EarthConsole® Stories: GPU-powered SAR back-projection fully focused processor for enhanced S6 altimetry data processing
EarthConsole® Stories are experiences about how we helped universities, research centres or service developers to leverage Earth Observation data to extract valuable insights for their research, educational or pre-commercial projects.
The Project
In the frame of the Sentinel-6 Poseidon-4 Ground Processor Prototype project, a Synthetic Aperture Radar (SAR) back-projection fully focused processor for Sentinel-6 (S6) was developed which processes all the chains starting from the altimeter source packets up to the Level 1B.
This processor had a major limitation: its run-time performance was too slow to be used for operational purposes. To address this, the research team decided to employ a dual strategy: optimizing the algorithm and leveraging the computational power of Graphics Processing Units (GPUs).
They successfully developed this improved processor in November 2023, which was a big step forward for the project.
Initial tests showed very promising results, showing that the processor running on GPUs has a run-time performance improvement of over 50 times with respect to the one running on CPUs. In the final phase of the project, the need arose to thoroughly test both the scientific accuracy and the run-time performance of the processor. This required a dedicated machine with specific configurations, a need which was met by EarthConsole®.
The Need
The main need was to find a robust and efficient processing environment to fully utilize the processor’s potential. The team required a high-performance computing solution to finalize the software and run tests using Sentinel-6 data. Following this phase, the intention was to perform a full 10-day cycle of S6 data with multiple configurations optimized for various surface types to verify how well the processor worked across various conditions. Finally, the generated data had to be analysed and assessed in comparison with existing operational data.
Why EarthConsole®
The decision to resort to EarthConsole® was driven by its ability to meet our computational needs, offering GPU-equipped virtual machines essential to test our processor’s optimized performance. These virtual machines with GPUs are not yet available within the European Space Agency’s (ESA) existing infrastructure. This enabled us to achieve improvements in run-time performance and scientific accuracy for our SAR back-projection processor
Marco Fornari, Ground Segment Engineer at RHEA for ESA – European Space Agency
The Impact
The impact of this project extends far beyond the immediate improvements in processing efficiency for Sentinel-6.
The knowledge acquired with this project will have significant implications for future missions as well, such as CRISTAL and Sentinel-3 Next Generation Topography (S3NG-T), where similar processing capabilities will be essential.
Marco Fornari, Ground Segment Engineer at RHEA for ESA – European Space Agency
This project has been supported via the ESA Network of Resources initiative.