Introduction

Land subsidence is a global phenomenon and is defined as:

“a gradual settling or sudden sinking of the Earth's surface due to removal or displacement of subsurface earth materials”  - National Oceanic and Atmospheric Administration (2021)

Subsidence can result from numerous natural and anthropogenic processes, many of which are water-related (Figure 1). A study of 290 subsidence cases by Bagheri-Gavkosh et al (2021) found that human-induced subsidence accounted for 76.92% of global cases, and that groundwater extraction contributed 59.75% of these! 

Subsidence can result from numerous natural and anthropogenic processes, many of which are water-related (Figure 1). A study of 290 subsidence cases by Bagheri-Gavkosh et al (2021) found that human-induced subsidence accounted for 76.92% of global cases, and that groundwater extraction contributed 59.75% of these! Figure 2 highlights this positive correlation.

Subsidence is intensifying due to a combination of factors including population growth, climatic conditions, increased economic activity, urbanisation, and lack of pumping regulations. All are contributing to increased demand for groundwater supplies. Heavy extraction of groundwater causes subsequent falling water levels that lead to deformation. The deformation is sometimes unnoticeable until devastating consequences like buckling bridges, cracks on buildings, earth fissures, sinkholes, or frequent flooding have occurred (Figure 3).

Where does subsidence occur?

Those areas most susceptible to subsidence (especially groundwater induced) tend to be low-lying, flat areas with unconsolidated sediments in alluvial basins or coastal plains. Urban areas and agricultural regions especially in drylands are at risk due to high population density and heavy use of groundwater for irrigated agriculture, respectively. (Herrera-García et al. 2021). Some places are underlain by susceptible to dissolution rocks, i.e., carbonate rocks (limestones and dolomites) or salt rocks that are particularly at risk of sinkholes.

A map of potential at-risk areas of subsidence was created by Herrera-García et al (2021) (Figure 4). The cumulative potential subsidence area amounted to 2.2 million km2 (1.6% of global land), affecting 1.2 billion people (19% of the global population) (Herrera-García et al. 2021).

Groundwater and subsidence

As already alluded to, ground deformation is closely related to the groundwater condition. Groundwater can be linked to subsidence in two ways: through the lowering of the water table and by the erosion of underground rock and soil.  

Lowering of the water table and subsequent decrease in fluid pressure in aquifers can lead to the gradual lowering of a land surface as sediments compact and also to sinkholes as the decrease in pressure can make the rooves of any underlying cave systems (caverns) susceptible to collapse. This lowering of the water table can be caused by over-abstraction of groundwater, drought, or dry seasons (Nelson 2016).

Jakarta, Indonesia’s capital, for example, is subsiding at an average of 1-15cm/year, with models estimating that by 2050, 95% of the north will be submerged below sea level (Figure 5). Subsidence is already damaging infrastructure,  increasing flooding areas, destroying local groundwater systems, and increasing inland seawater intrusion (Ng et al. 2012). This sinking is in part due to excessive groundwater extraction coupled with the underlying unconsolidated swamp land and exacerbated by very poor extraction regulations (Mei Lin and Hidayat 2018). 

The other cause of subsidence is the erosion of the underlying geology by water. It is the most common cause of the ground's sudden collapse. Leaking sewer/water pipes or the dissolution of carbonate rocks (CaCO3) or salt rocks (NaCl) can create such underground cavities. Both carbonate rocks and salt are susceptible to dissolution through chemical weathering by groundwater or injected fluids from mining which can open up large cavities. Sinkholes can subsequently form by roof collapse either by enlargement of caverns over time or by the lowering of the water table (Nelson 2016).

Managing subsidence 

Whatever the origin of subsidence, it requires accurate monitoring to mitigate the consequences. This requires detection and monitoring of subsiding areas, understanding the factors that control the phenomenon, characterising and modelling its evolution and predicting its impact especially in densely populated urban areas.

Measurement, monitoring and mapping land subsidence uses various ground-investigation methods (incl. modelling, levelling, geological surveying, (Global Positioning System (GPS)) and remote sensing methods (Interferometric synthetic aperture radar (InSAR) and Light Detection and Ranging (LiDAR)). Due to the time consuming, labour intensive and costly nature of traditional ground-based methods, remote sensing techniques (satellite and radar), developed over recent decades have revolutionised the study of land subsidence.

Space technologies to monitor subsidence

Earth observation techniques provide myriad of opportunities for subsidence monitoring, from detecting land deformation, compiling sinkhole inventories, monitoring the evolution of active sinkholes, and identifying environmental risk factors to land subsidence (Theron and Engelbrecht 2018). 

Earth observation for compiling sinkhole inventories

Sinkholes are one of the most dangerous potential consequences of land subsidence. Predicting and mitigating the risk of sinkholes requires sinkhole inventories. This relies on two aspects: the detection of buried cavities that may cause future collapse and an inventory of sinkholes that formed in the past. In-situ methods, including geophysics and interviews, have typically been used. However, remote sensing, which covers larger areas, much faster has applications here.

The remote mapping of sinkholes relies on the detection of their distinct geomorphological characteristics (incl. circular/oval shape, concentric cracks along the outer edges, basin-like depressions on the surface), either by feature extraction on optical imagery or terrain models analysed for such features. Remote sensing images are the simplest way to detect such features. However, when such geomorphological features are not detectable on photographs due to erosion or sedimentation fill, infra-red sensing on drones can be used.This use of infrared takes advantage of the fact that depressions are colder than their surrounding environment at night, with drones being useful for allowing close proximity sensing.

Terrain models have also been used to extract the characteristic geomorphic expressions of sinkholes. They can be created many ways, from topographic maps generated by surveying techniques, terrestrial airborne laser scanning, drone-based Structure from Motion (SfM), and also from InSAR. This latter satellite technology was used to generate the global elevation models from Shuttle Radar Topography Mission (SRTM) and the World Digital Elevation Model (WorldDEM), the latter of which has a high enough spatial resolution (12m) and absolute vertical accuracy (6m) to compile sinkhole inventories more successfully.  

Whilst space-based technologies have their place in sinkhole inventories, their use is limited, firstly where erosion and sediment fill have masked the normal geomorphological expressions of these events and secondly due to their spatial resolution and absolute vertical height accuracy generally being lower than typical sinkhole feature dimensions. This means that airborne platforms are often favoured.

Earth observation for identifying subsidence risk factors

Although space-based technologies may be limited in sinkhole inventories, they are very useful for monitoring potential causes. Risk factors for subsidence, as already discussed include leaking pipework, groundwater abstraction, mining, and tectonism. Remote sensing can be used to assess the presence of certain risk factors, and their relationship with gradual subsidence and sinkhole formation.

Pipe leaks for example can be detected by monitoring vegetation growth using vegetation indices (VI) on high resolution optical data. A study in the “Pyla” area of Cyprus used visual interpretation of a high resolution QuickBird image of a major rural pipeline to determine the location of leakage by detecting and studying vegetation formed as a result of the leakage (Figure 6). The Normalised Difference Vegetation Index (NDVI) in the area was calculated and definitions of thresholds for the NDVI values were defined. Using other vegetation indices and Principal Component Analysis (PCA), the researchers could then define the problem areas (Agapiou et al. 2016).

Furthermore, groundwater storage changes, can be estimated using GRACE water mass estimation data. A study in Mexico, showed the potential of mapping groundwater depletion by combining GRACE data with InSAR-derived ground displacement data. GRACE’s resolution is too low and the inversion of InSAR data into volume of groundwater storage loss requires extensive, often unavailable data. InSAR data was used successfully to focus GRACE-derived groundwater storage maps up to a resolution close to the groundwater management scale, providing quantitative information about the occurrence of groundwater depletion. Whilst its validity is limited, the study shows promising first steps towards creating fully-geodetic groundwater depletion maps (Castellazzi et al. 2018).

Earth observation for precursory deformation 

The final and perhaps most widespread way in which Earth observation techniques can be used in subsidence management is by detecting and monitoring small scale land deformation. Such movement can enable mitigating and remedial measures to be put in place, and can also forewarn of potential collapse, allow evacuation prior to collapse, and predict future damaging scenarios.  

Whilst ground deformation has traditionally been monitored in-situ using spirit levels/inclinometers, extensometers, geodetic measurements, laser transmitters and receptors, they are very spatially limited and labour intensive (Gutiérrez, Cooper, and Johnson 2008). LiDAR and SAR(Synthetic Aperture Radar) are two alternative, remote sensing technologies that can be used for monitoring deformation over large areas, whilst GPS/Global Navigation Satellite System (GNSS) can also allow 3D monitoring of points based on continuous or repeated surveys and are often used to complement SAR observations. Table 1 highlights the merits of some of these various technologies.

LiDAR is an active remote sensing technique that generates topographical information and has shown promise for detecting land movement. However, its use for deformation monitoring is challenging as a.) typical rates of subsidence are within the error limits of LiDAR collections and b.) the instrument needs to be within 10 – 100m from the ground to reach necessary accuracy, limiting its wide area applicability (Intrieri et al. 2015). Whilst LiDAR’s use may be limited, the use of SAR technology has a long history, with its first practical use demonstrated in 1992 when the ERS-1 satellite capture surface deformation caused by the Landers, California earthquake. SAR technology includes ground-based InSAR airborne SAR, and especially space-borne SAR data focussing on DInSAR (differential SAR interferometry) (Theron and Engelbrecht 2018). InSAR has resulted in more detailed images because it uses interferograms that are made by differencing successive SAR images taken from the same orbital position but at different times. InSAR can measure ground motion in high spatial resolution over a large coverage area, able to monitor movements of the land surface in the scale of millimeters. Usually, the use of InSAR in groundwater monitoring is combined by GPS and extensometer observation (Feifei et al., 2015).

There are several examples of where InSAR and LiDAR have been used successfully together to detect land deformation. One of these is in the Dead Sea, Israel, (Figure 7) where over 6000 sinkholes have been identified along the shores. The water level here is falling at a rate of >1m/year, exposing a 10,000 year old halite (salt) layer beneath the shoreline to water unsaturated with halite, leading to dissolution, cavities, and eventual collapse (Nof et al. 2019). Deformation rates of some subsidence basins here, monitored using DInSAR, have been recorded at rates of over 60mm/year (Baer et al. 2002)!

A lineament of sinkholes along a major transport route running parallel to the Dead Sea was studied by Nof et al. (2019). InSAR monitoring showed minor subsidence of <0.2mm/day in 2012, accelerating to 2mm/day during the following three years, until cracks and depressions began to appear (Figure 8). In-situ investigation revealed a 9m deep sinkhole beneath the road. This early warning provided by the InSAR data, meant that an alternative road could be constructed and no severe disruptions occurred, highlighting the importance of such technology (Nof et al. 2019).

Similarly, Venice is one the most extensively studied cities for ground deformation. Monitoring was initiated in the 1960s using spirit leveling, yet over the last decades, space technologies have been increasingly deployed to observe this phenomenon (Tosi et al., 2018). Between 1950 and 1970, subsidence reached 10 cm in Venice, and was caused predominantly by groundwater exploitation. However, in the 1970s, Venice started to regulate groundwater extraction and diversify their water supply. Artesian consumption declined from 500 l/s in 1969 to 170 l/s in 1975, allowing the land to rebound by 2cm  in 1975 (Gatto and Carbognin, 1981). Whilst Venice still struggles with ground deformation and sea-level rise, this city is an example of how simple measures such as the regulation of groundwater extraction can have a positive result in combating land subsidence.

Conclusion

Subsidence is a global phenomenon and therefore a global concern in need of global solutions. Whilst sometimes natural in origin, human-induced subsidence is by far the most common, with human activities sometimes even contributing to the so called natural causes. In order to manage subsidence and mitigate the risks, traditionally in-situ methods have been used (see column on the right in Table 2).

However, as technology has progressed, remote sensing is able to provide more accurate and timely monitoring. Monitoring deformation now largely relies on integrating reliable InSAR data with sparse but highly accurate GNSS measurements. As many causes of subsidence are preventable, the continued use and improvement of space-based technologies would be an asset, especially for the mining sector, water managers, and construction industry to reduce the risks currently faced.