Access to SAR data

Use of SAR data for marine monitoring in the Northern Sea Route, the Black Sea and the Caspian Sea has until now been limited to a few research and demonstration projects (i.e. ICEWATCH, ARCDEV [1]) and commercial projects (i.e. ice monitoring for oil industry). The results of SAR ice analysis in the Northern Sea Route are presented in a recent book [2].  Regular observation of sea ice by SAR in the Arctic has not been feasible because the access to wide-swath images has been limited.  However, from 2005 SAR data from ENVISAT can be obtained from ESA’s rolling archive for use in RTD projects and in development of GMES services. The rolling archive can deliver SAR data in near real time, which is an important milestone in developing SAR monitoring services, In the ICEMON [3] project (GMES consolidation study funded by ESA GSE 2003-2004) and the OSCSAR [4] project (ESA-IAF GMES networking with Russia and Ukraine) a large number of ERS and ENVISAT SAR scenes have been acquired and made available for the three study areas.   New ENVISAT SAR data will be ordered from ESA. The near real time capability of ESA rolling archive will be exploited during the monitoring exercise, showing the possibility to deliver daily services to users based on SAR data from ENVISAT. Also, near real time SAR data from RADARSAT will be provided by the Kazakhstan partner (CAR).

Retrieval algorithms

The most recent state-of-the-art review of SAR observation of the marine environment is Synthetic Aperture Radar Marine User’s Manual, (eds. Jackson and Apel) issued by NOAA, September 2004. Based on several decades of studies, it demonstrates the powerful capabilities of spaceborne SAR systems to observe and quantify a number of oceanographic, atmospheric boundary layer and sea ice processes.

Existing SAR system from ERS-2 (C-band, VV), RADARSAT-1 (C-band, HH) and ENVISAT (C-band, VV/HH/VH/HV) providing high resolution images have already demonstrated their capability to detect oil spills in many sea areas, as reported by e.g. Gade et al. (1998), Lu et al. (1999), Espedal and Johannessen (2000), Assilzadeh and Mansor (2002), Alpers and Espedal (2004). In addition to oil spills, there are many others surface features (such as current features, natural films, surface temperature patterns, rain cells) that provide similar radar signatures as oil, as shown by e.g. Beal et al., (1997) and Espedal (2001). Many studies have pointed out that the distinction between oil spills and their look-alikes can be done by means of synergetic analysis of radar images in different bands and polarizations. However, the access to such images has been very limited.  In MONRUK, use of alternating polarization images from ENVISAT will be used to improve slick classification.

Most of the efforts to detect and classify oil spills in SAR images have focused on analysis of images only. This means that it is difficult to identify and classify real oil slicks from many types of natural slicks which can be very similar to oil slicks ("look-alikes").  Operational oil spill monitoring systems using SAR data have therefore a problem with identification of many false positives.  Presently, little efforts have been devoted to study the physical properties of various slick types and how these properties have impact on the SAR signature. In two ongoing projects, OSCSAR and SIMP [5], there are efforts to investigate the physical processes responsible for generating slicks in SAR images.  In MONRUK, further studies will be undertaken using air-sea interaction models and radar scattering models in combination with field experiments to improve the knowledge of the mechanisms responsible for slick signatures.

SARTool is software package to process SAR images from ERS, ENVISAT and RADARSAT for retrieval of oceanographic parameters: surface wind and waves, ship and oil spill detection.  SARTool uses a number of auxiliary data files providing bathymetry, shoreline, wind data, SST, surface currents from models and others. It also contains a number of image processing and statistical tools.  SARTool is developed by BOOST Technologies [6], which is partner in MONRUK. SARTool will be further developed in MONRUK to include improved discrimination and quantification of surface slicks, surface wave spectra, propagation direction, significant height of dominant waves, wind fields near coasts and in the marginal ice zone, current velocity gradients and sea ice parameters.  Existing algorithms for detection of oil spills developed in other projects, usually based on statistical information in SAR, will be investigated.

SAR images, with a typical pixel size of 30-100 m, allow observation of a number of sea ice parameters such as floe parameters [Soh et al., 1998], concentration [Haverkamp and Tsatsoulis, 1999; Sandven et al., 2001], ice drift [Kwok et al., 1998], ice type classification [Kwok et al., 1992, Wackermann and Miller, 1996], leads [Van Dyne et al., 1990], and ice edge processes [Johannessen et al., 1992].  All these parameters can be retrieved through ice analysis performed by human interpretation, for example at operational ice centres using SAR data for ice monitoring [Bertioa et al., 1998]. Various algorithms are used to retrieve the ice parameters, but only a few can be run automatically. Most of the algorithms need human input as well as supporting data from other sources. Sea ice drift retrieval and classification of multiyear versus first year ice are examples of parameters that can be retrieved by automatic algorithms.

Several methods for sea ice classification have been developed and tested [i.e. Wackermann and Miller 1996, Sandven et al., 1999]. The straightforward and physically plausible approach is based on the application of sea ice microwave scattering models for the inverse problem solution [Fang, 1994]. This is, however, a difficult task because the SAR signature depends on many sea ice characteristics, as shown by Winebrenner et al., [1992]. A common approach in classification is to use empirically determined sea ice backscatter coefficients obtained from field campaigns, as shown by Onstott et al. [1979] and Johannessen et al., [1992]. Classical statistical methods based on Bayesian theory [Fukunaga, 1990] are known to be optimal if the form of the probability density function (pdf) is known and can be parameterised in the algorithm. A Bayesian classifier, developed by Kwok et al., (1992) for Alaska SAR Facility (ASF), assumes a Gaussian distribution of sea ice backscatter coefficients [Fetterer et al., 1994]. Utilization of backscatter coefficients only, limits the number of ice classes that can be distinguished, and decreases the accuracy of classification, because backscatter coefficients of several sea ice types and open water overlap significantly [Sandven et al., 1999]. The RADARSAT Geophysical Processor System (Kwok, 1998) has been developed to process large amounts of SAR data and retrieve information about ice drift, ice formation, ice deformation and other derived parameters. Systematic use of SAR data for ice observation can provide much better quantitative data and scientific understanding of sea processes.  The most recent reviews of SAR ice observations have been provided by Johannessen et al (2005), Onstott and Shuchman (2004) and Shuchman et al. (2004).

In MONRUK, improvement of sea ice algorithms will be done as follows: ENVISAT ASAR wide-swath images will be used to produce mosaics and derive ice drift that will be used to investigate the ice dynamics in the Northern Sea Route. Furthermore, ice type classification methods based on neural network and use of multi-source data, following the procedure proposed by Bogdanov et al., (2005) will be done. The advantage of using alternating polarization images from ENVISAT, and later from RADARSAT-2, for classification will be investigated. Other ice parameters such as ice concentration, leads, polynyas, ridges, shear zones and fast ice areas will be investigated.  Validation of the SAR observations will be done by field observations obtained from icebreaker expeditions, where scientists from NIERSC will participate. Finally, sea ice algorithms for ice classification and ice drift will be implemented in the SARTool computer program.

SAR ocean models

Most of the work in SAR ocean modelling has focused on improving forward models to predict the SAR signatures of regions with changing surface conditions, i.e. ocean currents interacting with the surface waves (Lyzenga and Bennet, 1988; Romeiser and Alpers, 1997). NIERSC has recently developed an advanced Radar Imaging Model (RIM) which predicts background radar scattering from the sea surface at arbitrary atmospheric stratification, wind and wave conditions (including cases of limited fetch relevant to coastal zone and marginal ice/water region), as well as radar signature of current features, surface temperature fronts and natural slicks or oil spills (Kudryavtsev et al., 2005; Johannessen et al., 2005). The RIM model has been verified against available well-controlled field experiments. Another model developed at NIERSC is a model of atmospheric boundary layer (ABL) transformation at abrupt changes of surface roughness and temperature relevant to coastal and marginal ice/water zones. This model predicts spatial distribution of wind velocity profiles and the sea surface wind stress "offshore". The former defines wind wave evolution while the latter - short wind wave spectrum. Thus, combination of RIM and ABL model can be used for development of advanced wind retrieval algorithms for fetch limited conditions.

In the MONRUK project, the RIM and ABL models will be combined in order to improve wind retrieval in coastal zones and marginal ice zones. The RIM model will be further advanced to allow retrieval of the general surface current features, like their divergence. It is suggested that wind field and surface temperature available from SAR and NOAA images will be the input for the current retrieval algorithm. One of the output parameters of the RIM model is mean square slope of the sea surface that is responsible for the forming of optical images of the sea surface. Thus an advantage of synergy of SAR and optical images for the sea surface current features retrieval will be assessed and implemented. Furthermore, discrimination of oil spills and their look-alikes as well as quantification of oil spill will be included. Current velocity fronts will also be retrieved where oil spill look-alikes often are observed in SAR images. The improved capabilities of RIM and ABL will be implemented in an updated version of the SARTool.

Testing of service chain and user interaction

A service chain to deliver SAR-based information to users of marine information in the context of GMES has not yet been established for the Russian, Ukrainian and Kazakhstan (RUK) areas.  Pre-operational demonstrations of SAR ice monitoring in the Northern Sea Route has been performed by NERSC through projects such as ICEWATCH 1995-1996 (Johannesset et al, 2000) and ICEMON 2003-2004.  Oil spill monitoring has been performed only as part of research projects such as OSCSAR and SIMS.  In MONRUK, it is proposed to set up a service chain that runs a pre-operational monitoring service for a test period of a few months.  The service chain will consist of the following components which will be used both for open ocean and sea ice monitoring: (1) SAR data provision will be provided by NERSC using ENVISAT data from ESA's rolling archive.  RADARSAT data from CAR will be used as additional data source in the Caspian Sea. (2) Service provision including data analysis and product delivery will be done by NIERSC in the Northern Sea Route, by MHI for the Black Sea and by FSUE RISDE in cooperation with CAR for the Caspian Sea. (3) Reception and validation of products by users will be organised by the service providers for each region, where selected users will provide feedback and validation of the service.  The results of the monitoring exercise will be used to plan and implement operational monitoring services in the RUK areas. 

Web map server

A prototype web map server system (DISPRO) is under development in the DISMAR project. DISPRO is capable of integrating distributed multi-source data and as well as numerical model simulations from several providers. The DISPRO architecture is consistent with INSPIRE's general model of an SDI (Spatial Data Infrastructure).  DISPRO is a multi-tier system with four main groups of components: user applications, geo-processing and catalogue services, catalogues and content repositories. Implementation is based on INSPIRE, OpenGIS and W3C standards, using Open Source software where available. Metadata plays a central role in DISPRO. All data products and services are described in an accompanying metadata file. The latter is a profile of the geographic metadata standard (ISO 19115) restricted mainly to the core 'discovery' metadata elements. Metadata are provided in XML format and validated against an XML Schema. The metadata are stored in a native XML database and transformed to HTML for presentation using XSLT stylesheets.

The DISPRO system, an OGC [7] conformant web map service system, will be extended through MONRUK to handle data served by oceanographic centres, typically netCDF, served through the OpeNDAP/DODS [8] Data Access Protocol (DAP). OpeNDAP/DODS has developed a software framework that simplifies all aspects of scientific data networking, allowing simple access to remote data. Local data can be made accessible to remote locations regardless of local storage format by using servers. Existing, familiar data analysis and visualization applications can be transformed into clients (i.e. applications able to access remote served data. The OpeNDAP Data Access Protocol (DAP) is a protocol for requesting and transporting data across the web. DAP 2.0 uses HTTP to frame the requests and responses (http://www.opendap.org/faq/whatIsDods.html).

This enhancement to the DISPRO system will allow it to integrate with application servers developed as part of the Mersea project, thus allowing DISPRO to play a full role in GMES type services. Furthermore, the DISPRO system will be extended to include service providers in Russia, Ukraine and Kazakhstan.

Links

[1] ICEWATCH (1995-1996) and ARCDEV (1997-1999) were demonstration projects were NERSC performed near real time ice monitoring by use of SAR data in the Northern Sea Route.

[2] POLAR SEAS OCEANOGRAPHY Remote Sensing of Sea Ice in the Northern Sea Route: Studies and Applications, Editors O. M. Johannessen et al., Praxis Springer. http://www.springer.com/west/home/environment?SGWID=4-198-22-2223586-0