Conference Agenda

The Italian Space Agency (ASI) and the Japan Aerospace Exploration Agency (JAXA) have partnered to advance the understanding of the Earth’s surface and atmosphere from space through the use of Synthetic Aperture Radar (SAR) satellites in the fields of Earth sciences and Earth observation applications. This long-standing cooperation started with the Memorandum of Understanding concerning the Feasibility Study and Joint Research Activities for the Anticipated Mutual Cooperation in the Satellite Disaster Monitoring signed on 18 September 2009, and fruitfully continued with the “Implementing Arrangement concerning Mutual Cooperation for Satellite Support to Disaster Risk Management” (IA) that was signed on 25 November 2016 and is still ongoing.

Since ASI and JAXA have both developed and operated SAR missions – COSMO-SkyMed First and Second Generations (CSK/CSG) on one side, and ALOS-2 and ALOS-4 on the other –, both the agencies have recognized the value to share an important experience in the operational use of X-band and L-band SAR data, and intend to increase the benefits of synergies in the use of virtual SAR constellation by combining L and X-band space-borne assets. In addition, both the agencies are jointly collaborating on new and innovative research on disaster management and other climate change study area using space-borne SARs.

The bilateral cooperation includes three main activities:

1. establishment of a COSMO-SkyMed archive over Japan and an ALOS-2 archive over Italy for Disaster Risk Management activities;

2. acquisition of SAR data by COSMO-SkyMed constellation and ALOS-2 in response to emergency requests made by the other Party;

3. joint SAR research activities related to Disaster Risk Management.

The present paper provides an overview of the ASI – JAXA cooperation for disaster risk management. In particular, the focus is on the synergies and results that have been achieved in order to provide mutual institutional support during emergencies, as well as building X and L-band data collections over the respective territories and hotspots of scientific interest to promote innovative research. In this respect, further stimulus to strengthen the cooperation came from the full operations of the COSMO-SkyMed Second Generation satellites and ALOS-4.

Accurate characterization of land subsidence is increasingly important for hazard assessment and infrastructure risk management. Interferometric Synthetic Aperture Radar (InSAR) is widely used to measure surface deformation; however, transforming Line-Of-Sight (LOS) observations into vertical and horizontal components remains a fundamental challenge. Full three-dimensional (3D) decomposition using data from two satellite tracks typically requires additional assumptions regarding the direction of the horizontal deformation component. Although several approaches have been proposed in the literature, these assumptions are not always objectively justified and may introduce bias.

Recently, a gradient-based directional constraint has been introduced, assuming that the direction of horizontal deformation aligns with the horizontal gradient of the vertical displacement field (i.e., the tilt direction). In this study, we focus specifically on subsidence processes and systematically investigate the validity of this assumption in such environments. Importantly, the proposed framework is not intended for general deformation mechanisms, although it may also be applicable to other contexts (e.g., volcanic deformation). Our primary objective is to determine under which physical and spatial conditions the gradient–direction constraint is valid in subsidence areas.

We first provide an analytical examination of the underlying assumptions and then evaluate their validity using synthetic simulations designed to represent realistic subsidence scenarios. These analyses identify the spatial configurations, deformation patterns, and boundary conditions under which the horizontal deformation direction can be reliably inferred from the vertical deformation gradient. The results demonstrate that the assumption is conditionally valid and depends on the mechanical and geometric characteristics of the subsidence field.

Building upon this assessment, we propose an iterative strategy for full 3D decomposition of dual-track InSAR LOS data under the gradient–direction constraint. The method jointly estimates vertical and horizontal deformation components while enforcing consistency between the horizontal motion direction and the vertical displacement gradient. Particular attention is given to computational stability, especially in cases where the inferred horizontal direction approaches the satellite heading direction, leading to ill-conditioning. Strategies to mitigate numerical instability are incorporated into the estimation procedure.

The proposed approach is applied to several subsidence areas in Iran and validated using available Global Positioning System (GPS) observations. The results demonstrate improved consistency in separating vertical and horizontal components and provide quantitative evaluation of reconstruction accuracy. We further discuss potential sources of bias, including model assumptions, noise effects, and geometric limitations.

The Tibetan Plateau is characterized by an extensive periglacial landscape. In such environments, cyclical expansion and contraction of the ground surface have been observed in numerous field surveys and associated with seasonal soil freezing and thawing. On Tibetan slopes, beyond observing these movements perpendicular to the surface, recent InSAR studies have also revealed systematic, short-wavelength millimeter to centimeter annual downslope displacement (Daout et al., 2020; Lemrabet, 2022; Watine et al., 2025). These movements were interpreted as solifluction, a slow but widespread downslope mass-movement process driven by seasonal freeze-thaw cycles and associated with water phase changes (Matsuoka, 2001).

However, previous studies focusing on these downslope movements are limited to regional permafrost areas in the northeastern Tibetan Plateau, leaving the large-scale spatio-temporal variability of solifluction across the Tibetan Plateau largely unknown, particularly in regions dominated by seasonally frozen ground (~53% of the Tibetan Plateau surface; Zou et al., 2017). Furthermore, as climate warming alters ground thermal regimes, solifluction dynamics may evolve, potentially destabilizing slopes and increasing the risk of rapid failures. Yet the extent and timing of these changes remain unconstrained.

Currently, the massive availability of SAR data makes it possible to reconstruct continental-scale ground deformation over periods of up to a decade. This allows InSAR approaches to explore surface displacement across the entire Tibetan Plateau and identify any acceleration in deformation. However, to date, no large-scale study has systematically investigated surface deformation processes across the Tibetan Plateau. Such processing requires particularly careful phase unwrapping strategies, especially on rapidly deforming slopes, in order to prevent signal aliasing and ensure reliable deformation estimates. Once these methodological challenges are overcome, large-scale InSAR observations offer a unique opportunity to map and quantify solifluction processes, and to investigate their sensitivity to climatic, morphometric, and geological controls.

Here, we present the first continental-scale mapping and characterization of systematic slope deformation across the western Tibetan Plateau using the FLATSIM (ForM@Ter LArge-Scale Multi-Temporal Sentinel-1 InterferoMetry) service, operated by ForM@Ter and the CNES (Thollard et al., 2021). Our analysis is based on the New Small Baseline Subset (NSBAS) processing chain (Doin et al., 2011, 2015; Grandin, 2015). We used approximately 13,153 interferograms acquired between October 2014 and May 2022 over 7 ascending and 10 descending Sentinel-1 tracks, covering ~600 000 km² in 8-looks.

Reconstructing reliable InSAR time series over the Tibetan Plateau is challenging due to its periglacial nature, and automatic processing is not always suitable. Rapid frost heave can generate strong phase gradients that can lead to phase unwrapping errors when the unwrapping path crosses highly deforming areas, potentially introducing modulo 2π ambiguities in the unwrapped phase. In addition, rapid soil moisture variations introduce systematic phase contributions, particularly in short temporal baseline interferograms, which bias SAR time series and reduce the reliability of deformation estimates. To ensure reliable InSAR time series, we implement a dedicated processing strategy in order to minimize phase aliasing and biases in InSAR time series over periglacial terrains. Note that the interferograms used are already corrected for tropospheric delays using ERA-5 atmospheric models.

First, a seasonal deformation model is subtracted prior to the critical phase unwrapping step, reducing phase variability and minimizing errors during the unwrapping procedure. The reduced interferograms are then unwrapped following the collinearity of the filtered phase, starting from areas of high temporal coherence across the interferometric network. After unwrapping, the seasonal signal is added back to reconstruct the full unwrapped signal. An initial time series inversion is performed without automatic correction of network misclosures to identify inconsistencies in the network and the associated interferograms. These interferograms are either removed or reprocessed, and the inversion is repeated iteratively until network consistency is achieved. A final inversion is then performed, weighting interferograms by their temporal baseline to reduce moisture-related biases. Pixels with large residual errors, low mean temporal baseline, and high mean interferogram misclosure weighted by the temporal baseline are masked out after inversion to limit the propagation of errors and biases in the following post-processing.

After these steps, time series displacement maps are spatially high-pass filtered using a Gaussian filter to remove residual atmospheric and large-scale tectonic contributions. Line-of-sight time series are temporally decomposed into linear velocity, acceleration, seasonal, and its temporal evolution. To study slope processes and, more particularly, to go in detail in their mechanism, products of the temporal decomposition are spatially inverted into slope-parallel and slope-normal components using multiple viewing geometries. This approach enables the direct correlation of slope-parallel and slope-normal components with morphometric and environmental variables such as air temperature or slope angle, providing a framework for analyzing solifluction processes. Pixels with slopes < 2° or > 30°, slopes oriented within ±30° of the north–south direction, and high RMS after spatial inversion are excluded to ensure robust analysis.

The resulting products consist of large-scale maps of slope-parallel and slope-normal velocities and accelerations, as well as maps of slope-normal seasonal amplitudes and their temporal increase. They reveal widespread active downslope deformation across the western Tibetan Plateau, with velocities on the order of millimeters to centimeters per year. The extensive presence of solifluction landforms, observed in both satellite imagery and on the field, supports the interpretation that these movements correspond mainly to active solifluction processes.

At first order, these large-scale measurements reveal that the spatial distribution of these processes is controlled by mean annual air temperature (MAAT), with more than 90% of areas exhibiting velocities greater than 5 mm/yr located in permafrost regions. Maximum velocities occur around a MAAT of approximately −8 °C and decrease toward warmer conditions. Although solifluction can occur in regions with only seasonally frozen ground, very little movement is observed there. This difference is likely explained by the generally dry conditions of the Tibetan Plateau limiting the water availability for free-thaw processes in seasonally frozen ground. In contrast, areas under permafrost retain sufficient water to sustain solifluction processes.

At second order, solifluction velocity co-varies with slope angle and the amplitude of movements normal to the slope (contraction/expansion cycle). This relationship provides insight into the mechanisms controlling downslope motion. The observed velocities exceed those expected from frost creep alone, one of the main solifluction mechanisms, which results from slope-normal frost heave followed by near-vertical thaw settlement (Washburn, 1979). This suggests that gelifluction, the slow gravitational shear deformation during seasonal thawing, likely dominates the downslope motion. The relative contribution of these mechanisms varies with lithology: frost creep contributes up to 30% of the downslope movements in unconsolidated sediments such as moraines. However, frost creep is more limited compared to gelifluction in bedrock hillslopes covered by colluvium. These variations can be explained by differences in grain size, which affect frost susceptibility, and by variations in bedrock depth beneath the colluvium.

Finally, analysis of the slope-normal component reveals increasing seasonal amplitudes and widespread irreversible ground subsidence within permafrost areas, indicating ongoing degradation. This degradation evolves through time but with a high spatial heterogeneity. Some regions show stabilization, with subsidence rates decreasing over time, locally by up to 80% between 2014 and 2022, whereas others exhibit acceleration. These trends directly influence solifluction dynamics, leading respectively to reduced or enhanced downslope velocities.

These results demonstrate that the spatial distribution of solifluction is primarily controlled by the permafrost distribution on the western Tibetan Plateau, and that its ongoing degradation directly modulates solifluction dynamics.

Daout, et al. "Ice loss in the Northeastern Tibetan Plateau permafrost as seen by 16 yr of ESA SAR missions." Earth and Planetary Science Letters 545 (2020): 116404.

Doin, et al. "Presentation of the small baselin NSBAS processing chain on a case example: The Etan deformation monitoring from 2003 to 2010 using Envisat data." Fringe symposium. 2011.

Doin, et al. "InSAR measurement of the deformation around Siling Co Lake: Inferences on the lower crust viscosity in central Tibet." Journal of Geophysical Research: Solid Earth 120.7 (2015): 5290-5310.

Grandin. "Interferometric processing of SLC Sentinel-1 TOPS data." FRINGE’15: Advances in the Science and Applications of SAR Interferometry and Sentinel-1 InSAR Workshop, Frascati, Italy, 23-27 March 2015. 2015.

Lemrabet, et al. "Referencing of continental‐scale InSAR‐derived velocity fields: Case study of the eastern Tibetan Plateau." Journal of Geophysical Research: Solid Earth 128.7 (2023): e2022JB026251.
Matsuoka. "Solifluction rates, processes and landforms: a global review." Earth-Science Reviews 55.1-2 (2001): 107-134.

Thollard, et al. "Flatsim: The form@ter large-scale multi-temporal sentinel-1 interferometry service." Remote Sensing 13.18 (2021): 3734.

Watine, et al. "Downslope solifluction movements and permafrost degradation in the northeastern Qinghai-Tibetan Plateau revealed by InSAR." Remote Sensing of Environment 329 (2025): 114926.

Zou, et al. "A new map of permafrost distribution on the Tibetan Plateau." The Cryosphere 11.6 (2017): 2527-2542.

These days, InSAR is a key technology for monitoring land subsidence and assessing its associated hazards and risks. While InSAR displacement maps provide essential information, effective subsidence risk evaluation requires accurate estimation of differential subsidence, e.g., spatial gradients and curvature, which are more directly related to infrastructure damage and ground instability than absolute deformation.

All methodologies for estimating spatial gradients from InSAR data—whether implicitly or explicitly—rely on some form of local spatial modelling/interpolation of the deformation field. Conventional approaches typically assume a spatially linear deformation model within a fixed moving window, which implies constant strain over that area. The accuracy and reliability of this strategy depend on two key factors: the choice of the basis function and the selected window size. For robust gradient estimation, the true deformation field must be adequately represented by the adopted model within each local window.

However, many subsidence systems are inherently multi-scale. Regional deformation driven by groundwater withdrawal may extend over broad areas, yet its spatial extent can vary significantly depending on hydrogeological conditions and extraction patterns. At the same time, localized subsidence may develop at much smaller scales, such as individual wells, buildings, infrastructure loads. These processes often coexist and overlap, resulting in heterogeneous displacement fields characterized by spatially varying gradients and curvature. In such environments, a fixed window size combined with a linear deformation model may either oversmooth localized features or become overly sensitive to noise. This mismatch between deformation complexity and model assumptions can lead to systematic underestimation of differential subsidence, thereby affecting hazard interpretation and risk zoning.

To address these limitations, we propose an adaptive multi-scale framework for InSAR-based differential subsidence estimation. The proposed approach integrates two key components. First, instead of relying solely on a linear model, a second-order polynomial basis function is used within each local window, increasing flexibility to represent curvature and spatial variability. Second, an iterative adaptive window selection strategy is implemented. The procedure begins with an initial window size; model adequacy is assessed using a statistical consistency criterion. If the deformation is not sufficiently represented, the window size is reduced, and the process is repeated until the model assumptions are satisfied. This ensures that the spatial scale of analysis is consistent with local deformation characteristics.

The method is first evaluated using controlled synthetic deformation scenarios, allowing quantitative testing under known ground-truth conditions and varying noise levels. The framework is subsequently applied to several subsidence-affected regions in Iran using Sentinel-1 data, covering different deformation mechanisms and spatial scales. These include areas dominated by groundwater-related subsidence with both regional and localized components in urban areas. The results demonstrate improved representation of spatial gradients compared to conventional fixed-window linear approaches, particularly in areas with strong deformation variability.

Importantly, the method provides an additional scale-diagnostic capability. The final selected window size represents the spatial extent over which the chosen basis function remains valid. The spatial distribution of optimized window sizes therefore provides insight into the intrinsic scale of deformation processes, helping to distinguish between large-extent regional subsidence and localized deformation sources. This byproduct enhances geophysical interpretation and supports more robust subsidence hazard assessment.

In conclusion, the proposed adaptive multi-scale methodology improves differential subsidence estimation from InSAR data, reduces bias in heterogeneous deformation fields, and introduces a novel indicator of deformation scale. The approach strengthens the link between satellite-derived deformation measurements and hazard-oriented subsidence analysis.

The Cascadia Subduction Zone, which stretches from northern California to southern British Columbia, has the potential to produce large hazards like megathrust earthquakes and tsunamis. This area is therefore heavily monitored by the US Geological Survey (USGS) with onshore and offshore instruments including seismometers, Global Navigation Satellite Systems, strainmeters, tiltmeters, and more recently Distributed Acoustic Sensing cables. Interferometric Synthetic Aperture Radar (InSAR) techniques provide a complementary dataset with dense spatial coverage over the entire Cascadia region.

InSAR data have been critical for understanding and monitoring hazards. For example there has been success in studying slow slip events in Cascadia using Sentinel-1 C-band data. These data, however, decorrelate on longer time scales due to vegetation growth throughout the Pacific Northwest. As a result, long term tectonic monitoring and constraining small signals related to different parts of the earthquake cycle is challenging. NASA-ISRO’s NISAR mission uses an L-band radar, which has a longer wavelength that better penetrates through densely vegetated regions and maintains coherence on longer temporal scales. Therefore, more interferograms can be formed with longer temporal baselines, which are necessary for deriving accurate long term time series. Furthermore, these data, in combination with other and future satellites, e.g., ALOS-2 and ROSE-L, provide for more imaging geometries enable better constraints on horizontal and vertical displacements, which can ultimately improve fault slip solutions in the region.

We perform a coherence analysis and comparison with the most recent ALOS-2 ScanSAR and Sentinel-1 data. Preliminary results using ALOS-2 ScanSAR data (path 170, frame 2800) from March 17 – 31, 2025 show that the vegetated area along the coast maintains coherence, ranging from 0.2 – 0.5, and the wrapped and unwrapped phase is spatially coherent. For the 2-month interferogram from January 20 – March 17, 2025 there is a drop in coherence along the coast but that the phase is still coherent in the wrapped interferogram. Despite lower coherence in the 2-month interferogram, the phase measurements at L-band are still usable for time series analysis.

We download and process all available NISAR data in the southern Cascadia region and compute phase statistics to determine the impact of residual noise in interferograms on future derived time series. We demonstrate that NISAR data enables long-term monitoring of vegetated regions and that L-band InSAR data provide critical surface deformation measurements that help us better understand the earthquake cycle and our preparedness for future hazards in Cascadia.