Poster Session/Demos/Drink

The increase in plastic pollution is advancing micro level pollution and the total weight of microplastics (<0.33mm-1.00 mm) in the ocean column. The amount of plastic in the North Pacific is estimated as 21 x 108 tons and in the North Atlantic as 10.4 x 108 tons. The plastic in marine environment will eventually degrade and it will be promptly colonized by bacteria releasing surfactants. Such surfactants will have the effect of damping the capillary and small gravitational waves on the ocean surface. Since SAR is sensitive to roughness induced by capillary waves, it may be exploited to detect bacterial activities related to plastic pollution.

In this work we used Sentinel-1A and COSMO SkyMed radar images acquired in the Atlantic and Pacific gyres to detect surfactants that may be associated to plastic pollution. We are using SAR, because the damping properties of surfactants produce dark areas in images. Since area of low backscattering in SAR images could also be produced by other oceanographic/meteorological event, we exploited geophysical remote sensing products associated to time and locations synchronised to SAR acquisitions. Among other products we considered sea surface temperature, surface wind, chlorophyll, surface reflectance, turbidity and wave heights. Additionally we made sure that the areas were not within busy shipping routes. The result of the analysis is that, including effects due to colocation errors of SAR and meteorological data, we could identify a large amount of linear slicks in SAR images that were not directly related to apparent meteorological conditions. Such slicks in the gyres have the appearance of oil slicks, however in some areas they are in large amount and they are not connected to large ship traffic. At the moment these slicks seems to only be visible when the wind conditions are moderate (e.g. 6m/s) as it happen for ordinary oil slicks.      

Besides the work on radar data, we are making controlled experiments with micro-plastic pollution in sea water, to understand the amount and type of surfactants produced by microbes colonising plastics.

The conclusion of our study is that radar remote sensing has the potential to detect plastic pollution areas under sorter meteorological conditions.  

Using a case study approach, the utility of spatially coincident (overlapping and/or adjacent) and quasi-simultaneously collected (within ~25 minutes of each other) Sentinel-1 and RADARSAT-2 imagery is demonstrated for enhanced ocean surveillance off of Canada’s east coast. We show how ship detection and characterization is improved by using spatially coincident RADARSAT-2 OSVN and Sentinel-1 IW imagery. Specifically, the higher resolution of the IW imagery allows small ships to be identified with higher certainty and allows for more accurate extraction of vessel characteristics (such as length and width estimates). Further, the displacement of the ship targets between image acquisitions allows for the direction and speed to be estimated, providing additional information that is not (necessarily) available when utilizing a single image. From an environmental monitoring standpoint, these quasi-simultaneous images allow for high confidence delineation of ocean surface features associated with thermal fronts, eddies, water mass boundaries, and atmospheric fronts. For example, we demonstrate the ability to map the movement of a strong storm system as it tracks between temporally separated, but spatially coincident, SAR images. Further, we show how Sentinel-1 Level-2 datasets, such as the Radial Velocity Product (RVL), provides additional contextual information that aids in the interpretation of RADARSAT-2 image products (such as for identification of the Gulf Stream North Wall). Finally, we show how spatially adjacent and simultaneously collected Sentinel-1 and RADARSAT-2 images can be used to expand coverage over which ocean surveillance can be undertaken by providing the equivalent of an ~800 km swath. This work provides practical examples of the benefits of collecting these types of SAR datasets for ocean surveillance, and suggests that planned observation scenarios could, in the future, be developed to fully maximize these opportunities. Looking further into the future, these examples demonstrate opportunities that may become routinely feasible after the launch of the RADARSAT Constellation Mission.

Icebergs represent a danger to navigation in cold waters. Detection and tracking of large icebergs using space-borne scatterometers, altimeters and synthetic aperture radar (SAR) systems have seen a large amount of work in the last decades. However, the identification of small icebergs is still challenging especially when these are embedded in sea ice.

In this work, a recently proposed iceberg detector the intensity Dual-Pol Ratio Anomaly Detector (iDPolRAD) is tested using ALOS-2 quad-polarimetric L-band data. The detector was originally designed for dual-polarized incoherent SAR images and it was optimised and validated for the use with Ground Detected Sentinel-1 data [1].

This work with quad-pol L-band has two main purposes: 1) we want to investigate the capability of the same detection formalism (the DPolRAD) when this is adapted to use quad-polarimetric data (for possible future mission); 2) we want to understand how the original DPolRAD detector (i.e. using the same typology of dual-pol data) comperes when we use L-band instead than C-band.  

The ALOS-2 data were acquired on the East Coast of Greenland, around the Kangerlussuaq glacier where we expect a large number of grounded icebergs visible in the images.

The results show that different polarisations could bring different information. In particular, the use of two coherent co-channels (HH and VV) could allow to set up the detector to identify anomalies in the horizontal reflections, rather than oriented reflection and volume as can be done with the incoherent HH and HV. Therefore the availability of quad-pol would allow the possibility to have a bank of DPolRAD detectors, each one focused on a different scattering mechanism. This would improve the detection performance. However, a quad-pol algorithm would not be operational at the moment due to the limited size of the quad-pol swath. We also observed that L-band data allows to better see some scattering mechanisms associated to reflections. This is probably due to the lower backscattering from the ice body compared to C-band which reduces the volume scattering at advantage of reflections from iceberg surfaces and ice-water interfaces. 

Reference:

[1] A. Marino, W Dierking, and C. Wesche, “A depolarization ratio anomaly detector to identify icebergs in sea ice using dual-polarization SAR images,” vol. 54, no. 9, pp. 5602–5615, 2016.

Synthetic aperture radar (SAR) instruments on board satellites are valuable for high-resolution wind field mapping. However, there is still no consensus on the optimal resolution for wind speed retrieval from either co-polarized or cross-polarized SAR images. This paper presents a comparison strategy for investigating the influence of spatial resolutions on sea surface wind speed retrieval accuracy from SAR images.

In order to evaluate the spatial scale effect on wind speed retrieval from co-polarized SAR images, we makes a quality assessment of wind speed retrieval accuracy based on four commonly used C-band geophysical mode functions (CMOD4, CMOD-IFR2, CMOD5 and CMOD5.N) at spatial resolutions ranging from 50 m to 50 km using Sentinel-1 Interference Wide Swath (IW) mode images. In comparison with the buoy wind measurements, our results show that the CMOD5 function is the most effective wind speed retrieval algorithm, whose retrieved wind speeds have a low RMSE of 1.17 m/s and a bias of -0.28 m/s at a spatial resolution of 500 m. What’s more, a phenomenon has been reveal that the variance of wind speeds retrieved from co-polarized SAR images decrease exponentially with the decrease of spatial resolutions. For Sentinel-1 IW mode images, the variance of wind speeds retrieved with CMOD5 decrease rapidly from 50 m to 500 m, with a drop in RMSE of 40%, and thereafter levels off. Thus, a spatial resolution of 500 m, in concert with CMOD5 function, could be recommended for wind speed retrieval using Sentinnel-1 IW mode data.

In terms of assessing the effect of spatial resolution on the accuracy of wind speeds retrieved from cross-polarization images, due to the lack of in situ measurements, the wind speeds retrieved from VH-polarization images were compared with those from VV-polarization images at different spatial resolutions using RADARSAT-2 fine quad polarization images. Firstly, for wind speeds retrieved from VV-polarized images, the optimal geophysical C-band model (CMOD) function was selected among the four CMOD functions. Secondly, the most suitable C-band cross-polarized ocean (C-2PO) model was selected between two C-2POs for the VH-polarized image dataset. Then, the VH-wind speeds retrieved by the selected C-2PO were compared with the VV-polarized sea surface wind speeds retrieved using the optimal CMOD, which served as reference, at different spatial resolutions. Results show a similar accuracy trend as the co-polarized SAR data that the variance of VH-polarized wind speed decreased rapidly with the decrease in spatial resolutions from 100 m to 1000 m, with a drop in RMSE of 42%. However, the improvement in wind speed retrieval accuracy levels off with spatial resolutions decreasing from 1000 m to 5000 m. This demonstrates that the pixel spacing of 1 km may be the compromising choice for the tradeoff between the spatial resolution and wind speed retrieval accuracy with cross-polarized images obtained from RADASAT-2 fine quad polarization mode.

The fracture of the Larsen C ice shelf, which had been regularly observed in the last months of 2016, started to rapidly grow in 2017 so that, in February, only an ice strip, 20 km long, kept attached a huge section of the shelf to the Antarctic Peninsula. The final collapse, expected in 2017, occurred indeed between July 10 and July 12, with a loss of an area of some 6,000 km2, corresponding to about 9-12% of the entire shelf. According to the US National Ice Center (NIC) criteria, the calved iceberg was named A-68. In the frame of the ASI "COSMO-SkyMed Open Call for Science Initiative", we followed the initial phase of the A-68 iceberg drifting trajectory and melting process. The analysis covers a period of 6 months, with the database composed of a set of COSMO-SkyMed ScanSAR Huge images.

The global bathymetry at 1 km resolution is already known and available through multiple different data sets (e.g., GEBCO [1]). However, small-scale shallow water topographic features, like sand banks, reefs and bars, can change due to storms, and may not be correctly marked in the official charts. These changes can be relatively fast and are not easily measured by traditional sound surveying methodologies and, therefore, methods that rely on Earth Observation from space can be valuable for the monitoring of the coastal bathymetry. 

In this study, coastal bathymetry is derived with a wave-tracking algorithm through wave parameters retrieved from SAR images from the Sentinel-1 satellites. The wave tracking algorithm starts by applying the Fast Fourier transformation algorithm (FFT) at an offshore squared area with 1 km length to estimate the mean wavelength and wave direction. The mean depth of this offshore area is obtained by an independent source (GEBCO), to mark the limit of the intermediate waters. The algorithm then tracks down the shoaling waves until the wave breaking zone, using the wave direction retrieved from the 2D spectrum centered at each tracking point. Then, through the linear wave dispersion relationship, changes in depth are calculated based on changes in the wavelength between each position of the wave track. The output of the algorithm is a smoothed 2D bathymetry field that results from the interpolation of the depth computed at each tracking point.

This work is a part of a research application developed within the EU H2020 Coastal waters Research Synergy Framework (Co-ReSyF) project, to demonstrate the operational capabilities of the platform. 

Several case studies at different locations off the Portuguese Coast will be presented, using high resolution SAR images from Sentinel-1. Several issues regarding the synergy between SAR and Optical data to retrieve high-resolution bathymetry in shallower waters will be briefly discussed.

 

References:

[1] https://www.gebco.net/

Offshore wind farms influence the atmosphere downwind for considerable distances and SAR has been used to demonstrate this effect in earlier studies. SAR has also been used to characterize winds in the coastal zone influenced by the proximity to the land. The novelty of the present study — focusing on the Anholt wind farm in the Kattegat Strait between Denmark and Sweden — is the interplay between the wind farm wake effect and a strong horizontal wind speed gradient from the coastline and further offshore. The wind farm was constructed after the Envisat ASAR mission ended and prior to the Sentinel-1 mission.

The study is based on the SAR wind archive available at https://satwinds.windenergy.dtu.dk . The SAR data are processed at DTU using CMOD5.N and wind directions from the Climate Forecast System Reanalysis (CFSR) from 2002 to 2010 and the Global Forecasting System (GFS) from 2011 to 2017. The pixel size is 500 m.

The offshore winds observed by Envisat are compared to Weather and Research and Forecasting (WRF) model results. More specifically, the variation in wind speed along the first (western) row of wind turbines (stretching 20 km from north to south located closest to the Danish coastline) for winds coming from 265° (±15°) show good agreement between SAR and WRF. The wind speed variation along this row of turbines is also assessed based on the Supervisory Control And Data Acquisition (SCADA) data from the wind turbines kindly provided by Ørsted and Partners. The data are 10 minute values from January 2013 to June 2015. Interestingly, the average variation in wind speed along the western row for winds coming from 265° (±15°) show around 1 m/s higher winds at the northernmost turbines compared to the southernmost turbines. This difference is most likely caused by the varying distance to the coast. The fetch is up to 50 km in the north and down to 16 km in the south. Wind speeds relative to the center turbine from SAR and SCADA agree within 0.1% while WRF over-predicts around 1% as compared to SCADA. All data sets quantify a significant coastal wind speed variation.

The comparison of winds from SAR and SCADA (extrapolated from 81.6 m hub-height with logarithmic profile to 10 m) in non-waked and waked conditions give results of R² of 0.97 with RMSE of 1.80 and 1.70 m/s and bias -0.12 and -0.52 m/s, respectively.

An investigation of the wind farm wake effects is completed using the Envisat data versus the Sentinel-1 data to provide the difference between free-stream and wind farm wake conditions. Various horizontal transects aligned with the wind direction and perpendicular to the wind direction are analysed. The uncertainty of the average wind speed is also assessed and significant variations are mapped (Ahsbahs et al. in review, Wind Energy Science).

The conclusions are that SAR enables quantification of spatial horizontal wind speed gradients and wind farm wakes. The large SAR wind archive can be used to explore complex cases providing a measurement independent of modelling results.

The importance of monitoring internal waves (IWs) for oceanographic studies is hard to overestimate. IWs are responsible for transferring energy between large-scale tides and small-scale mixing. On the continental shelf IWs affect coastal areas through nutrient mixing in the euphotic zone, transfer of phytoplankton rich waters, coupling of benthic and pelagic systems, sediment re-suspension and cross-shore pollutant transportation. Observation of IWs provides insight into the mechanisms of their generation, propagation and dissipation. It helps to evaluate and predict the impact that IWs may have on the environment and ecology in UK shelf seas.
Traditionally, measurements of IW fields have been carried out using instruments deployed in the ocean. They are time consuming, expensive and unsuitable for large-scale observations. Synthetic Aperture Radar (SAR) remote sensing method provides an alternative to in-situ observations of IWs. SAR is insensitive to cloud cover and solar illumination and provides excellent spatial coverage and consistency of observations. SAR sensors produce maps of sea-surface roughness, within which the presence of underlying IW crests and troughs are revealed due to their effect on small capillary waves, seen as alternating light or dark lines in the radar backscatter. While SAR imagery has been frequently used to study particular IWs, usually the linear features are detected manually due to their indistinctness within the speckled noise inherent in such images. Manual analysis would not be practical or appropriate for objectively inferring the long-term distribution of IWs across a large sea area.
We evaluated many potential methods for automated IW detection from SAR data, reviewed methods for generating distribution maps of IWs, and established the existence and feasibility of detecting IWs within the UK continental shelf. We have identified methods from the literature that can be used to identify regions where IWs are likely to occur, and reviewed ways in which IWs can affect the seabed. We implemented and applied the new methods for automated detection of IW features to a 7-year time-series of ENVISAT ASAR sensor data acquired in 2006-2012 by the European Space Agency, to build detailed maps of IW occurrence and climatology for the UKCS. Up to a hundred SAR scenes per month covered the region of interest, over 3,400 in total. Regions with high likelihood of seabed disturbance were identified by combining the IW occurrence data with mixed layer depth and bathymetry. Monthly and annual climatology maps of the UKCS have been produced showing the spatial and temporal variability of IW interaction with the seabed.

Altimeters measure the topography of the surfaces they fly over, but in the Arctic context where sea ice can be found, it is important to determine whether the measurement corresponds to open ocean, ice floe or leads within the sea-ice. The knowledge of the surface type is important for data interpretation and the computation of the freeboard obtained by difference between the open water level and the floe surface level. Ice thickness estimate is then derived from the freeboard estimate. The assessment and then the optimization of altimeter-based sea ice lead detectors is thus a prerequisite to further improve the quality of ice thickness maps. This can be carried out using a set of references cases provided by Earth Observation images, such as the ones provided by Copernicus Sentinel-1 (S-1) images

 

In this study, a dedicated S-1 lead detector has been implemented. Its general principle is to detect local minima thanks to morphological reconstruction filter, followed by watershed algorithm using the sigma0 image gradient.

 

For the first time, a robust and consistent methodology for the joint assessment of SAR and Altimeter lead detector has been developed. Compared to the existing state of the art, it is proposed to fully account the 2-D geometric problem when comparing 1-D altimeter track and 2-D SAR image. The distance from nadir and the surface of leads within the altimeter footprint are especially considered when building the SAR-based reference database.

 

We collocated S-1 / SRAL Sentinel-3 data with time lag less than 8 minutes, corresponding to 169 S-1 images. The overall accuracy between the S-3 and S-1 lead maps is good with more than 90 % of good agreement depending upon how the SAR-based reference dataset is built. The False Positive Rate is very low with less than 3-6 %. A new criterion, the so called Matthews coefficient, is introduced, and provide new parametrization for the SRAL S-3 lead detector. A tradeoff between false alarms and good detection is numerically found.

 

A comparison with collocated French-Indian SARAL AltiKa altimeter and S-1 data is also provided. A similar analysis is performed using AltiKa SARAL altimeter data, showing enhanced performance for SRAL S-3 data acquired in Delay-Doppler mode with reduced off-nadir returns.

This study has been partly funded by the European Space Agency (ESTEC/ESA) via the S3-CD studies, by the PEACHI SARAL contract and the French Space Agency CNES.

The Barents Sea is a region where maritime operations will be more frequent in the coming years. Icebergs can be a threat to these operations which makes it crucial to have knowledge about their position at any time. Most methods for automatic iceberg detection are focused on the detection of icebergs much larger than the typical size found in the Barents Sea. Therefore they are not applicable to this area. They also tend to cover only one specific environment of the iceberg.

Marino et al (2016) proposed a method called DPolRAD that utilizes the polarization properties of the SAR signal to distinguish icebergs from sea ice. It uses the different polarimetric behaviour in cross- versus co-polarized channels to reduce surface scattering (i.e. sea and sea ice) and enhance volume scattering (i.e. icebergs). Here this method has been applied to Sentinel-1 EW SAR images over Svalbard and Franz Josef Land to detect icebergs in the Barents Sea, with the modification of applying a gaussian filter instead of a boxcar filter in the calculations. Together with a statistical detector called the constant false alarm rate (CFAR), automatic iceberg detection is performed in three cases; open water, fast ice, and pack ice.

Four probability density functions (PDFs) are tested as input to the CFAR detector; the gamma-, generalized gamma-, normalized inverse gamma-, and the gaussian distribution. The parameters for each PDF are calculated using the background data surrounding the pixel under test, and the choice of variables for the DPolRAD and CFAR methods are based on consistency, scatter plots, and confusion matrices. The Google Earth Engine is used to colocate SAR and optical images for validation of iceberg pixels and for manual detection. This is useful to verify the detection results.

The results show that the normalized inverse gamma-PDF is most representative for the data calculated by the DPolRAD method. This is also the PDF that most easily can distinguish iceberg pixels from non-iceberg pixels, and by applying the PDF to the DPolRAD and CFAR methods we obtain high detection rates.

With this detector it may be possible to do systematic surveillance of icebergs in the Barents Sea, both using past and future data, which can give information about their distribution and path.

Reference:

A. Marino, W Dierking, and C. Wesche, “A depolarization ratio anomaly detector to identify icebergs in sea ice using dual-polarization SAR images,” vol. 54, no. 9, pp. 5602–5615, 2016.

The polar regions are distinctly different from other geographic areas and share a common set of characteristics, including remoteness; limited accessibility; low population densities; sensitive ecosystems of global importance; and increasing economic activity. In recent years, the uniqueness of these polar regions and their importance to the world has been recognized. This is made more urgent by the fact that the Arctic is warming much more rapidly than other regions of the world, and the summer sea ice is thinning and shrinking significantly. Such physical change can best be monitored by remote sensing from space. Although a number of satellite derived approaches exist, they need to be adapted to the new generation of sensors and the data provided need to be further validated by ground truth.

The early autumn voyage of RV Sikuliaq to the southern Beaufort Sea (Oct-Nov 2015) and the PIPERS winter cruise in the Ross Sea (May-Jun 2017), offered very favorable opportunities for observing the properties and thicknesses of frazil-pancake ice types. The operational regions were overlaid by a dense network of retrieved satellite imagery, including SAR (synthetic aperture radar) imagery from Sentinel-1 and COSMO-SkyMed (CSK). This enabled us to fully test and apply the SAR- waves technique, first developed by Wadhams and Holt (1991), and experimentally verified by Wadhams et al. (2004) for deriving the thickness of frazil-pancake icefields from changed wave dispersion. A line of sub-images from a main SAR image is analysed running into the ice along the main wave direction. Each sub-image is spectrally analysed to yield a wave number spectrum, and the change in the shape of the spectrum between open water and ice, is used to retrieve frazil-pancake ice thickness. For each of the case studies considered here, there was good or acceptable agreement on thickness between the extensive in situ observations and the SAR-wave calculation.

In this study, the sensitivity of environmental and sensor parameters that rule microwave backscattering of sea oil seeps is investigated by contrasting analytical models with actual satellite synthetic aperture radar (SAR) measurements.

The topic that will be addressed is of paramount importance in many marine and maritime applications as remotely sensed oil seep observation, that is very interesting not only from a scientific point of view, but also from an economical and environmental perspectives being related to oil and gas exploration and extraction activities. Moreover, the problem of sea oil seep backscattering, i. e., the backscattering from reserves of hydrocarbons naturally coming up from the bottom of the ocean to sea surface, is rather challenging and needs further investigation since the processes that lie at the basis of the escape, the development and the lifespan of oil seeps are extremely variable and irregular. The latter result in a quite different backscattering modeling approach if compared to conventional scattering models that deal with sea surface backscattering with and without oil slicks.

Hence, in this study, the two scale Boundary Perturbation Method (BPM) is adopted as a reference scattering model to predict sea surface backscattering and, then, the effects of the presence of the oil seep, that include damping properties, reduction of friction velocity and changes in the dielectric permittivity are taken into account. The firsts are considered including the rheological parameters of different kinds of surfactants, i. e., weak-damping biogenic films and strong-damping crude oil; the second is included by applying a penalty factor to slick-free sea surface friction velocity and the third is accounted for by modeling the heterogeneous oil seep according to the effective medium approximation theory, i. e., different kinds of mixture of seawater and oil droplets are considered.

Then, the influence of all of those parameters to sea oil seep backscattering is evaluated for different SAR acquisition parameters, i. e., incident wavelength and angle of incidence, sea state conditions, i. e., wind speed, and surfactants’ properties, i. e., damping and concentration. Preliminary results, obtained by contrasting model’s predictions and actual backscattering SAR measurements collected at L-, C- and X-band over well-known sea oil seeps, demonstrated that wind speed, damping properties and incidence angles most affected the backscattering from sea oil seeps, while incident wavelength and the adopted oil/water mixture have a negligible impact on their backscattering.

Oil spills and discharges represent a significant ecological risk to the marine environment. As a result, having timely and accurate information on the location and evolving state of slicks can aid in clean-up efforts as well as to inform future preventive measures. For this purpose, Synthetic Aperture Radar (SAR) imagery has proven to be an indispensable tool given its insensitivity to overlying cloud cover. A notable feature of oil in SAR images of the ocean surface is that oil will appear as a dark patch in relation to the surrounding sea surface if wind speed is sufficiently high.  This is due to the damping of the wind induced capillary waves which causes the oil infested region to become smoother than the surrounding sea surface. By applying theoretical scattering models to SAR imagery of oil infested waters it is investigated in this paper, if, and under which conditions, variations in the estimated dielectric constant can be detected over an oil slick. The dielectric constant is a quantity that is intrinsically related to the material in question and has a larger value for sea water than mineral oil. By inverting for the dielectric constant, it is believed that specific internal zones within slick will be observed that correspond to regions where emulsions of various oil/water ratios are present. Inverting for the dielectric constant in this way can yield information on the quantity of oil present within each resolution cell.

A series of images of oil spills of various types and emulsions were taken during oil-on-water exercises for the years 2011-2016 off the Norwegian coast using a variety of sensors such as Radarsat-2 (RS-2), TerraSAR-X (TSX) and UAVSAR. These sensors correspond to the X, C and L bands of the electromagnetic spectrum. For the derivation of the dielectric constant the Small Perturbation Model (SPM), the Extended Bragg Model (X-Bragg) and the Composite model are inverted using a look-up table approach. The linear mixture model, which describes the variation in the complex dielectric constant as the oil emulsifies is used to invert for the volumetric content of oil within slick.

Current work done on quad polarimetric RS-2 observations for the oil-on-water exercises for the year 2011 using the SPM for a period of time when the wind speed was within the validity range of this model suggests that an inversion for the dielectric constant may be possible over the oil infested areas but that the dielectric constant may be underestimated over ocean regions. This is likely due to the insensitivity of the model for high values of the dielectric constant. Current work is being conducted to reliably remove the effects of thermal noise which is induced by the operation of the sensor and manifests itself as an additive signal in the form of radar cross section.

Future work involves inverting for the dielectric constant using the other two methods cited above and using acquisitions that were taken for different wind speeds and incidence angles. 

The routine monitoring of oil spill discharges from spaceborne SAR observation was demonstrated in the early 2000’s and is since an operational service managed by the European Maritime Safety Agency (EMSA) as part of the CleanSeaNet service. The CleanSeaNet service is based on radar satellite images, covering all European sea areas, which are analyzed to detect possible oil spills on the sea surface. When a possible oil spill is detected in national waters, an alert message is delivered to the relevant country. Analyzed images are available to national contact points within 30 minutes of the satellite passing overhead [1].

The definition and evolutions of this operational service were supported by the realization of a set of R&D studies and projects ranging (ESA / Roses, FP7 Marcoast [2], FP7 Sea-U [3]). The initial challenge was first to demonstrate and improve the responsiveness of such service for the oil spill detection. It then evolved to a more complete service supporting the polluter identification as well, by applying drift models to the detected spills and by correlating with AIS information.

For now, approximately 2,000 images are ordered and analyzed per year [1] on behalf of EMSA. This does not include all the potential images to be acquired over the European areas of interest. The service is performed by several Service Providers applying a panel of automatic processing required to collect the appropriate knowledge of the maritime situation over the area of interest for this service (vessel detection, wind measurement, swell measurement) and supervised oil spill detection and polluter identification. This latest supervised step implies that each single image under analysis needs to be inspected by a specially trained SAR image analyst, which then create constraints and costs on the operations. This supervised step is required for ensuring the quality of service in terms of detection and false alarm rates compatible that cannot be achieved by operational method so far.

This paper will present the challenges of full automatic oil spill detection including: the need of a reference data base of pollutions, a stringent requirement on computation time for near real time operational service, and the various ways to tackle the problem (ad hoc techniques, machine learning, deep learning…). Some preliminary results and the program of work to achieve a complete automatic service will be exposed as well.

[1] Ten year of CleanSeaNet service: http://emsa.europa.eu/csn-menu.html

[2] The Marcoast project: http://www.copernicus.eu/projects/marcoast

[3] The Sea-U project http://www.copernicus.eu/projects/seau

Vessel detection and route characterization is of interest for various services ranging from fishing monitoring, border control and traffic monitoring. The ship wake detection is usually applied as a post processing to the vessel detection itself and is a mean to estimate the route of the vessel together with its speed by measuring the Doppler shift between the turbulent and the vessel itself. Various wake detection techniques are proposed usually based on Radon transform [1,2,3].

We previously defined and developped a detection of wake signature in the Hough transform domain. The Hough transform aims to accumulate evidences of presence of the object of interest. The Hough transform for the detection of linear features is very similar to the Radon transform. In fact, both the Radon transform, and the Hough transform correspond to the same transformation.

In this paper we present an optimized Hough/Radon wake detection algorithm based on practical considerations of the SAR imaging mechanism. More specifically, the Doppler shift between the vessel and its wake is driven by the radial speed of the vessel, that is itself strongly correlated to the orientation of the wake. It is then possible to strongly reduce the domain of detection of the wake signature in the Hough/Radon transform domain to reject wakes in the proximity of the vessel of interest but that cannot physically relate to it. This is of typical interest for dense traffic areas, low incidence angle and high resolution, were multiple wakes may be observed.

In addition to this considerations on radial velocity, we introduce the use of statistics on AIS messages (speeds, heading) as well to further constrain the detection of the wake in the Hough/Radon domain. Indeed, in some regions of the maritime domain the maritime traffic is strongly driven by a set of Traffic Separation Schemes (TSS) imposing specific vessels heading. For such areas, it is then possible to precompute histograms of speeds and heading from databases of AIS data, that can be used as a priory probability density functions to be considered in the wake detection scheme.

In this paper, both classical Hough/Radon and improved (using statistical AIS information) are evaluated in terms of probability of detection and false alarms and compared to a recent paper dealing of wake detection on Sentinel-1 images [3].

[1] Rey MT et al, Application of Radon Transform Techniques to Wake Detection in Seasat-A SAR Images. IEEE Trans. Geosci. Remote Sensing 1990, 28, 553-560

[2] Copeland A.C et al, Localized Radon Transform-Based Detection of Ship Wakes in SAR Images. IEEE Trans. Geosci. Remote Sensing. 1995, 33, 35-45

[3] Graziano MD, et al, Performance Analysis of Ship Wake Detection on Sentinel-1 SAR Images, Remote Sensing, MDPI, 2017 

Vessel detection and route characterization is of interest for various services ranging from fishing monitoring, border control, traffic monitoring [2] and oil spill monitoring [1]. The first step of vessel detection is to detect the bright echoes in the SAR images [3]. However, it is then of interest to reject the bright points that are not directly related to a vessel. Such kind of points can be related to various phenomena or object. For instance, cities may produce range or nadir ambiguities in the maritime domain observed by SAR that may impeach the detection of small vessels in the impacted areas. Such artefacts are more frequently observed on very high-resolution products, but we also characterized this on Sentinel-1 Interferometric Wide Swath products. In addition, some objects like oil rigs, or fish farms shall not be confused to vessels by automatic processing.

In this paper we present a method of detection and discrimination of both range and azimuth ambiguities of coastal objects and permanent echoes. This method is based on the multi temporal stacking of SAR images acquired on the same area of interest. Some methods were previously proposed to detect such specific points, but they are usually applied as a post processing of vessel detection. Our method applies object detection after a full multi temporal stacking (not before), then enabling to enhance the contrast over the permanent points by taking advantages of wind variability with the time in the maritime domain. It then allows to characterize smaller objects than the one spotted by standard techniques based on post-processing of large number of vessel detection reports.

The main result of this paper is a set of maps illustrating the permanent artefacts and objects over the Mediterranean seas for various Sentinel-1 configurations. Those maps can be used to filter out permanent echoes from standard vessel detection techniques. Furthermore, the method allows to produce maps of range/azimuth and permanent points that can be used as denoising information before application of vessel detection algorithm.

 

[1] Ten year of CleanSeaNet service: http://emsa.europa.eu/csn-menu.html consulted on 26 January 2018

[2] The FP7 Dolphin project: http://gmes-dolphin.eu/

[3] G.Hajduch et al, Ship Detection: from processing to instrument characterization, SeaSAR 2008

Accurate high resolution measurements of ocean surface stress are required to improve predictions in areas of strong atmosphere-ocean coupling. Scatterometer observations have shown that there is significant covariability between wind stress and sea surface temperatures and that 10-m winds are affected by ocean currents. These studies used relatively low resolution satellite scatterometer and microwave radiometer data along with temporal averaging and high pass filtering to resolve the atmosphere-ocean coupling. The reduced wind speed over cooler side of a front produces increased curl for wind parallel to the front and increased divergence for wind perpendicular to the front. Errors are introduced since the 10 m derived winds are relative to the moving ocean and the local sea state and surface currents are neglected.

The Western Arabian Sea is characterized by a seasonally reversing circulation influenced by strong monsoon winds. During the winter monsoon, the southward flowing Somali Current merges with the northward flowing East African Coastal Current to form the eastward flowing South Equatorial Counter Current. This pattern shifts during the summer monsoon as the East African Coastal Current feeds the Southern Wedge, Somali Current and Great Whirl. The summer monsoon winds are influenced by the ocean currents and coastal upwelling over the Western Arabian Sea. This covariability influences the moisture and heat transfer into the troposphere where it can affect the rainfall over India.

Sentinel-1 synthetic aperture radar data were acquired in early June over the Somali Current during the 2016 and 2017 summer southwest monsoons. The upwelling of cold water along the Somali coast, the Southern Wedge, and Great Whirl are typically well developed by June. Wind stress curl computed from the SAR derived wind field clearly show sub-mesoscale features along the Somali Current and the Great Whirl that are correlated with patterns in the SST data. The SAR data shows that during the 2017 Summer Monsoon, the Great Whirl and Southern Wedge developed later than in 2016. In 2017, the Somali Current and the Southern Wedge were broader and the Great Whirl was shifted significantly to the southwest. The result was that the covariability was less distinct in early June 2017 than it was in 2016. The wind stress curl also shows a bifurcation in the Somali Current in 2017, where part of the current forms the Great Whirl. Sentinel-1 data for the 2017 monsoon season were supplemented with Radarsat-2 and TerraSAR-X data to characterize the intraseasonal variability of the ocean-atmosphere coupling.

Application of the Sentinel satellites for tracing surface slicks along the Irish Western Shelf

 Mullins, M.¹and Croot. P.L¹

¹ Earth and Ocean Sciences, School of Natural Sciences, National University of Ireland Galway

 

A key challenge facing the maritime industry at present is developing the capability to distinguish in real time, maritime accidents from naturally occurring phenomena such as phytoplankton blooms and natural oil seeps. The ability to correctly determine the nature of the observed surface slick will facilitate applications in maritime safety, oil exploration, harmful algal blooms/ecosystem services and other marine based activities. This project aims to link detection and monitoring of these natural and manmade surface slicks using satellite based observations incorporating the new ESA sentinel series of satellites.  For example, the Sentinel 1 satellites, incorporating SAR bands, allow for the determination of surface slicks even under the cloudy conditions typically found along the Irish west coast. However SAR by itself is unlikely to accurately predict the nature of the observed slick and so predictive assessments will be made by combining SAR with ocean colour (Sentinel 3) and scatterometer winds (Sentinel 1). Overall our work has 4 main objectives:

Determine criteria for Good Environmental Status (GES) as per the EU Marine Strategy Framework Directive regarding the optical properties of seawater.

Identify CDOM components produced by natural blooms of phytoplankton in Irish waters. 

Develop a spectral library of the optical properties of common maritime pollutants in seawater.

Develop an informed data analysis system for the identification of surface slicks from satellite data.

This publication/presentation* has emanated from research supported in part by a research grant from Science Foundation Ireland (SFI) under Grant Number 13/RC/2092 and is co-funded under the European Regional Development Fund and by PIPCO RSG and its member companies.

Gaofen-3 (GF-3), the first Chinese civil C-band synthetic aperture radar (SAR), has been successfully launched by the China Academy of Space Technology (CAST) on 10 August 2016. Among its 12 imaging modes, wave mode is designed to monitor the ocean surface wave over the open ocean. An empirical retrieval algorithm of significant wave height (SWH), termed QPCWAVE_GF3, is developed for quad-polarized SAR measurements from GF-3 in wave mode. QPCWAVE_GF3 model is built using six SAR image and spectrum related parameters. Based on a total of 2,576 WaveWatch III (WW3) and GF-3 wave mode match-ups, 12 empirical coefficients of the model are determined for 6 incidence angle modes. The validation of QPCWAVE_GF3 model is performed through comparisons against independent WW3 modelling hindcast, and observations from altimeters and buoys from January to October in 2017. The assessment shows a good agreement with RMSE between 0.5 m to 0.6 m, and SI around 20%. In particular, application of QPCWAVE_GF3 model on SWH estimation for two storm cases from GF-3 data in wave mode and Quad-Polarization Strip I (QPSI)mode are presented respectively. Results indicate that the proposed algorithm is suitable for SWH estimation from GF-3 wave mode, and promising for other similar data.

Wind information retrieval over the sea surface from microwave sensor data has a long and successful history. The accuracy of geophysical model functions (GMFs) applied to relate the microwave backscatter to wind speed and direction in 10m height has been continuously improved from early scatterometer missions to contemporary spaceborne SAR sensors like the current Sentinel-1 (S-1) pair. The statistical accuracy of SAR-based wind speeds is nowadays better than 2m/s and therefore the data is appreciated by met-ocean modellers to validate and improve their models.

State of the art high resolution SAR systems can provide wind fields with a resolution of less than 100 m which have been successfully validated against co-located LiDAR data. SAR-based data is particularly interesting for the offshore wind industry as it combines a resolution near to LiDAR capabilities and a large cross-track coverage of up to 250km for e.g. the two S-1 satellites. The S-1 systems work in TOPS SAR imaging mode, which has been designed to reduce the scalloping effect in burst mode SAR imaging, e.g. ScanSAR. The large spatial coverage of these systems allows wind park operators to get regular snapshots of entire marine regions such as the German Bight and cross-validate model data especially with focus on the impact of wind shadows of adjacent turbine clusters.

A close inspection of the Sentinel-1 Interferometric Wide Swath (IW) mode data over the ocean reveals antenna beam patterns remaining in the calibrated and noise-corrected data, resulting in discontinuous wind speed values across the beam boundaries. While of minor importance when calculating coarse wind fields of several kilometres resolution, high resolution wind fields as desired by offshore wind operators suffer from this uncertainty. A compensation for these errors contained in S-1 IW wind fields is believed to further increase the acceptance of SAR-bases wind data in both, industrial users of the data and the Numerical Weather Prediction (NWP) community for assimilation in met-ocean models.

We present an analysis of the impact of this effect on estimated wind fields based on more than 8000 Sentinel acquisitions with level-2 OCN data available and compare to high resolution ECMWF model data. We propose a correction scheme to minimize the beam pattern impact. The method can be applied to both, level-2 OCN data products and level-1 original IW image data.

The study of intertidal zone close to coastal areas is of paramount importance due to both anthropomorphic activities and natural phenomena, which threaten the stability of land and safety of the people. However, the monitoring of intertidal zone is not trivial due to the presence of mud flats and aquacultures. In fact, field surveys are difficult to perform in mud flat areas because of weather, tidal conditions and other natural factors.

Within this context, remote sensing plays an important role for intertidal zone monitoring. Optical images have the great advantage of being simple to interpret and they are easily obtainable. However, optical radiation is severely affected by cloud cover, solar illumination, and other adverse meteorological conditions. These problems can be solved using radar sensors, which guarantee all-day and almost all-weather acquisitions, together with a wide area coverage. In particular, the Synthetic Aperture Radar (SAR) can be very useful for intertidal zone monitoring purposes, because of its fine spatial resolution.

The main goals of this study are to develop multi-polarimetric methods to analyze scattering mechanisms that characterize mud flat area using full- and dual-polarimetric SAR data.

The first part of this study is based on the detection of the mud flat area undertaken according to the Polarimetric Notch Filter and the change detection approach proposed in [1] and [2], respectively. The extended Bragg model [3] is adopted as a reference to model sea surface polarimetric covariance / coherence matrix.

Experiments, undertaken on actual SAR data collected over the intertidal zone near Jiangsu, China, by the C-band RadarSAT-2 and X-band TerraSAR-X missions show that the proposed methodologies, well detect the mud flat area and the hard objects deployed on them. Furthermore, a detailed analysis shows that full-polarimetric channels performs best in terms of land / water discrimination when compared to the dual-polarization channels.

 

[1] A. Marino, (2013), “A Notch Filter for Ship Detection With Polarimetric SAR Data", IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 6(3), 1219-1232.

[2] A. Marino; S.R. Cloude, and J. M. Lopez-Sanchez, (2013), “A New Polarimetric Change Detector in Radar Imagery”, IEEE Transactions on Geoscience and Remote Sensing, 51(5), 2986 -3000.

 

[3] I. Hajnsek, E. Pottier, and S. R. Cloude (2003), “Inversion of surface parameters from polarimetric SAR”, IEEE Transaction on Geoscienze and Remote Sensing, 41(4), 727–744.

For an adequate security of the maritime domain is necessary to be aware of the objects location and their activities, e.g. ship traffic, in relevant areas of the sea. TerraSAR-X, a satellite radar imager operating at X-band, is a powerful tool to detect maritime objects, e.g. ships, oil platforms, icebergs etc., which can raise potential risk for maritime traffic and environment. The use of satellite radar to create such awareness has the benefits, among many, that can detect also no self-reporting objects; it operates almost independently of cloud cover and beyond coastal ranges. However, radar images are less easy to interpret and carry no direct objects identification information, like it happens in optical images or messaging report systems such as Automatic Identification System (AIS) and Long Range Identification and Tracking (LRIT). Therefore, the challenge is to classify the different types of maritime objects based on the radar signal only. Different machine learning techniques are here investigated to this end.

In this paper the problem at hand is restricted to the classification of 5 common type of maritime object in Synthetic Aperture Radar (SAR) images, i.e. tanker ship, cargo ship, windmill turbine, offshore platform and harbour structure. Cargo and tanker represent are the most correlated in terms of size and mechanical structure, so the most challenging to discriminate. Indeed windmill and platform are yet static object but with the increasing offshore energy production the amount and geographical location is not always available, i.e. new structure being built, removed or moved. Land masking process in high resolution SAR ship detection might suffer from false alarms of harbour structure not included in medium resolution landmask database and therefore being able to identify this class gives the opportunity to reduce this risk.

The Multi Layer Perceptron (MLP) and Convolutional Neural Network (CNN) models are the two feature extractors and classifiers used to accomplish this task. The performance of the two models are evaluated depending on the depth of the network layers and resolution of the input image using a relatively large classification dataset composed of real SAR samples extracted from TerraSA-X images.

The ocean surface circulation is responsible for significant transport of heat, salt, passive tracers and ocean pollutants. On basin scales, surface currents and their variations are major players in climate and weather fluctuations. Surface currents impact the steepness of surface waves and are thus important for generating reliable marine sea state forecasts for maritime operations including fisheries and ship routing. Because of their significance in advecting passive particles, knowledge of surface currents is also important for oil spill and marine debris response actions, as well as search and rescue operations. Assimilation of surface current data in ocean circulation models is a common practice today, thanks mostly to the regular availability of satellite altimetry data and surface drifter data.

During the last decades surface current observations have also emerged from coastal high-frequency (HF) radar and satellite synthetic aperture radar (SAR) observations, thanks to the Doppler shift measurements. HF-radar systems derive hourly surface current observations in the near-coastal regions at a radial distance of up to 200 km offshore with a spatial resolution from 500 m to 6  km.  Between 2005 and 2010, an HF-radar (CODAR) was operating at the Fedje island on the west coast of Norway. The range coverage of this system was nearly 50 km.

In comparison, the range Doppler shift obtained with SAR can be used to derive a signal related to the range directed surface current across the swath (~500 km) at a spatial resolution of 1-10 km. According to previous theoretical analyses, the velocity associated to the geophysical Doppler shift can be approximated as a linear sum of wind-wave induced sea surface velocities and the sea surface current in the range direction. The wave-state contribution can be estimated using an empirical relationship between the range Doppler velocity and the near surface wind field. When the wave-state contribution is then subtracted from the observed geophysical Doppler shift, a measure of the surface current in the range direction can be obtained.

In this presentation, the surface current retrieved from Envisat ASAR data is assessed in comparison to the CODAR SeaSonde measurements at Fedje.

In the Mediterranean Sea, the mauve stinger Pelagia noctiluca is probably the most frequent jellyfish species, which appears to be blooming with increased frequency, and many others jellyfish species appear to be thriving, with significant impacts on fishing and other human activities.

The main current practical way of assessing the presence and the abundance of gelatinous macrozooplankton is through visual detection, in particular through skin diving, boats, airplanes, beaches, video cameras and citizen science campaigns. It is also possible to detect jellyfish blooms through indirect methods such as echosounding records and radio tracking.

The large spatial extent of jellyfish blooms suggests that some form of remote sensing might be an efficient way to provide an early warning system for such phenomena. A recent study has been performed using cloud-free EO ocean colour scenes.

This study aims to explore the potential of Synthetic Aperture Radar for detecting jellyfish blooms. The advantages of SAR are that it obtains high-resolution remote sensing synoptic images covering broad marine domains, and it can be used day and night under all weather conditions.  The detection is based on the damped radar return produced by biogenic surface slicks such as those produced by the presence of plants and animals in the ocean.

Firstly, jellyfish blooms datasets in the Mediterranean Sea have been acquired using information collected during EU-funded projects such as the MEDJELLYRISK one (www.jellyrisk.eu) and also through ad hoc citizen science campaigns as the Spot the Jellyfish one in Malta (www.ioikids.net/jellyfish). The sighting datasets are well-supplemented with metadata, including information about geographic position, species and abundance.

When available, SAR Sentinel 1 images corresponding to ongoing jellyfish bloom events are acquired and pre-processed in order to remove speckle noise that constrains image interpretation and further processing of the image. Finally the potential identification of jellyfish blooms is attempted through an adaptive thresholding and a clustering through the closest pixels method, similar to the one used to detect oil spills. A case study is presented, relating to a Pelagia noctiluca bloom identified along the south-eastern coast of the Maltese Islands in June 2016.

We propose a feature-tracking algorithm for sea ice drift retrieval from a pair of sequential Sentinel-1 wide-swath images. The method is based on feature tracking comprising feature detection, description, and matching steps. The approach exploits the benefits of nonlinear multiscale image representations. We evaluated several state-of-the-art feature-based algorithms, including A-KAZE, Scale Invariant Feature Transform (SIFT), and a very fast feature extractor that computes binary descriptors known as Oriented FAST and Rotated BRIEF (ORB) on dual polarized Sentinel-1A C-SAR extra wide swath mode data over the Arctic. The A-KAZE approach outperforms both ORB and SIFT up to an order of magnitude in ice drift. The experimental results showed high relevance of the proposed algorithm for retrieval of ice drift at subkilometre resolution from a pair of SAR images with 100-m pixel size. We found that feature tracking using nonlinear scale-spaces is preferable due to its high efficiency against noise with respect to image features compared with other existing feature tracking alternatives that make use of Gaussian or linear scale spaces.

The developed ice tracking system perfomance is demonstrated by an operational ice drift and deformation analysis over Ob's Bay (SW part of Kara Sea) for winter season of 2016-2017 using Sentinel-1A data to support navigation in sea ice. Some new ice regime characteristics on ice kinematics is obtained.

Sea surface wind is one of the key components in the study of waves, currents, ocean circulation, and atmospheric-ocean interactions, providing us with a overall understanding of complex marine phenomena. As interest in climate change has increased, the importance of observing the global wind field has been emphasized and researchers have been producing global wind fields data using scatterometer. These scatterometer data are accurate to around ±2 m/s for wind speed and 20˚ for wind direction, but they have disadvantages that are deficiency of coastal wind field data and impossibility of analysis of small scale marine phenomena due to low spatial resolution. In this study, the sea surface wind in the coastal region of Korea was calculated from Sentinel-1A/B Interferometric Wide swath (IW) mode VV polarization data and the accuracy of each algorithm was assessed. From May 1, 2015 to September 30, 2017, 395 Sentinel-1A/B data were processed, land masking was performed using Shuttle Radar Topography Mission (SRTM) Digital Elevation Model (DEM) data, ship was detected and removed by applying the adaptive threshold method, and in order to remove speckle noise of the Sentinel-1A/B image itself, an ensemble average was calculated by setting a moving window. CMOD4, CMOD_IFR2, CMOD5, CMOD5.N, and CMOD5.Na algorithms were applied to the preprocessed Sentinel-1A/B data. The results show that the root mean square error (RMSE) and bias are 1.83 m/s and –0.64 m/s for CMOD4, 1.82 m/s and –0.59 m/s for CMOD_IFR2, 1.69 m/s and –0.38 m/s for CMOD5, 1.68 m/s and 0.31 m/s for CMOD5.N, and 1.65 m/s and 0.14 m/s for CMOD5.Na, and CMOD5.Na algorithm showed the best simulation of sea surface wind in the seas around the Korean Peninsula. By region, the Yellow Sea showed an increase in bias compared to other region, which is interpreted as a result of the change in sea surface roughness due to low water depths.

The Mediterranean sea is known for the high spatial variability of its surface wind. The fluxes are constrained by the numerous reliefs very near the coasts and the effect of the islands, creating sustained local winds. Moreover autumn and winter seasons bring heavy storms and Medicane (Mediterranean hurricane) that enhance the surface fluxes and make even more complex wind fields.

The atmospheric mesoscale models are more and more able to simulate a variable surface wind by a fine representation of the relief and a capacity to simulate marine breeze due to diurnal cycle. The non-hydrostatic model of Meteo-France, Arome, run for instance operationally at a resolution of 1,3 km. Its surface fields are very realistic but can suffer of an incorrect localisation of the processes with a mismatch greater than its fine scale. In particular during a convective event, the very fine stormy cells can be shifted of some kilometers compared to reality.

The hig resolution SAR winds present the advantage to give a very precise measurement of surface wind intensity. The use of such satellite wind observation has a very important advantage for waves forecast in coastal areas in the Mediterranean sea. In this work the quality of the SAR winds are evaluated by comparison of Arome wind as input in a wave model. For that purpose, we use a MFWAM configuration at 2,5 km and a configuration of Wavewatch III around the french Mediterranean coasts. The WW3 configuration has been developped and implemented in operations at Météo-France in 2015 in the framework of the HOMONIM project, by Météo-France and Shom. It uses an unstructured grids up to a resolution of 200 meters.

The SAR winds are integrated in the Arome wind fields. Both simulations with Arome and merged wind forcing are conducted on particular events that happened in the french Mediterranean coast in 2016 and 2017. The sea state outputs are compared with altimeter and buoy observations. The comparison shows the good skill of SAR winds and the benefit of using more accurate winds than 1,3 km for a high resolution wave model like the WW3 configuration.

Further conclusions and results will be discussed in the final paper.

Space borne Synthetic Aperture Radar (SAR) is used for operational surveillance of ocean areas and oil spill detection [1]. Oil spills are frequently detected around oil platforms due to the releases of so-called produced water (see, e.g., [2]). This is water from the reservoir that has been separated from the oil and gas at the platform, and which is in part released into the sea. Release of produced water is legal, within given limits. The water still contains oil and can form surface slicks, similar to other oil spills.

Understanding the signatures of produced water and how they are related to, e.g., the relative oil volume and/or concentration, and to environmental conditions, can be helpful for the operational services. For example, distinguishing a “normal” release of produced water from an “abnormal” release (elevated amounts) in a SAR image is currently an unsolved problem. Very little research on these topics have been done before.

In this study, we investigate SAR data acquired over the oil platform Brage, located in the North Sea. In addition to the SAR data, in situ information on wind conditions and oil releases have been obtained from the platform operator. This type of data allows for comparison of SAR signatures of produced water of the same oil type, acquired under varying conditions. The aim of the study is to evaluate how the SAR signatures vary with, e.g., the amount of oil in the release and wind speed. Both wide swath low resolution ScanSAR imagery (Radarsat-2 and Sentinel-1) and high-resolution quad-polarization data (Radarsat-2) are available. A comparison between the two data types, in order to evaluate any potential advantage of the high-resolution data in characterization of the releases, is also of interest.

[1] Solberg, A. H. S. (2012), Remote Sensing of Ocean Oil-Spill Pollution, Proc. of the IEEE, 100(10), 2931-2945, doi: 10.1109/JPROC.2012.2196250.

[2] Espedal, H. A., and Johannessen, O. M. (2000), Detection of oil spills near of shore installations using synthetic aperture radar (SAR), Int. J. Rem. Sens., 21(11), 2141-2144, doi: 10.1080/01431160050029468.


The grounding of sargassum rafts is affecting multiple regions in the Caribbean Sea and in the Gulf of Mexico. Large sargassum rafts are drifting over long distance, probably from the Bresilian coast of Gulf and Guinea. The precise cause of this phenomena is not well known, even if anthropic sources (pollution from the Amazon) are suspected. The impact of this grounding on the coastal area is extremely bad as the decomposition of the algae produces hydrogen sulfide: a nauseous gaz that can be toxic at high concentration. The grounding of sargassum rafts induces significant decrease of touristic activities in the impacted area, with a noticeable impact on the local economy. Monitoring the sargassum rafts before their grounding is of interest in order to deploy preventing actions (catching the rafts at sea before a predicted grounding).

In this paper we present an operational demonstration of sargassum monitoring deployed in April 2016. This service is based on:

(1) observation and detection of the rafts on both Sentinel-1 SAR data and Landsat-8 optical data. The signature of Sargassum in SAR images is an increase of the local backscattering due to an increase of the surface roughness. The spectral signature of the sargassum in the optical domain was studied and panchromatic images were produced considering a dedicated composition of blue, near infra red (NIR) and  short wave infra red (SWIR). Both types of sensors (SAR and Optical) were used as complementing means of observation considering the capacity of the SAR to observe under the clouds and by night, while the optical sensor provide opportunities to capture signatures at higher resolution and different mean solar time.

(2) prediction of their drift: A drift model considering lagrangian advection under the effect of wind and current was tuned and used in order to predict the drift

(3) diffusion of the observed and predicted drift over a period of 72h ahead on web portal. This portal allows displaying the various satellite images, the delineation of sargassum rafts, the predicted drift, the generation of formal reports, etc

This demonstration allowed to evaluate the capacities of sargassum detection using SAR and optical imagery. More specifically, the detection performances with respect to incidence angle and polarization were assessed, the requirement on revisit frequency was refined so as the requirement on delay of delivery of analysis of each satellite image. This kind of service can benefit from additional medium resolution observation for instance from Modis and Sentinel-3 OLCI.

Aknowledgement: this demonstration service was partially funded by CNES.

Ocean surface wind is an essential parameter for ship navigation and the study of ocean. Satellite sensors such as synthetic aperture radar (SAR), microwave radiometer, radar scatterometer and radar altimeter can be used to retrieve information of ocean surface winds. SAR sensors are the ones that achieve finest spatial resolution independently of weather and environmental conditions. The availability of C-band SAR data from satellite, and the direct interactions of C-band wavelengths with ocean surfaces has made C-band the favorite frequency for SAR wind estimation.

In this study, results from two regression models [1,2] for retrieval of ocean surface wind speed from C-band dual-polarization SAR images are compared with wind speeds from the commonly used C-band geophysical model function CMOD5.N [3]. The first model uses co-polarization (VV) normalized radar cross-sections (NRCS) and antenna beam incidence angle as independent variables. The second one uses VV NRCS, cross-polarization (VH) NRCS, instrument noise floor, and incidence angle as predictors. To train and evaluate the models, a database containing wind measurements from 107 NDBC/NOAA buoys along the east and west US coast collocated and coinciding with Sentinel-1 IW mode images throughout the year of 2017 was constructed. In total, there are over 3000 samples in the database, among which 75% are used for training and 25% for evalution. Results of the latter model proved to give a higher accuracy than the former due to the use of additional VH variables. Furthermore, the two regression models (without wind direction models) demonstrate better accuracies in the retrieved wind speed than what is achieved with CMOD5.N, which uses wind direction information.

 

References:

[1] Komarov, S., Komarov, A. and Zabeline, V., 2012. Marine wind speed retrieval from RADARSAT-2 dual-polarization imagery. Canadian Journal of Remote Sensing, 37(5), pp.520-528.

[2] Komarov, A.S., Zabeline, V. and Barber, D.G., 2014. Ocean surface wind speed retrieval from C-band SAR images without wind direction input. IEEE Transactions on Geoscience and Remote Sensing, 52(2), pp.980-990.

[3] H. Hersbach, 2010. Comparison of C-band scatterometer CMOD5. N equivalent neutral winds with ECMWF. Journal of Atmospheric and Oceanic Technology, 27(4), pp.721-736.

The range Doppler method is a unique application of SAR, that can extract instantaneous velocity field at low extra cost in all kinds of existing SAR system. The main potential applications of this method are within near-surface wind and sea surface current retrieval, and identification and speed measurement of moving targets, including sea ice. As such, this study applies the range Doppler method on sea ice drift retrieval. The main challenges in applying the method to sea ice drift are (1) the relatively low speed of general sea ice drift, ranging from, e.g., 2 to 23 cm/s in the Fram Strait, and (2) uncertainties in the estimation of the geophysical contribution to the range Doppler shift. Regarding the first challenge, the measurement uncertainty must be lower than the actual speed of the object at hand for the method to be useful in drift retrieval from a single scene. In terms of Doppler frequency, an error of 1 Hz would translate to an error in the ground velocity of 6.7 cm/s at 25° look angle using the Envisat ASAR system. Thus, the accuracy of the geophysical Doppler shift should be better than 2-3 Hz for detecting moderate ice drift. The second challenge is related to the correction of non-geophysical contributions to the range Doppler centroid shift. The main sources of uncertainties are, here, the estimation of the geometric Doppler shift and the antenna electronic mispointing. Previous studies were only able to suppress the measurement uncertainty down to 5 Hz for single scenes. This is not good enough for application of the SAR range Doppler method in sea ice drift retrieval.

We therefore examine an extensive set of Envisat ASAR ScanSAR data from January 2010 to improve the Doppler calibration. Following this, we evaluate the feasibility of using Doppler measurements for sea ice drift retrieval. This is done by comparing the Doppler measurements with results from the conventional cross-correlation (CC) based sea ice drift estimation method. In the Doppler calibration, we propose a comprehensive global data analysis rather than focusing on individual scenes only. For estimation of the geometric Doppler shift, the antenna misalignment along the axes of the satellite body and incomplete attitude knowledge are examined. Corresponding calibration procedures are then suggested. This correction results in an RMSE is 2.8 Hz for non-moving land pixels. Following the geometric calibration, electronic antenna mispointing is estimated.  This is done by finding an empirical relationship between land covering range Doppler centroid shift retrievals corrected for geometric contributions, and the antenna look-angle. This relationship is then used to compensate the antenna mispointing in image, at any location. Following the signal corrections, a preliminary comparison of SAR Doppler-based instantaneous velocity and the CC-based mean velocity shows an offset of 0.09 m/s between the two. In general, the SAR-Doppler based velocity estimates are about 1.2 times higher than the ones resulting from the CC-method. This can be understood by considering the rotating nature of sea ice motion. 

The synthetic aperture radar (SAR) Doppler centroid shift has been demonstrated to contain geophysical information about sea surface wind, waves and current at an accuracy of 5 Hz and pixel spacing of 3.5-9x8 km2. This corresponds to a horizontal surface velocity of about 20 cm/s at 35 degrees incidence angle. The ESA Prodex ISAR project aimed to implement new and improved SAR Doppler shift processing routines to enable reprocessing of the wide swath acquisitions available from the Envisat ASAR archive (2002-2012) at higher resolution and better accuracy than previously obtained, allowing combined use with Sentinel-1 and Radarsat-2 retrievals to build time-series of the sea surface velocity in the Nordic Seas.

New geophysical Doppler shift retrievals from Envisat ASAR acquisitions in January-March 2010 are used together with normalized radar cross section (NRCS) measurements and model wind forecasts to provide high resolution near-surface wind in a Bayesian scheme, following Mouche et al. (2012). In contrast to previous analyses, the new method takes into account spatial variability of the wind field. The results are assessed in comparison to buoy measurements from NOAA NDBC standard meteorological buoys, scatterometry data, and the results obtained by Mouche et al., 2012.

The retrieval of Doppler shift information from satellites, and subsequent signal partitioning into wind, wave, and current information, has potential for application within both operational oceanography (e.g., ship routing, offshore operations) and climate research.

The new quality of observations of marine processes ensured by the Sentinel sensors was demonstrated by an analysis of three consecutive satellite images captured over the Gulf of Lion in the Mediterranean Sea on the 19th of June 2017 with time intervals of 5 and 7 hours.  Sentinel-1A,B SAR-C and Sentinel-2A MSI images were used to investigate in detail the evolution of the oil slick occurred as a result of ship discharge of waters containing petroleum products.  Joint processing and analysis of the satellite data was performed using the tools of the See the Sea portal developed at the Space Research Institute RAS. Wind parameters were provided by the Meteoblue meteorological service.

The evolution of the slick was not the same along the ship path. Different parts of it were most likely influenced by different hydrometeorological conditions along the distance of about 35 km. This was confirmed by current modeling and analysis of the wind field. South-southeast wind varied from 4,39 to 8,16 m/sec. However, in the immediate region of initial discharge, near the coast, no considerable displacements of the slick was observed, only spreading. It was suggested that here the effects of wind were offset by the coastal current with opposite direction shown by the modeling. Or, this portion of the discharge could be too massive to be influenced either by wind, which was probably not so strong at the coast, or current. Further out from the coast, the other parts of the slick, as expected, drifted northwest at an average speed of 15–19 cm/sec, under the south-southeast wind.

This and many other examples show, that today, with the instruments onboard the Sentinel satellite family, we can examine sea surface phenomena at scales down to a few meters, compare radar and optical data obtained almost simultaneously over the same region at high and comparable spatial resolutions; observe drift and evolution of sea surface pollution and water dynamics due to smaller time intervals between data acquisitions.

Applications of different radar and optical methods for detection of oil pollutions based on the effect of damping of short wind waves by surface films have been extensively studied last decades. The main problem here is poor knowledge of physical characteristics of oil and oil derivatives films which are responsible for wave damping and respectively for possibilities of their remote sensing. In this paper results of laboratory studies of damping of gravity-capillary waves on the water surface covered by crude oil and oil emulsion, diesel fuel and kerosene films are presented. A laboratory method based on measuring the damping coefficient and wavelength of parametrically generated standing waves has been applied for determination of the film characteristics. The investigations were carried out in a wide range values of film thicknesses (from some hundreds millimetres to a few millimetres) and in a wide range of surface wave frequencies (from 10 to 27 Hz). The selected frequency range corresponds to the operating wavelengths of microwave, Х - to Ka -band radars typically used for the ocean remote sensing. The studied range of film thickness covers typical thicknesses of routine spills in the ocean. To retrieve parameters of the oil films from the experimental data the surface wave damping was analyzed theoretically in the frame of a model of two-layer fluid with viscous-elastic properties of the upper and lower boundaries of the upper layer. Physical parameters of oil derivative films were estimated when tuning the film parameters to fit theory and experiment. It is obtained that characteristics of waves, measured in the presence of different oil films are different, in particular, because their volume viscosity is strongly different and because in some case (oil emulsion, kerosene, diesel fuel) it is necessary to take into account the film solubility. Comparison between wave damping due to crude oil, kerosene and diesel fuel films have shown some capabilities of distinguishing of oil films from remote sensing of short surface waves. The work was supported by the Russian Science Founfdation (Project 18-17-00224).

Sea surface wind is one of the most important variables of atmospheric-ocean interaction, and the exchange of heat, momentum, and materials is performed through the sea  surface wind. To understand the global climate system, which is large in spatial and temporal scale, it is essential to study complex relationship between the two. Winds in coastal area cause a variety of marine phenomena, especially Eddy, which frequently occurs in the spring on the Korean peninsula East sea, and has physical and biological variability. Using the Sentinel-1A/B SAR data, we calculated high-resolution sea surface winds in the East Sea and simultaneously collected NOAA AVHRR sea surface temperature and COMS/GOCI chlorophyll-a data around Eddy. We also confirmed the presence of eddy in the East Sea using satellite altimetry data. As a result, lower sst was appeared in the eddy than in the surrounding area, and the SAR image in the same area showed low sigma nought value, that is, weak sea surface wind. In addition, the magnitude of the wind speed varies according to the stability of marine-atmospheric boundary layer. The steady state shows low wind speed, while the unstable state shows strong wind speed. When the wind was blowing over Eddie, the bloom position of Eddie was changed according to the wind direction. The bloom of the eddy occurred on the left side in the wind direction, which is the influence of the rotation direction inside eddy and the wind stress curl on the eddie boundary. This study analyzed the spatial variation of chl-a around eddy when the winds were blowing on coastal area, suggesting the impact of air-sea impacts caused by wind fields.

Sea surface wind is one of the most important variables in the ocean to understand physical processes in the ocean associated with various marine phenomena such as surface waves, currents, ocean circulation, air-sea heat transfer, biological features, and so on. Satellite scatterometers have long provided us with a good quality of sea surface wind field. Due to the limitations of a satellite scatterometer near the coast, however, it has not been possible to obtain the surface wind field essential for understanding coastal features. Recently, more frequent observations of Synthetic Aperture Radar (SAR) has made it possible to obtain high-resolution wind field near the coastal area including numerous islands. In particular, such a high-resolution wind field can be used to understand spatial distinction of differential wind forcings and its impact on air-sea-land interactions near the complicated coastlines and many of islands around Korean peninsula. Various oceanographic phenomena have been affected by coastal topography and the wind forcings due to air-sea interaction in marine atmospheric boundary layer. In this study, we showed the dominant change of wind field due to the orographic effect of island and the effect of wind field change on oceanic ecosystem using Sentinel-1A/B Interferometric Wide swath (IW) mode data. Land masking was performed using the Shuttle Radar Topography Mission Digital Elevation Model data, the adaptive threshold method was applied to detect and remove the ship in the ocean as one of error sources, and to remove the speckle noises by applying a moving window and an ensemble average. The sea surface winds were retrieved by applying the CMOD5.N algorithm to the preprocessed Sentinel-1A/B IW mode data, and then wind stress curl and Ekman pumping were also calculated to understand wind forcings on sea water. In order to investigate a relationship between wind speed and sea surface temperature (SST), sea surface temperatures (SST) are estimated based on National Oceanic and Atmospheric Administration (NOAA) Advanced Very High Resolution Radiometer (AVHRR)  with 1.1 km resolution prior to and post Sentinel-1A/B observations. As a result, the wind field was revealed to be modified by SST change as well as the island effect. Time series of Geostationary Ocean Color Imager (GOCI) chlorophyll-a (chl-a) concentrations confirmed  important biological impacts on the sea surface by the change in EKP field. Utimately, this study addresses the role of air-sea-land interaction on physical process as well as biological process of low-level oceanic ecosystem.

Wind measurements derived from SAR acquisitions provide a unique high resolution and wide view of coastal regions at day and night time, independently from the cloud cover. Classical estimation methods rely on Bayesian inversion using ancillary wind outputs from model together with the SAR co-polarized channel, as implemented for the Level-2 Wide Swath OCN products delivered by Sentinel-1 ESA ground segment.

To improve this wind inversion, a pre-processing step is introduced. Its objective is twofold:

-          First, better qualify and flag the SAR image region where the sea surface backscattering processes are not related to the wind stress and potentially impact by other processes such as rain impacts, ships or low winds. This qualification is based on the analysis of the SAR image roughness heterogeneity at multiple resolutions.

-          Second, based on this quality mask, the wind direction is estimated using gradient methods in presence of wind streaks and an associated confidence index. Integrating this estimation in the overall wind inversion is particularly motivated by the absence of good quality Doppler information that could help further constrain the wind direction.

This additional wind direction information is derived from co- and cross-polarized channel, when available, and integrated in the overall wind inversion using a new Bayesian scheme. Using a large SAR/buoy co-located dataset, it is validated and compared to the performances of the ESA processor. Applying the pre-processing step and using the heterogeneity mask shows a clear performances improvement.

Currently, the ESA ground segment inversion is based on CMOD-IFR2 Geophysical Model Function (GMF). The performances of this GMF in the wind inversion are compared to other up-to-date GMFs.

Finally, the NRCS cross-pol channel is integrated in the overall inversion scheme together with a dedicated cross-pol GMF ina two step Bayesian scheme. This inversion method is specifically designed for extreme winds assessed using a SAR dataset of hurricane SAR observations acquired during past SHOC campaign. In the end, the capability to retrieve good quality wind maps remains partly limited by the quality of the Sigma0 noise correction, particularly affecting the cross-pol channel and the availability of good quality Doppler measurements.

This study has been partly funded by the ESA SEOM Ocean studies, and the Mission Performance Center (MPC) activities.

Several initiatives and projects contribute to support Group on Earth Observation's (GEO) global priorities including support to the UN 2030 Agenda for sustainable development, the Paris Agreement on climate change, and the Sendai Framework for Disaster Risk Reduction. Running until 2020, the NextGEOSS project evolves the European vision of a user driven GEOSS data exploitation for innovation and business, relying on the three main pillars:

engaging communities of practice,

delivering technological advancements,

advocating the use of GEOSS.

These 3 pillars support the creation and deployment of Earth observation based innovative research activities and commercial services.

This presentation will emphasize how the NextGEOSS project uses a pilot-driven approach to consolidate the system in a pragmatique way, integrating the complexity of the existing global ecosystem, leveraging previous investments, adding new cloud technologies and resources and engaging the diverse communities to address all types of Sustainable Development Goals (SDGs).

 

A set of 10 initial pilots have been defined by the project partners to address the main challenges and include as soon as possible contributions to SDGs associated with Food Sustainability, Bio Diversity, Space and Security, Cold Regions, Air Pollutions, Disaster Risk Reduction, Territorial Planning, Energy.

 

This presentation will emphasize the working process and NextGEOSS assets developing marine pilots using Sentinel-1 data.

 

The first pilot aims at showcasing examples of products for selected areas in Cold Regions, combining remote sensing data, in situ data and model products from relevant European infrastructures and international frameworks. The pilot focuses on three areas: (1) the Arctic/Svalbard region, (2) Antarctica, and (3) the Himalayan glaciers. Within each region, products will be developed for users and stakeholders in the GEO community, polar research and education. An initial list of products includes, among others, sea ice type and drift maps for the Fram Strait, based on Sentinel-1 SAR data; sea ice concentration for the Arctic, based on satellite altimeter and passive microwave data; near-surface atmospheric aerosol properties (near-real-time) at Zeppelin station, Svalbard and at Troll station, Queen Maud Land, Antarctica.

 

The second pilot implements a sargassum identification and tracking demonstration. The detection is based on Sentinel-1 and Landsat data. Coupled with a proprietary drift model, the sargassum can be then tracked to identify precisely the impacted shorelines. From a technical point of view, this pilot shows an example of integration between S1 data and optical data. This activity is expected to be of high interest for institutional user communities working on the preservation of beach areas, several industrial actors interested in the processing of algae as well as the scientific community who is considering the sargassum as a possible indicator for climate change.

All initiatives with an interest in and need of Earth Observations (data, processes, models, ...) are welcome to candidate to become a new pilot initiatives.

NextGEOSS is a H2020 Research and Development Project from the European Community under grant agreement 730329.

Studies showed that changes in ocean state parameters affect the roughness of the ocean surface and scatterometry techniques like SAR can measure these changes in the sea surface roughness and thus the sea surface wind speed effectively. Changes in sea surfaces features like Lee waves, wind rolls due to atmospheric rolls, wind shadows, swell waves are associated with change in sea surface roughness. Synthetic Aperture Radar (SAR) is sensitive to ocean surface roughness because it can observes the changes in Bragg backscatter, tilt and sea surface movement of the radar signal as the resonant of the ocean capillary waves changes. Sentinel-1 Wave Mode data is regularly acquired over open oceans in order to produce oceanographic products (level-2 OCN) giving information on sea surface wind speed and 2D ocean wave information.. Sentinel satellites data products are available free of charge to all users and can be analyzed with the SNAP toolbox. We tested the Sentinel-1 A and B products for use at a meteorological service. We acquired the wave mode images from the Sentinel-1 data hub over 2017. We analyzed its performance over NOAA bouys and compare to the results of the German wea ther Service (DWD) global ocean wave model. We improve the performance of the wave mode product by filtering wave mode images showing sea surface features such as sea ice, land, ships and dark surface patterns. We further tested a deep learning model, which connects the imagery space directly to the meteorological information.

Synthetic Aperture Radar (SAR) is a key sensor for detection of oil spills by virtue of its ability to perform in almost all-weather, in day or night conditions. As SAR progressed from single-polarization (single-pol) to fully (or multi-) polarimetric SAR (pol-SAR) imagery, oil spill analysis and detection methodologies also progressed from conventional single-pol to pol-SAR images. A new architecture is compact polarimetry (CP) SAR, which is coherent dual-pol SAR, whereby one polarization is transmitted, and two orthogonal polarizations are received, each with relative phase, with respect to the other [Raney, 2007]. The CP SAR can obtain as much information as possible with the wider swaths and overcome the limitation of quad-polarimetric (QP) SAR with narrow swath. The potential of CP SAR makes it urgent to study oil spill monitoring with this new SAR mode.

There are essentially two ways to study oil spill detection and monitoring with the CP data. One method is related to the reconstruct a pseudo QP signature from CP SAR [Shirvany et al., 2012]. And the other approach is through the Stokes scattering matrix or the sample coherence matrix [Li et al., 2016; Espeseth et al., 2017]. It is shown that the CP mode is comparable with the QP mode in its ability to distinguish the various slicks from open water [Espeseth et al., 2017].

All these study concentrated on detection and classification of oil spills, and paid less attention to oil-water mixing ratio. It is an important parameter for volume estimation of oil spills and effectively emergency response to oil spills. Collins et al. [2013] investigated the reconstruction of pseudo QP covariance matrix from simulated HP data and computed the oil-water mixing index suggested in [Minchew, 2012].  However, these algorithms do not contribute additional information to the original CP data [Raney, 2016]. Moreover, these algorithms need prior assumptions, which cannot be satisfied in the real remotely sensed scenes, thereby limiting potential applications [Raney, 2016].

Here, we will study oil-water mixing ratio of oil spills on the ocean surface with respect to the CP mode, as simulated from existing QP UAVSAR data. Our approach is focused on using the backscatter parameters to obtain polarimetric information from the CP coherence matrix directly. The relationship between the elements of CP coherent matrix and QP scattering matrix will be established based on tilted-Bragg scattering model. The elements ratio of CP coherent matrix is only related to the incidence angle and the dielectric constant, independent of the surface roughness or ocean waves. This characteristic is similar to that of the co-pol ratio [Minchew, 2012], and suggests that the parameter is a potential candidate for measurement of the oil-water mixing ratio with CP data. The nonlinear and non-constrained optimization and look up table methods are employed to calculate the oil-water mixing ratio from CP data. And the results will be compared with those from QP observation. 

SAR images display a characteristic intensity variation along the range dimension, caused by the side-looking geometry and the fact that the backscattered intensities depend on the incidence angle.

In the case of wide-swath images, where the incidence angle range is large, this intensity variation from near range to far range is significant enough to affect image segmentation performed on absolute intensity values. The effects are an over-segmentation which creates banding in the range direction, as well as the dillution of the real class distinction. In addition, it has been shown that the decay rates vary for different classes, thus reducing the efficiency of previously proposed global-correction methods.  

As a solution, we propose an unsupervised general-purpose segmentation (clustering) algorithm that incorporates the incidence-angle variation into the standard mixture modeling. We consider the simplified case, where, per class, the intensity variation in the log-domain is assumed to be Gaussian, and the incidence angle variation is considered to be log-linear.

We demonstrate the algorithm’s efficiency with UAVSAR images containing oil spills and ships on an open-water background, acquired in the North Sea during 2015 (the NORSE2015 experiment). By considering the intensities of the HH channel and an incidence angle range spanning from 30 to 60 degrees, we obtain a clear improvement over the non-correcting segmentation algorithm. The banding effect induced by intensity variations within the water class is removed, and the main image structures (water, oil slicks and ships) are grouped into distinct classes.

The success of our preliminary modeling is encouraging for further development. The simple, approximative incidence-angle relation can be improved, in order to account for the near-range behavior of wide-swath instruments such as the UAVSAR. The Gaussian distribution can be replaced with a more appropriate model that accounts for heavy tails or texture. At last, we are considering extensions for polarimetry, provided that the different incidence-angle dependent behaviors can be accurately modeled for each channel.

This work deals with the classification of sea-ice in Sentinel-1 images. One of the main issues of using SAR images for sea ice classification is backscattering variation due to the local incidence angle and different normalization methods have been proposed including linear, iterative and class based normalization. In this research we propose a multi-scale classification independently applied in different incidence angle blocks of 5 degrees. Support Vector Machine (SVM) is used as a classifier of Sentinel-1 SAR images and is tuned to be relaxed on noise and misclassified training data. This classifier is applied iteratively at different scales and the output of each step is used as prior knowledge for the next classification iteration. Operational egg shape ice charts are used to validate the performance of this method. This work is being performed in the frame of the Cryo-SEANICE project funded by ESA and it complements another communication where we show an example of synergistic use of altimetry data and SAR imagery through a dedicated tool.

There has been done several studies using sub-aperture coherence methods (i.e. IHP) for ship detection. The method is utilizing that the sea surface have shorter de-correlation time than solid targets. In this study such a method has been tested for detecting the edge between sea ice and sea surface in SAR images. The study has looked in detail hoe this method works also in the marginal ice zone.

Combining restituted Sentinel-1 attitude processed on ground from telemetry measurements and Doppler centroid estimates, a data-driven approach is proposed to reconstruct the geometric Doppler along Sentinel-1 orbit. Together with a land scene based measurement of the electronic miss-pointing induced Doppler shift as function of elevation angle, the geophysical Doppler (LEVEL2 RVL product) can be calibrated to an accuracy compatible to the requirements to observe ocean surface currents. A new innovative method is used to estimate the sea state contribution to the geophysical Doppler based on an Intregral Spectral Value from the image cross spectra imaginary part. Preliminary results will be shown of Sentinel1 derived surface current in the line of sight direction for both WV and IW mode. Resulting performance is estimated by comparison with in-situ surface current measurements.

.