Final Report Summary - MAIRES (Monitoring Arctic Land and Sea Ice using Russian and European Satellites)
Executive Summary:
The objective of the MAIRES proposal is to develop methodologies for satellite monitoring of Arctic glaciers, sea ice and icebergs. A major effort has been to compile satellite data from ESA, Russian Space Agency, and other satellite systems and establish data bases where satellite data can be searched, collocated and downloaded for further use. Methodologies have been developed to retrieve quantitative information on land ice, sea ice and icebergs from various satellite data, supplemented by in situ and other data. Examples of satellite derived products for each of the three thematic areas have been presented at conferences and workshops. The main satellite data have been Synthetic Aperture Radar (SAR), optical and infrared images, radar altimeter data, passive microwave data and geoid data from GOCE. Results from MAIRES can contribute to improved understanding of climate change processes in the Arctic cryosphere. This information will be useful for scientists, policy-makers, operators, various other stakeholders and the general public. MAIRES has demonstrated the possibilities of combining Earth Observation data from European and Russian satellites for operational mapping, interpretation and forecast of land and sea ice variations in the Eurasian Arctic. Satellite-derived data collected over many years have shown seasonal and interannual variability of land ice, sea ice and icebergs. The benefit of using satellite for climate monitoring in the Arctic will increase in the comings as more data and longer time series of crysopheric parameters will be available. The results have been disseminated to users groups including climate research and operational users through many presentations at workshops, conferences and in scientific publications.
Project Context and Objectives:
Satellite Earth Observation (EO) data provides unique opportunities to studying Arctic climate processes and in particular changes in the cryosphere. There have never been more polar orbiting satellites in operation, providing EO data for environmental monitoring on global and regional scale. Use of EO data is necessary in several climate-related scientific disciplines, providing methods for environmental monitoring, support to marine operations, resource management and contribution to education. The user requirements for land and sea ice information in the Arctic region is growing as a result of climate change and its impact on the environment and human activities. The expected growth in ship traffic, oil and gas exploration, fisheries and tourism in the coming years will increase the risk of accidents affecting the environment, health, safety and economy of this unique and vulnerable region. The EU has started developing the Arctic policy with the environmental protection and sustainable use of resources considered as issues of the highest priority (EU Communication, 2008).
The objectives of MAIRES are
• to establish cooperation between ongoing GMES projects and Russian actors in the area of Arctic ice observation from space;
• to develop a method for precise overall modelling of glacier elevation changes by use of differential interferometry and altimetry data;
• to test and validate sea ice drift data derived from SAR images in combination with other ice satellite derived ice drift data
• to develop iceberg detection methods using a combination of high-resolution SAR and optical images;
• to document inter-annual and decadal changes in land and sea ice variables based on the EO-products developed in the project;
• to disseminate EO-based products for/of monitoring land and sea ice to users and stakeholders.
To achieve these objectives, a number of satellite-based products from European and Russian data are used supported by US and Canadian data. For landice studies, the following data are used: SAR from for interferometry, IceSat lidar altimetry, CryoSat-2 radar altimetry, GOCE gravity field data, high-resolution optical mages and Russian digital elevation data and topographic maps. For sea ice and icebergs, the main data sources are SAR from ENVISAT, RadarSat and TerraSAR, optical images from Landsat, MODIS, and several Russian satellites, passive microwave data for regional sea ice studies as well as in situ data for ice drift and ice thickness.
Project Results:
WP1: User requirements and case study definition
In the first six months of the project the consortium reviewed user requirements for land and sea ice data with emphasis on Russian Arctic regions (Fig. 1). The study area contains sea ice, icebergs and many glaciers that are well covered by satellite data over the last decades. Access to data from satellites, both from ESA, Russian Space Agency and other space agencies have been established, providing overview of data that can be used in the project. Furthermore, a set of case studies for land ice, sea ice and icebergs to be conducted in the project were planned. The case studies were identified based on previous and ongoing research work in the area and on availability of satellite data over several years. The user requirements for providing more data on the cryosphere in this area are growing because of shipping, offshore operations, and the general interest in climate change data. However, the possibility to provide new climate related products depends on the spatial and temporal coverage of satellite data. This topic has been investigated in WP2.
The user requirements have been reviewed from literature, other previous and ongoing projects and through direct contact with some users. A key document describing the requirements for cryospheric observations from space is the IGOS Cryosphere Theme report published in 2007 (http://igos-cryosphere.org.). Observational requirements for climate research are also defined by GCOS (Global Climate Observing System) in a series of documents (http://www.wmo.int/pages/prog/gcos/index.ph). At present, the ESA CCI programme runs a series of projects where the climate observation requirements are analyzed in detail (http://www.esa-cci.org/). The results of this analysis will be used in the project. For sea ice and icebergs, there are also many requirements from operational users such as Arctic shipping, offshore industry, weather and ice services, due to the growing human activities in the western Russian Arctic.
A user workshop was organized by NIERSC in St. Petersburg on 12 April 2013. The workshop was attended by 14 participants representing different institutes. The final user workshop was organized by NIERSC in May 2014.
WP2: Data acquisition
A data procurement plan was made at the beginning of the project for land ice, sea ice and icebergs, describing the types of data needed, time periods and areas of data coverage, access to data and quantities of data planned for use in the project. A main effort has been to establish systems for browsing, downloading and archiving satellite data for the study areas, based on data from ESA, Russian Space Agency (RKA) and other agencies. Data were collected and stored on several websites where ENVISAT ASAR, Radarsat Wideswath SAR, Landsat, ASTER, and others were made available and could be used for the project. Example of the web-based data search and retrieval system is shown in Fig. 2 (http://web.nersc.no/project/maires/catalog.php).
The Research center for Earth operative monitoring (NTs OMZ) under Russian Space Agency (Roscosmos) participated in the project as subcontractor to provide EO data from Russian satellites. The main tasks of NTsOMZ have been to retrieve and pre-process archived optical, IR and radar data that have been acquired by the Russian satellites since 1992 over the MAIRES study areas. Also new data from Russian operational EO satellites have been provided. All the selected data from the study areas were included in a database with access through a web portal web portal, making data available for users. This is shown in Fig. 3.
A specific task has been to find collocated data from Russian and European satellites as well as in situ data, which can be used to improve the research of land and sea ice processes. In order to manage these data, NERSC has established a database with web interface (Figure 2), which makes uses for the above web sites and facilitate for users to find data to be used in the case studies of the project. The database will continue to operate after the end of MAIRES where new data are included daily. As of 01 July 2014, the database contains several tens of thousands of images, mostly SAR images from ENVISAT and RADARSAT-2.
WP3: Land ice processes
A major achievement has been to generate a new suite of glacier change models for the largest glaciers in the Russian Arctic, including Vestfonna – Ahlmanfonna (2.500 km²) in Svalbard, Northern Ice Cap (2.200 km²) in Novaya Zemlya and Semenov-Tyanshaskiy – Leningradskiy Glacier Complex (2.600 km²) in Severnaya Zemlya and several smaller glaciers. Standard quality control procedures including a positional accuracy test and a content review were performed. The position of the equilibrium line (EQL) was precisely delineated with solid yellow or dashed red lines depending on the complexity of the change pattern, and the EQL altitude, multi-year orographic ELA (Equilibrium Line Altitude) was measured for each ice cap in a point-, section-, and polygon-based approach. Example of data flow from several types of satellite data used to calculate the mass balance of glaciers is shown in Fig. 4.
In the project methodological modernization of the dual-sensor INSARAL technique has been demonstrated in order to improve the operational mapping and measurement of glacier elevation changes and mass balance characteristics (ELA, AAR, ΔV, etc.) in semi-automatic mode. The method performance was tested using real EO data (lidar and radar altimetry, radar interferometry and gradiometry) on several ice caps, as shown in Fig. 4.
The overall mapping of glacier elevation changes and quantification of mass balance characteristics in the study region was performed by comparing reference elevation models of study glaciers derived from Russian topographic maps 1:200,000 (CI = 20 or 40 m) representing the glacier state as in the 1950s‐1960s with modern elevation data obtained from satellite radar interferometry and lidar altimetry. In total, 14 ERS and 4 TanDEM‐X high‐quality SAR interferograms of 1995/96 and 2011 were acquired, processed in the standard 2‐pass DINSAR manner, geocoded, calibrated, mosaicked and interpreted using reference elevation models and co‐located ICESat altimetry data of 2003‐2010. The DINSAR analysis revealed the existence of fast‐flowing outlet glaciers at Arthur, Rudolph, Eva‐Liv and Bennett islands. The calculation of separate mass‐balance components is complicated in this case because of generally unknown glacier velocities and ice discharge values for the mid‐20thcentury. Hence only net balance values were determined for those ice caps. Other ice caps belong to the category of slow‐moving or passive glaciers with simpler estimation of mass balance characteristics.
For quantification of glacier changes a precise measurement of glacier changes in linear, areal and volumetric terms was carried out at both local and regional level. The result revealed a reduction in glacier area and general lowering of the glacier surface on most ice caps. However, we also registered a dozen ice caps which are growing under the present climatic conditions in the region. The maximum growth with an elevation change rate of 0.6 m/a and a cumulative volume change of +35 km³ was detected at Northern Ice Cap in Novaya Zemlya (Fig. 5). Many other ice caps, such as Kvitøya, has shown a significant decrease in volume (Fig. 6).
Glacier elevation change and mass balance products were inter‐compared and controlled against the results from ICESat altimetry, field surveys, meteorological observations and other remote sensing techniques (Fig. 3). 30‐year long records of daily precipitation obtained from 38 coastal stations were involved in causality analysis. The output products can be accessed at http://dib.joanneum.at/MAIRES/index.php?page=products. They improve the understanding of glacier reaction to climate changes, local perturbations and associated effects. Practical use of the designed technique for the joint processing of satellite altimetry, interferometry and reference elevation data will further widen the range of CryoSat applications and will lead to more intensive use of satellite data stored in the ESA and third‐party archives.
In contrast to previous studies utilizing quick proxy estimates of glacier mass balance characteristics it is now possible to determine and deliver the accumulation area ratio, the outlines of accumulation-ablation areas and multi-year orographic ELA at different spatial levels with higher confidence. By analysing data for selected ice caps in the Russian Arctic, including Svalbard, it has been estimated that nearly 1,005±20 km³ (-4.5%) of the ice volume and 2,030±100 km² (-2%) of the ice area has been lost in the last 60 years (Table 1). The cumulative mass budget in the study region is negative while individual rates of volume change vary from -3 km³/a to +0.6 km³/a and the glacier change pattern commonly remains heterogeneous. The explanation of the main causes for this heterogeneity is not trivial and some insular ice caps seem to increase their volumes for no apparent reason, as in many cases such “anomalies” cannot be explained by local orographic, oceanographic and/or rheological conditions.
WP4: Sea ice processes
Sea ice occurs in a wide range of types and forms, and affects significantly and directly marine transport and navigation. Economic and social developments are engendering significant increases in international shipping, particularly in areas susceptible to sea ice. SAR image reflects the geometric and electromagnetic properties of the ice surface, because the backscatter from sea ice in C-band is dominated by the surface backscatter.
Sea ice classification gives an ice charts where the ice classes reflect the degree of deformation of an ice field. The primary problem is naturally to discriminate between open water and ice covered areas. The classification algorithm developed by the NERSC\NIERSC derives an ice map from an dual-polarization (HH+HV) RADARSAT-2 (RS2) SAR images with the following qualitative categories: open water, new thin ice (nilas) or calm open water, level ice/fast ice, and slightly ridged/rafted ice as ice.
RADARSAT-2 ScanSAR Wide mode at dual-polarization, i.e. HH (horizontally transmitted and horizontally received) and HV (horizontally transmitted, vertically received) assembles wide SAR image from several narrower SAR beams, resulting to an image of 500 × 500 km with 50 m resolution. The automated ice\water classification was based upon a technique originally used for same goal using ENVISAT ASAR images in frame of MyOcean project, and extended the use of the cross-polarization information available from RADARSAT-2.
SAR images contain much useful information about the ice conditions, especially in cold conditions. In open water the higher wind increases the contrast of open water area on different polarizations (HH and HV), that shows bright areas at HH against dark background for HV channel, and the water distinguish become more reliable. In the same case the higher wind reduces the contrast between open water and sea ice, which gives an ambiguity of these classes. In the interpretation of SAR images in addition to image gray levels or the backscatter values, texture provides important information. Detecting any typical textural structures for different sea ice types is extremely difficult due to the highly dynamic and variable nature of ice. Texture characteristics are computed in a moving window using a Gray Level Co-occurence Matrix (GLCM).
The RADARSAT-2 images were calibrated to sigma0 values and speckle and thermal-noise (for HV polarization data only) affects removed by extracting the background thermal noise level and smoothing the image with Gaussian filter. Land area was masked. We create, test and apply the Neural Network and the Support Vector Machines algorithms for delineation the ice\water on RADARSAT-2 images using texture as additional SAR information. The algorithm use backscatter values and several textural characteristics -energy, correlation, inertia (or contrast), entropy, homogeneity, entropy, cluster prominence, 3rd and 4th statistical moments, calculated for HH and HV channels, in order to identify sea ice and open water classes.
The Neural Network (NN) and Support Vector Machines (SVMs, also support vector networks) methods are supervised learning models with associated learning algorithms that analyze data and recognize patterns, used for classification and regression analysis. The both methods take a set of input data and predicts, for each given input, which of two possible classes forms the output. A NN and SVM are applied at the last step for processing array of the most informative texture characteristics and ice/water classification.
We have developed and demonstrated sea ice \ open water retrieval methods for exploitation of RADARSAT-2 data, where several SAR modes can be used (Fig. 7). Dual-polarized SAR images with 50 m resolution are used to produce high-resolution maps with ice edge, leads and open water areas (later also with ice classes). Available SAR data from RADARSAT-2 are used during the project. Validation of the ice information retrieved form the SAR data is done using Met.no ice charts, in situ observations and visual sea ice expert's analysis. The high-resolution products are provided in near-real time for areas required by the user (e.g. Fram Strait and Arctic Basin).
Validation of the ice classification is done by comparison with Norwegian Meteorological Institute (Met.no) ice charts and OW/sea ice SVMs classification. MET.no sea ice type data were ranged into 2 classes: OW and sea ice. SVMs classification: sea ice as sea ice; OW – OWcalm + OW rough. Met.NO ice charts are used as “apriory” (correct) classification and we find the accuracy of SVMs classification agreement with MET.no data.
The overall accuracies were estimated for all received Radarsat-2 images for the period from March – November 2013 (2778 images in total). The average values of total accuracies for each month are presented in Fig. 8. The errors in ice and water classification are also shown for each month. The data of the last table presented as a graph are shown in figure 6. The graph demonstrates that the accuracy is lower in summer months. The reason is that the classification algorithm was tuned using winter data. Validatio of the classification results can also be done by use of MODIS optical images on days where there are little clouds in the area. Optical images gives good discrimination between open water an d sea ice if the cloud cover is limited (Fig. 9).
A new approach of automatic sea ice drift retrieval have been tested over the Fram Strait region. The method based on original implementation of feature-based tracking technique and allows to retrieve more valid ice drift vectors compared to cross-correlation approach. We have started test calculations for winter and summer months from ASAR and Radarsat2 data. A geo-spatial database of in-situ data was prepared (IAPB dataset for 1979-2012) for validation. SAR images generated by NANSAT system . The ice drift algorithm uses anovel combination of feature-based tracking technique based on Harris-corner detector. Each feature point has a descriptor (multi-dimensional vector to describe a point) invariant to rotation, noise and scale. The algorithm is characterized by:
• wavelet responses in horizontal and vertical direction for a neighbourhood of size 6s for orientation assignment
• gaussian weights are applied
• the dominant orientation is estimated by calculating the sum of all responses within a sliding orientation window of angle 60 degrees’
• wavelet response can be found out using integral images very easily at any scale
• A neighbourhood of size 20sX20s is taken around the keypoint where s is the size
• It is divided into 4x4 subregions
• For each subregion, horizontal and vertical wavelet responses are taken and a vector is formed like this
• Represented as a vector gives Speeded Up Robust Features (SURF) descriptor with total 64 dimensions, (Bay, et al., 2006).
Example of ice drift retrievals from a pair of SAR images is shown in Fig. 10.
In order to validate the ice drift retrievals form the automated algorithm, manual analysis has been performed for an image pair in the Kara Sea, obtained on 01 and 03 February 2011. In Fig. 11 the green vectors are from the algorithm and the red vectors are from manual analysis. By comparing more than 100 manual with co-located algorithm-based vectors, the mean difference was found to be 0.1 km, corresponding to a RMSE error of 5 %.
The ice drift algorithm will be implemented for operational use in the Nansat system at the Nansen Center. With daily SAR coverage from RADARSAT-2 and Sentinel-1 the automated algorithms will be very important for processing ice classification and drift from many images. These products will be a contribution to the Copernicus marine services based on Sentinel-1 data.
In addition to algorithm development for SAR data and validation of these algorithms, extensive work has been carried to analyse long-term data on ice drift in the whole Arctic area. For climate studies it is necessary to analyse data covering several decades.
NIERSC has conducted statistical analysis of several ice drift data sets available for the last two decades, including data from the International Arctic Buoy Proramme, the satellite-based ice drift data from Ifremer based on passive microwave data and scatterometer data (Fig. 12), and additional data from SAR and optical data provided by the National Snow nd Ice Data Center (http://nsidc.org). The analysis the so-called vector-algebraic method, which has been developed and applied in Russia on time series of 2D vector data such a wind, currents and ice drift. The technique allows us to significantly compress the initial information and most adequately describe the vector time series of full-scale and model data restricted by a set of statistical characteristics in the invariant form. The work has focused on analysis of the seasonal and interannual variability of the sea ice drift conditions. The MAIRES project has contributed to the study of long-term variability of the drift fields for the entire Arctic Ocean, which has been carried out for the first time. The main idea of the study was to search for links between the dynamics of sea ice (drift velocity and variance, change of ice outflow, etc.) with the spatial structure of the ice drift fields, that can form conditions for growing or melting of sea ice in the Arctic.
Analysis of average seasonal maps of total variance and mean vectors for thirty years has shown substantial variability, but two main type of a spatial structure of the large-scale circulation of the ice were determined (Fig. 13). These two types coordinated with the classifications of Prof. Z.M. Gudkovich (1972). Type A is characterized by a pronounced both Anticyclonic Gyre and Transarctic current (the traditional “scheme), while type B is characterized by the almost complete absence of the Anticyclonic Gyre and very intensive Transarctic current that is creating an increased outflow of ice from the Arctic Basin.
In last three years (2011-2012-2013) there is a tendency to restore "traditional" structure of the field of ice drift - strengthening Anticyclonic Gyre (Fig. 14) on the background of increasing of the repeatability of synoptic processes of B-Group that probably can create conditions for a rehabilitation of the Arctic sea ice cover. The first results of this work is published by Volkov et al., in Journal of Operational Oceanography, Vol. 5, No. 2, August 2012, pp. 61-71.
In conclusion, it should be noted that in the period of the project implementation has established an integrated database, for the first time developed a methodology of analysis of vector fields, the analysis was made and the conclusions on the current state of ice processes in the Arctic basin formed.
Retrieval of ice thickness from satellites altimeters has become an important research topic in the last few years. The ESA CryoSat-2 satellite has provided altimeter data over sea ice since 2011 and significant efforts have been done to calculate ice thickness from these data. In MAIRES some work has been done to analyse in situ data, mainly the Russian Sever expeditions from the 1980’s and North Pole Drifting Station data from the last decade. These data includes measurements of ice thickness, ice freeboard and snow thickness data, as well as estimation of ice density (Fig. 15). All these data are needed for validation of ice thickness retrievals from satellite radar altimeter data.
WP5: Iceberg detection and monitoring
Icebergs are scattered pieces ice floating in the ocean originating from glaciers. Icebergs are found in large quantities both in the Arctic Ocean and in the Antarctica. Icebergs are often found in the same areas as sea ice, which is generated by freezing of the surface layer of the ocean. Icebergs and sea ice have highly different physical properties (size, shape, density, thickness) as well as distribution. Icebergs have small horizontal scale but are much thicker. In the Antarctica, the size of icebergs can become very large, several km, but in the Arctic the normal size of icebergs is typical between 20 and 100 m in horizontal extent. The small horizontal scale makes detection more difficult and the large thickness represent huge masses of ice, which is a threat to ships and operating platforms in polar regions.
Icebergs can be detected from high-resolution visible images from Landsat, Aster, “Monitor-E” and other satellites if there are no clouds and light conditions are favourable. The iceberg identification is difficult when their size is about the sensor resolution, and when their size corresponds to 2-3 pixels (60-90 м), they are identified using one of the following features:
• presence of shadow,
• light patch – light reflection from sunny part of iceberg,
• texture of large tabular icebergs,
• presence of fracture near the iceberg due to its motion relatively to drifting sea ice.
Weather conditions, first, cloudiness and sun height, which determine a length of iceberg shadow and contrast between sunny side of iceberg and its surroundings, influence manifestation of these features. Icebergs are better detected in early spring, when they are surrounded with level ice and have long shadow due to small sun angles. The valuable complementary information for their detection provide tracks in sea ice in the form of ellipse, caused by tidal currents or inertial movements.
Use of SAR data can improve iceberg detection compared to optical data since SAR data are independent of cloud and light conditions. SAR images with resolution better than 100 m can improve the detection of larger icebergs of size of 100 m or more. However, in the Barents Sea and Kara Sea area, many icebergs are smaller, typically 10 – 50 m, and this will require SAR images with resolution of about 10m, provided by for example TerraSAR-X, RADARSAR-2 and CosmoSkymed. An example of iceberg detection in SAR and optical images is shown in Fig. 19, and analysis result from one SAR image is shown in Fig. 20.
In this project, several hundred high-resolution optical and SAR images have been compiled and analyzed for the following regions: Svalbard, Franz Josef Land, Novaya Zemlya and Severnaya Zemlya. Examples of data coverage are shown in Fig. 21).
Analysis of images was done by use of ARC GIS to read and geolocate several images, perform simple enhancement (contrast enhancement) and identify icebergs by visual interpretations. When images were overlapped the iceberg is demarcated only one time from the first image since most of the icebergs are embedded in fast ice, which means that they do not move. A raster database containing the images in tiff format was created in the Arc GIS. Fig 22 shows a zoom-in part of one ASTER image in the eastern part Severnaya Zemlya where many icebergs were found.
For Franz Josef Land, a total of 2600 icebergs were detected and demarcated from the ASTER images (Fig. 23). The icebergs were distributed over the whole area and they are in difference sizes. The smallest detectable icebergs were about 15 m, corresponding to the pixel size of the images.
From the database where each iceberg is registered, the distribution of icebergs is calculated and presented as a density (Fig. 24). The highest density is around Hoffmann Island (there are no main calving areas here but maybe they were drifting from nearby calving glaciers.
For Severnaya Zemlya a total number of 80 images were analyzed and 4470 icebergs were detected and demarcated. Most of the icebergs are detected in the east part of islands (Laptev Sea) and Krasnoi Armii Strait, where the out let glaciers calving, that maybe due to the fact there few (5 to 6) slow-moving outlet glaciers flow towards the Kara Sea and 26 fast-moving outlets reach the Laptev Sea (Fig. 25). The main source is Matusevich Ice Shelf which is the largest remaining ice shelf in the Russian high arctic. The iceberg density map prepared from the database where all icebergs were registered (Fig. 26 a) has been compared with historical Russian iceberg maps (Fig. 26b).
WP6 Integrated presentation
For land ice studies, integration of data from several sources is an essential part of producing maps of glaciers, ice caps and ice sheets. Precise maps depicting long‐term elevation changes of land ice masses and seasonal changes of the sea ice cover represent an important instrument and indispensable basis for assessing Earth’s cryosphere and obtaining quantitative data on the extent, accumulation, ablation, movement, calving flux and mass balance of ice bodies in documentary and human perceivable form.
New extensive, albeit detailed remote sensing studies devoted to overall glacier change mapping and regional estimates of geodetic mass balance in the Eurasian Insular Arctic were carried out in the MAIRES frameworks using a synergetic combination of satellite altimetry and interferometry. Apart from high sensitivity to changes in glacier topography and independence of natural illumination, themajor advantage of combining radar interferometry and altimetry, referred to as dual‐sensor INSARAL technique, is the enhanced glacier‐wide coverage with elevation change data and the high precision of elevation measurements achieved even in the case of insufficient ground control typical of glacial areas. Example of multisatellite data for landice studies are shown in Fig.27. This is important for the reliable modelling of topographic changes in glacier accumulation areas characterized by relatively sparse coverage with altimetric transects and corresponding underestimation of the accumulation signal by simplified mono‐sensor techniques, such as those offered by Moholdt et al. (2010). Our 41 resultant maps of uniform quality cover the entire Eurasian Arctic Basin from Svalbard in the west to Wrangel Island in the east. Web versions of ice change maps accompanied with meta‐data are brought together in the “Online Atlas of Ice Cover Fluctuations in the Eurasian Arctic” (http://dib.joanneum.at/MAIRES/index.php?page=atlas).
The synergetic combination of CryoSat and ICESat altimetry with differential ERS and TanDEM‐X radar interferometry provided technically efficient geodetic solution to measuring and mapping elevation changes in accumulation areas and determining the altitude of multi‐year equilibrium line on large Eurasian ice caps. Unprecedented series of glacier change products continuously representing elevation changes in linear, areal and volumetric terms, showing ice divides and their migration, delineating the equilibrium line position and ELA trends and estimating mass balance characteristics in accumulation areas of large Eurasian ice caps for the period of 1950s – 2010s were generated, interpreted and validated. Example of integrated information presented in a map is shown in Fig. 28.
Sea ice is an essential climate component which is very sensitive to climate change. In the Arctic, sea ice area and thickness have been reduced during the last three decades affecting the heat flux causing enhanced warming in this region. The reduction of the Arctic sea ice cover in the last decades can only be documented by regular monitoring from satellite passive microwave data. Sea ice observation from satellites has been carried out for more than four decades and is one of the most important applications of satellite data in climate change studies. Regular monitoring of sea ice with passive microwave data has been done continuously since 1979 and these data therefore represent one of the longest satellite data records on climate variables. The sea ice extent in the Arctic is shown in Fig. 29, where the last extreme minimum was observed in September 2012. Several sensors and retrieval methods have been developed and successfully utilized to measure various sea ice variables such as concentration, thickness, motion, type, albedo, snow cover, surface temperature, duration of melt season, leads/polynyas and ridges. Remote sensing can contribute to retrieve quantitative measurements of most of these parameters, but data from other observing systems are needed to assess the accuracy of the satellite observations. In particular for ice thickness observation, it is necessary to collect data from other platforms and instruments such as airborne surveys, in situ measurements and upward-looking sonars from submarine cruises, anchored moorings and autonomous underwater vehicles. Attention has to be paid to the different scales involved, as ice properties often vary over several scales which cannot be covered with one instrument or one type of measurement, e.g. air-borne, alone. This applies in particular to the validation of satellite products.
Another sea ice climate variable is the ice drift flux through the Fram Strait. By analyzing ice drift vectors obtained from series of SAR images the ice area flux can be calculated and the cumulative flux over an ice season can be estimated. Analysis of 10 years of ice drift data are shown in Fig. 31.
In MAIRES studies have focused on algorithm development and validation of satellite derived products such as ice classification, ice drift and ice thickness. All these parameters will benefit from integrating data from several satellite sensors.
Potential Impact:
The main results from the MAIRES project are the following:
An on-line database has been created at NERSC and populated with data from European, Russian and other satellite data covering the study area in the western Russian Arctic. The database contains mostly Synthetic Aperture Radar images from ENVISAT and RADARSAT-2 and high-resolution optical images from ASTER, Landsat, MODIS and Russian satellites. These data have been used as input to the sea ice and iceberg studies in the project. The database is updated daily with new satellite dada form the Arctic and will be maintained after the end of MAIRES project, as a contribution to Copernicus marine services.
An atlas of landice products has been created by Joanneum Reserch and made available on a website, where the results all the glacier change analyzed are made available. The analysis shows that most of the glaciers have lost mass over the last 5 – 6 decades, while some have increased their mass.
For seas ice studies, new algorithms for estimation of ice classification and ice drift retrieval from SAR data have been developed and validated. These products will be used after MAIRES to develop operational sea ice services from Sentinel-1. Furthermore, ice drift data collected from several satellite systems have been analyzed for the whole Arctic, showing that the ice drift pattern varies from year to year, depending on the atmospheric wind systems in the Arctic.
For icebergs studies, intensive data collection and analysis of high-resolution optical images (ASTER images) was carried for 2012, for the Franz Josef Land, Novaya Zemlya and Severnaya Zemlya areas, where most of the iceberg in the Russian sector of the Arctic are generated. The analysis of iceberg density in 2012 was compared with Russian iceberg Atlas, showing iceberg distribution from aircraft surveys in the 1970 – 1980s. The iceberg distribution varies from year to year and it is not possible to assess any trend without more systematic data collection over many years.
The studies of land ice sea ice and icebergs conducted in MAIRES are all contributions to understand cryospheric process in the Arctic and how climate change has impact on the cryosphere. The Arctic has experienced a significant increase in surface air temperature over the last 3 – 4 decades (Fig. 32). This increase is connected to the reduction in both land ice masses and sea ice, although there are significant regional and interannual variabilities.
Satellite observations are generally obtained over relatively short-term periods for climate research. Except for sea ice extent, which has been obtained systematically for more than 35 years, satellite data records are rarely longer than a decade, but they can provide very detailed information about seasonal and regional processes. Such processes are important to understand and quantify in order to improve climate models and climate predictions. It is envisaged that satellite data will play an increasingly important role in cryospheric research and climate studies in the coming years.
List of Websites:
http://maires.nersc.no
The objective of the MAIRES proposal is to develop methodologies for satellite monitoring of Arctic glaciers, sea ice and icebergs. A major effort has been to compile satellite data from ESA, Russian Space Agency, and other satellite systems and establish data bases where satellite data can be searched, collocated and downloaded for further use. Methodologies have been developed to retrieve quantitative information on land ice, sea ice and icebergs from various satellite data, supplemented by in situ and other data. Examples of satellite derived products for each of the three thematic areas have been presented at conferences and workshops. The main satellite data have been Synthetic Aperture Radar (SAR), optical and infrared images, radar altimeter data, passive microwave data and geoid data from GOCE. Results from MAIRES can contribute to improved understanding of climate change processes in the Arctic cryosphere. This information will be useful for scientists, policy-makers, operators, various other stakeholders and the general public. MAIRES has demonstrated the possibilities of combining Earth Observation data from European and Russian satellites for operational mapping, interpretation and forecast of land and sea ice variations in the Eurasian Arctic. Satellite-derived data collected over many years have shown seasonal and interannual variability of land ice, sea ice and icebergs. The benefit of using satellite for climate monitoring in the Arctic will increase in the comings as more data and longer time series of crysopheric parameters will be available. The results have been disseminated to users groups including climate research and operational users through many presentations at workshops, conferences and in scientific publications.
Project Context and Objectives:
Satellite Earth Observation (EO) data provides unique opportunities to studying Arctic climate processes and in particular changes in the cryosphere. There have never been more polar orbiting satellites in operation, providing EO data for environmental monitoring on global and regional scale. Use of EO data is necessary in several climate-related scientific disciplines, providing methods for environmental monitoring, support to marine operations, resource management and contribution to education. The user requirements for land and sea ice information in the Arctic region is growing as a result of climate change and its impact on the environment and human activities. The expected growth in ship traffic, oil and gas exploration, fisheries and tourism in the coming years will increase the risk of accidents affecting the environment, health, safety and economy of this unique and vulnerable region. The EU has started developing the Arctic policy with the environmental protection and sustainable use of resources considered as issues of the highest priority (EU Communication, 2008).
The objectives of MAIRES are
• to establish cooperation between ongoing GMES projects and Russian actors in the area of Arctic ice observation from space;
• to develop a method for precise overall modelling of glacier elevation changes by use of differential interferometry and altimetry data;
• to test and validate sea ice drift data derived from SAR images in combination with other ice satellite derived ice drift data
• to develop iceberg detection methods using a combination of high-resolution SAR and optical images;
• to document inter-annual and decadal changes in land and sea ice variables based on the EO-products developed in the project;
• to disseminate EO-based products for/of monitoring land and sea ice to users and stakeholders.
To achieve these objectives, a number of satellite-based products from European and Russian data are used supported by US and Canadian data. For landice studies, the following data are used: SAR from for interferometry, IceSat lidar altimetry, CryoSat-2 radar altimetry, GOCE gravity field data, high-resolution optical mages and Russian digital elevation data and topographic maps. For sea ice and icebergs, the main data sources are SAR from ENVISAT, RadarSat and TerraSAR, optical images from Landsat, MODIS, and several Russian satellites, passive microwave data for regional sea ice studies as well as in situ data for ice drift and ice thickness.
Project Results:
WP1: User requirements and case study definition
In the first six months of the project the consortium reviewed user requirements for land and sea ice data with emphasis on Russian Arctic regions (Fig. 1). The study area contains sea ice, icebergs and many glaciers that are well covered by satellite data over the last decades. Access to data from satellites, both from ESA, Russian Space Agency and other space agencies have been established, providing overview of data that can be used in the project. Furthermore, a set of case studies for land ice, sea ice and icebergs to be conducted in the project were planned. The case studies were identified based on previous and ongoing research work in the area and on availability of satellite data over several years. The user requirements for providing more data on the cryosphere in this area are growing because of shipping, offshore operations, and the general interest in climate change data. However, the possibility to provide new climate related products depends on the spatial and temporal coverage of satellite data. This topic has been investigated in WP2.
The user requirements have been reviewed from literature, other previous and ongoing projects and through direct contact with some users. A key document describing the requirements for cryospheric observations from space is the IGOS Cryosphere Theme report published in 2007 (http://igos-cryosphere.org.). Observational requirements for climate research are also defined by GCOS (Global Climate Observing System) in a series of documents (http://www.wmo.int/pages/prog/gcos/index.ph). At present, the ESA CCI programme runs a series of projects where the climate observation requirements are analyzed in detail (http://www.esa-cci.org/). The results of this analysis will be used in the project. For sea ice and icebergs, there are also many requirements from operational users such as Arctic shipping, offshore industry, weather and ice services, due to the growing human activities in the western Russian Arctic.
A user workshop was organized by NIERSC in St. Petersburg on 12 April 2013. The workshop was attended by 14 participants representing different institutes. The final user workshop was organized by NIERSC in May 2014.
WP2: Data acquisition
A data procurement plan was made at the beginning of the project for land ice, sea ice and icebergs, describing the types of data needed, time periods and areas of data coverage, access to data and quantities of data planned for use in the project. A main effort has been to establish systems for browsing, downloading and archiving satellite data for the study areas, based on data from ESA, Russian Space Agency (RKA) and other agencies. Data were collected and stored on several websites where ENVISAT ASAR, Radarsat Wideswath SAR, Landsat, ASTER, and others were made available and could be used for the project. Example of the web-based data search and retrieval system is shown in Fig. 2 (http://web.nersc.no/project/maires/catalog.php).
The Research center for Earth operative monitoring (NTs OMZ) under Russian Space Agency (Roscosmos) participated in the project as subcontractor to provide EO data from Russian satellites. The main tasks of NTsOMZ have been to retrieve and pre-process archived optical, IR and radar data that have been acquired by the Russian satellites since 1992 over the MAIRES study areas. Also new data from Russian operational EO satellites have been provided. All the selected data from the study areas were included in a database with access through a web portal web portal, making data available for users. This is shown in Fig. 3.
A specific task has been to find collocated data from Russian and European satellites as well as in situ data, which can be used to improve the research of land and sea ice processes. In order to manage these data, NERSC has established a database with web interface (Figure 2), which makes uses for the above web sites and facilitate for users to find data to be used in the case studies of the project. The database will continue to operate after the end of MAIRES where new data are included daily. As of 01 July 2014, the database contains several tens of thousands of images, mostly SAR images from ENVISAT and RADARSAT-2.
WP3: Land ice processes
A major achievement has been to generate a new suite of glacier change models for the largest glaciers in the Russian Arctic, including Vestfonna – Ahlmanfonna (2.500 km²) in Svalbard, Northern Ice Cap (2.200 km²) in Novaya Zemlya and Semenov-Tyanshaskiy – Leningradskiy Glacier Complex (2.600 km²) in Severnaya Zemlya and several smaller glaciers. Standard quality control procedures including a positional accuracy test and a content review were performed. The position of the equilibrium line (EQL) was precisely delineated with solid yellow or dashed red lines depending on the complexity of the change pattern, and the EQL altitude, multi-year orographic ELA (Equilibrium Line Altitude) was measured for each ice cap in a point-, section-, and polygon-based approach. Example of data flow from several types of satellite data used to calculate the mass balance of glaciers is shown in Fig. 4.
In the project methodological modernization of the dual-sensor INSARAL technique has been demonstrated in order to improve the operational mapping and measurement of glacier elevation changes and mass balance characteristics (ELA, AAR, ΔV, etc.) in semi-automatic mode. The method performance was tested using real EO data (lidar and radar altimetry, radar interferometry and gradiometry) on several ice caps, as shown in Fig. 4.
The overall mapping of glacier elevation changes and quantification of mass balance characteristics in the study region was performed by comparing reference elevation models of study glaciers derived from Russian topographic maps 1:200,000 (CI = 20 or 40 m) representing the glacier state as in the 1950s‐1960s with modern elevation data obtained from satellite radar interferometry and lidar altimetry. In total, 14 ERS and 4 TanDEM‐X high‐quality SAR interferograms of 1995/96 and 2011 were acquired, processed in the standard 2‐pass DINSAR manner, geocoded, calibrated, mosaicked and interpreted using reference elevation models and co‐located ICESat altimetry data of 2003‐2010. The DINSAR analysis revealed the existence of fast‐flowing outlet glaciers at Arthur, Rudolph, Eva‐Liv and Bennett islands. The calculation of separate mass‐balance components is complicated in this case because of generally unknown glacier velocities and ice discharge values for the mid‐20thcentury. Hence only net balance values were determined for those ice caps. Other ice caps belong to the category of slow‐moving or passive glaciers with simpler estimation of mass balance characteristics.
For quantification of glacier changes a precise measurement of glacier changes in linear, areal and volumetric terms was carried out at both local and regional level. The result revealed a reduction in glacier area and general lowering of the glacier surface on most ice caps. However, we also registered a dozen ice caps which are growing under the present climatic conditions in the region. The maximum growth with an elevation change rate of 0.6 m/a and a cumulative volume change of +35 km³ was detected at Northern Ice Cap in Novaya Zemlya (Fig. 5). Many other ice caps, such as Kvitøya, has shown a significant decrease in volume (Fig. 6).
Glacier elevation change and mass balance products were inter‐compared and controlled against the results from ICESat altimetry, field surveys, meteorological observations and other remote sensing techniques (Fig. 3). 30‐year long records of daily precipitation obtained from 38 coastal stations were involved in causality analysis. The output products can be accessed at http://dib.joanneum.at/MAIRES/index.php?page=products. They improve the understanding of glacier reaction to climate changes, local perturbations and associated effects. Practical use of the designed technique for the joint processing of satellite altimetry, interferometry and reference elevation data will further widen the range of CryoSat applications and will lead to more intensive use of satellite data stored in the ESA and third‐party archives.
In contrast to previous studies utilizing quick proxy estimates of glacier mass balance characteristics it is now possible to determine and deliver the accumulation area ratio, the outlines of accumulation-ablation areas and multi-year orographic ELA at different spatial levels with higher confidence. By analysing data for selected ice caps in the Russian Arctic, including Svalbard, it has been estimated that nearly 1,005±20 km³ (-4.5%) of the ice volume and 2,030±100 km² (-2%) of the ice area has been lost in the last 60 years (Table 1). The cumulative mass budget in the study region is negative while individual rates of volume change vary from -3 km³/a to +0.6 km³/a and the glacier change pattern commonly remains heterogeneous. The explanation of the main causes for this heterogeneity is not trivial and some insular ice caps seem to increase their volumes for no apparent reason, as in many cases such “anomalies” cannot be explained by local orographic, oceanographic and/or rheological conditions.
WP4: Sea ice processes
Sea ice occurs in a wide range of types and forms, and affects significantly and directly marine transport and navigation. Economic and social developments are engendering significant increases in international shipping, particularly in areas susceptible to sea ice. SAR image reflects the geometric and electromagnetic properties of the ice surface, because the backscatter from sea ice in C-band is dominated by the surface backscatter.
Sea ice classification gives an ice charts where the ice classes reflect the degree of deformation of an ice field. The primary problem is naturally to discriminate between open water and ice covered areas. The classification algorithm developed by the NERSC\NIERSC derives an ice map from an dual-polarization (HH+HV) RADARSAT-2 (RS2) SAR images with the following qualitative categories: open water, new thin ice (nilas) or calm open water, level ice/fast ice, and slightly ridged/rafted ice as ice.
RADARSAT-2 ScanSAR Wide mode at dual-polarization, i.e. HH (horizontally transmitted and horizontally received) and HV (horizontally transmitted, vertically received) assembles wide SAR image from several narrower SAR beams, resulting to an image of 500 × 500 km with 50 m resolution. The automated ice\water classification was based upon a technique originally used for same goal using ENVISAT ASAR images in frame of MyOcean project, and extended the use of the cross-polarization information available from RADARSAT-2.
SAR images contain much useful information about the ice conditions, especially in cold conditions. In open water the higher wind increases the contrast of open water area on different polarizations (HH and HV), that shows bright areas at HH against dark background for HV channel, and the water distinguish become more reliable. In the same case the higher wind reduces the contrast between open water and sea ice, which gives an ambiguity of these classes. In the interpretation of SAR images in addition to image gray levels or the backscatter values, texture provides important information. Detecting any typical textural structures for different sea ice types is extremely difficult due to the highly dynamic and variable nature of ice. Texture characteristics are computed in a moving window using a Gray Level Co-occurence Matrix (GLCM).
The RADARSAT-2 images were calibrated to sigma0 values and speckle and thermal-noise (for HV polarization data only) affects removed by extracting the background thermal noise level and smoothing the image with Gaussian filter. Land area was masked. We create, test and apply the Neural Network and the Support Vector Machines algorithms for delineation the ice\water on RADARSAT-2 images using texture as additional SAR information. The algorithm use backscatter values and several textural characteristics -energy, correlation, inertia (or contrast), entropy, homogeneity, entropy, cluster prominence, 3rd and 4th statistical moments, calculated for HH and HV channels, in order to identify sea ice and open water classes.
The Neural Network (NN) and Support Vector Machines (SVMs, also support vector networks) methods are supervised learning models with associated learning algorithms that analyze data and recognize patterns, used for classification and regression analysis. The both methods take a set of input data and predicts, for each given input, which of two possible classes forms the output. A NN and SVM are applied at the last step for processing array of the most informative texture characteristics and ice/water classification.
We have developed and demonstrated sea ice \ open water retrieval methods for exploitation of RADARSAT-2 data, where several SAR modes can be used (Fig. 7). Dual-polarized SAR images with 50 m resolution are used to produce high-resolution maps with ice edge, leads and open water areas (later also with ice classes). Available SAR data from RADARSAT-2 are used during the project. Validation of the ice information retrieved form the SAR data is done using Met.no ice charts, in situ observations and visual sea ice expert's analysis. The high-resolution products are provided in near-real time for areas required by the user (e.g. Fram Strait and Arctic Basin).
Validation of the ice classification is done by comparison with Norwegian Meteorological Institute (Met.no) ice charts and OW/sea ice SVMs classification. MET.no sea ice type data were ranged into 2 classes: OW and sea ice. SVMs classification: sea ice as sea ice; OW – OWcalm + OW rough. Met.NO ice charts are used as “apriory” (correct) classification and we find the accuracy of SVMs classification agreement with MET.no data.
The overall accuracies were estimated for all received Radarsat-2 images for the period from March – November 2013 (2778 images in total). The average values of total accuracies for each month are presented in Fig. 8. The errors in ice and water classification are also shown for each month. The data of the last table presented as a graph are shown in figure 6. The graph demonstrates that the accuracy is lower in summer months. The reason is that the classification algorithm was tuned using winter data. Validatio of the classification results can also be done by use of MODIS optical images on days where there are little clouds in the area. Optical images gives good discrimination between open water an d sea ice if the cloud cover is limited (Fig. 9).
A new approach of automatic sea ice drift retrieval have been tested over the Fram Strait region. The method based on original implementation of feature-based tracking technique and allows to retrieve more valid ice drift vectors compared to cross-correlation approach. We have started test calculations for winter and summer months from ASAR and Radarsat2 data. A geo-spatial database of in-situ data was prepared (IAPB dataset for 1979-2012) for validation. SAR images generated by NANSAT system . The ice drift algorithm uses anovel combination of feature-based tracking technique based on Harris-corner detector. Each feature point has a descriptor (multi-dimensional vector to describe a point) invariant to rotation, noise and scale. The algorithm is characterized by:
• wavelet responses in horizontal and vertical direction for a neighbourhood of size 6s for orientation assignment
• gaussian weights are applied
• the dominant orientation is estimated by calculating the sum of all responses within a sliding orientation window of angle 60 degrees’
• wavelet response can be found out using integral images very easily at any scale
• A neighbourhood of size 20sX20s is taken around the keypoint where s is the size
• It is divided into 4x4 subregions
• For each subregion, horizontal and vertical wavelet responses are taken and a vector is formed like this
• Represented as a vector gives Speeded Up Robust Features (SURF) descriptor with total 64 dimensions, (Bay, et al., 2006).
Example of ice drift retrievals from a pair of SAR images is shown in Fig. 10.
In order to validate the ice drift retrievals form the automated algorithm, manual analysis has been performed for an image pair in the Kara Sea, obtained on 01 and 03 February 2011. In Fig. 11 the green vectors are from the algorithm and the red vectors are from manual analysis. By comparing more than 100 manual with co-located algorithm-based vectors, the mean difference was found to be 0.1 km, corresponding to a RMSE error of 5 %.
The ice drift algorithm will be implemented for operational use in the Nansat system at the Nansen Center. With daily SAR coverage from RADARSAT-2 and Sentinel-1 the automated algorithms will be very important for processing ice classification and drift from many images. These products will be a contribution to the Copernicus marine services based on Sentinel-1 data.
In addition to algorithm development for SAR data and validation of these algorithms, extensive work has been carried to analyse long-term data on ice drift in the whole Arctic area. For climate studies it is necessary to analyse data covering several decades.
NIERSC has conducted statistical analysis of several ice drift data sets available for the last two decades, including data from the International Arctic Buoy Proramme, the satellite-based ice drift data from Ifremer based on passive microwave data and scatterometer data (Fig. 12), and additional data from SAR and optical data provided by the National Snow nd Ice Data Center (http://nsidc.org). The analysis the so-called vector-algebraic method, which has been developed and applied in Russia on time series of 2D vector data such a wind, currents and ice drift. The technique allows us to significantly compress the initial information and most adequately describe the vector time series of full-scale and model data restricted by a set of statistical characteristics in the invariant form. The work has focused on analysis of the seasonal and interannual variability of the sea ice drift conditions. The MAIRES project has contributed to the study of long-term variability of the drift fields for the entire Arctic Ocean, which has been carried out for the first time. The main idea of the study was to search for links between the dynamics of sea ice (drift velocity and variance, change of ice outflow, etc.) with the spatial structure of the ice drift fields, that can form conditions for growing or melting of sea ice in the Arctic.
Analysis of average seasonal maps of total variance and mean vectors for thirty years has shown substantial variability, but two main type of a spatial structure of the large-scale circulation of the ice were determined (Fig. 13). These two types coordinated with the classifications of Prof. Z.M. Gudkovich (1972). Type A is characterized by a pronounced both Anticyclonic Gyre and Transarctic current (the traditional “scheme), while type B is characterized by the almost complete absence of the Anticyclonic Gyre and very intensive Transarctic current that is creating an increased outflow of ice from the Arctic Basin.
In last three years (2011-2012-2013) there is a tendency to restore "traditional" structure of the field of ice drift - strengthening Anticyclonic Gyre (Fig. 14) on the background of increasing of the repeatability of synoptic processes of B-Group that probably can create conditions for a rehabilitation of the Arctic sea ice cover. The first results of this work is published by Volkov et al., in Journal of Operational Oceanography, Vol. 5, No. 2, August 2012, pp. 61-71.
In conclusion, it should be noted that in the period of the project implementation has established an integrated database, for the first time developed a methodology of analysis of vector fields, the analysis was made and the conclusions on the current state of ice processes in the Arctic basin formed.
Retrieval of ice thickness from satellites altimeters has become an important research topic in the last few years. The ESA CryoSat-2 satellite has provided altimeter data over sea ice since 2011 and significant efforts have been done to calculate ice thickness from these data. In MAIRES some work has been done to analyse in situ data, mainly the Russian Sever expeditions from the 1980’s and North Pole Drifting Station data from the last decade. These data includes measurements of ice thickness, ice freeboard and snow thickness data, as well as estimation of ice density (Fig. 15). All these data are needed for validation of ice thickness retrievals from satellite radar altimeter data.
WP5: Iceberg detection and monitoring
Icebergs are scattered pieces ice floating in the ocean originating from glaciers. Icebergs are found in large quantities both in the Arctic Ocean and in the Antarctica. Icebergs are often found in the same areas as sea ice, which is generated by freezing of the surface layer of the ocean. Icebergs and sea ice have highly different physical properties (size, shape, density, thickness) as well as distribution. Icebergs have small horizontal scale but are much thicker. In the Antarctica, the size of icebergs can become very large, several km, but in the Arctic the normal size of icebergs is typical between 20 and 100 m in horizontal extent. The small horizontal scale makes detection more difficult and the large thickness represent huge masses of ice, which is a threat to ships and operating platforms in polar regions.
Icebergs can be detected from high-resolution visible images from Landsat, Aster, “Monitor-E” and other satellites if there are no clouds and light conditions are favourable. The iceberg identification is difficult when their size is about the sensor resolution, and when their size corresponds to 2-3 pixels (60-90 м), they are identified using one of the following features:
• presence of shadow,
• light patch – light reflection from sunny part of iceberg,
• texture of large tabular icebergs,
• presence of fracture near the iceberg due to its motion relatively to drifting sea ice.
Weather conditions, first, cloudiness and sun height, which determine a length of iceberg shadow and contrast between sunny side of iceberg and its surroundings, influence manifestation of these features. Icebergs are better detected in early spring, when they are surrounded with level ice and have long shadow due to small sun angles. The valuable complementary information for their detection provide tracks in sea ice in the form of ellipse, caused by tidal currents or inertial movements.
Use of SAR data can improve iceberg detection compared to optical data since SAR data are independent of cloud and light conditions. SAR images with resolution better than 100 m can improve the detection of larger icebergs of size of 100 m or more. However, in the Barents Sea and Kara Sea area, many icebergs are smaller, typically 10 – 50 m, and this will require SAR images with resolution of about 10m, provided by for example TerraSAR-X, RADARSAR-2 and CosmoSkymed. An example of iceberg detection in SAR and optical images is shown in Fig. 19, and analysis result from one SAR image is shown in Fig. 20.
In this project, several hundred high-resolution optical and SAR images have been compiled and analyzed for the following regions: Svalbard, Franz Josef Land, Novaya Zemlya and Severnaya Zemlya. Examples of data coverage are shown in Fig. 21).
Analysis of images was done by use of ARC GIS to read and geolocate several images, perform simple enhancement (contrast enhancement) and identify icebergs by visual interpretations. When images were overlapped the iceberg is demarcated only one time from the first image since most of the icebergs are embedded in fast ice, which means that they do not move. A raster database containing the images in tiff format was created in the Arc GIS. Fig 22 shows a zoom-in part of one ASTER image in the eastern part Severnaya Zemlya where many icebergs were found.
For Franz Josef Land, a total of 2600 icebergs were detected and demarcated from the ASTER images (Fig. 23). The icebergs were distributed over the whole area and they are in difference sizes. The smallest detectable icebergs were about 15 m, corresponding to the pixel size of the images.
From the database where each iceberg is registered, the distribution of icebergs is calculated and presented as a density (Fig. 24). The highest density is around Hoffmann Island (there are no main calving areas here but maybe they were drifting from nearby calving glaciers.
For Severnaya Zemlya a total number of 80 images were analyzed and 4470 icebergs were detected and demarcated. Most of the icebergs are detected in the east part of islands (Laptev Sea) and Krasnoi Armii Strait, where the out let glaciers calving, that maybe due to the fact there few (5 to 6) slow-moving outlet glaciers flow towards the Kara Sea and 26 fast-moving outlets reach the Laptev Sea (Fig. 25). The main source is Matusevich Ice Shelf which is the largest remaining ice shelf in the Russian high arctic. The iceberg density map prepared from the database where all icebergs were registered (Fig. 26 a) has been compared with historical Russian iceberg maps (Fig. 26b).
WP6 Integrated presentation
For land ice studies, integration of data from several sources is an essential part of producing maps of glaciers, ice caps and ice sheets. Precise maps depicting long‐term elevation changes of land ice masses and seasonal changes of the sea ice cover represent an important instrument and indispensable basis for assessing Earth’s cryosphere and obtaining quantitative data on the extent, accumulation, ablation, movement, calving flux and mass balance of ice bodies in documentary and human perceivable form.
New extensive, albeit detailed remote sensing studies devoted to overall glacier change mapping and regional estimates of geodetic mass balance in the Eurasian Insular Arctic were carried out in the MAIRES frameworks using a synergetic combination of satellite altimetry and interferometry. Apart from high sensitivity to changes in glacier topography and independence of natural illumination, themajor advantage of combining radar interferometry and altimetry, referred to as dual‐sensor INSARAL technique, is the enhanced glacier‐wide coverage with elevation change data and the high precision of elevation measurements achieved even in the case of insufficient ground control typical of glacial areas. Example of multisatellite data for landice studies are shown in Fig.27. This is important for the reliable modelling of topographic changes in glacier accumulation areas characterized by relatively sparse coverage with altimetric transects and corresponding underestimation of the accumulation signal by simplified mono‐sensor techniques, such as those offered by Moholdt et al. (2010). Our 41 resultant maps of uniform quality cover the entire Eurasian Arctic Basin from Svalbard in the west to Wrangel Island in the east. Web versions of ice change maps accompanied with meta‐data are brought together in the “Online Atlas of Ice Cover Fluctuations in the Eurasian Arctic” (http://dib.joanneum.at/MAIRES/index.php?page=atlas).
The synergetic combination of CryoSat and ICESat altimetry with differential ERS and TanDEM‐X radar interferometry provided technically efficient geodetic solution to measuring and mapping elevation changes in accumulation areas and determining the altitude of multi‐year equilibrium line on large Eurasian ice caps. Unprecedented series of glacier change products continuously representing elevation changes in linear, areal and volumetric terms, showing ice divides and their migration, delineating the equilibrium line position and ELA trends and estimating mass balance characteristics in accumulation areas of large Eurasian ice caps for the period of 1950s – 2010s were generated, interpreted and validated. Example of integrated information presented in a map is shown in Fig. 28.
Sea ice is an essential climate component which is very sensitive to climate change. In the Arctic, sea ice area and thickness have been reduced during the last three decades affecting the heat flux causing enhanced warming in this region. The reduction of the Arctic sea ice cover in the last decades can only be documented by regular monitoring from satellite passive microwave data. Sea ice observation from satellites has been carried out for more than four decades and is one of the most important applications of satellite data in climate change studies. Regular monitoring of sea ice with passive microwave data has been done continuously since 1979 and these data therefore represent one of the longest satellite data records on climate variables. The sea ice extent in the Arctic is shown in Fig. 29, where the last extreme minimum was observed in September 2012. Several sensors and retrieval methods have been developed and successfully utilized to measure various sea ice variables such as concentration, thickness, motion, type, albedo, snow cover, surface temperature, duration of melt season, leads/polynyas and ridges. Remote sensing can contribute to retrieve quantitative measurements of most of these parameters, but data from other observing systems are needed to assess the accuracy of the satellite observations. In particular for ice thickness observation, it is necessary to collect data from other platforms and instruments such as airborne surveys, in situ measurements and upward-looking sonars from submarine cruises, anchored moorings and autonomous underwater vehicles. Attention has to be paid to the different scales involved, as ice properties often vary over several scales which cannot be covered with one instrument or one type of measurement, e.g. air-borne, alone. This applies in particular to the validation of satellite products.
Another sea ice climate variable is the ice drift flux through the Fram Strait. By analyzing ice drift vectors obtained from series of SAR images the ice area flux can be calculated and the cumulative flux over an ice season can be estimated. Analysis of 10 years of ice drift data are shown in Fig. 31.
In MAIRES studies have focused on algorithm development and validation of satellite derived products such as ice classification, ice drift and ice thickness. All these parameters will benefit from integrating data from several satellite sensors.
Potential Impact:
The main results from the MAIRES project are the following:
An on-line database has been created at NERSC and populated with data from European, Russian and other satellite data covering the study area in the western Russian Arctic. The database contains mostly Synthetic Aperture Radar images from ENVISAT and RADARSAT-2 and high-resolution optical images from ASTER, Landsat, MODIS and Russian satellites. These data have been used as input to the sea ice and iceberg studies in the project. The database is updated daily with new satellite dada form the Arctic and will be maintained after the end of MAIRES project, as a contribution to Copernicus marine services.
An atlas of landice products has been created by Joanneum Reserch and made available on a website, where the results all the glacier change analyzed are made available. The analysis shows that most of the glaciers have lost mass over the last 5 – 6 decades, while some have increased their mass.
For seas ice studies, new algorithms for estimation of ice classification and ice drift retrieval from SAR data have been developed and validated. These products will be used after MAIRES to develop operational sea ice services from Sentinel-1. Furthermore, ice drift data collected from several satellite systems have been analyzed for the whole Arctic, showing that the ice drift pattern varies from year to year, depending on the atmospheric wind systems in the Arctic.
For icebergs studies, intensive data collection and analysis of high-resolution optical images (ASTER images) was carried for 2012, for the Franz Josef Land, Novaya Zemlya and Severnaya Zemlya areas, where most of the iceberg in the Russian sector of the Arctic are generated. The analysis of iceberg density in 2012 was compared with Russian iceberg Atlas, showing iceberg distribution from aircraft surveys in the 1970 – 1980s. The iceberg distribution varies from year to year and it is not possible to assess any trend without more systematic data collection over many years.
The studies of land ice sea ice and icebergs conducted in MAIRES are all contributions to understand cryospheric process in the Arctic and how climate change has impact on the cryosphere. The Arctic has experienced a significant increase in surface air temperature over the last 3 – 4 decades (Fig. 32). This increase is connected to the reduction in both land ice masses and sea ice, although there are significant regional and interannual variabilities.
Satellite observations are generally obtained over relatively short-term periods for climate research. Except for sea ice extent, which has been obtained systematically for more than 35 years, satellite data records are rarely longer than a decade, but they can provide very detailed information about seasonal and regional processes. Such processes are important to understand and quantify in order to improve climate models and climate predictions. It is envisaged that satellite data will play an increasingly important role in cryospheric research and climate studies in the coming years.
List of Websites:
http://maires.nersc.no