Applications: Subsidence and Landslides (1)

SAR interferometry (InSAR) is the most precise technique to measure possible range changes using SAR images. By measuring phase variations, the sensitivity of InSAR is proportional to a fraction of the radar wavelength, providing a tool for monitoring surface deformation phenomena with centimeter or millimeter accuracy. However, to accurately measure range variations, the area of interest should exhibit high coherence values in all available images. When the lack of coherence is due to temporal and geometrical decorrelation, some advanced InSAR algorithms, such as SqueeSAR, can be applied to mitigate the effect of decorrelation noise, by exploiting the coherence matrix associated to each radar target.
In some cases, however, the loss of coherence is actually due to very high displacement rates that, even if the reflectivity of the terrain does not change with time, prevent any phase unwrapping procedure. When phase data cannot be exploited, but radar returns are consistent in time, a second possibility to measure displacements using SAR images is based on the exploitation of amplitude data, providing an estimation of displacement with an accuracy ranging from about 0.01 to 0.1 pixel size, depending on the signal-to-clutter ratio (SCR) of the target area. This method, usually referred to as speckle tracking, provides the slant range and azimuth offset fields between a pair of SAR images by maximizing a normalized cross-correlation of small patches. Whenever available, even point-wise scatterers characterized by high SCR values can be exploited to provide displacement time series with centimeter accuracy, at least using high resolution X-band imagery.
The cross-correlation of amplitude images has been already used for some applications of satellite radar data, such as the monitoring of glaciers and the estimation of co-seismic deformation fields, where very high displacement values between two successive SAR acquisitions do not allow a successful exploitation of conventional InSAR algorithms. Results have already confirmed the theoretical analysis concerning the accuracy of cross-correlation techniques.
Moreover, whenever ascending and descending passes are available, amplitude data make it possible to retrieve a full 3D displacement field affecting the area of interest with immediate benefit for data interpretation and/or inversion. It should be noted that the possibility to use correlation techniques as a standard tool for displacement time-lapse analysis is becoming more and more appealing, as high-resolution SAR imagery (in range and/or in azimuth) becomes progressively available.
Finally, whenever neither InSAR nor speckle (or point) tracking works, change detection algorithms can, nevertheless, provide useful information to final users, highlighting abrupt reflectivity changes that can be of interest for different applications and explaining the lack of information provided by both phase and amplitude time-lapse analysis.
In this paper we show a gallery of examples on how amplitude data can be used in synergy with InSAR results to provide a much more valuable time-lapse analysis of the area of interest, compared to InSAR (or PSI) results only. Too often, radar specialists focus their attention either on phase data or on amplitude data only. We provide evidence of the importance of using a more synergistic approach, taking advantage of complex radar data.

The water level of the Dead Sea (Israel and Jordan) and its associated groundwater have been dropping at an increasing rate since the 1960s, exceeding a meter per year during the last decade. The extremely high density of the Dead Sea water (1.24 kg/L) induces a very shallow interface between the fresh groundwater and the Dead Sea brine. This shallow interface responds to the sea level drop and the eastward shift of the western Dead Sea shoreline by increased flushing of the coastal aquifer by fresh groundwater, dissolution of a subsurface salt layer, gradual land subsidence, and formation of sinkholes. The major source of fresh groundwater along the Dead Sea shorelines is the Judea Mountain aquifer, which provides a relatively constant eastward flux of water year-round and is replenished in the highlands 10-30 kmto the west. A secondary source of water is seasonal flash-floods in the major riverbeds, occurring several times a year during the rainy season. In the past few years a significant increase in the number of sinkholes was noticed after each flood event. The extremely large number and density of sinkholes (a few hundreds in an area of about 1 square km) that formed over short periods pose a major threat to the potash-production industry infrastructure in this area.

In this study we combine COSMO SkyMed InSAR measurements with high resolution (0.5 m/pixel) airborne LiDAR digital elevation models, sinkhole mapping, time-lapse camera documentation, and chemical and isotopic analyses of surface- and ground-water, to uncover and analyze a new mode of "seasonal" sinkhole formation. The regional long-term sinkhole forming process involves a relatively stable dissolution of a 10-25 m deep and up to 10 m thick halite layer by aquifer-fed fresh groundwater. This process accelerates and a new mode of sinkhole formation is introduced as flash-floods are drained into existing or newly formed sinkholes. Consequently, the subsurface salt layer dissolves rapidly, the overlying ground surface subsides, and salt-saturated water seeps out downstream of the drained sinkholes. The rates and dimensions of subsidence increase significantly during and immediately after the flood events, and decay exponentially thereafter. Our study demonstrates the formation of an active real-time karst system by the sinkhole-drained floodwater, an immediate (hours to days) subsidence of the ground surface in response to the subsurface salt dissolution, and subsurface water flow at velocities comparable to the flow of surface water. Inactive subsidence sites and sinkholes along the Dead Sea shorelines are reactivated following flash-flood events. The subsiding volumes at specific sites increase by more than 20 cubic meters a day following the flood events. The post-flood subsidence is restricted to the close vicinity of incised riverbeds (up to tens of meters). Finally, we illustrate how groundwater level rise during several months at the end of the winter is correlated with accelerated dissolution of the subsurface salt layer, ground subsidence and subsequent formation of new sinkholes in an area that was almost inactive for more than a decade.

While subsidence has affected Mexico City for over a century, other cities in central Mexico have been subjected to subsidence since the '80, as a result of their large urban expansion, population increase and aggressive groundwater extraction rates. The continuous subsidence results in severe damage to urban infrastructure and civil structures. Unfortunately the damage cost assessment and vulnerability are difficult to evaluate, because of the variable geographic extent and the continuous nature of the process, which have different characteristics than other phenomena characterized by localized, short duration events such as earthquakes or floodings with better media coverage.

 In Mexico, subsidence induced damage to houses or other urban infrastructure is not eligible for federal emergency relief funds, because it is not considered a naturally occurring hazard but rather an anthropogenically induced process. Furthermore, this phenomenon is localized and usually managed only at the local city or county-level administrations. Most other large impact natural hazards, as earthquakes, volcano eruptions, or flooding events, are usually managed by federal agencies.  These agencies are responsible for hazard monitoring efforts, disaster response, and also for vulnerability and risk assessments. This work is aimed towards a better recognition of subsidence as a major hazard in Mexico, assess the number of inhabitants affected and the spatial extent of the subsiding areas affected by this process. Space geodetic techniques, most notably InSAR are quite suitable for this task.

 A recent InSAR analysis in central Mexico shows that land subsidence can be observed in 17 cities. InSAR provides an unsurpassed synoptic view of the earth's dynamic surface. However, different satellite sensors and sometimes widely spaced data availability make it difficult to derive long-term time series, or measure rapid changes or nonlinear variations of subsidence velocities. To alleviate this situation, higher temporal resolution subsidence observations of associated fault motion has been pursued using continuously operating GPS stations. We have developed a GPS network that covers 6 urban centers to detect short duration variations using different processing schemes that include both real-time solutions using RTNet as well as daily solution using Gipsy-Oasis.

InSAR time-series derived ground subsidence velocity maps (Figure 1a) were processed using a horizontal subsidence gradient analysis approach (Figure 1b) that provides an excellent tool to discriminate areas with high potential of surface faulting. Further data merging with population, hydrology and meteorology data sets allow the creation of risk maps using a simple risk matrix (Figure 1c). The results are cartographic products that are integrated into a geographical information system suitable for distribution to local, state and federal as a geocoded, high-resolution product suitable for detailed spatial analysis and better hazard assessment. These products provide decision elements for water resource management agencies, or risk management purposes, such as vulnerability for shallow faulting, and land use zonations. We will provide examples of these hazard assessments for several cities, including Mexico City, Aguascalientes and Morelia, and the challenges encountered to integrate these results with other data sets from federal and state organizations.

Grass is perhaps the most difficult land cover when it comes to measuring surface deformation. Especially pasture on drained peat soils suffers from very strong temporal decorrelation. Yet, in low-lying coastal regions the deformation of peat soils is important. Peat is composed of organic materials which oxidize and emit greenhouse gases when exposed to the air. In the Netherlands, peat oxidation contributes for 1 - 3% of the annual greenhouse gas emissions. Peat soils with a subsidence rate of 10 mm/y equate to an emission of about 22 tons of CO2 per hectare/y.

Oxidation of peat soils results in volume reduction and subsequent subsidence. As a result, the thickness of the vadose zone decreases, as the land surface gets closer to the phreatic zone or groundwater level.

Consequently, to keep the land sufficiently dry to be used as pasture, the soil needs to be drained, resulting in an increased vadose zone thickness, in more oxidation, and therefore more subsidence. This loop is bound to continue until the peat soils have disappeared completely. In addition to oxidation, the peat soils are consolidating when drained, as the weight of the increased thickness of the vadose zone increases the effective stress of the saturated peat. Many parts of these regions are currently situated below sea level and are expected to subside significantly in the coming decades. In addition, there seems to be seasonal deformation caused by the change of the groundwater level between the seasons; low in summer and high in winter.

Measuring subsidence rates in pasture on drained peat soils is difficult, if not impossible, with conventional geodetic means as soft soils make it impossible to install fixed benchmarks for repeated surveying. Moreover, conventional InSAR cannot maintain coherence due to the very fast temporal decorrelation in the area. The absence of coherent point scatterers prohibits the use of PSI, and even when buildings are found in the area, it is likely that their deformation will not be representative of the subsidence in the center of pasture fields.

Here we tackle this problem using a combination of adaptive coherence estimation, to optimally estimate the interferometric phase, and a parametric deformation model combined with a multisensor SBAS approach for the joint estimation of the deformation parameters in section. The method is tested on a multisensor data set over a pasture area in the Netherlands.

Application of the method on real data shows that the pasture area is subsiding on average 36.6 mm/y, albeit with significant local variability, that seasonal variation has an amplitude of 10.6 mm, and that the maximum of the seasonal height occurs very homogeneously in time. Application of this technique over larger areas in the Netherlands will have important consequences for water management and sustainability.

During the round table, seed questions proposed by the chairs will be discussed with the audience.

The ongoing pattern of groundwater induced land subsidence in major valleys and agricultural regions of Iran has been recently documented by several studies using C-band Interferometric Synthetic Aperture Radar (InSAR) observations. In this article we present the results of our research in which we evaluated the performance of C-band, L-band and X-band SAR data, using time-series method of small baseline subset (SBAS), to retrieve long time series of ground subsidence in agricultural regions in the country. Two major groundwater basins have been selected for this purpose: (1) Rafsanjan Valley in the Kerman province of central Iran and (2) Tehran Plain (capital of Iran). We also report on our experience using dual-polarimetry (HH/VV) X-band SAR data for Persistent Scatterer (PS) deformation analysis in natural terrains subject to high rate of deformation.