PSInSAR Multi-Sensor Fusion-Based Deformation Analysis for Long Term Monitoring of Urban Areas
Patrascu, Carmen1; Datcu, Mihai2

Long term monitoring of interest regions for displacement assessment using Persistent Scatterer (PS) techniques is based on the existence of large image data stacks, usually spread over several years [1]. In this paper we perform such long term monitoring of urban areas in order to estimate deformation trends. The proposed study will be performed over a period of 20 years, using data acquired with multiple sensors (ERS 1/2, TerraSAR-X). The fusion of the results for the overlapping time periods commands special attention, due to the differences in both resolution and wavelength of the satellites [2].

Bucharest is the capital and largest city in Romania, also ranking among the top 10 largest cities by population in the European Union. Since the beginning of the '80s extensive construction has been carried out in the region, comprising of apartment buildings, several industrial sites, the Palace of Parliament and an underground metro network that is still expanding. Moreover, since the beginning of the Bucharest metropolitan area project in 2003, the number of construction sites has rapidly increased, due to new residential areas (Figure 1), overpasses and subway stations being built. Another important fact that needs to be considered is the geographical localization of the city. Being traversed by the Dambovita River, with some construction being carried out in previously drained regions and with high underground water levels, the risk of subsidence in the area is significant. Furthermore, due to its closeness to the Vrancea seismic area, Bucharest is also prone to seismic induced deformation.

Figure 1 Changes can be observed in the Bucharest area. ERS 1 (a) and TerraSAR-X (b) images acquired on 28.08.1992 and 31.08.2011, respectively, show an important increase of the inhabited perimeter of the city

The deformation analysis is performed by extracting displacement profiles using the Persistent Scatterer (PS) approach. Unlike traditional differential interferometric techniques, this method performs well even for data stacks spanning long time intervals, as is our case.

Stable point candidates will be selected using two different approaches, both having a stack of co-registered SLC products as input. The first method exploits the spectral behavior of stable points and its differences from normal distributed targets'. The second approach selects low intensity variation points, which is the normal response of single dominant scatterers in a resolution cell. Due to the fact that urban areas are analyzed, PS techniques represent a good choice for the maximization of the number of acquisitions, without affecting the targetsí density. For this case, a-priori information related to the region of interest will be used to develop structural models for the estimation of the deformation signal. The data used for this study consists of two independently acquired datasets. The first image stack is formed of ERS 1/2 images acquired between 1992 and 2011, while the second one contains TerraSAR-X StripMap images acquired between 2007 and 2013, as can be seen in Figure 2.

Figure 2 Time intervals covered by available data

Two deformation trends will be computed. The first one covers a period of 19 years and corresponds to the ERS data, while the second one spans a 5 year interval and is derived using the available TerraSAR-X acquisitions. An extended analysis over the overlapping time period will be performed with the purpose of linking the previously obtained deformation trends, thus providing a continuous monitoring of the interest regions over a period of 20 years. The fusion of ERS and TerraSAR-X products for long term monitoring will also establish an important headstone for the integration of upcoming datasets that will become available with the planned launch of Sentinel-1 in 2013.