Report 2008
Task 1.2 Ice types and properties
Progress: 55%
Task 1.2 aims at a better understanding for the use of satellite data for remote sensing the Arctic sea ice types and properties because in situ data are difficult to obtain from these regions with hostile living conditions for men. The resulting maps will support field campaigns, data analysis in case studies and studies to assimilate these data into oceanic and atmospheric models in WP 4.
D1.2-1 Sea ice parameters
b) Ice concentration and fractionation
UB: providing daily sea ice concentration data for drifting station TARA since August 2006.
c) Incidence-angle adjusted backscatter fields from ASCAT
The main activity at IFREMER has been to built the first ASCAT/MetOp daily incidence-adjusted backscatter maps. First uncalibrated data were available in March 2007. The backscatter maps show that the incidence-correction is a must to get geophysical information. First drift maps are also processed, the comparison with QuikSCAT drift maps shows a good agreement. Ice type discrimination would be greatly improved using both ASCAT C-band and QuikSCAT Ku-band data, this will be analysed during the 2007-2008 winter.
D1.2-2 Emissivity studies
a) Empirical studies
UB:
Deliverable completed. Summary: Knowledge of the surface emissivity is essential for retrieving both surface and atmospheric parameters from satellite observations with passive microwave sensors. The emissivity of open ocean is low and can be predicted from physical models. Also land emissivity can be constrained rather narrow from a priori knowledge. However, the emissivity of sea ice is so high, variable and difficult to predict from environmental parameters that up to date no reliable forward models and the microphysical parameters required as input are not known to an accuracy sufficient to allow for the retrieval of atmospheric parameters over sea ice (except few special cases where not the full emissivity information is needed). Therefore, here an attempt is made to gather empirical information on the surface emissivity of sea ice to the best possible accuracy. This information is intended to be used for the retrieval of atmospheric parameters such as total water vapour, and cloud liquid water path or temperature profiles. Better emissivity knowledge will also be helpful in order to better constrain surface parameters such as ice type and properties. Spaceborne microwave radiometers operate at typical frequencies so that here the retrieval is constrained to them rather to determine emissivity at all frequencies. The different frequencies and scanning schemes (conically and cross track) are here covered in two separated reports attached to the deliverable, which at the same time have been submitted as manuscripts to peer reviewed journals. As a third attachment to the deliverable, and as an example for the usefulness of the obtained emissivity information, a procedure to improve the temperature profile retrieval from AMSU-A data by including a priori emissivity information has been developed.
met.no:
met.no: Work is ongoing to modify the observation processing in the NWP model HIRLAM data assimilation to allow for improving the treatment of the sea ice emitting temperature and emissivity. The developments consist of modeling of the efficient emitting temperature of the sea ice based on work at University of Bremen, and of refining and testing first guess emissivity values based on developments by DAMOCLES partners. This will allow for evaluating the schemes for emissivity and emitting temperature using colocated HIRLAM and microwave satellite data (AMSU-A). It will later on be used in an improved data assimilation system for microwave observations over sea ice.
b) Microwave emissivity model
DMI: A sea ice and snow microwave emissivity model has been developed. It treats an n-layer snow and ice profile and deals with scattering from new and re-crystallised snow and brine or air inclusions for each individual layer. A validation experiment in Kangerlussuaq/ Greenland was carieed out April 2007. The emission model relates physical snow and ice properties such as density, temperature, snow crystal and brine inclusion size to microwave attenuation, scattering and reflectivity, brightness temperature, emissivity for v or h polarisation at oblique incidence angles, effective thermometric temperature. Output from a mass and thermodynamic sea ice model can be used as input to the emissivity model. Snow layering is very important for the microwave signatures therefore the thermodynamic model treats snow layers related to individual precipitation events. It also has a growth rate dependent salinity profile. A report describing the thermodynamic model and one describing the emissivity model has been written. The emission model is now used in other parts of WP1.2 at DTU and UB. Next the model will be validated with the field experiment data.
c) Freeze/melt state
The deliverable has been completed. Summary: The onset of melt of Arctic sea ice is an important geophysical parameter, together with the subsequent freeze-up in the autumn. Melting induces a drop of the surface albedo which is about 0.8 for dry snow to values less than 0.5 for mixtures of bare ice and melt ponds. Moreover, the freeze-up ends the input of freshwater to the upper layers of the ocean. As the albedo drives the surface energy budget, changes in the timing of freeze-up and melt may also modulate the mean annual ice thickness. The exact location and timing of melt and freeze-up is also needed when parametrizing the surface albedo, which is currently rather crude in climate models. Finally, a global warming will affect and be visible in an extended melt period of the Arctic sea ice. In the report, the state of previous and current activities is reported. As data set it is suggested to use a data set which has been analysed in a study published in 2007 and which is currently being extended at the NSIDC.
d) Snow parameters on sea ice
The sea ice and snow microwave emissivity model required for the sensitivity study has been developed, implemented, and validated; see section D1.2-2 b) above. Long-term field work aboard the drifting vessel Tara has started. Sensitivity studies have been carried out in order to assess the feasibility of deriving snow and ice parameters by an optimal estimation procedure from AMSR-E and AMSU passive microwave measurements.
D1.2-3 Ice dynamics from satellite observations
a) Low resolving
Met.no: An algorithm has been designed to retrieve sea ice drift from low resolving sensors (currently the SSM/I). It is able to measure the ice displacement vectors inside a pair of daily satellite composite images. Efforts are being made to also retrieve the drift uncertainty. This is of importance for assessing the quality of the drift products and to aid design of a multisensor algorithm. It can also help ocean models assimilate such ice drift products more efficiently. A prototype sea-ice drift processing chain is ran and monitored every day at met.no to assess the overall quality of the algorithm. Recently also work on derivation of ice drift from ASCAT has started. The work takes place in coordination with DMI and IFREMER. A validation against drifting buoys is also in preparation.
b) Medium resolving
DMI: A maximum cross correlation algorithm for deriving sea ice drift from sequential AVHRR data has been developed and tested. A 9 month data set of estimated ice drift in the Greenland seas is generated and uploaded to the Damocles data repository. The ice drift data set is based on both VIS and TIR AVHRR data received at the NOAA receiving facility in Kangaarlussuaq in Greenland. Early 2008 an operational NRT ice drift estimation chain will be running on test basis. This test setup is based on 1km METOP AVHRR swath data and will cover both the southern ocean and the Arctic ocean.. Documentation for the NRT AVHRR ice drift setup is being prepared.
DTU: Daily medium resolution ice drift data (3-day drift) from most of the Arctic Ocean is operationelly produced (every day).
b) High resolving
Collection and analysis of ASAR Wideswath data has continued during 2007, with focus on the Tara region, the Fram Strait and the area around Svalbard. For the Tara region Wideswath images were collected once per week on average during autumn – winter of 2006-07. From mid-April SAR data were collected every day for a month period, but Tara drifted north of the ENVISAT SAR coverage area after 07 April. From mid August the Tara drifted back into the coverage area of ASAR, south of 87 N. The Wideswath coverage of Tara as it drifted southwards to the Fram Strait has been quite scarce during the last few months.
Preliminary analysis of SAR ice drift has been performed and presented in the D1_2-03c report delivered in June. Ice drift from SAR over weekly time scale has been compared with Tara drift data and AMSRE drift data, showing good agreement. Also daily SAR ice drift has been calculated for a two week period and compared to 2- and 3-daily AMSRE ice drift. A few examples of convergence/divergence zones have been detected in the SAR ice drift data, but most of the ice drift analysis shows quite uniform ice velocity fields around Tara. The ice drift data from SAR and other satellite data will be compared to ice drift from models. The SAR data will furthermore be used for ice classification and mapping of leads and thin ice areas.
DTU: A special version of our ice drift detection algorithm is being applied to derive 24h ice drifts from high resolution ENVISAT ASAR data from the Lincoln Sea and the Tara drift area. Since June 1, 2007, the algorithm has been run operationally on all available ENVISAT ASAR WSM data from the Arctic. The algorithm performs very well, even during the Arctic Summer, and a large dataset of ice drift vectors from the European Arctic has been derived.
D1.2-04: Report on ice type classification based on satellite data
One of the objectives of NIERSC in WP1.2 is to improve the multiyear retrieval algorithm for passive and active microwave satellite data. The most consistent and the longest available quantitative means to monitor the Arctic sea ice cover is from satellite-borne passive microwave sensors. However a number of studies demonstrated problems of retrieving multi year ice concentrations form passive microwave data (growth of calculated by passive microwave algorithms multiyear ice coverage during winter). An improvement of passive microwave calculations has been suggested based on using scatterometer data. QuikSCAT scatterometer data was used as a complimentary source of information that assisted in separating first year and multi year ice, improving the algorithm of multiyear ice retrieval and describing multiyear ice extent changes in the periods when passive microwave retrievals cannot provide a stable picture. NORSEX calculations and QuikSCAT-based retrievals were validated using available ASAR images and NIC and AARI ice charts. Capabilities and limitations of using passive microwave data for estimating the relative coverage of first year and multi year ice in the Arctic are discussed and quantitatively established in some cases.
Another approach to improve ice type classification that is under development in NIERSC uses a Neural-Network (NN) approach. The data for numerical experiment have been selected. It has been decided to use Era-40 Reanalysis data on atmospheric parameter profiles and sea ice temperature for the numerical integration of the radiative transfer of the microwave emission in the Atmosphere-Ocean-Ice System. The set of programs for data reading with the following quality control for discarding non-physical data is elaborated. The data on cloud liquid water content and cloud boundaries are modeled basing on the results of Arctic SHEBA experiment. The geophysical radiation transfer model for non-precipitating conditions (developed earlier in NIERSC) is upgraded for the Arctic conditions using the latest models of interaction of electromagnetic radiation with medium. Computer simulations of the radiometric brightness temperatures (BTs) are carried out for SSM/I instrument characteristics. Neural Networks-based algorithms for the first year and multiyear ice concentration retrieval from SSM/I data are trained, using theoretically simulated values for brightness temperatures as inputs. Optimal NN configurations are derived. Report D1.2-04 first version is completed.
D1.2-06: Russian and non-Russian satellite data for ice type and properties retrieval
FSUE RISDE has made an overview report on Russian satellite data of Arctic sea ice, with focus on high-resolution optical and IR images covering the Barents and Kara Sea region (Deliverable D1.2-06). The images are mainly obtained from the satellite series RESURS, METEOR and MONITOR over the last decade. The images are available for use in further studies of sea ice in WP1. FSUE RISDE also provides archives of AVHRR and MODIS data for the European and Russian sector of the Arctic. The IR channels of these images are useful for estimating ice surface, sea surface temperature, heat flux from ocean to atmosphere and for estimation of thin ice thickness.
D1.2-07: Preliminary version of SGPS (Snowgrain and Pollution) algorithm
Deliverable completed. IP NASB's main goals in WP 1.2 are to develop and validate an analytical algorithm to retrieve Snow Grain Size and Pollutions (SGSP) from data of satellite optical instruments. Snow fields have potentially significant effects on the planetary albedo and climate. Development of satellite remote sensing of snow is of a great importance particularly for monitoring of snow age, pollution, and grain sizes over the Polar Regions that are difficult to access.
The preliminary version of the SGSP algorithm to retrieve the snow grain size and soot pollution concentration from satellite spectral data has been developed and delivered. The underlying theory uses the realistic model of snow as consisted of non-spherical close-packed grains instead of the customary model with independent spherical particles used in state of the art; the analytical solutions for the inherent optical characteristics of snow in terms of microphysical parameters obtained in the framework of geometrical optics instead of time-consuming computations; the analytical asymptotic solution instead of the radiative transfer computations to relate radiative characteristics of a snow layer to its inherent optical parameters.
Note that shapes of snow grains in a pixel are not known, any a priori model may lead to non-controlled errors in the retrieval. Substantial effort was made to develop the data base of optical properties of snow grains as non-spherical particles with rough surfaces and air and soot inclusions. This base has already been used for the preliminary validation of the developed algorithm and will be a part of special computer tool that will simulate the spectral response of atmosphere-snow system which is going to be developed during the next stage. The main advantage of the proposed algorithm over any currently used ones is the rejection of any a priory model use rather than the use of any model of snow particles. The additional advantage is an extremely high processing technique due to analytical nature of the developed algorithm. In the current version this algorithm is addressed to process the MODIS data, but can be easily adjusted to process data of any satellite optical instrument with appropriate spectral channels. The SGSP analytical retrieval (instead of the conventional LUT technique) provides extremely fast satellite data processing.
The results have been presented at an International Symposium on Snow Science (Moscow, Russia, 2007) and submitted for publication in the Annals of Glaciology.
D1.2-08: Processing chain for NRT MODIS data including SGPS algoritm
UB: Installing of the operational processing chain has started. The selection of the appropriate MODIS data sets ('granules', each comprising 5 minutes of sensor data) of given geographic locations has been completed.
