Subsidence/Landslides 2

Our study aims at assessing the potential in combining Synthetic Aperture Radar (SAR) interferometry data and time series of high-precision (sub-mm) leveling measurements to generate maps of land subsidence with a high spatial coverage, spatial resolution, and accuracy. The study is based on ERS-2 and Sentinel-1 imagery and offers highly important information for climate adaptation measures in Denmark.

We focus on the urban town Thyborøn, which is situated on a sandy coastal barrier on the high-energy North Sea coast by the entrance to the Limfiord. The town is low-lying and for the most part resting on landfills. Starting in 1954, the Danish Geodata Agency has conducted repeated leveling measurements to a number of height benchmarks throughout the town and surrounding area. Derived land subsidence rates have shown an acceleration between 1954 – 2012 and 2006 – 2012, and recent rates yield subsidence of up to 7 mm/yr. Combined with a rising sea level and presumably more frequent storm surge events, this calls for measures to be taken to prevent future flooding incidents. Variations in land subsidence at local scales have a significant impact on the performance of the underground waste water pipeline system as the slopes of the pipelines may become inverted. Additional effects are changing surface patterns of water accumulation and run-off.

Leveling measurements are carried out over a number of reference points distributed throughout Thyborøn. Elevation differences are estimated relative to stable reference points, for which the vertical velocities correspond to the modelled glacial isostatic adjustment. The measurements have a sub-mm accuracy hence making out an ideal dataset for the derivation of land subsidence. However, such data are limited in time and space and thus may not be applicable for change detection throughout the observation area. A high spatio-temporal coverage is important in order to implement the correct counter measures to the changes.

One solution for achieving this may be subsidence rates derived from time series of SAR interferometry data. Hence, we derive such rates from ERS-2 SAR Image Mode and Sentinel-1 Interferometric Wide Swath (IWS) Mode data using Persistent Scatterer Interferometry. The inclusion of Envisat data to ensure a near-continuous time record might have provided important knowledge on the change signal. However, too few images were available to generate a reliable signal.

ERS-2 data over Thyborøn are acquired in the period 1995 – 2001, while Sentinel-1 data are available from 2014. The SAR images have a spatial resolution of 30 m (IM) and 5 x 20 m (IWS), respectively, while the repeat cycle is increased from 35 days for ERS-2 to 12 days for Sentinel-1. Hence, the inclusion of data from both sensors allow for assessing not only a change in trends but also the advantages achieved with the higher spatio-temporal resolution.

Validating the results against trends derived from leveling data will demonstrate whether an equally high accuracy may be obtained with SAR interferometry; if this is the case, results from the two types of measurements may be merged, yielding maps of subsidence rates with both a high spatial coverage and accuracy. With the launch of Sentinel-1B further increasing the repeat cycle to six days, there is therefore great potential in a future, continuous mapping of present-day subsidence rates. This will greatly assist the authorities in their implementation of climate change adaptation measures.

Satellite radar interferometry is a well-documented technique for the characterisation of ground motions over large areas. However, the measurement density is often constrained by the land use, with best results obtained over urban areas. The Intermittent Small Baseline Subset (ISBAS) method was conceived to improve the coverage of the standard SBAS methods over non-urban classes so that areas exhibiting intermittent coherence are considered alongside those displaying full coherence. The result provides improved coverage and  characterisation of ground motion over areas affected by instabilities.  The greatest impact of the ISBAS implementation is in rural areas by showing ≥20% increase in coverage of the derived ground motion information if compared to SBAS conventional methods which provided limited results due loss of coherence.

The technique has been already successfully tested over the UK, Italy, Netherlands and Indonesia and won the 2014 Copernicus Masters competition as innovative solution for European business and society based on Earth observation data.

The applications encompass mining, civil engineering works, tectonic and natural hazards, and exploit L-band (ALOS), C-band (ERS and ENVISAT) and X-band (COSMO-SkyMed) Synthetic Aperture Radar sensors. For these case studies the outcomes supported research carried out by using both traditional and new mapping technologies, including digital stereoscopic aerial photo interpretation, digital field data capture, terrestrial LiDAR and differential GPS.

Furthermore, ISBAS is currently exploiting the free and open access data from Sentinel-1A satellite which carries a C-band radar sensor suitable for InSAR analyses and has recently started acquiring images over many areas of Earth. It is poised to provide  an invaluable source of information on land motions and geohazards, and the ISBAS technique will help to ensure that InSAR is possible over a greater portion of the Earth’s surface.

Ground subsidence induced by water overexploitation of underground reservoirs is a common problem happened in many cities over the word. In Ho Chi Minh City (HCMC), the rapid increase of ground water use started in late 1990s. This results in the water table has been lowering up to 2 m/year, leading to the serious subsidence of some areas. In response to this problem, the knowledge of the ground subsidence such as their spatial extent and their temporal evolution is essential.

This work has presented an advanced Persistent Scatters Interferometry analysis, to provide unprecedented spatial extent and continuous temporal coverage of the subsidence in HCMC by using 40 stripmap Cosmos SkyMED (CSK) X-band and TOPS Sentinel-1 C-band 20 images acquired from 2014 to present.

The analysis was carried out by using the Irstea TomoSAR toolkit, which supports the entire processing from SAR, Interferometry to Tomography (so called TomoSAR), created by D. Ho Tong Minh [1]. The TomoSAR toolkit is designed to push the services of Irstea GEOSUD and the THEIA Land Data Center. For external, it opens to provide service solutions under collaborations.

The study shows that the performance of stripmap CSK and TOPS Sentinel-1 is quite similar and effective to detect the subsidence phenomena. Subsidence is most severe in the Holocene silt loam areas along Sai Gon river and in the Southwest of the city. The urbanization which has been considered through groundwater extraction and urban growth factors, is mainly responsible for subsidence. The estimation of the average subsidence rate is -8.0 mm/yr with the maximum value up to -30 mm/yr, similar with the previous study using ALOS PALSAR [2].  

This work was supported in part by Agenzia Spaziale Italiana (ASI), in which Cosmos SkyMED images were provided under project COSMO-SkyMed-Open ID226.

[1] D. Ho Tong Minh, N. Baghdadi, P. Maurel, “TomoSAR:  the Irstea toolkit for Stimulating the use of TOPS Sentinel-1 and BIOMASS data”, ESA LPS 2016

[2] D. Ho Tong Minh, Le Van Trung, T. Le Toan. "Mapping ground subsidence phenomena in Ho Chi Minh City through the radar interferometry technique using ALOS PALSAR data". Remote Sensing, vol.7, pp.8543-8562, Jul. 2015

Sinkholes (also known as dolines) are a hazard across many parts of the world, with considerable impacts on local communities and infrastructure, and the potential for fatalities. Commonly, these are related to karst dissolution processes in areas underlain by carbonates or evaporites. They may also be caused by suffosion or piping of sediment, or collapse of existing void spaces. Although most sinkholes occur naturally, others can be caused or influenced by human activity, for example water leaks or collapse of abandoned mine workings.

Regions prone to sinkhole formation can often be identified based on underlying geology and groundwater conditions, however prediction of exact sinkhole locations and occurrences is exceedingly difficult. Given the potential risks to lives and property, there are obvious benefits to any advanced warning of sinkhole formation.

InSAR measurements have been used to successfully detect subsidence around a number of existing sinkholes. These show the extents of instability often extend well beyond the edge of visible surface features, and continue for years or even decades after the initial collapse.

In a number of locations, archive InSAR studies have also revealed precursor subsidence signals, prior to sinkhole formation. Understanding these signals, and the processes which affect the progression and timing of sinkhole development, is key for understanding and mitigation of the hazards.

A key case study comes from the Permian Basin, Texas. In 2002 a sinkhole formed outside the town of Wink, with an approximate diameter of 250 m. This is located less than 2 km from a previous sinkhole, which formed in 1980. No suitable archive SAR imagery is available spanning the period of sinkhole formation, however ALOS PALSAR data are available spanning the period 2007-11, in addition to a pre-collapse dataset from ERS spanning the period 1993-2000. A small stack of recent SAR images from Sentinel-1 are also available.

The post-collapse ALOS dataset shows significant displacement detected across an area of approximately 2 by 1.2 km, extending primarily to the north and northeast of the sinkhole. The rate of deformation is variable, but reaches some tens of centimetres per year during some periods.

Analysis of recent Sentinel-1 data shows this subsidence is still occurring, more than a decade after sinkhole formation. Furthermore, 25 years later there is also evidence for ongoing subsidence surrounding the original sinkhole, formed in 1980, showing the longevity of the associated subsidence phenomena.

Importantly, the pre-collapse data from the 1990s show significant precursor deformation across the sinkhole area and surroundings, dating back at least as far as 1993, nine years before sinkhole formation.

InSAR clearly has the potential to detect precursor subsidence which may correlate with sinkholes, identifying locations with high potential risk. However, sinkholes may also pose some challenges for InSAR monitoring, particularly those occurring over small spatial scales, or with fast or strongly variable subsidence rates. In addition, some sinkholes have been shown not to exhibit precursory subsidence, instead occurring as a sudden collapse.

The Wink study will be compared and contrasted with InSAR work across a number of other sinkholes across the world, to highlight examples of the opportunities and challenges of sinkhole monitoring using InSAR.