Conference Agenda

The present study was aimed at comparing vertical and horizontal surface displacements derived from the Cosmo-SkyMED, TerraSAR-X and Sentinel-1 satellite missions for the detection of oil extraction-induced subsidence in the Tengiz oilfield during 2018–2021. The vertical and horizontal surface displacements were derived using the 2D decomposition of line-of-sight measurements from three satellite missions. Since the TerraSAR-X mission was only available from an ascending track, it was successfully decomposed by combining it with the Cosmo-SkyMED descending track.

Vertical displacement velocities derived from 2D Decomposition showed a good agreement in similar ground motion patterns and an average regression coefficient of 0.98. The maximum average vertical subsidence obtained from the three satellite missions was observed to be −57 mm/year. Higher variations and deviations were observed for horizontal displacement velocities in terms of similar ground motion patterns and an average regression coefficient of 0.80. Fifteen wells and three facilities were observed to be located within the subsidence range between −55.6 mm/year and −42 mm/year.

The spatial analyses in the present studies allowed us to suspect that the subsidence processes occurring in the Tengiz oilfield are controlled not solely by oil production activities since it was clearly observed from the detected horizontal movements. The natural tectonic factors related to two seismic faults crossing the oilfield, and terrain characteristics forming water flow towards the detected subsidence hotspot, should also be considered as ground deformation accelerating factors.

The novelty of the present research for Kazakhstan’s Tengiz oilfield is based on the cross-validation of vertical and horizontal surface displacement measurements derived from three radar satellite missions, 2D Decomposition of Cosmo-SkyMED descending and TerraSAR-X ascending line-of-sight measurements and spatial analysis of man-made and natural factors triggering subsidence processes.

The strongest instrumentally recorded earthquake in the region of the Komandorsky Islands occurred on July 17, 2017 at 23:34 GMT. This seismic event had the magnitude Mw = 7.8 and its epicenter was located southeast of Medny Island, 200 km from village Nikolskoe (Bering Island) and had the coordinates 54.44° N, 168.86° E. This earthquake is particularly interesting because of the three following reasons.

(1) The earthquake occurred in the vicinity of the Kamchatka-Aleutian triple junction. In the large-scale tectonic plate models this is the meeting point of the Pacific Plate, the Okhotsk Plate, and the North American Plate. More recently, the seafloor part north of the Aleutian arc has been identifyid as the Beringia plate based on the geological and seismological data. It presumably spans the entire area of the Bering Sea and some coastal regions. In the eastern part of the Aleutian arc, the Pacific plate is subducting at a rate of 66 mm/yr almost perpendicular to the strike of the island arc. Further westward, the ratio of the shear component gradually increases, and in the western part of the arc the Pacific plate moves parallel to the arc at a rate of 75 mm/yr. The study of the ruptures of the earthquakes at the periphery of the Beringia plate, including the methods of SAR interferometry, is important for testing the hypothesis of the existence of this microplate because it still remains the subject of debate.

(2) Similar other strongly oblique parts of subduction zones, in the western termination of the Aleutian only a small portion of the relative displacement of the lithospheric plates arc is accommodated by their contact. Most displacements occur along the back-arc shear zone called the Bering fault. Studying the displacement distribution along the fault system in this region is important, inter alia, for forecasting seismic activity.

(3) The 2017 Komandorsky Islands earthquake occurred in a seismic gap - a region where no strong seismic events occurred for a long time despite the high velocities of the relative plate motion.

To date, several models of the Komandorsky Islands earthqauke rupture have been published. These models are based on waveform inversion, on seismological data, on GPS and tide gauge data (Lay et al., 2017), and on seismological and GPS data (Chebrov et al., 2019). The difficulty in building a rupture model in the case of this earthquake is that most data used to construct the source model are from remote stations. In particular, in the vicinity of the earthquake there are only two GPS stations where horizontal displacements are above the noise level and can be used to constrain the source model (Lay et al., 2017).

We present here a new model of the Komandorsky Islands earthquake rupture based on satellite geodesy and InSAR data. For the first time, we managed to construct the displacement fields on the Bering and Medny Islands located in the epicentral zone of the earthquake using the Sentinel-1B images. Given the insufficient density of the GNSS network in the study region, the displacement fields estimated from InSAR data provide new information about the structure of the earthquake source.

Among the interferogram pairs calculated from the images covering the period from June 17 to August 28, 2017, the most reliable displacement fields were obtained from the image pair July 11–July 23, 2017. These displacements include coseismic and part of postseismic displacements. The inversion also involved the displacement data recorded by the GNSS GPS stations on the Kamchatka Peninsula, Komandorsky Islands, and the closest to the epicenter Aleutian Islands. Due to the fact that displacements substantially exceeding the noise level were only recorded at two GPS stations on the Bering and Shemya islands, the use of the InSAR data substantially refines the existing earthquake source models.

In our models, the seismic rupture zone is approximated by a plane with a length of 370 km along the strike and the width of 19 km along the dip, respectively. Three models have been tested: (1) a model of uniform displacement across the entire rupture surface; (2) a model in which the rupture surface is divided strikewise into five elements; and (3) a model divided into four elements along the strike and into two levels along the dip, with a total of eight elements. All models demonstrate the same displacement type: right-lateral strike-slip faulting with a relatively small thrust component. According to the constructed models, the displacements in some areas of the rupture surface are slightly smaller than average but, generally, they occur all over the source zone. The models based on satellite geodetic data and on waveform inversion largely agree. The discrepancy between the models based on different data types can probably be due to the fact that seismological data characterize the part of the source process that is accompanied by the generation of seismic waves. Surface displacements estimated from InSAR data do not characterize only the mainshock but also contain contributions that may reflect various creep processes. The period covered by the radar images includes the foreshocks with magnitudes up to 6.3 as well as more than 100 aftershocks with magnitudes between 4 to 5.5. Perhaps that is why the displacements obtained in our models are more uniformly distributed over the 370-km rupture surface than in the models based on the waveform analysis.

The study was carried out in partial fulfillment of the State Contract of Schmidt Institute of Physics of the Earth of the Russian Academy of Sciences and Interdisciplinary Scientific and Educational School “Fundamental and Applied Space Research” of the Lomonosov Moscow State University.

REFERENCES

Chebrov, D.V., Kugaenko, Yu.A., Lander, A.V., Abubakirov, I.R., Gusev, A.A., Droznina, S.Ya., Mityushkina, S.V., Ototyuk, D.A., Pavlov, V.M., and Titkov, N.N., Near Islands Aleutian earthquake with MW = 7.8 on July 17, 2017: I. Extended rupture along the commander block of the Aleutian island arc from observations in Kamchatka, Izv., Phys. Solid Earth, 2019, vol. 55, no. 4, pp. 576–599.

Lay, T., Ye, L., Bai, Y., Cheung, K.F., Kanamori, H., Freymueller, J., Steblov, G.M., and Kogan, M.G., Rupture along 400 km of the Bering Fracture Zone in the Komandorsky Islands earthquake (Mw 7.8) of 17 July 2017, Geophys. Res. Lett., 2017, vol. 44, no. 24, pp. 12161–12169.

A disastrous earthquake of magnitude 7.8 struck southern and central Turkey and northern and western Syria followed by a 7.7 magnitude earthquake on February 6^th, 2023, that caused tens of thousands of fatalities and widespread damage to buildings and infrastructure. The earthquake is considered to be one of the deadliest seismic events worldwide in the 21^st century, and various countries and humanitarian organisations provide support for earthquake victims, including humanitarian aid. A rapid damage assessment of the buildings and infrastructure provides valuable information to humanitarian organisations. So far, mainly optical Earth observation (EO) data has been used for building and infrastructure damage assessments. However, they are limited to daytime acquisitions and the availability of cloud-free scenes is not guaranteed. Using synthetic aperture radar (SAR) data can provide an alternative to avoid these constraints (Aimaiti et al., 2022). However, the utilisation of SAR data is mainly focused on deformation analysis using differential interferometry SAR (DInSAR) techniques. Whereas using SAR backscatter and coherency for building damage assessments after the earthquake could provide valuable information (Plank, 2014). Therefore, we aim to use Sentinel-1 SAR (C-band) data to (1) explore intensity and coherency information for assessing damages and building destructions after the Turkey-Syria earthquake; and (2) to assess the reliability of the damage assessments to support rapid humanitarian actions.

We selected the city of Jindires in the Afrin district (Aleppo governorate), located close to the border of Turkey and Syria, which suffered major building and infrastructure damages. The damages reach from minor changes on walls and roofs to fully collapsed buildings. Affected buildings are spread across the entire city, with some damage clusters in the inner city and individual buildings affected in the outer parts of the city.

We used the following two pre-event scenes (i.e., before the earthquake), 2023/01/16, & 2023/01/28 and one post-event scene (i.e., after the earthquake) from 2023/02/09, all in IW mode, in GRD and SLC format from the ascending orbit, and path 14. As reference data, we took footprints from a building damage assessment, which were digitized manually based on expert knowledge using very-high-resolution Pleiades imagery captured on 2023/02/10 and that found 142 structures damaged, 161 destroyed, and 18 possibly damaged. Additionally, we included the damage assessment provided by United Nations Satellite Center (UNOSAT, 2023), which reports 233 damaged and 323 possibly damaged buildings. The GRD data was pre-processed by applying orbit files, calibration, thermal noise, and terrain corrections. SLC data was pre-processed, including TOPSAR split, applying orbit file, back-geocoding layer stacking, coherency formation, debursting, and terrain corrections; and two coherency images were created, using 2023/01/16 & 2023/01/28 (pre-event Sentinel-1 data) and using 2023/01/28 & 2023/02/09 (pre- and post-event Sentinel-1 data). Additionally, we created ratios between post- and pre-event VV and VH polarisations, and between the generated coherency layers. Moreover, we calculated different texture measures to leverage spatial texture information using the grey-level co-occurrence (GLCM) matrix, which has been used in literature to distinguish between collapsed and intake buildings (Akhmadiya et al., 2021). As the focus was on buildings within the city, we masked out non-building areas. The damage detection was done based on the expectation of a decrease in backscatter due to the structural change of damaged buildings compared to intact buildings. However, when a building is partially damaged, remaining walls, debris and grounds may cause corner reflections, resulting in strong double-bounce effects and increased backscatter intensity (Aimaiti et al., 2022). We also expect a drop in coherency measures due to out-of-phase signals caused by the damage or destruction of buildings. Therefore, we considered both decrease and increase in values derived from backscatter, coherency and texture information within the building footprints to derive the final damaged buildings. The results derived from Sentinel-1 include destroyed and damaged buildings, while false positives were identified with the help of the reference data. Although the results reveal the potential of Sentinel-1 data for building damage assessments after the earthquake, further studies should investigate errors related to false positives, and building damage categorisations, e.g., total or partial damage.

We presented a simple, nevertheless robust workflow to derive and combine different information layers derived from Sentinel-1 data, which can provide valuable information in rapid building damage assessments and support humanitarian actions. Nevertheless, a level of uncertainty needs to be acknowledged and, if possible, accounted for; e.g, related to the temporal baseline between analysed and reference data. The latter was fairly low in this study as both the Sentinel-1 post-event data and Pleiades image used for creating the reference layer were only a single day apart. Another uncertainty lies in the capabilities and expertise of the interpreter, which influences the quality of reference layers. Although the detection of major damages such as fully collapsed buildings might be unambiguous, detecting partially damaged buildings is challenging and contributes to uncertainty in the reference layer. Further studies shall focus on utilising SAR EO-based damage assessments in an automated workflow and improve on the mentioned uncertainties.

Aimaiti, Y., Sanon, C., Koch, M., Baise, L. G., & Moaveni, B. (2022). War Related Building Damage Assessment in Kyiv, Ukraine, Using Sentinel-1 Radar and Sentinel-2 Optical Images. Remote Sensing, 14(24), 6239.

Akhmadiya, A., Nabiyev, N., Moldamurat, K., Dyussekeyev, K., & Atanov, S. (2021). Use of Sentinel-1 Dual Polarization Multi-Temporal Data with Gray Level Co-Occurrence Matrix Textural Parameters for Building Damage Assessment. Pattern Recognition and Image Analysis, 31(2), 240–250. https://doi.org/10.1134/S1054661821020036

Plank, S. (2014). Rapid damage assessment by means of multi-temporal SAR—A comprehensive review and outlook to Sentinel-1. Remote Sensing, 6(6), 4870–4906.

Whilst subduction earthquakes are sudden dislocations at the plate interface involving seismic slip, there are also transient phenomena characterized by slow motion under aseismic slip, which are known as Slow Slip Events SSEs (Draguert et al., 2001, Schwartz & Rokosky 2007).Globally, SSEs are found to occur predominantly in the deeper part of the seismogenic zone, where the transition from unstable to stable sliding takes place (Kano et al., 2018). Recently, a SSE located at the deeper part of the megathrust starting in the middle of 2014 was detected north of Chile, close to Tal-tal area (Klein et al., 2021, Pastén-Araya et al., 2022) (FIG.1). GPS observations suggest that a similar aseismic process has occurred in 2005 and 2009, implying a recurrence time of approximately ~5 years, which has been confirmed recently with the detection of a new SSE on the region (Klein et al., 2021,2023). Importantly, the Tal-Tal area has been shown to be a mature seismic gap involving high seismic risk (Metois et al., 2016). Hence constraining the time-space evolution of SSEs in this region, and to explore how they might be influencing the stress build-up on locked asperities becomes crucial. In this context, Interferometric Synthetic Aperture Radar (InSAR) could significantly improve the characterization of the 2014-2020-SSEs and of SSEs that will follow them. Because of its regional-scale observation, and its regular repeat time, InSAR is an incredible tool to get spatially dense time series of Earth surface deformation from the 90s (using ERS and Envisat archives) until now (using Sentinel data) (Bürgmann, 2000, Jolivet et al., 2012).

In this work, we will investigate the feasibility of InSAR time series to detect transient slip at the plate interface. Notably, the reported deformation pattern of the 2014-2020-SSEs is characterized by displacements on the plate interface around ~200 mm, whose magnitude has been shown to be possibly detected by InSAR measurements (Rouet-Leduc et al., 2021) (FIG.1). The raw data processing of InSAR has been performed by the FLATSIM in the framework of ForM@Ter Large-scale multi-Temporal-Sentinel-1-InterferoMetry project (Thollard et al., 2021). The transient slip can be associated with a deformation amplitude of ~5 cm on the surface, which is much lower than atmospheric noise (FIG.1). The later is mostly expected to dominate at large-scale wavelengths, therefore masking the SSE signature. Here we will explore different signal analysis tools to decompose the InSAR time series on the multiple sources to isolate that part only related to SSE deformation. To do so, blind separations methods as Principal Component Analysis (PCA), Independent Component Analysis (ICA), will be performed, enhanced by low-pass filtering and tectonic corrections. Additionally, available GNSS time series will be used to define the amplitude and timing of the SSEs expected to be found on the InSAR data (FIG.1). Notably, the timing of SSE defined by GNSS can then be used for a parametric decomposition or for a joint GNSS -InSAR decomposition. Thereby, we will show whether the InSAR data can be applied to the detection of transient slip on the Chilean subduction margin to then characterize the temporal and spatial evolution of the fault behavior on the area. Further, our results may offer and opportunity to highlight how SSEs and large earthquakes might be interacting, and therefore giving insights on seismogenesis physics.

The western side of the Balkans is one of the most tectonically active area in Europe even if
many unknowns remain to properly estimate the seismic hazard there. It has recently experienced two shallow Mw 6.4 crustal destructive earthquakes : the 2019, Dürres thrust fault earthquake in the external Albanides, and the 2020 Petrinja transpressive event that stroke Croatia on the eastern flank of the Dinarides. In order to quantify and explore the current day strain accumulation and release modes in th western Balkans and the coseismic displacement associated with these two moderate earthquakes, we analyse InSAR time-series provided by the FLATSIM service developed by the Data and Services center for solid earth ForM@Ter and operated by CNES, based on Sentinel-1 data acquired from 2014 to 2021.5 (Thollard et al. 2021). The spatial resolution is 240 m (16 looks processing).
The Dürres area is covered by 3 tracks (2 ascending, 1 descending) that we analyze to assess the interseismic loading, coseismic jump and potential postseismic motion associated with the 2019 earthquake. For each track, we jointly invert the FLATSIM time series for the linear trend, coseismic jump and annual seasonal signal for each pixel independently using a least-square optimized trajectory model. We combine different quality criteria (misclosure of the interferometric network, number of unwrapped interferograms per pixel) to mask areas that are poorly constrained and provide conservative estimates of the coseismic jump. Both the residuals of the inversion and the analysis of the postseismic time series do not show any clear postseismic signal neither in space or time, while some significant seasonal signal is observed in the Dürres and Tirana sedimentary basins. In particular, we check whether the LOS time series are in agreement with claimed GNSS-detected SSE that may have occurred postseismically.

In order to better understand which fault is involved, we conduct a joint inversion of the coseismic slip using the coseismic maps obtained from the inversion of the InSAR time series on the three independent tracks, the coseismic jumps estimated from high-rate GNSS stations, and teleseismic observations.
At a broader regional scale, we aim at comparing the InSAR-derived interseismic strain field (built assuming a purely horizontal motion) with the GNSS derived strain rate.

On February 6th, 2023, a 7.8 magnitude earthquake hit the southern and central regions of Turkey, as well as the northern and western regions of Syria. This earthquake was one of the largest earthquakes ever recorded causing extensive damage to the buildings and infrastructures in the affected regions and more than 50000 casualties. The disaster management authorities have been struggling to assess the damages and prioritize rescue and relief operations due to the widespread nature of the damages.

In this context, remote sensing can provide valuable insights into the extent and severity of the damages. In particular, the joint use of high-resolution Synthetic Aperture Radar (SAR), and Multi-spectral optical sensors can provide complementary information and improve the accuracy and reliability of Earth Observation applications for damage mapping purposes.

This study presents the application of multi-sensor and multi-frequency change detection methods for detecting damage in the aftermath of the Turkey earthquake, in a semi-automatic procedure, for pre-operational use. Kahramanmaras city has been chosen as the test site since it was one of the most damaged by the earthquake.

We performed a quantitative analysis of earthquake-induced damage by using a short time series of SAR and optical imagery collected before and after the seismic event using X-band COSMO-SkyMed 2nd generation and Planetscope sensors, respectively.

The SAR change detection approach is based on the Intensity Correlation Difference (ICD) which estimates the changes in the spatial distribution of the scatters, and their SAR intensity value, within a user-defined window.

Planetscope constellation, consisting of approximately 130 small Dove satellites, provides daily coverage of the entire land surface of the Earth at 3m spatial resolution in 4 spectral bands and more recently, with the new superDove satellites, in 8 spectral bands from coastal blue to near-infrared. We here employ the spectral signature difference on a pixel-per-pixel basis in the 8 bands to evaluate the damaged areas by analyzing several acquisitions before and after the seismic sequence.

We will test supervised and unsupervised data fusion methods, based on Machine Learning approaches (e.g. Neural Networks), to merge the information coming from SAR and Optical data, aiming at improving the reliability and accuracy of damage assessment.

Our results will be compared and validated with the products provided by the Copernicus Emergency Mapping Service.

The final goal of this study relies on demonstrating the effectiveness of the joint use of SAR and optical change detection methods in detecting damage to buildings and infrastructure that can be used for disaster management authorities to prioritize rescue and relief operations in the affected regions.

Postseismic deformation occurs due to stress relaxation following earthquakes and has been widely captured by space geodetic observations. The main mechanisms proposed to explain the postseismic deformation include afterslip, viscoelastic relaxation, and poroelastic rebound. Coseismic stress changes have been shown to drive afterslip on fault interface surrounding coseismic asperities. Viscoelastic behavior in the lower crust and upper mantle can lead to more widespread deformation. Poroelastic rebound caused by fluid migration could explain some of the early postseismic deformation. Understanding the contributions from these mechanisms provides important information about the frictional, rheology, and porous structures of the seismogenic fault and surrounding crust.

The 2021 Mw 7.4 Maduo earthquake ruptured ~150 km of the Jiangcuo fault, a previously-poorly known NWW-trending, sinistral strike-slip fault which lies within the Bayan Har block of the eastern Tibetan Plateau. This earthquake provides valuable opportunity to study the mechanisms responsible for postseismic deformation of the intrablock earthquakes. Here we use ~2-years of Sentinel-1 interferometric synthetic aperture radar (InSAR) data to study the postseismic deformation following the Maduo earthquake. We first produce descending and ascending interferograms using the “Looking into Continents from Space with Synthetic Aperture Radar” (LiCSAR) system. We then perform the small baseline subset (SBAS) InSAR analysis using an open-source time series analysis package LiCSBAS. The atmospheric noise is modeled by the Generic Atmospheric Correction Online Service. We identify the unwrapping errors using the baseline loop closure and residuals of the SBAS inversion, and correct them by integers of 2pi. Long-wavelength noise including the ionospheric phases, orbital inaccuracies and tectonic plate motion were reducted by fitting a linear ramp for each interferogram. For other short-wavelength signals such as unmodeled atmospheric delays and topography-related noises, we adopt independent component analysis to separate these signals and to obtain the postseismic signals. Both our descending and ascending data reveal notable localized deformation in the middle segment of the seismogenic fault suggesting shallow afterslip, and diffused deformation in the far field implying either deep afterslip or viscous flow, or their coupled contributions.

In our study, we will compare kinematically-inverted afterslip versus stress-driven afterslip to infer the potential contribution to postseismic surface deformation from other mechanisms such as viscoelastic relaxation. We will model the viscoelastic contribution using Maxwell, Burgers and power-law rheologies, and compare the best-fit results with a mechanically-coupled model that combines afterslip and viscoelastic relaxation. We will discuss the constraints on depth-dependent rate-strengthening frictional parameters and lateral variation of viscosity beneath the fault provided by this event, and discuss the implications of the results for the assessment of future seismic hazard and the understanding of the crustal rheology structure.

InSAR is an increasingly important tool for the assessment of earthquakes in the continental crust, which is crucial to understanding continental deformation process and the associated seismic hazard. The South American plate experiences deformation induced by stress transfer in response to the subduction of the Nazca slab beneath it, and the interaction of Cocos-Caribbean plates in the north. As a result, shallow and complex networks of active faults are found near some heavily populated areas. Seismic risk analysis indicates that shallow earthquakes with moderate magnitudes (M_w 6.0-7.5) occurring near major cities can lead to significantly greater damage and fatalities when compared to large but distant interplate events with magnitudes of M_w 8.0 or higher. In this study, we use Sentinel-1 InSAR to build a catalogue of moderate magnitude earthquakes in the South American plate. We then investigate data-driven approaches to improve our ability to resolve the source parameters of moderate magnitude earthquakes, and compare our results to those from seismic methods.

To select all potential candidate earthquakes, we choose earthquakes in South America from the Global Centroid Moment Tensor (GCMT) and United States Geological Survey (USGS) catalogues. We filter the earthquakes to find those with i) Sentinel-1 coverage (between 2016 and 2023), ii) magnitude range 5.0-7.0, iii) absolute focal depth < 20 km, and iv) relative depth to slab > 15 km. We have identified 31 earthquakes that fit these criteria, including the 2019 M_w 6.0 Mesetas (Colombia) earthquake, the 2020 M_w 5.8 Humahuaca (Argentina) earthquake, and the 2021 M_w 5.7 Lethem (Guyana) earthquake, which are respectively strike-slip, normal, and reverse faulting. We then process Interferometric Synthetic Aperture Radar (InSAR) from Sentinel-1 TOPS images for each event.

Moderate magnitude earthquakes produce small and localized surface displacements which can be obscured or distorted by various noise sources, making it challenging to determine the fault parameters and slip distribution accurately. In particular, spatially correlated noise from the turbulent atmosphere results in a low signal-to-noise ratio (SNR). Therefore, we explore two statistical methods to enhance the SNR - stacking and time-series - and successfully reconstruct the earthquake signal displacements for each case. We also test external corrections based on weather model data from Generic Atmospheric Correction Online Service (GACOS) to reduce the atmospheric signals.

We assess the impact of each of these methods on our ability to model the earthquake source parameters, by performing a non-linear inversion for the fault geometry using the Geodetic Bayesian Inversion Software (GBIS). For the studied earthquakes we evaluate the robustness and consistency of each approach in comparison to using individual interferograms. After accounting for InSAR uncertainties, we compare the source parameters derived from InSAR with those from the global seismic catalogues (USGS and GCMT). The InSAR-derived solutions - location, focal mechanism, magnitude and depth - demonstrate the reliability of this strategy for constraining moderate shallow earthquakes.

This study provides a new framework for analysing InSAR deformation signals associated with moderate magnitude intraplate earthquakes. Furthermore, it provides new insights into the seismic cycle of crustal faults within the South American plate. The InSAR methodology applied could be extended to other regions of the world with similar geological and tectonic settings, where shallow crustal earthquakes are frequent and pose a threat to human life and infrastructure. 

In magma-rich rift settings, most medium-to-large magnitude, normal slip earthquakes are induced by dikes, while purely tectonic normal faulting is less common. For example, in the magma-rich rifts of Ethiopia (Afar and the Main Ethiopian rift (MER)) all the geodetically measured examples of normal faulting (i.e., since the onset of InSAR measurements in the area in 1994) have been induced by dike intrusion. An earthquake sequence starting with a Mw 5.5 earthquake occurred between 26-28 December 2022 in northern Afar (Bada region), with several earthquakes recorded globally. Here we use InSAR measurements of the seismic sequence to show that the deformation was caused by purely tectonic normal faulting without involvement of magma. We processed pre- and co-seismic interferograms from ascending (track 014) and descending (track 079) acquisitions made by the European Space Agency (ESA) satellite Sentinel-1a, using the InSAR Scientific Computing Environment (ISCE) software package. We co-registered the SLCs and removed the topographic phase using a 1 arc-sec (∼30 m resolution) DEM and unwrapped the interferograms using the ICU branch cut algorithm and geocoded them using the 1 arc-sec DEM. Satellite acquisitions made at different times during the seismic sequence allow us to discriminate which fault segments moved during the initial and the later part of the sequence. To explain the observed deformation patterns, we inverted the interferograms for the best-fit fault parameters (Okada shear dislocation), assuming an elastic half space with a Poisson’s ratio of 0.25 and a shear modulus of 30 GPa. Our best-fit InSAR models show that different fault segments of a conjugate system forming a graben ruptured during the seismic sequence with mainly normal dip-slip, corresponding to a single Mw 5.7 event, and in agreement with the seismic moment release from global and local seismic recordings. Our models show that purely tectonic faulting accommodates 26 cm of extension corresponding to ~30 years of plate spreading without any link to magma. This mode of deformation differs from past geodetically observed occurrences of normal slip earthquakes in Afar which have to date been mainly dike-induced, and therefore directly shows that extensional faults in magma-rich extensional settings can potentially slip without being modulated by magmatic processes. The occurrence of both magma-assisted and purely tectonic fault growth in a single rift can be explained by spatial and/or temporal variations in magma-supply.

Two strong earthquakes occurred in eastern Turkey on 6 February 2023 within nine hours. The strong doublet took place on the south section of the East Anatolia Fault Zone (EAFZ) and a nearby fault 20 km away to the west. The doublet killed more than 40,000 people in Turkey and Syria. The first mainshock occurred on the southern section of the East Anatolian fault, but it actually initiated on a short branch at the east of the mainshock fault and then propagated to the main fault. The mainshock produced a surface rupture of ~360 km. The second major earthquake occurred on the Sürgü fault which is located on the northwest side of the mainshock fault and also effectively ruptured the surface as long as ~153 km. The doublet shows no physical connections between the two ruptures with the second major event delayed ~9 hours, and no ample aftershocks occurred between the two ruptures. We use the Synthetic Aperture Radar (SAR) images collected by JAXA’s ALOS-2 and ESA’s Sentinel-1 satellites to extract the surface deformation of the doublet. Due to heavy ground shaking and damages that lead to heavy unwrapping issues, we use only amplitude offset data for coseismic deformation mapping. Both the range and azimuth offset measurements on two separated faults show a high signal-to-noise ratio, but the azimuth offset results of the ALOS-2 data suffer from a strong ionosphere disturbance and also inaccuracy of the preliminary orbit information. Hence, we do not use the ALOS-2 azimuth data in the following slip inversion work.

With the SAR deformation data and the coseismic GPS results from the Nevada Geodetic Lab., we adopt a vertical fault geometry with the fault-top fixed on the surface according to the offset data and invert for the slip-distribution on the fault planes using the Steepest Descent Method (SDM) (Wang R., 2011). Both the homogenous (Okada) and layered medium models are adopted in the inversion, with the realistic velocity model and crustal thickness from the receiver function inversion (Tezel et al., 2013). A large fault width of 35 km (local Moho depth) is adopted in the inversion so that we can infer the maximum possible rupture depth with the layered crustal structure adopted.

Both the Okada and layered inversion models indicate strong shallow ruptures with a maximum slip of ~8.5 and ~9.0 meters respectively, the larger slip in the layered model may be due to weaker materials relative to a homogenous model. In addition, the second major event shows more continuous and concentrated slip reaching the maximum slip at its middle section, and only localized slip reaching the maximum slip on the mainshock fault rupture. The most prominent differences between the two kinds of models are at their rupture bottoms. In the Okada model, the slips terminated at the 25 km depth on the mainshock fault, and at the 35 km depth on the second rupture fault. The feature is consistent with the aftershock distribution on the two faults. But in the layered model, we see some clear slip of ~1.5 m at the 35-km depth on the middle section of the mainshock fault. The slips are more broadly distributed in the layered model in contrast to the Okada model, though we adopt the same level of smoothing constraint in both the Okada and layered models. We confirm the existence of the deeper slips in the layered model because their rake angles are consistent with the upper-crust slips though we allowed a +/-50-degree rake variation in the inversion. The deeper slips with realistic velocity models could excite different postseismic relaxations and help objectively resolve the rheology properties of the lower crust and upper mantle.

Besides the slip-distribution model inversion, we also calculated the static coulomb stress changes, poroelastic stress changes, and viscoelastic stress changes between the two fault ruptures, so that we can quantitatively assess the triggering effects between the two events by considering their realistic crustal structures.

Seismic hazard assessment is challenging in remote regions such as the Tibetan Plateau where there is little available data due to a lack of near-field seismic stations, leading to large uncertainties in seismic source models. Interferometric synthetic aperture radar (InSAR) provides a method by which these areas can be monitored remotely and efficiently to better constrain source parameters without the need for additional seismic data.

The Tibetan Plateau’s Qiangtang Block, and surrounding blocks have been the location of several Mw 5 - 6 earthquakes in the past 20 years. Despite their frequency, these events are less well studied than large events such as recent Mw 7+ events in the Bayan Har Block and its bounding faults (e.g. the Kunlun fault to the North). Additionally, some moderate events occur on unmapped faults and at shallow depths, where seismic solutions can often have large uncertainties. Investigating and cataloguing these events will enhance our understanding of crustal dynamics in the Eastern Tibetan Plateau and similar tectonic settings.

This study aims to statistically compare earthquake source models calculated using geodetic measurements from InSAR with fault plane solutions from seismic catalogues such as the Global Centroid Moment Tensor Project.

The seismogenic fault geometry is constrained for most recent moderate magnitude earthquakes that have occurred in the Eastern Tibetan Plateau to build on previous catalogues of events in the region, and quantify the accuracy of existing seismic fault solutions using statistical methods.

InSAR with Sentinel-1 data is used to obtain coseismic interferograms for several Mw 5 - 6 earthquakes in the Eastern Tibetan Plateau. A Bayesian inversion approach is applied to constrain fault parameters from the InSAR data and characterise their uncertainties from the posterior probability density functions, based on an Okada model for a rectangular dipping fault with uniform slip in elastic half-space.

We will present preliminary models based on geodetic data for Mw 5 – 6 earthquakes in the Eastern Tibetan Plateau from July 2020 to January 2023. Results from this study are combined with geodetic fault solutions from previous studies and analysed statistically to quantify the accuracy of the existing seismic catalogue for the region.

In this research article, two methods, namely pixel offset tracking and interferometric phase, were employed to monitor ground movements in the Turkish earthquake. Despite their varying levels of accuracy, both techniques produced a consistent pattern (source: https://www.facebook.com/groups/radarinterferometry/permalink/6202173836515733). While pixel offset tracking can provide insight into potential shifts, the interferometric phase is more precise. However, the interferometric phase has a limitation in tracking spectral shifts within its bandwidth, making it more suitable for detecting slow-motion targets. Further information on the results of phase unwrapping is available in this Youtube video: (https://youtu.be/qQqmwBJgHj8).

The use of radar technology in monitoring ground movements during the Turkish earthquake is a significant milestone in disaster management. This technology has enabled researchers to detect and measure displacements caused by earthquakes, providing valuable insights that can inform future disaster response efforts. The two methods employed in this study, pixel offset tracking and interferometric phase, have shown to be effective in identifying ground movements, despite their varying levels of accuracy.

Pixel offset tracking is an incoherent technique that permits us to envision possible movements. It is not as precise as the interferometric phase but excels at identifying substantial displacements. The pixel offset method detects changes in the position of pixels between two images taken before and after the earthquake. By analyzing the changes in pixel position, researchers can estimate the amount of ground movement that occurred during the earthquake.

On the other hand, the interferometric phase is a more precise technique that uses the phase information of the radar signals to detect ground movements. The technique works by comparing the phase of radar signals reflected off the ground before and after the earthquake. By analyzing the phase changes, researchers can estimate the amount of ground movement that occurred during the earthquake. However, the interferometric phase has a limitation in tracking spectral shifts within its bandwidth, making it more suitable for detecting slow-motion targets.

Despite their varying levels of accuracy, both techniques produced a consistent pattern of ground movement during the Turkish earthquake. The findings align with previous reports, indicating an average displacement of approximately 4 meters across the East Anatolian Fault (EAF) and the adjacent Surgu fault. These results provide valuable information for disaster management efforts, as they help identify the areas most affected by the earthquake and the extent of the damage.

However, the process of unwrapping interferometric phase data can be challenging, particularly in areas with high levels of decorrelation or incoherence. In this study, the fault running through the entire scene divided the image into two parts, and each side was unwrapped individually. This approach helped to reduce the amount of decorrelation or incoherence, particularly near the fault. The affected areas were masked and later interpolated to provide a more accurate picture of the ground movements.

In conclusion, radar technology in disaster management has revolutionized how we respond to natural calamities. The pixel offset tracking and interferometric phase techniques have shown to be effective in detecting and measuring ground movements during the Turkish earthquake. While pixel offset tracking is not as precise as the interferometric phase, it excels at identifying substantial displacements. The interferometric phase, on the other hand, is more precise but has a limitation in tracking spectral shifts within its bandwidth. The findings from this study provide valuable insights that can inform future disaster response efforts and improve our understanding of geological events.

On February 6^th a strong M_w 7.9 earthquake hit the south-eastern sector of the Anatolia region (Turkey), close to the boundaries with Syria, followed by several aftershocks and another strong M_w 7.5 seismic event several hours later located some Km to the north. These two main events were generated by the dislocation of two different faults, the Eastern Anatolian Fault and the Sürgü faults. Both of them are characterized by left lateral strike-slip faulting mechanism which produced a prevalent horizontal coseismic surface displacement of several meters causing large damages to the infrastructures, building collapses and unfortunately more than 50.000 casualties.

In order to image the coseismic displacement field and to constrain the seismic sources responible for the two main events, the INGV GEOSAR Laboratory exploited several pairs of Synthetic Aperture Radar (SAR) images acquired by both Sentinel-1 and ALOS-2 space missions.

Satellite data were processed by SAR Interferometry (InSAR) [Massonnet et al., 1998] and Pixel Offset Tracking (POT) [Joughin, 2002] techniques to retrieve the full displacement field both along the satellite Line-of-Sight (LoS) and the Line-of-Flight (LoF).

By means of InSAR data, the LoS displacement due to the two events was estimated based on the phase difference between two images, i.e. the radar-to-target different travel times. InSAR analysis returns a phase differences map, called interferogram, where the LoS displacement is represented by several interferometric color fringes each one indicating a deformation proportional to the radar wavelength. The drawback is that, due to the strong displacement in the proximity of the epicenters, there is such a large number of fringes to produce phase ambiguity effects and causing signal loss. Such problem can be partially reduced by using L band data thanks to its larger wavelength of about 24 cm. Regarding the standard two-steps InSAR analysis, two pairs of L-band ALOS-2 SAR data acquired in SCANSAR WD mode along ascending and descending track were exploited. The ascending pair consists of images acquired on 05/09/2022 and 20/02/2023 and charcterized by a normal baseline of 20 m and a temporal baseline of 168 days. Instead, the descending one is formed by images acquired on 16/09/2022 and 17/02/2023 with a normal baseline of 48 m and a temporal baseline of 154 days.

Several fringes due to ionospheric artifacts were present along both the ascending and descending wrapped interferogram. They have been removed by estimating a planar ramp computed considering a narrow Region of Interest located along the borders of the frame and far from the expected displacement field as well. Furthermore, the coeherence maps along the two causative faults were masked to make easier the unwrapping step.

The obtained results are quite satisfactory even if some unwrapping errors are still present but the main patterns are well reproduced showing displacement values larger than ±2 meters across the left-lateral faults. Moreover, the availability of ascending and descending data allowed to move from LoS to E-W and U-D component of the displacement field.

On the other hand, POT techniques can be applied also on the amplitude of SAR signal which is not affected by phase problems as InSAR thus recovering displacement values also in the proximity of the causative faults. Such technique estimates pixel-by-pixel the shifts between pre- and post-event image both along the Line-of-Sight (Look Direction or Range) and the Line-of-Flight (Flight direction or Azimuth) of the satellite. The POT analysis was applied to the pair of Sentinel-1 descending data acquired on 29/01/2023 and 02/10/2023 which best cover both the seismic events.

Experimental results highlight a deformation pattern along both directions peaking at more than 2 m consistent with the left lateral strike-slip fault mechanism of the two structures responsible for the two main seismic events of 6 February. The accuracy of the measurements is inversely proportional to the pixel posting, which for S1 is about 3x15 m along the range and azimuth directions, respectively. In order to cross-validate the measurements and to be confident with the results, POT outcomes were compared with the E-W and U-D displacement component retrieved form InSAR along a NE-SW profile crossing the fault responsible for the M_w 7.5 event obtaining a good agreement in terms of displacement values and trend.

Further analysis concerning the pre-seismic phase have been also performed considering two SAR datasets from the Sentinel-1 mission. Indeed, 124 images acquired between January 2019 and January 2023 along ascending orbit (Track 14) and 147 images along descending orbit (Track 21) were processed using the P-SBAS approach. The P-SBAS processing service was accessed on the Geohazard Exploitation Platform (https://geohazards-tep.eu) operated by Terradue (www.terradue.com).

Finally, all the retrieved displacement maps were exploited as input for the modelling algorithms so to calculate the parameters of the seismic sources. Also a Coulomb Failure Function calculation was performed to estimate the stress transfer from the fault responsible for the first event to the nearest ones.

We utilize interferometric synthetic aperture radar (InSAR) observations to investigate the fault geometry and afterslip within ~4.5 years after a the 2017 Sarpol-e Zahab earthquake. Initially, we explore postseismic deformation sources using analytical models and determine that afterslip dominated the postseismic deformation while the viscoelastic response is negligible. Then we investigate the afterslip fault geometry and frictional properties by kinematic and stress-driven afterslip modeling. Our findings suggest that a multisegment, stress-driven afterslip model (hereafter called the SA-2 model) with depth-varying frictional properties better explains the spatiotemporal evolution of the postseismic displacements than a two-segment, stress-driven afterslip model (hereafter called the SA-1 model). Such a multisegment fault (SA-2 model) with depth-varying friction also is more physically plausible because of the depth-varying mechanical stratigraphy in the region. Compared to the kinematic afterslip model, the stress-driven afterslip models with friction variation tend to underestimate early postseismic deformation to the west, which may indicate more complex fault friction and/or more complex structure (splay fault) triggered during the postseismic period. Thus, we attempt to model the postseismic deformation using varied fault friction and more complex fault geometries from the perspective of 2-D finite element models. We incorporate ~4.5 years of InSAR measurements after the mainshock and 2-D numerical modeling to investigate the kinematic and mechanical afterslip models based on a series of planar, ramp-flat and splay faults, which could provide us some new insights into the postseismic physical process after the earthquake. Form the analytical and numerical modeling, the results are presented and discussed to understand the role of 2017 Sarpol-e Zahab earthquake to the crustal shortening, interaction between the sedimentary cover and basement in the Zagros Mountain Belt as well as the frictional properties of the complex seismogenic faults.

3D displacements are important for understanding and modelling surface-deforming events. Decomposing range and azimuth offsets from satellite data measured in different lines-of-sight into the standard Cartesian displacement fields allows easy integration of InSAR and optical pixel-tracking offsets with data from different sources for further modelling and applications.

We present ~100 m resolution 3D displacements, horizontal strain and surface slip distributions from the 2023 February Türkiye-Syria Earthquakes (Ou et al., 2023). The current 3D displacement field is jointly inverted from four tracks of Sentinel-1 range and azimuth offsets and a set of north and east displacements from Sentinel-2 pixel tracking.

We generate Sentinel-1 azimuth and range offsets as a rapid response of the COMET LiCSAR Earthquake InSAR Data Provider by cross-correlating 128x64 pixel windows (range x azimuth) over 2x oversampled deramped low-pass filtered intensity data. We also derive optical pixel tracking east and north offsets from L1C Sentinel-2 data using COSI-Corr's (Leprince et al., 2007) frequency correlator (two iterations with an initial and final window size of 64 and 32 pixels respectively) applied to the near-infrared band.

All the offset data are referenced to a distribution of dummy zero points away from the co-seismic ruptures by removing a planar ramp. We estimate empirical uncertainties of the offset data as mean absolute deviation in 4x4 pixels windows of the offset data, assuming nan-values are zeros. These uncertainties are used to weight the 3D motion inversion and are propagated to the uncertainties of the decomposed displacements through a model covariance matrix.

We also calculate horizontal displacement magnitude as a vector combination of the east and north motion fields, each masked by respective uncertainties. This horizontal displacement field allows us to extract surface slip distribution along the two faults ruptured during the Mw7.8 and Mw7.5 earthquakes. We further present the second invariant of horizontal strain resulting from these two earthquakes from the horizontal displacement gradients of east and north motions, after applying a median filter with 30 km windows at ~1 km intervals, which highlights the surface ruptures caused by the two earthquakes. The Mw7.8 earthquake generated over 310 km of surface rupture with a peak surface slip of 6.6 ± 1.2 m, whereas the Mw7.5 earthquake generated over 150 km of surface rupture with a peak surface slip of 7.5 ± 1.7 m.

We will present updated products including additional or reprocessed source data from Sentinel-1 data after their re-coregistration using rubber-sheet resampling (Yun et al., 2007), particularly co-seismic Sentinel-1 along-track displacements extracted by spectral diversity of burst overlaps and interferograms unwrapped after flattening phase gradients by spatially filtered range pixel offsets.

We have made the data available to the community for use in modelling. The data can be downloadable from https://catalogue.ceda.ac.uk/uuid/df93e92a3adc46b9a5c4bd3a547cd242.

References:

Ou, Q.; Lazecky, M.; Watson, C.S.; Maghsoudi, Y.; Wright, T. (2023): 3D Displacements and Strain from the 2023 February Turkey Earthquakes, version 1. NERC EDS Centre for Environmental Data Analysis, 14 March 2023. doi:10.5285/df93e92a3adc46b9a5c4bd3a547cd242.

Leprince, S.; Barbot, S.; Ayoub, F. and Avouac, J. -P. (2007): Automatic and Precise Orthorectification, Coregistration, and Subpixel Correlation of Satellite Images, Application to Ground Deformation Measurements, IEEE Transactions on Geoscience and Remote Sensing, vol. 45, no. 6, pp. 1529-1558, doi: 10.1109/TGRS.2006.888937.

Yun, S.-H., H. Zebker, P. Segall, A. Hooper, and M. Poland (2007), Interferogram formation in the presence of complex and large deformation, Geophys. Res. Lett., 34, L12305, doi:10.1029/2007GL029745.

On November 14, 2021, two earthquakes of magnitude 6.1 struck the Fin region in southern Iran. The first earthquake occurred at 12:7 GMT, while the second one occurred a minute later at 12:8 GMT. The earthquakes are located in the Simply Folded Belt, in the southeasternmost part of the Zagros Mountains. The focal mechanism solutions for both earthquakes from the CMT catalog suggest almost pure reverse slip on the E-W striking fault planes, either a low-angle (34°) north-dipping or a high-angle (62°) south-dipping nodal plane. This region is characterized by historical and instrumental earthquakes, including the 2006 March 25 Fin seismic sequences 40km west of the 2021 Fin earthquakes. A study of this 2006 seismic sequences based on the geodetic and waveform analysis reported that either both north or south dipping faults can be attributed to the earthquakes. Moreover, based on previous studies, reverse-faulting focal mechanisms are dominant in the region.

We investigate the coseismic surface displacement by processing Copernicus Sentinel-1 space-borne Synthetic Aperture Radar (SAR) data covering the study area, in ascending (A57) and descending (D166) geometries. During the interferogram generation process, the topographic and flat-earth phase contributions were removed from the differential interferograms using the 30 m Shuttle Radar Topography Mission Digital Elevation Model. The turbulent component of the tropospheric delay was corrected using atmospheric parameters of the global atmospheric model ERA-Interim provided by the ECMWF. Finally, the generated interferograms were filtered using Goldstein's filter and unwrapped with a branch-cut algorithm. Both interferograms exhibit a maximum displacement of ~40 cm in the line-of-sight direction of the satellite. To obtain improved source parameters and centroid depths for both earthquakes, teleseismically recorded P and SH body waves were modeled. Uniform slip modeling was applied using a Bayesian bootstrap optimization nonlinear inversion method to find earthquake source parameters. These parameters include length, width, depth, strike, dip, rake, slip, location of the fault plane, rupture nucleation point, and origin time. Search grids were specified based on the LOS displacement map and focal mechanism solutions for each fault parameter to find the best solutions. In the next step, we extend fault length and width and try to find slip distribution in different patches of derived fault from uniform slip modeling. The slip distribution, geological information, and relocation of seismic sequence help us to understand the relation between folding and the Fin doublet earthquakes.

On 6 February 2023, at 01:17 UTC, a Mw 7.8 earthquake struck southern and central Turkey and northern and western Syria. The epicenter was 37 km west–northwest of Gaziantep. A second earthquake of Mw 7.7 magnitude followed at 13:24, causing extensive destruction in both countries. This second earthquake was centered 95 km north-northeast from the first one. There were widespread damages to infrastructures and buildings and tens of thousands of fatalities. The two major earthquakes were followed by hundreds of smaller aftershocks and the seismic sequence was the result of shallow strike-slip faulting.

The damage caused by the earthquakes affected an area of ​​350,000 km². These earthquakes and the following aftershocks are the worst to strike the region in almost a century. Tens of thousands of people have been killed with many more injured in this tragedy and United Nations estimated that about 1.5 million people were left homeless [1].

Damaged roads, winter storms and disruption to communications hampered the Disaster and Emergency Management Presidency's rescue and relief effort.

The Italian Space Agency (ASI) has been activated by Istituto Nazionale di Geofisica e Vulcanologia (INGV) to provide satellite images over the seismic-affected areas to define the extent of the disaster and support local teams with their rescue efforts.

Radar imagery from satellites allows scientists to observe and analyze the effects that earthquakes have on the land. The COSMO-SkyMed constellation carries a radar instrument that can sense the ground and can ‘see’ through clouds, whether day or night.

The COSMO-SkyMed constellation in its initial configuration consisted of four identical satellites, each equipped with a high-resolution microwave Synthetic Aperture Radar (SAR) operating in the X-band and positioned in a sun synchronous orbit at ~ 620 km above the Earth's surface. Following the four First Generation satellites, the mission is continuing with two Second Generation COSMO-SkyMed satellites also based on identical satellites equipped with an X-band SAR payload and positioned on the same orbital plane of the First-Generation satellites.

Thanks to the “COSMO-SkyMed Background Mission” planned by ASI since 2008 on the mission satellites, images of almost all the cities that have been hit by the seismic swarm are present in the COSMO archives.

The “Background Mission” has been conceived by ASI to maximize and optimize the use of the COSMO-SkyMed system with the aim of collecting data acquisitions all over the world and populating the image archive. This planning is intended to guarantee the availability of reference datasets for future mapping projects, emergency mapping and change detection applications. Data collected are stored and made available when required. The acquisition plan is kept as simple as possible so that it can be exploited with low priority modality (for example, using right-looking acquisitions as default configuration).

The Background Mission implements the lowest level of priority plan, i.e. it is performed when no further activity (so called foreground activity) is defined.

Following the activation, about 300 pre-event images, acquired by COSMO-SkyMed satellites in STRIPMAP mode (3x3 m resolution) on various cities affected by the seismic swarm, and about 100 post-event images were delivered. The two sets of images (pre and post-event) can be used to generate damage and situation maps to help estimate the hazard impact and manage relief actions in the affected areas.

Furthermore, a dedicated acquisition plan was required and planned to monitor the fault.

All the activities have been coordinated within the Working Group on Disaster (WGD) of Committee on Earth Observation Satellites (CEOS), that has been working from several years on disasters management related to natural hazards through pilots, demonstrators, recovery observatory concepts, Geohazard Supersites, and Natural Laboratory (GSNL) initiatives (https://ceos.org/ourwork/workinggroups/disasters/). In detail, the “Kahramanmaraş Event” Supersite has been put in place in coordination among Marmara Supersite users, CEOS WGD and space agencies (https://ceos.org/news/kahramanmaras-event-supersite/), to ensure Earth observation data for recovery efforts and to provide scientific information about this devastating hazard. In this framework, taking benefits also of the ASI-CONAE “SIASGE” cooperation, ASI is supporting Supersite users by providing COSMO-SkyMed and SAOCOM products over the earthquake areas of interest (AOIs).

Satellite data are being used to help emergency aid organizations in assisting earthquake-affected people, satellite analysis is aiding risk assessments that authorities will use as they plan recovery and reconstruction, as well as long-term research to better model such events.

[1]: "1.5 million now homeless in Türkiye after quake disaster, warn UN development experts". United Nations Office at Geneva. 21 February 2023. Retrieved 23 February 2023.

ABSTRACT:

Major earthquakes events are common around the global. They can cause severe damage to both human lives and sending weakened structures crashing down. Remote sensing data and methods are nowadays widely deployed to produce damage maps after natural disasters. Therefore, this study aims to explore the potential application of the state-of-the-art satellite technology is to the rapid mapping of damage after caused by multiple earthquakes occurrence. In order to achieve the goal, a rapid damage mapping approach is proposed combining deep learning using Interferometric Synthetic Aperture Radar (InSAR) observations of an impacted region due to earthquake. The case study is a region near Pazarcık City in south-central Turkey, that at 04.17 on 6 February 2023, an Mw 7.8 earthquake struck followed by an Mw 7.5 event about 9 hours later. These earthquakes More than 50,000 dead and thousands injured across Turkey and Syria, are the largest earthquakes to hit Turkey in recent years. In this research, land surface changes are are calculated using time series of displacement and radar coherence, then use a long short-term memory network (LSTM) in order to real time anomaly detection. The LSTM is first trained on pre-event displacement and coherence time series, and then predict a probability distribution of the displacement and coherence between before and after synthetic aperture radar (SAR) images. SAR can map damage in any weather condition even under thick cloud cover. The analysis of displacement and radar coherence time series of many interferograms is performed using Sentinel-1 SAR data to investigate the conditions pre- and post-event the earthquake. The time series of displacement and radar coherence extracted from SAR images have strong responses to damage due to earthquakes which is expressed by a sudden changes in the values of displacement and coherence. Also, pre- and post-event Sentinel-2 optical images is used to confirm the destructive effects of earthquakes in the region. Through this review, the consequences of earthquakes for structures and buildings in terms of various types of damage and warnings are reported and some new insight will be provided for potential use of remote sensing for the mitigation to reduce damages.

Keywords: Multiple earthquakes, Sentinel-1, Coherence, Turkey, Sentinel-2, InSAR

SAR techniques, including InSAR and PolSAR, are well-established and are employed in natural and anthropogenic hazard monitoring, as well as land use and land cover classification. As an increasing number of dedicated SAR missions are launched, the community of SAR users is also expanding. There are now more than 863,000 SAR-related journal articles, published since the 1990s. Yet, due to the complex nature of SAR imagery and the limited availability of labelled SAR datasets, SAR products are less widely used than optical remote sensing imagery for machine learning applications. Open-access SAR benchmark datasets along with detailed specifications that can facilitate such applications are, therefore, strongly needed.

To this end, the AlignSAR project will: 1) design a generic procedure for the creation of SAR benchmark datasets; 2) develop a reference, quality-controlled, documented, open benchmark dataset of SAR spatial and temporal signatures of complex real-world targets. These will be highly diverse, to serve a wide number of applications with societal relevance, and respecting FAIR (findable, accessible, interoperable, reproducible) and Open Science principles; 3) create the database considering both ongoing and complete SAR missions, maximization of the geographical and temporal coverage, and integration and alignment of multi-SAR images, and other geodetic measurements, in time and space; 4) define a specification of the SAR signatures and their associated descriptors so that they can be easily indexed and programmatically searched and retrieved; 5) develop an open-source software library, with associated documentation, to create, describe, test, validate and publish SAR signatures, and expand the SAR benchmark datasets.

We will present the latest progress of the AlignSAR project, funded by ESA, and led by the University of Twente in collaboration with the University of Leeds, AGH University of Science and Technology, and RHEA group. We will introduce the first version of the Open SAR library encompassing representative SAR benchmark datasets, signatures, specifications and software tools. We will describe the procedure and methods for the creation of SAR benchmark datasets. We will also demonstrate, test and validate this library on two test sites in the Netherlands and Poland, using Sentinel-1 SAR data, legacy SAR data, and geodetic measurements applied to machine learning-based land use, land cover, and surface dynamics classification.

The Interferometric Synthetic Aperture Radar (InSAR) technique allows the measurement of ground surface displacements over wide areas. InSAR data are used in a variety of fields, including displacement monitoring in mining areas. Continuous observations of subsidence and the prediction of impacts caused by underground mining operations are an important issue for the protection of buildings and infrastructure located in areas affected by mining activities. Machine learning methods are effective in analysing significant amounts of data to explore patterns and make predictions. This study aims to assess the feasibility of applying InSAR data and machine learning algorithms to the prediction of displacements in a mining area.

The study was carried out in an area of underground copper mining in south-western Poland. The Small Baseline Subset (SBAS) InSAR method was used to measure ground surface displacements based on Sentinel-1A and 1B imagery. The analysis covered the period from May 2016 to October 2020. The displacement study used data from the ascending and descending satellite tracks to account for horizontal displacement and to determine the time series of vertical displacement in the study area. InSAR results were processed by selected time series forecasting methods and machine learning models to develop a forecasting model. The prediction horizon of six months was assumed. Traditional methods (ARIMA, Exponential Smoothing), machine learning models (linear regression, decision trees and ensemble models), and neural network models (N-BEATS model, Recurrent Neural Network and BlockRNN model) were used in the study. The machine learning methods and neural networks were designed in a global approach, with the aim for a single model to predict displacements on a set of time series over a given area. The performance of the models was compared with the naive baseline model using the MAE, RMSE and MAPE accuracy metrics.

The SBInSAR technique determined the time series of vertical displacements in the study area, which allowed the identification of subsidence zones corresponding to the locations of the underground mining operations. The time series of vertical displacement values were validated with levelling measurements, with an R-squared value of 0.94, indicating strong agreement between the SBInSAR measurement and the field measurement. Machine learning models trained on the displacement time series showed an increase in performance of approximately 20 to 40% over the baseline models, depending on the region in which the displacements were forecast. Among the models tested in the study, the regression ensemble model proved to be the most effective, based on the accuracy metrics. The main limitation of this method is the inability of the models to account for rapid changes in the time series, resulting from e.g. mining-induced seismicity.

The study demonstrated the feasibility of using InSAR time series to predict displacement in mining areas using machine learning algorithms. The data processing scheme applied in the study enabled global models predicting displacements in a given area to be developed. Further research should consider applying new machine learning models and using additional data, to create more complex models able to measure the impact of various factors on deformations.

In the era of big SAR data, it is urgent to develop dynamic InSAR processing method, especially for landslides that occur successively in mountainous areas, which require dynamic monitoring. In addition, the dense vegetation coverage in mountainous areas causes severe decorrelation, which requires both the accuracy and efficiency of phase optimization processing. For time-series interferometric phases optimization of distributed scatterers (DSs), the SqueeSAR technology used the phase linking (PL) to extract the equivalent single-master (ESM) interferometric phases from the multilooking time-series coherence matrix. The highest achievable estimation accuracy of the ESM phases depends on the number of looks and the time-series coherence matrix. With the abundance of time-series polarimetric SAR data, many scholars have studied the coherence magnitude-based polarimetric optimization methods for optimizing the DS’s time-series interferometric phases. However, traditional polarimetric optimization algorithms cannot work satisfactorily because of the unstable statistical characteristics and low efficiency, which limits the application of multi-polarization phase optimization methods in large-scale and long-sequence scenarios. Furthermore, variations in the scattering characteristics of terrain in actual SAR scenes may result in less identification of homogeneous areas, which directly affects the accurate estimation of the coherence matrix.

To achieve efficient InSAR time series analysis dynamically, we combined the sequential estimator with polarization stacking method, named SETP-EMI. In terms of homogeneous point identification, we identify and update the homogeneous pixels after each new sub-dataset of SAR images acquired to prevent the loss of long-term consistency. In terms of interferometric phase optimization, all the Time Series interferometric coherency matrices (TSIn) of three Pauli basis (full polarization) or two Pauli basis (dual polarization) polarimetric channels can be taken as statistical samples, and the Time Series Total Power (TSTP) coherency matrix can be constructed by stacking all available PolSAR scattering vector. To pursue the efficient stacking scheme, Sequential Estimator was combined with TSTP, which starts with a mini-image stack with a predetermined size. Then we process the mini stack with phase estimation and compression. In the phase estimation, EMI is used to enhance the phase SNR based on the TSTP coherency matrix of this mini stack. Next in compression, the first mini stack is compressed into a one-rank subspace by linear transformations, which can retrieve the coherence via formation of artificial interferograms between the compressed and the newly acquired data.

In order to illustrate the advantages of the new method in terms of accuracy and efficiency compared with traditional methods, we conducted simulation experiments and real data tests respectively.

In the simulated experiments, the proposed algorithm can better improve the optimization performance of time-series interferometric phases of DSs than the single channel SAR algorithms in terms of interferometric phase restoration. In addition, the computational efficiency and storage burden are relatively low. With the accumulation of time-series data in the data set, the calculation time consumption of the single-polarization EMI method increases exponentially, which is much longer than that of the SETP-EMI method. See Fig. 1.

In the real data experiment, we selected two landslides in the reservoir area of Lianghekou hydropower station (E 101°0′20″, N 30° 21′20″) as example. These two ancient landslides were reactivated by the impoundment of the reservoir. we collected 174 scenes of Sentinel-1 dual polarization data (VV and VH) from January 24, 2017 to November 23, 2021. Fig. 2 shows the optical image map of the study area and the corresponding amplitude image of Sentinel-1 before and after water storage.

Compared with the single-polarization EMI method, the phase optimization accuracy of the SETP-EMI method is significantly improved, and the computational burden is also lower. In terms of efficiency, when a single CPU core is used to process 174 scenes of 500 by 1200 pixels, the single-polarization EMI takes 28 h, while the sequential multi-polarization EMI takes only 6 h. In terms of accuracy, SETP-EMI obtained interferograms with better spatial coherence. After the phase optimization of EMI, the average spatial coherence of the interferometric phase increases from 0.21 to 0.45. However, the phase optimization process of SETP-EMI improves the coherence to 0.71. See Fig. 3.

Landslide and erosion processes are causes of major concern to population and infrastructures on Reunion Island. These processes are led by the tropical climate of the island. The hydrological regime of the rivers is distinct owing to the coexistence of several major parameters that predispose it to extreme vulnerability. Holding almost all the world records for rainfall between 12 h (1170 mm) and 15 days (6083 mm), the island has a marked relief with a peak at 3,069 m, with exceptional cliffs that reach 1500 m in height.Cirque de Salazie (CdS) is the rainiest of the large erosional depressions on Reunion Island with an average annual cumulative rainfall of approximately 3,100 mm since 1963; a minimum of 698 mm was recorded in 1990, and a maximum of 5,893 mm was recorded in 1980.This depression is surrounded by steep rock cliffs and filled with epiclastic material. Intense river erosion incises deep valleys and has produced several isolated plateaus across the cirque.

This study examined the results of an interferometric Synthetic Aperture Radar (InSAR) and SAR Offset Tracking (OT) study on Cirque de Salazie, Reunion Island, France, within the context of the RENOVRISK project, a multidisciplinary programme to study the cyclonic risks in the South-West Indian Ocean. Despite numerous landslides on this territory, CdS is one of the denser populated areas in Reunion Island. One of the aims of the project was to assess whether Sentinel 1 SAR methods could be used to measure landslide motion and/or accelerations due to post cyclonic activity on CdS. We concentrated on the post 2017 cyclonic activity. We used the Copernicus Sentinel 1 data, acquired between 30/10/2017 and 06/11 2018. Sentinel 1 is a C-band SAR, and its signal can be severely affected by the presence of changing vegetation between two SAR acquisitions, particularly in CdS, where the vegetation canopy is well developed. This is why C-band radars such as the ones onboard Radarsat or Envisat, characterized by low acquisition frequency (24 and 36 days, respectively), could not be routinely used on CdS to measure landslide motion with InSAR in the past. In this study, we used InSAR and OT techniques applied to Sentinel 1 SAR. We find that C-band SAR onboard Sentinel 1 can be used to monitor landslide motion in densely vegetated areas, thanks to its high acquisition frequency (12 days). OT stacking reveals a useful complement to InSAR, especially in mapping fast moving areas. In particular, we can highlight ground motion in the Hell-Bourg, Ile à Vidot, Grand Ilet, Camp Pierrot, and Belier landslides.

eo4alps-landslides is a community-tailored application giving access to several on-line geo-information services for landslide ground motion analysis and hazard modelling. It allows the exploitation of satellite imagery time series, use of advanced InSAR and optical ground motion services and advanced modelling capacities for the assessment of gravitational hazards. The application aims at ensuring that satellite-based Earth Observation (EO) products in combination with models are increasingly and more efficiently used in practice for both science and operational landslide analyses. The application has been designed by an hybrid consortium of research centres and geological engineering companies, and with the support of more than 20 active users (state authorities, stakeholders responsible for landslide disaster risk management). The presentation targets the presentation of the portfolio of “eo4alps-landslides” services and products in order to create ground motion maps, harmonised and advanced landslide inventories and susceptibility/hazard maps with examples in the French, Swiss and Italian Alps. The EO-based services and products can be complemented by local datasets and terrain data from the end users.

The products include 1) automatic landslide detection using satellite optical and InSAR-based services, 2) harmonised and advanced landslide catalogues resulting from the satellite based detection and local inventories, 3) susceptibility/hazard maps consisting of possible landslide source areas and landslide type-specific runout modelling. Specifically landslide-tailored SqueeSAR datasets have been created for large regions of the European Alps. and will be presented and discussed. The services are generic in order to be used at several spatial scales. The application is accessible on the Geohazards Exploitation Platform (GEP) and a sustainability plan will be presented.

The detection and monitoring of active landslides on populations and their livelihoods in high mountain areas are important to stablish and to mitigate associated hazards, territorial spatial planning and to determine criteria in case of relocation of populations. Differential Interferometric SAR (DInSAR) and Persistent Scatterers (PS- DInSAR) are powerful remote sensing tools to identify the spatial distribution of landslides and the deformations that occur as a result of their activity. This research used 42 SAR images from the SENTINEL-1 satellite in ascending and descending mode to determine the spatial extension, the deformation rate and the hot spots of maximum soil deformation occurred by the active landslide in Chango Population Center (CPC), in the department of Cerro de Pasco, Peru. The DInSAR results in the ascending orbit showed that the accumulated ground deformation at the CPC had a minimum value of -31.3 mm and a maximum value of 56.6 mm along the satellite line-of-sight (LOS) for the study period, although this value could be affected by atmospheric disturbance. Regarding PS-DInSAR, the results allowed to determine that, both in the descending and ascending geometry in CPC there are slow and extremely slow landslide phenomena in the Cruden and Varnes range. The application of both geometries allowed estimating the east-west (E-W) and vertical deformation. For E-W component, soil displacements have been found in the range of [-60 to -70]mm/y and a vertical component of soil displacement that is between [-25 to 30]mm/y. In addition, the total ground deformations are in the range of [-613 – 687]mm on average during the study period for ascending and descending orbit. In addition, the PS application made it possible to map 14 areas of active landslides with the maximum deformations (hotspot) of the soil in the study area. It was also found that the greatest soil deformations caused by the active landslide occured in the wet season and were located in the CPC close to the main escarpment of the landslide, evidenced by a high concentration of PS-DInSAR maximums in the descending and ascending orbit, identification of the extreme cold and hot spots by means of statistical cluster analysis (Gi-Bin) and recording the highest values of deformation in the E-W and vertical direction and the total deformations.

A comparison was made between the results of the cumulative sum of the interferograms unwrapped with DInSAR and PS-DInSAR in the ascending orbit, the results show a robust correlation R²=0.74 and identification of deformation patterns of uplift and subsidence of the soil in the entire extension of the rural area of the CPC.

Finally, the DInSAR and PS techniques allowed to determine the soil deformation caused by the CPC landslide. In addition, it allowed to identify the spatial distribution, the soil deformation rates and the hot spots where the greatest soil deformations occurred. This research leads to optimizing resources and implementing a focused soil deformation monitoring system and performing engineering controls and risk assessment, although it is true that the effectiveness of risk control works could be inappropriate given the extent of the deformation found in the study area even the called old landslide presents movement, therefore, the relocation of the study area must be consciously evaluated. In the future, it is feasible to carry out a near-real-time alert system based on SAR applications for prevention and monitoring purposes.

A-DInSAR (Advanced Differential Synthetic Aperture Radar Interferometry) is widely acknowledged as one of the most powerful remote sensing techniques for measuring Earth’s surface displacements over wide areas such as subsidence, landslides and seismic activity. Thanks to the large number of applications in several scenarios, A-DInSAR techniques became a common tool to understand and quantify deformation processes, monitor and preserve several man-made structures and mitigate natural hazards.

Characterization and interpretation of land-deformation processes can greatly benefit from the application of A-DInSAR post-processing analyses, especially when a complex deformation behaviour cannot be easily highlighted and understood. Therefore, NHAZCA Srl has developed and designed the software “PSToolbox” a set of post-processing plugins for the open-source software QGIS, with the aim to enhance spatial and temporal deformation trends of the A-DInSAR results, as well as visualize the differences between multi-satellite datasets. Indeed, in a complex scenario, such as vertical structures or landslides in areas with complex topography, the geometric distortions and the site coverage percentage can lead to a lack of information and difficulties to interpret the interferometric multi-image results.

On that account, the post-processing plugins allow to derive information about the kinematic of displacement processes applying specific analyses where the principal functionalities are the following:

• Vectorial decomposition permits to quantify the displacement along the vertical (Up/Down) and horizonal axis (West/East);
• Interferometric section allows to visualize the displacement velocity of measurement points along a section;
• Create a 3D model of measurement points in order to study deformation that affects structures or slopes;
• Classification of linear features (i.e., pipelines, aqueducts, highways etc.) by the estimated displacement trend with the aim to highlight hazardous sectors;
• Highlight changes in the deformational trend of displacement time series.

Therefore, in this study we present several cases of application of post-processing analyses to enhance the A-DInSAR data spatial information and derive a more detailed behaviour model of the investigated processes applying the “PStoolbox” plugins. In order to get to this outcome, we used the measurement points derived from the processing of the SAOCOM (L-band, Comisión Nacional de Actividades Espaciales – CONAE), Sentinel-1 (C-band, European Space Agency – ESA) and COSMO-SkyMed (X-band, Agenzia Spaziale Italiana – ASI) SAR data acquisitions. These measurement points encompass various Italian complex scenarios affected by landslides in Southern and Northern Italy where natural hazards affect some principal economic assets.

Rates of land subsidence in the Samoan Islands rapidly increased after the 2009 Samoa-Tonga earthquake, exacerbating environmental hazards from sea level rise in a region already strongly exposed to climate hazards [1, 2]. Understanding and predicting future trends of vertical land motion (VLM) from current observations requires both a first-order estimate of the rates of subsidence but also an understanding of changes in those rates over time. However, deriving high-resolution estimates of VLM trends in the Samoan Islands is difficult given the challenging terrain: heavy vegetation, rugged topography, and thick cloud cover over small island landmasses. Previous work has shown the ability of InSAR to resolve estimates of VLM over a span of 6 years given a large data stack of Sentinel-1 imagery processed with an innovative time-series method that fuses advantages of SBAS and PS methods [3]. Still, resolving second-order rate changes introduces further challenges and more stringent accuracy requirements, given the reduced size of the available data stack.

In this presentation, we detail current successes and challenges in constraining temporal changes in subsidence rates on the Samoan Islands. Specifically, we focus on the island of Upolu in the Independent Nation of Samoa and the island of Tutuila in American Samoa; both islands have one independent and permanent ground GPS/GNSS station that can be used a reference point for tying InSAR measurements to the geodetic frame. For each island, we also analyzed VLM measurements derived from differenced tide gauge/altimetry data but found them too noisy for shorter time periods to serve as a useful comparison with InSAR data. For InSAR data, we analyzed all available data from Sentinel-1 between 2016-2023 and subset into two separate time periods (2016-2019, 2020-2023), corrected and geocoded the data using a backprojection processor [4], and applied redundant PS-InSAR time-series analysis to the data stack as described in [3]. To improve the accuracy of estimated rates for our 3-year time-series compared to the 6-year time-series, we integrated new corrections in our processing workflow to address inconsistencies arising from phase misclosure and updated the primary selection methods to include consideration of phase misclosure inconsistencies. Initial results are promising, leading to VLM estimates that are more spatially realistic as well as more consistent with GPS/GNSS data and about a 10% reduction in estimated time-series error. We also discuss ongoing studies into optimal methods to compensate for DEM error and atmospheric phase contamination. By continuing to improve the accuracy of InSAR techniques at shorter time scales over challenging terrain, these developments will enable fine-scale temporal analysis of rate change with InSAR data using workflows that can be deployed more quickly – by requiring analysis of fewer acquisitions for the same accuracy – and will be less data-intensive than current methods.

[1] Han et al., “Sea Level Rise in the Samoan Islands Escalated by Viscoelastic Relaxation After the 2009 Samoa-Tonga Earthquake,” Journal of Geophysical Research: Solid Earth, vol. 124, no. 4, pp. 4142-4156, 2019.

[2] Martínez-Asensio, et al., “Relative sea-level rise and the influence of vertical land motion at Tropical Pacific Islands,” Global and Planetary Change, vol. 176, pp. 132-143, 2019.

[3] S. Huang, J. Sauber, and R. Ray, “Mapping Vertical Land Motion in Challenging Terrain: Six‐Year Trends on Tutuila Island, American Samoa, With PS‐InSAR, GPS, Tide Gauge, and Satellite Altimetry Data,” Geophysical Research Letters, vol. 49, no. 23, 2022.

[4] H. Zebker, “Sentinel-1 Analysis Ready Data – A Convenient and Easy to Use System Producing Common-coordinate Timeseries,” AGU Fall Meeting Abstracts 2022, G42D-0257, 2022.

The detection and the measurement of transient aseismic slip events allow a better understanding of the seismic cycle on major seismogenic faults and their associated seismic hazard. In this study, we focus on the last ruptured segment of the western North Anatolian Fault (NAF) in Turkey, the Izmit segment affected by two large Mw 7.6 and Mw 7.2 earthquakes in 1999. We use the Interferometric Synthetic Aperture Radar (InSAR) products (interferograms and time series), automatically processed by the FLATSIM project developed as part of the ForM@Ter Solid Earth data and services center, and supported and operated by CNES, following the NSBAS processing chain and using the Sentinel-1 data, acquired over the period 2015-2021 (Thollard et al., 2021). From the mean velocity field, we first estimate a creep rate around 5 mm/yr at a depth between 0 and 10 km along the Izmit segment. By comparing with geodetic measurements from InSAR, Global Navigation Satellite System (GNSS) and creepmeters from previous studies, we confirm a logarithmic decay of the postseismic afterslip, which is still active more than 20 years after the mainshock. Second, we analyze the temporal dynamics of creep on the Izmit segment. We study the seasonal signals on all the tracks and decompose them into horizontal and vertical components to characterize potential annual creep modulations. We then test the ability of these InSAR time series for the detection and the quantification of transient slip events, in addition to the post-seismic signal. To do so, we adapted for InSAR time series, a geodetic matched filter approach dedicated to the automatic detection of small slip events (equal or lower than the noise level), first developed for GNSS time series datasets. We conducted an analysis on synthetic time series calculated using realistic noises and transient slip events in order to evaluate the resolution of the potential transient events detection, in terms of depth and size/magnitude. For the atmospheric noise of the region and the geometry of both the NAF strike-slip fault and SAR acquisitions, we show that events with Mw 4.9 close to the surface and Mw 5.5 at 5 km depth can be detected. Further work is needed to validate the method on real InSAR time series.

Physics-based models of volcanic eruptions coupled with parameter estimation and uncertainty quantification methods are one of the most promising tools to improve our forecasting capabilities of volcanic hazards. During an eruption, the surface is affected by two main changes. The first is due to variation in pressure in the plumbing system, and the second due to the spreading of viscous lavas flows. InSAR datasets are becoming more and more important, as they can provide two key observables: surface deformation time-series, and net topographic change, obtained through DEM differencing. In this work, we focus on the latter and perform a sensitivity study to understand how the spatio-temporal sampling of different InSAR missions affect the features and errors of the derived DEMs, and how in turn these propagate in the accuracy and uncertainty of eruption model forecasts. To achieve this, we generate synthetic datasets covering a wide range of scenarios, from the extrusion of small viscous domes to the eruption of large lava flows, and consider different noise model and acquisition strategies. In a first step, we use simple models which simply predict the total erupted volume as function of time, but later include more sophisticated cases in which also the shape of the flow is predicted. We finally compare these findings with real data of recent eruptions, including the ones of Kilauea in 2018 and of Mauna Loa in 2022.

Our results highlight the necessary trade-off between noise levels and resolution, which is mainly controlled trough the choice of the looks number during processing and provide guidelines for future InSAR missions targeting volcanic hazard monitoring and mitigation.

The Observational Products for End-Users from Remote Sensing Analysis (OPERA) project at the Jet Propulsion Laboratory aims to enhance the accessibility of Sentinel-1 synthetic aperture radar (SAR) data by creating a surface displacement product of North America using interferometric synthetic aperture radar (InSAR) . To achieve this, the OPERA displacement algorithm will use a hybrid persistent scatterer (PS)/ distributed scatterer (DS) approach to produce updated displacement time-series each time a new SAR image is available. Unlike other wide-area processing products , the OPERA product will use geocoded single-look complex (SLC) images for the coregistration step . Since most existing persistent scatterer algorithms and software assume the SLCs are in radar geometry, new accommodations must be developed for differences in the geocoded domain. Moreover, the PS and DS selection algorithms should be amenable to online, incremental adjustments without re-downloading and analyzing the entire historical archive.

PS candidate pixels are often selected from single-look complex (SLC) images using amplitude dispersion, Da.

For InSAR studies considering one time period, Da is usually computed using all available SAR images, as a larger N leads to a more reliable estimate of σ and μ. However, when the time span of available SAR imagery grows beyond several years, the land surface may have undergone temporal changes (for example, due to construction/destruction of buildings). Incorporating too many SLCs to compute Da can cause near-real-time systems to be insensitive to changes.

In this study, we investigate modifications to PS and DS candidate selection algorithms that balance a low false positive rate with the ability to adapt to temporal changes in surface scattering. As a case study, we used 162 Sentinel-1 images acquired between 2015 and 2021 over downtown Miami, Florida. To determine the minimum number of SLC images required to reliably estimate , we varied the stack size from 5 SLCs to all 162 SLCs (Figure 1). We found that when the stack size is small (N < 20), many false positive PS pixels are selected.

To determine the effect of changes to the scattering surface on the PS density, we identified an urban area which saw a large increase in the number of candidate PS after a high-rise construction project finished in 2018 (Figure 2). By considering separate stacks of 50 SLCs from the beginning of the study period (Sep. 2015 to Jul. 2018, Figure 2, top row) and the end of the period (Mar. 2020 - Dec. 2021, Figure 2, middle row), we observe high in the construction zone (green box). After the construction finished, the new building created a region in the image with 40 new PS candidates (Figure 2(i)).

Since long-running deformation monitoring projects should be able to identify these PS changes, we developed an algorithm which incrementally updates a map of using a modification of Welford’s online algorithm . Given a mean and variance computed from pixel amplitudes at times , Welford’s algorithm gives the new mean and variance using only . However, when N is very large, the new has little effect on or . We can modify the original algorithm to calculate a moving mean and variance which only considers a fixed window of the last data points.

Since this requires only the new amplitude and a single old amplitude , we avoid re-pulling large stacks of data in the computation. We also compared the use of an exponentially-decaying mean and variance to weigh recent acquisitions more heavily. Additionally, we incorporate a sequential -test to detect sudden large changes in backscatter . This test is used to exclude formerly high-amplitude pixels which are no longer useful measurement points, as well as pixels which undergo seasonal variations in their backscatter.

When analyzing DS pixels using phase-linking algorithms, the covariance matrix is typically estimated by averaging a neighborhood of statistically homogeneous pixels (SHPs) . The SHP neighborhood can be found using a statistical test, such as the Kolmogorov-Smirnov test (KS test) , which involves comparing the empirical cumulative distribution function (CDF) of a pixel with that of its neighbors. However, computing the empirical CDF for each KS-test can be time-consuming. To address this issue, we have developed a faster alternative method for selecting SHPs. This method uses the Kullback-Leibler (KL) distance and only requires the values of σ_N and μ_N.

Under the assumption that the amplitudes of each pixel can be approximated as Gaussian, we compute the KL distance between the pixels by plugging in their means and variances into Equation [eq:KL]. We label them as SHPs when the distance between the estimated PDFs is low. We compared the KL distance method for finding SHPs to the KS-test and the -test methods using a stack of 20 Sentinel-1 SLCs over the Island of Hawaii. We found that the KL distance method chose similar SHP neighborhoods as the KS- and -test methods (Figure 3), but computed the results over two orders of magnitude faster than the KS-test method.

Combining the PS and DS improvements, we demonstrate the performance of the proposed algorithms for estimating ground displacement time-series using Sentinel-1 at native SAR resolution and in near-real time with short latency.

Permafrost is a region where the ground temperature remains below 0°C for more than two years and is formed by ice combined with various types of soil, sand, and rocks. Recently, permafrost thawing has occurred due to global warming, and ground motion caused by natural hazards such as earthquake can cause instability of the permafrost. On August 12, 2018, an earthquake of Mw 6.4 (mainshock) occurred in the Sadlerochit Mountains of the Brooks Range on the southern North Slope of Alaska, which is a permafrost region. Six hours later, an earthquake of Mw 6.0 (aftershock) followed. This series of earthquake event is called Kaktovik earthquakes. In this study, the Small BAseline Subset (SBAS) interferometric SAR (InSAR) technique was applied to 31 Sentinel-1 SAR images acquired from February 2018 to February 2019 to measure the pre-, co-, and post-seismic surface displacements of the earthquakes in the time series. A total of 87 interferograms were generated from the Sentinel-1 SAR images, among which 15 interferograms that were difficult to observe surface displacement due to low coherence were excluded from the time series displacement analysis. During the six months before the earthquakes, there was no surface displacement in the study area. Immediately after the earthquakes, three regions with different magnitudes (−27 to 12 cm) and directions of the displacement in line-of-sight (LOS) were clearly distinguished. Among the three regions, LOS displacement of −4 cm was observed in the rock glacier region for six months after the earthquakes, while little displacement was observed in the other regions. This suggests that mountain permafrost may be vulnerable to earthquakes. In the permafrost of the North Slope, 20–30 km from the epicentre of the Kaktovik earthquakes, very small LOS displacement (±1 cm) was observed for six months after the earthquakes. In our future works, time series surface displacements in different radar look directions will be measured by using Sentinel-1 SAR images acquired in ascending and descending orbits over the area of Kaktovik earthquakes. The directions and magnitudes of pre-, co-, and post-seismic displacements will be determined from the time series displacements in multiple radar look directions, and the behaviour of ground deformation by the earthquakes in the permafrost region will be analyzed.

The measurement of ground displacement over large geographic areas is made possible with Interferometric Synthetic Aperture Radar (InSAR). The availability of modern satellites has resulted in the routine generation of a significant amount of InSAR data. Consequently, there is a need for an automated process to detect deformation signals that appear as fringes in wrapped interferograms. Machine learning methods with transfer learning strategy have been successful in detecting these fringes [1,2], but they are limited to detecting ground deformations that have similar characteristics to the training dataset. This means that ground deformations with different characteristics from the training dataset might go undetected. Therefore, our study explores the potential of improving detection performance using semi-supervised learning [3]. In this approach, global feature representation of InSAR data is learned through unsupervised contrastive learning [4], and the detection task is performed through a fine-tuning process on a limited number of labelled samples. Specifically, the first part utilises the DetCo [5] technique with a ResNet architecture, which learns discriminative representations from global images and local patches through contrastive learning. The ResNet model is subsequently trained and used as a backbone for the Faster-RCNN [6] to perform detection. To evaluate our method, we test it on images that were missed by the supervised learning method proposed in [2].

References:

[1] N Anantrasirichai, J Biggs, F Albino, P Hill, D Bull, Application of Machine Learning to Classification of Volcanic Deformation in Routinely Generated InSAR Data Journal of Geophysical Research: Solid Earth, 2018

[2] N Anantrasirichai, J Biggs, F Albino, D Bull, A deep learning approach to detecting volcano deformation from satellite imagery using synthetic datasets, 2019

[3] T Yang, N Anantrasirichai, O Karakuş, M Allinovi, A Achim, A Semi-supervised Learning Approach for B-line Detection in Lung Ultrasound Images. IEEE International Symposium on Biomedical Imaging, 2023

[4] R. Hadsell, S. Chopra, and Y. LeCun, Dimensionality reduction by learning an invariant mapping, IEEE/CVF International Conference on Computer Vision and Pattern Recognition, 2006

[5] E. Xie, J. Ding, W. Wang, X. Zhan, H. Xu, P. Sun, Z. Li, and P. Luo, Detco: Unsupervised contrastive learning for object detection, IEEE/CVF International Conference on Computer Vision, 2021

[6] S. Ren, K. He, R. Girshick, and J. Sun, Faster R-CNN: Towards real-time object detection with region proposal networks,” Advances in neural information processing systems, 2015.

The Small BAseline Subset InSAR (SBAS-InSAR) approach gains scientific popularity in the last decades in the estimation of ground surface changes due to its high interferogram redundancy, which increases the precision of displacement estimates. The selection of interferograms to be used in the SBAS approach is conventionally calculated based on pre-defined and fixed criteria of temporal and spatial baselines. However, in many areas, various other aspects such as snow cover, vegetation growth but also high rate of displacement, influence the interferometric coherence and the quality of the unwrapped interferograms. Therefore, the conventional approach with fixed temporal and spatial baselines does not guarantee that all interferograms are bringing valuable information for SBAS inversion. Moreover, in areas of active mining with significant displacement rates, even limited decorrelation can lead to phase unwrapping errors. To address this problem, we propose a two-stage optimisation procedure to find the most appropriate SBAS-InSAR network: 1) coherence-based image selection and 2) phase unwrapping error detection using Machine Learning. This new approach was applied over a test Area of Interest (AOI) with a high deformation gradient caused by active underground mining in the Upper Silesian Coal Basin, Poland. The land cover of the AOI is characterised by mainly rural fields and sparse forests. For the SBAS analysis, a one-year stack of C-band Sentinel-1 SAR images between July 2018 and July 2019 was acquired in three various geometries involving ascending and descending geometries with relative orbit numbers 175, 51 and 124. The SBAS-network optimisation procedures were carried out with Python while the SBAS processing was carried out using the SARScape software.

In the first stage of the optimisation procedure, we set a temporal baseline threshold of 24 days and a spatial threshold equal to the value of 5% of the critical baseline. Based on these values, we calculated the initial interferometric coherence between more than 200 SAR pairs. Afterward, considering a commonly used coherence threshold of 0.2 for phase unwrapping (PhU), we calculated the percentage of pixels within the AOI that meets this coherence criterion. Then, only the pairs for which at least 80% of the pixels within the AOI have a coherence above the 0.2 threshold are used for further InSAR processing. With this approach, the SBAS network was reduced by approximately 10% (depending on the SAR datasets).

The chosen combinations of SAR images were further processed following the conventional DInSAR processing by applying the Delaunay Minimum Cost Flow unwrapping algorithm with a coherence threshold of 0.2. The visual analysis of the resulting unwrapped interferograms indicated that many of them have PhU errors, even after the reduction of unreliable pairs, done in the first stage of the optimisation. Therefore, we developed the second stage of the SBAS optimisation procedure, at which automation of the identification of the interferograms with PhU errors was created. For that step, we built a Random Forest (RF) model to automatically identify the interferograms to be removed from further SBAS processing. For this purpose, we built a feature space that consists of 16 input layers, which were calculated based on the previously generated unwrapped interferograms. To build this automatic approach as well as evaluate its performance, a visual inspection of the unwrapped interferograms was made by the user. The RF model was trained based on the descending dataset with relative orbit number 124 and evaluated by the other two datasets: descending dataset with relative orbit 51 and ascending dataset with relative orbit number 175.

After the RF model training, automatic detection of interferograms, which should be removed or preserved in further SBAS processing was carried out. Various accuracy metrics were calculated to assess the model performance. For instance, the F1-scores for the results for the ascending 175 and descending 51 datasets were found on the level of 0.85 and 0.92, respectively. Within the second stage SBAS-optimisation procedure, approximately 34-37% of the unwrapped interferograms, depending on the dataset, have been classified for removal for further processing.

Considering this, for rural areas with substantial decorrelation effects (i.e., vegetation growth, snow coverage) and significant displacement rates, aiming for the highest redundancy of the SBAS network is not always the best choice, as the huge contribution of low-quality interferograms adversely influences the SBAS estimates and also unnecessarily increases the SBAS processing time.

Although the use of InSAR to measure surface uplift is a conceptually simple task, how we can relate these instantaneous measurements to long term mountain growth remains challenging. England and Molnar (1990) defined the relationship between the surface uplift, uplift of rocks and exhumation, where exhumation is equal to rock uplift – surface uplift. They considered this relationship over geological timescales (i.e., over multiple earthquake cycles). However, to incorporate modern geodetic techniques such as GNSS and InSAR, we consider a fourth term – the Geologically Instantaneous Surface Uplift.

Taken in isolation, a GISU measurement provides no information on the exhumation rates in a region. We show, however, that when placed into the context of an orogeny deforming under tectonic equilibrium (where the long-term rock uplift rate is equal to the erosion rate), then a GISU measurement of exposed bedrock can be used as a proxy measurement for the interseismic component of exhumation. This can then be combined with coseismic displacement models to provide an estimate of the exhumation rate over an entire seismic cycle.

We take South Island, New Zealand as a study area, as it represents one of the highest straining onshore regions in the world, where oblique motion of the Pacific Plate has resulted in the formation of the Southern Alps. The central Southern Alps is an extensively studied region, where studies of river terrace uplift, seismicity, apatite and zircon fission track thermochronometry, landsliding and sediment load analysis, and vertical GNSS velocities have shown that this region is in tectonic equilibrium.

By generating 3-component velocity fields using Sentinel-1 InSAR, we measure the spatial distribution of interseismic uplift across South Island. We invert these velocities using the Geodetic Bayesian Inversion Software (GBIS) to determine the structure and slip rates of the Alpine Fault in this region. By comparing these slip rates to the measured fault slip rates, we use the slip deficit to place a lower bound of Mw 7.9 on the magnitude of earthquake that can be accommodated on the fault, and combine the resulting coseismic uplift with our GISU measurements to show the distribution of exhumation in the central Southern Alps. We show that exhumation is focused in the Whataroa region at ~ 8 mm/yr, with local maximum rates of 10—12 mm/yr, with 6 mm/yr of exhumation at the fault. We provide further evidence for a structural control on the location of the locus of exhumation, due to a shallowing of dip of the Alpine Fault caused by a bend in the deep fault.

Recent and near-future increases in the availability, resolution, and revisit frequency of coherent ground-track-repeat SAR data have seen an explosion in monitoring solutions being proposed and implemented, across a variety of applications. Natural and anthropogenic disasters such as tailings dam failures, volcanic eruptions, bridge collapses, landslides and tunneling related building damage have all been the focus of attention for follow-up studies that demonstrate the utility of InSAR as a monitoring technology. The majority of such published studies analyze the available data after the fact to identify precursory data that they posit could be used to prevent, control or otherwise limit the impact of the event. However, few studies ask the reverse question that is critical to making the technology viable for near real-time monitoring - if the data were being provided in real time to a decision maker, when would they identify a threat and therefore be able to take action?

In this work we take three case-studies of recent catastrophic events that have been studied extensively in the literature, and discuss how the presented data would appear to a decision maker in real time. We draw examples from three prominent tailings dam failures in the past 5 years (Brumadinho, Brazil; Jaegersfontein, South Africa; and Cadia, Australia). The extensive literature focus on these tailings dam failure events make them well suited for meta-analysis, but the conclusions of this study are applicable much more widely to most applications of InSAR to near real-time monitoring.

In most case studies in the existing literature, results have been presented as evidence of how InSAR can be used to provide early warning, yet what they actually show is that when we know an incident has occurred we can detect precursory displacement, or ascribe variations in the data to precursory displacement. We compare the published displacement estimates with successive displacement estimates prepared without inclusion of the future data and note that while the data appear to show significant precursory displacement when one knows where to look, the utility in a real-time monitoring situation is much less clear. In all cases, simply estimating the displacement is not sufficient to raise the alarm prior to the event. We find that both the spatial and temporal sensitivity of InSAR results is highly variable through both space and time - e.g. a distributed target coherent over only part of the timeseries versus sparse but very high quality persistent targets that are coherent through the full dataset. In order to assess whether any estimated displacement is significant relative to the noise we must have a quantitative assessment of the spatial and temporal variance and covariances within the data.

In particular, we focus on asking the question of what information could a decision maker use to take action in a near real-time setting? Many case studies focus very intently on a particular landslide or parcel of ground displacement and refine an excellent estimate of that specific location. The power of remote sensing, however, is in the monitoring of larger areas than are practical on the ground. The metric for InSAR to be useful to decision makers should instead be: was this region of displacement and/or acceleration detected with low enough false-positive and false-negative rates across the entire monitored area of interest. The acceptable false positive rate will vary by application and even within an area of interest, but in general such events are low-probability, high consequence, so the requirements for monitoring are very stringent. False negatives (undetected displacement/acceleration) generally pose a direct risk to environmental or human safety, while false positives (erroneous detections) may lead to complacency and ignoring of the detections. A quantitative assessment of the sources of error is the only way to quantify the false positive and negative rates and establish a measure of significance to the results. Such a measure of significance is essential for InSAR to be used as a near real-time monitoring data stream in a decision making environment.

Tropospheric delays (TDs), resulting from spatiotemporal variation in pressure, temperature, and humidity between SAR acquisitions, limit the efforts to obtain precise ground displacements from InSAR phase measurements. Although regional/global weather models have been exploited to reduce the delay influence on InSAR displacements, spatial resolution and temporal gap between auxiliary data and SAR acquisitions make the weather models less applicable in all cases. The phase-based methods for TDs correction are gaining popularity, but their performance cannot be guaranteed due to varying tropospheric properties across topographic divides. Meanwhile, the possible topography-related deformation signals degrade the performance of phase-elevation linear regression.

This study presents a quadtree-based joint model to simultaneously tackle the atmospheric heterogeneity and deformation-elevation coupling phenomena. Considering the spatial correlation of tropospheric property in the local area, we propose a segmentation strategy that allows a division of the interferogram into quadtree windows according to the statistics of phase variations. In each local window, we use spatial polynomials to parameterize the elevation-dependent and latent phases of TDs. The low-pass component of displacements is accounted for by a cubic function in time series. Because the TDs and the deformation signals possess distinct spatiotemporal features, they can be jointly estimated and separated in each divided window. The phase discontinuity between adjacent windows is further smoothed by integrating the corrected phase difference of the common point in the overlapping area.

The performance of the new method is verified at Bali island using a one-year-long ascending and descending Sentinel-1 SAR data sequence before a volcanic eruption on 21 November 2017. The experiment results show that the quadtree segmentation can reduce the standard deviation (STD) of the complex phase variation due to varying tropospheric properties by ~50%, as opposed to the weather model and traditional terrain-related linear correction over the whole image. A semi-simulation experiment is conducted to demonstrate the effectiveness of isolating TDs from elevation-correlated deformation signals. We further test the new method at Hawaii island, where the largest active volcano on Earth, Mauna Loa, erupted on 27 November 2022 for the first time in nearly 40 years. Through the proposed TDs correction, the misfit STD between InSAR and ground GPS displacements decreases from 25.1 mm to 4.1 mm. The corrected displacements from ascending and descending orbits illuminate consistent inflation at the summit of Mauna Loa from 2014 to 2022. The denoised deformation measurements by the new TDs method not only help for the detection of ground movement boundaries in space but also improve the retrieval of movement evolution in time.

Agricultural regions pose a challenge to InSAR displacement time series production due to abrupt transitions in land use over short spatial scales, such as at the edges of fields, and rapid temporal changes associated with different stages of the agricultural cycle, such as tilling, irrigation, and harvest. Some of these processes simply add decorrelation or a random component of the noise, with mean zero, to the time series, but some could add a bias that may either reduce or increase the apparent subsidence signal derived from such data. We analyze a full-resolution, multi-year SLC stack over California's San Joaquin Valley, an intensely cultivated region producing a wide variety of crops such as cotton, almonds, grapes, and pistachios. This region contains a well-documented subsidence signal on the order of 30 cm/yr associated with groundwater extraction, as has been recorded in numerous studies using InSAR, GPS, and ground truth measurements. Using independent information about land cover and crop type from the USDA Cropland Data Layer Program and vegetation structure inferred from NDVI analysis of Sentinel-2 optical imagery, we isolate the effects of differing land cover and land use conditions on backscatter amplitude, interferometric phase change, and interferometric coherence over space and time. We determine the statistical behavior of the phase changes associated with several key crop types by comparing the phase of pixels categorized as a given crop type to the phase values of pixels in nearby roads and developed areas. We perform our comparisons over distances of a few tens of pixels, or within 100 meters in ground range coordinates. Comparisons over short spatial scales allow us to reduce the impact on our analysis from larger-scale signal sources associated with atmospheric properties or regional subsidence associated with groundwater withdrawal. Based on the statistical characteristics of this set of crop types, we generate synthetic data that only contains the biases and noise levels from our analysis with no deformation signal. We infer the average secular rate and seasonality from this synthetic data using the same approaches that are typically applied to data from this region, including filtering, unwrapping, and downsampling. This approach allows us to quantify the contribution of land use on the inferred secular displacement rate and assess the potential bias that can occur when heterogeneous land cover is filtered and processed using standard techniques.

The country of El Salvador has suffered destructive earthquakes in the past. In 2001, two seismic events, a subduction zone earthquake followed one month later by another one at the crustal faults, caused great damage to both people and infrastructure across the region. Besides, landslides triggered by the January Mw7.7 subduction earthquake proved particularly fatal. Understanding the tectonic kinematic behaviour in detail is critical for future seismic hazard studies in the area.

El Salvador is located on an active, convergent, tectonic margin, where the Cocos plate subducts under the Chortís block of the Caribbean plate. The subduction interface is thought to be weakly coupled, with the Cocos plate advancing orthogonally towards the trench. The country is traversed by the El Salvador Fault Zone (ESFZ), comprising a set of right-lateral, strike-slip faults that run through the Central American Volcanic Arc. The Volcanic Forearc sliver, located between the ESFZ and the trench, presents a differential movement of ~12 mm/yr with respect to the Chortís Block (to the north of the ESFZ).

The long-wave, broad tectonic deformation has been constrained by past GNSS studies in the area. Nonetheless, recent GNSS campaigns have been carried out and new continuous stations have been installed. Moreover, due to the scarcity of the GNSS network, the complex behaviour of the individual faults and the intra-fault basins within the ESFZ is not yet well understood.

Here we present the first combined results of GNSS and InSAR data in El Salvador, together with preliminary results of a new, higher-resolution tectonic block model for the area.

We have processed and updated GNSS data in over 110 campaign and continuous stations in the region. We used ALOS PALSAR L-band images acquired between 2006 and 2011, in both ascending and descending tracks, to form interferograms following a Small Baseline (SBAS) approach. We computed the time series and average LOS velocity, while assessing the atmospheric effects on the signal. We used both datasets (together and independently) to build kinematic models with TDEFNODE that explain the tectonic deformation in El Salvador. We compare those results with past studies.

This work is supported by the SARAI project (Project PID2020-116540RB-C22 funded by MCIN/ AEI /10.13039/501100011033), as well as by Grant FPU19/03929 (funded by MCIN/AEI/10.13039/501100011033 and by “FSE invests in your future”).

Phase unwrapping is an essential processing step in SAR interferometry, which estimates the absolute phase from the wrapped phase within (- 𝜋, 𝜋]. Phase unwrapping is an essential data processing procedure for synthetic aperture radar interferometry. Accordingly, a lot of traditional unwrapping algorithms have been developed. Phase unwrapping is still a challenging problem in the presence of steep phase gradients and a noisy area. Recently, deep-learning-based phase unwrapping approaches have been proposed, and they show superior performance than conventional phase unwrapping algorithms. However, recent studies have not considered 1) the locally different noise, and 2) the data balance of phase gradient and noise. In addition, although, the unwrapped phase is estimated by accumulating relative phase differences between adjacent pixels from the reference point on the entire wrapped phase image, conventional model structures for semantic segmentation were adopted as it is without consideration of the phase unwrapping process. Therefore 3) the models have difficulty exploiting the phase information of the entire image together. In this study, training data and model structure were optimized for the performance enhancement of deep-learning-based phase unwrapping. For that, the training data was simulated with simple and local noise. And data augmentation was applied for balancing the phase gradient and noise level. Besides, the multi-encoder U-Net regression model structures are suggested, which have different kernels of 3X3, 5X5, and diliated 3X3. Also, the best model structure was determined by comparing the unwrapping performance according to the numbers of pooling layers and encoders. Finally, we found that optimizations of training data and model structure are a valid approach for enhancing deep-learning-based phase unwrapping. The mean absolute errors for applying suggested models, which were trained by simple and local noise, to real synthetic aperture interferograms were 0.592 and 0.445 respectively. Single-kernel model trained by local noise showed only a mean absolute error of 0.542. For the same phase data, mean absolute errors of minimum cost flow and statistical-cost, network-flow algorithm for phase unwrapping were 0.953 and 0.861 respectively. We expect that this study will contribute to designing the model structure and training data simulation approaches for the phase unwrapping, and also help to clarify earth internal processes and mechanisms.

The Okavango Rift System is an extensional tectonic structure located in northern Botswana, at the southwestern terminus of the East African Rift System. The surface expression of the tectonic deformation in this region consists in an active hemi-graben, the Okavango Graben, and a series of normal faults located in the Makgadikgadi Basin, southeast of the graben (McCarthy, 2013). Previous Global Navigation Satellite System (GNSS) based studies show extensional to dextral strike-slip displacements on both sides of the Okavango Graben with a rate of around 1 mm/yr (Pastier et al., 2017), when no displacement studies exist yet in the Makgadikgadi Basin. In order to map the ground displacement field over the whole Okavango Rift System, we analyze regional-scale Interferometric Synthetic Aperture Radar (InSAR) data produced by the ForM@Ter LArge-scale multi-Temporal Sentinel-1 InterferoMetry service (FLATSIM), developed as part of the ForM@Ter Solid Earth data and services center and supported and operated by CNES (Thollard et al., 2020). FLATSIM uses the New Small temporal and spatial BASelines (NSBAS, Doin et al., 2011; Grandin et al. 2015) algorithm to automatically compute interferograms from Sentinel-1 SAR data and invert them into displacement time series over wide areas. The products cover the period between years 2016 and 2021 with a 12-days temporal resolution on five ascending tracks covering a more than 430 000 km² area over the Okavango Graben and the Makgadikgadi Basin. Our preliminary analysis shows that the resulting signal has a strong seasonal component with a loss of coherency of the interferograms during the wet season (between November and April). By comparing the FLATSIM products with rain (IMERG data), we propose a methodology to clean the interferograms and mitigate the impact of the presence of rainy clouds on the time series analysis. We then evaluate the impact of the rain on the ground condition changes (vegetation phenology and moisture fluctuations) and on the signal using field data, Sentinel-1 Ground Range Detected SAR data and Sentinel-2 optical images. By following these approaches, we access to the spatial distribution of the annual vertical oscillations, reaching 2 cm measured at the GNSS stations and related to the flexural response of the crust to hydrological loading combining rainfall in Angola during the wet season and the flood reaching the graben during the dry season (Dauteuil et al., in press). Among those seasonal signals, we estimate the slip rate of the faults to eventually bring new insight on the propagation of the East African Rift System at its southwestern terminus.

Although InSAR is a powerful tool for measuring tectonic and anthropogenic ground deformation, many other types of signals also exist that contribute to the InSAR signal and can result in errors in InSAR-based estimates of surface displacement. One source of noise that affects InSAR data stems from fluctuations in soil moisture due to evaporation, precipitation and/or watering of agricultural fields. This soil moisture-related noise can cause cm-scale errors in interferograms and hinders our ability to constrain small-magnitude deformation signals with InSAR in places with high soil moisture variability. Soil moisture variations and subsequent downlooking and/or spatial filtering of the complex-valued phase data introduce a nonzero “triplet phase closure“ (TPC) term that has been observed in many places around the world and sometimes has a resolvable bias. The relationship between this TPC bias and the inferred underlying ground deformation signal is still poorly constrained. Many dry salt lakes/playas/salt flats around the world, which often occur in areas of active tectonic deformation, show unrealistic InSAR-derived uplift rates relative to the surrounding area. Some of this signal may be associated with soil moisture variability, but another potential contributor to this suspicious signal is evaporite crystal growth (uplift) during dry/drying out periods, followed by dissolution (subsidence) during precipitation or flooding events. Standard InSAR time series processing schemes favor coherent time periods, and decorrelated time periods tend to be ignored or minimized. At Laguna Salada, dry time periods are usually coherent, and wet wet/flooded time periods are decorrelated. Therefore, standard processing of InSAR time series may result in an erroneous extrapolation of the rates during the dry time periods to the full history of the region.

The Mexicali-Imperial Valley of southern California and northern Baja California, Mexico contains a diverse set of land cover and land use types, including agriculture, geothermal fields, fault systems, and a dry salt lake called Laguna Salada. A previous study using Sentinel-1 data from 2014-2019 inferred unrealistically high (>1 cm/yr) uplift rates within Laguna Salada. The expected surface displacement rate between the salt lake and surrounding alluvial fans region is near-zero, so we propose that TPC bias and evaporite crystal growth/dissolution are two processes that may contribute to these unrealistic uplift rates. Parsing out the contributions of these signals is challenging without ground-based observations, but using full resolution analysis on a small region can help us to assess what signals are being affected by filtering/downlooking. Phase closure is a feature of filtered or downlooked data, so if we still observe the same uplift rates when using the full resolution data, we can rule out the processes (like changes in soil moisture) that lead to non-zero phase closure and biases in the displacement rate in other regions. We use Sentinel-1 data from 2017-2022 to explore surface displacement time series of full resolution data and compare them to those using filtered data, including multiple methods of regularizing the inversion for velocity/displacement histories. We found that uplift rates remain unrealistically high within Laguna Salada, indicating that evaporite crystal precipitation and dissolution cycles are likely occurring causing real uplift and subsidence of the ground surface.

Interferometric Synthetic Aperture Radar (InSAR) is used to measure deformation rates over continents to constrain dynamic tectonic processes. InSAR measurements of ground displacement are relative, due to unknown integer ambiguities introduced during propagation of the signal through the atmosphere. However, these ambiguities mostly cancel when using spectral diversity, allowing measurements to be made with respect to a terrestrial reference frame. Such “absolute” measurements can be particularly useful for global velocity and strain rate estimation where GNSS measurements are sparse, or in specific cases where it is difficult to unwrap phase with respect to reference areas, such as volcanic islands. Furthermore, exploiting spectral diversity of overlapping regions of Sentinel-1 TOPS mode bursts gives ground displacements with a significant component of northwards motion, overcoming low sensitivity for this direction for conventional line-of-sight InSAR.

Here, we calculate along-track ground displacement velocities for a global dataset of Sentinel-1 acquisitions as processed by the COMET LiCSAR system, extending previous work primarily focused on the Asian part of the Alpine-Himalayan Belt (around 80,000 samples). Estimating along-track velocities from the azimuth subpixel offsets, including spectral diversity, we find good agreement with model values from ITRF2014 plate motion model and averaged estimates from GPS measurements, although we identify an overall offset from this data. By combining data from ascending and descending orbits we can estimate northwards and eastwards velocities over 250 x 250 km blocks, with estimated average accuracy of 4.2 and 22.8 mm/year, given as 2x median of RMSE estimates, respectively.

Application of solid-Earth tide corrections improves the average accuracy estimate of the northwards direction from 5.2 to 4.4 mm/year. Further improvement to an accuracy of 4.2 mm/year is achieved with ionospheric corrections, using gradients of ionospheric total electron content from the IRI2016 ionospheric model. This correction is strongest in near-equatorial regions and for the dusk acquisitions of ascending tracks. Finally, we evaluate that the change of precise orbit determination (POD) products definition in mid-2020 improves precision of measurements by 12% and introduces an azimuth offset of -39 mm.

This contribution will present current improvements, particularly in the ionospheric correction, and discuss findings relevant to the community. We will show results using updated global LiCSAR dataset of azimuth offsets (over 230,000 samples) and will also investigate large-scale range offsets that should help improve accuracy of the eastwards velocities.

The India-Eurasia collision has created the Tibetan plateau that exhibits a complex deformation pattern and is characterised by widespread active faulting and associated earthquakes. Particularly, one of the most intriguing observations is the clockwise rotation of southeastern Tibet around the eastern Himalayan syntaxis (EHS). Various models have been built to interpret the deformation for the region, such as lateral extrusion and rotation of blocks along major faults, or a continuum driven by gravitational spreading or ductile flow of lower crust. How best to understand the deformation field has been a subject of extensive debate.

Creeping faults slip aseismically at shallow depths and have been revealed in a variety of tectonic environments. Various temporal behaviours of creep have been observed among a few fault systems including steady-state creep, creep triggered by postseismic afterslip, quasi-periodic creep, or episodic transient creep, etc. Characterising the spatio-temporal evolution of fault creep is essential as it affects the slip budget along a fault, and hence the seismic hazard. With the increasing volume of InSAR data and improvements in data quality and processing techniques, we are able to measure surface creep with high resolution and accuracy.

The 350-km-long left-lateral Xianshuihe fault is one of the major faults in southeastern Tibet. The fault is tectonically active and considered to have substantial earthquake potential. Creeping behaviour has been reported along some sections of the fault. However, the temporal evolution of the creep is not well characterised. In this study, we use 9 years of Sentinel-1 SAR interferometry, processed by COMET-LiCSAR system, to obtain large-scale interseismic velocity and strain rate fields for southeastern Tibetan plateau. We employ a multiscale unwrapping procedure to improve unwrapping results. Unwrapped interferograms multilooked by a factor of 10 are used as a coarse estimate for the following higher resolution unwrapping step; this avoids some unwrapping errors due to isolated components. Time series inversion is performed using the LiCSBAS approach, correcting atmospheric artefacts using GACOS. We combine InSAR velocities and published GNSS data to simultaneously invert for surface velocities on a triangular mesh and reference frame adjustment parameters following the VELMAP approach. We then decompose the referenced InSAR data to east-west and vertical velocities. The strain rate fields reveal localised shear strain along the Xianshuihe and eastern Kunlun faults. Most of the region are experiencing extension, whereas the Longmen Shan thrust belt and the Jiali fault around the EHS show clear contraction. We observe continued postseismic transient associated with the 2008 Wenchuan earthquake. We explore the relationship between creep and seismic behaviour of the Xianshuihe fault. The creep rate is higher along the Kangding segment, which is likely due to postseismic relaxation of the 2014 Mw 5.9 Kangding earthquake. The 2022 Mw 6.7 Luding earthquake correlates with highly locked zones. We will investigate the temporal evolution of creep of the Xianshuihe fault. We will also examine deformation associated with other faults in the region and possible hydrological and anthropogenic factors. We discuss the implications for earthquake cycle and seismic hazard, and regional kinematics and dynamics of southeastern Tibet.

The Makran Subduction Zone (MSZ) extended east-to-west along the southern Iran and Pakistan coasts, where the oceanic portion of Arabian plate underthrusts northward beneath the Eurasia, is one of the least studied subduction zones. This is mainly due to the lack of dense and continues geodetic measurements in this area. In particular, the western Makran has received less attention due to its lower seismicity with no historical earthquake in the last 500 years compared to the eastern part in which large earthquakes has been documented. In addition to the limited seismic and geodetic data, the geometry of the Makran megathrust makes it difficult to be monitored by satellite InSAR data. The east-west elongation of the megathrust results in the main interseismic deformation component in the south-north direction, the least sensitive deformation component for InSAR. Consequently, in InSAR data, we expect very low amplitude with long wavelength interseismic deformation signal. To isolate and extract such a signal, long timeseries of data are required. With Sentinel-1 data, more than seven years continuous SAR data is now available over Makran for the first time. Here, by a sensitivity analysis, we show that the length of this timeseries with the given satellite geometry of Sentinel-1 in both ascending and descending orbits is sufficient to isolate and estimate the interseismic strain accumulation associated with plate coupling on the west Makran megathrust. This is provided that a proper atmospheric mitigation to be applied on the data. In this study, we design and apply an efficient atmospheric mitigation following by series of other corrections (e.g., removing non-tectonic local processes, correction for the reference frame motion) on the sentinel-1 data. The input interferograms has been obtained from the operating system: Looking Into Continents from Space with Synthetic Aperture Radar (LiCSAR). The LiCSAR products covering western Makran were selected from two frames, including ascending and descending passes. In order to isolate interseismic deformation signal, we employ a time series analysis method with focusing on estimating and filtering atmospheric effects from interferograms. Subsequently, the interseismic rates estimated by this method are inverted to assess the magnitude and down-dip extent of plate coupling along different trench-perpendicular profiles. The results reveal important characteristics about the kinematics of the plate coupling on the western Makran megathrust. The obtained results are helpful for further quantitative assessments of seismic and tsunami hazards in this area.

Volcano deformation happens across many orders of magnitude in terms of duration (seconds-centuries), deformation rate (mm/yr – meter/day), and deformation area (50 m - >100 km) and can be a precursor to volcanic eruptions. Currently, over half of the world’s active volcanoes do not have sufficient ground monitoring instruments, thus differential Interferometric Synthetic Aperture Radar (InSAR) has become an integral tool to measure and monitor displacement at volcanoes. However, InSAR is unable to observe large surface displacement gradients exceeding 1 fringe per pixel due to loss of coherence. This makes InSAR unsuitable for monitoring rapid, high-magnitude, small-footprint deformation that can occur before and during volcanic crises.

Pixel Offset Tracking (POT) – measuring displacements between matching pixels in the SAR intensity data – is often used for measuring rapid movement (e.g. glaciers, large magnitude earthquakes, mining subsidence) as it allows us to use large baseline image pairs and measure displacement gradients beyond the limits of InSAR. The precision and resolving power of POT are primarily dependent on the pixel size of the SAR data, with a detection limit and precision ranging from 1/10th – 1/30th of the pixel dimensions, depending on the acquisition and surface conditions. Time series POT processing (similar to Small Baseline InSAR time series processing) has been shown to reduce this limit to 1/50th of the pixel dimensions. Therefore, high-spatial-resolution data (≤ 1 m) like those from COSMO-Skymed (CSK), TerraSAR-X (TSX), and PAZ, should allow for the detection of large-magnitude displacements with centimetric accuracy. POT has the additional benefit of measuring displacement along both slant range and azimuth directions, thus only requiring 2 viewing geometries to reconstruct 3D surface displacement.

We explore methods that incorporate multi-look interferograms, high-spatial-resolution interferograms, and (high-spatial-resolution) POT from (staring) spotlight CSK, TSX, and PAZ data to accurately measure rapid, large-magnitude displacements at volcanoes. We test our method on Merapi volcano, which experienced complex meter-scale displacement near its summit leading up to and during the 2021-present lava-dome-building eruption. We find that high-resolution InSAR only performs well with a sufficiently small baseline (≤200 m) and only over areas where the displacement gradient is small (<< 1 fringe per pixel). We find that high-resolution POT (azimuth resolution 17 cm, slant range resolution 45 cm) can reliably measure large displacements (≥ 1 m) with a standard deviation of ≤ 1.3 cm over the old lava domes/lava flows surrounding the summit but is less accurate over vegetated areas (standard deviation of ≥ 5 cm). Combining ascending and descending azimuth and range offsets should allow us to estimate the full 3-D motion, but ascending acquisitions have severe foreshortening or overlay on the deforming western flank. Nevertheless, the sense of motion observed in the ascending acquisitions allows us to determine that motion of the old lava flows is directed approximately down-slope. Time series processing of the descending data reveals displacement occurs in a start-stop-like manner on some of the old lava flows. The start-stop displacement could indicate a response to time-varying magmatic processes (e.g. intrusion/extrusion rate, lava dome loading) or show stick-slip behaviour.

We conclude that high-resolution InSAR–POT is able to measure rapid, small footprint displacement to (sub-)cm precision even near the summit of steep-sided stratovolcanoes. It is, therefore, a useful tool for the detection and monitoring of flank instability, especially when no or limited ground monitoring is present.

In magma-rich rift systems, how propagating rifts interact and what is the role of magma during rift linkage remains a matter of debate. In the Afar depression of Ethiopia, the plate divergence between Nubia, Arabia and Somalia plates has lead to the formation of a series of rift segments that currently accommodate extension through magmatic activity and faulting. In particular the Central Afar is a 250-by-100 km zone where the deformation links through overlapping grabens. Here, several studies proposed that rift linkage occurs either by means of ‘bookshelf’ faulting or a combination of extension and shear. However, the contribution of magma has never been addressed. To study the kinematics of Central Afar, we formed a series of interferograms, using the InSAR Scientific Computing Environment (ISCE) software package and Sentinel-1 acquisitions spanning the 2014-2021 period. We processed ascending (track 014) and descending (track 006) interferograms by stitching three adjacent frames together and getting spatially continuous phase observations across the region at a 30 m resolution. We selected interferometric pairs by adopting the small baselines approach, yet excluding 6 and 12 days interferograms to avoid short-term phase biases. We also excluded noisy interferograms and created two final datasets of 104 and 151 interferograms for ascending and descending tracks, respectively, with temporal baselines between 24 and 144 days. We then calculated time-series of cumulative LOS displacement and maps of average LOS velocity in both ascending and descending orbits and covering the entire Central Afar zone, using the pi-rate software. Finally, we jointly inverted the ascending and descending average LOS velocities and GNSS measurements available in literature to obtain the 3D velocity field with the aid of a regular triangular mesh at high spatial resolution, 3 km-spacing. Our new high spatial resolution 3D velocity map of Central Afar shows how horizontal and vertical deformation is accommodate across the study area. In particular the plate boundary extension, previously considered as distributed over the entire Central Afar zone, is instead accommodated discretely in the single overlapping grabens where we observe clear velocities increases. Such increase occur in correspondence of major tectonic structures. We also observed vertical focused deformation which is interpreted as induced by magma ponding at depth.

remotIO is a modular service ecosystem for performing a variety of InSAR engineering tasks with operational use of InSAR geodesy. The system is based on a geodetic estimation theory approach [1] and offers a validation framework for measurements with artificial radar reflectors [2] and integration with ground-infrastructure for quality-control. remotIO functions are split into InSAR processing module, geodetic quality control toolboxes, data mining module and web-based visualisation platform [3] for easy access to results and for continuous monitoring tasks.

The system offers auxiliary toolboxes for geodetic quality control. The first toolbox facilitates a standard procedure to design the network of artificial radar reflectors and analyse their radar cross section, signal-to-clutter ratio and displacement time-series. The second toolbox is aimed at post-processing InSAR results with data-mining and machine-learning techniques to classify time-series, detect outliers, mitigate systematic errors and filter the final deformation maps [3].

Artificial radar reflectors are successfully employed in Slovakia for InSAR accuracy validation, geodetic positioning improvement and absolute geodetic referencing of displacement time-series. In this work, we characterise the first experiences building geodetic InSAR-GNSS collocation stations and results from the operational use of remotIO system in different monitoring scenarios (landslides, dams, mining subsidence, urban areas) are presented.

The ultimate goal of remotIO is to build an integrated service ecosystem for improved structural stability monitoring while keeping the system flexible and allowing for custom setup per customer, so modules could be replaced and services tailored.

[1] van Leijen, F. (2014), Persistent Scatterer Interferometry based on geodetic estimation theory, Delft University of Technology.

[2] Czikhardt, R.; van der Marel, H.; Papco, J. GECORIS: An Open-Source Toolbox for Analyzing Time Series of Corner Reflectors in InSAR Geodesy. Remote Sens. 2021, 13, 926. https://doi.org/10.3390/rs13050926

[3] Bakon, M.; Czikhardt, R.; Papco, J.; Barlak, J.; Rovnak, M.; Adamisin, P.; Perissin, D. remotIO: A Sentinel-1 Multi-Temporal InSAR Infrastructure Monitoring Service with Automatic Updates and Data Mining Capabilities. Remote Sens. 2020, 12, 1892. https://doi.org/10.3390/rs12111892

The grounding line positions of Antarctic glaciers are needed as an important parameter to assess ice dynamics and mass balance in order to record the effects of climate change to the ice sheets as well as to identify the driving mechanisms for these. In order to address this need, ESA’s Climate Change Initiative (CCI) produced interferometric grounding line positions as ECV for the Antarctic Ice Sheet (AIS) in key areas. Additionally, DLR’s Polar Monitor project focuses on the generation of a near complete circum-Antarctic grounding line. Until now these datasets have been derived from interferometric acquisitions of ERS, TerrasSAR-X and Sentinel-1. Especially for some of the faster glaciers, the only available InSAR observations of the grounding line have been acquired during the ERS Tandem phases (1991/92, 1994 and 1995/96).

In May 2021, a joint DLR-INTA Scientific Announcement of Opportunity was released which offers the possibility of a joint scientific evaluation of SAR acquisitions of the German TerraSAR-X/TanDEM-X and the Spanish PAZ satellite missions. These satellites are almost identical and are operated together in a constellation therefore offering the possibility of combining their acquisitions to SAR interferograms.

The present study harnesses the interferometric capability of joint TSX and PAZ acquisitions in order to reduce the temporal decorrelation between acquisitions. The revisit times are reduced from 6 days (Sentinel-1 A/B) or 11 days (TSX) to 4 days (TSX-PAZ). Together, the higher spatial resolution than Sentinel-1 and the reduced temporal baseline allows imaging the grounding line at important glaciers and ice streams where the fast ice flow causes strong deformation. These are often the glaciers where substantial grounding line migration has taken place or is suspected (e.g Amundsen Sea Sector) but where current available SAR constellations cannot preserve enough interferometric coherence to image the grounding line. The potential of short temporal baselines was already shown with data from the ERS Tandem phases in the AIS_cci GLL product and more recently but only in dedicated areas with the COSMO-SkyMed constellation [Brancato, V. et al. 2020, Milillo, P. et al. 2019]. In some fast-flowing regions, InSAR grounding lines could not be updated since.

For the derivation of the InSAR grounding line, 2 interferograms (PAZ-TSX) with a temporal baseline of 4-days will be formed. It is not necessary, that the acquisitions for the two interferograms fall in consecutive cycles but is advantageous to acquire the data with limited overall temporal separation to be able to assume constant ice velocity. The ice streams where potential GLLs should be generated were identified with focus on glaciers in the Amundsen Sea Sector (e.g. Thwaites Glacier, Pine Island Glacier) but also glaciers in East Antarctica (e.g. Totten, Lambert, Denman). Besides filling spatial or temporal gaps in the circum-Antarctic grounding line, the resulting interferograms will also be used for sensor cross-comparison to Sentinel-1-based grounding lines in areas where both constellations preserve sufficient coherence.

Brancato, V., E. Rignot, P. Milillo, M. Morlighem, J. Mouginot, L. An, B. Scheuchl, u. a. „Grounding Line Retreat of Denman Glacier, East Antarctica, Measured With COSMO-SkyMed Radar Interferometry Data“. Geophysical Research Letters 47, Nr. 7 (2020): e2019GL086291. https://doi.org/10.1029/2019GL086291.

Milillo, Pietro, Eric Rignot, Paola Rizzoli, Bernd Scheuchl, Jérémie Mouginot, J. Bueso-Bello, und P. Prats-Iraola. „Heterogeneous Retreat and Ice Melt of Thwaites Glacier, West Antarctica“. Science Advances 5, Nr. 1 (1. Januar 2019): eaau3433. https://doi.org/10.1126/sciadv.aau3433

Global warming due to greenhouse gas emitted into the atmosphere is triggering a climate crisis, the impacts of which can already be felt in current times with more frequent extreme weather events such as flooding, heatwaves or wildfires. Another consequence of the global warming is the rise of the sea level (SLR). The SLR amplitude will depend on the Representative Concentration Pathway (RCP) emission scenario we will follow. It is thus estimated for 2100 between 0.29 m and 0.59 m for a low emission scenario (RCP 2.6) or between 0.6 m and 1.1 m for a high emission scenario (RCP 8.5). Even if the mean Earth temperature increase is kept below 2°C (compared to the pre-industrial period) within the next decades, sea level will continue to rise for several centuries or more due to the system inertia. This estimation is worrying as coastal area can be tremendously biodiverse and host a substantial part of the world population and many critical infrastructures.

However, sea rise is just one factor in the relative sea level changes and vertical ground motions can significantly amplify or reduce the effect of the global SLR. Indeed, sinking ground along the shoreline greatly magnifies the effects of sea level rise because both processes work together to worsen the situation. Indeed, uplift or subsidence along the coast are generated either by natural phenomena (sediment compaction, global isostatic adjustment, or tectonics) or by human activities (ground water/hydrocarbon extraction, or land reclamation).

In this study, we investigate the vertical movements of the Nice-Côte d'Azur airport that has been built on reclaimed land over a narrow coastal shelf (1-2 km wide) in the Var river delta (French Riviera, France). This critical economical infrastructure has been a permanent concern since the partial collapse of the platform in 1979 that caused the death of 11 people. Although engineers and workers managed to stabilize the runways and finished the construction in early 1980s, Envisat InSAR measurement revealed in a previous study the on-going subsidence of the airport.

Here, we process 28 years of SAR data from three satellite generations (ERS, Envisat, and Sentinel-1) to comprehensively monitor the dynamics of the airport subsidence. We observe that the spatial displacement pattern is steady through the whole observation. However, the maximum downward motion rate is slowing down from 16 mm/yr in the 1990s to 8 mm/yr today. We thus observe a deceleration of 50% of the subsidence rates over 28 years, revealing a transient non-linear deformation that is expected for ground layer compaction. Actually, soils and rocks can exhibit creep behavior, which is the development of time-dependent strains at a state of constant effective stress. Creep behavior influences the long-term stability of grounds and movement of slopes. This time-dependent material behavior exhibits viscoelastic or viscoplastic characteristics that can be reproduced with different creep models of increasing complexity depending on the type of material and loading conditions (Jaeger and Cook, 1979). Several constitutive laws have been introduced in the past to study creep and this still is an active field of research in the rock physics labs and geophysical field studies.

We used, thus, a simple analytical Burger’s creep model to constrain the mechanisms and rheology at play. The data are properly explained by the primary and secondary creep phases, highlighting a slow viscoelastic deformation at multiyear timescales. Although the subsidence rate decelerates, at least for 28 years, our results show that the compaction of the sediment is still active and its future evolution is uncertain and still at stake. Indeed, if compaction zones are developing under the airport platform, creep process could potentially lead to accumulated material damage toward failure.

Our study demonstrates the importance of remotely monitoring of the platform to better understand coastal land motions, which will ultimately help evaluate and reduce associated hazards.

Multi-temporal SAR interferometry (MTInSAR), by providing both mean displacement maps and displacement time series over coherent objects on the Earth’s surface, allows analysing wide areas, identifying ground displacements, and studying the phenomenon evolution on long time scales. This technique has also been proven to be very useful for detecting and monitoring instabilities affecting both terrain slopes and man-made objects. In this contest, an automatic and reliable characterization of MTInSAR displacements trends is of particular relevance as pivotal for the detection of warning signals related to pre-failure of natural and artificial structures. Warning signals are typically characterised by high rates and non-linear kinematics, so reliable monitoring and early warning require a detailed analysis of the displacement time series looking for specific trends. However, this detailed analysis is often hindered by the large number of coherent targets (up to millions) required to be inspected by expert users to recognize different signal components and also possible artifacts affecting the MTInSAR products, such as, for instance, those related to phase unwrapping errors.

This work concerns the development of methods able to fully exploit the content of MTInSAR products, by automatically identifying relevant changes in displacement time series and to classify the targets on the ground according to their kinematic regime. We introduced a new statistical test based on the Fisher distribution with the aim of evaluating the reliability of a parametric displacement model fit with a determined statistical confidence [1]. We also proposed a new set of rules based on the statistical characterization of displacement time series, which allows different polynomial approximations for MTInSAR time series to be ranked. The method was applied to model warning signals. Moreover, in order to measure the degree of regularity of a given time series, an innovative index was introduced based on the fuzzy entropy, which basically evaluates the gain in information by comparing signal segments of different lengths [2]. This fuzzy entropy index, without postulating any a priori model, allows highlighting time series which show interesting trends, including strong non linearities, jumps related to phase unwrapping errors, and the so-called partially coherent scatterers.

The work introduces the theoretical formulation of these two selection procedures and show their performances as evaluated by simulating time series with different characteristics in terms of kinematic (stepwise linear with different breakpoints and velocities), level of noise, signal length and temporal sampling. The proposed procedures were also experimented on real MTInSAR datasets. We show results obtained by processing both Sentinel-1 and COSMO-SkyMed datasets acquired over Southern Italian Apennine (Basilicata region), in an area where several landslides occurred in the recent past [3]. The MTInSAR displacement time series were analysed by using the proposed methods, searching for nonlinear trends that are possibly related to relevant ground instabilities and, in particular, to potential early warning signals for the landslide events. The index based on the fuzzy entropy was able to recognize coherent targets affected by phase unwrapping errors, which should be corrected to provide reliable displacement time series to be further analyzed. The procedure based on the Fisher distribution was used for classifying targets according to the optimal degree of a polynomial function describing the displacement trend. This allowed to select targets showing nonlinear displacement trends related to the several ground and structure instabilities.

Specifically, the work presents an example of slope pre-failure monitoring on Pomarico landslide, an example of slope post-failure monitoring on Montescaglioso landslide, and few examples of structures (such as buildings and roads) affected by instability related to different causes. Our analysis performed on COSMO-SkyMed MTInSAR products over Pomarico was able to capture the building deformations preceding the landslide and the collapse. This allows the understanding of the phenomenon evolution, highlighting a change in velocities that occurred two years before the collapse. This variation probably influenced the dynamics of the landslide leading to the collapse of an area considered to be at a medium-risk level by the regional landslide risk map. Results from the analysis performed on Sentinel-1 MTInSAR products were instead useful to identify post-failure signals within the Montescaglioso landslide body. The selected trends confirm the stability of the landslide area with some local displacements due to restoration works. In this case, the value of the MTInSAR displacement time series analysis emerges in the assessment phase of post-landslide stability, resulting in a useful support tool in the planning of safety measures in landslide areas.

References

[1] Bovenga, F.; Pasquariello, G.; Refice, A. Statistically‐based trend analysis of MTInSARdisplacement time series. Remote Sens. 2021, doi:10.3390/rs13122302.

[2] Refice, A.; Pasquariello, G.; Bovenga, F. Model-Free Characterization of SAR MTI Time Series. IEEE Geosci. Remote Sens. Lett. 2020, doi:10.1109/lgrs.2020.3031655.

[3] Bovenga, F.; Argentiero, I.; Refice, A.; Nutricato, R.; Nitti, D.O.; Pasquariello, G.; Spilotro, G. Assessing the Potential of Long, Multi-Temporal SAR Interferometry Time Series for Slope Instability Monitoring: Two Case Studies in Southern Italy. Remote Sensing, 2022, 14(7): 1677, 2022, doi.org/10.3390/rs14071677 2021

Acknowledgments

This work was supported in part by the Italian Ministry of Education, University and Research, D.D. 2261 del 6.9.2018, Programma Operativo Nazionale Ricerca e Innovazione (PON R&I) 2014–2020 under Project OT4CLIMA; and in part by Regione Puglia, POR Puglia FESR-FSE 204-2020 - Asse I - Azione 1.6 under Project DECiSION (p.n. BQS5153).

Synthetic aperture radar interferometry (InSAR) offers a cost-effective and accurate way to study the deformation dynamics of surfaces (Fernández-Torres et al., 2020; Ferretti et al., 2001; Gabriel et al., 1989). InSAR has been used extensively to analyze land deformation resulting from seismological, volcanological, soil and geologic factors, as well as anthropogenic factors such as water withdrawal and construction. Although there are many case studies on this topic worldwide, research on Central American countries is scarce. Therefore, Guatemala City is an excellent candidate for remote sensing techniques, particularly InSAR. Guatemala City is situated in a highly seismic area and has been affected by many destructive earthquakes in the past (Lang et al., 2009).The study area includes important faults and volcanic features, such as the Mixco and Pinula Fault structures and the Santiaguito, Fuego, and Pacaya volcanoes (Pérez, 2009).

The Department of Guatemala, which encompasses Guatemala City and 16 other municipalities, is home to 20.2% of the country's population (Instituto Nacional de Estadística Guatemala, 2019). As groundwater population grows, the exploitation has increased, causing water levels to drop in some well fields (Herrera Ibáñez, 2018). Studies have shown that one of the main factors leading to land subsidence is water pumping for urban and agricultural use .(Chaussard et al., 2014; Engi, 1985; Koudogbo et al., 2012; Normand & Heggy, 2015; Zhu et al., 2015)

In this analysis, 226 synthetic aperture radar (SAR) images from both satellites of the Sentinel-1 constellation (A and B) were used, for the period of time between January 2017 and September 2021. Persistent Scatterers were generated using the Stanford Method for Persistent Scatterers (StaMPS) software (Foumelis et al., 2018). SAR images were preprocessed using the SeNtinel Application Platform(SNAP) developed by the European Space Agency. GPS GUAT (Blewitt et al., 2018) was selected as a reference point.

A total of 580,872 persistent scatterers were obtained for the ascending geometry and 360,828 for the descending geometry, along with the deformation time-series. The decomposition in vertical and east-west deformation was calculated for 211,455 points. The results allow identifying eight “hotspot” areas with subsidence velocity values between 5 and 16 mm/year, indicating clear subsidence processes during the study period.

The preliminary results identify the location of these areas affected by subsidence and quantify their evolution in the period analysed. These results have revealed that 11.60% (2,651 hectares) of the urbanized area within the study area experienced deformations greater than 5 mm/year, reaching up to 11 mm/yr in some locations. Administrative zones (neighborhoods) 4, 5, 8, and 9 had more than half of their surface area affected by subsidences, whose velocities are over 5 mm/yr.

Bilbiography

Blewitt, G., Hammond, W., & Kreemer, C. (2018). Harnessing the GPS Data Explosion for Interdisciplinary Science. Eos, 99. https://doi.org/10.1029/2018EO104623

Chaussard, E., Bürgmann, R., Shirzaei, M., Fielding, E. J., & Baker, B. (2014). Predictability of hydraulic head changes and characterization of aquifer‐system and.pdf. https://doi.org/10.1002/2014JB011266

Engi, D. (1985). Subsidence Due to Fluis Withdrawal: A Survey of Analytical Capabilities (p. 114). Sandia National Laboratories.

Fernández-Torres, E., Cabral-Cano, E., Solano-Rojas, D., Havazli, E., & Salazar-Tlaczani, L. (2020). Land Subsidence risk maps and InSAR based angular distortion structural vulnerability assessment: An example in Mexico City. Proceedings of the International Association of Hydrological Sciences, 382, 583-587. https://doi.org/10.5194/piahs-382-583-2020

Ferretti, A., Prati, C., & Rocca, F. (2001). Permanent Scatterers in SAR Interferometry. IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, 39(1), 13. https://doi.org/0196–2892/01

Gabriel, A. K., Goldstein, R. M., & Zebker, H. A. (1989). Mapping small elevation changes over large areas: Differential radar interferometry. Journal of Geophysical Research, 94(B7), 9183. https://doi.org/10.1029/JB094iB07p09183

Herrera Ibáñez, I. rodolfo. (2018). Sobreextracción de las aguas subterráneas en la cuenca norte de la ciudad de Guatemala. XXXVI(2).

Instituto Nacional de Estadística Guatemala. (2019). XII Censo Nacional de población y VII de vivienda. INE.

Koudogbo, F. N., Duro, J., Arnaud, A., Bally, P., Abidin, H. Z., & Andreas, H. (2012). Combined X- and L-band PSI analyses for assessment of land subsidence in Jakarta (C. M. U. Neale & A. Maltese, Eds.; p. 853107). https://doi.org/10.1117/12.974821

Lang, D. H., Sergio, M., Crempien, J., & Erduran, E. (2009). Reducción de Resgo Sísmico en Guatemala, El Salvador y Nicaragua con Cooperación regional a Honduras, Cota Rica y Panamá.

Normand, J. C. L., & Heggy, E. (2015). InSAR Assessment of Surface Deformations in Urban Coastal Terrains Associated With Groundwater Dynamics. IEEE Transactions on Geoscience and Remote Sensing, 53(12), 6356-6371. https://doi.org/10.1109/TGRS.2015.2437368

Pérez, C. L. (2009). Estructura geológica del Valle de la Ciudad de Guatemala interpretada mediante un modelo de cuenca por distensión. Revista Geológica de América Central, 41. https://doi.org/10.15517/rgac.v0i41.4179

Zhu, L., Gong, H., Li, X., Wang, R., Chen, B., Dai, Z., & Teatini, P. (2015). Land subsidence due to groundwater withdrawal in the northern Beijing plain, China. Engineering Geology, 193, 243-255. https://doi.org/10.1016/j.enggeo.2015.04.020

The US Atlantic coastal communities, due to their low-lying elevation, large population density, and high economic importance, are highly susceptible to coastal flooding hazards. Over the last decade, Southeast Florida's coastal communities have experienced a significant surge in coastal flooding events, leading to severe harm to the environment, economy, and society. Coastal subsidence is a crucial factor in amplifying the coastal flooding hazard by decreasing the coast's elevation compared to sea level rise. Therefore, monitoring coastal subsidence is vital in developing necessary mitigation measures and improving coastal flooding hazards. The objective of this study is to monitor coastal subsidence in Southeast Florida, identify the factors contributing to it, and evaluate its impact on the increased coastal flooding hazard. We used two geodetic techniques, interferometric synthetic aperture radar (InSAR) and global navigation satellite system (GNSS), to investigate coastal subsidence. We carried out a time-series analysis on InSAR observations from Sentinel-1 data provided by the European Space Agency (ESA) to generate a vertical land motion (VLM) map at a spatial resolution of 50m. Our initial findings for the observation period of 2016-2022 showed that most of the Southeast Florida region is stable, with localized subsidence occurring in a few areas at a rate of 3-5 mm/year. We also compared the observed InSAR vertical displacement rate with the GNSS dataset provided by the Nevada Geodetic Laboratory, and the comparison revealed good agreement between the two datasets, indicating the reliability of the InSAR results. Overall, our study suggests that although the contribution of local land subsidence is limited to small regions along the Southeast Florida coast, within these regions, the risk of coastal flooding is significantly higher than in non-subsiding regions.

The Synthetic Aperture Radar Interferometry (InSAR) technique can quickly obtain millimeter-level surface deformation in urban areas with high coherence. However, expanding the application of time series InSAR in non-urban areas is an important research focus. An improved SBAS-InSAR analysis approach is applied in this study to present the surface displacement of highways under construction. The density and accuracy of Point-like targets are improved by a foreground-background scattering-based PTs identification method. Taking the Kejiao Highway in the Shenzhen-Shantou Special Cooperation Zone as an example, the deformation along the highway under construction and the surrounding ground objects is revealed.

The Synthetic Aperture Radar Interferometry (InSAR) technique can quickly obtain millimeter-level surface deformation in urban areas with high coherence [1-3]. However, expanding the application of time series InSAR in non-urban areas is an important research focus. Summarizing the current research progress, the current problems lie in the accurate identification and integration of structural PTs in non-urban areas, and detailed deformation analysis of different areas around under-constructing highways [4-6]. Firstly, the coherence of highways under construction in non-urban areas is influenced by continuous construction and complex non-urban environment, making it difficult to select dense and accurate point-like targets (PTs) along the highway structure. Secondly, the previous studies always ignore the environment-structure coupling analysis, leaving the detailed deformation analysis of different highway construction periods still unclear.

An improved SBAS-InSAR analysis approach is applied in this study to present the surface displacements along the Kejiao Highway in the Shenzhen Shantou Special Cooperation Zone under construction. The density and accuracy of Point-like targets are improved by a foreground-background scattering-based PTs identification method. The results show that the settlement rate of the spoil ground along the highway generally reached -40 ~ -60mm/yr. The surrounding artificial slope and building zone are generally lifted after soil backfill, while the bare soil and foundation pit showed more serious settlement. We also interpret the mechanism behind the different surface displacements of different ground objects by combining time series displacement and local data. The analysis shows that the difference in displacement rate is the result of the comprehensive influence of many factors, such as temperature, rainfall, ground property, construction technology, and formation time.

The time-series InSAR deformation monitoring results revealed by the traditional InSAR method and our method are shown in Fig. 1. It can be seen that the PTs in Fig.1(a) are just distributed upon some sparse artificial buildings. While, by analyzing the foreground-background scattering characteristics of the highway, as well as adapting the interferometry combination according to the number of temporal-coherent points, the number of PTs selected by our method has been significantly increased especially along the highway (as shown in Fig. 1(b)), which will support a more reliable deformation analysis and interpretation.

According to Fig. 1(b), the deformation exactly upon the highway is small, however, two serious subsidence areas, including a spoil ground and a slope (shown in Fig.2 and Fig. 4, respectively), with subsidence velocities of about -60 mm/yr along the highway are identified.

One of the most serious subsidence areas is a large soil ground on the north part of the highway, which is shown as the red rectangle in the left picture of Fig. 2. Comparing the deformation distribution map and the Google Map, the deformation of the buildings is relatively stable, which is within -10 and 10 mm/yr. However, a serious subsidence area is identified in the south of the buildings, which is spoiled ground. The subsidence velocity of the spoil ground is about -60mm/yr, which would threaten the stability of the highway and its surrounding buildings. Therefore, it is worth further attention.

Based on the above deformation velocity analysis, we further calculate the time-series displacement of the spoil ground, as shown in Fig. 3. It can be seen that the cumulative subsidence during the observation time is about 100 mm. Considering the continuous subsidence of the spoil round, remedial measures such as soil backfilling are conducted in September 2021 and July 2022, which are shown in the orange and green rectangles in Fig. 2, respectively. According to the time-series displacements, the subsidence has temporarily slowed down after the soil backfilling. However, due to the continued construction of the highway, the subsidence of the spoiled ground increased again.

Another serious subsidence area is a slope along the highway as expressed in Fig. 4 (the red rectangle in the left picture). As for the slope, the maximum subsidence velocity also reaches up to -60 mm/yr. The subsidence mainly occurred on the east side of the highway. According to the survey, landslides have occurred here. Therefore, slope maintenance has been conducted during the observation period to guarantee construction safety.

Based on the above deformation velocity analysis, the time-series displacements of the slope are calculated and expressed in Fig. 5. The accumulative subsidence in this area from January 2021 to October 2022 is about 100 mm, which is worth further monitoring. Moreover, during the two maintenance period, slight uplifts have been observed (see the deformation near the orange and green rectangle), which indicate that the maintenance has, to a certain extent, mitigated the settlement trend. However, such maintenance didn’t show a long-term effect on the subsidence caused by the highway construction. Therefore, the deformation of this slope still needs more attention.

The uncertainty estimated for the line-of-sight (LOS), vertical and horizontal (E-W) velocities from the InSAR measurements is useful information for practical applications. For example, the comparison of terrestrial and InSAR measurements needs the uncertainties of both components to test the statistical significance/non-significance of differences between measurement results. I review the approach by following the propagation of uncertainty (JCGM, 2011) to estimate the uncertainties of vertical and horizontal components from the InSAR measured LOS uncertainties. As an example, the vertical stability of benchmarks in Tallinn city center were evaluated with the help of repeated leveling (in 2007-2019) and multi-temporal InSAR analysis (in 2016-2022) of Sentinel-1 data. The comparison of long-term vertical velocities at 116 benchmarks of Tallinn height network has shown that the differences between leveled and InSAR results were statistically significant (within 2σ confidence interval) only for the 10% of benchmarks. Thus a good agreement between leveled and InSAR derived vertical displacements can be concluded. Furthermore, it illustrates the high efficiency of InSAR measurement technique in monitoring the geodetic infrastructure in urban environment.

References

JCGM. (2011). Evaluation of measurement data. Supplement 2 to the “Guide to the expression of uncertainty in measurement”, Joint Committee for Guides in Metrology (JCGM) 102:2011.

The Vadomojón reservoir is located between the municipalities of Baena (Córdoba) and Alcaudete (Jaén), southern Spain, and constitutes an environment of special importance for the neighboring regions. This reservoir belongs to the Guadalquivir Hydrographic Confederation and has a capacity of 163 hm³. It occupies an area of 782 ha, making it one of the most significant reservoirs in the Guadalquivir basin. Due to the availability of data, it is proposed as a pilot case study for the project SIAGUA which is mainly devoted to the development of a new generation of surveillance systems for water cycle infrastructures. These systems will integrate satellite data, in-situ monitoring, and expert judgment.

Many dam managers have access to diverse information about these infrastructures derived from in-situ topographic surveys, analyses, and measurements from geotechnical and hydraulic sensors, storage volume, etc., which they use on a daily basis in their inspection and maintenance tasks. These tasks can be less efficient at times if all this information is not interconnected. Furthermore, MT-InSAR techniques provide another valuable source of data for monitoring infrastructure movements and adjacent areas, often not used by dam managers, and even less integrated with the rest of the dam information. Therefore, our proposal for integrating all this information is presented in the following steps. First, the documentation database is created for the selected dam. The documents, plans, and reports of the dam starting from its design and construction are collected, classified, and integrated into a common general dam database. Secondly, we proceed with the integration of monitoring records from MT-InSAR, geotechnical and hydraulic instrumentation, and geodetic in-situ surveys. This task requires the standardization of data with a common temporal origin and the selection and identification of measuring points, including persistent scatterers detected with MT-InSAR (PSI), dam instrumentation, and topographic references. The third phase is the cross-validation of MT-InSAR, geotechnical, and geodetic records through GIS analysis and geostatistics.

Finally, we implement an integrated monitoring system that includes the interpretation of monitoring variables for managers. The outcome is displayed in an accessible web platform linked to the main database through an API that includes many tools designed for the convenient handling of all the data.

Rapid land subsidence accelerates relative sea-level rise and can expose larger land areas and populations to significant risks of flooding and extreme weather events. Land subsidence has been commonly observed at rates over tens of millimetres per year in localized parts of coastal cities – an order of magnitude faster than other major factors of relative sea-level rise such as ocean mass and thermal expansion, and glacial isostatic adjustment. However, land subsidence effects are not well considered in global relative sea-level assessments due to the high spatial variability and a lack of data that is comparable across cities and regions. Globally consistent data are mostly based off point measurements from Global Navigation Satellite System and tide gauge networks which do not capture local variabilities in land subsidence. On the other hand, spatially continuous measurements such as from Interferometric Synthetic Aperture Radar (InSAR) are mostly limited to local or regional settings where a disparity of processing techniques have been used across studies. This warrants the need for large-scale and accurate monitoring of land subsidence.

Here, we provide self-consistent, high spatial resolution land subsidence rates with coverage of the 48 largest coastal cities, representing 20% of the global urban population. The rates are derived at 90 m pixel spacing using C-band Sentinel-1 data from a single look direction between 2014 and 2020 for each coastal city. We employ a standardized, semi-automated processing workflow using the Advanced Rapid Imaging and Analysis system for interferogram generation and the Miami INsar Time-series software in Python for Small BAseline Subset time series analysis. Spatial data gaps due to decorrelation are filled with kriging, where rates with lower temporal uncertainty are given higher weights during kriging. We show that cities experiencing the fastest land subsidence are concentrated in Asia. The fastest peak rate of subsidence is -42.9 mm/year (Tianjin, China) and more than 10 times faster than climate-driven global mean-sea level rise of 3 to 4 mm/year. The median rate of each city ranges from -16.2 (Ho Chi Minh City, Vietnam) to 1.1 (Nanjing, China) mm/year and is wider than that of the total vertical land motion estimated in the Intergovernmental Panel on Climate Change Sixth Assessment Report (IPCC AR6) derived solely from tide gauge data based on point measurements. The latter ranges from -5.2 (Manila, Philippines) to 4.9 (Kolkata, India) mm/year. We suggest that total vertical land motion is likely to have higher global variability than estimated in the IPCC AR6, and thus highlight the need to integrate these InSAR-based land subsidence rates in future relative sea-level assessments.

Land subsidence is mostly attributed to excessive groundwater extraction. As per the block-wise ground water resources assessment report released by the Central Ground Water Board (CGWB) for the year 2017, 2020 and 2022, various regions in India fall under the over-exploited category for all these three years. Less availability of ground water for longer period affects crop yield and production from that particular region and also causes land subsidence. Measuring and understanding the spatio-temporal extent of subsidence is crucial to mitigate its hazards. In this study, we employed an Interferometric Synthetic Aperture Radar (InSAR) technique to measure the temporal subsidence (in dry season i.e. March to June) for the years 2020 and 2021 over the Salem region of Tamil Nadu, India. Salem is the fifth largest city in Tamil Nadu and mostly influenced by Information Technology, Steel and Textile industries. In this study, we have used small baseline subset (SBAS)-InSAR technique with Sentinel-1 data which has been widely used to estimate surface displacement of millimeter scale. Sentinel-1 satellite data operating in C-band (5.405 GHz central frequency f, and 5.547 cm wavelength λ) has been widely used because it offers comprehensive geographic coverage, frequent acquisitions, and free access. In this study, the data were acquired along descending track in dual polarization and the employed acquisition mode was the Interferometric Wide (IW) swath. Total 9 SBAS pairs were processed to generate time-series subsidence maps. The temporal baseline for each SBAS pair ranges from 24 to 36 days whereas perpendicular baseline varies from 5 m to 154 m. For the dry season of the year 2021, we found 3.3 mm of line-of-sight (LOS) displacement (in East-West direction) close to the center of the city whereas 46 mm LOS displacement was observed for the year 2020. These are the preliminary results, further detailed analysis is in progresswhere subsidence has been correlated with the available CGWB ground water information and the GRACE satellite. Initial obtained results are promising. The results of this study are very important for government organisations as they need to create new regulations to prevent the overuse of groundwater and to support resilient and sustainable agricultural practises.

HARMONIA is a European funded project that is focussed on developing integrated solutions for urban environments, tailored to the European Cities needs of security, health, prosperity and wellbeing, with regards to the impact of Climate Change (CC). HARMONIA wants to combine multiple Earth Observation (EO) datasets - including GEOSS and Copernicus datasets and services - with ensemble modelling, socio-economic and in-situ data at the spatial and temporal scales.

In the framework of the HARMONIA, Development of a Support System for Improved Resilience and Sustainable Urban areas to cope with Climate Change and Extreme Events based on GEOSS and Advanced Model-ling Tools, we have adopted remote sensing data acquired by SAR sensors to study possible ground movements for selected urban areas, i.e., the pilot sites of the project. HARMONIA will test modern Remote Sensing (RS) tools, Machine Learning (ML)/Deep Learning (DL) AI techniques to develop a modular scalable data-driven multi-layer urban areas observation information knowledge base, using Satellite data time series, spatial information and auxiliary data, which will also integrate detailed information on local level. In this work, we will show the retrieved results of the monitoring of the surface in urbanized sites through multi-temporal InSAR technique. We have adopted Persistent Scatterers Interferometry (PSI) to map de mean ground velocity and related dime series of deformation, on Milan (Italy), Sofia (Bulgaria), and Piraeus (Greece). SAR images acquired by ESA mission Sentinel-1. Data from ascending and descending orbits have allowed estimate the vertical and horizontal components of the ground motion. For the three pilot city, we have compared InSAR with the velocities from continuous GNSS stations. The three cartesian components of GNSS measurements (North, East, Up) have been projected along the ascending and descending LOS of the SAR acquisitions, respectively.

The PS maps highlight some patterns of ground and infrastructure deformation. In particular, within the Sofia metropolitan area have been declared 4 vulnerable zones along the Sofia region rivers. The main river that poses a potential flood hazard is the Iskar River, crossing the so-called ECOZONE Sofia – East (207 km²), appointed as a pilot area, integrating urbanised, nature and agricultural areas in a complex region for industrial production, logistics, services, agriculture, living, culture and leisure activities. The zone expresses all main environmental and risk prevention challenges in their full complexity. A buffer zone covering two kilometres on both sides of the riverbed is selected for the development of some of the HARMONIA services, especially in flooding/flash floods and landslides domains, as well as critical infrastructure and environmental quality impacts.

In addition, for Sofia pilot site, a further InSAR analysis on surface displacement was conducted on a peatland area of national importance, located outside the city of Sofia, to assess the restoration peat and ecosystem conditions.

GDM-SAR "Ground Deformation Monitoring using SAR data" is an on-demand service for processing InSAR products from Sentinel-1 radar imagery. This service has been developed by ForM@Ter (the Solid Earth data and services center of the French Research Infrastructure Data Terra) in connection with the Thematic Core Service "TCS Satellite data" of the European Research Infrastructure EPOS and since the end of 2019, with the support of CNES (French Space agency). Based on the NSBAS processing chain using a small baseline approach, GDM-SAR allows an automated computation of single interferogram or a network of interferograms with its associated unwrapped phase time series giving access to measurement of ground deformations worldwide and with a revisit time down up to 6 days. This service allows non-expert users to run processing with simple option choices without having to worry about setting up and maintaining a complex processing chain on a computing cluster. It also offers expert users a simple and fast way to explore a new area or a specific phenomenon such as a volcanic or seismic crisis, while keeping a certain flexibility in the choice of processing parameters. Users access the service through a web interface specifically designed for radar interferometry usage. The interface allows the user to interactively choose the study area and the Sentinel-1 data suitable for InSAR processing and to follow the progress of the processing. The generated products are available for download for a limited period of time (a few weeks). A preview of the products is possible directly on the interface. The generated products are similar to those of the FLATSIM service of ForM@Ter (see https://formater.pages.in2p3.fr/flatsim). Most of the products are provided in both radar and ground geometry (in geotiff format), interferograms are available in different versions (wrapped/unwrapped, filtered/unfiltered, with/without atmospheric correction from global model) allowing for user-customized post-processing. A time series of the unwrapped phase is also provided as well as many other auxiliary products allowing advanced analysis of ground displacements by the user. Products are compatible with the catalog and data formats of EPOS and ForM@Ter. The service scheduled to open mid-2023, initially to researchers from French research institutions and universities. A wider opening to the community of EPOS users is planned, but its modalities and its economic model are under discussion.

In southwestern Taiwan, approximately 10 cm/yr of ultra-rapid uplift rate and 1 cm/yr convergence rate on part of the block between two active reverse faults, the Chegualin fault to the west and the Chishan fault to the east, has been detected by geodetic observations. This ultra-rapid deformation rate is larger than the plate convergence rate of ~8.2 cm/yr across the Taiwan Island between the Eurasian and Philippine Sea plates. Because of the presence of an ~4-5 km thick mudstone formation in SW Taiwan, which extends from inland to off-shore where mud diapirs have been proposed to form anticlines, we therefore hypothesize that activity of mud diapirs may be a dominating process responsible of the crustal deformation pattern in SW Taiwan. Based on the limited geodetic data in previous studies, three segments with different present-day kinematics were proposed along the Chishan and Chegualin faults. The Chishan fault shows left-lateral motion in its northern and southern segments and the right-lateral motion in its central segment. The Chegualin fault shows the right-lateral motion in its northern and southern segments and left-lateral motion in its central segment. In addition, significant uplifts are revealed on the blocks between the two faults in the northern and central segments. However, according to geological investigation, looking at cumulated displacement at longer time-scale the Chishan fault is a reverse fault with left-lateral strike-slip component, while the Chegualin fault is a reverse fault with right-lateral strike-slip component. Furthermore, the horizontal and vertical velocity profile in the neighboring area cannot be well modeled by an elastic dislocation fault model only. Additional structures or physical processes may exist to cause the observed deformation in this region. We hypothesize that the significant uplift and the different fault components of the motion in the central segments of the Chishan fault and the Chegualin fault are caused by mud diapirs. To verify the proposed fault kinematics in the central segment of this fault system, we have installed 12 continuous GNSS stations across the two faults since 2022. We also started InSAR processing using 9 ascending ALOS-2 images from 2016 to 2021 to improve the spatial resolution of surface deformation measurements in this region. We expect to show preliminary comparison of GNSS measurement with several ALOS2 interferograms.

For future work, a high spatial resolution 3D velocity field will be estimated by inverting the GNSS data and InSAR results using velocity inversion. Furthermore, the strain rate field, including the principal strain rates and maximum shear strain rates, will also be calculated to understand the detailed deformation pattern in this region. The strain rate field will help to test whether the faults or mud diapirs dominate the lateral extrusion in SW Taiwan.

The Kerguelen islands (South Indian Ocean), at a latitude of 49° in the southern hemisphere, constitute the emerged part of a 1500 km-long oceanic block formed in the Oligocene by hotspot activity. During the last glacial maximum, the Kerguelen islands were largely covered by an ice cap, whose remains are today reduced to the 25 km-wide Cook ice cap, in the west part of the main island. The Cook ice cap is subject to accelerated melting since the 2000s, a likely consequence of global climate change. The Kerguelen islands also host recurrent low-magnitude seismicity, clustered in swarms, resulting from the possibly combined effect of structural inheritance, residual volcanic activity and glacio-eustatic adjustment.

To investigate the present-day deformation field of the Kerguelen islands, using the full archive of Sentinel-1 SAR imagery acquired since 2015, we conduct a small-baseline InSAR time-series analysis. We find that the Kerguelen islands are affected by a broad pattern of crustal uplift, peaking at ~ 5 mm/yr, centered on the Cook ice cap, with a spatial wavelength of ~ 100 km. This result is confirmed by independent analysis of two overlapping Sentinel-1 tracks. The spatial distribution of the LOS deformation can be explained by elastic rebound of the crust in response to unloading at the surface in the area of the Cook ice cap.

Using the same Sentinel-1 dataset, we also isolate the coseismic deformation field of an earthquake doublet in October 2017 (with magnitudes M=4.6 and M=4.7), and another earthquake in June 2015 (with M=4.7). A joint seismological-geodetic analysis of the deformation pattern and seismic wavefield of these events shows that all three events occurred near the surface (depth < 2 km), and all involve normal faulting, albeit with contrasting azimuths. Proximity of the 2017 earthquake doublet with the melting Cook ice cap is suggestive of a causal link between ongoing surface unloading and fault slip. However, the overall low level of seismic activity in the area and the intrinsically ambiguous causes of earthquake triggering in general lead to an unverifiable valuation of this hypothesis. Nevertheless, the existence of shallow earthquakes and ongoing uplift in the Kerguelen islands, first documented here, may reveal a transient process that would justify a more regular monitoring by in-situ and satellite observations.

Qaidam basin in the Tibet Plateau is known as the highest and most evaporative basin in China. It is located in a crescent valley bounded by highlands and the mountains of Altyn-Tagh, Qilian, and Kunlun. Extending 350 kilometers from the north to the south and 800 kilometers from the east to the west, the basin covers an area of 250,000 square kilometers at an altitude of 2,600-3,000 meters with annual evaporation up to 3,700 millimeters. The basin is divided into three blocks by the form of its substrate: the Mangya depression, the northern-margin fault block zone, and the new Sanhu depression, and all the underground buried structures and above-ground structures are distributed in these blocks. In the basin, the Cenozoic sedimentary rocks are up to 15,000 meters in thickness, and abundant oil and gas resources are contained in the oil- and gas-bearing Jurassic and Tertiary formation series and the gas-bearing Quaternary formations from the bottom up.

According to the report published by China National Petroleum Corporation (CNPC), the Qaidam basin is China’s highest onshore base of oil production and one of the essential petroliferous basins that CNPC’s oil and gas operation has mainly focused on. The large-scale hydrocarbon exploration and development began in the 1950s in the basin. From the 1960s to 1970s, it witnessed advances in hydrocarbon exploration with the discovery of the Sebei Gas Field and Gasikule Oil Field. After more than five decades of development, the Qaidam’s oil and gas development uniquely promotes social and economic growth on the Qinghai-Tibet Plateau. The Sebei Gas Field stands by the Senie Lake in the east Qaidam Basin with an average altitude of 2,750 meters. As CNPC’s 4th largest onshore gas field, it is the gas source of the Sebei-Xining-Lanzhou Pipeline and one of the West-East Gas Pipeline's primary strategic replacement gas sources. Through the development initiated in 1974, the gas field has developed a yearly gas capacity of 4.963 billion cubic meters (bcm) and accumulatively produced 11.659 bcm of gas.

Anthropogenic activities, such as the massive exploitation of oil and gas in reservoirs, are resulted in infrastructure insecurity leading to surface deformation of offshore. The subsidence rate depends on many variables, e.g., the amount of fluid removed, pore pressure decline, depth, and volume change. In contrast, the uplift of offshore happens depending on the amount of fluid injection, pore pressure increase, the reservoir layers expansion, and geological setting (depth, thickness, and area extent). Intense surface deformation of offshore may result in loss of life and assets, environmental implications, and significant influences on the industry's image. Here, we used the Interferometric Synthetic Aperture Radar (InSAR) and Sentinel-1 dataset from 2014 to 2022 to explore the surface deformation over the Sebei Gas Field and observed three circular subsiding features with rates up to 158 mm/yr. Our data also discovered westward motion up to 55 mm/yr and eastward motion up to 64 mm/yr on the eastern and western sides of subsiding areas, respectively, resulting from the radial strain variations across the subsiding zones.

Thanks to the vast amount of free continuous satellite SAR images, diverse Multi-Temporal Interferometry Synthetic Aperture Radar (MT-InSAR) approaches have been developed to estimate surface deformation in sub-centimeters accuracy. These techniques aim at limiting target decorrelation that reduces the accuracy of displacement estimation. MT-InSAR approaches can be categorized into three groups corresponding to two types of scatterers that are Permanent Scatterer (PS) and Distributed Scatterer (DS). The first group called as Permanent Scatterer Interferometry (PSI) uses high coherent point-wise scatterers (PS) to diminish the signal decorrelation. The spatial resolution is preserved with a cost of sparse estimation points coverage, especially in non-urban areas. In order to increase the estimation density, Distributed Scatterer Interferometry (DSI) was introduced. Contrary to PSI, to raise Signal-to-Noise Ratio (SNR), most DSI approaches lose spatial resolution due to the use of multi-looking. This results in changes in statistical properties of interferograms leading to phase inconsistency. A redundant network of interferograms is thus formed in most MT-InSAR approaches to retrieve the phase consistency.

Phase Linking (PL), or Phase Triangulation Algorithm (PTA)[1, 2, 3] is based on the principle where all the possible interferograms from a time series of SAR images are exploited. Recently, a study [4] demonstrates the presence of fading signals in multilooked interferogram, especially in case of short temporal baseline interferograms. The Small BAseline Subset (SBAS) algorithm is thus limited by a systematic phase bias. The study also points out that the use of all temporal combination network (i.e. PL) can mitigate substantially the phase bias, leading to an improvement of the phase estimation accuracy.

Generally, phase linking is driven by a maximum likelihood estimation (MLE) approach which requires reliable prior information on the coherence. The quality of the coherence used consequently determines the performance of the estimation [5]. This plug-in coherence, in most phase linking algorithms, is built upon the sample covariance/coherence matrix [2, 3, 6], following the assumption of an underlying Gaussian data distribution. This assumption can be inaccurate in the case of high resolution SAR data or when a spatially heterogeneous study area (e.g., urban area) is under consideration. As a result, the improvement in phase estimation accuracy can be expected if we take the non Gaussian data distribution into account in the covariance matrix estimation. Another problem is the spatial resolution degradation. In principle, the size of the multilooked (spatial) window should be twice the number of acquisitions in the time series to guarantee the accuracy of the covariance estimation. This implies a large multi-looking window, thus a significant degradation of the spatial resolution when the time series size is large.

To account for the two aforementioned issues, we introduce robust statistical models, in which a combination of a scaled Gaussian model with a low-rank structure covariance matrix is used to fit with non-Gaussian data and to address the spatial resolution degradation problem. To perform phase linking with the proposed robust statistical models, we propose a block coordinate descent (BCD) and majorization minimization (MM) algorithm to solve a joint maximum likelihood estimation of the covariance matrix and interferometric phases. The performance of the proposed algorithms is compared to that of the state-of-the-art PL with both synthetic simulations and real data applications (Sentinel-1 SAR images over the Mexico City, acquired from 03 Jul 2019 to 18 Dec 2019). The results obtained highlight that scaled Gaussian models allows for a significant improvement in terms of noise reduction, and low rank structure supports to reduce multilooked window size, especially in the context of long time series [7].

References

[1] A. M. Guarnieri and S. Tebaldini, “On the Exploitation of Target Statistics for SAR Interferometry Applications,” IEEE Trans. Geosci. Remote Sens., vol. 46, no. 11, pp. 3436–3443, 2008.

[2] A. Ferretti, A. Fumagalli, F. Novali, C. Prati, F. Rocca, and A. Rucci, “A New Algorithm for Processing Interferometric Data-Stacks: SqueeSAR,” IEEE Trans. Geosci. Remote Sens., vol. 49, no. 9, pp. 3460–3470, 2011.

[3] N. Cao, H. Lee, and H. C. Jung, “Mathematical Framework for Phase-Triangulation Algorithms in Distributed-Scatterer Interferometry,” IEEE Geosci. Remote Sensing Lett., vol. 12, no. 9, pp. 1838–1842, 2015.

[4] H. Ansari, F. De Zan, and A. Parizzi, “Study of Systematic Bias in Measuring Surface Deformation With SAR Interferometry,” IEEE Trans. Geosci. Remote Sens., vol. 59, no. 2, pp. 1285–1301, 2021.

[5] P. V. H. Vu, F. Brigui, A. Breloy, Y. Yan, and G. Ginolhac, “A New Phase Linking Algorithm for Multi-temporal InSAR based on the Maximum Likelihood Estimator,” in Proc. IEEE Geoscience and Remote Sensing Symp. (IGARSS), 2022, pp. 76–79.

[6] H. Ansari, F. De Zan, and R. Bamler, “Sequential Estimator: Toward Efficient InSAR Time Series Analysis,” IEEE Trans. Geosci. Remote Sens., vol. 55, no. 10, pp. 5637–5652, 2017.

[7] P. V. H. Vu, A. Breloy, F. Brigui, Y. Yan, and G. Ginolhac, “Robust Phase Linking in InSAR,” IEEE Trans. Geosci. Remote Sens., 2022 [under review].

Surface subsidence is a common phenomenon in undercast mining areas, as well as the uplift of the surface after mine closures and the ceasing of water withdrawal. Both types of these surface deformation can cause significant building damage if the mine shafts were close enough to the inhabited areas. Mining-induced deformations are usually measured using different techniques, such as conventional leveling, differential GPS (DGPS) surveys, and nowadays more often with SAR sensors. Differential interferometric techniques (DInSAR) and advanced DInSAR stacking tools (PS, SBAS, and hybrid solutions) can provide high-quality measures of the spatial and temporal evolution of the deformations. Moreover, even the deformations and damages of individual buildings affected by the mining activity can be monitored. InSAR-based building damage assessments, vulnerability-, and building health mapping are mainly based on the velocity of PS scatterers (Pratesi et al. 2015, 2016), or more specifically on differential settlement, relative rotation values (Peduto et al. 2019, Nappo et al. 2021). All the aforementioned techniques require dense PS point clouds with high spatial resolution, therefore, freely available sensors are less suitable for this kind of detailed analysis. On the contrary, there are mining-related sites where besides the conventional field surveys, only medium-resolution sensors are available for deformation monitoring and vulnerability mapping. Taking into account the limitation of these sensors, a new methodology is needed for vulnerability mapping.

Our research aimed to investigate the post-mining surface deformations of a former mining area near Pécs (southern Hungary) and compile the post-mining vulnerability map of the town. For determining the spatial and temporal evolution of the mining site and its surroundings, the full stack of ERS, Envisat, and S1 SAR images (between 1992 and 2022) were evaluated considering both ascending and descending geometries. Images were downloaded from the Vertex server of the Alaska Satellite Facility and ESA Online Dissemination portal, and they were processed using the PS algorithm of the ENVI SARscape 5.6.2.1. (sarmap SA, Caslano, Switzerland) software. The spatial evolution PS-based deformation was statistically analyzed with the R software and compared to the time series of more than 100 leveling points, geological and geomorphological maps, and more than 800 residential complaints (mining damage complaints between 1993 and 2021). Regression analysis of the above-mentioned factors highlighted the ineffectiveness of the on-site leveling on the monitoring of post-mining surface uplift at the study site. While ERS, Envisat, and S1-based PS data provided a solid base even for vulnerability mapping. The vulnerability model contained the spatiotemporal descriptors of PS scatterers and the geomorphological, and geological conditions of the site.

Observing how the surface deforms in time and space in volcanic regions is crucial for a better understanding of subsurface magmatic processes, but it also plays a significant role for hazard assessment, risk reduction, and crisis management. In recent years, Mount Etna, one of the most active volcanoes in the world and surrounded by densely populated areas, has experienced a period of intense activity characterized mainly by continuous degassing and recurrent lava fountains. Ground- and space-based systems are continuously monitoring the ground deformation caused by this activity.

In the final months of 2020, the summit craters showed vigorous activity along with increasing seismicity. In December 2020, a period of paroxysms with powerful and brief bursts of lava fountains began. This period intensified in February 2021 and lasted until April 1, during which 17 lava fountain episodes with heights of several hundred meters occurred along with tall columns of ash and steam rising several kilometers above the crater and with a rapid increase of volcanic tremor signal. Lava flows were observed descending to the south and east towards the Valle del Bove. The frequency of the events ranged from a few hours to a few days. By constraining the sources of the observed paroxysms, this study aims to understand the dynamics of the near-surface feeding system. We used Sentinel-1 data from the second half of 2020 to mid-2021, together with analysis of GNSS permanent network data, to examine the surface deformation on Mount Etna even before the increase in volcanic activity in order to locate and define the time-dependent ground deformation. For combining the two data sets, we have applied the Simultaneous and Integrated Strain Tensor Estimation From Geodetic and Satellite Deformation Measurements (SISTEM) algorithm, which allows to estimate three-dimensional ground displacements by integrating sparse GNSS measurements and Differential Interferometric Synthetic Aperture Radar (DinSAR) displacement maps.

A period of deflation during the paroxysm episodes and the occurrence of an inflation phase before the initial onset of the paroxysms suggest a link between the volcano activity and the observed deformation. The findings could serve to further the discussion on the distribution and dynamics of the magma reservoirs that shape the conduit system of Mount Etna and how those reservoirs interact with the regional tectonic regime.

Given the global importance of understanding natural hazards, the availability of synthetic aperture radar interferometry (InSAR) has proven invaluable for monitoring ground deformation from space. As InSAR phase is recorded modulo 2π , the unwrapping process to return continuous phase values is essential:

Φ_i,j = Ψ_i,j + 2πk_i,j (1)
where Φ_i,j is unwrapped phase, Ψ_i,j is wrapped phase and k_i,j is the ambiguity number.

The ill-posed nature of the unwrapping process necessitates the use of Itoh’s condition, an assumption where the absolute difference between the phase of adjacent pixels is generally less than absolute π. Traditionally methods have utilised residue information to guide the integration pathways during unwrapping (Goldstein et al., 1988), using Lp norm methods (Ghiglia et al., 1996) to reduce error occurrence and subsequent error propagation across an interferogram. Such methods are often successful when unwrapping interferograms of high coherence and where phase gradients are within the constraint of one phase cycle change. In circumstances where these conditions do not hold, isolation of areas with low signal to noise ratio and difficulties unwrapping high fringe
densities with steeper phase jumps greater than one phase cycle, can result in unwrapping errors.

With deep learning’s success in other fields, the application of deep learning to improve phase unwrapping has increased in popularity. Generally methods utilise a supervised approach, providing wrapped phase as input with target data ranging from the unwrapped phase itself (Wu et al., 2020), the ambiguity number (Spoorthi et al., 2019) or the ambiguity gradient (Chen et al., 2023). Whilst more successful than traditional unwrapping methods, limitations have been shown when applied to unwrap interferograms of average coherence less
than 0.5 and in places of no-zero gradient pixels (Chen et al., 2023).

Here we present a deep learning model which allows an improved distinction between noise and dense fringe regions. By doing so, improved unwrapping of interferograms with lower noise-to-signal ratios, where average interferogram coherence is less than 0.5 is possible. Using a training dataset containing synthetically generated interferograms, a multi-output supervised model has been trained to label the ambiguity gradient in the x and y directions when given a wrapped interferogram as input. The inclusion of a classification map as a target output improved the model performance. Output prediction certainty levels combined with the classification map are used to guide the order of unwrapping using an L1 norm method to return the ambiguity number of each pixel. Unwrapped phase is then calculated per (1).

References
Chen, Xiaomao, Chao He, and Ying Huang (2023). “An error distribution-related function-trained two-dimensional insar phase unwrapping method via U-GauNet”. In: Signal, Image and Video Processing. doi: 10.1007/s11760-
022-02482-y.
Ghiglia, D and L Romero (1996). “Minimum Lp-norm two-dimensional phase unwrapping”. In: Journal of the Optical Society of America A 13.10, pp. 1999–2013. doi: https://doi.org/10.1364/JOSAA.13.001999.
Goldstein, R, H Zebker, and C Werner (1988). “Satellite radar interferometry: Two-dimensional phase unwrapping”. In: Advancing Earth and Space Science 23.4, pp. 713–720. doi: 10.1029/RS023i004p00713.
Spoorthi, G, S Gorthi, and R.K Sai Subrahmanyam Gorthi (2019). “PhaseNet: A Deep Convolutional Neural Network for Two-Dimensional Phase Unwrapping”. In: IEEE Signal Processing Letters 26.1, pp. 54–58. doi: 10.1109/LSP.2018.2879184.
Wu, Zhipeng, Heng Zhang, Yingjie Wang, Teng Wang, and Robert Wang (2020). “A Deep Learning Based Method for Local Subsidence Detection and InSAR Phase Unwrapping: Application to Mining Deformation Monitoring”. In:
IGARSS 2020 - 2020 IEEE International Geoscience and Remote Sensing Symposium, pp. 20–23. doi: 10.1109/IGARSS39084.2020.9323342.

Inland still waters, such as lakes, wetlands and reservoirs, provide key ecosystem services to humans. Since freshwater supply, storage, and water power for electricity are the most relevant services for humans, the waters providing these services are monitored. However, almost 120 million water bodies worldwide remain unmonitored, and the costs to monitor all water bodies are enormous. Monitoring is essential as unmonitored still waters are already facing accelerated Earth system change, driven by human activities and climate change, with unknown consequences. Differential Synthetic Aperture Radar (DInSAR) is a promising technology for observing these resources from space. It employs the differences in the path length of two satellite acquisitions taken from the same orbital to generate maps of spatial and temporal changes of the water or land surfaces. Despite its potential, DInSAR also faces limitations for monitoring regarding resolution, water resources, and scope of application. Here we test two outrageous hypotheses concerning its application: First, against the common belief, InSAR can be used to track water level changes not only in wetlands but also in lakes. Second, DInSAR can not only help identify connectivity in wetlands but also hydrological barriers to sheet flow. For the first, we use DInSAR to track water level changes in lakes in Sweden and Ecuador, validating them against in-situ observations or hydrological patterns, respectively. We find that DInSAR can detect water level changes based on the phase differences of coherent pixels located on the shores of some lakes. For the second, we develop a convolutional neural network to identify hydrological barriers based on InSAR interferograms. We train and test this model in three tropical and subtropical wetlands; The Everglades and the Louisiana Wetlands in the United States and Ciénaga de Zapata in Cuba. The model can successfully locate flow barriers by seeing abrupt patterns of differences in phase, enabling mapping of the hydrological barriers to flow in wetlands such as roads, ditches or embankments. In this time of rapid Earth system change and the availability of SAR sensors increasing worldwide, we show the unknown potential of DInSAR for the monitoring and hydrological assessment of the functioning of surface water resources. This potential increases under the light of the new and upcoming missions SWOT and NISAR.

As the significant node connecting the subway network, the deformation monitoring of metro hubs is essential to ensuring urban transportation safety. Previous studies have often utilized InSAR technology to detect deformation in established subway stations and their surrounding areas. However, the deformation evolution process during the construction period is also important but is often difficult to be measured only based on the InSAR technique due to the decoherent effect and limited penetration capability. Therefore, there is currently a lack of comprehensive strategy that can reveal the detailed deformation evolution process from underground structures to the ground surface of metro hubs during the construction period.

This paper combines InSAR monitoring and on-site sensors to address these issues. The PS-InSAR method is used to monitor the surface deformation of the construction impact area, while on-site sensors are used to monitor the settlement both upon and under the subway station, which can increase the density of observation points in the low coherent regions and inside the structure. Through this approach, a settlement funnel overlooked by traditional monitoring methods (leveling and GNSS) was discovered, and precise settlement within the construction area was obtained. Moreover, the longitudinal and cross profiles along the subway station are calculated to reveal the influence scope of the construction. Finally, the evolution process of subsidence from underground structures to the ground surface is observed and analyzed.

The study area of this research is the Shapu metro hub under construction on Line 12 in Shenzhen. To conduct a comprehensive deformation analysis, the third-party monitoring data, Sentinel-1 measurements, and machine vision observations are combined. The machine vision sensor was installed on the top inside of the prefabricated subway station, which can monitor the vertical deformation of the main structure (Figure 1).

Due to the impact of construction, the PS (permanent scatterers) points obtained from Sentinel-1 data are mainly distributed on the structures near the construction area, and few PS points are identified inside the construction fence. The cross-validation is conducted by comparing groundwater level monitoring data during the construction period with nearby PS points (see Figure 2). The results show that: in the first stage (①: between Jul. and Oct. 2021), the groundwater level fluctuated around zero and a slight settlement of about 5mm was observed on the PS points; then, the groundwater level dropped rapidly during the second stage (②: from Nov. 2021 to Jun. 2022), and the deformation of the PS points increased to more than 20mm; in the third stage, after the groundwater level stabilized (③: since Jun. 2022), the deformation of the PS points became more stable.

The cumulative settlement of PS1, PS2, and PS3 is approximate -25mm, and there are many other PS points in this area showing similar cumulative settlement. A settlement funnel is found in the Langxia Industrial Park (in figure 3(a)). We can infer that the construction of the subway station had a great impact on the area. However, this was ignored by third-party monitoring.

We further analyze the influence scope of the subway construction, as shown in Figure 3(b), a 500m longitudinal section and a 600m cross-section are selected to analyze the cumulative deformation. The left side of the cross-section is more seriously affected than the right side (see red lines), which is the location of the Langxia Industrial Zone. For the longitudinal section, the subsidence area is mainly within -150 and 150m (see blue lines). We can infer that the construction has the most serious impact on the roadside buildings of Langxia Industrial Park, where the cumulative deformation of many PS points reaches 25mm, and a settlement funnel is formed. Secondly, the cumulative deformation of the surrounding municipal roads due to construction also reached about 20mm, while the residential areas in the southeast were relatively less affected.

Machine vision sensors are very sensitive to the vertical deformation caused by soil covering construction, as shown in figure 4. The results of machine vision showed highly consistent with the on-site construction process. Throughout the entire time series, the average settlement of the station structure first increased and then stabilized with the progress of the soil covering. The middle part had a larger average settlement, while the two ends had smaller average settlements, with a settlement variation range of 0 to -5mm.

The subsidence observations derived from InSAR and Machine vision sensors from Sep. to Dec. 2022 are compared to reveal the evolution process of subsidence from underground structures to the ground surface. Kriging interpolation is used to calculate the time series of the InSAR surface subsidence profile along the subway station as shown in Figure 5(a). Moreover, the time series settlement profiles of machine vision with a sampling interval of approximately 12 days are shown in Figure 5(b).

It can be seen that the subsidence first occurred on the underground structure before Sep. 13 2022 and then almost keep stable within 5mm. However, basically, no subsidence was observed on the ground surface in Sep. 2022. The subsidence of the ground surface begins in Oct. 2022 which is at least one month later than the underground structure and gradually increased to about 5mm. After Dec. 2022, the profiles derived from the two datasets showed a similar deformation trend with larger subsidence (about 5mm) on the Ring No.1 to No.66 and smaller (about 3mm) on the other part. Therefore, the subsidence of Shapu Metro Station first occurred on the underground structure and one month later gradually transmitted to the ground surface. After three months of soil consolidation and compression, the subsidence of the underground structure and ground surface become almost consistent.

According to our results, fortunately, the cumulative settlement of the subway structure is less than the standard setting of 8mm, and currently, the top structure of the station is basically safe.

In summary, the combination of InSAR and on-site sensors can be used for detailed surface and underground deformation monitoring of subway stations during the construction period. Among them, PS-InSAR can monitor the surface construction-affected area, while on-site sensors can accurately monitor structural deformation and supply surface observations. The evolution process of subsidence from underground structures to the ground surface is further modeled and revealed.

The North African region is known for its transpressional tectonic regime, which is primarily controlled by the ongoing oblique convergence between the Nubian and Eurasian plates. The relative plate motion increases eastwards from ~2 mm/yr to ~7 mm/yr and involves both offshore and inland tectonic structures distributed within a broad zone. Despite the relatively low strain rates, significant crustal seismicity and destroying earthquakes have been recorded in the region (Morocco, 1960; Algeria, 1954 and 1980; and Tunisia, 1977 and 1989). The current kinematic models involve discrepant implications for the role between inland and offshore structures in accommodating the relative plate motion. Therefore, better constraints on the quantification of the strain partitioning and the interseismic behavior of inland tectonic structures are critical for the seismic hazard assessment of the region.

To improve our understanding of the current active tectonics in northernmost Africa, we present the first large-scale map of current interseismic velocities over the whole Maghreb region. Our velocity field combines over 7 years of Sentinel-1 SAR imagery and was produced using the New Small Baseline Subset processing chain (NSBAS, Doin et al. 2011). To retrieve the near-vertical and horizontal components of present-day motions, SAR data from 10 tracks in descending and 11 tracks in ascending orbit were integrated. Interferogram networks included image pairs with temporal baselines equaling 6, 12, 24, 48, 96 days, and 1 year to optimize for temporal sampling while minimizing the signal bias resulting from processes like seasonal vegetation growth. However, because of the large spatial scale of the study area, several interferograms were discarded because they were highly affected by snow, vegetation, and fast-moving sand dunes. For the time-series estimation, the remaining interferograms (over 1800 per track) were multi-looked to 32x8 looks to increase the signal-to-noise ratio, filtered using a gradient-based filter, unwrapped in the spatial domain by region growing with starting point in the most coherent area and corrected from orbital ramp residuals. Furthermore, prior to the unwrapping step, delay maps derived from the ERA5 atmospheric model reanalysis were applied to the interferograms to correct for the Atmospheric Phase Screen (APS).

The estimated deformation maps reveal multi-scale present-day motions, with large- and small-scale signals suggesting tectonic origin and ground response to anthropogenic activity or landslides, respectively. The massive data and our processing strategy allowed us to capture both the near-horizontal and vertical components of the millimeter-level interseismic displacement fields. Using these results, we investigate the main tectonic structures in the area and propose an updated map of the active faults. Finally, we use our regional deformation field to test whether it supports for most of the present-day relative plate motion in northernmost Africa being absorbed by inland structures along the Atlas Mountains or if offshore deformation plays the main role.

The information extracted from Time-Series Interferometric Synthetic Aperture Radar (TInSAR) nowadays is routinely used for studying of the earth surface dynamics of different deformation mechanisms. The increasing use of TInSAR-derived products (provided particularly by free availability of ESA Copernicus SAR data) induces a necessity of proper and standard quality control methods to assess the precision and accuracy of the InSAR-based products. Despite many studies and developments regarding such quality description in terms of precision and noise structure, the quantification of the TInSAR uncertainties (or biases) induced by phase unwrapping errors has been remarkably overlooked so far. Although some initial efforts have been made (either for some limited methodologies and scenarios, or by extensive simulation algorithms), still there is no analytical criterion for assessment of such uncertainties.

It should be noted that the presence of unwrapping errors in TInSAR products is always probable. Particularly, in areas with high level of noise or with a peculiar deformation pattern, there is always a chance (even small) for unwrapping errors to be occurred. TInSAR algorithms usually try to somehow identify and mitigate the unwrapping errors either by a trial-and-error or by an experimental approach based on the skills of InSAR experts. Nevertheless, the performance of such heuristic methods is always case-study dependent. The main reason is that there are different factors, differing from case to case, which contribute to the success of the phase unwrapping. Examples of these factors are different spatio-temporal behavior of deformation mechanisms, different initial assumptions used in the phase unwrapping, different landscape characteristics, different processing settings, and so on. The impact of these factors on the correctness of the phase unwrapping needs to be assessed and delivered to the final users. In other words, there is a need for a quality-description approach capable of digesting the effect of the aforementioned factors to quantify the probability of correct phase unwrapping or its success-rate.

In this study, we introduce a new analytical approach for quantification of InSAR uncertainties induced by phase unwrapping errors. The concept of the method is based on the quality description criteria, such as Success-Rate and Ambiguity Dilution of Precision (ADOP), that are used in GNSS applications for describing the uncertainties of integer ambiguity resolution methods. It should be noted that these criteria have been already exploited in some TInSAR studies, however all the studies so far have been limited to relative phase unwrapping of pair of close-by pixels (called arc). Here, we extend this idea to spatio-temporal phase unwrapping in a network approach. The main challenge to address is how the quality (or success-rate) of individual arcs in a network of pixels should be propagated to the success-rate of the final estimated time series of all the pixels. By such propagation, both the noise characteristics and also the spatio-temporal network structure of the data are taken into account. At the end for each individual point, we estimate a success-rate indicator, which provides the probability of correct phase unwrapping for that point. This new indicator can be used together with the final TInSAR products and other quality measures to describe not only the precision of the data but also their accuracy. The proposed approach is also flexible to quantify the phase unwrapping uncertainties induced by wrong initial assumptions about deformation mechanisms (Note that all the phase unwrapping methods require such assumptions about spatial or temporal behavior of deformation signals). The proposed approach provides a quantitative tool (called Biased-Success-Rate) to assess the effect of wrong deformation assumptions on the accuracy of TInSAR phase unwrapping. In this way, it can improve the falsifiability of the TInSAR products.

We validate the introduced method in a simulation manner for different scenarios. The results confirm that the method is capable to describe the probability of occurrence of unwrapping errors with sufficient correctness. Also the performance of the method is demonstrated for different real case studies, from small-scale applications (e.g., infrastructure monitoring) to large-scale studies (e.g., subsidence monitoring in urban and semi-urban areas). The introduced quality indicator can be considered as the first quantitative/analytical measure of accuracy of TInSAR data in respect of unwrapping errors.

The geodynamics of Ecuador (northwestern South America) are directly related to the subduction of the oceanic Nazca plate beneath the edge of the South American continent (at a rate of 55-58 mm/yr beneath the Ecuadorian coastal margin). Ecuador has a large continental transcurrent fault system starting at the active margin in Gulf of Guayaquil oblique to the Andes Cordillera, through to the Colombian border known as the Chingual-Cosanga-Pallatanga-Puná (CCPP) fault system. In addition, the subduction has created important fault systems inside the country (e.g. Quito, Latacunga-Pujili, El Angel fault systems) as well as magmatic systems (e.g. Sangay, Cotopaxi volcanoes) that have been showing ongoing deformation in recent years. This is in addition to active surface deformation related to other types of natural phenomena (e.g. landslides and land subsidence) and anthropic events throughout the country.

We use the technique of Synthetic Aperture Radar Interferometry (InSAR) for monitoring the large-scale surface deformation and these observations provide an essential complement to GNSS network ground-based instruments in Ecuador. Here, we use 4.1 years (between 2017 and 2021) of Sentinel-1 InSAR time series analysis across the country territoy. We produce interferograms every 6, 12, 24 days and 3, 6, 9 and 12 months between epochs using the LiCSAR system (divided in 8 descending and 7 ascending LiCS frames), and the LiCSBAS software to perform the time series analysis. We tested GACOS weather correct models to mitigate atmospheric contributions to phase, and examined the effect of the phase bias (fading signal) due to short period interferogram networks.

In terms of InSAR coherence we have identified zones with poor or near-zero coherence due to the dense vegetation that is prevalent either in the coastal region (west) or in the Amazon (east) where the measured velocity is restricted only to patchy areas (e.g. urban areas and stretches without vegetation). However, in the mountain range (centre of the country) where most of the important fault systems and volcanic centres are located, as well as large urban areas, the coherence is good and allows us to have reliable deformation measurements.

We estimate the average north-south and east-west velocity between 2017 and the end of 2021 for the GPS time series network of the National Geodesy Network (RENGEO) of the IG-EPN.

We combine our InSAR line of sight velocities and GNSS north-south motion to decompose into vertical and horizontal motion to develop a velocity field for Ecuador in order to identify surface deformation, However, much of the territory is a challenge to work in due to the lack of coherence with C-band SAR. We present some examples associated with the active and erupting volcanoes as Sangay and Cotopaxi , active tectonic areas around Quito and Chiles Cerro Negro volcanoes and anthropogenic processes related to mining activities in southern Ecuador.

Introduction

In its most essential form, InSAR (SAR Interferometry) can be used to provide displacement estimates for an arc, formed by two sufficiently coherent point scatterers. The displacement estimate, which is usually a parametric description of the displacement as a function of time, needs to be estimated from the original observations, which are the double-differenced (DD) phases for the arc, i.e., the phase difference between two point scatterers (PS), relative to a reference epoch.

Both a proper functional and stochastic model are essential to accurately estimate the displacement parameters. However, the intrinsic problem of InSAR is that both are unknown. Especially in the built environment, it is generally never known exactly from what object the main signal originates, resulting in an unknown kinematic behavior, and thus functional model. For example, there can be a major difference between a signal originating from a building, compared to that of the road right next to the building, even though these signals are spatially close.

Regarding the stochastic model, the quality of a phase observation at a single epoch is intrinsically unknown since each individual PS has its own unique scattering properties. Additionally, the quality of the observed phases is likely to change over time. Therefore, different phase observations should receive different weights in the stochastic model.

In current PSI approaches the quality of the observations is typically determined by evaluating the residuals between the observations and the model evaluated from the estimates, given a pre-selected parameterization. This method is highly reliant on the correctness of the functional model, as using a different model will result in different estimates, residuals and thus estimated quality. Likewise, under-parameterization of the model will lead to an overly pessimistic quality estimate, resulting in, e.g., overly pessimistic minimal detectable displacements. Most importantly, the residue-based quality assessment is epistemologically equivalent to circular reasoning, and therefore a fallacy: in order to estimate residuals, we need to have estimated the parameters, but to estimate the parameters unbiasedly, we need to know the quality of the observations, which we derive from the estimated residuals.

Ideally, the stochastic model should be known prior to the estimation since it influences the result. Lower quality observations should receive a lower weight when the displacement, ambiguities, heights, and atmospheres are estimated. The absence of a proper stochastic model may lead for instance to different estimated ambiguities and thus in significantly incorrect displacement parameter estimates. Moreover, an independent stochastic model is essential when InSAR is used for monitoring purposes. To test whether a significant change in the displacement behavior of a scatterer has occurred, we need to know the quality of that observation.

Method

Here we present a method to estimate the Variance-Covariance Matrix (VCM) , Qφ_i,j, for the double-difference (DD) phase observations of an arc between point scatterers i and j, starting with the VCM for the Single Look Complex (SLC) phases of one single point scatterer (PS), Qψ_i, where ψ_i are the SLC phases of point i.

The Normalized Amplitude Dispersion (NAD) can be used to fill the diagonal of Qψ_i, which is assumed to be loosely related to the quality of the SLC phases with:

σ_ψ ≈ μ_A / σ_A = NAD,

where μ_A and σ_A are the mean and standard deviation of the amplitude of the PS respectively. The assumption σ_ψ ≈ NAD only holds when NAD < 0.2 (Ferretti et al., 2000). Therefore, we derived an empirical relation between σ_A and NAD based on simulations. Note that the amplitude of a single PS may change over time and so does σ_ψ. Therefore, we used the Pelt (Pruned exact linear time) change point detection algorithm to detect different temporal partitions in the amplitude time series (Truong et al., 2020). Per partition, we estimate the NAD and subsequently σ_ψ based on the derived empirical relation.

So, when we detect p partitions, we estimate p values for σ_ψ, and all phase observations within one partition are assigned the same value for σ_ψ. These values are used to fill the diagonal of Qψ_i and the off-diagonal elements are set to zero, since there is no correlation in time. Note that a coherent (ant thus correlated) signal is required to get proper estimates. However, the coherent signal is part of the functional model. With the stochastic model, we only want to describe the variability of the observations, and this variability is not correlated in time. With both Qψ_i and Qψ_j it is possible to derive the VCM of the single difference phases in time, given a chosen mother (reference) image and consequently it is possible to combine the two points, take the difference, and compute Qφ_i,j.

Results, Impact and Conclusion

We applied our approach on real data to estimate displacement models as a function of time. We found that using a proper VCM improves the results, where the fitted models with the VCM are a better approximation of the displacement data. Moreover, using a proper stochastic model allows us to make improved statements on the precision and reliability of the estimated parameters, which is essential when the results are used for monitoring purposes.

A key characteristic of our method is that we do not only use the phase data in the estimation, but that we include more information in the form of the amplitude data. Utilizing various partitions is particularly advantageous, as the quality of the observations often changes over time. Moreover, we regularly observe that the kinematic behavior of the arc also changes between partitions. We need to take advantage of this information, which we can do since we know when a new partition starts.

References

Ferretti, A., Prati, C., & Rocca, F. (2001). Permanent scatterers in SAR interferometry. IEEE Transactions on geoscience and remote sensing, 39(1), 8-20.

C. Truong, L. Oudre, N. Vayatis. Selective review of offline change point detection methods. Signal Processing, 167:107299, 2020.

A common practice with InSAR results is to classify “persistent” or “permanent” scatterers, which are ambiguous terms about the properties of the scatterers in space (scatterer strength) versus time (phase coherence). According to the Delft InSAR scatterer taxonomy (Hu et al., 2019), scatterers range from Point Scatterers (PS) to Distributed Scatterers (DS), of which both vary from continuously coherent, to temporary coherent, and incoherent. The distinction between PS and DS is clear from the definition perspective, but not so much from the estimation perspective. InSAR results classified as “PS” may contain sub-mainlobes, sidelobes, and other scatterers, obscuring the superior geopositioning and high phase signal-to-noise ratio potential of a dominant PS. On the other hand, pruning the scatterers to only a dominant PS may have a negative impact on the point coverage, and, consequently, interpretability.

Here we propose an estimation procedure for extended classification of the InSAR scatterers in the spatial domain. We classify every pixel in the SAR image based on its scattering in the following steps, always excluding the already identified category from further classification. First, we find PS candidates based on the amplitude peak estimator. The sidelobes are then removed from them. The remaining dominant PS fall within the highest quality category, for which a sub-pixel position can be estimated (Yang et al., 2020). DS pixel patches are then identified based on the amplitude neighborhood estimator. For this group, phase refinement by means of phase linking is possible (Ansari et al., 2018). The remaining category contains sub-mainlobe pixels, weak point scatterers not identified as peaks, and distributed scatterers without sufficiently homogeneous neighbors. We refer to this group as “Weak Scatterers” (WS). The result is a classification of detected scatterers in three categories (PS, DS, WS), each with its own quality characteristics regarding estimated (displacement) parameters and geopositioning. Depending on the objective of a particular project, the most suitable selection of one or more scatterer categories can be used for interpretation and further data analysis, thereby optimizing the outcomes.

We show the added value of the classification on two different industry projects. For a building infrastructure project, the attributability of the scatterers to objects (Dheenathayalan, 2016) is essential for drawing conclusions about the stability of the building. Limiting the interpretation to dominant PS with superior positioning accuracy, these object-related conclusions can be drawn more reliably. Knowing whether the scatterer is on a building roof, or the road next to it is of critical importance to drawing conclusions about the stability of the building. For a different type of project focusing on wide area displacement patterns, both PS and WS carry useful displacement signals. Especially when aggregating scatterers on assets, the point density, including WS, has an impact on the derived statistics. An increase in sample size lowers the standard deviation of average displacement rates and increases the reliability of derived insights.

Hu, F., Wu, J., Chang, L., & Hanssen, R. F. (2019). Incorporating temporary coherent scatterers in multi-temporal InSAR using adaptive temporal subsets. IEEE transactions on geoscience and remote sensing, 57(10), 7658-7670.

Yang, M., Dheenathayalan, P., López-Dekker, P., van Leijen, F., Liao, M. & Hanssen, R. F. (2020), ‘On the influence of sub-pixel position correction for PS localization accuracy and time series quality’, ISPRS Journal of Photogrammetry and Remote Sensing 165, 98–107.

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

Dheenathayalan, P., Small, D., Schubert, A. & Hanssen, R. F. (2016), ‘High-precision positioning of radar scatterers’, Journal of Geodesy 90(5), 403–422.

The high Tibetan plateau is marked by large fault systems accommodating the deformation generated by the India-Asia collision. Large sedimentary basins, affected by strong seasonal hydrological loads, surrounded by mountain ranges subject to erosion, also mark the landscape. Measuring the deformation of the ground surface associated with fault activity or of non-tectonic origin is one of the elements to better understand the different deformation processes of the Tibetan plateau and quantify the kinematics of the faults, some essential steps to progress in the understanding of the seismic cycle and hazards' assessment.

Thanks to their high temporal resolution and wide spatial coverage, the radar images provided by the Sentinel-1 satellites offer the possibility to measure surface deformations with an unprecedented spatio-temporal resolution. This study is based on a massive and automated InSAR processing service developed by the french Solid Earth Data and Services center, ForM@Ter, and operated by CNES (FLATSIM, doi : Thollard et al., 2021). We analyze time series produced by FLATSIM (doi:10.24400/253171/FLATSIM2020) using a small baseline approach (NSBAS, Doin et al., 2011, Grandin, 2015), based on Sentinel-1 images covering the eastern Tibetan Plateau over the period 2014-2020 (1200 km long swaths in seven ascending and seven descending orbits, covering an area of 1,700,000 km2, with a spatial resolution of 120 m).

We propose a new time series analysis methodology to characterize deformations at a continental scale, notably via a referencing of InSAR surface velocities in a pseudo-absolute reference frame, with a low dependence on GNSS data. We decompose the line-of-sight time series into a linear term (whose horizontal and vertical components are inverted) and a seasonal term. The latter is dominated by hydrological motions in large sedimentary basins and deformation associated with permafrost freeze-thaw cycles; atmospheric delay residuals are also observed. The vertical component of the mean velocity map is dominated by permafrost degradation (and other non-tectonic phenomena). The horizontal velocity is dominated by tectonic deformation associated with active faults, and ubiquitous small scale downslope movements. These gravitational signal are filtered based on a local slope velocity correlation analysis. The corrected velocity map then highlights the slip transfers between different fault systems (Altyn Tagh to Haiyuan, Kunlun to Xianshuihe) and the secondary structures accommodating the deformation within the large recognized tectonic blocks. Finally, we jointly invert InSAR velocity maps and published GNSS velocity fields using an elastic block model (TDEFNODE, McCaffrey, 2009) to discuss the interseismic velocities of major active faults, the degree of localization and partitioning of tectonic deformation in the eastern Tibetan Plateau, and the limitations of such a modeling approach.

The 2023 Kahramanmaras earthquake sequence was devastating for the densely populated nearby regions within Türkiye and Syria. The main source of seismic hazard in Türkiye has historically been from the right-lateral North Anatolian Fault system. However, the February 6, 2023, M7.8 main shock occurred on the shorter, left-lateral East Anatolian Fault and the M7.5 aftershock occurred roughly 10 hours later on a secondary fault. InSAR data can provide valuable insights into these complex ruptures by capturing the full field of surface deformation in response to the events. ESA’s Sentinel-1 and JAXA’s ALOS-2 missions both acquired ascending and descending scenes which spanned the two earthquakes, however, neither platform made acquisitions between the M7.8 and M7.5 earthquakes, making the contributions from each individual earthquake difficult to separate in the resulting interferograms. In this work, we use InSAR data to illuminate the complex ground deformation patterns resulting from the 2023 events and constrain their rupture properties. We use the GMTSAR software to process the raw data (Sandwell et al., 2011; Wessel et al., 2013; Xu et al., 2017) and construct interferometric products. We unwrap the phase using the statistical cost, network flow algorithm for phase unwrapping (SNAPHU). We utilize cross-correlation to validate the results to be more resistant to decorrelation. Our analysis of Sentinel-1 and ALOS-2 InSAR data highlights: 1) the broad coseismic deformation field from both earthquakes; 2) the presence of secondary fault structures highlighted by phase gradient processing which is sensitive to sharp changes in surface deformation. This type of feature has been linked to the activation of secondary fault structures during major events; and 3) a comparison between the abilities of Sentinel-1 (C-band) vs. ALOS-2 (L-band) to capture the large surface offsets produced by these events, particularly in the near-fault region. We find that L-band data handles the large offsets more easily and is also generally less decorrelated by vegetation and snow resulting in cleaner unwrapping results, while the C-band data’s frequent repeat passes allow for time dependent analysis of mid- and far field motion but might underestimate coseismic offsets in the near field region. The data produced in this study are free and openly available at topex.ucsd.edu.