Monitoring of the Arctic Sea Ice and Icebergs:Satellite Data Processing and Analysis of Climate Change in the Cryosphere
Volkov, Vladimir1; Sandven, Stein2; Bobylev, Leonid1; Demchev, Denis3; Korzhikov, Alexander4; Zakhvatkina, Natalia3

Presented are current results received under the MAIRES project (Monitoring Arctic land and sea Ice using Russian and European SatellitesFP7-SPACE-2010-1, ref. 1263165, 2011-2014) dedicated to develop methodologies for satellite monitoring of the Arctic sea ice and icebergs identification. The developed methodologies to retrieve quantitative information from the ESA and RKA (Russian Space Agency) data, and examples of satellite derived products and its analysis are described. The main initial satellite data were from the Synthetic Aperture Radar (SAR), optical and infrared images, passive microwave data and others. Results from the project can contribute to improved understanding of climate change and provide useful data for scientists, policy-makers and the general public. Sea ice classification An algorithm, elaborated for ice type recognition and calculation of multi-year ice (MYI) concentration in the Central Arctic, is based on the Bayesian classification, which accounts for differences in the probability density functions, and allows minimizing the error probability. This algorithm is pixel-based and uses an a priori probability of occurrence of ice types. The decision is made in favor of the sea ice class, which has a maximum a posteriori probability. Conditional probabilities were assessed from calibrated SAR images, where areas typical of each ice type were delineated visually, and histograms of the backscattering coefficients were calculated. Statistically significant estimates of conditional probabilities were derived for MYI, deformed first-year ice (DFYI) and level first-year ice (LFYI) during winter. Due to the absence of reliable estimates of conditional probabilities for new ice and young ice we made an assumption that three ice types (MYI, LFYI, and DFYI) are observed in the Central Arctic and hence used the partial concentration of these ice types. For using this technique the distribution densities and a priori probabilities of the ice types have to be estimated. The Bayesian algorithm has been developed for ENVISAT ASAR sea ice classification in the Central Arctic for winter. A simple algorithm has been proposed enabling ENVISAT ASAR sea ice type classification (MYI, LFYI, and DFYI) to be adapted using RADARSAT-2 SAR data. Classification of RADARSAT-2 SAR images is made for three ice classes. Sea ice drift analysis Analysis of the ice drift fields variability in the Arctic Ocean during two last decades at the turn of the 20th and 21st centuries using microwave and radar satellite was performed. Two daily satellite derived data sets presented at regular grid points with a step of about 32 km over the entire Arctic Ocean area were used as initial data. One of them (for the winter period from 1992 up to now) was prepared by the Center for Satellite Exploitation and Research (CERSAT) at IFREMER (France), an other data set (as well as for winter and summer seasons from 1979 up to now) was developed by the National Snow and Ice Data Center - NSIDC (USA). Also original drift data calculated using our algorithm of ice drift retrieval from ASAR images were used. The full potential of the algorithm could be realised with continuously operational wide swath SAR instrument such as the forthcoming Sentinel-1. The vectorial-algebraic approach was chosen as a fundamental method for analysis of seasonal and year-to-year variability of the ice drift series. This approach allows significantly compressing the initial information and most adequately describing the vector time series of full-scale and model data restricted by a set of statistical characteristics in invariant form. The results of such a statistical analysis make possible to describe a detailed fields variability, to detect some zones with uniform dynamics, to access an intensity of the water and ice outflow and temporal variability of the status of circulation systems in the course of time. A joint analysis of the drift data and large-scale weather processes during the current warming epoch demonstrates a determinative role of the global atmospheric circulation in forming ice conditions. The study was developed keeping in mind the classification of large-scale weather processes developed at the Arctic and Antarctic Research Institute (AARI). The classification includes 26 types, divided into six groups: A, B, V, G, D and K. The correspondence between the prevailing type of atmospheric circulation and the specific properties of the drift field was determined. Thus, the degree of influence on the process of ice formation in the Arctic most significantly manifests itself through group B. The processes of this group are characterized by the development of an anticyclonic field over most of the Arctic basin and hence by the lack of strong advection of warm air masses from middle-latitude zone, the predominance of eastern air flows and minimal cloud cover. In this case the most favorable conditions for increasing the sea ice cover in the Arctic basin are formed. And contrarily, V-group processes are characterized by the development of the cyclonic field over the Western Arctic and the cyclonic field - over the Eastern Arctic, and could be resulte in decreasing the ice cover in the Arctic Ocean. The gradual increase in frequency of large Arctic anticyclones over the Arctic Basin, observed since 2000, may affect the subsequent growth of ice area. This may serve as a prelude to the subsequent increase of the polar sea ice area. Based on the Pathfinder Sea Ice Motion dataset we obtained, a new scheme of Arctic zones with heterogeneous ice circulation regime was established. An original method of an objective-zonation was adapted to ice drift fields. Icebergs The methods of identification of Arctic icebergs in the visible Landsat, "Monitor_E", Terra, Aqua (ASTER and MODIS) satellite images and Envisat ASAR images have been developed, and features for iceberg detection are determined. The conducted analysis revealed that iceberg monitoring by satellites should be based on using SAR and optical images at a resolution of 10 meter or better. The technological scheme of a combined use of different types of satellite information depending on ice and hydrometeorological conditions for iceberg identification are presented. The methodology consists of the following procedures: definition of the area under study, determination of iceberg detection conditions (open water, fast ice, drifting ice), selection of image type (coverage and resolution) and search for available image in the archive, image analysis, estimation of iceberg features and iceberg identification, validation of identification results using sequential images, a combined analysis of identification results from satellite images in different spectral bands, composition of iceberg distribution maps, supplement of iceberg data base.