The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008
To validate the results obtained, the ALOS PALSAR DInSAR
results are compared against ground survey data for both mine
sites. The DInSAR subsidence profile extracted from the
ground survey points (Figure 7), compared with the ground
survey data measured on 19 September 2007 and 12 November
2007 at the Westcliff Mine, is shown in Figure 8. Figure 8
shows that the DInSAR result matchs well with the ground
survey data. The subsidence profile along point LOOl - L021
has an RMSE of 0.6cm. The result suggests that this DInSAR
technique has the capability to deliver sub-centimetre accuracy.
Figure 9 shows the DInSAR height change result from ALOS
images acquired on 27 Dec 2006 and 11 Feb 2007 and overlaid
on Appin Mine plan and ground survey points (N042-N142). In
order to assess the quality of the DInSAR result it is important
to have good spatial and temporal overlap between the DInSAR
result and the ground survey data. Unfortunately the difference
between the date of the ground survey and the two ALOS
PALSAR image acquisitions is quite significant in the case of
Appin. The closest ground survey dates before and after the two
acquisitions are in 19 Oct 2006 and 15 Jan 2007 for ALOS
images 27 Dec 2006 and 06 Feb 2007 and 20 Feb 2007 for
ALOS images 11 Feb 2007. In order to compare the ground
survey data with DInSAR result, the deformation is assumed to
change linearly between the ground survey dates before and
after each ALOS acquisitions. The estimated height at 27 Dec
2006 (estimated from 19 Oct 2006 and 15 Jan 2007) is then
subtracted from the estimated height at 11 Feb 2007 (estimated
from 06 Feb 2007 and 20 Feb 2007) and the comparison to the
DInSAR result is shown in Figure 10. It can be seen that the
DInSAR result and the ground survey data agree well with most
survey points, including the area with the highest rate of
subsidence. However, most of the error has been found from
survey point number N80-N105, which is at the east of the
subsidence bowl. Up to 4cm difference has been observed from
point N80-N84. Apart from these points, the DInSAR result
follows the ground truth very well and an RMSE of 1.7cm has
been calculated.
Figure 7. DInSAR height change derived from ascending ALOS
images acquired on 29 Sep 2007 and 14 Nov 2007 (46 days
apart) and overlaid on Westcliff Mine plan.
Figure 8. Validation of DInSAR-derived subsidence profiles
against ground truth.
It is interesting to note that the subsidence bowl from the
displacement maps derived from ascending (Figure 6e) and
descending (Figure 6f) pairs are slightly different. For the
ascending pairs (Figure 6a-e), the subsidence bowl seems to
have higher subsidence in the west, whereas the subsidence
bowl seems to have higher subsidence in the east for the
descending pair (Figure 6f). This maybe caused by different
displacements vector and incidence angle between two tracks.
This may explain why the most of the error has been found on
the east of the subsidence bowl.
The RMSE 0.6cm and 1.7cm between the DInSAR-derived
results and ground survey data have been observed in the two
sites. The inconsistency between the ground survey data and the
DInSAR result can be due to several reasons: (1) difference
between the date of survey and the date of satellite image
acquisitions, (2) the uncertainties in the georeferencing, (3)
errors in the DInSAR processing, and (4) errors in the ground
surveys.
Figure 9. DInSAR height change derived from ascending ALOS
images acquired on 27 Dec 2006 and 11 Feb 2007 (46 days
apart) and overlaid on Appin Mine plan.