ICE DRIFT IN THE FRAM STRAIT FROM ENVISAT ASAR DATA Stein Sandven (1), Kjell Kloster (1), and Knut F. Dagestad (1) (1) Nansen Environmental and Remote Sensing Center (NERSC), Thormøhlensgte 47, N-5006 Bergen, Norway. E-mail: stein.sandven@nersc.no ABSTRACT Since 2003 ENVISAT Advanced SAR (ASAR) images have been produced over many sea ice areas in the Arctic. This paper shows preliminary results of ice drift retrieval from ASAR Wideswath images in the Fram Strait starting in February 2004. The ice drift vectors from SAR are compared to from scatterometer and passive microwave, showing that SAR can provide more accurate ice drift throughout the year. The ice drift are used to estimate the ice area flux and validate ice models in the strait. 2. METHOD ASAR Wideswath images of about 400 km width are used to map the Fram Strait (Fig. 1a.) where one image is sufficient to cover the ice area in the strait. The area can in principle be covered with new images almost every day because of the high latitude. 1. INTRODUCTION Data from polar orbiting satellites play an increasingly important role in monitoring sea ice and other parts of the marine and polar environment (Johannessen and Sandven, 2002; Johannessen et al., 2007). From 2003 European Space Agency (ESA) started to produce Wideswath images for sea ice monitoring in various regions of the Arctic. The development of operational ice monitoring services under GMES requires that satellite SAR are used to produce ice drift and other sea ice parameters (Sandven et al., 2004). The Fram Strait is the main area for sea ice and liquid freshwater export from the Arctic Ocean to the Nordic Seas and the North Atlantic. The sea ice flux is a key climate parameter which is monitored on fixed moorings across 79 N in the Fram Strait and by drifting ice buoys when they are advected through the strait. Ice drift from satellite has been estimated for many years using passive microwave and scatterometer, but retrievals have mainly been obtained for the winter months. Furthermore, the resolution and accuracy of the ice drift estimates from these have not been compared to similar retrievals from Wideswath SAR. This paper investigates regular use of wideswath SAR in order to achieve more accurate estimates of ice drift and ice area flux in the strait. Figure 1. (a) SAR image of 25 February 2004, with ice drift vectors from 22-25 February; (b) interpolated ice drift vectors across 79 N. Proc. Envisat Symposium 2007, Montreux, Switzerland 23 27 April 2007 (ESA SP-636, July 2007)
In practice, it has been possible to collect SAR Wideswath every 3-4 days almost continuously from February 2004 to present. In this study the ice drift estimation is focused on covering the 79 N latitude, where oceanographic moorings with current meter and Upward-Looking Sonar are used to monitor ice thickness and ice drift at fixed locations. An ice drift algorithm is presently tested to retrieve ice drift automatically (Kloster et al., 1992; Sandven et al., 1999), but manual analysis is used to ensure that the accuracy of the ice drift vectors is 10 % or better. typical mean velocity in winter is 0.20 ms -1 and in summer 0.10 ms -1 (Fig. 3a). Maximum mean velocities have been up to 0.40 ms -1. In July and August periods of no ice in the strait have been observed. Estimation of the ice drift from satellite in the outer Marginal Ice Zone can be difficult, because persistent ice features cannot be recognised. The reason is that the icepack is mostly broken up in small ice floes with rapidly changing SAR signatures. In addition, the ice drift velocity is at a maximum in the core of the East Greenland Current, which often coincide with the ice edge (Johannessen et al., 1994). In the outer Marginal Ice Zone the vectors are extrapolated to the ice edge. Near the Greenland coast there is usually a zone of zero ice flux because of landfast ice. The shear zone between landfast and drifting ice can readily be determined from the SAR images. The ice drift vectors retrieved from the SAR images are interpolated/extarpolated across the 79 N latitude at a resolution of about 20 km, with zero velocity in the fastice region and a constant value towards open water. In order to estimate the ice area flux in the Fram Strait, profiles of ice concentration from passive microwave satellite (SSMI) are obtained at the same time as the SAR images are used to retrieve ice displacement vectors. The southward component of the ice drift vectors are multiplied by the ice concentration profile, resulting in profiles of the area flux for each time interval (Fig. 2). These fluxes are then accumulated to monthly and annual fluxes which can be compared with other observations (i.e Kwok et al., 2004) as well as model simulations. 3. RESULTS 3.1 Ice area flux estimation Starting in February 2004, NERSC has produced ice area flux profiles across 79 N using ice drift vectors retrieved from Wideswath SAR images and ice concentration profiles from SSMI. The time interval between the images is normally 3 days, but in some cases 6 days. There is considerable short-time variability in the velocity field, superimposed on a seasonal cycle, where Figure 2. (a) Profiles across 79 N of ice concentration from SSMI and north component of ice drift from two SAR images February 22 and 25, 2004; (b) profile of ice area flux for the three day period. The monthly mean area flux, which is estimated from February 2004 to September 2006 based on the SAR ice drift vectors and the SSMI ice concentration is shown in Fig. 3b. There is a pronounced maximum in the winter months and a minimum in the summer months, which is in agreement with previous studies (Vinje et al., 1998, Kwok et al., 2004). The annual area flux from the SAR is estimated to be 0.73 x 10 6 km 2 /year for 2004 and 0.66 x 10 6 km 2 /year for 2005. Area fluxes retrieved from passive microwave from 1978 to 2002 showed a mean value of 0.86 x 10 6 km 2 /year with a standard deviation of 0.10 x 10 6 km 2 /year (Kwok et al. 2004). A maximum area flux of 1.11 x 10 6 km 2 /year was observed in 1994-95, while minimum fluxes of about 0.72 x
10 6 km 2 /year were observed in 1978-79, 1984-85 and 1990-91. It is noteworthy that the retrievals from 2004 and 2005 are near or below the previously observed minima. Use of passive microwave has provided time series of area flux only for the months October to May. For the summer months the area fluxes were modelled. Kwok et al (2004) have also used SAR from RADARSAT to observe area fluxes in the period 1996 to 2001 to enhance and validate the passive microwave retrievals as well as to study the cross strait profiles of the ice drift. centre (http://www.ifremer.fr/cersat/en/index.htm). Ifremer provides different ice drift products based on scatterometer only and merged scatterometer and passive microwave. Ice drift vectors from CERSAT have been extracted for the same 3-day intervals as the SAR-derived vectors, covering the winter season 2005-2006. Examples of simultaneous ice displacement vectors in the Fram Strait from SAR and merged Ifremer retrievals are shown for late March early April 2006, using a three day interval (Fig 4). a b Figure 3. (a) Ice drift from SAR averaged across 79 N as function of date of year for 2004 2006; (b) Monthly area flux across 79 N retrieved from SAR ice velocity and ice concentration profiles. Negative flux means northward drift. 3.2. Comparison with ice drift from Quikscat-SSMI Global ice drift are produced operationally from scatterometer and passive microwave (SSMI and AMSR- E) by Ifremer (e.g. Ezraty and Piollé, 2003). These also cover the Fram Strait, and it is interesting to compare the SAR-retrieved ice drift vectors with the Ifremer ice drift products, available at CERSAT satellite Figure 4. Examples of simultaneous ice drift retrievals from the merged scatterometer passive microwave and the ASAR Wideswath for (a) 23 26 March, (b) 26 29 March, (c) 29 March 01 April, and (d) 01 04 April 2006.
There is reasonable agreement between the two sets for this period, taking into account that gridding and interpolation is done differently. The SAR retrievals are focused on 79 N latitude with manual control of the vectors, while the Ifremer retrievals are done automatically without specific treatment for the Fram Strait. The SAR retrievals can discriminate better the border between fastice and drifting ice on the Greenland shelf, and between open water and drifting ice in the centre of the strait. At low ice drift velocity, the SAR can also obtain more reliable drift estimates than the merged product. The SAR-based vectors can be used to validate the automated products from scatterometer and passive microwave. The SAR can also fill in the gap in the Ifremer between May and October. 3.3 Quantitive comparisons A quantitative comparison between the merged Quikscat- SSMI and the SAR ice drift vectors has been done for the winter season October 2005 May 2006. For each time interval, a mean velocity was estimated from the Quikscat and SAR-derived vectors, respectively. The time series of the mean north south and east-west components in the two sets are presented in Fig. 5. The SAR shows generally higher southward velocity compared to the Quikscat-SSMI. The mean northsouth velocity for the time period shown in Fig. 5 is -0.04 m/s for the Quikscat-SSMI and -0.13 m/s for the SAR. For the east-west component, which is much smaller than the north-south, the SAR tends to show higher westward velocity than the Quikscat-SSMI. This preliminary result shows that Quikscat-SSMI underestimate the ice drift in the Fram Strait compared to SAR, which are quality controlled by manual analysis and therefore considered to be more accurate. Comparison of ice drift from SAR and Quikscat- SSMI has also been done in the Barents Sea during deployment of three drifting ice buoys. These ice buoys operated in sea ice for about one month in March 2006, providing accurate positions from ARGOS and GPS. Drifting ice buoys give the best validation for ice drift from satellite and should be used systematically to check the quality of satellite-retrieved. An example of comparison of two satellite ice drift products and from drifting buoys for the period 15 18 March 2006 is shown in Fig. 6. Figure 6. Comparison of ice drift in the northern Barents Sea from three different sources: merged Quikscat- SSMI, Wideswath ASAR, and three drifting ice buoys. The period of comparison is 15 18 March 2006. This is an example of very good ice drift retrieval from the Quikscat-SSMI, and good agreement with the ASAR Wideswath retrievals. The validation of the two satellite-derived ice drift sets by the ice buoys is also very good. A quantitative comparison of ice displacement and direction for the three sources is shown in Table 1. Figure 5. Ice drift from Quickscat-SSMI merged product and SAR for the same three-day periods from October 2005 to May 2006. The upper graph shows the northsouth component and the lower graph shows the eastwest component. Table 1. Ice displacement (distance and orientation) from three sources in the northern Barents Sea. Period: 15-18.03 ASAR Wideswath Quikscat- SSMI Drifting buoys Displacement (km) 63.0 56.9 64.4 Direction ( ) 190 192 196
More ice drift, however, need to be compared before definite conclusions can be drawn on the quality of various ice drift in the Fram Strait and other areas. 4. CONCLUSION This study has shown that ASAR Wideswath images can provide accurate estimates of ice velocity and ice area fluxes in the Fram Strait. The SAR-retrievals have been compared to ice drift from Quikscat-SSMI, showing that the latter tend to underestimate the ice drift in the strait. Validation of ice drift from SAR and Quikscat-SSMI has been done by comparison with drifting ice buoys in the Barents Sea. However, more validation of various satellite-derived ice drift products is needed in various parts of the Arctic in order to assess the error variance of the as input to sea ice modeling and forecasting. Assimilation of sea ice products from satellite in operational ice-ocean models will be an important part of the implementation of Marine Core Services under GMES in the coming years. 5. ACKNOWLEDGEMENT The study is supported by the ESA GSE ICEMON project (Cat-1 no. 2363), ESA AO 1260, the EU MERSEA Integrated project, the Norwegian Space Centre s SatOcean project and Statoil ASA. 6. REFERENCES Ezraty, R. and J. F. Piollé (2003). Sea-ice drift in the Central Arctic estimated from SeaWinds/QuikSCAT backscatter maps, User Manual, Version 1.0. CERSAT V. G. Smirnov and E. U. Mironov (2007). Polar Seas Oceanography, Remote Sensing of Sea ice in the Northern Sea Route: Studies and Applications. Praxis Springer, 2007, 472 pp. Kloster, K. et al. (1992). Ice motion from airborne SAR and satellite imagery. Adv. Space Res. Vol.12, No.7, pp.(7) 149-153. Kwok, R., G. F. Cunningham and S. S. Pang (2004). Fram Strait sea ice outflow. J. Geophys. Res., 109, C01009. Rothrock, D. A., R. Kwok, and D. Groves (2000). Satellite views of the Arctic Ocean freshwater balance. In The Freshwater Budget of the Arctic Ocean (ed. E. L. Lewis), NATO Science Series 2. Environmental Security, Vol. 70, pp. 409 451, Kluwer Academic Publishers. Sandven, S, O. M. Johannessen, Martin Miles, Lasse H. Pettersson and K. Kloster (1999). Barents Sea seasonal ice zone features and processes from ERS- 1 SAR. J. Geophys. Res. Vol. 104, No. C7, p. 15843 15857. Sandven, S., K. Kloster, H. Tangen. T. S. Andreassen, H. Goodwin and K. Partington (2004). Sea ice mapping using ENVISAT ASAR Wideswath images. IN Proceedings of the Second Workshop on Coastal and Marine Applications of SAR, 12 15 September 2003, Svalbard, Norway. ESA SP-565. Vinje, T., N. Nordlund and Å. Kvambekk (1998). Monitoring ice thickness in the Fram Strait. J. Geophys. Res., 103, pp 10437 10449. Johannessen, O. M., S. Sandven, W. P. Budgell, J. A. Johannessen and R. Shuchman (1994). Observation and Simulation of Ice Tongues and Vortex-Pairs in the Marginal Ice Zone. Nansen Centennial Volume, American Geophysical Union Monograph 85, pp. 109 136. Johannessen, O. M. and S. Sandven (2002). Monitoring of the Arctic Ocean. In Operational Oceanography, Implementation at the European and Regional Scales. Proceedings from the Second international Conference on EuroGOOS, 11 13 March 1999, Rome, Italy, pp. 165 177, Elsevier Oceanography Series, 66. Johannessen O. M., V. Yu Alexandrov, I. Ye. Frolov, S. Sandven, M. Miles, L. P. Bobylev, L. H. Pettersson,