Conference Agenda

The exploitation of Sentinel‑1 Interferometric Synthetic Aperture Radar (InSAR) data has become increasingly important for Earth Observation (EO) applications, particularly for monitoring surface deformation and related natural hazards. However, the processing of Sentinel‑1 Single Look Complex (SLC) data, especially for multi-temporal techniques such as Persistent Scatterer Interferometry (PSI) and Small Baseline Subset (SBAS), remains computationally demanding. These techniques require handling large volumes of SAR data and must rely on robust, scalable computing infrastructures, which often exceed the capabilities of many users and organizations. Cloud‑based processing environments offer a transformative alternative by providing elastic scalability, robust data management, and direct access to EO data archives.

To address these challenges, we developed an open-source, fully cloud-native InSAR processing solution integrated into openEO within the Copernicus Data Space Ecosystem (CDSE). While openEO previously supported Sentinel‑1 GRD backscattering workflows, it lacked native capabilities for SLC‑based interferometric processing. We closed this gap by enabling the execution of complete InSAR workflows directly in the cloud, ensuring modularity, reproducibility, and user-driven customization.

New openEO processes were developed to cover all key steps of the Sentinel‑1 SLC interferometric workflow, including burst selection, co‑registration, interferogram generation, coherence estimation, and phase unwrapping. All processes are fully open-source and implemented using the Common Workflow Language (CWL), providing a standardized and portable way to define complex, container-based processing chains. These CWL-based processes can be seamlessly combined with existing openEO building blocks to form standardized, reusable process graphs. This modular architecture allows users to adapt workflows, integrate additional datasets, and perform scalable analyses over large spatial and temporal domains.

The processing system is based on burst-level operations, a design that significantly enhances computational efficiency in cloud environments. By exploiting the Sentinel‑1 TOPS acquisition structure, the workflow minimizes data transfers, optimizes resource allocation, and enables large‑scale parallelization.

To validate the implementation, we applied the workflow to two real-world use cases: (1) multi-temporal coherence analysis for debris‑covered glacier detection, and (2) interferogram time series generation for permafrost‑related deformation monitoring.

In detail, in the first use case, we produced long-term coherence time series for all the glaciers in South Tyrol, processing 25 bursts across 5 acquisition geometries. Persistent coherence patterns associated with constant‑moving debris enabled the automated delineation of debris-covered glacier areas. These results were validated against existing glacier inventories, derived from high‑resolution orthophotos and LiDAR acquisitions, demonstrating the reliability of cloud‑native coherence time series and their suitability for operational cryosphere applications.

In the second use case we generated consistent interferogram time series over permafrost areas in alpine terrain. Special attention was given to ensuring compatibility with MintPy, a widely used tool for multi‑temporal InSAR analysis. The interferograms produced with openEO were successfully ingested into MintPy confirming that the outputs meet the necessary technical standards for downstream deformation analysis. This demonstrates the capability of the system to support hybrid cloud‑desktop workflows and strengthens interoperability with established InSAR processing ecosystems.

Overall, the results confirm that integrating cloud-native InSAR processing into openEO and CDSE enables scalable, reproducible, and user-friendly workflows for advanced SAR applications. The developed solution expands openEO and CDSE into a fully interferometric processing environment, supporting both research and operational applications.

Operational InSAR services are often limited by two practical bottlenecks: reliable, automated access to large Sentinel-1 archives and reproducible, scalable processing that can be deployed beyond a single workstation. We present an end-to-end processing environment that couples “phidown” for programmatic discovery and downloading of Copernicus Sentinel data from the Copernicus Data Space Ecosystem with “sarpyx”, a Python SAR toolkit that orchestrates ESA SNAP Graph Processing Tool (GPT) workflows for InSAR. The system is designed around burst-level processing: users specify an area of interest, acquisition constraints, and a time range; the required Sentinel-1 TOPS bursts are retrieved, cached, and processed through configurable SNAP graphs. “sarpyx” manages batch execution and tiling to support large product volumes and produces geocoded outputs suitable for GIS and downstream analytics.

We focus on the engineering choices required for operationalization: declarative pipeline configuration, capture of processing provenance (input product IDs, auxiliary orbit/DEM dependencies, and graph versions), deterministic product naming, and automated quality indicators (e.g., coherence statistics and coregistration diagnostics). The environment generates standard intermediate and final data products including coregistered SLCs, interferograms, coherence and amplitude mosaics, and geocoded rasters, packaged to integrate with catalogues and service endpoints. A demonstration on representative Sentinel-1 stacks illustrates how burst-centric acquisition and tiling reduce unnecessary I/O and accelerate iteration from research prototypes to repeatable processing services.

The launch of radar satellites like the European Copernicus Sentinel-1 mission (S1) paved the way for free of charge massive amounts of SAR data availability. This huge influx of spatial images requires adapting the way we work with the data and associated derived products. Indeed, new specific tools and services must be developed to access, store, manage and use them, including by non-expert users. To address this challenge in France, the FLATSIM (FormaTerre LArge-scale multi-Temporal Sentinel-1 InterferoMetry) service provides the French community with ground motion measurements through large-scale processing of Sentinel-1 data using multi-temporal Interferometric Synthetic Aperture Radar (InSAR). Developed within the framework of the national data hub Data Terra FormaTerre, FLATSIM is based on the New Small temporal and spatial Baseline (NSBAS) chain. This service emerged from a fruitful partnership between the French Space Agency (CNES, currently operating the service) and scientific teams from French laboratories involved in the development of NSBAS and InSAR tools, which also led to the creation of the on-demand InSAR processing platform (GDM-SAR-In). FLATSIM's primary focus is to support Earth science research projects covering regions larger than 250,000 km² using the Sentinel-1 data archive (available since October 2014). This service aims to address scientific research topics such as seismology, tectonics, volcanology, and the hydrological cycle, as proposed by the French scientific community. Three calls for proposals have selected 20 projects in 2020, 2022, and 2025, resulting in more than 800 To of InSAR-derived products.

We are pleased to announce that the restrictions on the FLATSIM products from the first call of proposals (2020) will soon be lifted, making over 70k products available to the research community. This release covers 7 regions: Afar, Andes, Ozark, Balkans, Okavango, Tarim, and Türkiye from 2014 to 2021. The dataset includes wrapped and unwrapped interferograms, spatial and temporal coherence maps, displacement time series and mean velocity fields, together with their associated auxiliary files. Products are provided in radar and ground geometry, filtered or not and at multiple resolutions (from 20m to 160m). FLATSIM's processing approach offers several key advantages over alternative chains and services, including predictive atmospheric corrections applied before phase unwrapping to reduce artefacts and ease subsequent unwrapping. Then, unwrapping is performed iteratively, guided by the amplitude of the filtered signal, ensuring robust phase recovery over large areas. The approach also relies on long temporal interferometric baselines to reduce phase biases and improve noise integration in time series analysis, as well as the processing of a large number of bursts to provide a GNSS-independent long-wavelength solution essential for tectonic and geodynamic deformation studies.

Building on this wealth of results, we will present the product architecture and coverage areas, showcase key analyses derived from these data as published in the literature, and provide updates on the next steps and future developments for the service.

The increasing availability of high-frequency satellite radar acquisitions has enabled the transition of Interferometric Synthetic Aperture Radar (InSAR) from retrospective deformation analysis to operational ground motion monitoring for industry applications. One such application is in the energy sector, where ground deformation directly reflects subsurface pressure and mass redistribution processes from practices such as fluid injection, hydrocarbon production, and underground gas storage. However, conventional InSAR products remain largely velocity-based, emphasizing long-term linear trends. Such representations are insufficient for operational environments where timely detection of deviations from expected behavior is critical.

This work presents the integration of a Cumulative Sum (CUSUM) change-detection framework into an InSAR monitoring service designed for continuous surveillance of energy assets. The objective is not merely to quantify displacement rates, but to detect statistically significant changes in deformation trends at the earliest possible stage. CUSUM, widely used in statistical process control [1], accumulates sequential deviations from an expected model, thereby amplifying small but persistent departures that would otherwise remain masked by noise or gradual transitions. As such, it has been successfully used for detecting volcanic unrest [2].

Within the proposed processing environment, deformation time series derived from Persistent Scatterer and Small Baseline Subset (SBAS) workflows are continuously updated as new acquisitions become available. An expected deformation model, typically linear or linear and seasonal, is estimated over a reference interval. Residual deviations are subsequently evaluated using one-sided and two-sided CUSUM statistics, with thresholds calibrated to site-specific noise characteristics. The approach is computationally lightweight and scalable to millions of measurement points, enabling portfolio-level deployment across geographically distributed assets.

The method has been operationally tested in multiple industrial settings. In California, for example, oil production fields such as the Belridge field produce from highly compressible diatomite reservoirs, where fluid injection is used to enhance recovery and maintain reservoir pressure. Because diatomite is mechanically weak and prone to compaction, subsurface pressure must be carefully managed. Both underpressure (leading to compaction and subsidence) and overpressure (increasing risks to well integrity and potential leakage pathways) require close monitoring.

CUSUM statistics, derived from multi-year InSAR time series, clearly reflect large-scale changes in injection/extraction strategy, while remaining nearly imperceptible in the overall InSAR displacement time series. Thus, it provides previously unseen feedback on the response to subsurface pressure changes. Additionally, through parameter tuning, we can highlight short-term deviations as well, which is useful for detecting underground leaks, also known as blisters.

From a service perspective, CUSUM enhances standard deformation products by introducing a temporal anomaly statistic that converts descriptive displacement information into actionable monitoring indicators. The algorithm allows for parameter tuning to highlight relevant displacement rate anomalies. This enables InSAR-based services to move beyond static velocity maps toward dynamic, process-aware monitoring frameworks that integrate smoothly with Subsurface pressure management and proactive risk mitigation.

[1] Chang, T. C., & Gan, F. F. (1995). A Cumulative Sum Control Chart for Monitoring Process Variance. Journal of Quality Technology, 27(2), 109–119.

[2] Albino, F., Biggs, J., Yu, C., & Li, Z. (2020). Automated methods for detecting volcanic deformation using Sentinel‐1 InSAR time series illustrated by the 2017–2018 unrest at Agung, Indonesia. Journal of Geophysical Research: Solid Earth, 125(2), e2019JB017908.

The growing volume and diversity of SAR acquisitions require processing chains capable of producing InSAR time series at scale. Operational ground deformation monitoring, in particular, demands workflows that handle large numbers of interferometric pairs from multiple sensors and orbital geometries, run incrementally as new acquisitions become available, and operate with minimal manual intervention. AMSTer (SAR & InSAR Automated Mass processing Software for Multidimensional Time series) is an open-source toolbox that provides a complete, automated workflow from SAR data download to multidimensional deformation time series and web-based dissemination of results.

AMSTer integrates three main components. The AMSTer Engine is a command-line InSAR processor derived from the CSL InSAR Suite (Centre Spatial de Liege). It performs coregistration, interferogram computation, filtering, phase unwrapping, and geocoding for a wide range of SAR sensors: ERS-1/2, EnviSAT, ALOS/ALOS-2/ALOS-4 (upcoming), RadarSAT, CosmoSkyMed, TerraSAR-X, TanDEM-X, Sentinel-1 A/B/C/D, Kompsat-5, PAZ, SAOCOM, ICEYE, and NISAR. The Engine also includes split-band interferometry capabilities for ionospheric correction and absolute phase unwrapping, multiple unwrapping algorithms (SNAPHU, a branch-cut method, and the exploratory pre-unwrapping tool DetPhun), adaptive filtering, and a multi-level masking strategy. The MSBAS (Multidimensional Small Baseline Subset) module performs time series inversion by combining deformation maps from multiple satellites and acquisition geometries. In its standard configuration, it retrieves displacement components in two dimensions (east-west and vertical). A three-dimensional inversion mode is available either when sufficient viewing geometry diversity exists (e.g. combining right-looking and left-looking acquisitions) or when displacement is assumed to occur along the steepest slope, using slope gradients as an additional constraint. The AMSTer Toolbox consists of bash and python scripts that orchestrate the full processing chain, from data download through interferometric processing, MSBAS inversion, and web-based product dissemination.

Beyond mass processing for deformation time series, AMSTer can also be used for individual differential interferogram computation (for deformation measurement or DEM generation), as well as for producing coregistered time series of coherence or amplitude maps suitable for pixel tracking, for example. Several pair selection strategies are available to build the interferometric network: baseline criteria with dual temporal and spatial thresholds (to accommodate changes in the orbital tube, e.g. after the loss of Sentinel-1B), Delaunay triangulation, k-shortest connections, and an optimization procedure based on a coherence proxy. Sentinel-1 ETAD (Extended Timing Annotation Dataset) data can be integrated to apply tropospheric, ionospheric, and geodetic corrections to the interferometric products and to improve the geometric accuracy of the geocoding to the centimetric level. Recent developments include support for NISAR data processing, a diagnostic toolbox for quality control, and tools for exporting products as Cloud Optimized GeoTIFFs.

AMSTer is structured for fully automated operational monitoring. The processing chain is organized in sequential steps scheduled through cron jobs: data download, image reading and coregistration on a Global Primary image, interferometric mass processing, and MSBAS inversion with time series extraction. Each step is incremental, processing only newly available data without reprocessing the existing archive. Built-in verification procedures check image integrity, detect conflicts between concurrent processing runs, quarantine problematic acquisitions, and identify duplicate products. At the European Center for Geodynamics and Seismology (ECGS), AMSTer is operationally deployed for the continuous monitoring of multiple targets worldwide, including South Kivu (Democratic Republic of the Congo), Piton de la Fournaise (La Reunion, France), Domuyo and Laguna del Maule (Argentina), Galeras (Colombia), the Comoros archipelago, Luxembourg, the Himalayan Mountains (Nepal), and so on. For each target, the toolbox processes several thousands of interferograms across ascending and descending geometries. Results are disseminated through automatically updated web pages, which can be password-protected when needed, displaying baseline plots, velocity maps, amplitude time series, and (double-difference) deformation time series at predefined points of interest. Differential deformation maps and additional time series can be generated on demand. This web interface allows both project partners and monitoring agencies to access regularly updated deformation products without requiring InSAR expertise.

AMSTer is distributed on GitHub under the GNU AGPL v3 license, together with an installer script, a user manual, installation guides for Linux and macOS, processing flowcharts, and training course materials. The software has been under continuous development since the early 2010s and is actively maintained to keep up with new SAR missions and evolving data access infrastructure. On a best effort basis, the European Center for Geodynamics and Seismology (ECGS) can provide assistance for installation and training.

The Philippines is one of the most vulnerable countries in the world, due to its geological situation, human activities and the impact of climate change. Earthquakes, volcanic eruptions, landslides, subsidence, floods, typhoons, and sea level rise in its coastal regions are examples of resultant geohazard and environmental challenges. In order to strengthen the Philippines’ response capability and resilience to such natural and man-made hazards through the strategic use of space-based data, and help reduce vulnerabilities of the people of the Philippines to climate hazards, support climate adaptation, and environmental protection, the European Commission signed a contribution agreement with the European Space Agency (ESA) for the implementation of a national Copernicus data centre in the Philippines in 2023 (https://copphil.philsa.gov.ph). This agreement initiated the ESA-funded CopPhil project: Earth Observation Service Development & Transfer, which officially commenced in January 2024. One of the thematic objectives of CopPhil project is to promote the uptake of Copernicus data in the Philippines through the development of a Ground Motion Monitoring Service (GMS). The GMS enables the local stakeholders and partners to monitor, analyze, interpret and further predict ground motion related to those hazards, through a nationwide, automated, user-friendly, standardized workflow, supported by cloud computing.

The CopPhil-GMS is based on time series InSAR (Interferometric Synthetic Aperture Radar) techniques, including two approaches: SBAS (Small BAseline Subset) [1] and PSI (Persistent Scatterer Interferometry) [2], and is built entirely upon open-source software and tools. Over the past two years, we have developed a generic GMS processing workflow using the improved GMTSAR software for interferogram generation and SBAS processing [3] and STAMPS processing tool [4] for PSI processing. This workflow is implemented within a Docker environment in which all necessary software and dependencies are automatically pre-installed, and can 1) automatically access Sentinel-1 SAR data through Creodias (https://creodias.eu) without local downloading by mounting the data directly into the processing environment, and seamlessly provide DEMs and orbital data using conda-installed packages within the Docker container; 2) perform end-to-end tooling by completing the whole SBAS or PSI processing executed with a single command, while also exposing key processing parameters (such as region of interest, temporal baseline, multilooking factor in range and azimuth direction, master acquisition date, amplitude dispersion index and atmospheric phase removal method) for end-user customization; 3) generate three types of InSAR products, taking EGMS (European Ground Motion Service, https://egms.land.copernicus.eu) as a reference: P1.1 - Basic: Line of sight velocity maps and deformation time series with annotated quality measures per measurement point. P1.2 - Calibrated: Line of sight velocity maps and deformation time series in ascending and descending orbits referenced to a model derived from GNSS (Global Navigation Satellite System) time-series data. P1.3 - Ortho: components of motion (horizontal East-West and vertical Up-Down) retrieved by combining the ascending and descending products [5, 6]; 4) offer the products in CSV, GeoPackage and KMZ (Keyhole Markup Language Zipped), which are codefined with local stakeholders. The latest version of our open-source GMS tool, named GMTSAR+, is will be publicly available on Github. For details, refer to [7].

As a GMS demonstration, we processed three years of Sentinel-1 SAR images that cover ~50% of the Philippines’ territory, focusing on Luzon and Mindanao, using the Geoville cloud computing platform. This results in the generation of P1.1 and P1.3 GMS products –SBAS deformation time series at a 100 m spatial resolution. The P1.2 GMS product is still in the pipeline, pending the collection of sufficient GNSS observations from local stakeholders, while the corresponding processing workflow has already been fully integrated and tested within the GMS framework.

References:

1. Berardino, P., Fornaro, G., Lanari, R., Sansosti, E., 2003. A new algorithm for surface deformation monitoring based on small baseline differential SAR interferograms. IEEE Transactions on geoscience and remote sensing 40, 2375–2383. doi:10.1109/TGRS.2002.803792.
2. Ferretti, A., Prati, C., and Rocca, F., 2001. Permanent scatterers in SAR interferometry. IEEE Transactions on Geoscience and Remote Sensing, 39(1):8–20.
3. Sandwell, D., Mellors, R., Tong, X., Wei, M., Wessel, P., 2011. Open radar interferometry software for mapping surface deformation. doi:10.1029/2011EO280002.
4. Hooper, A., Zebker, H., Segall, P., Kampes, B., 2004. A new method for measuring deformation on volcanoes and other natural terrains using InSAR persistent scatterers. Geophysical research letters 31. doi:10.1029/2004GL021737.
5. Hanssen, R.F., 2001. Radar interferometry: Data interpretation and error analysis. Springer. doi:10.1007/0-306-47633-9_4.
6. van Leijen, F.J., 2014. Persistent Scatterer Interferometry based on geodetic estimation theory. 86, Netherlands Geodetic Commission.
7. Li, B., Feng, W., Chang, L., 2026. GMTSAR+: An extended GMTSAR workflow streamlining SBAS products with standardized geospatial outputs. Submitted to Environmental Modelling & Software.