underground longwall mining subsidence monitoring in New. South Wales, Australia. The mining subsidence (vertical surface deformation) was measured using ...
Radar Interferometry for 3-D Mining Deformation Monitoring Hsing-Chung Chang1, Linlin Ge1, Hua Wang2, Chris Rizos1 and Tony Milne3 Email 1
CRCSI & School of Surveying and SIS The University of NSW, Sydney, Australia 2
3
CRCSI & School of Biological, Earth and Environmental Sciences The University of NSW, Sydney, Australia
School of Geodesy and Geomatics Wuhan University, China
Abstract— Geodetic information of terrain can be measured using remote sensing techniques such as photogrammetry, airborne laser scanner (ALS) and interferometric synthetic aperture radar (InSAR). They are considered to be relatively more cost-effective than and complementary to conventional ground-based surveying methods. Our previous studies demonstrated the capability of using differential InSAR for underground longwall mining subsidence monitoring in New South Wales, Australia. The mining subsidence (vertical surface deformation) was measured using DInSAR with the assumption of negligible horizontal deformation. However, the ground surveying data shows that the underground mining activity may induce horizontal surface deformation. Therefore, both ascending and descending orbits and different swath modes of ENVISAT/ASAR data are used in this paper to quantify the vertical and horizontal vectors of mining deformation. Keywords-DInSAR; 3D; deformation; mining
I.
INTRODUCTION
Geodetic information of terrain can be measured using remote sensing techniques such as photogrammetry, airborne laser scanner (ALS) and interferometric synthetic aperture radar (InSAR). They are considered to be relatively more cost-effective than and complementary to conventional ground-based surveying methods. Our previous studies demonstrated the capability of using differential InSAR (DInSAR) for underground longwall mining subsidence monitoring in the State of New South Wales, Australia [1]. As in many other DInSAR studies, the mining subsidence (vertical surface deformation) was measured using DInSAR with the assumption of negligible horizontal deformation. However, the ground surveying data show that the underground mining activity may also induce horizontal surface deformation, or so-called strain. In contrast to mining subsidence which is at the order of tens of centimeter, the induced strain is normally sub-centimeter per meter. This paper used both ascending and descending orbits, and various swath modes of the ENVISAT/ASAR data to quantify the vertical and horizontal vectors of mining deformation.
Cooperative Research Centre for Spatial Information
Some previous research [2, 3] used DInSAR with both ascending and descending orbits of ERS-1/2 SAR imagery. Due to the small incident angle of ERS-1/2 (230), its DInSAR results are more sensitive to vertical displacement of the land surface but less sensitive to horizontal vector [3]. The variation of the swath mode of ENVISAT/ASAR provides the opportunity to investigate the 3-D surface deformation vectors due to underground mining with a range of satellite look angles. For example, ENVISAT swath mode 1~4 have various look angles ranging from 150 to 360. II.
TEST SITE
A. Southern Highlands of New South Wales, Australia In Australia, most of underground coal mines employ longwall mining technique in order to have the maximum recovery of coal. Mining depth at the test site ranges around 420m ~ 470m and it is comparatively shallow to other underground mines where the mining depths are over 1km. The reported maximum subsidence occurred during 24 hours measured by ground survey conducted by the mining company is about 1cm which is also detected in our previous study [4] using ERS-1/2 tandem DInSAR. The local land-use includes built-up areas, typical Australia bushes, farms and rugged and hilly river gorges. Due to site accessibility and compensation issues for damaged surface or near surface infrastructures, the subsidence is field surveyed weekly over urban area and less in density and coverage over other areas. This study is to apply DInSAR to derive the mining deformation map in 3D. III.
METHODOLOGIES
In the last two decades, InSAR and DInSAR demonstrated their capability to create high resolution digital elevation model (DEM) and measure ground deformation caused by either natural hazards, e.g. seismic and volcanic activities, and man-made activities, e.g. underground mining and groundwater pumping. DInSAR mainly measures the ground deformation along the look direction of radar, which is also referred as line-of-sight (LOS) or slant range direction of the radar system. In most of the studies, the assumption of
zero or negligible horizontal movement of the deformation was made so that the vertical deformation, or the subsidence, could be converted directly from the slant-range deformation. This assumption is not always true when applied to mining subsidence where the small amount of horizontal movement, or so-called strain, is detected by field surveys. As mentioned earlier, DInSAR measures land deformation along its LOS. Also, due to its small look angle, it is more sensitive to vertical than horizontal movement of the land surface. For example, one full cycle of phase change in differential interferogram is equivalent to a deformation of λ/2 along the slant range direction (where λ is the wavelength of the microwave). For ERS-1/2 and ENVISAT data, it is equivalent to a vertical displacement of 3cm or a horizontal displacement of 7.1cm. In order to resolve the horizontal deformation vector, the DInSAR results derived from both ascending and descending data are combined together. If the deformation is purely vertical, the results derived from both orbits should be the same. Otherwise, the horizontal displacement exists. ENVISAT ASAR provides the acquisition of same region from different viewing angles. So the same surface deformation can be measured not only in both ascending and descending orbits, but also at different viewing angles. This study combines the DInSAR results from 1 ascending and 2 descending orbits to reveal the 3-D deformation vectors. The deformation vector along LOS of radar system (DLOS) is a composite of up (DU), easting (DE) and northing (DN) deformation vectors. Here, DE and DN are re-projected to ground range first, then to LOS. The deformation measured in differential interferogram is the sum of vertical and horizontal deformations projected to LOS. The contributions of DU, DE and DN towards to DLOS is shown in (1).
[− cos(θ )
DU sin(θ ) cos(α ) sin(θ ) sin(α )] DE = [ DLOS ] (1) D N
where θ is the radar incident angle and α is the azimuth of the satellite heading vector (positive clockwise from north). Please note that DU is defined as negative for subsiding movement. It is preferable to choose SAR data with larger incident angle to ease high phase gradient problem in differential interferogram for monitoring large subsidence. This is demonstrated in the next section. IV.
INPUT DATA
The recent ENVISAT acquisitions over the test site are listed in Table I. The pairs with a temporal separation of 35 and 70 days (1 and 2 repeat cycles) are tested. There is only 3 day difference between the first and third pairs. Therefore,
the temporal coverage of the pairs is very suitable for our purpose for 3D deformation analysis.
TABLE I. Pass
CHARACTERISTICS OF INTERFEROMETRIC PAIRS Pair
descending ascending descending descending ascending descending
1 2 3 4 5 6
TABLE II. Pair 1, 4 2, 5 3, 6
Master dd/mm/yyyy 8/12/2006 10/12/2006 11/12/2006 12/01/2007 14/01/2007 15/01/2007
Slave dd/mm/yyyy 12/01/2007 14/01/2007 15/01/2007 23/03/2007 25/03/2007 26/03/2007
Bperp (m) 234 238 310 54 46 60
Btemp (day) 35 35 35 70 70 70
ENVISAT SCANNED MODE OF THE INTERFEROMETRIC PAIRS Swath mode I4 I3 I2
V.
Incident angle range 310 ~ 36.30 260 ~ 31.40 19.20 ~ 26.70
RESULTS
DInSAR interferograms of pair 1 to 3 in Table I are shown in Fig. 1. The upper concentric fringes in the interferograms are induced by mining subsidence; where the lower phase changes indicate the surface height change at local mine stockpile. Despite the phase noise caused by spatial decorrelation of the pair 1~3, high phase gradients within the concentric fringes induced by the mining deformation are noticed. As shown in Table II, the pair 1~3 have the decreasing incident angles from approximately 340 to 230. It is evident that the phase fringes are more distinguishable in the DInSAR result with a larger incident angle (pair1) than the ones with smaller incident angles (pair 2 and 3). Even though pair 4 ~ 6 in Table I have short baselines (
