellite images in which ice bands emerges over a broad area.
Figure 13a shows a satellite image of the Sea of Okhotsk off Sakhalin Island on 26 March 2003. Although there were ice packs of various scales over the sea ice area, there were no major ice-band patterns on this day. Then, a low pressure system passed through this area from 27 to 30 March (Fig. 13b). After that the ice-band patterns were clearly seen (on 2 April) in Fig. 13c, with a spread of over 300 km. The wind vectors from 1 day earlier (on 1 April) are superimposed on the ice-band pattern because the ice-band development should have occurred during the passage of the low pressure system. This implies that the ice-band pattern emerged within a week. Instability modes due to resonance are likely responsible as they are capable of generating such a large-scale ice-band pattern that emerged within a short period of time.
Satellite images of the ice bands off Sakhalin in the Sea of Okhotsk from Advanced Very High Resolution Radiometer (AVHRR) were retrieved by the Kitami Institute of Technology. (a) Distribution of sea ice on 26 Mar 2003. (b) Distribution of sea ice and clouds on 29 Mar 2003. The 10-m wind vectors of ERA-Interim (Dee et al. 2011) are superimposed. (c) Distribution of sea ice on 1 Apr 2003. The 10-m wind vectors of ERA-Interim (Dee et al. 2011) on 31 Mar are superimposed. (d) Sea ice drift vectors on 1 Apr 2003 (courtesy of Dr. N. Kimura). Unit of the wind vector in (b) and (c) is m s−1.
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Satellite images of the ice bands off Sakhalin in the Sea of Okhotsk from Advanced Very High Resolution Radiometer (AVHRR) were retrieved by the Kitami Institute of Technology. (a) Distribution of sea ice on 26 Mar 2003. (b) Distribution of sea ice and clouds on 29 Mar 2003. The 10-m wind vectors of ERA-Interim (Dee et al. 2011) are superimposed. (c) Distribution of sea ice on 1 Apr 2003. The 10-m wind vectors of ERA-Interim (Dee et al. 2011) on 31 Mar are superimposed. (d) Sea ice drift vectors on 1 Apr 2003 (courtesy of Dr. N. Kimura). Unit of the wind vector in (b) and (c) is m s−1.
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Satellite images of the ice bands off Sakhalin in the Sea of Okhotsk from Advanced Very High Resolution Radiometer (AVHRR) were retrieved by the Kitami Institute of Technology. (a) Distribution of sea ice on 26 Mar 2003. (b) Distribution of sea ice and clouds on 29 Mar 2003. The 10-m wind vectors of ERA-Interim (Dee et al. 2011) are superimposed. (c) Distribution of sea ice on 1 Apr 2003. The 10-m wind vectors of ERA-Interim (Dee et al. 2011) on 31 Mar are superimposed. (d) Sea ice drift vectors on 1 Apr 2003 (courtesy of Dr. N. Kimura). Unit of the wind vector in (b) and (c) is m s−1.
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
We also examine the relation between the band spacing and the wind speed over the Bering Sea shelf discussed by Muench and Charnell (1977, see their Fig. 4). Since the higher vertical-mode waves for continuous stratification may develop over the shallow Bering Sea shelf, cI of a 1.5-layer model in (28) is replaced by eigenvalues of higher vertical modes, which are slower than cI (e.g., Sjørberg and Mork 1985; FO11). The dashed line in Fig. 14 denotes the relationship λmax = 2π
Relationship between wind speed and band spacing in the Bering Sea (redrawn from Muench and Charnell 1977). Dots denote ice bands observed from satellites. Numbers 1, 2, 3, 4, and 5 in the figure denote the five gravest modes. Dashed line denotes the 
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Relationship between wind speed and band spacing in the Bering Sea (redrawn from Muench and Charnell 1977). Dots denote ice bands observed from satellites. Numbers 1, 2, 3, 4, and 5 in the figure denote the five gravest modes. Dashed line denotes the
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Relationship between wind speed and band spacing in the Bering Sea (redrawn from Muench and Charnell 1977). Dots denote ice bands observed from satellites. Numbers 1, 2, 3, 4, and 5 in the figure denote the five gravest modes. Dashed line denotes the
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
The resonance condition may be modified by the presence of the background flow. If the flow is purely barotropic, then the resonance condition is basically unchanged because both internal waves and ice bands are advected by the uniform flow. If the flow is baroclinic, however, the internal wave phase speed may be modified by a vertically sheared flow, and hence the resonance condition can be affected. Coupling between ice bands and internal waves in a vertically sheared flow is an important subject for the quantitative applicability of this theory to reality.
b. Nonhydrostatic effects
The present theory is based on the hydrostatic approximation. Therefore, the hydrostatic approximation is likely to be valid for this band scale. Then, the next question is what is expected if the hydrostatic approximation is relaxed so that nonhydrostatic effects are involved?
The dispersion relation is displayed in Fig. 15 by using N = 30f, which is likely a typical value of N in polar seas; we estimated N ≃ 60f in the Greenland Sea from McPhee et al. (1987) and N ≃ 20f–30f in the Sea of Okhotsk from Fukamachi et al. (2004), Mizuta et al. (2004), and Ohshima et al. (2001), whereas N ≃ 15f in the Southern Ocean from Rintoul and Bullister (1999) and Wong and Riser (2011). The line ω = (N/kz)kx, which is a long-wave limit (kx → 0) of the internal gravity waves on a nonrotating frame (i.e., f = 0), is also plotted in Fig. 15. The dispersion curve of the nonhydrostatic internal inertial–gravity waves (Fig. 15) has an inflection point because ω → f as kx/kz → 0, while ω → N as kx/kz → ∞. Therefore, this dispersion curve gives an additional intersection (or a resonance point) with the line 
Solid line denotes the dispersion relationship of the internal inertia–gravity wave without adopting the hydrostatic approximation, where N = 50f and kz = 0.01 m−1 in (28). The vertical axis is normalized by f, and the horizontal axis is normalized by L0 (=1000 m). The dashed line represents baroclinic phase speed with a nondispersive limit.
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Solid line denotes the dispersion relationship of the internal inertia–gravity wave without adopting the hydrostatic approximation, where N = 50f and kz = 0.01 m−1 in (28). The vertical axis is normalized by f, and the horizontal axis is normalized by L0 (=1000 m). The dashed line represents baroclinic phase speed with a nondispersive limit.
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Solid line denotes the dispersion relationship of the internal inertia–gravity wave without adopting the hydrostatic approximation, where N = 50f and kz = 0.01 m−1 in (28). The vertical axis is normalized by f, and the horizontal axis is normalized by L0 (=1000 m). The dashed line represents baroclinic phase speed with a nondispersive limit.
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Based on the above argument, the ice-band formation could appear at the high wavenumber. For example, a number of 2–3-km ice bands were observed in the Southern Ocean (Ishida and Ohshima 2009), although these narrow bands could be generated by various other factors such as melting processes in the boundary layer below ice (McPhee 1983; FO11) and wave radiation stress (Wadhams 1983).
c. Nonrotational system
d. Considerations on variability in the ice drift speed depending on ice concentration and wind stress variations
Schematics of an ice-band evolution when the ice drift speed depends on the ice concentration.
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Schematics of an ice-band evolution when the ice drift speed depends on the ice concentration.
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Schematics of an ice-band evolution when the ice drift speed depends on the ice concentration.
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
We also assumed that the wind is uniform in time and space in order to extract the self-organized band pattern emerging from a homogeneous background field. In reality, the wind has various structures associated with the atmospheric boundary layer. The perturbation in the wind stress can force the organization of sea ice patterns. For example, streaky surface winds with roll-like cells, such as those in Fig. 1 visualized by clouds, may cause an ice-band pattern by the convergence/divergence of the Ekman drift of sea ice. In this case, however, the long axis of the band should be parallel along the downwind direction (e.g., Muench and Charnell 1977). Apparently, effects of direct wind forcing on ice can be seen in a numerical result in Fig. 11b (on day 3.5), showing streaks that are parallel to the wind. Interaction between the atmospheric boundary layer and the ice bands requires further study.
5. Conclusions
In this paper, we discussed the formation mechanisms of ice-band patterns from a point of view of the resonant interaction using an ice–ocean coupled system. The following two main questions have been discussed: 1) What determines the ice-band scale? 2) Is there a favorable wind direction with respect to ice-band formation? To answer these questions, we formulated a 1.5-layer ocean model coupled with a sea ice model.
We have found that resonant interaction between internal inertia–gravity waves and sea ice pattern propagation well explains the formation of the ice bands with 10-km-scale band spacing. In this coupled system, the internal waves are generated as a result of the difference between the stress on the open ocean and the stress on the ice-covered ocean as well as the turning angle between these two stresses. The generated internal waves in turn affect the evolution of the sea ice concentration so that the ice band and internal waves are mutually interacting with each other. Assuming a plane wave solution, we obtained a characteristic equation that describes instability representing a positive feedback between the ice-band formation and internal wave growth. The coupling strength ϵ is characterized by a ratio between the surface layer thickness and the typical amplitude of internal waves.
In this paper, we assumed that ϵ1/2 is a small parameter and we expanded the characteristic equation with respect to ϵ1/2. The O(ϵ0) solution represents both the phase speed of the internal waves and the ice drift speed in the direction of ice-band propagation. Resonance occurs when these speeds coincide with each other, and it is this condition that determines the band spacing of 10 km. The O(ϵ1/2) solution represents the growth rate as well as modification to the phase speed. We have found that the growth rate is maximum when the band propagation direction is nearly parallel to the wind. The wind for the maximum growth may turn to the left slightly in the Northern Hemisphere, owing to the Coriolis force acting on the ice.
In the present study, we did not deal with thermodynamics because we focused on the dynamical process of band pattern formation. However, thermodynamics should also be important for the formation and melting of the ice bands. For example, intense surface buoyancy flux by rapid melting may reduce the frictional coupling as ice drifts over relatively warm water so as to increase ice speed, and then separation of a band may result (McPhee 1979, 1982, 1983; FO11). Therefore, the thermodynamic process associated with ice-band formation should be considered in future studies.
In this study, we show a mechanism of how the ice-band pattern, whose spacing is regular, is organized even from a random ice field in the MIZ. Streaky winds with roll-like cells in the atmospheric boundary layer may also initiate generating perturbations in the MIZ. Once a regular pattern forms, inhomogeneous ice drift velocity that is dependent upon various ice conditions (Steele et al. 1989), as well as the radiation stress by wind waves (Wadhams 1983), may enhance the ice-band pattern further. In the future, we will evaluate these contributions to the evolution of the MIZ because the ice-band formation and melting is a ubiquitous phenomenon.
Acknowledgments
We thank Drs. A. Manome, T. Toyota, N. Kimura, J. Ukita, M. Leppäranta, T. Nakamura, T. Nakanowatari, and S. Aoki for comments and discussion. Support from Pan-Okhotsk Research Center members is greatly appreciated. We thank four reviewers for comments that helped to improve the manuscript. Numerical results were obtained with the Pan-Okhotsk Information System. Satellite images from the Kitami Institute of Technology and National Aeronautics and Space Administration (NASA) were used to analyze the data. Ice drift data were produced by Dr. N. Kimura, which is greatly appreciated. This work was supported by KAKENHI 22106010 and 26247074.
APPENDIX A
Generation of Vertical Motion Associated with δθ
Let us consider an up-ice-wind parallel to the band’s long-axis direction. The stress τaw generates the Ekman transport in the off-ice direction. On the other hand, τiw generates the Ekman transport that turns to the right by δθ from the off-ice direction. Therefore, the off-ice component of the Ekman transport becomes smaller by the factor of sin(δθ). This can cause divergence (convergence) in front of (behind) the ice band as in Fig. A1. This schematic diagram represents the effect of the fourth terms of the right-hand side of (14).
Schematic diagram of the generation of vertical motion by the turning angle between stress on the ice-covered water and stress on the open water. This schematic diagram represents the effect of the fourth term of the right-hand side of (10).
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Schematic diagram of the generation of vertical motion by the turning angle between stress on the ice-covered water and stress on the open water. This schematic diagram represents the effect of the fourth term of the right-hand side of (10).
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
Schematic diagram of the generation of vertical motion by the turning angle between stress on the ice-covered water and stress on the open water. This schematic diagram represents the effect of the fourth term of the right-hand side of (10).
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
APPENDIX B
Ice–Ocean Coupling in Case the Wind is Parallel to the Long-Axis Direction of an Ice Band
An along-ice wind in the positive y direction causes the Ekman transport in the positive x direction. Therefore, Δτ causes the downwelling (upwelling) in front of (behind) each ice band as in Fig. 4b. This is not a favorable phase relationship between the internal wave and the ice band for mutual enhancement. For example, if the internal waves are in a favorable location for the ice band to aggregate (see Fig. B1a), the downwelling (upwelling) in front of (behind) each ice band, generated by Δτ, tends to reduce the amplitude of the internal wave. Nevertheless, a neutral wave mode is possible in case the vertical velocity associated with the internal wave and that associated with the ice band are in phase (Fig. B1b). This implies that the growth rate is zero for the along-ice wind in the positive y direction if δθ = 0.
As in Fig. 8, but for the relation between the ice-band and the internal wave flow fields generated by the Ekman transport difference caused by Δτ. (a) Phase relationship when the ice band damps the internal wave. (b) Neutral mode when the vertical velocity due to the ice band and vertical velocity associated with the internal wave are in phase.
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
As in Fig. 8, but for the relation between the ice-band and the internal wave flow fields generated by the Ekman transport difference caused by Δτ. (a) Phase relationship when the ice band damps the internal wave. (b) Neutral mode when the vertical velocity due to the ice band and vertical velocity associated with the internal wave are in phase.
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
As in Fig. 8, but for the relation between the ice-band and the internal wave flow fields generated by the Ekman transport difference caused by Δτ. (a) Phase relationship when the ice band damps the internal wave. (b) Neutral mode when the vertical velocity due to the ice band and vertical velocity associated with the internal wave are in phase.
Citation: Journal of Physical Oceanography 46, 2; 10.1175/JPO-D-14-0162.1
REFERENCES
Baines, P. G., and H. Mitsudera, 1994: On the mechanism of shear flow instabilities. J. Fluid Mech., 276, 327–342, doi:10.1017/S0022112094002582.
Dee, D. P., and Coauthors, 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553–597, doi:10.1002/qj.828.
Fujisaki, A., and L. Y. Oey, 2011: Formation of ice bands by winds. J. Geophys. Res., 116, C10015, doi:10.1029/2010JC006655.
Fukamachi, Y., G. Mizuta, K. I. Ohshima, L. D. Talley, S. C. Riser, and M. Wakatsuchi, 2004: Transport and modification processes of dense shelf water revealed by long-term moorings off Sakhalin in the Sea of Okhotsk. J. Geophys. Res., 109, C09S10, doi:10.1029/2003JC001906.
Gill, A. E., 1982: Atmosphere–Ocean Dynamics. Academic Press, 662 pp.
Hakkinen, S., 1986: Ice banding as a response of the coupled ice-ocean system to temporally varying winds. J. Geophys. Res., 91, 5047–5053, doi:10.1029/JC091iC04p05047.
Hunke, E. C., and J. K. Dukowicz, 1997: An elastic–viscous–plastic model for sea ice dynamics. J. Phys. Oceanogr., 27, 1849–1867, doi:10.1175/1520-0485(1997)027<1849:AEVPMF>2.0.CO;2.
Ishida, K., and K. I. Ohshima, 2009: Ice-band characteristics of the Antarctic seasonal ice zone observed using MOS MESSR images. Atmos.–Ocean, 47, 169–183, doi:10.3137/OC300.2009.
Leppäranta, M., 2005: The Drift of Sea Ice. Springer, 350 pp.
Martin, S., P. Kauffman, and C. Parkinson, 1983: The movement and decay of ice edge bands in the winter Bering Sea. J. Geophys. Res., 88, 2803–2812, doi:10.1029/JC088iC05p02803.
McPhee, M. G., 1979: The effect of the oceanic boundary layer on the mean drift of pack ice: Application of a simple model. J. Phys. Oceanogr., 9, 388–400, doi:10.1175/1520-0485(1979)009<0388:TEOTOB>2.0.CO;2.
McPhee, M. G., 1982: Sea ice drag laws and simple boundary layer concepts, including application to rapid melting. U.S. Army Corps of Engineers CRREL Rep. 82-4, 17 pp.
McPhee, M. G., 1983: Turbulent heat and momentum transfer in the oceanic boundary layer under melting pack ice. J. Geophys. Res., 88, 2827–2835, doi:10.1029/JC088iC05p02827.
McPhee, M. G., G. A. Maykut, and J. H. Morison, 1987: Dynamics and thermodynamics of the ice/upper ocean system in the marginal ice zone of the Greenland Sea. J. Geophys. Res., 92, 7017–7031, doi:10.1029/JC092iC07p07017.
Mellor, G. L., M. G. McPhee, and J. D. Steels, 1986: Ice–seawater turbulent boundary layer interaction with melting or freezing. J. Phys. Oceanogr., 16, 1829–1846, doi:10.1175/1520-0485(1986)016<1829:ISTBLI>2.0.CO;2.
Mellor, G. L., S. Hakkinen, T. Ezer, and R. Patchen, 2002: A generalization of a sigma coordinate ocean model and an intercomparison of model vertical grids. Ocean Forecasting: Conceptual Basis and Applications, N. Pinardi and J. D. Woods, Eds., Springer, 55–72.
Mizuta, G., K. I. Ohshima, Y. Fukamachi, M. Itoh, and M. Wakatsuchi, 2004: Winter mixed layer and its yearly variability under sea ice in the southwestern part of the Sea of Okhotsk. Cont. Shelf Res., 24, 643–657, doi:10.1016/j.csr.2004.01.006.
Muench, R. D., 1983: Mesoscale oceanographic features associated with the central Bering Sea ice edge: February–March 1981. J. Geophys. Res., 88, 2715–2722, doi:10.1029/JC088iC05p02715.
Muench, R. D., and R. L. Charnell, 1977: Notes and correspondence observations of medium-scale features along the seasonal ice edge in Bering Sea. J. Phys. Oceanogr., 7, 602–606, doi:10.1175/1520-0485(1977)007<0602:OOMSFA>2.0.CO;2.
Muench, R. D., P. H. LeBlond, and L. E. Hachmeister, 1983: On some possible interactions between internal waves and sea ice in the marginal ice zone. J. Geophys. Res., 88, 2819–2826, doi:10.1029/JC088iC05p02819.
Nakayama, Y., K. I. Ohshima, and Y. Fukamachi, 2012: Enhancement of sea ice drift due to the dynamical interaction between sea ice and a coastal ocean. J. Phys. Oceanogr., 42, 179–192, doi:10.1175/JPO-D-11-018.1.
Ohshima, K. I., G. Mizuta, M. Itoh, Y. Fukamachi, T. Watanabe, Y. Nabae, K. Suehiro, and M. Wakatsuchi, 2001: Winter oceanographic conditions in the southwestern part of the Okhotsk Sea and their relation to sea ice. J. Oceanogr., 57, 451–460, doi:10.1023/A:1021225303621.
Rintoul, S. R., and J. L. Bullister, 1999: A late winter hydrographic section from Tasmania to Antarctica. Deep-Sea Res. I, 46, 1417–1454, doi:10.1016/S0967-0637(99)00013-8.
Sagawa, G., 2007: Development of ice dynamic model that takes account of floe collision and its validation in numerical sea ice forecast in the Sea of Okhotsk (in Japanese). Ph.D. dissertation, University of Tokyo, 104 pp.
Sjørberg, B., and M. Mork, 1985: Wind-induced stratified ocean response in the ice edge region: An analytical approach. J. Geophys. Res., 90, 7273–7285, doi:10.1029/JC090iC04p07273.
Steele, M., J. H. Morison, and N. Untersteiner, 1989: The partition of air-ice-ocean momentum exchange as function of ice concentration, floe size, and draft. J. Geophys. Res., 94, 12 739–12 750, doi:10.1029/JC094iC09p12739.
Wadhams, P., 1983: A mechanism for the formation of ice edge bands. J. Geophys. Res., 88, 2813–2818, doi:10.1029/JC088iC05p02813.
Wadhams, P., 2000: Ice in the Ocean. Gordon and Breach, 351 pp.
Wong, A. P. S., and S. C. Riser, 2011: Profiling float observations of upper ocean under sea ice off the Wilkes Land coast of Antarctica. J. Phys. Oceanogr., 41, 1102–1115, doi:10.1175/2011JPO4516.1.
