The Home Page of the GHRSST-PP Sea Ice Technical Advisory Group (SI-TAG)
The GHRSST-PP Sea Ice TAG (SI-TAG) is responsible for issues due to sea ice within the GHRSST-PP. The SI-TAG is chaired by Peter Minnett pminnett@rsmas.miami.edu.
GHRSST follows the development in the Arctic Ocean sea ice extent and sea surface temperature
The Arctic Ocean shows very large inter annual variations in the sea ice extent and in the sea surface temperature. Year 2007 showed a record low Arctic Ocean ice cover and with the presence of first-year ice at the North Pole this year, there is a possibility that the North Pole might be ice free later this year. At the moment, the sea ice extent in the Arctic is larger than last year but smaller than the climatological sea ice extent. The sea ice minimum in September will depend upon the coming months and is difficult to predict due to the large presence of first year ice.
The sea surface temperatures of the open waters in the Arctic show large positive anomaly in the Beaufort Sea at the mouth of the Mackenzie River, where sea surface temperatures are more than 5 degrees warmer than normal. In addition, the waters in the eastern part of the Baffin Bay are significantly warmer than normal.
To follow the situation closely, have a look at the GHRSST Level 4 products and at the GMPE web site.
--------------------------------------------------------------------------------------------------------------------------------------
The GHRSST SI-TAG
There are four specific issues associated with the accurate retrieval of sea surface temperature (SST) at high latitudes using infrared radiometry:
a) The discrimination between ice-free and ice-covered water at the resolution (temporal and spatial) of the GHRSST SST retrieval schemes.
b) The discrimination between ice-free and ice-covered water at the resolution (temporal and spatial) of the GHRSST SST global analyses schemes.
c) The accurate correction of the effects of the atmosphere on the infrared radiation as it propagates from the sea surface to the satellite radiometer.
d) The reconstruction of SST in the marginal ice zone based on ice concentration, typically needed in climate-related, long-term data sets such as HadISST and ERSST.
Detecting ice-free water
Within GHRSST, the requirements are for an ice mask are spatial resolution of 1, 4 or 10 km with a six hourly update. While polar orbiters with suitable sensors map polar regions sufficiently frequently for the temporal requirement to be met (or nearly met), the spatial requirement can not be currently achieved using microwave radiometry, which is the conventional technique for mapping sea-ice. However, the ice-mask requirement for SST does not require the retrieval of ice properties, but merely identification of the presence of ice in particular infrared pixels. Increasingly strict timeliness constraints apply with increasing resolution. Ice drift velocities may well exceed 0.1 ms-1, which implies that the ice may move in excess of 2 km (pixels) within one six hourly update period, Figure 1 shows an example. The most likely solution to the ice masking problem for the high resolution infrared sensors is therefore to mask based on contemporaneous information with a suitable resolution available within the data stream of the satellite platform. Currently, this is restricted to visible bands. However, it is very analogous to the identification of cloud, and the same techniques can be used to identify ice as are used to identify cloud with the addition that coarse resolution ice analyses may provide useful background or a-priori constraints.. For infrared SST retrievals, during the day, reflected sunlight provides a powerful mechanism for identifying open, cloud-free water. Figure 2 shows part of a MODIS swath extending from the northern Pacific Ocean to the northwest coast of Greenland. Areas of open water are clearly identifiable in the "true color" image to the right, even though the discrimination between cloud and ice cover may not be straightforward. The image to the right shows the result of application of cloud-mask tests to these data and indicates a qualitatively successful outcome. Those areas identified by eye in the image on the left are deemed to be cloud and ice-free in the image to the right.
Figure 1. An example of sea ice movement during 24 hours (north east coast of Greenland, AVHRR ch2,3A,4 color composite giving the ice red colors). Ice sheet A has moved about 15km in the time between the two images (from Eastwood and Andersen, 2006).
During the polar night the problem of identifying ice becomes more difficult, but a simple temperature threshold test might be adequate. Surface temperature retrievals from spacecraft infrared radiometers below -1.8°C, the freezing point of sea water, can be classified as ice cover. However, this is prone to miss-classification as a) there is noise, perhaps systemic, in the satellite-derived surface temperature so that ice-free retrievals could fall below the threshold, and ice-covered pixels fall above the threshold; b) when melting, sea ice, especially if covered by snow, may remain frozen at temperatures above the threshold. In these cases, microwave retrievals of ice cover may have a role to play, even though they lack the high spatial resolution.
For microwave SST retrievals, side-lobe contamination from the microwave emission from sea ice prevents accurate SST measurements within several pixels of the ice edge. The large emissivity contrast between ice and open water means the contamination of the microwave SST retrievals by sub-pixel ice can be severe. This emissivity contrast means, however, that microwave radiometers are able to determine the presence of sea ice at a higher spatial resolution than the SST retrieval in both day and night conditions. The microwave ice retrieval has difficulties distinguishing sea ice in conditions that include high cloud and/or rain.
Figure 2. An example of cloud and ice cover at high latitudes. The swath is from MODIS and shows a "true color" image (left) and cloud-ice mask (right) based on a visible reflectance test. White is cloud and ice, and brown is a land mask; other colors represent open water under cloud-free skies.
Techniques that combine the information from microwave and infrared data should be explored. In that respect one should be very aware of the shortcomings of microwave observations, the most relevant of which are the reduced sensitivity to thin ice types and land contamination. These problems are of equal relevance to microwave SST retrievals. While land contamination is inherent to the poor resolution of the passive microwave observations, different microwave channels have different resolutions and the problem might be reduced by a more careful combination of channels, taking account of the increased atmospheric contamination in the high resolution microwave channels.
Figure 3 shows an example of the thin ice problem in the Davis Strait region, where the OSISAF ice edge product fails to detect a considerable amount of sea ice. The ice concentration product clearly does better and retrieves some, but not all, of the sea ice that follows the coast in the Southwesterly direction. This region was shown by Agnew and Howell (2003) to show particularly large discrepancies between ice charts and passive Microwave products. The Sea of Okhotsk is another prominent area affected by large concentrations of thin ice (Cavalieri, 1994). Andersen et al. (2006) shows that the atmospheric correction of the microwave radiances may improve the detection of thin ice. Figure 4 demonstrates this for the Odden Area off the Greenland East Coast (Wadhams & Comiso, 1999), where a patch of thin ice is detected by SAR observations on two occasions during a period in January 2004. During this period, the atmospherically corrected retrievals are more consistent and the weather filter, which is traditionally used in most existing sea ice analyses, fails to detect the feature on some occasions and fails to remove some artifacts that are most likely due to atmospheric contamination.
Figure 3. Ice products from 28 November 2006 in the Davis Strait region. From left to right: OSISAF ice edge product, OSISAF ice concentration transparently overlaid on ice edge product, and Canadian Ice Service ice chart. 
Figure 4. Series of daily ice concentration retrievals based on the DMSP F13 pass at about 11 UTC off the northeast coast of Greenland from January 20 to January 27, 2004. Location is shown in map at bottom right. Leftmost column is atmospherically corrected retrievals, middle column is weather filtered retrievals. Two Envisat ASAR wide swath (450 km swath width) quick-look images from January 22 and 26 are shown at the top right. Their coverage is marked by red polygons on the corrected retrievals for the corresponding day. (After Andersen et al., 2006).
Atmospheric Correction
The polar atmosphere is generally very dry and cold, and is an extreme in terms of the climatological distribution of atmospheric properties. As such it represents an anomalous set of conditions for routine atmospheric correction algorithms that are used to retrieve SST from infrared brightness temperatures measured from Polar Orbiting Satellites. It is expected, therefore, that systemic retrieval errors in the derived SSTs will result when they are obtained using standard atmospheric correction algorithms optimized for the global range of atmospheric variability (e.g. McClain et al, 1985). Such bias errors, usually resulting in an erroneously warm SST, are routinely observed and can be greater than 1K.
Recent work using AVHRR brightness temperature measurement collocated with ship-based radiometric skin SST measurements have shown that a simple, single channel retrieval algorithm (CASSTA - Composite Arctic Sea Surface Temperature Algorithm) can produce satisfactory accuracy in the measurement of skin SST and Ice Surface Temperature (IST; Key et al, 1997) (Vincent et al., 2007a, b). Figure 5 shows the residual SST errors using the new algorithm compared to the standard multi-channel retrieval; the reference measurements are those of the Marine-Atmospheric Emitted Radiance Interferometer (M-AERI; Minnett et al, 2001).
Figure 5. M-AERI ground truth data is compared to CASSTA, McClain SST (1985) and Key IST (Key at al, 1997) estimates. A significant gain in accuracy is evident with CASSTA, which closely follows the 1:1 line. (From Vincent et al, 2007a).
The explanation for the poor performance of the multi-channel approach is the loss of the correlation between the brightness temperatures measured at 10.5 and 11.5 µm with the atmospheric water vapor that occurs in very dry atmospheres. The brightness temperature differences, at the heart of the assumptions behind all multi-channel atmospheric correction algorithms, do not provide the appropriate information necessary to correct for the effects of the intervening atmosphere. A single-channel algorithm appears to be more appropriate.
Air-sea temperature differences
As with atmospheric water vapor, the air-sea temperature difference in polar regions manifests values that are seldom seen elsewhere over the oceans. Very large values are possible for off-ice airflow (Figure 6). The air-sea temperature difference is important in introducing uncertainties in the retrieved SSTs as it is closely related to the temperature difference between the ocean surface and the atmospheric gases that modify the infrared radiation on its passage to the satellite radiometer. Although less important than in moist atmospheres, the wide range of air-sea temperature differences encountered in polar regions introduce a source of uncertainty in the SST retrievals. It is not clear that the single-channel SST algorithms can account for such variability.
Figure 6. Air temperatures and surface skin temperatures measured by an M-AERI on the Pierre Radisson in the Amundsen Gulf of the Beaufort Sea. Very large air-sea temperature differences are found in the vicinity of the ice.
Sea ice in climate analyses
Climate analyses, such as HadISST and ERSST, use ice concentrations to estimate the SST in the marginal ice zone (e.g. Reynolds et al., 2002; Rayner et al., 2003). As such the primary requirement for ice concentration differs from the requirements in lower level SST products.
The GCOS SST&Sea Ice working group is working to compare different ice concentration products in order to obtain better knowledge of their error properties. In turn this work may complement the operational GHRSST community in its possible use of ice concentration as a background field in ice masking applications. Conversely, the GHRSST operational community focus on the ice edge and marginal ice zone may be of relevance to the GCOS SST&SI working group as these areas are known to be particularly complex (e.g. Meier, 2005).
Summary
While presenting particular problems to the accuracy of SST retrievals in the infrared, high latitude conditions can be addressed by applying standard approaches for cloud screening, possibly in combination with information such as ice concentration from passive microwave observations, to the need for discriminating between open water and ice cover. Shortcomings in the passive microwave products, primarily thin ice and land contamination need consideration and comparison of methods is probably beneficial.
Accuracy artifacts pertaining to the polar atmosphere can be addressed and very simple algorithms have been shown to function well in correcting for these effects .
Validation of all these approaches is hampered by the difficulties in obtaining accurate in situ measurements, which, in the case of SST, are only achievable by using instrumentation on ice-breaking research vessels.
The activities of the GCOS SST&SI working group and GHRSST are found to be complementary as the former has a distinct focus on ice concentration as a whole whereas GHRSST has a natural focus on the ice edge and marginal ice zone. The groups have a particular overlap in the fields of reanalysis and climate oriented analyses.
References
Agnew, T., S. Howel (2003), The use of operational ice charts for evaluating passive microwave ice concentration data, Atmosphere-Ocean, 41(4), 317-331.
Andersen, S., R.T. Tonboe, S. Kern, H. Schyberg (2006), Improved retrieval of sea ice total concentration from spaceborne passive microwave observations using Numerical Weather Prediction model fields: An intercomparison of nine algorithms, Rem. Sens. Environ., 104, 4, 374-392.
Cavalieri, D.J. (1994), A microwave technique for mapping thin sea ice, J. Geophys. Res., 99(C6), 12561-12572.
Eastwood, S., S. Andersen, 2006: Masking of sea ice for METOP SST retrieval. OSISAF Global SST progress report.
Key, J. R., J. B. Collins, C. Fowler, and R. S. Stone, 1997: High-latitude surface temperature estimates from thermal satellite data. Remote Sensing of Environment, 61, 302-309.
McClain, E., Pichel, W., and Walton, C., 1985. Comparative Performance of AVHRR-based multichannel sea surface temperatures. Journal of Geophysical Research, 90: 11587-11601
Meier, W.N. (2005), Comparison of passive microwave ice concentration algorithm retrievals with AVHRR imagery in Arctic peripheral seas, IEEE Trans. Geosci. Remote Sens., 43(6), 1324-1337.
Minnett, P. J., R. O. Knuteson, F. A. Best, B. J. Osborne, J. A. Hanafin, and O. B. Brown, 2001: The Marine-Atmospheric Emitted Radiance Interferometer (M-AERI), a high-accuracy, sea-going infrared spectroradiometer. Journal of Atmospheric and Oceanic Technology, 18, 994-1013.
Rayner, N.A., D.E. Parker, E.B. Horton, C.K. Folland, L.V. Alexander, D.P. Rowell, E.C. Kent, A. Kaplan (2003), Global analyses of sea surface temperature, sea ice and night marine air temperature since the late nineteenth century, J. Geophys. Res., 108(D14), doi:10.1029/2002JD002670.
Reynolds, R.W., N.A. Rayner, T.M. Smith, D.C. Stokes, W. Wang (2002) An improved in situ and satellite SST analysis for climate. J. Climate, 15, 1609-1625.
Vincent, R. F., R. F. Marsden, P. J. Minnett, K. A. M. Creber, and J. R. Buckley, 2007a: Arctic Waters and Marginal Ice Zones: Part 1- A Composite Arctic Sea Surface Temperature Algorithm using Satellite Thermal Data. In preparation.
Vincent, R. F., R. F. Marsden, P. J. Minnett, and J. R. Buckley, 2007b: Arctic Waters and Marginal Ice Zones: Part 2 - An Investigation of Arctic Atmospheric Infrared Absorption for AVHRR Sea Surface Temperature Estimates. In preparation.
Wadhams, P., J. C. Comiso (1999), Two modes of appearance of the Odden ice tongue in the Greenland Sea, Geophys. Res. Lett., 26(16), 2497-2500, 10.1029/1999GL900502.
(Last Updated: 18-07-2008)