Conference Agenda

Abstract:

We investigated ICEYE X-band SAR data over several sites in Europe, Asia and North America. For pairs with short spatial baselines the data were found to be well suited for interferometry (InSAR). Up to now, the satellites are not operated in narrow orbital tubes, and so pairs with short baselines (< 500m) are rare, except for the ICEYE X6 satellite that is operated in a one-day repeat orbit. ICEYE X6 data stacks could be successfully used for interferometric time series analysis, to derive terrain height corrections, displacement rates and atmospheric path delays.

1. Introduction

The objective of the Eurostars RAMON Project (2019-2022) was to design, develop and test an innovative radar-based landslides monitoring service to support different phases of the landslide risk management. The service combines existing, established elements such as landslide velocity maps derived from stacks of satellite synthetic aperture radar (SAR) data using differential interferometry (DInSAR) and Persistent Scatterer Interferometry (PSI) with near-real-time monitoring elements, as urgently required during crisis situations. The elements considered for this purpose are the use of a novel microsatellite constellation (ICEYE [1]), https://www.iceye.com) and terrestrial radars [2,3]. In this contribution, our experience with ICEYE DInSAR and InSAR time series analysis is presented.

ICEYE Polska [4] was one of the partners of the RAMON project and provided us with access to ICEYE data. For its first launched satellites, InSAR was not a priority of ICEYE. The orbit control was not optimal and thus initial pairs that were investigated clearly had baselines that were too long for InSAR. This changed with the launch of the ICEYE X6 satellite. ICEYE X6 is operated in a one-day repeat orbit. Over such a short time period, the orbital drift was typically on the order of 200m, making the data suitable for DInSAR and InSAR [NJ1] time series analysis. In the following, our processing workflow and results obtained over several sites are presented. The processing was done using the GAMMA Software [5].

2. Initial DInSAR tests

In 2020 we conducted several ICEYE DInSAR tests using data acquired by the first four ICEYE satellites. The most promising pair was between an ICEYE X2 and ICEYE X4 scene, acquired over the Brienz landslide in Switzerland, with a perpendicular baseline of 1730 m and a time interval of 10 days. At X-band, a spatial baseline of > 1km is very long and coherence was only obtained due to the broad range spectra of these spotlight mode scenes. However, the phase was very noisy and non-zero coherence was mainly observed for point-like scatterers in the village.

3. DInSAR and SBAS tests over Mojave, USA

With the launch of ICEYE X6 in late 2020 the situation improved. A stack of 51 ICEYE X6 scenes acquired with one-day time intervals and reasonably short spatial baselines (a few hundred meters between subsequent acquisitions) over the US City Mojave and the surrounding arid area could be used to assess DInSAR and InSAR [NJ2] time series analysis. Thanks to the short time intervals and reasonably short spatial baselines, the differential interferograms have high coherence over this arid site. The large stack of 2D differential interferograms was also used to assess the potential for time series analysis. For this we used a multi-reference stack of multi-look interferogram phases. During the acquisition period, the satellite orbit was first drifting about 200m per day in one direction. Towards the end of the period, the drift direction changed. Based on the stack we could determine terrain height corrections, atmospheric phases and displacement phases. The estimated terrain height corrections are found to be of high quality due to the relatively long baselines of some pairs. Despite not being able to provide a validation, the result looked adequate and clearly had a much better resolution than the 1-arc-second Copernicus DEM. The estimation error of the displacement rate, on the other hand, is high because of the short total time period covered by the data. While millimeters per year displacement rates could not reliably be determined, a salt lake area shows displacements at centimeter scale between the observed first and the last dates.

To document the coherence level, we calculated an average coherence over pairs with short baselines (up to 200m) and short time intervals (up to 3 days). This average coherence shows reduced values in the very near and far ranges of the scene. The plausible explanation of the reduced coherence is the lower antenna gain in these areas that results in a higher Noise Equivalent Sigma Zero (NESZ). This interpretation is also supported considering the backscatter level in areas where very low backscatter is expected, such as smooth water surfaces and radar shadow.

4. PSI test over Ichinomiya, Japan

A Persistent Scatterer Interferometry (PSI) test was conducted using a stack of 17 ICEYE X6 scenes acquired over Ichinomiya, Japan. The PSI result provided high quality topographic heights for the selected point-like scatterers. The calculated statistical height estimation error was between 0.1m and 0.2m. The heights could not be directly validated, but the positional accuracy of the georeferenced point-like scatterers confirms that the point heights are of sub-meter accuracy. The estimated deformation rate is not very useful in this case. Because of the short overall time span covered by the data and the high stability of the area, the estimation error is much larger than the expected deformation rates. Similar data over a fast-moving landslide or mining site with LOS displacements in the mm/day range would be very attractive.

5. SBAS test over Disko Island, Greenland

In search for a site with fast displacements within the potential coverage of the ICEYE X6 satellite, we proposed Disko Island to the west of Greenland. ICEYE collected a significant stack of scenes at one day intervals. For the interferometric time series analysis, we selected a stack of seven scenes with the sensor drifting for half of the time to one side and then back again to where it started, so that we had short baselines and varying time intervals between one and seven days. A multi-reference stack with 15 interferometric pairs was used for the time series analysis. Multi-look phases were also considered. Terrain height corrections, deformation rates and atmospheric path delay phases were estimated.

The retrieved terrain height corrections are relative to the Copernicus DEM that is based on TanDEM-X data acquired after 2010. The corrections were generally small except for the glacier and snow field areas where the height tends to reduce over time, especially near the tongue of the glaciers and lower end of the snow fields. Because of the one-day interval between the observations and the short total period considered, the analysis is mainly suited to map m/year displacement rates. On Disko Island, such displacement rates are observed for glaciers (including faster rates), rock glaciers, and fast-moving landslides. Considering individual differential interferograms confirms the location and shape of the identified displacement features. The high spatial resolution and short time interval are very useful in this case.

6. Conclusions and Recommendations

High resolution stripmap and spotlight mode data acquired by ICEYE X-band satellites is well suited for DInSAR, provided the data is acquired with reasonably short spatial baselines. So far, the satellites are not kept in narrow orbital tubes (< 500m), so the spatial baselines are short mainly for the X6 satellite that is operated in a one-day repeat orbit. Over one day, the daily orbit drift observed was typically on the order of 200m for the stacks we had access to. X6 acquisitions could be used for InSAR time series analysis. Furthermore, it was also possible to identify X6 pairs with longer temporal separation and short spatial baselines.

Stacks with short time intervals are well suited to estimate terrain height corrections relative to an existing DEM such as the Copernicus DEM. Furthermore, as clearly demonstrated for Disko Island, such stacks are also suitable for the mapping and monitoring of m/year scale displacement rates. Both multi-look and single pixel phases were successfully used in our interferometric time series analysis tests.

Using pairs with long perpendicular baselines (> 500m) is difficult and thus ICEYE data not acquired by X6 in the one-day repeat cycle did not provide useful interferometric results in our tests. Therefore, we strongly recommend to improve the orbit control, so that the satellites can be operated in narrow orbital tubes. Furthermore, we observed that the NESZ of the ICEYE SAR data is high, especially in the very near and very far range of the scene. As a consequence, the coherence tends to reduce over surfaces with low backscatter. We recommend to reduce the NESZ.

SAR sensors with meter scale spatial resolution tend to have a small swath width. Nevertheless, the combination of operating several satellites with the capability of acquiring data at different incidence angles permits obtaining data over an area of interest within a short time interval. This does not mean that interferometric data can be acquired. Between 2020 and 2022, for example, we failed to get a single useful interferogram over Switzerland.

6. Acknowledgements

ICEYE is acknowledged for providing the ICEYE SAR data used in the presented analysis. This work was supported through the Eurostars Projekt E!113220 RAMON (EUREKA, co-financed by Innosuisse).

7. References

[1] ICEYE company web-site: https://www.iceye.com.

[2] GAMMA Portable Radar Interferometer (GPRI) information brochure:
https://gamma-rs.ch/uploads/media/Instruments_Info/GAMMA_GPRI_information.pdf.

[3] GAMMA L-band SAR information brochure:
https://gamma-rs.ch/uploads/media/Instruments_Info/GAMMA_L-Band_SAR_information.pdf.

[4] ICEYE Polska company web-site: https://www.iceye.com/pl/o-firmie/iceye-w-polsce.

[5] Gamma Software Information Brochure:
https://gamma-rs.ch/uploads/media/Software_Info/GAMMA_Software_information.pdf.

We have mapped the deformation of Iceland using all available Sentinel-1 radar data (Summer/Fall 2015-2021) from three parallel and overlapping descending and three ascending orbit tracks, yielding a complete countrywide coverage for both look directions. The total number of satellite passes for each of the six orbit tracks is about 170, i.e., we used over 1000 data sets, from which we processed around 8700 interferograms (multilooked to 100 m ´ 100 m pixels). To improve the data, we employed a two-step atmospheric correction approach based on global atmospheric models and information about the stochastic characteristics of atmospheric signals. We then solved for time-series of each of the six data sets and inverted for near-east and near-vertical time-series, assuming that north ground displacements are small. Plate motions and glacio-isostatic adjustment dominate the large-scale displacements in Iceland. We can observe how the width of the plate-boundary zone varies from being relatively narrow in southwest Iceland to more distributed deformation in the east and northeast of the country. Uplift in central Iceland reaches ~3 cm/year, probably mostly due to glacio-isostatic adjustment, and it appears to increase in rate during the observation period. We modelled and removed the large-scale horizontal and vertical displacements, with a model of the plate motion, plate-boundary deformation and glacio-isostatic adjustment, to examine better local deformation signals. The residual displacement rate map shows deformation in many places, e.g., at central volcanoes and in areas of geothermal exploitation. We also observe widespread slope movements, with practically all east-facing slopes moving eastward and west-facing slopes westward. The slope movement typically amounts to a few mm/year, with faster rates at some known landslide bodies. This slope motion is seen all over the country, especially in northern and western Iceland. The signals are less clear in southern Iceland, where slopes are smaller and in eastern Iceland, where most slopes of the east-west trending fjords are either north- or south-facing. We furthermore inspect if rate changes can be observed at locations where recent slope failures have occurred, e.g., at two locations in north Iceland where mudslides caused road closures and some structural damage. However, these mudslides occurred following sudden and intense rainfall events and we see no clear speed-up on these slopes in the months before the failures. In Summary, our results show that InSAR data are effective to map country-wide ground velocities and velocity changes as well as local deformation signals and transients at volcanoes and geothermal areas. The results also show that slopes all over Iceland are subject to steady gravitational soil creep amounting to several mm/year, with higher rates observed in many areas where geomorphologically landslides can be identified in the landscape.

During the previous two decades, the DInSAR and Persistent Scatterer Interferometry (PSI)
approaches have advanced significantly in data analysis and processing techniques. This has been accompanied by an important increase in the SAR data acquisition capability by space-borne sensors.
European Union’s Sentinel-1 constellation makes it possible to monitor ground deformation over
entire continents with short revisiting time, high spatial resolution, and open data policy. This
information is used in a wide range of research areas like topography, ground surface deformation mapping, and infrastructure monitoring.
The article centers around the most significant ground motion initiative ever developed: The
European Ground Motion Service (EGMS) which provides consistent, standardized information
regarding ground deformation, related to both natural and anthropogenic hazards, at a European scale with a precision of a few millimeters and annual updating. The whole European Community can greatly benefit from a distinctive set of displacement maps, which contain mean annual velocities, displacement time series starting from 2015, line-of-sight information of both ascending and descending geometries, as well as horizontal and vertical components. A wide range of users, including public or governmental institutions, industry, academia, and citizens, can take advantage of it; however, dealing with and analyzing this massive data is difficult and time-consuming. The development of methodologies and new tools that can extract automatically information, make initial interpretations, and generate operational maps will improve the usage of this type of data.
Here we present a semi-automatic methodology to exploit and effectively utilize the wide-area
displacement maps of EGMS, with the final aim of automatically identifying buildings and urban
structures that may be at risk of damage. The potential risk of damage is based on the spatial gradients of displacement (differential deformation). In a built environment having a map of the spatial gradient of displacement is crucial because most of the significant damages to manmade structures and infrastructures are associated with high deformation gradient values. We propose two parallel approaches with different scales of analysis. The first approach starts from the automatic extraction of the Active Deformation Areas (ADA) to make the analysis at the ADA scale, using the whole information inside each ADA. The second approach, is a single-building analysis, exclusively based on the displacement information that belongs to the building. While the first approach can be widely applied over all moving areas, offering a low to medium level of information, the second approach can only be applied where single-building EGMS data is sufficient but yields a higher-detailed output.
The differential deformation is used as an intensity value to attribute potential damage classes to both ADA and buildings. Furthermore monitoring the spatial variations of deformation can support both the impact assessment of motion phenomena and the urban management and planning activities.
This work presents the methodology and the first results of its application over a region of Catalunya (Spain). Moreover, a tool that automates the process is being developed in order to apply it to the entire EGMS data across Europe.

Abstract:

This work is understood as a contribution in preparation for ROSE-L in the InSAR domain – directly addressing one of the scientific objectives of the Fringe’23 workshop. L-band satellite SAR data of the PALSAR missions are analyzed to investigate the potential of L-band Persistent Scatterer InSAR for ground displacement monitoring. We selected the Swiss Alps as the area of interest, where we have a good knowledge of ongoing processes, access to reference information and many results generated with other SAR sensors. The results obtained at L-band have clear advantages for monitoring fast-moving landslides and analyzing vegetated areas in comparison to results obtained at C- and X-band.

1. Introduction

Ground motion and infrastructure mapping and monitoring with satellite synthetic aperture radar interferometry (InSAR) has become a relevant technique in many research and professional fields. The experience gained with the ERS-1, ERS-2 and ENVISAT ASAR satellites, the wide availability of systematic Sentinel-1 acquisitions and the free and open data access all contributed to a wide acceptance of this technology. This excellent development will not only be continued but enhanced by similar datasets acquired at L-band (NISAR, ROSE-L, ALOS-4). To prepare for this, but also to complement present and past work with the C- and X-band sensors, we investigated currently available L-band data. The focus of our presented work is on landslides in the Swiss Alps using PALSAR-1 and PALSAR-2 data.

2. Data used

For the L-band data analysis we selected the Swiss Alps, an area prone to landslides and for which results are available from previous InSAR investigations. Technically, the focus was on the investigation of the potential of interferometric time series analysis methods such as persistent scatterer interferometry (PSI). Previous work [1-3] confirmed that L-band data of JERS, PALSAR-1 and PALSAR-2 are suitable for this purpose, including ScanSAR data [3]. The Swiss Alps are covered by small stacks of ScanSAR data in the descending tracks 094 and 095 (> 10 suitable scenes each) and by even smaller stacks in the ascending tracks 198 and 199 (4-5 suitable scenes). In addition, Stripmap data stacks are available for some areas, mainly for ascending orbits. Furthermore, the entire area of Switzerland is covered with small stacks (4-10 scenes, when only counting the scenes during the snow-free period) of ascending orbit PALSAR-1 Stripmap data.

PALSAR-2 ScanSAR data are synchronized (except in 2014). PALSAR-2 ScanSAR as well as combined PALSAR-2 ScanSAR and Stripmap time series can be processed [3]. ScanSAR data of PALSAR-1 are not used, as their bursts are generally not synchronized. Another relevant difference between the two sensors is that PALSAR-2 is operated in a narrow orbital tube, so that baselines are generally short (< 300m) while PALSAR-1 perpendicular baselines are up to more than 2 km.

As terrain height reference we used a combined elevation model considering SwissAlti3D heights over Switzerland and Copernicus DEM heights outside Switzerland.

3. Methods used

PALSAR-2 ScanSAR: Our objective was to consistently process multiple sub-swaths of a PALSAR-2 ScanSAR data stack to get a single displacement time series. Two alternative approaches were successfully used. In the first, the sub-swath single-look complex (SLC) images were first separately co-registered and geocoded and only after that mosaiced into a single dataset and then interferometrically analyzed. The preferred method was to first resample the sub-swath SLCs of the so-called “full aperture processor” to a common spatial grid. After that, they could be treated in the same way as Sentinel-1 data consisting of a dataset with multiple sub-swaths. Instead of multiple bursts, as in the Sentinel-1 case, the PALSAR-2 ScanSAR data set is treated as a dataset with a single burst. This permitted using available methodologies and formats. In spite of the typical ScanSAR spectrum of the SLC data, with 5 narrow azimuth bands, the spectral diversity criteria used to identify point scatterer candidates worked well. The PALSAR-2 ScanSAR mosaic SLC sizes, in FCOMPLEX format, were very large (~20 GByte). The time series analysis was done in vector data format using the GAMMA IPTA software [4] with point lists of about 50 million points. Instead of two-dimensional regressions considering the time differences and baselines, only one-dimensional regressions were used. No point height correction was estimated, which was not a problem considering the high quality of the used DEM heights and the short interferometric baselines of the considered pairs. Atmospheric path delays, consisting of a stratigraphic and a turbulent part, were estimated based on the data. In addition, an overall phase ramp removal was applied.

Combined PALSAR-2 Stripmap and ScanSAR: At a more local scale we also conducted time series analysis on ScanSAR data combined with Stripmap data acquired in the same orbit as the ScanSAR data. This worked similarly well with and without applying common band filtering – possibly because of the rather point-like scattering characteristics of the selected candidates.

PALSAR-2 Stripmap: Where available we also processed PALSAR-2 Stripmap stacks. While somewhat larger stacks were available in some cases, the spatial coverage is more limited. The processing was done in a standard approach. Steps such as phase unwrapping were typically less challenging than for C-band data.

PALSAR-1 Stripmap: Considering the good results obtained with PALSAR-2 Stripmap, we also investigated the potential of the available PALSAR-1 Stripmap mode data, as it offers similar information for a different time period. Previous results over other mountainous regions less affected by seasonal snow cover than the Swiss Alps already demonstrated the potential of PALSAR-1 using data stacks of > 15 scenes [2]. Over the Swiss Alps, this processing was nevertheless very challenging, due to the very small stacks available together with the long interferometric baselines and the SLCs in skewed geometry. Concatenating subsequent raw data files and processing the raw data with the GAMMA MSP SAR processor [4] solved the last challenge. The long interferometric baselines make the estimation of point height corrections necessary. Furthermore, the coherence over vegetated areas is lower for pairs with long baselines, resulting in more phase noise and complicating phase unwrapping procedures. For time series analysis in cases with very few scenes (e.g., 5), we used a multi-reference stack including all combinations (in this case 10). It remains unclear whether the two-dimensional regressions used to estimate point height corrections and linear deformation rates can successfully be done in this case. Using very stringent acceptance thresholds and subsequently rejecting outliers in the results permitted getting results even for these small stacks. Nevertheless, such results should be used with caution.

4. Results

A nearly complete coverage of the Swiss Alps with ground motion information could be obtained with the PALSAR-2 ScanSAR data of descending orbits, complemented in some areas with PALSAR-2 Stripmap mode results. A comparison with other results indicates a high quality for the descending orbit data, with many known landslides clearly identified. There are only few gaps for too rapidly moving landslides (> 5cm/year) and in particular for landslides characterized by strong acceleration or deceleration in recent years, such as the well-known Brienz and Moosfluh landslides, respectively. Overall, the achieved spatial coverage is excellent and even includes some forested areas. Comparing the results with the Sentinel-1 C-band-based results of the European Ground Motion Service [5] shows that faster movements (> 2 cm/year) and movements in vegetated areas are clearly better retrieved with the L-band data, in spite of the much smaller data stacks.

The PALSAR-2 ScanSAR and PALSAR-1 Stripmap mode ascending orbit results also provide nearly complete coverage of the Swiss Alps. In both cases the results are based on very small data stacks. As a consequence, these results are less reliable and include some errors. Careful reprocessing, e.g., at a more local scale, typically permits getting more reliable results. PALSAR-2 ScanSAR-Stripmap ascending orbit results were obtained only for selected regions, yet indicated a similarly high quality as for the descending orbit data. Many of the identified landslides are covered by ascending and descending orbit data, increasing the confidence in the results.

5. Conclusions and Recommendations

The results obtained over the Swiss Alps demonstrated that interferometric time series analysis methods such as PSI are applicable with PALSAR-2 ScanSAR and Stripmap mode L-band data.

The processing was found to be overall more straight-forward at L-band than at C-band. Our related explanation is that the phase error terms, such as the unmodelled atmospheric path delay, topographic phase errors, deformation phase errors, and phase noise, are rather small compared to one phase cycle, which facilitates filtering and phase unwrapping. In addition, coherence over vegetated areas is higher at L-band than at C-band. The processing was found to be quite robust even with relatively small stacks of > 10 scenes, provided the spatial baselines are relatively short < 500m. Furthermore, the ScanSAR data-based results have a very wide spatial coverage (~350 km).

Besides the simpler processing, L-band data clearly has the potential to obtain results for fast displacement rates of several cm/year and in vegetated and forested areas.

Identified limitations include a reduced precision for stable and slowly moving areas (few mm/year). Nevertheless, the work shows that lower phase standard deviation thresholds can be used compared to C-band data. Furthermore, the main error, namely uncompensated atmospheric path delay, is almost independent of the radar frequency when expressed in centimeters. The precision of the average displacement rate estimate roughly depends inversely on the total period covered by the available observations and inversely on the square root of the number of observations. A similar precision is thus expected for 36 scenes over one year and nine scenes over two years. Furthermore, the spatial resolution of ScanSAR data-based results is lower.

Based on the experience gained with the PALSAR-1 and PALSAR-2 data, we recommend to operate ROSE-L in a narrow orbital band. This optimizes the coherence, relaxes the accuracy requirements of the topographic reference, and facilitates phase unwrapping. Furthermore, we recommend an acquisition strategy that systematically acquires multi-temporal interferometric stacks over the entire mission duration for both ascending and descending orbits.

6. Acknowledgements

PALSAR-1 and PALSAR-2 data used in this work are copyright JAXA. The PALSAR-2 data were made available to us through ALOS PI Projects ER2A4N001 and ER3A4N003.

7. References

[1] Werner C., U. Wegmüller, A. Wiesmann, and T. Strozzi, „Interferometric point target analysis with JERS-1 L-band SAR data”, Proc. IGARSS 2003, Toulouse, France, 21-25 July 2003.

[2] Strozzi T., Klimeš J., Frey H., Caduff R., Huggel C., Wegmüller U., and Cochachin Rapre A., Satellite SAR interferometry for the improved assessment of the state of activity of landslides: A case study from the Cordilleras of Peru, Remote Sensing of Environment, Volume 217, November 2018, Pages 111-125, https://doi.org/10.1016/j.rse.2018.08.014.

[3] Strozzi T., R. Caduff, N. Jones, A. Manconi, and U. Wegmüller, “L-Band StripMap-ScanSAR Persistent Scatterer Interferometry in Alpine Environments with ALOS-2 PALSAR-2,” in Proc. IEEE Int. Geosci. Remote Sens. Symp., 2022, pp. 1644–1647. https://doi.org/10.1109/IGARSS46834.2022.9884743.

[4] https://gamma-rs.ch/uploads/media/Software_Info/GAMMA_Software_information.pdf, Gamma Software Information Brochure, 2022.

[5] European Ground Motion Service: https://land.copernicus.eu/pan-european/european-ground-motion-service.

Aquifers play an important role in addressing water needs worldwide, especially in countries with extensive arid regions, or with large spatial and temporal discrepancies in recharge/discharge rates, such as Mexico. When aquifers are overexploited, groundwater resources deplete, sandy layers lose storage and confining clay beds compact, causing land subsidence and induced impacts on urban landscapes. These include the development of topographic depressions, ground fissures and cracks, surface faulting, structural damage to private and public properties and infrastructure, loss of land-to-water bodies and increased flood risk.

Using ~1700 Sentinel-1 Interferometric Wide swath Synthetic Aperture Radar (SAR) images acquired over Central Mexico in 2019–2020, we perform the largest ever-made Interferometric SAR (InSAR) survey over this country, by covering a 700,000 km² area. This encompasses the whole Trans-Mexican Volcanic Belt and several major states, including Puebla, Federal District, México, Hidalgo, Querétaro, Guanajuato, Michoacán, Jalisco, San Luis Potosí, Aguascalientes and Zacatecas, and hosts a total of >85.2 million inhabitants (i.e., ∼68% of the Mexican population). By implementing the parallelized Small BAseline Subset (SBAS) InSAR approach in ESA’s Geohazards Exploitation Platform (GEP), we estimate present-day subsidence rates for ~35.7 million coherent targets and identify yet unmapped and well-known hotspots, e.g.: −45 cm/year vertical rates in Mexico City, −22 cm/year in Chaparrosa, −19 cm/year in Villa de Arista, −17 cm/year in Aguascalientes Valley, −16 cm/year east of Fresnillo, and −15 cm/year in Ciudad Guzmán. Via spatial integration within aquifer-system boundaries, we also compute yearly compaction volume rates at 321 aquifer-systems (e.g., up to −60 hm³/year at Mexico City Metropolitan Area, −43 hm³/year in the Aguascalientes Valley and in Texcoco). InSAR-derived aquifer-system compaction generally correlates well with the modelled and/or measured groundwater deficits, extractions and storage changes provided by the National Waters Commission (CONAGUA) in the latest aquifer-system management reports.

We finally derive semi-theoretical relationships between groundwater balance parameters and land subsidence for the whole Central Mexico and 3 of its main hydrological-administrative regions (i.e. VII Cuencas Centrales del Norte, VIII Lerma-Santiago-Pacífico, and XIII Aguas del Valle de México), thus enabling the assessment of ground compaction rates and volumes resulting from groundwater exploitation. These relationships could be used to inform groundwater management strategies towards adaptation to climate change and future needs of a growing population. Furthermore, we discuss how the subsidence maps derived at country scale, alongside risk maps produced at city level, prove valuable not only to achieve a holistic understanding of this geohazard that may support governmental institutions to issue new policies or assess the performance of existing ones, but also to help regional authorities in quantification of properties and population at risk, and thus optimization of groundwater resource management in light of existing and future water demands.

Reference:

Cigna F. & Tapete D. (2022). Land subsidence and aquifer-system storage loss in Central Mexico: A quasi-continental investigation with Sentinel-1 InSAR. Geophysical Research Letters, 49(15), e2022GL098923, https://doi.org/10.1029/2022GL098923