Hostname: page-component-7b7567c8c-2rsb6 Total loading time: 0 Render date: 2026-11-12T22:14:33.929Z Has data issue: false hasContentIssue false
  • English
  • Français

Synoptic-scale ice-motion case-studies using assimilated motion fields

Published online by Cambridge University Press:  14 September 2017

Walter N. Meier and
James A. Maslanik
Walter N. Meier
Affiliation:
National Ice Center, FOB-4, Room 2301, 4251 Suitland Road, Washington, DC20395, U.S.A.
James A. Maslanik
Affiliation:
Colorado Center for Astrodynamics Research, Aerospace Engineering Sciences Department, University of Colorado, Boulder, CO 80309−0431, U.S.A.
Rights & Permissions [Opens in a new window]

Abstract

Observed and modeled sea-ice motions, combined via an optimal-interpolation assimilation method, are used to study two synoptic events in the Arctic. The first is a convergence event along the north Alaska coast in the Beaufort Sea during November 1992. Assimilation indicates stronger convergence than the stand-alone model, in agreement with Advanced Very High Resolution Radiometer-derived ice motions and Special Sensor Microwave/Imager-derived ice concentrations. The second event pertains to ice formation and advection in Fram Strait and the Barents and Greenland, Iceland and Norwegian Seas. Assimilation indicates export of thick, less saline ice out of the central Arctic into the East Greenland Sea. However, the model indicates little flow through Fram Strait, instead showing strong flow of thin, more saline first-year ice from the Barents Sea westward into the Greenland Sea. These results indicate that assimilation is a useful tool for investigating synoptic events in the Arctic and may be useful for both climate studies and operational analyses

Information

Type
Sea-Ice Motion and Deformation
Information
Annals of Glaciology , Volume 33 , 2001, pp. 145 - 150
Copyright
Copyright © the Author(s) [year] 2001

Introduction

Ice motion is an important component in the evolution of the polar ice cover and its role in Arctic climate. Divergent motion opens leads where heat and moisture transfer between the ocean and atmosphere is facilitated. Convergent motion results in ridged ice that can be several meters thicker than surrounding ice, affecting the ice mass balance and impeding navigation. Patterns of ice circulation in the Arctic also result in a transport of relatively fresh water, which is important for the formation of ocean water masses.

Long-term mean ice motion depends primarily on the average ocean currents and the wind forcing (Reference Thorndike and ColonyThorndike and Colony, 1982). On short time-scales, wind is the primary forcing component of ice motion. Thus, ice motion exhibits a transient component where ice direction and speed can dramatically deviate from the mean motion due to synopticscale weather systems.

Because of its transient nature, ice motion is particularly relevant for operational ice analyses and forecasts in the Arctic. Knowledge of when and where ice will exist in a given locale is crucial to a variety of human activities, such as military maneuvers, shipping and fishing endeavors and petroleum operations. Application of data assimilation is well suited for these types of operational ice-forecasting situations because it constrains model forecasts with observed data.

Here we apply a new data-assimilation methodology, combining ice-motion observations from buoys and passive-microwave imagery with motion estimates from a sea-ice model to investigate two synoptic-scale events in the Arctic. The first is a convergence event in the Beaufort Sea, along the north coast of Alaska. The second deals with ice export out of the Barents Sea and Fram Strait.

Background

Three primary approaches have been used to describe sea-ice motion: in situ measurements, remote-sensing observations and model simulations. Each approach has yielded valuable insights into the polar regions. However, each approach has limitations and none can fully describe Arctic processes at all scales.

In situ observations, while accurate, are sparse and often cover only short time series. This particularly limits their use for describing detailed spatial patterns of motion that can result from synoptic-scale systems, where only one observation, or none, may be available to represent the entire affected region (Reference Kwok, Schweiger, Rothrock, Pang and KottmeierKwok and others, 1998).

Satellite remote sensing allows the entire basin to be described on at least a daily basis. Visible and infrared Advanced Very High Resolution Radiometer (AVHRR) imagery has been used to map ice motion (e.g. Reference Ninnis, Emery and CollinsNinnis and others, 1986; Reference Emery, Fowler, Hawkins and PrellerEmery and others, 1991), but is limited by clouds. Synthetic aperture radar (SAR) imagery provides high-resolution ice-motion fields under all sky conditions (e.g. Reference Kwok, Curlander, McConnell and PangKwok and others, 1990, Reference Kwok, Rothrock, Stern and Cunningham1995), but is limited by the swath width and orbit repeat cycle to 3−6 day intervals.

Passive-microwave imagery from the Special Sensor Microwave Imager (SSM/I) provides daily basin-wide coverage under all sky conditions (e.g. Reference Agnew, Le and HiroseAgnew and others, 1997; Reference Emery, Fowler and MaslanikEmery and others, 1997; Reference Kwok, Schweiger, Rothrock, Pang and KottmeierKwok and others, 1998; Reference Liu and CavalieriLiu and Cavalieri, 1998). However, SSM/I-derived ice motions suffer from low resolution (12.5 km) and problems with surface melt and atmospheric moisture during the polar summers.

Sea-ice models use dynamic and thermodynamic balances, and atmospheric and oceanic forcing to simulate ice motion (e.g Reference HiblerHibler, 1979). Though temporal resolution can be quite high (typically 4−6 h time-steps or finer), spatial resolution (a maximum of about 10 km in the interior ice pack for continuum-based models) has generally been poor. In addition, the model is constrained by the accuracy of the forcing and the assumptions of the model physics.

Data assimilation was first employed in the geosciences within atmospheric models (e.g. Reference Charney, Halem and JastrowCharney and others, 1969). It has proven successful at constraining models to physical reality and improving forecast effectiveness (Reference Ghil and Malanotte-RizzoliGhil and Malanotte-Rizzoli, 1991). More recently, data-assimilation methods have been applied to ocean modeling (e.g. Reference KanthaKantha, 1995; Reference Ullman and WilsonUllman and Wilson, 1998).

Assimilation studies involving sea-ice applications have been relatively few to date. The primary work was a series of papers that used a Kalman filter method to assimilate passive-microwave data with a simple model of ice growth and transport (Reference Thomas and RothrockThomas and Rothrock, 1989, Reference Thomas and Rothrock1993; Reference Thomas, Martin, Rothrock and SteeleThomas and others, 1996). Though employing a simple model, the results demonstrated the ability of assimilation to produce more realistic estimates of ice concentration than from observations alone. Assimilation of observed ice motions into a high-resolution sea-ice model via a simple direct insertion demonstrated the potential of assimilating remotely-sensed ice velocities within standard, two-dimensional, dynamic-thermodynamic models (Reference Maslanik, Maybee and SteinMaslanik and Maybee, 1994). The implementation of a more complex, optimal-interpolation method substantially reduced biases and errors in modeled and observed motions and produced a long time series of complete and more accurate assimilated ice motions (Reference Meier, Maslanik and FowlerMeier and others, 2000). Here, we apply the ice-motion fields from these optimal-interpolation assimilation products to examine two synoptic events.

Methodology

Observed motions

Two sources of observed motions are used in this research. Ice motions are derived from daily composite SSM/I 85 GHz imagery obtained from the U.S. National Snow and Ice Data Center (NSIDG) using a cross-correlation method (Reference Emery, Fowler, Hawkins and PrellerEmery and others, 1991). This method tracks a feature on a consistent grid over two succeeding images. The image resolution is 12.5 km, which limits the ability to detect small motions on daily time-scales and results in a relatively "noisy" motion field. Oversamphng and filtering methods were employed to improve the effective resolution. The final SSM/I motions are produced on a 60 km resolution grid. When compared to buoy observations, the daily rms motion error for the Arctic is on average 5−6 cm s−1, with a bias of <0.5 cms−1 (Reference Meier, Maslanik and FowlerMeier and others, 2000). (Similar comparisons for the Southern Ocean (Reference Heil, Fowler, Maslanik, Emery and AllisonHeil and others, 2001) indicate larger biases in the Antarctic.)

Buoy motions are computed from daily buoy-location data obtained from the International Arctic Buoy Program (IABP). The rms motion error is dependent on the location error (±350 m) and is approximately 0.5 cm s–1 for daily motions (Reference Colony and RigorColony and Rigor, 1993).

Model motions

The model used in this research is a two-category sea-ice model with three ice classes based on age, an approximation of multiple thickness levels used in energy-budget calculations (Reference Walsh and ZwallyWalsh and Zwally, 1990), a detailed surface albedo treatment (Reference Ebert and CurryEbert and Curry, 1993) and a viscous-plastic ice rheology. The model is forced with daily fields of pressure, temperature, downwelling longwave and shortwave radiation, and geostrophic winds from the U.S. National Centers for Environmental Prediction (NCEP) re-analysis fields (Reference KalnayKalnay and others, 1996). The model is centered at the North Pole and has 80 km gridcells (Reference Maslanik and DunnMaslanik and Dunn, 1997). Ice motions are calculated at 8h time-steps and output daily.

Assimilation method

An optimal-interpolation assimilation method is employed to combine the observed and modeled ice motions using a weighted linear combination described by Equation (1):

(1)

Here, u is the assimilated parameter (in this case the u component of velocity), k is the model gridpoint, and j is one of n observations in the neighborhood of the κth model gridpoint. The weights, akj, are calculated from model and observation error covariances to minimize the estimate error. The Kalman filter method used in previous sea-ice assimilation studies (Reference Thomas and RothrockThomas and Rothrock, 1993) is a generalized optimal-interpolation method where the error covariances are allowed to evolve with the model state. The simplification of using constant error covariances limits the effectiveness of the assimilation, but reduces the computational burden considerably and makes the use of detailed ice models practical.

The method described here is implemented with seasonally and regionally prescribed error covariances to improve the assimilation’s effectiveness. It could have been implemented using daily error covariances for these synoptic, hindcast case-studies. However, the scheme was designed to be flexible for applications to longer time series and possible forecast studies (where the daily error covariances would not be known). Therefore, the methodology was not changed from its original form. Buoy observations are assimilated with their own, seasonally and regionally invariant, error covariances that reflect the more accurate error characteristics of the buoys. For more details on the assimilation methodology, see Reference Meier, Maslanik and FowlerMeier and others (2000).

Results

Beaufort Sea, November 1992

A previous research study on applications of combining multiple forms of data for study of the Arctic considered variability in sea-ice concentration as a function of ice transport in the Beaufort Sea in November 1992 using SSM/I and AVHRR imagery (W.N. Meier and others, 1997, http://earthinteractions.org). On 19 November, AVHRR-denved ice motions combined with SSM/I-derived ice concentrations show a strong convergence event along the Canadian coast due to northerly winds arising from low pressure in the eastern Beaufort Sea (Fig. 1). Sea ice is advected toward the coast on the 19th, causing an increase in ice concentration. By the 20th, the coastal zone appears to be filled with ice, such that the southward drift into the Mackenzie River delta is deflected eastward and westward.

Fig. 1.

AVHRR-derived ice motion and percentage change in ssm/iice concentration from previous day for (a) 19 november 1992, and (b) 20 november 1992 the dotted line denotes a line of convergence where ice motion suddenly decreases. adapted from w.n. meier and others (1997, http://earthinteractions.org).

Here, we revisit that study using the assimilation methodology, comparing ice motions from the stand-alone model simulation with motions from the assimilation mode of the model. On 19 November, the stand-alone model shows some convergent motion along the Canadian coast (Fig. 2). However, there is a clear eastward component of the flow. This is because the NCEP pressure field indicates that the low pressure is substantially north of the coast. By 20 November, the model shows the convergent flow being retarded and more of the flow being deflected to the east and west. The assimilation case shows more southerly, convergent motion on the 19th than the standard model, with motion due south towards the coast and little eastward or western deflection of ice (Fig. 3). The assimilated motions indicate that the center of the pressure system is on the eastern Alaskan coast. By the 20th, the assimilated motions show the flow being deflected eastward. Thus, the assimilated motions yield better agreement with the higher-resolution AVHRR motions and the changes in the SSM/ I ice concentrations. This is mostly due to the correction of errors in the NCEP atmospheric forcings by the assimilated motions, although errors in ocean currents and the model ice dynamics are also likely to be affected. The qualitative improvement shown by the assimilated motions (from SSM/I, buoys and model) here in comparison with the AVHRR motions recapitulates previous statistical analyses between SSM/I-model assimilated motions with buoy-observed motions (Reference Meier, Maslanik and FowlerMeier and others, 2000).

Fig. 2.

Model ice motion in the beaufort sea on (a) 19 november 1992 and (b) 20 november 1992

Fig. 3.

Assimilated ice motion in the beaufort sea on (a) 19 november 1992 and (b) 20 november 1992

Another interesting aspect of the flow occurs just north of the Bering Strait. While the model shows westward motion in the region on the 19th, the assimilation case indicates a strong northward flow. This is the result of the apparent misplacement of high pressure in the western Beaufort by the NCEP forcings. By the 20th, both the assimilation and the model indicate southwestward flow near the Bering Strait.

Because the model is driven by NCEP pressure fields, the model’s simulated ice advection does not produce as much convergence on 19 November as the assimilation case (Fig. 4). Since such convergence of the ice pack is a threat to navigation, coastal operations and military submarine activities, accurate simulation of such events is of particular interest.

Fig. 4.

Ice convergence in the beaufort sea on 19 november 1992for (a) model and (b) assimilation. contour levels have units of 106 s–1.

Barents Sea/Fram Strait, March 1993

Ice export through Fram Strait is an important component of the Arctic sea-ice mass balance since most ice advection out of the Arctic occurs through Fram Strait. Ice that forms in the marginal ice zones in the coastal seas tends to be circulated into the central Arctic, growing and aging as it moves (Reference Colony and ThorndikeColony and Thorndike, 1984). Eventually, ice that does not melt during the Arctic summers is advected out of the Arctic, primarily through Fram Strait. After passing through Fram Strait, the ice enters the Greenland, Iceland and Norwegian (GIN) Seas where it meets warmer water from the North Atlantic and melts.

The Barents Sea is one such coastal region where a great deal of ice production occurs. Ice produced in the Barents may be advected northward into the central Arctic or westward into the GIN seas, depending on conditions. It is an important region in setting up the Arctic halocline, a layer of cold, saline stratified water that insulates the mixed layer and the overlying sea ice from the warm, saline Atlantic water mass (Reference Aagaard, Coachman and CarmackAagaard and others, 1981). Changes in this region have been observed in the 1990s that suggest that the character of the formation and evolution of the halocline may be changing (Reference Steele and BoydSteele and Boyd, 1998). This could result in or be the result of changes in sea-ice cover, heat fluxes in the Arctic, and ocean circulation. In addition, the region encompasses important shipping routes and fishing sites. Knowledge of the drift of ice and the character of the ice (thin first-year vs thick multi-year) is important for route planning.

Here, we investigate ice motion in this region on 18 March 1993. A large discrepancy exists between the model and observed pattern of ice motion. SSM/I indicates typical ice motion in the region, with export of ice from the northern coast of Greenland through Fram Strait (Fig. 5a). This would primarily be thick, multi-year ice with low salinity. Thinner ice formed in the Barents Sea is shown to advect northward into the pack-ice region.

Fig. 5.

Ice motion in the fram strait/barents sea region on 18 march 1993for (a) ssm/i, (b) model and (c) assimilation.

On the other hand, the model shows little export through Fram Strait (Fig. 5b). Flow is primarily westward from the Barents Sea and from the west coast of Svalbard. Thus, compared to the observed motions, much thinner and more saline ice is being advected into the GIN seas.

Through optimal-interpolation assimilation, the buoy and SSM/I observations constrain the model closer to observations (Fig. 5c). The assimilation case shows substantial outflow of ice from North Greenland through Fram Strait. The westward motion out of the Barents Sea is diminished, and northward motion is apparent, in agreement with the observations.

Because of increased outflow from the central Arctic, assimilation yields greater ice export through Fram Strait. The model yields an export of 0.10 Sv on 18 March, while assimilation yields over twice as much export, 0.21 Sv.

Though this is only a 1 day case-study, it illuminates the ability of assimilation to use improved ice-motion fields to study synoptic-scale motion patterns. These synoptic events could have seasonal consequences for the character of the ice cover in the region. A previous study indicates that the model consistently underestimates ice export through Fram Strait compared to the assimilated case over the course of several years (Reference Meier, Maslanik and FowlerMeier and others, 2000). Models are commonly used to study ice export through Fram Strait (e.g. Reference Harder, Lemke and HilmerHarder and others, 1998) and halocline and deep-water formation (e.g. Reference Goosse and FichefetGoosse and Fichefet, 1999). Assimilation indicates that use of observations may help to remove model biases and errors in forcings, constraining them closer to physical reality and yielding better insights into important ice-formation and melt processes in the region.

Conclusion

Synoptic-scale events in the Arctic Ocean can cause sudden changes in the patterns of ice motion. These changes can result in the formation of leads where heat and vapor transfer is facilitated. Convergent motion can cause ridging, affecting the ice mass balance and impeding operational activities.

Here, we demonstrated, via two example case-studies, the utility of data assimilation to combine observed and modeled motions to better characterize such events. The assimilated model simulates a convergence event along the north Alaskan coast that is indicated by AVHRR-derived ice motions and SSM/I ice concentrations. In the Fram Strait/Barents Sea region, the assimilated model simulates high inflow into the GIN seas from the central Arctic through Fram Strait and low inflow from the Barents Sea.

In both of these case-studies, assimilation substantially reduces errors in the estimates of ice motion and improves derived products of ice convergence and Fram Strait ice-volume export. The more accurate representation of these synoptic events demonstrates the utility of a data-assimilation approach for a variety of regional Arctic studies.

Acknowledgements

This work has been supported by the NASA Mission to Planet Earth Fellowship Program (NASA NGT-5−30034), the University Corporation for Atmospheric Research Visiting Scientist Program and NASA grant NAG-6820. SSM/I and buoy data were obtained from the NSIDC and the IABP The U.S. National Oceanic and Atmospheric Administration/NASA Polar Pathfinder projects provided AVHRR ice-motion and additional data. Thanks to the reviewers for their helpful comments.

References

Aagaard, K., Coachman, L. K. and Carmack, E. C.. 1981. On the halocline of the Arctic Ocean. Deep-Sea Res., Ser. A, 28, 529−545.Google Scholar
Agnew, T.A., Le, H. and Hirose, T.. 1997. Estimation of large-scale sea-ice motion from SSM/I 85.5 GHz imagery. Ann. Glacial., 25, 305−311.Google Scholar
Charney, J., Halem, M. and Jastrow, R. 1969. Use of incomplete historical data to infer the present state of the atmosphere. J. Atmos. Set., 26, 1160−1163.Google Scholar
Colony, R. L. and Rigor, I. G.. 1993. International Arctic Ocean Buoy Program data report for 1 January-Si December 1992. Seattle, WA, University of Washington. Applied Physics Laboratory. (Tech. Mem. PL-UWTM 29−93.)Google Scholar
Colony, R. and Thorndike, A. S.. 1984. An estimate of the mean field of Arctic sea ice motion. J. Geophys. Res., 89(C6), 10,623−10,629.Google Scholar
Ebert, E. E. and Curry, J. A.. 1993. An intermediate one-dimensional thermodynamic sea ice model for investigating ice-atmosphere interactions. J. Geophys. Res., 98(G6), 10,085−10,109.Google Scholar
Emery, W. J., Fowler, C. W., Hawkins, J. and Preller, R. H.. 1991. Fram Strait satellite image-derived ice motions. J. Geophys. Res., 96(C3), 4751−4768. (Correction: J. Geophys. Res., 96(C5), 1991,8917−8920.)Google Scholar
Emery, W. J., Fowler, C.W. and Maslanik, J. A.. 1997. Satellite-derived maps of Arctic and Antarctic sea-ice motion. Geophys. Res. Lett., 24(8), 897−900.CrossRefGoogle Scholar
Ghil, M. and Malanotte-Rizzoli, P.. 1991. Data assimilation in meteorology and oceanography. Adv. Geophys., 33,141−266.Google Scholar
Goosse, H. and Fichefet, Th.. 1999. Importance of ice-ocean interactions for the global ocean circulation: a model study. J. Geophys. Res., 104(C10), 23,337−23,355.Google Scholar
Harder, M., Lemke, P. and Hilmer, M.. 1998. Simulation of sea ice transport through Fram Strait: natural variability and sensitivity to forcing. J. Geophys. Res., 103(C3), 5595−5606.Google Scholar
Heil, P., Fowler, C.W., Maslanik, J. A., Emery, W.J. and Allison, I.. 2001. A comparison of East Antarctic sea-ice motion derived using drifting buoys and remote sensing. Ann. Glacial, 33 (see paper in this volume).Google Scholar
Hibler, W D. III. 1979. A dynamic thermodynamic sea ice model. J. Phys. Oceanogr., 9(7), 8l5−846.Google Scholar
Kalnay, E. and 21 others. 1996. The NCEP/NCAR 40−year reanalysis project. Bull. Am. Meteorol. Sac, 77(3), 437−471 Google Scholar
Kantha, L. 1995. Barotropic tides in the global oceans from a nonlinear tidal model assimilating altimetric tides. 1. Model description and results. J. Geophys. Res., 100(C12), 25,283−25,308.Google Scholar
Kwok, R., Curlander, J. C., McConnell, R. and Pang, S. S.. 1990. An ice-motion tracking system at the Alaska SAR facility. IEEE J. Oceanic Eng., OE-15(1), 44−54.Google Scholar
Kwok, R., Rothrock, D. A., Stern, H. L. and Cunningham, G. F.. 1995. Determination of the age distribution of sea ice from Lagrangian observations of ice motion. IEEE Trans. Geosci. Remote Sensing, GE-33(2), 392−400.Google Scholar
Kwok, R., Schweiger, A., Rothrock, D A., Pang, S. and Kottmeier, C.. 1998. Sea ice motion from satellite passive microwave imagery assessed with ERS SAR and buoy motions. J. Geophys. Res., 103(C4), 8191−8214 Google Scholar
Liu, A. K. and Cavalieri, D. J.. 1998. On sea ice drift from the wavelet analysis of the Defense Meteorological Satellite Program (DMSP) Special Sensor Microwave Imager (SSM/I) data. Int. J. Remote Sensing, 19(7), 1415−1423.Google Scholar
Maslanik, J. and Dunn, J.. 1997 On the role of sea-ice transport in modifying Arctic responses to global climate change. Ann. Glacial., 25, 102−106.Google Scholar
Maslanik, J. A. and Maybee, H.. 1994. Assimilating remotely-sensed data into a dynamic-thermodynamic sea ice model. In Stein, T. I., ed. IGARSS’94. Surface and Atmospheric Remote Sensing: Technologies, Data Analysts and Interpretation, Pasadena, California, USA, August 8−12,1994. Proceedings, Vol. 3. Piscataway, NJ, Institute of Electrical and Electronics Engineers, 1306−1308.Google Scholar
Meier, W. N., Maslanik, J. A. and Fowler, C.W.. 2000. Error analysis and assimilation of remotely sensed ice motion within an Arctic sea ice model. J. Geophys. Res.,W5(G2), 3339−3356.Google Scholar
Ninnis, R. M., Emery, W.J. and Collins, M.J.. 1986. Automated extraction of pack ice motion from advanced very high resolution radiometer imagery. J. Geophys. Res., 91(C9), 10,725−10,734.Google Scholar
Steele, M. and Boyd, T.. 1998. Retreat of the cold halocline layer in the Arctic Ocean. J. Geophys. Res., 103(C5), 10,419−10,435.Google Scholar
Thomas, D.R. and Rothrock, D. A.. 1989. Blending sequential scanning multichannel microwave radiometer and buoy data into a sea ice model. J. Geophys. Res., 94(C8), 10,907−10,920.Google Scholar
Thomas, D. R. and Rothrock, D. A.. 1993. The Arctic Ocean ice balance: a Kalman smoother estimate. J. Geophys. Res., 98(C6), 10,053−10,067 Google Scholar
Thomas, D., Martin, S., Rothrock, D. and Steele, M.. 1996. Assimilating satellite concentration data into an Arctic sea ice mass balance model, 1979−1985. J. Geophys. Res., 101(C9), 20,849−20,868.Google Scholar
Thorndike, A. S. and Colony, R.. 1982. Sea ice motion in response to geostrophic winds. J. Geophys. Res., 87(C8), 5845−5852.Google Scholar
Ullman, D. S. and Wilson, R. E.. 1998. Model parameter estimation from data assimilation modeling: temporal and spatial variability of the bottom drag coefficient. J. Geophys. Res., 103(C3), 5531−5549.Google Scholar
Walsh, J. E. and Zwally, H.J.. 1990. Multiyear sea ice in the Arctic: model- and satellite-derived. J. Geophys. Res., 95(C7), 11,613−11,628.Google Scholar
Figure 0

Fig. 1.AVHRR-derived ice motion and percentage change in ssm/iice concentration from previous day for (a) 19 november 1992, and (b) 20 november 1992 the dotted line denotes a line of convergence where ice motion suddenly decreases. adapted from w.n. meier and others (1997, http://earthinteractions.org).

Figure 1

Fig. 2.Model ice motion in the beaufort sea on (a) 19 november 1992 and (b) 20 november 1992

Figure 2

Fig. 3.Assimilated ice motion in the beaufort sea on (a) 19 november 1992 and (b) 20 november 1992

Figure 3

Fig. 4.Ice convergence in the beaufort sea on 19 november 1992for (a) model and (b) assimilation. contour levels have units of 106 s–1.

Figure 4

Fig. 5.Ice motion in the fram strait/barents sea region on 18 march 1993for (a) ssm/i, (b) model and (c) assimilation.