Conference Agenda

Sentinel-1 (S-1) single-look complex (SLC) synthetic aperture radar (SAR) images allow for a very high geolocation accuracy down to centimetric levels, if one makes use of the mission’s precise orbit products and if the known perturbing effects are compensated for in post-processing. These perturbing effects involve the path delays of troposphere and ionosphere, the Earth’s tidal deformation signals, and residual systematic effects associated with the S-1 SAR data processor. All of these effects have been studied and validated extensively by the S-1 SAR Mission Performance Cluster (SAR-MPC) [1]. The SAR-MPC is an international consortium of SAR experts and is in charge of the continuous monitoring of the S-1 instruments status, as well as the monitoring of the quality of the L1 and L2 products. This monitoring also includes the geolocation quality of S-1 mission. Therein, non-critical but nevertheless noteworthy residual offsets of 0.1 to 0.3 meters are still present to various degrees in azimuth geolocation residuals of reference corner reflectors (CRs) measured with S-1A/B/C satellites [1]. They are considered to be at least partly related to SAR antenna characteristics of the particular spacecraft [2]. Investigation and possible mitigation of these effects is an ongoing effort at SAR-MPC.

As part of SAR-MPC’s contribution to the S-1C and S-1D in-orbit commissioning activities, geolocation analyses have been performed over a multitude of calibration sites, including El Sosneado and Casleo in northern Argentina, which are operated by CONAE for the SACOM mission [3]. These two sites offer several unique features which make them very interesting for our S-1 data analysis: Each site is equipped with eight or more very large CRs of 3 m inner-leg size, which are differently oriented to support both the ascending and the descending passes. Even with the medium-resolution S-1 IW C-band data, they enable precise azimuth geolocation measurements of 0.1 m or better due to their large dimensions. Moreover, the sites are situated in the high-altitude terrain of the Andes, which is located some 30 degrees south of the geomagnetic equator, a region that this strongly affected by ionosphere dynamics. Finally, the sites happen to show up at burst edges or in the burst overlap area of the tightly maintained S-1 IW data footprints, where the SAR Doppler gets largest and thus any TOPS data focussing limitations or antenna effects become more evident. All these features make it challenging to perfectly accommodate the S-1 azimuth geolocation results of these CRs for the various available pass geometries, exposing any shortcomings in the S-1 TOPS data processing and the ionospheric corrections.

Currently, we are investing three different effects that should allow us to even further enhance the azimuth geolocation accuracy of S-1 TOPS data and possibly resolve the dcm-level residual azimuth shifts among the sub-swaths: Firstly, there is the impact of the azimuth antenna pattern. The loss of isolated transmit-receive-modules (TRMs), which at some point occurred for each of the S-1 satellites, or of an entire antenna tile such as for S-1A in June 2016 [2], can alter the azimuth antenna phase pattern and therefore introduce sub-swath-dependent shifts if not considered during the processing. Secondly, hyperbolic modelling of the sensor-to-ground distance function, which is part of the SAR-IPFs focussing scheme, deviates from the true distance function. For S-1 TOPS data, the deviation towards the edge of the bursts becomes large enough to cause residual shifts of up to 0.1 m. Thirdly, there are the azimuth shifts introduced by the ionosphere if strong along-track ionospheric gradients are present in the scene. In principle, this effect is well understood but the coarse resolution of ionospheric total electron content (TEC) maps makes it difficult to accurately determine the effect. SAR-based interferometric methods such as split-spectrum analysis offer a mean to derive relative ionospheric azimuth shifts, but they operate on image pairs and require good coherence for stable results [4]. Again, the CONAE sites are favourable in that regard, as they are located in the arid regions of the Andes that provide high long-term-stable coherence in SAR data, which allows us to compare both methods.

In summary, the large CRs of the two CONAE sites provide a mean to test the modelling of these three azimuth shift effects against ground truth. In our conference contribution, we plan to present them in more detail and report on the status of our analysis. Eventually, the outcomes will also contribute to further advancement of the S-1 Extended Timing Annotation Dataset (ETAD) product, which offers post-processing corrections for all the consolidated S-1 perturbation effects in a comprehensive product in-line with S-1 level-1 data [5].

[1] G. Hajduch et al.: S-1 Annual Performance Report for 2024. Technical note by SAR MPC, SAR-MPC-0715, issue 2.1, 02/04/2025. Available on SentiWiki: https://sentiwiki.copernicus.eu/web/document-library

[2] MPC-S1-team: Sentinel-1A Tile #11 Failure. Technical note by SAR MPC, MPC-0324, issue 1.2, 13/10/2026. Available on SentiWiki: https://sentiwiki.copernicus.eu/web/document-library

[3] M. Thibeault, CONAE Target Sites. Proceedings of CEOS SAR Cal/Val Workshop 2021. Available Online: https://calvalportal.ceos.org/sarcv

[4] G. Gomba, F. Rodríguez González and F. De Zan: Ionospheric Phase Screen Compensation for the Sentinel-1 TOPS and ALOS-2 ScanSAR Modes, in IEEE Transactions on Geoscience and Remote Sensing, vol. 55, no. 1, pp. 223-235, Jan. 2017, doi: 10.1109/TGRS.2016.2604461

[5] Gisinger, C., Libert, L., Marinkovic, P., Krieger, L., Larsen, Y., Valentino, A., Breit, H., Balss, U., Suchandt, S., Nagler, T., Eineder, M., Miranda, N.: The Extended Timing Annotation Dataset for Sentinel-1 - Product Description and First Evaluation Results. IEEE Transactions on Geoscience and Remote Sensing, vol. 60, pp. 1-22, 2022. doi: 10.1109/TGRS.2022.3194216

Acknowledgement

The results presented here are outcome of the ESA contract Sentinel-1 / SAR Mission Performance Cluster Service 4000135998/21/I BG, funded by the EU and ESA. The views expressed herein can in no way be taken to reflect the official opinion of the European Space Agency or the European Union.

Synthetic aperture radar (SAR) data is now a well-known source of surface displacement data for earthquakes under or near land. The NASA-ISRO SAR (NISAR) mission is a new SAR satellite with two radar systems, the L-band (24 cm wavelength) SAR provided by NASA and the S-band (12 cm wavelength) SAR provided by ISRO and has a 12-day repeat cycle. Both are left-looking instrument. The L-band SAR images all land areas between 77.5°N and 87.5°S in 240 km wide swaths, while the S-band SAR is operated over South Asia and a few other locations worldwide. NISAR entered the science operations phase at the beginning of January 2026. NISAR is the first SAR satellite to provide standard products that include level 2 geocoded unwrapped interferograms (GUNW) for the adjacent-in-time pairs of NISAR scenes with 80-meter pixel spacing. The GUNW products include an estimate of the ionospheric correction (phase screen). The analysis-ready data (HDF5) files for the GUNW products also include geocoded wrapped interferograms with 20-meter pixel spacing. The NISAR L-band SAR data products will be available soon through the NASA EarthData system, easily searchable through the Alaska Satellite Facility Vertex search tool or through applications programming interfaces. The NISAR S-band products will be available through the ISRO data portal, Bhoonidhi. The standard product latency provides the processed L2_L_GUNW data within two days after acquisition by the satellite, when the most accurate satellite orbits are available. The NISAR mission has a special urgent response system that can give higher priority to the data downlink from the satellite and expedited L-band processing using preliminary orbits with a goal of products available, including GUNW, within 2-6 hours after data downlink from the satellite. The system is triggered automatically for earthquakes and volcanic eruption, or manually through requests from vetted U.S. agency end users. For earthquakes, it is triggered automatically from the USGS earthquake feed based on magnitude (near USA or India magnitude 7.0 with depth less than 50 km) or on the PAGER estimate (yellow/orange/red in USA or India and orange/red in rest of the world). The preliminary L-band diata products will be available immediately after processing.

Digital Elevation Models (DEMs) are static expressions of elevation. InSAR products provide estimates of elevation change over time. In highly dynamic and/or subsiding low-lying regions the combination of both parameters is required. We propose a Dynamic Digital Elevation Model, or D-DEM, describing elevation in a uniform geodetic datum as a function of time rather than as a static value. The D-DEM defines for individual geographic positions or areas a model of elevation as a function of time, and specifies alongside the estimated parameters their quality descriptions as well.

We create a D-DEM by combining five geodetic techniques: leveling, GNSS, Airborne Laser Scanning (ALS), InSAR, and absolute gravimetry, over uniform grassland parcels, i.e., areas that were until recently considered to be incoherent for time series C-band InSAR applications. The D-DEM utilizes datasets ranging from the early 1950s (so-called surface leveling) up to recent Sentinel-1 SAR acquisitions.

Instantaneous elevation data sets have been acquired at five epochs; one from surface leveling and four from ALS. We supplement these DEMs with four tracks of Sentinel-1 imagery from 2015 onwards.

The subsequent DEMs exhibit biases due to reference frame differences and a shifting vertical reference frame (VRF). Vertical surface motion cannot directly be estimated reliably from subsequent DEMs, as the possible biases are sufficiently large to introduce major biases in the subsidence rate estimates. These biases are therefore estimated by using collocation of leveling benchmarks located in buildings of which the rooftop ridge line is visible in the ALS DEMs. Assuming internally rigid buildings, the vertical motion of the benchmark and the roof ridge line are assumed to be identical, allowing for estimation of the relative biases between the DEMs. The leveling benchmarks and DEMs are corrected for potential vertical motion of the VRF reference benchmarks estimated from time series of absolute gravimetry, resulting in five DEMs referenced to the same datum, yet this datum may also still be biased with respect to the Amsterdam Ordnance Datum we intend to use. The bias of the reference DEM with respect to ITRF is estimated using Integrated Geodetic Reference Stations (IGRS), collocating a GNSS module, an ALS reference plane, and InSAR corner reflectors in one instrument.

For InSAR the IGRS allows us to eliminate the motion of the reference point, which is known from GNSS. From the five corrected DEMs the elevation per spatial unit at the reference epoch is estimated. A kinematic model based on the expected temporal behavior is estimated for all observations.

We applied the methodology over highly dynamic peat soils in the ’Green Heart’ region in the Netherlands. Over the entire Green Heart we estimate an average mean subsidence rate of −4.1 mm/y between the 1960s and 2023. Validation with reference datasets in small regions of the Green Heart shows that our models are statistically the same as the reference datasets. Because of the combination of elevation and elevation change in the D-DEM, elevation at any epoch can be derived anywhere, even in such highly dynamic environments.

The Copernicus SAR mission Sentinel-1 is currently providing systematic global data acquisition, offering the scientific community both extensive historical archives and continuous new observations at an unprecedented volume. ROSE-L, an L-band SAR mission to be launched in the coming years, will further extend these capabilities complementing the characteristics of Sentinel-1. To ensure that data from both the operational Sentinel-1 mission and the future ROSE-L mission can be exploited efficiently, particularly for time series analysis, the definition of Analysis Ready Data (ARD) products and the development of corresponding processors are essential.

The Committee on Earth Observation Satellites (CEOS) [1] is promoting the definition of SAR ARD with a set of data standards and guidelines aimed at improving quality, interoperability, and accessibility of EO data acquired from various satellite SAR missions.

In this framework, ESA funded the project “Prototype Processor for ARD of Copernicus SAR Missions” to analyze candidate ARD products and to implement prototype processors suitable for both Copernicus SAR missions.

The main objective of this project is to define advanced algorithms and implement and validate the corresponding processors and products for a new generation of Analysis-Ready Data (ARD). This abstract presents the development and utility of these products, which are engineered to provide end-users with easy-to-use data while maintaining state-of-the-art quality. By prioritizing an intuitive user experience, we have created solutions that are highly accessible yet fully capable of delivering professional-grade results.

The project is organized in two phases. The first phase focused on analysis and on generating demonstration products derived from Sentinel-1 and from existing L-band SAR missions. These demonstrators were designed to represent a set of use cases aligned with ROSE-L’s primary scientific objectives.

At the end of Phase 1, two ARD products were selected for prototype implementation:

• CSLC: Co‑registered Single Look Complex
• GMLP: Geocoded MultiLooked Phase

The selected products are suitable for timeseries analysis and enable advanced interferometric processing either at full resolution in radar geometry (CSLC), preserving the complete information content, or in multi-looked and geocoded form (GMLP), which simplifies processing while still maintaining rigorous information quality. Similar, though not fully equivalent, products are foreseen in the NISAR mission, based on ISCE 3 software [2].

Products

The CSLC product provides geometrically aligned SAR images on a common reference radar grid, compensating for geometric distortions, atmospheric effects (tropospheric and ionospheric), and long-wavelength artefacts such as tidal motion. It preserves the full fidelity of the original SAR measurements: amplitude is calibrated to Beta Nought, or to Sigma Nought [3] and the complex phase is maintained without distortion. DEM-based distorsion compensation [4] and state-of-the-art model-based [5] and data-driven corrections for geometric and atmospheric contributions, including Enhanced Spectral Diversity (ESD) [6] and split-spectrum methods [7], are integrated in the processing. As data are preserved in native radar geometry, the CSLC product is ideally suited for high-precision interferometric applications, including the Small BAseline Subset (SBAS), the Persistent Scatterer (PS), and concept of Distributed Scatterer (DS) appraoches, coherence analysis, and phase unwrapping [8-11].

The GMLP product is an innovative ARD product that inherits all CSLC corrections while simplifying interferometric processing for both single interferometric pairs and full time series [12-13]. It is geocoded to a standard cartographic coordinate system (UTM), enabling seamless integration into GIS software. The phases are referenced to a unique temporal reference, inherently enforcing phase closure and allowing immediate interferogram generation from any pair of acquisition dates. The processing workflow incorporates advanced bias correction strategies to minimize long-term phase drifts without requiring exploitation of the full covariance matrix, facilitating efficient updates as new acquisitions arrive. Auxiliary coherence layers accompany the product to support phase unwrapping and provide information on long-term phase stability.

Prototype processor

State-of-the-art algorithms are adopted throughout the processing chain. The prototype processors are implemented in Python, leveraging the scientific software stack Xarray, Dask, and Zarr to enable cloud-native, scalable, and parallel geospatial data processing. This architecture supports chunked and lazy computation, distributed processing on high-performance or cloud infrastructures, and the generation of analysis ready outputs in interoperable formats. The design emphasizes modularity, allowing integration of alternative correction modules or new sensors, and ensures that the processing can scale to the data volumes expected from Sentinel-1 and ROSE-L systematic acquisition strategy.

ARD products and processors are validated using both real and simulated datasets to cover the full range of representative use cases. The validation methodology also anticipates the future availability of ROSE-L data and defines clear procedures to verify ARD product performance once the mission is operational.

Validation approach

The validation strategy for CSLC products focuses on assessing their suitability for advanced interferometric applications. A primary metric is co-registration accuracy, verified using artificial corner reflectors (CRs) to ensure compliance with stringent requirements for image alignment. The quality of the radar signal is further evaluated through the amplitude stability of these same CRs and the phase stability of persistent scatterer (PS) points extracted from the scene. These empirical measurements are complemented by a theoretical model comparison, where the product's statistical properties are checked against expected distributions. Finally, an external software comparison using other tools like ISCE [2] (Interferometric synthetic aperture radar Scientific Computing Environment) provides a benchmark by analyzing the quality of the coherence of generated interferograms, thereby confirming the product's consistency.

For the GMLP, the validation approach is tailored to its multi-looked and geocoded nature, emphasizing its readiness for large-scale deformation studies. A direct comparison with an independent implementation like EMI [12] (Eigendecomposition based Maximum-likelihood-estimator of Interferometric phase) is performed, focusing on distributed targets through the analysis of short-term coherence. For point-like targets, the validation examines the long-term coherence and compares the measured phase with the average phase at PS locations. Importantly, this process validates the quality of this phase product without requiring the execution of a complete interferometric PS or SBAS chain. The analysis incorporates a theoretical coherence model and evaluates short-term noise characteristics using SLCs from end-to-end simulations. To ensure geophysical signal fidelity, long-term trend consistency is assessed by comparing phase-derived trends from the L-band GMLP product with independent measurements from C-band sensors at compatible spatial scales.

Conclusions

The analysis and prototype development carried out as part of this project demonstrate the feasibility and scientific value of CSLC and GMLP products as ARD products for current and future Copernicus SAR missions. The processors, based on cloud-native geospatial technologies, successfully address the challenges posed by large-scale systematic SAR acquisitions and provide a scalable basis for generating high-quality ARD products. The validation results confirm that ARD products meet the accuracy and stability requirements for advanced interferometric applications, supporting a wide range of scientific and operational use cases. These developments are the basis for the future integration of ROSE-L and Sentinel-1 data and will enable more efficient, robust, and interoperable SAR time series analysis across the Copernicus program.

REFERENCES

[1] CEOS ARD Data Definition Team (2025). Combined CEOS-ARD for Synthetic Aperture Radar. Product Specification Document https://ceos.org/ard/files/PFS/SAR/v1.2/ CEOS-ARD_PFS_Synthetic_Aperture_Radar_v1.2.pdf.

[2] Fattahi, H., “ISCE3: InSAR Scientific Computing Environment Enhanced Edition”, vol. 2023, no. 488, Art. no. G23C-0488, 2023.

[3] Shiroma, G. H. X., Lavalle, M., and Buckley, S. M. (2022). An area-based projection algorithm for SAR radiometric terrain correction and geocoding. IEEE Transactions on Geoscience and Remote Sensing, 60:1–23.

[4] Sansosti, E., Berardino, P., Manunta, M., Serafino, F., and Fornaro, G. (2006). Geometrical SAR image registration, IEEE Transactions on Geoscience and Remote Sensing

[5] Sentinel-1 ETAD Product evolution: new processor version (v3.0) https://sentinels.copernicus.eu/web/sentinel/-/sentinel-1-etad-product-evolution-new-processor-version-v3.0

[6] Yagüe-Martínez, N., Prats-Iraola, P., Rodríguez González, F., Brcic, R., Shau, R., Geudtner, D., Eineder, M., and Bamler, R. (2016). Interferometric processing of Sentinel-1 TOPS data. Geoscience and Remote Sensing, IEEE Transactions on, 54(4):2220–2234.

[7] Gomba, G., Parizzi, A., De Zan, F., Eineder, M., and Bamler, R. (2016). Toward operational compensation of ionospheric effects in SAR interferograms: The split-spectrum method. IEEE Transactions on Geoscience and Remote Sensing, 54(3):1446–1461.

[8] Berardino, P., Fornaro, G., Lanari, R., and Sansosti, E. (2002). A new algorithm for surface deformation monitoring based on small baseline differential SAR interferograms. IEEE Transactions on Geoscience and Remote Sensing, vol. 40, no. 11

[9] Ferretti, A., Prati, C., and Rocca, F. (2000). Nonlinear subsidence rate estimation using permanent scatterers in differential SAR interferometry. Geoscience and Remote Sensing, IEEE Transactions on, 38(5):2202–2212.

[10] Ferretti, A., Fumagalli, A., Novali, F., Prati, C., Rocca, F., and Rucci, A. (2011). A new algorithm for processing interferometric data-stacks: Squeesar. Geoscience and Remote Sensing, IEEE Transactions on, 49(9):3460–3470.

[11] Vecchioli, F., Costantini, M., Minati, F., and Zavagli, M. (2023). A Novel Algorithm for Point Coherence Estimation in SAR Interferometry. IGARSS 2023 - 2023 IEEE International Geoscience and Remote Sensing Symposium, 7868–7871.

[12] Ansari, H., De Zan, F., and Bamler, R. (2018). Efficient phase estimation for interferogram stacks. IEEE Transactions on Geoscience and Remote Sensing, 56(7):4109–4125.

[13] De Zan, F. (2022). Recursive and robust InSAR phase estimation. In EUSAR 2022; 14th European Conference on Synthetic Aperture Radar, pages 1–5.

CEOS Analysis-Ready Data (CEOS-ARD) is a joint effort by the Committee on Earth Observation Satellites (CEOS) to streamline data flows and enable interoperable products between sensors and data providers, and, specifically, to broaden the Earth Observation user community by provision of data products that do not require expert knowledge to ingest and analyse. This last point is perhaps particularly relevant for Synthetic Aperture Radar (SAR), where the potential to contribute to today’s great environmental challenges with unique information is significant, but with the SAR user community remaining small and expert-oriented even after 30 years of operational SAR missions. CEOS-ARD is an opportunity to bridge that gap.

In a coordinated effort by the CEOS Land Surface Imaging Virtual Constellation (LSI-VC) and the CEOS Working Group on Calibration and Validation (WGCV) SAR Subgroup, four SAR-specific specifications, included in a single unified “CEOS-ARD for SAR” Product Family Specification (PFS) have been developed:

• Normalised Radar Backscatter (NRB).
• Polarimetric Radar (POL).
• Ocean Radar Backscatter (ORB)
• Geocoded Single-Look Complex (GSLC).
• Interferometric Radar (INSAR).
• Multi-Source Backscatter (MSB) – under development

The CEOS-ARD GSLC product describes the complex radar reflectivity on the surface with all propagational phases removed, so that the amplitude and phase values represent properties of the surface and not the instrument. GSLC data are presented in a common, often user-defined, ground-based coordinate system (e.g. UTM, geographical coordinates, etc.), rather than in radar slant range coordinates, to facilitate use by non-radar-specialists.

The CEOS-ARD INSAR product specification covers a suite of three products generated by InSAR processing of (at least) two images captured of the same geographic area at different times:

• Wrapped interferogram: Image of differential phase signals between two SLC images
• Unwrapped interferogram: Image of differential phase signals where the wrapped fringes are summed (“unwrapped”) to give a continuous phase signal across the image
• Interferometric coherence: Image of phase coherence between the two images.

The NRB, POL, ORB, GSLC and INSAR specifications have been endorsed by CEOS LSI-VC and can be accessed on the CEOS ARD website (ceos.org/ard). The MSB product specifications is being prepared for endorsement in April 2026.

There has been a significant interest in CEOS-ARD by space agencies and public and private data providers. The first CEOS-ARD products for the InSAR community have been developed by NASA JPL and ISRO. The JPL OPERA project has had its Sentinel-1 based InSAR surface displacement products passed as compliant against the PFS, with both Sentinel-1 and NISAR GSLC products currently under development for assessment. Furthermore, ISRO is developing NISAR GSLC products. In this presentation we will show some examples of these new InSAR ARD products.