Monitoring Ground Subsidence in Shanghai Maglev Area Using PALSAR and ASAR Data

Shanghai maglev is a very fast traffic tool, so it is very strict with the stability of the roadbed. However, the ground subsidence is a problem in Shanhai because of the poor geological condition and some human-induced factors. So it is necessary to monitor ground subsidence in the area along the Shanghai maglev precisely and frequently. Traditionally, precise levelling method is used to survey along the track. It is expensive and time consuming, and can only get the ground subsidence information on sparse benchmarks. Recently, small baseline differential SAR technique comes into playing a valuable part in monitoring of ground subsidence, which can extract ground subsidence information in a wide area and with high spatial resolution. In this paper, L-band ALOS PALSAR data and C-band ENVISAT ASAR data are used to extract ground subsidence information using SBAS method in Shanghai maglev area. The results show that the general pattern of ground subsidence from InSAR processing of two differential bands of SAR images is similar. Both results show that there is no significant ground subsidence on the maglev line. Near the railway line, there are a few places with significant local subsidence rates at about -20mm/y or even more, such as Chuansha town, the junction of the maglev and Waihuan road. * Corresponding author. 1. INTRDUCTION Shanghai maglev is the first successful commercial maglev transportation system in the world. Since 2005, Shang maglev has normally given commercial service to public. The whole length of Shanghai maglev is 36km and the maximum running speed is 403km/h. As a kind of very fast traffic tools, the running of Shanghai maglev requires a stable ground support bases. However, the ground subsidence in Shanghai is a problem because of the poor geological condition and some human-induced factors. Nowadays, the average annual subsidence velocity is nearly 10mm/y in the centre downtown. So it is very important for detection of potential ground subsidence near the railway to ensure the safety of the maglev line. However, spirit levelling is very time consuming and laborious. Recently, differential synthetic aperture radar interferometry (DInSAR) technology offers a convenient and efficient method for monitoring ground subsidence. DInSAR is a new technique for earth observation, with features of large-scale (100km × 100km), high spatial resolution (20m × 20m), high accuracy (mm level). DInSAR is widely used in the field of earth sciences such as seismic default, ground subsidence, volcanic activity, land slide and so on. As a kind of very slow deformation, ground subsidence is easy to be affected by the phase decorrelation and atmospheric inhomogeneities if using DInSAR method. The emergence of time series analyzing method provides a new idea to solve these problems. The time-series processing concentrates on those points called coherent targets that maintaining good coherence even during a long observation interval. There are many methods of time series analysis in the application of DInSAR. In this paper, we choose small baseline subsets(SBAS) method which based on different master images for DInSAR. SBAS divides the SAR images of the same region into several subsets. The interferometric baselines are small in each subset, but bigger than critical baseline between every two subsets. By filtering and other method, the influence of atmospheric delay and topographic height can be removed. Then we can use SVD decomposition to get the mean velocity in Line of Sight. In this paper, L-band ALOS PALSAR data and C-band ENVISAT ASAR data are used to extract ground subsidence information in Shanghai maglev area. At first, basic rational of the SBAS algorithm is introduced. Then 17 scenes of ALOS PALSAR 1.0 data and 27 scenes of ASAR SLC data are used for taking InSAR processing respectively. The results show that the general pattern of ground subsidence from InSAR processing of two differential bands of SAR images is similar. Both results show that most places of the maglev area have no significant ground subsidence. There are a few places with significant local subsidence rates at about -20mm/y or even more, such as Chuansha town, the junction of the maglev and Waihuan road, and so on. 2. BASIC RATIONAL OF THE SBAS ALGORITHM DInSAR uses the phase difference information of two complex radar images to determine the small deformation of ground targets . The observed interferometric phase difference can be written as follows : def topo orb trop noise            (1) where def  is the phase related to surface motion, topo  is the topographic phase corresponding to curve surface of the earth and ground elevation, orb  is the phase caused by orbit errors, trop  is the phase related to atmospheric delay, and noise  is random noise due to other reasons. For two-pass DInSAR, external DEM data should be used to remove the topographic phase. ASTER GDEM products are used in this paper. The spatial resolution of ASTER GDEM is about 30m and vertical accuracy is about 20m in flatten area such as Shanghai. The phase error caused by DEM errors can be expressed as [12] 4 sin dem dem B R          (2) From this formula we can see long perpendicular baseline will introduce more topographic errors, and only images with short baseline are used to take interferometry, so as to weaken DEM influences. By equation (2), topographic errors can be removed if we know errors in height. The atmospheric delay may result in great error in InSAR results. There is currently no promising way to eliminate the influence entirely from the interferograms. As the atmospheric error is correlated in space but uncorrelated in time, most part of the influences can be reduced using filtering. Moreover, the orbit errors influences can be reduced using a polynomial fitting method. We suppose that the deformation between the two SAR images is very small so that it can be ignored to some extent. Then the phase can be expressed as: def noise      (3) Suppose we have N interferometric pairs, phase i  is acquired between two SAR images with time interval i t (in unit of year), and then the phase rate i v can be expressed as: