USING PERMANENT SCATTERER INSAR TO MONITOR LAND SUBSIDENCE ALONG HIGH SPEED RAILWAY- THE FIRST EXPERIMENT IN CHINA

Daqing Ge(1), Yan Wang(1), Ling Zhang(1), Ye Xia(2) ,Yi Wang(1) and Xiaofang Guo(1)

(1).China Aero Geophysical Survey & Remote Sensing Center for Land and Resources (AGRS), 100083, Beijing, China. Email:survery79@yahoo.com.cn

(2). GeoForschungsZentrumPotsdam, Telegrafenberg A17, D-14473 Potsdam, Germany

ABSTRACT

We present in this contribution the results of the first experiment to monitor land subsidence along JingJin High Speed Railway (JingJin HSR), which connects Beijing city and Tianjin city and is the first high speed railway in China,by using the Permanent Scatterer InSAR (PSI). Since the length of the railway is usually beyond the standard scenes coverage of the ongoing radar satellite, i.e. ENVISAT and RADARSAT SAR images which has a ground coverage of 100×100 km, we have developed techniques in this work for land subsidence monitoring of very extended areas with multi-track long strip SAR images. In particular, we have studied the method for integrating multi-track PSI for large coverage subsidence mapping. We applied the ENVISAT ASAR data acquired from Jan, 2007 to Dec, 2008 inneighboring tracks (Track-218 and track-447, Frame 2799 to 2817) to perform the PSI analysis. The PSI estimated subsidence velocity map in adjacent track have been integrated to generate the large coverage subsidence velocity map; these result are relevant to an area in the North China Plain that extends for about 200×150 km. For detailed investigation of JingJin HSR, a buffer zone with a width of 3 km in each side of the railway have been generated and the subsidence values have been extracted from the merged velocity map. The results clearly indicate the distribution and magnitude of the subsidence zone along Jing-Jin high speed railway and they demonstrate the approach to analyze the subsidence of very extended areas.

Index Terms— Land subsidence, high speed railway, Permanent Scatterer, small baseline, multi-track

1. INTRODUCTION

The deformation of large-scale linear infrastructure, such as roads, pipelines and highways etc, has become a serious problem of those man-made features. Since the damage of these features are not only loss of financial investment but also lives at many places, as where deformation was caused by underground mining of liquid or solid resources or construction of subways or reservoirs. It attracts people's

increasing attention in recent years since most of them are landmarks in the world. Conventional surveying techniquesare not enough to help because of the costs and infeasibilities at many situations. The developed SAR interferometry (InSAR), which is a microwave remote sensing technique, allows people to map large coveragetopography and monitor surface deformation by exploiting the phase difference between different SAR data acquisitions of the same interesting areas [1]. Differential SAR interferometry is the further step of InSAR technique which was firstly used for single deformation event monitoring with a short time span. Since the limitations of phase delay [2] caused by atmospheric disturbance and coherence loss induced due to large geometric difference and longer time span between images, The InSAR analysisis more suitable for cases with good coherence and slightlyatmosphere influences. The developed multi-temporal differential InSAR techniques, i.e. the Permanent Scatterer InSAR (PSI) [3-5], the Small BAseline Subset (SBAS) [6-8], overcome the difficulties in the conventional D-InSAR through analyzing the interferometric phase time series on those coherent targets that can be monitored by satellites.Since these targets remains relative high coherence within a long temporal data set, it is possible to detect and measure millimeter variations in the sensor-targets distance.High speed railway is a kind of linear feature characterized with large extent and irregular distribution. Long term of land subsidence along high speed railway threatens its stability and safely running. For well understanding the deformation of railway due to regional land subsidence, the first task is to investigate the subsidence along the full length of it. So, monitoring techniques with high precision and large coverage is necessary in practice. In this paper,we presented the first experiment study of land subsidence monitoring along JingJin High Speed Railway (JingJin HSR), which is the first high speed railway in China and connects Beijing city and Tianjin city, by using the advanced technique-Permanent Scatterer InSAR (PSI). Considering the large coverage of JingJin HSR, we proposed in this work the multi-track PSI technique to perform the investigation. The PSI result of adjacent tracks

_____________________________________________________ Proc. ‘Fringe 2009 Workshop’, Frascati, Italy, 30 November – 4 December 2009 (ESA SP-677, March 2010)

have been analyzed and the integrated velocity map significantly demonstrated the distribution and magnitude of the subsidence along the JingJin HSR. The impact of .non-uniform subsidence to railway has been evaluated.

2. STUDY AREA AND DATA SETS

JingJin HSR is the first 300 km/h trainsets for Chinese Railways, which have been entered service on the Bejing-Tianjin line in time before the 2008 Olympic Games. The total length of Jingjin HSR is more than 110km. JingJin HSR across the major subsidence zone in North China Plain(NCP). During the past several decades, the NCP has witnessed the most excessive pumping of groundwater in the world and covers the largest subsidence area with the most funnels on the planet[9]. In Beijing areas, over-pumping of the groundwater is often considered to be the dominant causes of subsidence but the impact of high-rise building and major engineering works on ground deformation can't be neglected. In recent years subsidence areas in Beijing grew rapidly since the urbanization and industrial development, especially in the east of the city where the JingJin HSR across. In Tianjin area land subsidence originated from 1920s and till now formed the most severe subsidence zone with the amount exceeds 3.1m in China[9]. Long term over-exploitation of groundwater is the major cause of this result.

Fig.1 shows the location of the JingJin HSR and the coverage of the ENVISAT SAR data. Since the railway

length is beyond the width of a standard SAR images, it needs two tracks ENVISAT data at least for the complete coverage. To this purpose, we collected the ENVISAT ASAR data acquired from Jan, 2007 to Dec, 2008 with track number 447 and 218, frame number 2799 to 2817 for the subsidence investigation.

3. METHODOLOGY

3.1 PSI processing of single track

For surface deformation monitoring application, the InSAR analysis should be the same scale in spatial and temporal domain with the deformation phenomena. Therefore, in case of large coverage land subsidence mapping, a long strip PSI processing are necessary since the coverage of single imageis limited to 10,000 or more km2. To this purpose, long strip SAR images processing are the basic step in this work. Generally we can archive this task through the PSI analysis on long strip RAW data stack with the same DopplerCentroid[10].Several steps are common in PSI processing from the initialacquisition of a set of SAR data to the final deformation map. Here we focus on the interferometric pairs stack combination and also the PS candidates (PSC)identification.In comparison with the reference PSI techniques[3-5], which utilize one common master image to generate differential interferogram stack and to carry out the estimation of the deformation parameters, we adopt a small baseline strategy which allows a relative free combination of all available images for the interferograms generation. Meanwhile small baseline combination minimize the effect of the external DEM inaccuracy and the spatial decorrelation. Moreover, the usage of multiple small baseline combination allows the redundancy of differential interferograms which increase the number of samples for the model coherence maximized and enables a better estimation of the linear velocity and DEM error from a reduced set of images.For PS candidates identification, many PSI techniques take the point-based criterion refer to Amplitude Dispersion Index [4] from a large number of SAR images. Pre-condition for the use of this method are all the well focused SAR data are accurate co-registered and well radiometric calibrated. In this work we proposed a spectral variability measure to identify point targets[5,11]. Since a band-limited SAR system acquires one part of the constant point scatter spectrum in slant range direction，the main idea to identify point scatters is to make a spectral correlation involving different looks with fractional azimuth and range bandwidth. This provides the advantage to identify PS candidate from a single SAR images as well as stack of images and relievethe dependency on data number.

Figure. 1. Study area and ENVISAT ASAR data

Once the differential interferometric phase of each coherent pixel is computed the following step is phase unwrapping soas to retrieve the real phase gradient of deformation. For properly estimation of the deformation parameter, theobvious orbit error phase trend across the interferogramhave to be removed firstly. Following that, all the PSCs were connected with the Delaunay triangle network. 2-D regression is performed to estimate the deformation difference and DEM error between each "arc" (neighboring coherent point pairs) of the network with respect to the spatial and temporal baseline[12]. Once the difference of deformation velocity have been estimated, the deformationvelocity map can be derived by integration of phase difference between each connection of the PS pair.

3.2 Integration of multi-track PSI subsidence velocity map

Since the deformation parameters (deformation velocity and time series ) of each track have been estimated in the previous step, the following is the integration of multi-trackvelocity maps. The PSI estimates of each track have their own coordinate system (datum) that is defined by the acquisition geometry of the master strip. Hence, the multi-track PSI estimates need to be referenced to a common datum before they can be integrated. Moreover, due to the redundancy is introduced in the PSI estimates, since the same deformation signal is observed from independent overlapping satellite tracks, multiple overlapping tracks can be utilized to assess the reliability of the PSI deformation estimates[13,14]. Thus, the datum connection procedures consists of the conversion of a common coordinate system and the connection of the PSI parameter estimates in the adjacent track. For the common coordinate conversion, the first step is geocoding of each single track. This is carried out by using the external DEM to transform the SAR images from radar coordinate to map coordinate. Moreover, a polynomial transformation is performed between the geocoded single track by using the tie points in the overlapped areas to transform the adjacent track to the master track we selected in advance. PSI parameter estimates connection is the merging of the estimated deformation velocity in different track into a same datum. In the previous step the coordinate conversion is completed the major task in this step is to calculate the difference between the master track and the adjacent track and then compensate the difference by using a transform function. Since in the overlapped areas in multi-track map the deformation refers to the same regimes, the PSI estimated velocity should theoretically only differ a translation because each of the track has the different reference in the procedure of deformation map generation.

Considering this fact in the PSI processing the same reference PS in the overlapped area have been identified for the calculation of deformation map of the adjacent tracks. For the deformation parameters computation over the same area, PS within resolution cell distance are weighted averaged to form the deformation to represent the same deformation regime.

4. RESULTS AND ANALYSIS

To illustrate the proposed method for the investigation of subsidence along JingJin HSR, two tracks of ENVISAT ASAR images with frame number from 2799 to frame 2817in track 447 and track 218, acquired from Jun, 2007 to Nov, 2008 have been processed. All these data have been imaged in descending track. The integrated area have been analyzed covers an area of 150×200 km in relevant to area located in the north part of the NCP, covering Beijing, Tianjin and Hebei province. The total number of SAR data for PSI processing in each track is 18 scenes and 19 scenes respectively, and each of them covers area of 100×200km. Fig. 2 presents the integrated PSI velocity map of track 218 and track 447. Datum connection between the adjacent tracks is performed in an average sense since the number of tracks is limited in this case. From this map it is evident that PS density in urban areas is very high, allowing very accurate spatial sampling, while in rural area is relatively low. Nevertheless, the subsidence map clearly describes the variation of the subsidence in this region. In particular, to Beijing area most subsidence zone is in the east and north of city and to Tianjin areas in the south and west. The factis that most of the fast subsidence occurs around the big cities. This resulted from the fact that most of subsidence zones are the developed or fast developing middle or small towns, where water consuming factories are densely distributed. Long term of ground water pumping is the major trigger of the this result. To the JingJin HSR, it just across the fast subsidence area in the east of Beijing and west of Tianjin, and non-uniform subsidence in these regions would be unavoidable to influence the stability of the railway.For well analyze the subsidence along the railway, a buffer zone have been generated. Fig. 3. indicates the subsidence buffer zone along the JingJin HSR with 3 km width in each side. PSs inside the buffer zone have been extracted from the large integrated velocity map. From this map it is evident that full coherent coverage of the railway can not be expected with low band width sensors as ENVISAT, as SAR resolutions are relatively too poor, especially in range direction. Consequently, east-west oriented linear structures are imaged poor due to the incidence orientation of thedescending image mode. while north-south oriented line aresampled best due to the high azimuth resolution (Fig 2).

Figure. 2. Integrated multi-track subsidence velocity map. ENVISAT descending images in track 218 and track 447 have been preformed PSI analysis. The integrated multi-track subsidence map is related to the north China Plain, including Beijing, Tianjin and Hebei province.

In this study, since the JingJin HSR has been constructed completely last year, many points would not be imaged coherently during the period of data acquisition. Nevertheless, Fig. 3 clearly demonstrates the subsidence along the full JingJin HSR. Two subsidence center have been found in the region JingJin HSR crossed. The maximum velocity is near 60-70 mm /a. Although the fact that full coherent coverage of the railway is not true, the impact of the regional subsidence on the railway can be investigated. By interpolating the PSI velocity value within

the buffer zone, the subsidence profile of the railway can be generated (fig.4). Since the damage of linear infrastructure is independent of the absolute subsidence, we calculated the second derivation of surface subsidence- curvature to analyze the real impact of the non-uniform subsidence. Fig. 5. indicates the curvature changes due to non-uniform of the full length. The max curvature occurs in the changes of the subsidence slope and the value is near 15 mm/km. This fact verified that non-uniform subsidence is the major reason for the potential damage to the railway.

Figure.3. The subsidence velocity map along the JingJin HSR. PSs inside the buffer zone have been extracted to generate this map. Two subsidence zone have been found and one is in Tongzhou, Beijing area and another is in Beichen district , Tianjin area.

Figure.4. Subsidence profile of the JingJin HSR. Two major subsidence zone could be found in the distance and their location is in distance 10km to 20 km and distance 95- 105 km.

Figure.5. Curvature of the JingJin HSR. This profile shows the potential deformation of subsidence impact on the railway. Non-uniform subsidence is the major reason for this result.

For validation of the integrated PSI along the railway, a comparison of leveling and PSI subsidence velocity have been carried out. Fig. 6 and tab.1 shows the significantagreement of the PSI velocity with the nearby points. Note that in this case, the small number of the available leveling data limits the comparison.

dInSAR（mm） dLeveling (mm) △d

-31.248 -27.49 -3.758

-56.045 -48.58 -7.465

-59.990 -59.735 -0.265

-56.912 -51.52 -5.392

-32.117 -33.74 1.623

-39.881 -41.89 2.009

5. DISCCUSSION

From the results presented above it is clear that subsidence along the JingJin HSR could be investigated by integrating the multi-track PSI results. The impact of the regional subsidence could be well evaluated, which provides useful

information for the safety running of the high speed railway. Meanwhile, for man-made linear features, the successful application of PSI technique depends on many factors, such as the orientation and slope of the feature, the radar look direction and amount of data acquisitions available[15].Fig.7. gives a clearly comparison of linear features with different orientation in this work. From it we can conclude that linear feature with south-north orientation is subject to be coherent in the SAR images, this is the fact that its orientation is relatively parallel with the look direction of ENVISAT descending data. Furthermore, due to corner reflector effect in many cases, man-made large features with certain height could be imaged coherently even in a long time interval. It should be noted that the dimension of the scatter and the bandwidth of the sensors are still important factors for a target being coherent.

6. CONCLUSIONS

We presented in this work the results of the first experiment of PSI for the monitoring land subsidence along the Jingjin high speed railway. The proposed multi-track PSI results integration has been applied for large coverage land subsidence mapping. The datum connection procedure resulted in a consistent set of PS velocity map in the framework of the master track, which provide the large coverage and high precision subsidence map for the analysis of the JingJin HSR. It has been indicated that the PSI is able to provide a detailed description of the subsidence along linear structures, such as high speed railways. Such observations can be used to evaluate the stability of many kinds of linear features, so as to improve hazard assessment in relation to safety running. This study verify the potential of satellite InSAR technique for deformation mapping of man-made features, due to their advantages of frequent revisits, wide area coverage, dense sampling, and high precision displacement monitoring. this is in particular in for the application of high resolution radar data since many coherent point could be identified and analyzed.

7. ACKNOLEDGEMENT

This work was jointly supported by the China Geology Survey and China National Ministry of Science and Technology in the framework of the project “Large scale land subsidence monitoring with SAR Interferometry in North China Plain” and “Research on the Monitoring of Large Scale Land Subsidence in Low Coherence Area” under the “National High Technology 863 Program” respectively. The ENVISAT ASAR data used for interferometric analysis were partially provided by ESA under CAT-1 project 3863.

Figure.6. PSI velocity and the available leveling data. The PS nearby the leveling the point have been compared.

Table.1. Comparison between PSI velocity and leveling measurement.

S

E

S

E

Figure. 7. The appearance of north-south vs west-east oriented linear feature in the PSI velocity map.

8. REFERENCES

[1] R. F. Hanssen “Radar Interferometry-Data Interpretation and Error Analysis "[M]. New York, Kluwer Academic Publishers, 2002. [2] H. A. Zebker, P A. Rosen, S. Hensley. “Atmospheric Effects in Interferometric Synthetic Aperture Radar Surface Deformation and Topographic Maps”. Geophysical Research Letter, vol. 102,pp. 7547-7563. 1997.[3] A.Ferretti, C.Prati, F Rocca, “Permanent Scatterers in SAR Interferometry”. IEEE Trans. Geosci. Remote Sensing. vol.39(1) ,pp.8-20. 2001.[4] A. Ferretti, C. Prati, F. Rocca, “Nonlinear Subsidence Rate Estimation Using Permanent Scatterers in Differential SAR Interferometry”. IEEE Trans. Geosci. Remote Sensing , vol.38(5), pp. 2202-2212, 2001.[5] W.Charles, W.Urs, W.Andreas. “Interferometric Point Target Analysis with JERS-1 L-band SAR Data”. IGARSS'03, Toulouse, France, 21-25, July, 2003.[6]O.Mora, J.Mallorqui, A.Broquetas. “Linear and Nonlinear Terrain Deformation Maps from A Reduced Set of Interferometric SAR Images”. IEEE Trans. Geosci. Remote Sensing, vol.41(10), pp.2243-2253, 2003.[7]R.Lanari, O.Mora, M.Manunta, et al. “A Small-Baseline Approach for Investigating Deformations on Full-Resolution Differential SAR Interferograms”. IEEE Trans. Geosci. Remote Sens, vol. 42 (7), pp.1377-1386, 2004.

[8] C. Francesco, M. Mariarosaria, Antonio P, et al. SBAS-DInSAR Analysis of Very Extended Areas: First Results on a 60,000-km2 Test Site. IEEE Geosci. Remote Sensing Letters, vol. 5, No. 3, July 2008.[9] Q. C, He, Land Subsidence in the North China Plain(NCP). in Proc. the 7th International Symposium on Land Subsidence, 2005, pp:18-29..[10] C. Francesco, M. Mariarosaria, Antonio P, et al. SBAS-DInSAR Analysis of Very Extended Areas: First Results on a 60,000-km2 Test Site. IEEE Geosci. Remote Sensing Letters, vol. 5, No. 3, July 2008.[11] Z. S. Rafael, P. Kostas, H. Irena, et al, “Analysis of Coherent Scatterers over Urban Areas”. http://earth.esa.int/workshops/ polinsar2005/particiants/124/paper_Paper_PolinSAR_2005_Schn-eider.pdf.[12] W. Yan, G, Daqing, H. Qiong, et al, Surface Subsidence Monitoring with Coherent Point Target SAR Interferometry. in Proc. IGARSS, 2008, pp: 1205-1208.[13] G. Ketelaar, F.van Leijen, P. Marinkovic, et al, Multi-track PS-InSAR datum connection, in Proc. IGARSS, 2007, pp: 1205-1208.[14] G. Ketelaar, Satellite Radar Interferometry Subsidence Monitoring Techniques [D]. Springer, 2008.[15] R. F. Hanssen, F. J. Van Leijen. Water Defense System Monitoring using SAR Interferometry. http://www.hansjebrinker..net/publicaties/2008.038.Eusar_Hanssen.pdf.