Search

Poster presentation at the ESA Polinsar Biomass 2023 conference, 19.06.23 - 23.06.23 in Espaces Vanel, Toulouse. https://polinsar-biomass2023.esa.int/.

Source at https://www.vde-verlag.de/buecher/proceedings/. The Polarimetric Synthetic Aperture Radar (PolSAR) backscatter from a target is dependent on the incidence angle. Consequently, the associated roll invariant parameters are affected by changes in incidence angle. In this work, we identify a few of these parameters that remain robust in identifying sea ice features even under large incidence angle variations. We conclude that the helicity angle and the degree of purity are preferable over the scattering type angle in this respect. We utilize two overlapping RADARSAT-2 C-Band full polarimetric images, with a time difference of less than 2 hours, but with significant incidence angle difference.

Source at https://www.vde-verlag.de/proceedings-en/455457132.html. We apply a polarimetric two-scale model to multifrequency synthetic aperture radar imagery of verified oil slicks measured by DLRs F-SAR instrument, which can acquire high spatial resolution and high signal-to-noise data. The purpose, is to determine the permittivity of the scattering surface via an inversion procedure. The ocean surface is modelled as an ensemble of randomly orientated, tilted facets for which the small perturbation model is assumed valid under the condition of no tilt. Only co-polarization channels are utilized due to system noise, hampering the signal within the cross-polarization channels. Results yield realistic results for the permittivity of oil slick.

ExtremeEarth is a European H2020 project; it aims at developing analytics techniques and technologies that combine Copernicus satellite data with information and knowledge extraction, and exploiting them on ESA’s Food Security and Polar Thematic Exploitation Platforms. In this publication, we focus on the Polar case which requires the selection of validation areas, the generation of a training dataset, the development and testing of deep learning algorithms, and the demonstration of regional results. During the development of deep learning algorithms, a key activity is to establish a large amount of referenced Earth Observation data. They need to be sufficiently diverse to cover the major target areas of satellite images under varying imaging conditions and across all seasons. For doing this, we propose to select overlapping target areas from Synthetic Aperture Radar and multispectral images acquired with rapid succession. Such a combination approach already demonstrated its applicability for monitoring seasonal snow cover [1]. By applying an already established active learning approach based on a Support Vector Machine with relevance feedback [2], we can limit ourselves to a limited number of typical satellite images to extract their information content, and to generate semantic annotations for them. This approach is also a simple way to generate benchmarking datasets that can be used for testing and validating different algorithms, and for creating additional bigger datasets for large-scale demonstrations. The proposed methodology uses new paradigms from Recurrent Neural Networks and Generative Adversarial Networks, supported by Bayesian and Information Bottleneck concepts. References 1. Monitoring sesonal snow cover. Available: https://ieeexplore.ieee.org/document/8518203. 2. EOLib project. Available: http://wiki.services.eoportal.org/tiki-index.php?page=EOLIB+Project.

Copernicus is the European programme for monitoring the Earth. It consists of a set of systems that collect data from satellites and in-situ sensors, process this data and provide users with reliable and up-to-date information on a range of environmental and security issues. The data and information processed and disseminated puts Copernicus at the forefront of the big data paradigm, giving rise to all relevant challenges, the so-called 5 Vs: volume, velocity, variety, veracity and value. In this short paper, we discuss the challenges of extracting information and knowledge from huge archives of Copernicus data. We propose to achieve this by scale-out distributed deep learning techniques that run on very big clusters offering virtual machines and GPUs. We also discuss the challenges of achieving scalability in the management of the extreme volumes of information and knowledge extracted from Copernicus data. The envisioned scientific and technical work will be carried out in the context of the H2020 project ExtremeEarth which starts in January 2019.

Space-borne synthetic aperture radar (SAR) systems are essential for operational monitoring of sea ice in polar regions. Radar polarimetry is a powerful niche within remote sensing, allowing investigations of various sea ice types’ scattering properties. Full-polarimetric and compact polarimetry SAR measurements may hence allow improved sea ice type characterization and discrimination capabilities aiding operational sea ice services. In this talk,we disseminate results from remote sensing measurements of sea ice collected under the Norwegian Young Sea Ice 2015 (N-ICE2015) field campaign that took place January to June 2015 in the Fram Strait. In particular, we focus on multifrequency SAR studies, involving acquisitions from ALOS-2 (L-band), Radarsat-2 (C-band), and TerraSAR-X (X-band), and give some directions for future research. First, an analysis of full-polarimetric L-, C-, and dual-polarization (HH/VV) X-band SAR data recorded over lead ice revealed that by combining the scattering entropy and co-polarization ratio we can successfully separate newly formed sea ice from open water and thicker sea ice within all three frequencies throughout the winter and spring season. The polarization difference exhibits less incidence angle dependency and shows to provide additional discrimination support. X- and L-band SAR acts complementary to the more regular acquisitions in C-band in terms of characterizing the newly formed sea ice types and surface structure. Second, asemi-coherent backscatter model is used to interpret the space-borne SAR data acquisitions of the Fram Strait sea ice. Specifically, full-polarimetric L-,C-, dual-polarization (HH/VV) and full-polarimetric X-band SAR is compared to model output. Constrained to simultaneous in-situ observations from the campaign, the model is able to reproduce the backscatter from lead ice and ice floes well for the individual frequency bands. For open water leads, unexpectedly high backscatter values are observed in L-band compared to C-band. Possible explanations relating to the sea ice formation process are discussed. Third, an automatic sea ice classification algorithm developed for near real-time services on full-polarimetric SAR measurements has been tested for X-, C-, and L-band data. Spatial and temporal coincident sea ice freeboard measurements of an airborne laser scanner as well as sea ice thickness data were used to validate the classification results. It was found that the number of multipolarization SAR parameters could be reduced from18 to 9, for all three frequencies, whilst still maintaining the 96.9% sea ice classification accuracy. The set of parameters that were found most useful in L-band was slightly different compared to those for the other two frequencies. Next, exploring a supervised classification scheme, we which to includemore full-polarimetric ALOS-2 L-band SAR scenes from the N-ICE campaign. Our goal is to investigate which multipolarization SAR parameters that are most useful for operational services, aiming at high accuracy sea ice classification under various environmental conditions and imaging geometries.

Source at https://www.vde-verlag.de/proceedings-en/454636136.html. In this paper, we investigated the capabilities of the Gaussian Process Regression (GPR) algorithm in predicting of two quad-polarimetric parameters (relevant for sea ice analysis) from 6-dimensional dual-polarimetric input vectors. The GRP is trained on few hundred samples selected randomly from an image subset, and tested on the entire image. The performance is assessed by visual comparisons, and by quantifying two regression performance statistical measures. The results of the regression showed big variations from scene to scene, and between the estimated output parameters, but the overall assessment is that the method gave surprisingly good correspondence to the real quad-polarimetric parameters.

Compared to widely used C-band SAR imagery, L-band SAR is a valuable complement in monitoring the Arctic and Antarctic sea ice covers, since it is very sensitive to ice deformation structures. The continuous observation of sea ice and the retrieval of sea ice properties from satellite data has received increased attention from the science community after the extreme minimum of summer arctic sea ice extent observed in 2007, which was followed by an even lower minimum in 2012. Also in operational sea ice monitoring there is an interest in investigating the potential of L-band SAR for sea ice mapping, although spatial coverage and timeliness of L-band data provided by recent satellite missions are still a critical issue for operational services. The concept of the TANGOSat mission will it make possible to address the potential of satellite tandems in gaining additional insights into the interaction between longer microwaves and sea ice. In particular the tomographic and bistatic phase may offer possibilities for specific investigations such as depth of scatterers and structures in thick multi-year ice (as found north of Greenland), quantification of ice surface roughness due to the occurrence of deformation structures, potential improvements of ice type classification, or snapshots of ice drift components by using along-track interferometry. In the presentation, different possibilities of investigations will be discussed.

This work extends upon our simple feature-based multichannel SAR segmentation method to incorporate highly desirable statistical properties into a computationally simple approach. The desirable properties include Markov random field contextual smoothing and goodness-of-fit testing to automatically obtain the significant number of classes. To achieve this we need to find an explicit class model to fit these non-Gaussian, non-symmetric or skewed feature space clusters. We take the skewed scale mixture of Gaussian scheme to model our classes and approximate it by a number of constrained Gaussians, thereby retaining much of the speed and simplicity of the original feature space method. The algorithm will be demonstrated on a real data and compared to an automatic Gaussian model.

Source at https://earth.esa.int/web/guest/document-library/browse-document-library/-/article/esa-living-planet-symposium-7438. Our aim is to use unsupervised, non-Gaussian clustering of Arctic glaciers for post-classification change detection. Firstly, we demonstrate the consistency of non-Gaussian clustering algorithms for Envisat ASAR images by characterizing the expected random error level for different SAR acquisition conditions (such as incidence angle). This allows us to determine whether an observed variation is statistically significant and therefore can be used for post-classification change detection of Arctic glaciers. Real significant change was not detected with mixed configurations during the time period of this study.

Publication in the conference proceedings of EUSIPCO, Florence, Italy, 2006

Publication in the conference proceedings of EUSIPCO, Florence, Italy, 2006