Detection of ground movements in Montjuïc (Barcelona) using TerraSAR-X data

Satellite-based radar interferometric techniques represent a powerful tool to measure and monitor ground displacements (Massonnet and Feigl 1998; Hanssen 2001). The focus of this paper is on a particular implementation of Persistent Scatterer Interferometry (PSI), a technique which is based on the use of a stack of images acquired by a satellite over a given area at different times. The term PSI is used in this work to indicate a number of different techniques including the Permanent Scatterers approach (the first PSI technique proposed by Ferretti et al. 2000), other techniques based on Persistent Scatterers (PSs), those based on distributed scatterers (Berardino et al. 2002), and other hybrid methods. PSI has been widely used for the detection, characterization, and monitoring of different phenomena such as subsidence (Ferretti et al. 2004; Dixon et al. 2006; Galloway and Burbey 2011), slope instability, and landslides (Farina et al. 2006; Colesanti and Wasowski 2006; Lu et al. 2012), deformation in urban areas (Ng et al. 2012; Lubitz et al. 2013), etc. The PSI technique allows obtaining a comprehensive outlook of the deformation phenomena occurring in wide areas while, at the same time, maintaining the capability to measure individual features such as buildings or infrastructures. However, there are some limitations of the PSI technique worth mentioning: (1) the PSI technique relies on coherence and, therefore, is opportunistic, which means that it is only able to estimate deformation over the available PSs, i.e., those points where PSI phases maintain good quality over time to get reliable deformation estimates, e.g., see Hanssen (2001); (2) PSI suffers limitations in its capability to measure “fast” deformation phenomena due to the ambiguous nature of PSI observations, which are 2π-wrapped. Although it is difficult to quantify what “fast” means, in the case of TerraSAR-X data, differential movements of up to 0.77 cm of displacement in 11 days between pairs of PSs can be measured (Crosetto et al. 2010); (3) the spatial sampling is variable and, as the availability of PSs depends on coherence, in urban areas the sampling will probably be reasonably good but PSI tends to fail in vegetated and forested areas (Ferretti et al. 2000; Berardino et al. 2002). In addition, PS locations are not known before processing; and 4) the deformation measurements are made in the direction of the Line-Of-Sight (LOS) of the satellite (Hanssen 2001).

In this study, 28 TerraSAR-X images spanning the period December 2007 to November 2009 were used. These images were processed using the PSI chain of the Geomatics (PSIG) Division of CTTC (Crosetto et al. 2011; Devanthéry et al. 2014). The PSIG processing was performed for the entire metropolitan area of Barcelona leading to the detection of several areas affected by superficial displacements, including the hill of Montjuïc where some areas affected by ongoing superficial displacement were found. The results of the PSIG processing of the TerraSAR-X data for this hill are shown and discussed in this article.

The hill of Montjuïc is a broad shallow hill (height 184.8 m a.s.l.) situated in the urban area of Barcelona (Catalonia, Spain), on the northeast coast of the Iberian Peninsula (Fig. 1). On a flat coastal area of approximately 5 km wide (NW–SE direction) by 30 km long (NE–SW direction), limited by the mountain range of Collserola to the west, by the Llobregat river delta to the southwest and by the Besòs river to the north. This plain covers an area of 170 km² of which 101 km² are occupied by the city itself.

From a geomorphological point of view, Montjuïc represents an “island” among the flat surrounding areas that are characterised by recent alluvial sediments, mainly of the Llobregat river delta in the southern part and of other minor streams coming from the surrounding hills (Serra de Collserola) in the northern part.

The slope angle map (Fig. 2b) and the profile A–B show the morphological aspect of the hill, characterized by steep slopes on the side facing the harbour and in the southern part and a relatively flat top.

The lithology of the hill is characteristic and different from the surrounding reliefs such as Serra de Collserola, the hills of Monterols, El Putget, El Carmel, and la Rovira, that are composed by metamorphic rocks and sandstones of the Paleozoic age (between 500 and 300 million years) (IGC 2010).

Montjuïc geology is characterized by marine facies of shallow waters (Alfaro et al. 2001), consisting of a thick strongly silicified detrital succession (Parcerisa Duocastella et al. 2001) in particular Miocenic arcose sandstones with levels of marls with abundant fossils of marine fauna and plants, locally covered by more recent fluvial terrace deposits. The thickness of these materials exceeds 200 m, as can be seen in the scarp facing the harbour.

The Miocenic sandstone is commonly named “Montjuïc Sandstone”(Lourenço et al. 2004), and it is mainly composed of quartz (80–90 % volume) and feldspars (orthoclase and plagioclase <10 % volume); to a minor degree it contains fragments of rock (schist, quartzite, etc.; between 0.5 and 2 % volume) and muscovite (Iglesias-Campos et al. 2014).

Due to its lithological composition, an intense mining activity was carried out on the hill, with the development of many sandstone quarries that were, for a long time, the main source of building material for the city. However, these quarries deeply changed the morphology of the hill until they were closed in the early seventies of the twentieth century.

The hill is considered a very important site in Barcelona because it currently hosts several monuments and tourist attractions such as the botanical garden Costa i Llobera, or sport venues built for the 1992 Olympic Games, such as the Palau St. Jordi sporting arena, the Pèrez de Rozas baseball stadium, and the Olympic Stadium.

In this study, 28 SAR scenes acquired in StripMap mode and ascending orbit by the X-band (3.1 cm wavelength) sensor mounted on the TerraSAR-X satellite were exploited (Table 1). The pixel footprint of these images is approximately 2.5 m in range (the direction perpendicular to the orbit) and 3.3 m along-track.

The scene covered by the processed images is approximately 35 by 30 km. The PSIG chain (Crosetto et al. 2011; Devanthéry et al. 2014) was used to generate a set of 56 full resolution (multilook 1 × 1), multiple-master interferograms spanning the period between 30/12/2007 and 22/11/2009, and to perform the PSIG processing over the full scene (Fig. 3a). The selection of the interferograms was based on minimizing the correlation between the perpendicular and temporal baselines in order to increase the sensitivity to topographic height, minimize the amount of decorrelation due to phase gradients, and maximize the effectiveness of the phase unwrapping. A perpendicular baseline threshold of 500 m was set for this analysis. A precision of 0.07 pixels in azimuth and range was achieved in the co-registration of the images used to generate the interferograms. A spatial low pass filter was used to estimate the atmospheric phase screening (APS), which was subsequently removed from the interferograms. The PSIG analysis was performed at full resolution and, at such, a dispersion of amplitude (DA) threshold of 0.3 was used to select the potential Persistent Scatterers (PSs).

The PSIG analysis generated more than 5.4 million PSs. For each PS the main estimated parameters include the average displacement rate or deformation velocity, the deformation time series, and the so-called residual topographic error. This latter parameter plays an important role in the accurate modelling of the PSI observations (i.e., the PSI phases) and for geocoding purposes. The magnitude of the topographic phase component in the PSI phase is reduced by simulating a synthetic topographic phase using a Digital Terrain Model (DTM) of the observed scene, in this study the 30-m DTM generated at the Cartographic Institute of Catalonia. However, any difference between the true height of a PS and the DTM height generates the Residual Topographic Error (RTE) component, which has to be properly modelled, estimated, and separated from the deformation phase component. The estimated RTE is used to obtain an improved geocoding of the PSI products. In fact, the standard PSI geocoding only employs the DTM to geocode its products and does not consider the RTE, which results in geocoding location errors. By using the estimated RTE this kind of error can be largely reduced, and a more precise geocoding achieved. The main parameter considered in this work is the deformation velocity.

To better highlight significant superficial displacements, a 7 classes colorbar ranging between −10 and +3 mm/year was chosen to display the PS velocities. Furthermore, based on field validation surveys and taking into account the overall distribution of velocity values in the area, values between −1.5 and +1.5 mm/year were conventionally associated to negligible or absent surficial movements.

Points with positive velocity values (blue) indicate displacements towards the satellite, measured along the line of sight, while points with negative velocity values (red) indicate displacements moving away from the satellite, measured along the LOS.

The PSIG analysis revealed the existence of some areas showing remarkable surface displacements in Montjuïc (Fig. 3), one near the Palau St. Jordi sporting arena and the Pérez de Rozas baseball stadium (Site 1 in Fig. 3), and another in the area of the Costa i Llobera Gardens (Site 2 in Fig. 3).

In such areas, displacements with velocity values of up to several millimetres per year were measured along the Line-Of-Sight (LOS) of the satellite, and moving away from the satellite (orange and reddish points). These results are discussed in the following paragraphs. Figure 4 shows the RTE map for the Site 1. The precision of the RTE estimates is of the order of 2 m (1 σ). Considering the acquisition geometry, this value results in a precision of the geocoding (affecting mainly the E–W direction) of 2.65 m (1 σ).

This study has been focused on the analysis of PSI results obtained for the hill of Montjuïc (Barcelona, Spain). In particular, the results were generated from 28 Stripmap TerraSAR-X images, covering the period from December 2007 to November 2009, processed with the PSIG chain.

The PSI velocity map has proven to be very effective in order to detect superficial deformation processes, allowing the detection of deformation processes on the top and the eastern flank of the hill of Montjuïc. Such deformation processes, ongoing during the period of acquisition of the satellite images, were subsequently characterized and interpreted by analysing the measured velocity values and performing field surveys. In the area surrounding the Palau St. Jordi sporting arena and the baseball stadium, named Site 1, the analysis of historical images has been crucial to understand the causes of the deformation, related to the compaction of the garbage landfill in an ancient sandstone quarry. Unfortunately, further information about the exact thickness of the landfill and the depth of the underlying bedrock were not available since they can be obtained only using geophysical methods, but they were not considered for this project.

The deformation process ongoing in Site 1 shows displacement, highlighted by cracks on the surface and damaged streetlights and road floors, with velocities of up to 12 mm/year (in the LOS direction).

In the second site (Site 2), near the gardens of Costa i Llobera, lower displacement rates (5–7 mm/year) were detected in two distinct areas related to different causes. Unfortunately, the presence of dense vegetation did not allow obtaining a uniform distribution of PSs on the entire areas of interest. Nevertheless, taking into account the information provided by the velocity map and performing a focused field survey, the two moving areas in Site 2 are interpreted as slope processes. The first one (Site 2a) is larger and mainly caused by instability of the slope deposits at the foot of a rock cliff, and the second one (Site 2b), smaller in size, is due to the instability of a landfill with a small car park on top.

On the whole, this work contributes to providing a further example of the effectiveness of the integration between high-resolution PSI and historical data and maps, to detect, measure, and interpret different types of ground deformation phenomena.