Sea ice concentration data from passive microwave satellite data (PMW), obtained from 1978 to present, is the most important data set to document the decrease of the Arctic sea ice area in the recent decades. There are many retrieval algorithms to calculate ice concentration from the satellite data, and these algorithms produce different results for sea ice area. In ArcticSIV eleven algorithms have been compared and the results have been analyzed to find seasonal and interannual variability, long-term trend and spatial variability. The standard deviation between the algorithms have been used to estimate error bars for the ice area and ice extent calculations.
The algorithms differ in annual sea-ice area by up to 1.3 million km2 (up to 1.00 million km2 for extent). In winter, the algorithms differ most in the Sea of Okhotsk, and the Bering, Labrador, Greenland and Barents seas. This is explained by the proximity of larger areas of open water in these regions, which increases the atmospheric effects and thus increases the algorithms uncertainties. In summer the differences are higher than in winter and are present everywhere where there is ice at this time of the year. During this period the surface is rather complicated because of the wet snow, melt-ponds and thin ice and water mixtures, which make the retrievals less reliable.
Linear trends were calculated and compared for sea-ice concentration, area and extent in order to estimate the change in these parameters during the 1979-2012 and 1992-2012 periods, including the seasonal change in trends. The annual Arctic sea-ice area decreased from 1979 to 2012 at the rate of 0.517-0.573 million km2 per decade for the 1978-2012 period. The decrease was higher when estimating for 1992 to 2012 period: 0.866 to 0.975 million km2/dec. The annual sea ice decrease is mostly determined by summer ice decline. The decrease was most pronounced in the East Siberian and Beaufort seas, northern Laptev, Kara and Barents seas. The algorithms ensemble and in particular the sea-ice concentration standard deviation describing part of the uncertainty is suggested in order to provide users with more insight into the uncertainties and potential biases of sea-ice concentration data.
In order to assess performance of the algorithms, PMW sea ice concentrations from all the algorithms were compared to SAR retrievals. The comparison was done for both total ice concentration (by all the algorithms) and multi-year ice concentration (which is possible to obtain only by four of the algorithms). It was also found that all the algorithms underestimate sea ice concentration over thin ice (below 40 cm) and in areas where melt ponds are present. Since all the algorithms are sensitive to weather effects, a weather filer should be applied. Performance of such a filter was tested and it was shown that, in the areas of sea ice concentrations of up to 25%, it filters out areas with actual sea ice. The algorithms were found to be very sensitive to the tie-points choice, and thus the dynamical tie-points solution was suggested.
A major scientific question is how much the total ice volume in the Arctic is changing. To answer this question, data on ice concentration must be combined with data on ice thickness. But ice thickness data is much more scarce and has not been collected in a systematic way compared to ice concentration data. A method to extract sea ice thickness from radar altimetry data has been developed for ERS-1/2, ENVISAT and CryoSat data from 1993 to present, and this method is presently validated in the ESA Climate Change Initiative project. ArcticSIV has contributed to this validation work where ice draft measurements from submarine surveys, moorings with Upward Looking Sonar data, drilling of holes during icebreaker expeditions, and airborne laser and radar altimeter measurements. Preliminary results include estimation of uncertainties in retrieving ice thickness from radar altimetry where several factors such as snow depth, snow and ice density, and penetration depth of the radar signal are important. The studies have also analysed the accuracy of both radar and laser altimetry data, suggesting that ice thickness retrievals have an uncertainty of order 0.5 m. The relative accuracy of ice thickness data is therefore in the range of 20 to 40 %, which is are much lower than for ice concentration, where relative accuracy is of order 5%. By combining ice extent and ice thickness data, the total ice volume can be determined, and studies have been performed to analyze uncertainties and trends in Arctic ice volume over the last decade. The results of this work are published in a two papers in The Cryosphere.
The aim is to implement and validate retrieval algorithms of ice concentration, types, thickness and volume in the Arctic from passive and active satellite microwave data, in order to improve the quantification of sea ice variability, trends and uncertain ties over the last 3 decades. The motivation for this is that all available passive microwave algorithms (11) show discrepancies up to 1 million km2 in retrieval of the sea ice extent and area and 20-40% differences in ice concentration. The project has f ive work packages: Implementation and validation - The optimal algorithm(s) - Sea ice thickness - Variability and trends - Arctic ROOS dissemination. For the first time 11 passive microwave sea ice algorithms will be compared and validated using the same brightness temperature data from SMMR, SSM/I since 1978 and recent AMSR-E and SMOS data. Validation will be done by SAR, Scatterometer and optical and infrared satellite data such as AVHRR and Landsat. Sea ice algorithms for thin and thick ice will be imp lemented using passive microwave, radar and laser altimeters, optical and infrared satellite data and validated by using in-situ observations. Based on improved sea ice concentration and thickness the total ice volume in the Arctic will be estimated for t he last two decades apart from summer months when the snow and ice are wet. The sea ice variability, trends, uncertainties and their relationship to atmospheric, ocean and greenhouse forcing over the last 3 decades will be analyzed supported by model simu lations. The project will cooperate with 4 national partners and 4 international partners. We expect that the result of the project will be useful for scientific community, climate models validation and initialization, operational oceanography (ice monito ring, data assimilation and forecasting), fisheries, navigation and offshore industry. All results including publications will be disseminated to the users via the Arctic ROOS (http://arctic-roos.org)