International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004
Whilst it was important to develop a methodology for the
continuous monitoring of subsidence, it is also valuable to
examine old photography for evidence of past subsidence. Two
sources of aerial photography were available for this research;
historical survey photography and photography flown for the
study by NERC. Historical photography for 1971 and 1996,
flown at an approximate scale of 1:10,000 was purchased from
commercial archives and photogrammetrically scanned.
Camera calibration certificates were also obtained but no
ground control points (GCPs) were available from the time of
the surveys.
The photography acquired during the September 2002 and
September 2003 surveys was controlled using phase processed
GPS. The surveys were controlled using two sources of GCPs.
Prior to the survey flights identifiable photo features were
surveyed on the ground and these were complemented by
circular white targets of Im diameter which were set out for the
duration of the overflights.
Whereas traditional photogrammetric processing relies on GCPs
to scale and orientate blocks of photography, this process is not
feasible using modern GCPs to control historical photography
in areas of potential surface instability. The historical blocks
were processed in a photogrammetric workstation to the relative
orientation stage and digital elevation models (DEMs) were
extracted using stereo autocorrelation. The absolute orientation
of these photogrammetric DEMs was achieved using a surface
matching algorithm (Mills et al, 2003). The rigorously
controlled contemporary DEM from 2002 photography
provided a control surface to orientate the historical DEMs.
The fully orientated DEMs have two applications in this
research; surface profiling and differencing. Surface profiling
involves the extraction of sections through a DEM to identify
the characteristic depressions that form when subsidence
occurs. This approach has proven particularly useful at
identifying the position of lines of fissuring. ^ Surface
differencing involves subtracting each of the historical DEMs
from the contemporary control surface. Areas of change,
including drops in the land surface due to subsidence, are
identified and can be quantified. The potential of this approach
for identifying potential pillar collapse requires further
investigation.
Spurious changes results from arcas of poor autocorrelation in
the DEM extraction process. It is most important to compare
“bare-earth’ surfaces to ensure that surface changes are real and
not artefacts of different land use, for example a field covered
with a mature cereal crop in one epoch may be ploughed on a
subsequent date giving an erroneous difference. These affects
can be minimised by user interaction in the DEM extraction
process and by careful photo-interpretation of the source
photography.
Acquisition of future aerial photography will enable the
monitoring of Houghton-le-Spring to continue and the accuracy
of the method to be quantified. It is anticipated that the
effectiveness of the surface matching process will reduce the
need for photo-control for future surveys (Mills et al., 2003).
4. SPECTRAL ANOMALIES
CASI-2 and ATM imagery from each of the three surveys was
geometrically corrected using simultaneously — acquired
navigation data and a 10 m post spacing DEM acquired by the
Ordnance Survey (OS) using NERC proprietary software. The
accuracy of this geometric correction was sufficient to locate
the footprint of individual pixels on the ground. As one aim of
this research is multi temporal analysis of the spectral response
of subsidence features it was necessary to correct the data for
atmospheric effects. At-sensor radiance values were converted
to apparent ground reflectance using the Empirical Line Method
(Smith and Milton, 1999). This process involves normalising
the spectral response of the calibration targets observed in
airborne imagery to surface reflectance measurements made
with the ASD spectroradiometer. The thermal band of the
ATM imagery was not corrected as no suitable calibration data
was available.
The mapping of soil moisture anomalies utilises the thermal
band of the ATM scanner. Exploiting the relationship between
soil moisture and temperature noted by Pickerill and Malthus
(1998), it is possible to identify arcas of anomalous soil
moisture in otherwise homogeneous areas. A qualitative
approach was applied, by extracting the pixels on a parcel by
parcel basis using field boundaries. A simple standard
deviation contrast enhancement was then applied to each parcel,
maximising the contrast and highlighting any anomalous area.
Thermal anomalies were recorded that are associated with
subsidence features observed in the field. However,
observation of the same areas on different dates demonstrated
that the thermal anomalies are affected by prevailing soil
moisture conditions and easily obscured by vegetation.
reflectance (%)
Re
|
|
45-
|
id 7
j
|
ta
/
ro. /
es y /
/
zn
ier m
P. d
|
450 550 650 750 850 950
wavelength (nm)
Typical CASI-2 vegetation reflectance spectrum
showing position of individual bands.
Figure 2.
The identification of vegetation anomalies associated with
subsidence exploits two features of vegetation spectra that are
found within the CASI-2 bandwidth of 405-945nm (Figure 2);
the red edge and the chlorophyll absorption feature.
Zarco-Tejada & Miller (1999) noted that spectral parameters
have the inherent advantage that they are relatively insensitive
to variations in illumination or inaccuracies in atmospheric
correction. This indicates that they are also an appropriate tool
for multi temporal studies. In order to investigate the most
appropriate techniques for defining red edge and chlorophyll
absorption parameters two test data sets were used; continuous
field spectra acquired using the ASD, and simulated CASI-2
band spectra derived from the ASD using published full width
at half maximum (FWHM) for CASI-2. The processing
Inter.
techn
acqui
2003
4.1
The
reflec
betwi
conci
al.,
maxi
positi
REP
senes
appro
veget
reflec
Figure
A mel
bands
bands
from
detrac
(1988)
Currar
et al. (
edge t
the RE
first de
to esti
intensi
the RE
Accure
spectra
calcula
appliec
noted 1
more tl
phenor
and rec
