Multi-temporal InSAR evidence of ground subsidence induced by groundwater withdrawal: the Montellano aquifer (SW Spain)
- First Online:
- Cite this article as:
- Ruiz-Constán, A., Ruiz-Armenteros, A.M., Lamas-Fernández, F. et al. Environ Earth Sci (2016) 75: 242. doi:10.1007/s12665-015-5051-x
- 265 Downloads
This study uses the InSAR technique to analyse ground subsidence due to intensive exploitation of an aquifer for agricultural and urban purposes in the Montellano town (SW Spain). The detailed deformation maps clearly show that the spatial and temporal extent of subsidence is controlled by piezometric level fluctuations and the thickness of compressible sediments. The total vertical displacement measured with multi-temporal InSAR, between 1992 and 2010, is 33 mm that corresponds with a decrease of 43 m in the groundwater level. This technique allows monitoring the evolution of settlement related to water level fall in an area where subsidence has not yet been reported by population or authorities through infrastructure damages and to discuss the effect of the aquifer recovery. This information is, therefore, valuable for implementing effective groundwater management schemes and land-use planning and to propose new building regulations in the most affected areas.
KeywordsInSAR Radar interferometry Subsidence Deformation SW Spain Aquifer
Natural hazards and their relationship with land-use planning have been the main focus of research together with the development of methodologies to assess them (Mitchel 1998). Among others, ground deformation has been reported as a main task in planning present-day human settlements (Wang et al. 2004). World population increases has compelled to the occupation of less suitable territories. In this context, ground subsidence induced by intensive aquifers exploitation to meet the needs of the rapid evolving industries and urbanization has become particularly relevant (Hu et al. 2004; Perissin and Wang 2011). This phenomenon, the consolidation of the aquifer system (aquifers and aquitards), has a huge socio-economic impact as it affects wide areas and many cities along the world—Mexico D.F., Shangai (China) or Bangkok (Thailand)—(Osmanoglu et al. 2011; Chai et al. 2004; Aobpaet et al. 2013). Precisely evaluating long-term response to pumping and recharge is of great importance to avoid heavy expenses to the local or national administrations due to damages to infrastructures, decrease in water resources or water contamination (Ortiz-Zamora and Ortega-Guerrero 2010).
InSAR data applied to ground displacements, when combined with information about groundwater levels and management practices, have contributed to the hydrogeologic understanding of aquifers (Amelung et al. 1999; Galloway and Hoffmann 2007; Davila-Hernandez et al. 2014). General subsidence due to water withdrawal has been previously described in Spain at the cities of Murcia (Rodríguez Ortiz and Mulas 2002; Tomás et al. 2005) and Granada (Sousa et al. 2007, 2010, 2011; Fernández et al. 2009), at rates of up to 10 mm/year.
The aim of this contribution is to determine the pattern and timing of the aquifer system response to pumping and successive natural recharge in an intensively exploited aquifer in the Montellano area (SW Spain). We focus on the relationship between lithology and differential settlement through the analysis of the results obtained with the multi-temporal InSAR (MTI-InSAR) technique. The dataset consists of 51 ERS-1/2 SAR and 20 Envisat ASAR scenes acquired from June 1992 to July 2000 and from February 2003 to September 2010, respectively. These results are compared with the current stress history of soils in the area, expressed as the preconsolidation stress and the overconsolidation ratio (OCR) to establish the future behaviour of soils related to the pumping activity.
Geological and hydrogeological setting
The Montellano aquifer (Fig. 2), with an outcropping surface of 23 km2, is divided into two sectors (Durán-Valsero et al. 2003). The eastern or carbonate one is constituted by up to ~600–800 m of Jurassic carbonate rocks (6 km2) confined toward the west by an aquitard mélange unit mainly constituted by clays and gypsum (IGME 1988). The western or detritic sector is formed by 17 km2 of Mio-Pliocene to Quaternary sediments in contact with clays and gypsum at its base. Its stratigraphic sequence starts with Upper Miocene sandy marls and calcarenites (~20 m) and Lower Pliocene yellow sands and calcarenites of up to 60 m thick. Over them, there are ~20 m of Middle Pliocene green clays superposed by ~20 m of Upper Pliocene limestones with gastropods. Finally, at the top of the sequence, there are alluvial and colluvial sand, silts and gravels of Quaternary age.
Meteorological and hydrogeological data
Time evolution of the piezometric level (Fig. 3a), together with meteorological data, rainfall and accumulated deviation of monthly precipitation with respect to the average (Fig. 3b), is analysed from 1992 to 2013 to infer interconnection with the deformation pattern. The data come from the nearest pluviometric station of the Junta de Andalucía network, located in Los Molares, 20 km northwestward of Montellano. The accumulated deviation of monthly precipitation (blue line) gives an estimate of the relative deficit or surplus of precipitation. To calculate it, the average monthly precipitation is obtained for a long-term period (here, 1992–2013) and then this value is successively subtracted to the precipitation value for each month.
Several drought periods could be easily recognized along the trend of the rainfall accumulated deviation: 1992–1995, 1998–2001 and 2004–2009. Concurrently, the piezometric evolution shows a particular hydrological behaviour. First of all, an evident decaying tendency for the piezometric level is observed in the piezometers located at the detritic aquifer (P1, P2 and P3) together with a fast recovery after rainy periods (1996–1998 and 2002–2004). Secondly, a very stable groundwater level is registered in the piezometers located at the carbonate aquifer.
For the first one drought period, the water level decrease was registered at P1 piezometer (13 m). During the other two drought periods, the decrease was 15 and 37 m, respectively, with a mimetic pattern in the P1, P2 and P3 wells. Over the whole period, from 1993 till 2009, the excessive pumping of the aquifer caused an average decrease of 45 m in the groundwater levels as observed at P1 (Fig. 4). For P2–5, piezometric data available are discontinuous because of the change in the water management agencies responsible of water level records (IGME, TRAGSA, CHG), thus we are not able to precisely address the contribution of each extraction well to subsidence.
ERS-1/2 SAR data for the Montellano area (Track 94 descending, Frame 2859)
Envisat ASAR data for the Montellano area (Track 94 descending, Frame 2859)
Geotechnical parameters of the sediments in the study area
% < 0.08 mm
Clays, sands and gravels
To estimate ground deformation Δϕdisp, the different phase contributions have to be determined by means of differential processing. Δϕflat and Δϕtopo depend on the perpendicular baseline and can be calculated accurately using precise satellite orbit data and a digital elevation model (DEM). However, large temporal and satellite separation between both SAR acquisitions (perpendicular baselines), as well as DEM errors, add decorrelation noise due to a change in scattering characteristics of the ground objects over time. On the other hand, atmospheric artefacts Δϕatm, that is, the atmospheric conditions in both troposphere and ionosphere, can vary considerably between both SAR acquisitions causing phase disturbances. The latter is only significant for dispersive frequencies, like L-band SAR systems, and is not applicable to our study in which C-band data from Envisat and ERS is used. The bibliography on SAR interferometry is extensive. A detailed review of the main principles can be found in Bamler and Hartl (1998), Massonnet and Feigl (1998), Bürgmann et al. (2000), Rosen et al. (2000), and Hanssen (2001).
Time series InSAR techniques address the problem of decorrelation by identifying a small subset of radar targets, called “Persistent Scatterer” (PS) pixels (also referred to as a “Permanent Scatterer”), that exhibit a stable phase characteristic in time, leading to an improved signal-to-noise ratio. PS pixels often correspond to point-wise scatterers or manmade objects on the Earth’s surface (building, metallic objects, exposed rocks, etc.) which dominate background scattering and maintain reflective characteristics along the series of SAR images. Time series InSAR methods include the persistent scatterer interferometry (PSI) methods and small baseline (SB) techniques.
Because PSI and SB approaches are optimized for resolution elements with different scattering characteristics, they are complementary, and techniques that combine both approaches are able to extract the signal with greater coverage than either method alone (Hooper 2008; Ferretti et al. 2011). In our study we use StaMPS-MTI. The Stanford method for persistent scatterers (StaMPS) is a software package that implements a PSI method developed to work even in terrains devoid of manmade structures and/or undergoing non-steady deformation (Hooper et al. 2012, 2013). In addition, StaMPS-MTI (multi-temporal InSAR) is an extended version of StaMPS that also includes an SB method and a combined multi-temporal InSAR method, allowing the identification of scatterers that dominate the scattering from the resolution cell (PS) and slowly decorrelation-filtered phase (SDFP) pixels, that is, pixels whose phase when filtered decorrelates little over short time intervals (Hooper 2008). The PSI technique identifies phase-stable PS pixels using primarily correlation of their phase in space, not requiring any approximate model of displacements. The SB method uses amplitude dispersion values and then identifies the SDFP pixels performing phase analysis in space and time. Finally, both selections (PS + SDPF) are combined and a 3D phase unwrapping algorithm is applied to isolate the deformation signal, based on these pixels. The inner workings of this software package are described in more detail in Hooper (2008, 2010, 2006) Hooper et al. (2004, 2007), and Sousa et al. (2010, 2011).
A DEM with 25 m resolution provided by the Instituto Geográfico Nacional de España was used to remove the contribution of the topographic phase to the interferometric phase. Using highly precise orbit data for ERS-1/2 and Envisat satellites calculated by TU Delft (Scharroo and Visser 1998) and ESA, the reference phase was computed and subtracted from the interferometric phase.
For the SB approach, Figs. 1 and 2 of the Appendix show the residuals between the unwrapped phase of the SB interferograms and the estimated SB interferograms when redundancy was removed by inverting to a single master network first. No spatially correlated residuals are identified indicating the absence of unwrapping errors. A sophisticated tropospheric correction like in Bekaert et al. (2015) is not being needed as delay variation over the city is expected to be small.
Geotechnical characterization and tense-deformational behaviour
Preconsolidation stress (σ′p) is the maximum effective stress that a soil has suffered throughout its history. From a geotechnical point of view, it is relevant because it differentiates elastic and reversible deformation from inelastic and partially irreversible deformation revealing the starting point of high compressibility. It also allows us to predict long-term consolidation and differential settlements (Jamiolkowski et al. 1985). The preconsolidation stress for 50 undisturbed samples of sediments and carbonates from the Montellano aquifer area has been studied using the uniaxial consolidation swelling test in accordance with the appropriate international standard ASTM D 4546. We have applied the Casagrande (1936) graphical method, while using the analytical procedure of Sridharan et al. (1991). In addition, to avoid subjective interpretations of the maximum curvature point (more important when more ductile is the material), we have applied the methodology described by Gregory et al. (2006). Triaxial water permeability test, due to aquitard character of most of the samples, was performed as defined by ASTM D 5084, except for the limestones. In the rest of the tests performed we have followed the procedures proposed by UNE-AENOR.
The overconsolidation ratio (OCR) is another useful parameter to complement the tense-deformational information of a soil. This value is the ratio of preconsolidation stress to current natural overburden stress and indicates whether the soil is overconsolidated (OCR >1), normally consolidated (OCR = 1) or underconsolidated (OCR <1). A soil is said to be overconsolidated when it has been exposed to vertical effective stresses higher than the ones acting at present. In this study, we performed consolidated and undrained (CU) triaxial tests in accordance with the appropriate international standard ASTM D 4767. The results show OCR ratio values varying from 0.89 to 1.19, with an average value of 0.90. Accordingly, soils seem to be normally consolidated to slightly underconsolidated close to the surface (unless the green clays samples that provide overconsolidated values). Taking into account the lithology distribution in the surroundings of the P1 well (where the maximum vertical displacement is observed) a subsidence of up to 30 cm is expected, one order of magnitude higher than the vertical displacement estimated with InSAR for the same location. However, it is important to note that this maximum expected subsidence should be higher than the real subsidence as it is evaluated considering surface conditions and for a total compression of the sediments.
Decrease in pore-water pressure leads to compression of the granular structure of the aquifers (coarse grain deposits) and/or aquitards (silt and clay), and indeed to consolidation of these compressible sediments (Leake 1990; Galloway et al. 1998). In the Montellano aquifer, the decline in pore-water pressure related to groundwater pumping likely controls the land subsidence revealed by the INSAR data in the period 1992–2010. Unfortunately no ASAR data have been acquired after 2010, due to the unexpected loss of contact with Envisat. Associated with this decrease, there was an increase in effective stresses in the subsoil, causing consolidation of the clayey sediments. Analysis of these results has allowed the identification of two main sectors with different ground surface behaviour. The southern sector, placed on recent, deformable quaternary colluvial sediments deposited over Triassic clays and the northern one, over colluvial cemented sediments overlying Middle Pliocene and Jurassic limestones.
Although deformation in an aquifer system is generally elastic and recoverable (Poland 1961), aquitard sediments respond both elastically and inelastically depending upon whether the maximum preconsolidation stresses are exceeded when the piezometric level decreases (Xue et al. 2005). They then undergo a slow, accumulative and irreversible rearrangement of their pore structures. With long-term water-level decline, the aquitards will continue to exhibit delayed drainage and residual compaction even though groundwater levels in the aquifers may have recovered (Helm 1984; Bell et al. 2008). Such delayed compression is not so evident in the pattern of residual subsidence found in the Montellano aquifer. In 1995, rainfall recharge produced the arising of the groundwater level and this effect was observed, with a small time lag, at the deformation time series.
In many of the affected areas, subsidence has been finally reported after the observation of some local problems as underground utility lines cracking, seawater intrusion or settlements of buildings and civil infrastructures (Tomás et al. 2010; Jiang et al. 2011). Anticipation to this problem will avoid the spending of great amounts of money, as direct and indirect economic costs, in addressing them. Indeed, much of the subsidence occurring in aquifers subjected to high pumping activity is irrecoverable, resulting in a loss of aquifer storage. This is why the multi-temporal InSAR technique could act as an alarm to draw attention to the excessive withdrawal of groundwater (Chai et al. 2004; Shi and Bao 1984) in areas filled with compressible deposits and thus prone to land subsidence and where these effects are not so evident yet, as in the Montellano area. However, the applicability of this methodology is subject to the spatio-temporal distribution of the piezometric data and the location of geological columnar sections. This information is frequently provided by the national Geological Surveys (as the Instituto Geológico y Minero of Spain in http://info.igme.es/catalogo/?tab=2) but its availability differs from one country to another.
Estimation of the spatio-temporal distribution of the settlement is crucial to locate the most damaged areas, to predict their evolution and to establish counter measures to eliminate, or at least mitigate, the causes (Bell et al. 2008; Chaussard et al. 2014). In the Montellano sector, subsidence rates are likely representative of the amount of water extracted and the thickness of the compressible deposits. The southern half of the village is affected by subsidence in the range of 33 mm. The first human settlements were located over the stable Middle Pliocene limestones, in the north part of the present-day village. However, expansion since the 1980s has extended the building area towards the South over uncemented colluvial sediments that overlie Triassic clays, causing consolidation of the clayey sediment layers. Characterization of the subsidence distribution will help to reshape the future urban development plans and better define the urban expansion and land uses.
We applied a multi-temporal InSAR technique to analyse ground subsidence due to intensive exploitation of an aquifer for agricultural and urban purposes during the 1992–2010 period in the town of Montellano (SW Spain). We demonstrate that this technique allows monitoring the evolution of settlement related to water level fall in an area where subsidence has not yet been reported by population or authorities through infrastructure damages. In time, we found a correlation between the piezometric level changes and subsidence as observed from InSAR. In addition, interrelation among subsidence and shallow rock distribution has been demonstrated showing that deformations are higher where Triassic clays are thicker in respect to areas where Middle Pliocene or Jurassic limestones crop out. Our study provides valuable information that can be used to implement effective groundwater management schemes, land-use planning, and to propose new building regulations in the most affected areas. The obtained results provide very useful spatial and temporal data about the incidence of intensive pumping at a low cost.
SAR data are provided by the European Space Agency (ESA) in the scope of 9386 CAT-1 project. This research was supported by PRX 12/00297, ESP2006-28463-E, Consolider–Ingenio 2010 Programme (Topo-Iberia project) CSD2006–0041 (Consolider), AYA2010-15501 projects from Ministerio de Ciencia e Innovación (Spain). In addition, it was supported by the RNM-282 and RNM148 research groups and the P09-RNM-5388 project from the Junta de Andalucía (Spain). The first author has been also funded by a Juan de la Cierva grant (JCI-2011-09178) from Ministerio de Ciencia e Innovación. Interferometric data were processed using the public domain SAR processor DORIS and StaMPS/MTI. The DEM is freely provided by © Instituto Geográfico Nacional de España. The satellite orbits used are from Delft University of Technology and ESA.