In this paper we work with truly multi-temporal change detection in multi-looked, multi-frequency polarimetric synthetic aperture radar (polSAR) data in the covariance matrix formulation. We apply recent general results on better approximations than the usual chi-squared distribution for the probability distributions associated with maximum likelihood ratio test statistics for equality of several block-diagonal covariance matrices with complex Wishart distributed blocks. We demonstrate the superiority of the new approximations by means of generated data as well as airborne EMISAR data from four time points covering an agricultural region in Denmark. Results from the generated data show the importance of applying the new approximations in the no change situation. This use is more important for low equivalent number of looks (ENL) and for long time series (i.e., high number of degrees of freedom). Results from the generated data example are confirmed by results from the case with EMISAR data.

We report on a complex Wishart distribution-based test statistic Q for block-diagonality in Hermitian matrices such as the ones analyzed in polarimetric synthetic aperture radar (polSAR) image data in the covariance matrix formulation. We also give an improved probability measure P associated with the test statistic. This is used in a case with simulated data to demonstrate the superiority of the new expression for P and to illustrate the dependence of results on the choice of covariance matrix, its dimensionality, the equivalent number of looks, and two parameters in the improved P measure. We also give two cases with acquired data. One case is with airborne F-SAR polarimetric data, where we test for reflection symmetry, another case is with (spaceborne) dual-pol Sentinel-1 data, where we test if the data are diagonal-only. The absence of block-diagonal structure occurs mostly for man-made objects. In the example with Sentinel-1 data, some objects (e.g., buildings, cars, aircraft, and ships) are detected, others (e.g., some bridges) are not.

This article describes the latest developments in our work on complex Wishart distribution-based detection of change in time series of multilook polarimetric synthetic aperture radar data in the covariance matrix representation. These developments include better approximations of the probability measures associated with the omnibus test statistics Q and R-} for block diagonal data in general, including the important cases with diagonal only Sentinel-1 data as obtained from Google Earth Engine and reflection symmetry data for full polarimetry. Additionally, the article introduces an omnibus version of the Loewner (or L & ouml;wner) order with visualization of change over time, where the omnibus change path shows significant difference. We also find the time point with the greatest change along the omnibus change path. The processing is illustrated with generated data and a series of 15 Sentinel-1A scenes covering Frankfurt Airport, Germany. Results show that the new and better approximations of the probability measures for the test statistics are important for the assignment of labels "change" or "no change" to a pixel or a patch, especially in "no change" regions. Furthermore, compared to the use of the full covariance matrix, the probability measures associated with the diagonal only test statistics incorrectly detect more change in these "no change" regions for the Sentinel-1 diagonal only data. Hence, the use of the full 2 x 2 covariance matrix if avalable is important. Finally, the omnibus Loewner order gives far fewer false detections than its pairwise counterpart.

In polarimetric synthetic aperture radar (SAR) images, speckle is removed by multilooking and the local covariance matrix is the main parameter of interest. In the covariance matrix from a backscatter with reflection symmetry, the terms ${\langle \boldsymbol{S}_{hh}\boldsymbol{S}_{hv}^*\rangle }$, ${\langle \boldsymbol{S}_{vv}\boldsymbol{S}_{hv}^*\rangle }$, and their complex conjugates are 0. The backscatter from natural covers, such as fields and forested areas, is typically reflection-symmetric, as these four elements have near-zero values. The backscatter from urban areas and man-made structures is substantially different, and the backscatter from buildings not aligned with the radar line of sight usually does not have reflection symmetry. A novel block-diagonality test statistic for reflection symmetry with a constant false alarm rate property is proposed. It is compared to an approximate test built on a change detection test statistic for Wishart-distributed covariance matrices. Their use on quad-polarimetric data in different situations shows their high potential for man-made structure detection. Applied after an orientation correction of the covariance matrices, these test statistics highlight with high-contrast buildings and urban areas. We also apply this test for ship detection at sea, and show that while the results are unconvincing at X-band, it can also be applied at longer wavelengths such as L-band.

The increased amount of information measured by fully polarimetric synthetic aperture radar (SAR) give additional knowledge about ground scatterers. Making the best use of the polarimetric information is crucial for target detection, among other applications. Several representations of the data, such as polarimetric decompositions, have been proposed to summarize the information into polarimetric features. The relation between these features with physical properties of the scatterers has been studied in depth. The different approaches to target detection proposed make use of different polarimetric features and different properties of the targets. The goal of this article is twofold: 1) to give a brief review of polarimetric features usually used for target detection, and 2) to combine them optimally for vehicle detection in open fields, in large natural scenes. The study's backbone is a large airborne dataset in X-, S-, and L-bands, in which several flights following different flight tracks were performed around a controlled area with a dozen vehicles. At first, a univariate study is performed to evaluate the contrast provided by individual polarimetric features between vehicles and different types of natural covers. Then, optimal subsets of polarimetric features for distinguishing vehicles in open fields from natural cover are determined using random forest classifiers. The multivariate approach yielded better detection results for all wavelengths but brought more significant improvement as the wavelength increases. At the X-band, the total received power is one of the best predictive parameters for vehicle detection while the scattering mechanism characterization becomes more important at S- and L-bands.

Dans ce chapitre, nous considèrerons le problème de la détection de changements dans une série chronologique d'images SAR (radar à synthèse d'ouverture) polarimétriques à l'aide de la représentation en covariance de données SAR polarimétriques multivisées. Les pixels seront alors représentés par des matrices hermitiennes complexes suivant une distribution de Wishart. La chaîne de détection de changements consiste en un test omnibus destiné à tester l'égalité sur la totalité du laps de temps, puis en une factorisation permettant d'évaluer individuellement les dates où un changement a eu lieu. La méthode est aisément étendue à la détection de changements de zones "homogènes", et nous introduisons enfin un concept pour le changement directionnel utilisant la relation d'ordre de Lowner.

We use the combined Wishart-Loewner method to successfully detect change and direction of change in truly multitemporal, multilooked quad/full polarization synthetic aperture radar image data in the covariance matrix representation. Based on in situ data interpretations of the obtained results are given for three selected fields. Histograms of the Wishart test statistics in a wooded no-change area shows good agreement with the theoretical distributions.

With a growing number of different satellite sensors, data fusion offers great potential in many applications. In this work, a convolutional neural network (CNN) architecture is presented for fusing Sentinel-1 synthetic aperture radar (SAR) imagery and the Advanced Microwave Scanning Radiometer 2 (AMSR2) data. The CNN is applied to the prediction of Arctic sea ice for marine navigation and as input to sea ice forecast models. This generic model is specifically well suited for fusing data sources where the ground resolutions of the sensors differ with orders of magnitude, here 35 km × 62 km (for AMSR2, 6.9 GHz) compared with the 93 m × 87 m (for sentinel-1 IW mode). In this work, two optimization approaches are compared using the categorical cross-entropy error function in the specific application of CNN training on sea ice charts. In the first approach, concentrations are thresholded to be encoded in a standard binary fashion, and in the second approach, concentrations are used as the target probability directly. The second method leads to a significant improvement in R 2 measured on the prediction of ice concentrations evaluated over the test set. The performance improves both in terms of robustness to noise and alignment with mean concentrations from ice analysts in the validation data, and an R2 value of 0.89 is achieved over the independent test set. It can be concluded that CNNs are suitable for multisensor fusion even with sensors that differ in resolutions by large factors, such as in the case of Sentinel-1 SAR and AMSR2.

Time series analysis of Sentinel-1 SAR imagery made available by the Google Earth Engine (GEE) is described. Advantage is taken of a recent modification of a sequential complex Wishart-based algorithm which is applicable to the dual polarization intensity data archived on the GEE. Both the algorithm and a software interface to the GEE Python API for convenient data exploration and analysis are presented; the latter can be run from a platform independent Docker container and the source code is available on GitHub. Application examples are given involving the monitoring of anthropogenic activity (shipping, uranium mining, deforestation) and disaster assessment (flash floods). These highlight the advantages of the good temporal resolution resulting from cloud cover independence, short revisit times and near real time data availability.

We describe the use of P-band data from DLR's F-SAR system to change detection between three time points. Change detection on data in the covariance matrix representation is performed by means of test statistics in the complex Wishart distribution on both multi- and single-look data. We successfully detect objects including humans buried under snow in Disko Island, Greenland.

Temporal filtering for speckle reduction of polarimetric SARimages is described. The method is based on a sequential complex Wishart-based change detection algorithm which is applied to polarized SAR imagery, including the dual-polarization intensity data archived on the Google Earth Engine (GEE). Software for convenient application and analysis is presented. Results compare favorably with, and improve upon, standard spatial and temporal filters for speckle reduction.

The present project presents the technical implementation, testing and validation of a soil moisture retrieval algorithm in Python using C-band Sentinel-1 data at high incidence angle (∼42°). The retrieval algorithm is based on the alpha approximation, first developed by [Balenzano et al. 2011]. The alpha approximation utilizes the dense temporal coverage of the Sentinel-1 mission, assuming that changes in backscatter between subsequent acquisitions are only due to variations in soil moisture, such that vegetation and roughness can be neglected. The area used for testing the algorithm was chosen to be the region surrounding the Foulum test center for agricultural studies in Denmark, due to the availability of time series from 2018 of in situ soil moisture measurements to be used for validation. Masking of too densely vegetated areas have been performed using the cross-polarized component of the SAR backscatter, which have been validated using NDVI maps. Auxiliary data, including land cover maps and parcel borders enable the computation of backscatter field means, significantly reducing the impact of speckle noise and thus decreasing uncertainty of the estimated soil moisture. Consequently, the results have field scale resolution (i.e. ∼0.1 km). The permittivity to soil moisture inversion is performed using a polynomial model by [Hallikainen et al. 1985], where a soil texture map provide the information necessary to obtain precise results. Further work will aim toward applying a change detection algorithm in order to detect sudden temporal changes in vegetation and surface roughness, as the alpha approximation is inherently sensitive to such sudden changes. The study has received partial funding from Innovation Fund Denmark, contract number: 7049-00004B (MOIST).

This paper investigates the sensitivity of Sentinel-1 (S-1) interferometric coherence to crop structure and near surface soil moisture (SSM) content. The study analyzes a data set collected in 2017 over the Apulian Tavoliere agricultural site (Southern Italy). The data set includes: i) in situ data over more than 600 agricultural fields monitored during the 2017 winter and spring growing seasons; ii) time-series of S-1 IW VV & VH backscatter & interferometric coherence; iii) time series of S-1 SSM maps. The temporal behavior of S-1 coherence and VH backscatter has been assessed over the monitored agricultural fields. Initial results indicate a stronger sensitivity of S-1 coherence than VH backscatter to crop geometric structure. In addition, an analysis at site scale, conducted before and after an important rain event, indicates a change of SSM from 0.18 to 0.30 m 3 /m 3 along with a change of S-1 coherence from 0.61 to 0.53.

Today, ice charts in Greenland waters are produced manually by the Danish Meteorological Institute (DMI) for selected regions depending on season and shipping routes. The project “Automated Downstream Sea Ice Products for Greenland Waters” or shorter “Automated Sea Ice Products” (ASIP) attempts to automate this process by means of fusion of data from instruments with different resolutions and modalities. As a part of this process data from the Advanced Microwave Scanning Radiometer (AMSR2) will be interpolated to the geometry of the SAR data acquired by Sentinel-1. In a preparatory leave-one-out cross-validation (LOOCV) study, different interpolation methods including ordinary kriging (OK) are compared. Using bias and root-mean-squared error (RMSE) as measures of precision, OK using 20-30 nearest neighbours outperforms other often used methods such as inverse distance (ID) weighting. This comes at a cost: more work needs to be done by both the operator and the computer. INTRODUCTION The project “Automated Downstream Sea Ice Products for Greenland Waters” or shorter “Automated Sea Ice Products” (ASIP) is a cooperation between the Danish Meteorological Institute (DMI), two departments at the Technical University of Denmark (DTU), the National Space Institute (DTU Space) and the Department of Applied Mathematics and Computer Science (DTU Compute), and Harnvig Arctic & Maritime. The project is funded by Innovation Fund Denmark. The objective of ASIP is to develop an automatic sea ice product service for Greenland waters which can meet the increasing demands for sea ice information coming from the growing group of users operating in Greenland waters. In the span from traditional, manually produced ice charts and daily downstream sea ice products at coarse resolution, there is a lack of high resolution products delivered in near-real time. ASIP intends to meet this demand by taking advantage of the vast amount of new data from the Copernicus Sentinel satellites and by using a new and innovative data fusion approach and state-of-the-art mathematical/statistical data processing methods: utilization of data from satellite sensors with different modalities/capabilities will facilitate the making of ice products that are reproducible, independent of operator, daylight, weather and season and will result in a significant increase in product temporal frequency and geographical coverage compared to existing ice products. The statistical algorithms work directly in the Sentinel-1 scene coordinate system. In order to make use of the information in the AMSR2 data along with the radar data an alignment of the AMSR2 data to the radar coordinate system is therefore necessary. In this process of interpolating AMSR2 data to the Sentinel SAR data, in a preparatory study six methods are compared by means of leave-one-out cross-validation (LOOCV) 1. nearest neighbour (NN, one neighbour only), 2. triangulated irregular network (TIN, three neighbours only), 3. local mean (LM), 4. inverse distance (ID), 5. inverse squared distance (ID2), and 6. ordinary kriging (OK). Bias and and root-mean-squared error (RMSE) are used as measures of precision. OK with 20-30 nearest neighbours obtains a LOOCV bias of around 0.001 K and RMSE of around 1.1 K. The second best of the six methods is ID2 which with 5-10 nearest neighbours gives a LOOCV bias of around 0.01 K and RMSE of around 3 K. When we use kriging we must estimate semivariograms and model them, this takes operator as well as computer time. The project homepage https://asip.dk will be launched soon.

When the covariance matrix formulation is used for multilook polarimetric synthetic aperture radar (SAR) data, the complex Wishart distribution can be used for change detection between acquisitions at two or more time points. Here, we are concerned with the analysis of change between two time points and the “direction” of change: Does the radar response increase, decrease, or does it change its structure/nature between the two time points? This is done by postprocessing/coprocessing the detected change with the Loewner order which calculates the definiteness of the difference of the covariance matrices at the two time points. We briefly describe the theory. Two case studies illustrate the technique on Sentinel-1 data covering the international Frankfurt Airport, Germany, and on Radarsat-2 data covering Bonn, Germany, and surroundings. We successfully demonstrate our “direction” of change approach to detected change areas.

For the wind energy application, updated information on aerodynamic surface roughness is important for an ac-curate prediction of the land surface effect on the atmosphere. Tall wind turbines are commonly sited in forested areas, and since the forest both increase turbulence levels and decrease the wind resource, the parametrization of forest roughness in wind models is of high relevance. Here, we investigate whether the Sentinel-1 SAR images can be used to identify high-roughness forested areas. The backscatter properties of the SAR images are compared to digital surface models and vegetation density maps derived from near-concurrent aerial lidar scans (ALS). These ALS products have previously shown good results in wind models for the wind energy application, but the scans are costly to perform and therefore typically only represent a snapshot in time, whereas the Sentinel mission SAR images provide frequent updated information. We investigate how the SAR images vary with season over both de-ciduous and needle-leaf forests and in addition test whether nearby meteorological observations can explain image to image differences in the backscatter level. In order to understand the backscatter level, new products from the ALS point cloud are derived and compared with the SAR images. This part of the work is focused on whether we also can quantify the roughness based on the SAR backscatter. Since SAR images are affected by speckle noise, they are averaged over monthly and bimonthly intervals after careful inspection of each image. The work is focused on the Østerild test site for large wind turbines in Northern Denmark, where extensive wind experiments have been performed.