Search

TU Wien’s flood mapping algorithm, used for global operations, utilizes harmonic functions to model the seasonal behavior of backscatter to improve flood classification. In earth observation (EO), temporal harmonic models have been used in various scenarios for vegetation and water mapping in the optical and, recently, synthetic aperture radar (SAR) domains. These models condense EO time series stacks to a few Fourier coefficient images that capture seasonal variability, allowing for variable estimation for each day of the year. TU Wien’s harmonic parameters consist of these coefficients plus the regression standard deviation and number of observations. However, generating harmonic models at large scales and high resolution presents significant logistical and technical challenges. Particularly for SAR, which requires special handling due to acquisition geometry considerations, implementation on a datacube infrastructure is necessary for agile filtering in metadata, temporal and spatial dimensions. In this work, we highlight our harmonic parameter dataset and our software stack of loosely coupled Python packages, which were deployed in a high-performance computing environment to generate these parameters globally.

Synthetic Aperture Radar (SAR) images acquired by Earth observation satellites often constitute the only source of information for monitoring the progression of flood events over larger regions. Particularly attractive are the SAR data acquired by the Copernicus Sentinel-1 satellites because they are free and open, and combine a short revisit time with a good spatial and radiometric resolution. In this contribution, we discuss how a Sentinel-1 data processing system should be designed to optimally benefit from the dense Sentinel-1 time series and advanced algorithms such as change detection or machine learning methods. This was one of the questions addressed by an expert group tasked by the Joint Research Centre of the European Commission to investigate the feasibility of an automated, global, satellite-based flood monitoring product for the Copernicus Emergency Management Service. Drawing from the expert group report, we distinguish three broad categories of data processing architectures, namely single-image, dual-image, and data cube processing architectures. While the latter architecture is the most demanding in terms of large storage and compute capacities, it is also the most promising to derive high-quality Sentinel-1 flood maps comprised not just of the flood mask but also of data fields describing the retrieval uncertainty and masks showing where Sentinel-1 cannot detect floods due to physical reasons. Therefore, we recommend to use data cube processing architectures and showcase the use of the Austrian Data Cube for monitoring a small-scale flood event that occurred in Austria in November 2019.

Radiometric Terrain Corrected (RTC) gamma nought backscatter, which was introduced around a decade ago, has evolved into the standard for analysis-ready Synthetic Aperture Radar (SAR) data. While working with RTC backscatter data is particularly advantageous over undulated terrain, it requires substantial computing resources given that the terrain flattening is more computationally demanding than simple orthorectification. The extra computation may become problematic when working with large SAR datasets such as the one provided by the Sentinel-1 mission. In this study, we examine existing Sentinel-1 RTC pre-processing workflows and assess ways to reduce processing and storage overheads by considering the satellite's high orbital stability. By propagating Sentinel-1's orbital deviations through the complete pre-processing chain, we show that the local contributing area and the shadow mask can be assumed to be static for each relative orbit. Providing them as a combined external static layer to the pre-processing workflow, and streamlining the transformations between ground and orbit geometry, reduces the overall processing times by half. We conducted our experiments with our in-house developed toolbox named wizsard , which allowed us to analyse various aspects of RTC, specifically run time performance, oversampling, and radiometric quality. Compared to the Sentinel Application Platform (SNAP) this implementation allowed speeding up processing by factors of 10-50. The findings of this study are not just relevant for Sentinel-1 but for all SAR missions with high spatio-temporal coverage and orbital stability.

We present a new perspective on Earth's land surface, providing a normalised microwave backscatter map from spaceborne Synthetic Aperture Radar (SAR) observations. The Sentinel-1 Global Backscatter Model (S1GBM) describes Earth for the period 2016-17 by the mean C-band radar cross section in VV- and VH-polarisation at a 10 m sampling. We processed 0.5 million Sentinel-1 scenes totalling 1.1 PB and performed semi-automatic quality curation and backscatter harmonisation related to orbit geometry effects. The overall mosaic quality excels (the few) existing datasets, with minimised imprinting from orbit discontinuities and successful angle normalisation in large parts of the world. Regions covered by only one or two Sentinel-1 orbits remain challenging, owing to insufficient angular variation and not yet perfect sub-swath thermal noise correction. Supporting the design and verification of upcoming radar sensors, the obtained S1GBM data potentially also serve land cover classification and determination of vegetation and soil states. Here, we demonstrate, as an example of its potential use, the mapping of permanent water bodies and evaluate against the Global Surface Water benchmark., (© 2021. The Author(s).)

Urban areas and their vertical characteristics have a manifold and far-reaching impact on our environment. However, openly accessible information at high spatial resolution is still missing at large for complete countries or regions. In this study, we combined Sentinel-1A/B and Sentinel-2A/B time series to map building heights for entire Germany on a 10 m grid resolving built-up structures in rural and urban contexts. We utilized information from the spectral/polarization, temporal and spatial dimensions by combining band-wise temporal aggregation statistics with morphological metrics. We trained machine learning regression models with highly accurate building height information from several 3D building models. The novelty of this method lies in the very fine resolution yet large spatial extent to which it can be applied, as well as in the use of building shadows in optical imagery. Results indicate that both radar-only and optical-only models can be used to predict building height, but the synergistic combination of both data sources leads to superior results. When testing the model against independent datasets, very consistent performance was achieved (frequency-weighted RMSE of 2.9 m to 3.5 m), which suggests that the prediction of the most frequently occurring buildings was robust. The average building height varies considerably across Germany with lower buildings in Eastern and South-Eastern Germany and taller ones along the highly urbanized areas in Western Germany. We emphasize the straightforward applicability of this approach on the national scale. It mostly relies on freely available satellite imagery and open source software, which potentially permit frequent update cycles and cost-effective mapping that may be relevant for a plethora of different applications, e.g. physical analysis of structural features or mapping society's resource usage., Competing Interests: None., (© 2020 The Author(s).)