Conference Agenda

Earth is a highly dynamic system where the transport and exchanges of energy and matter are regulated by a multitude of processes and feedback mechanisms. Untangling these complex processes to better understand how Earth works as a system is a major challenge. Together with observations from Sentinel-1, Harmony will deliver a wide range of unique high-resolution observations of motion occurring at or near Earth’s surface. Over land, Harmony will provide data to estimate small shifts in the shape of the land surface, such as those leading to and resulting from earthquakes, volcanic activities and possible other ground motion sources, hereby contributing to the assessment of geohazards over geologically active areas. It will also provide new information to study dynamic volume changes and 3D deformation at the rapidly changing marginal zones of the ice sheets for a better understanding of the contribution of ice mass loss on sea-level rise.

This presentation will discuss the scientific goals of the Harmony mission within the solid earth and land ice domains, namely:

H-C1 Providing a consistent and highly resolved global glacier mass balance, filling major spatial gaps in the current observation of mountain glaciers and outlet glaciers of the ice sheets.

H-C2 Give new insight on the physical processes associated with the coupling between glacier mass change and ice dynamics. Through that, substantially improve understanding and prediction of rapid or even abrupt glacier changes, and the balance between vertical ice flow and mass accumulation/ablation.

H-G1 Map all components of global tectonic strain and the deformation caused by volcanism and the earthquake cycle. These measurements are required to constrain the geometry and nature of the driving processes and improve forecasts of the associated geohazards.

H-G2 Understand cycles of topographic growth, mass transport and collapse at actively erupting volcanoes and improve forecasts of the associated geohazards.

We will explain how these goals will be addressed by the Level 3 products of the Harmony Mission, namely, Three-Dimensional Velocities (TDV), Three-Dimensional Time Series (TDTS) and Topographic Change (TOC) and also Level 2 products where appropriate. We will also explain the criteria used to generate the proposed observational masks and the levels of prioritisation.

This contribution addresses the current status of the science study of the Harmony mission for land applications, namely, solid Earth and land ice. In particular, the focus will be put on the latest developments of the Harmony End-to-End Performance Simulator for land (HEEPS/Terra) and on the dedicated L2 and L3 processing algorithms.

Short before Harmony’s system CDR, scheduled to take place by Q1 2027, an assessment of the scientific readiness level 6 (SRL-6) shall be conducted. For this reason, it is necessary to address several scientific aspects of the mission, namely, the consolidation of the L2/L3 processing algorithms, the improvement of the end-to-end simulator representativity, which shall also include a first version of the full processing chain (the so-called breadboard processors), and the execution of a performance campaign demonstrating the feasibility of the mission objectives. All these activities will be conducted in the frame of the aforementioned science study for land applications.

A first version of the HEEPS/Terra simulator exists, which was developed during the Phase 0 and Phase A of the mission (see, e.g., [1][2]), and was used to demonstrate SRL-5 for land applications in preparation for the User Consultation Meeting at ESA, which took place in July 2022. Some of the upgrades that are currently being implemented in the current context are: the improvement of the forward model for land ice, which needs to consider the bistatic nature of the measurement in terms of the penetration depth of the signal through snow and firn; the exact reverse image formation of extreme bistatic geometries; the inclusion of more accurate system and instrument models, being the bistatic synchronization a critical aspect; as well as the development and integration of a first version of the L1/L2/L3 processors in the end-to-end simulation chain.

Concerning the processing algorithms, and similar as with the end-to-end simulator, a first version of the algorithm theoretical baseline documents (ATBD) is available for the different products. These L2/L3 algorithms are currently being consolidated and implemented, and are briefly described in the following:

- L2 processor for DEM products: this corresponds to the single-pass interferometric processor for the generation of the Harmony DEMs. It includes the steps of coregistration, spectral filtering, phase unwrapping and geocoding. A critical aspect is the consideration of the extreme bistatic geometry and the non-zero-Doppler output geometry, which will require the modification of some of the processing steps. This processor is used to generate the coregistered products (CoSLC) for all land applications. In the XTI phase, single-pass DEMs will be generated for both land ice areas (mass balance estimation) and volcanoes. For land ice applications, a dedicated step will take care of handling the penetration bias. The current processor prototype is based on the eo-tools software [3], which has been extended to handle the bistatic geometries of Harmony.

- L2 and L3 processor for TOC (Topography Change) products: these processors compute the difference between two DEM products, whereby the L3 processor will perform the mosaicking of different tracks, including bundle adjustment and the calibration of residual trends.

- L2 processor for TDV (3-D velocity maps) solid Earth products: this processor corresponds to the PSI processor responsible for the retrieval of the deformation time series exploiting the Harmony and Sentinel-1 image stacks. The goal is the retrieval of the 3-D surface deformation in order to derive the tectonic strain. The current approach is based on the exploitation of both point-like targets and distributed targets [4]. A novel aspect that will be included in the chain is the consideration of the ionosphere as part of the PSI processing [5]. The current processor prototype is based on the StaMPS processor [6], which is currently being translated to Python outside the science study.

- L2 processor for TDV (3-D velocity maps) land ice products: the non-stationary scenarios of land ice, with a large variety of velocity regimes, ask for different processing strategies. Fast moving glaciers shall exploit incoherent cross-correlation to retrieve the displacements, whereas the interferometric phase can be exploited over slow moving glaciers and ice sheets. In the latter, the azimuth-dependent Doppler centroid of the TOPS mode requires particular attention to avoid the introduction of biases in the velocity products.

- L3 processor for TDV products: this processor will perform the 3-D inversion out of the individual line-of-sight measurements, which applies to both solid Earth and land ice products.

The presentation will address the above aspects in detail, reporting on the status of the study at the time the conference takes place.

[1] P. Prats-Iraola, A. Pulella, A. Benedikter, A. Hooper, J. Biggs, A. Kääb, et.al., “Performance Analysis of the Harmony Mission for Land Applications: Results from the Phase A Study,” FRINGE 2023.

[2] Pau Prats-Iraola, Marc Rodriguez-Cassola, Irena Hajnsek, Andy Hooper, Eric Rignot, Andreas Kääb, Juliet Biggs, Francesco De Zan, Ramon Brcic, Helmut Rott, Thomas Nagler, Andrea Pulella, Georg Fischer, Simon Trumpf, Andreas Benedikter, Dominik Richter, Marcus Bachmann, Paco Lopez-Dekker, Björn Rommen, “Status of the Harmony Mission on End-to-End Simulations, Products and Processing Algorithms for Land Applications,” Living Planet Symposium 2025.

[3] https://github.com/odhondt/eo_tools

[4] A. Ferretti, A. Fumagalli, F. Novali, C. Prati, F. Rocca and A. Rucci, "A New Algorithm for Processing Interferometric Data-Stacks: SqueeSAR," in IEEE Transactions on Geoscience and Remote Sensing, vol. 49, no. 9, pp. 3460-3470, Sept. 2011, doi: 10.1109/TGRS.2011.2124465

[5] Navarro Sanchez, V. D., Gomba, G., De Zan, F., & Kretschmer, K. (2021). Compensation of ionospheric effects for InSAR stacks by means of a split-spectrum method. In 13th European Conference on Synthetic Aperture Radar, EUSAR 2021 (pp. 898-901). VDE Verlag GmbH.

[6] https://homepages.see.leeds.ac.uk/~earahoo/stamps/

A central objective of the Harmony mission in the cryosphere domain is to improve the quantification of glacier and ice sheet mass balance through globally consistent measurements of surface elevation change [1]. During the cross-track InSAR (XTI) phase, Harmony will deliver single-pass digital surface models (DSMs) that form the basis for topographic change (TOC) products [1]. However, InSAR elevation measurements over dry snow, firn, and ice are known to be substantially biased due to partial penetration of the radar signal into the volume, which shifts the interferometric phase center below the physical surface [2]. Correcting this penetration-related elevation bias is therefore essential to meet the accuracy requirements of cryospheric elevation products. The correction of the penetration bias is primarily addressed through coherence-based inversion techniques that relate the magnitude of the interferometric coherence to the depth of the phase center, assuming a uniform volume with exponential power extinction [3]. While this approach has been successfully applied in monostatic InSAR over homogeneous glacier and ice sheet areas [2], its extension to bistatic geometries and its robustness in heterogeneous firn volumes remain open challenges. In addition, propagation effects in the dielectrically dense firn medium introduce further biases that are commonly not fully accounted for in standard processing chains.

In this contribution, we present advancements to standard interferometric processing approaches for providing accurately calibrated surface elevation measurements in Harmony XTI acquisitions over ice sheets and glaciers. First, we show that InSAR processing over dielectrically dense media is affected by additional geolocation offsets (beyond the well-known penetration bias), such that the height and range distance measured with InSAR are biased with respect to the true location of the phase center [4]. These effects originate from an uncompensated stretching of the vertical wavenumber inside the medium and from refraction at the surface interface, potentially introducing residual elevation biases of several meters in conventionally processed InSAR DEMs that assume propagation in free space. For a uniform volume model of the firn structure, we demonstrate that both the offset between the phase center and the surface, as well as the propagation effects within the volume, can be naturally accommodated within the interferometric processing by applying phase and slant range corrections to the interferograms prior to the phase-to-height conversion. This approach directly provides an elevation measurement of the surface, i.e., a DSM. Importantly, the phase and slant-range corrections can be applied without requiring prior knowledge of the dielectric permittivity (i.e., density) of the volume. Only if an estimate of the phase center depth is desired as an additional product, an estimate of the bulk dielectric permittivity is required.

Second, we extend existing penetration bias models to the bistatic Harmony XTI geometry, explicitly accounting for the large transmitter–receiver squint angle between Sentinel-1 and the Harmony satellites and its impact on the vertical wavenumber within the volume. The analysis shows that, for identical extinction properties, bistatic and monostatic configurations exhibit only small differences in effective penetration depth [5]. The bistatic extension of the geometry model can be naturally accommodated within the nominal InSAR processing, including the phase and range corrections described above.

Beyond single-interferogram approaches, we propose exploiting multiple XTI acquisitions of the same scene acquired with different baselines or incidence angles as a complementary measurement strategy. The resulting diversity in vertical wavenumber provides additional information that allows a joint estimation of surface elevation and penetration depth directly from the interferometric phases, potentially reducing sensitivity to errors in the estimation of the coherence magnitude. This multi-interferogram inversion may also offer a means to detect and potentially model more complex vertical backscatter distributions beyond the standard uniform volume assumption. Especially at high latitudes, Harmony XTI acquisitions of the same scene with different observation geometries can be collected from neighboring orbits while minimizing the temporal lag between acquisitions. Small temporal lags (not exceeding several days) are likely required to maintain constant scattering and volume properties of snow and firn. Initial tests using airborne and TanDEM-X acquisitions over Greenland are used to validate the concept and show promising results.

[1] Kääb A, Mouginot J, Prats-Iraola P, Rignot E, Rabus B, Benedikter A, Rott H, Nagler T, Rommen B, Lopez-Dekker P. Potential of the bi-static SAR satellite companion mission Harmony for land-ice observations. Remote Sensing. 2024; 16(16):2918.

[2] Rott H., Scheiblauer S., Wuite J., Krieger L., Floricioiu D., Rizzoli P., Libert L., Nagler T. Penetration of interferometric radar signals in Antarctic snow. Cryosphere 2021, 15, 4399–4419.

[3] Dall J. InSAR elevation bias caused by penetration into uniform volumes. IEEE Trans. Geosci. Remote 2007, 45, 2319–2324.

[4] Benedikter A., Rodriguez-Cassola M., Prats-Iraola P., Krieger G., Fischer G. On the processing of single-pass InSAR data for accurate elevation measurements of ice sheets and glaciers. IEEE Trans. Geosci. Remote 2024, 62, 4300310.

[5] Fischer G., Belinska K., Benedikter A., Papathanassiou K., Hajnsek I., Rott H., Nagler T., Prats-Iraola P. Bistatic signal penetration geometry over land ice for the HARMONY mission. EUSAR 2026, accepted.

The achievement of the Harmony scientific goals requires the calibration of system phase signatures with precision and accuracies approaching 1 deg. One of the critical elements in this context is the time and phase synchronisation between the radars, i.e., between the Harmonies, but also between Sentinel-1 and each of the Harmonies, with the sole use of a GNSS-based synchronisation subsystem [Rodrigues-Silva et al., TGRS 2024]. The current best estimates of the performance of the GNSS-based synchronisation suggest residual phase signatures within the band of the Harmony interferograms consistent with a fraction of the oscillator phase errors themselves, and about a factor close to one order of magnitude higher than the phase synchronisation goal after processing and calibration [Rodrigues-Silva et al., TGRS 2025].

It is in this context that the use of a data-based time and phase synchronisation approach appears as a sine qua non for the ground segment of the Harmony mission. In particular, its robust combination with other dynamic and static calibration algorithms (e.g., baseline, antenna patterns, pointing) or processing errors (e.g., topography, atmosphere) is expected to play a relevant role in the design, tuning and validation of the algorithms.

The use of data-based time and phase synchronisation (AutoSync) in the spaceborne environment was for the first time demonstrated by the authors during the commissioning phase of TanDEM-X [Rodriguez-Cassola et al., GRSL 2012], in which phase synchronisation estimations close to the observables of the synchronisation link were achieved and consistent digital elevation models (DEM) derived from the corrected interferograms were generated. We present in this contribution a robust approach for the clock synchronisation in the Harmony mission based on the classical AutoSync formulation of TanDEM-X. The approach is valid for single acquisitions with the Harmonies in close formation. Though heavily relying on the multisquint observables for the estimation of the clock-induced azimuth drifts [Prats et al., TGRS 2003], the approach avoids the typical random-walk residuals associated with multisquint by constraining the solution to the overall deformation of the bistatic images. In addition to the classical AutoSync, we suggest to take advantage of the Harmony instrument layout for complementing the estimation with additional observables allowing for the identification of residual contributions and the allocation of the individual phase synchronisation errors with respect to Sentinel-1, something which is expected to dominate phase synchronisation estimates in large baseline scenarios [Rodriguez-Cassola et al., IGARSS 2023]. This novel element opens the door to the synchronisation of the bistatic images acquired in the stereo phase of the mission. Additional elements will be given in the presentation on the possibilities for the synchronisation of ocean observations in the stereo phase and repeat-pass acquisitions. All results will be illustrated with data generated and processed with the HEEPS-Terra developed in the land science activities [Prats-Iraola et al., Fringe 2026].