Abstracts

(alphabetical order)

  1. Lars Axell, SMHI, "Assimilation of observed and unobserved ocean and sea ice": To make a good forecast, two things are required: (1) a model that can integrate the equations of motions forward in time, and (2) a good initial condition. The latter is usually obtained by assimilating observations into the model fields, such as S/T profiles, SST fields, and ice concentration. This talk will briefly cover the assimilation methods used at SMHI today and results from some recent experiments and analyses. However, data assimilation of sea ice usually requires assimilation of data which are not usually, nor easily, observed, such as level ice thickness and ridge density. The second part of this talk will briefly discuss how this problem currently is solved at SMHI.

  2. Lars-Anders Breivik, Steinar Eastwood, Mari Anne Killie, Jon Albretsen, met.no, Rasmus Tonboe, Leif Toudal Pedersen, DMI, “OSI SAF SEA Ice products and their use in ocean-ice-model data assimilation”: The OSI SAF Sea Ice products have been operationally available since July 2002 (http://saf.met.no/). The products are generated daily in near real time on 10 km polar stereographic grids covering the Northern and Southern hemisphere. The products are ice concentration, ice type and ice edge. The satellite data input for the current ice concentration are SSM/I brightness temperatures. Examples and validation of assimilation of these products in Met.no's operational ice-ocean model systems is given. An upgrade of the ice concentration algorithm using SSM/I 85 GHz is planned. In parallel AMSR-E data is used in a ice concentration test product. Both use of SSM/I 85 and AMSR-E gives higher resolution information and more details in particular for low ice concentrations near the edge and are therefore of particular interest for data assimilation.The ice OSI SAF edge product is using a Bayesian approach to estimate probabilities of sea ice. The operational product is using SSM/I data 19 and 37 GHz channels and thereby gives relatively coarse resolution. The near future plan is to utilize also scatterometer data and SSM/I 85 which will give more details on the edge. A high resolution regional products using AVHRR will also be available early 2009. Examples of these products are given to initiate a discussion on whether these products can be utilized in data assimilation.

  3. Mark Buehner, Tom Carrieres, Andrea Scott and Michael Ross, Environment Canada, "Towards the use of high-resolution remote sensing data in a three-dimensional variational analysis system: A collaboration with OSI-SAF": A new effort, funded by the Canadian Space Agency, to make effective use of high-resolution remote sensing measurements in an automated sea ice data assimilation system has recently started. The main components of the project are the application of approaches for extracting information on sea ice concentration (and possibly ice type) from SAR and AVHRR data within a 3D-Var data assimilation system. This will complement the assimilation of passive microwave data and eventually replace the current practice of assimilating manually derived sea ice concentration from SAR imagery. In addition, work is continuing on evaluating the use of a simple radiative transfer model to directly assimilate AMSR-E brightness temperature observations as compared with assimilating the ice concentration obtained with a standard retrieval algorithm. Finally, a new approach for modelling the background error covariances for sea ice fields within the 3D-Var is being examined. An important aspect of this project is the collaboration with OSI-SAF researchers on many of the activities described above.

  4. Tom Carrieres, Mark Buehner and Alain Caya, Canadian Ice Service, "An Overview of Some Environment Canada Sea Ice Data Assimilation Projects": Continuing in the pursuit of a more automated ice service operation, the Canadian Ice Service, in collaboration with the Meteorological Research Division of Environment Canada, have initiated a number of projects related to sea ice data assimilation. These include an International Polar Year project on sea ice data re-analysis, a Canadian Space Agency-Eumetsat joint project on assimilating passive microwave, weather satellite and Synthetic Aperture Radar data, and a Panel on Energy Research and Development funded project on improved ice forecasting. An outline will be provided on these projects as well as recent overview results. Alain Caya and Mark Buehner will present more in-depth information on specific results.

  5. Alain Caya, Mark Buehner, and Tom Carrieres, Environment Canada, "Sea ice thickness categories and their error covariances for assimilation with a three-dimensional variational data assimilation system": Sea ice models represent the sea ice in terms of a thickness category distribution, from which the moments (concentration, mean thickness) can be calculated. The main source of information about sea ice comes from observations from space-borne platforms, providing only information about the ice concentration. To initialize the ice model, information about the ice concentration has to be properly transfered into the different ice thickness categories of the model. This is achieved through the error covariances of the ice thickness distribution. In this study, an ensemble Kalman filter is used to estimate the homogeneous and static part of the error covariances of the ice thickness distribution. The data assimilation is then carried on by a three-dimensional variational data assimilation system to initialize a coupled ice-ocean model of the Canadian East-coast. An overview of the 3D-Var data assimilation system is first presented. Then assimilation experiments are described and the results are discussed.

  6. Ian Fenty, P. Heimbach, C. Wunsch Massachusetts Institute of Technology, Cambridge, USA, "Assimilation of sea ice observations in a coupled ocean sea ice state etsimate of the Labrador Sea using the adjoint method": Despite being one of the most abundant high-latitude observations of the world ocean, satellite-derived sea ice concentration (SIC) observations are often excluded in global ocean data assimilation efforts. In efforts where the adjoint method, or method of Lagrange multipliers, is utilized, such as in the Estimating the Climate and Circulation of the Ocean (ECCO) project, this exclusion has been due to the difficulty in generating useful adjoints of highly nonlinear sea ice models. Without a sea ice model adjoint, it is impossible to identify the physical pathways linking the sea ice state to the model control variables, such as the atmospheric state and initial ocean state. Now that a useful sea ice model adjoint has been created, SIC observations can be readily included in state estimation efforts such as ECCO. We demonstrate that additionally including SIC data quantitatively improves a regional one year ocean-sea ice state estimate compared with an estimate made using only observations of in situ ocean temperature and salinity taken from CTDs, profiling floats, and XBTs. The state estimate is made in a regional 1/3 degree coupled sea ice-ocean model of the Labrador Sea using the MIT General Circulation Model.

  7. Jean-Claude Gascard, LOCEAN-IPSL, “The DAMOCLES project and beyond - recent advances in oceanographic observations in arctic regions and the legacy of IPY oceanography”

  8. Nick Hughes, met.no, "Title": The Norwegian Ice Service provides sea ice charting for the European sector of the Arctic, covering the area between Cape Farewell in the west and the Kara Sea in the east.  Charts are produced by our ice analysts using a combination of remote sensed images and observations.  In recent years more detail has been provided in the Svalbard area through the use of SAR images.  Our electronic archive of ice charts cover the period 1967 to present and data can be customised to individual user requirements.

  9. Thomas Kaminski (1), Frank Kauker (2,3), Michael Karcher (2,3), Ralf Giering (1), Ruediger Gerdes (3,4), Michael Vossbeck (1), (1) FastOpt, (2) OASys, (3) Alfred Wegener Institute for Polar and Marine Research, (4) Jacobs University Bremen, "Variational assimilation of hydrographic and ice concentration observations into a coupled ocean sea ice model of the Arctic": NAOSIMDAS is a variational assimilation system around the coupled ocean sea-ice model NAOSIM, which is being developed within DAMOCLES. We present a first NAOSIMDAS experiment, which uses hydrographic and remotely sensed ice concentration observations to derive analysed fields for September 1979. We also outline a potential extension of NAOSIMDAS as a network design tool, suitable for evaluation of (potential or real) observing systems.

  10. Frank Kauker (1,3), Thomas Kaminski (2), Michael Karcher (1,3), Ralf Giering (2), Ruediger Gerdes (3,4), Michael Vossbeck (2), (1) OASys, (2) FastOpt, (3) Alfred Wegener Institute for Polar and Marine Research, (4) Jacobs University Bremen, "Adjoint analysis of the summer 2007 low in Arctic sea-ice area": The past two decades saw a steady decrease of summer Arctic sea ice extent. The 2007 value was yet considerably lower than expected from extrapolating the long-term trend. We present a quantitative analysis of this extraordinary event based on the adjoint of the coupled ocean-sea ice model NAOSIM. This new approach allows to efficiently assess the sensitivity of the ice-covered area in September 2007 with respect to any potential influence factor. We can trace back 86\% of the ice area reduction to only four of these factors: May and June wind conditions, September 2-meter temperature, and March ice thickness. Two thirds of the reduction are determined by factors that are already known at the end of June, suggesting a high potential for an early prediction.

     

  11. Christof König Beatty, Thierry Fichefet, Martin Vancoppenolle, Hugues Goosse, "Data Assimilation (and more) using the Ensemble
    Kalman Filter in LIM3":     We give a short introduction to how the Ensemble Kalman Filter works and what it can achieve. We point out that it can be used for parameter estimation (tuning) besides data assimilation, but also what its potential shortcomings are. Using a simple climate model we illustrate those points. We are in the process of implementing the Ensemble Kalman Filter with our sea-ice model LIM3 (Louvain-la-Neuve Ice Model 3), so we end with a presentation of our on-going and proposed activities.

  12. Thomas Lavergne, met.no, "Arctic Wide Ice Motion Estimation from Operational Satellite Imagery : improving the time span and accuracy of low resolution products": The estimation of ice motion vectors from sequence of satellite images has a long history of successful applications, both in research and, more recently, operational contexts. When assimilated in ocean and ice coupled models, such a dataset has been proved to enhance the quality of the forecasts with a strong feedback on ice thickness and internal stress. Two groups of such ice motion products exist, high resolution (from SAR or AVHRR imagery) and low resolution (from passive microwaves imagers, e.g. SSM/I and scatterometers, e.g. QuikSCAT). This second group of products offers a better coverage (daily to the Arctic Ocean extent) but does not allow tracking fine temporal or spatial features, being drift estimates over several days (3 to 6 for the IFREMER merged product). Such a long time span might be a critical issue for the operational assimilation of ice drift into ice models, as the integration period drastically lengthens.In this contribution, we present preliminary results from a research effort conducted during the Damocles, iAOOS-Norway, MERSEA-IP and OSI SAF projects at the Norwegian Meteorological Institute. An alternative method for retrieving the ice motion vector is introduced which allows tracking ice over shorter time spans (48 hours), yet from the same satellite data and without sub-sampling or using Resolution Enhancement techniques. Mainly, the CMCC (Continuous Maximum Cross Correlation) technique permits the removal of the quantization effect (aka tracking error) and thus provides spatially smoother and more accurate motion fields, which supposedly are less demanding for the ice mechanics part of the model, during assimilation or forcing. The method is described and validation results against IPY in-situ data (buoys, Tara, NP-35) are given. Remarks on what a satellite-derived ice motion product actually does (and does not) measure are finally made which might help the further refinement of assimilation methods.

  13. Christian Melsheimer, Institute of Environmental Physics, University of Bremen, "Satellite remote sensing of Arctic surface parameters at University of Bremen": We retrieve several Arctic surface parameters from data from several satellite instruments (AMSR-E, AMSU, MODIS), namely, sea ice concentration, multi-year ice fraction, sea and ice surface temperature, sea ice emissivity at microwave frequencies, snow grain size, soot concentration of snow. The multi-year ice fraction, ice surface and sea surface temperature are retrieved along with some atmospheric parameters from AMSR-E microwave radiances using an inverse method, the other parameters mentioned above are retrieved with direct methods based on physical or empirical models. Here we give an overview of the retrieval methods, their development status, the resulting data and their characteristics.

  14. An T. Nguyen, Dimitris Menemenlis, Ron Kwok, JPL, "Assessment of the ECCO2 regional optimized solution in the Arctic": One of the Estimating the Circulation and Climate of the Ocean, Phase II (ECCO2) project's objectives is to realistically estimate the Arctic Ocean circulation and sea-ice distribution during the ocean satellite era (1978-present).  Towards the release of a global optimized ECCO2 solution to the scientific community, an optimized Arctic solution has been obtained using the Green's function approach to minimize model/data misfits.  This paper provides a comprehensive assessment of this Arctic solution using satellite and in-situ measurements of sea-ice, freshwater and heat fluxes, and temperature/salinity vertical profiles.  Compared to the baseline, the optimized Arctic solution improved significantly with an overall model-data cost reduction of 54\%.  For the Arctic Ocean, the optimized solution successfully reproduces the Cold Halocline in the Canadian Basin and improved significantly Atlantic Water properties in the Arctic Ocean and Greenland Sea.  In addition, transport of freshwater across Fram and Bering Straits are compatible with derived estimates from previous published results.  For sea-ice, both thickness and velocity improve with cost reductions of 40\% and 68\%, respectively.

  15. Leif Toudal Pedersen, Gorm Dybkjær, Danish Meteorological Institute, Center for Ocean and Ice, "High resolution Arctic Sea Ice drift": In recent years high resolution radar and visible satellite images of the Arctic have advanced our knowledge about sea ice mechanics processes considerably.Synthetic Aperture Radar image coverage of large parts of the Arctic every 3 days has been available since the launch of RADARSAT in the mid 1990s, and for the last 15 months daily coverage of most of the European Arctic has been available from ESA's ENVISAT ASAR. A major result of the analysis of the radar images is the demonstrated capability to detect ice deformation at sub kilometre scale. The radar images allow detailed observations of the Arctic independent of sunlight and cloud cover. However, observations are available only every 1-3 days and for some areas even less frequently.At DMI we supplement the radar coverage with the more frequent (but cloud dependent) observations by the many AVHRR instruments in space. The wide swath and large overlap in the Arctic allows 14 observations per day of the Central Arctic with just one satellite, and since 3-4 instruments are in orbit we have observations at approximately every hour. The AVHRR data has a spatial resolution of ~1 kilometer which is sufficient for many sea ice dynamics studies.Recent analysis of 250m resolution MODIS  images from the east coast of Greenland reveals detailed small scale ice dynamics that potentially allows us to track also surface currents in the ocean. Examples of our use of satellite data for sea ice dynamics studies will be shown.

  16. Till Rasmussen, University of Copenhagen, Nicolai Kliem, DNMI: "Validation of a regional coupled ocean/sea ice model in the Nares Strait" : A simulation of the climate around Greenland has been made with HIRHAM (The Danish Meteorological Institute’s atmospheric climate model) for a period of 100 years. The result from this simulation has been used to force a regional coupled ocean and sea ice model (HYCOM+CICE) at the surface. The region considered is Nares Strait, Lincoln Sea and the Baffin Bay. The lateral boundaries of HYCOM and CICE have been relaxed towards a global model. The purpose of this study is to investigate the coupled systems ability to model the hydrography and the sea ice distribution and to investigate the freshwater flux through the Nares Strait. The Nares Strait is expected to be a demanding test area for both the ocean and sea ice model with mixing of different water masses, building up an ice bridge, and formation of polynias. This talk will focus on the validation of this model, mainly flux of water and sea ice through the Nares Strait.

  17. Joao Rodrigues, University of Cambridge, "Measurements of Arctic sea ice thickness with upward-looking sonars in 2004 and 2007": We present results for ice draft statistics in the areas of the Arctic Ocean covered by the British submarine Tireless cruises in the spring of 2004 and winter of 2007. The first voyage includes the regions of Fram Strait, the 85N parallel between 0 and 60W and the North Pole. The second voyage covered the regions of Fram Strait, N coasts of Greenland and Ellesmere Island and the Applis ice camp N of Alaska. Along-track single beam upward-looking sonar data were recorded using Admiralty-pattern 780 and 2077 echo sounders. Preliminary analysis for nearly-coincident parts of the tracks shows little difference between the averages values of the ice draft in 2004 and 2007.

  18. Katja Rollenhagen, Ralph Timmermann, Jens Schroeter, Alfred Wegener Institute, "Data assimilation in a regional finite-element Arctic sea-ice model": A finite-element Arctic sea-ice model is applied in a data assimilation study with the Singular Evolutive Interpolated Kalman (SEIK) Filter. We assimilate three-day mean ice drift fields derived from passive microwave satellite data. Based on multivariate covariances, the sea-ice drift data assimilation produces not only direct modifications of the ice drift but also updates for sea-ice concentration and thickness, which in turn yield sustainable corrections of ice drift. We use observed buoy trajectories as an independent dataset to validate the analyzed sea ice drift field. A good agreement between modeled and observed tracks is achieved already in the reference simulation. Application of the SEIK filter with satellite-derived drift fields further improves the agreement. Spatial and temporal variability of ice thickness increases due to the assimilation procedure; a comparison to thickness data from a submarine-based upward looking sonar indicates that the thickness distribution becomes more realistic.

  19. Rasmus Tonboe1, Gorm Dybkjær1, Leif Toudal Pedersen1, Steinar Eastwood2, Lars-Anders Breivik2, John Stark3, Walter Meier4,(1) DMI, (2) met.no, (3) Met Office, (4) NSIDC": The joint NSIDC and EUMETSAT sea ice re-analysis":The use of thermal microwave data for mapping the sea ice extent and area is perhaps the most successful application of satellite remote sensing for sea ice monitoring. Today time-series, covering the arctic regions daily from the early 1970s, are most significant for estimating inter-annual and decadal trends in this important climate parameter. Applications also include climate oriented coupled general circulation and numerical weather prediction models. These data are important input to numerical sea ice models where the ice thickness is estimated. Ice concentration is not directly linked with ice thickness, but the minimum ice extent in summer is a measure of the amount of thick multiyear ice. The reduction in the multiyear ice extent over the past decades is an indication of an ongoing climate change process that affects the ice thickness as well. From a climate change perspective, the key question is how fast the total volume of sea ice is changing. This requires reliable estimates of ice concentration for the derivation of the sea ice area. Therefore, ice concentration is an important ice cover parameter and must be estimated accurately.Seven of the most common radiometer algorithms, used to compute the sea ice concentration, were compared to ScanSAR data estimates of ice concentration in Andersen et al. (2007). The focus was on the near 100% ice cover in winter. On a climatological time scale the differences between algorithms amounts to 14% and 22% of the down-going trend in winter Arctic sea ice extent and area, respectively. The climatological changes in atmospheric and water surface emissivity primarily affect the extent trend while the changes in sea ice surface emissivity affect the sea ice area trend. In other words there is a climatic trend in the sea ice time series related to changes in the snow cover and sea ice surface properties and the Arctic atmosphere. Reliable estimates of atmospheric cloud liquid water and the ice brightness temperature variability are not readily available and it is therefore important to find ice concentration algorithms that are least sensitive to these atmospheric and surface properties. Other parameters, such as atmospheric water vapour and open ocean surface wind, are quantified rather well by numerical weather prediction models. It is therefore feasible to correct brightness temperatures for the influence of these effects using radiative transfer models before computing the ice concentration.The joint NSIDC and EUMETSAT reanalysis project is an extension of existing Ocean and Sea Ice SAF (OSI SAF) plans to reanalyse the SSM/I record (1987 to present). The entire level-1 data set was purchased from Remote Sensing Systems by EUMETSAT for use in the SAF network. Cooperation with NSIDC includes extension with SMMR back to 1978 and the collaboration further defines a common dataset and standards. The processing includes atmospheric correction of brightness temperatures for open ocean surface wind roughness and atmospheric water vapour. A set of three different ice concentration algorithms has been selected due to their low sensitivity to cloud liquid water and ice surface emissivity. A new procedure has been developed where tie points for the ice concentration algorithms are selected dynamically in time and space. This reduces inter-sensor, sensor drift and seasonal biases. Processing is ongoing and the first results including error-bars are expected in 2008 and will be used in both the ECMWF and the Hadley centre re-analysis and climate models. The level-3 dataset will be available to the community via the OSI SAF and NSIDC websites.

  20. Stiig Wilkenskjeld, Mads Hvid Riebergaard, DMI, “Development of operational coupled sea ice and ocean model at DMI”: Status and experience from development/implementation of HYCOM/CICE coupled ice/ocean model at DMI.