of applications (e.g. Mitchell and Chadwick, 1998; Mills et al.,
2005). This is an automated technique which minimises
differences between overlapping DEMs by establishing point-
to-point or point-to-surface correspondences. Previous research
at Newcastle University has resulted in the development of a
robust matching algorithm, which has been shown to perform
well in the context of glacier change estimation in the Svalbard
archipelago, Norway (Miller et al., 2009). The scale-stability of
DEMsS derived from ASTER satellite imagery appears to offer a
solution for the registration of DEMs extracted from archival
aerial imagery, overcoming the requirement for ground control
and, at the same time, allows for historical ice volume change to
be robustly determined (Miller et al., 2009).
In the research presented here, DEMs extracted from USGS
archive stereo-photography of the AP are registered to DEMs
derived from modern ASTER data and aerial photography by
means of robust surface matching. A case study examining
multi-decadal glacial volume change for two glaciers within the
AP is presented, and the results are assessed to evaluate the
potential of this approach for wider implementation through
future work.
2. DATASETS AND TEST SITES
The USGS Earth Resources and Observation and Science
(EROS) Data Centre holds a collection of over 330,000 single
frame records of Antarctica from 1946-present. The data is
freely available online via EarthExplorer
(earthexplorer.usgs.gov). From this archive a minimum of three
overlapping images was selected, for two sites in the western
Antarctic Peninsula, with sufficient image contrast and clearly
identifiable stable terrain to optimise the reliability of
subsequently extracted DEMs.
Figure 1: Antarctica Single Frame Record from January 1969
(nadir image, O USGS)
Most of these archive images were acquired by the U.S. Navy
with a trimetrogon sensor configuration. This configuration is
highly suited to rapid topographic mapping and consists of one
nadir looking photograph, which is of particular interest here
(Figure 1), and two oblique looking photographs, with all three
acquired simultaneously. To enable their use within modern
digital photogrammetric workstations and to preserve
information for future generations these single frame images
have been scanned and converted to digital image format by the
EROS Data Center. In this study the highest resolution scans,
25 microns (1000 dpi), were utilised. The original format size is
approximately 9 x 9 inches. Metadata information is available
but limited. Each frame has associated information relating to
the focal length, lens type and acquisition date. Although the
calibrated focal length and lens distortion parameters were
available, more detailed camera calibration information was
lacking. The only available calibration information was that
found in a report on the calibration of military cameras (Spriggs
1966). Ground control points (GCPs) were not available. Hence
the photogrammetric processing of this data is not
straightforward.
,
,
The selected USGS image frames cover the front of Nemo
glacier (January 1969; Figure 1), located approximately
67.33 °W and 67.71 °S on Pourquoi Pas Island, and the front of
Leonardo glacier (November 1968), located approximately
61.91 °W and 64.68 °S. Both glaciers can be classified as
marine terminating glaciers and are highly crevassed at the front.
The surrounding terrain is steep and mountainous, which is
typically for most of the Antarctic Peninsula.
As reference data for the for the Leonardo glacier site, modern
ASTER satellite imagery (November 2001) was used. ASTER
provides a favourable source of data for glaciological studies
because it allows mapping in the visible, the near-infrared and
the thermal parts of the electromagnetic spectrum. Importantly,
it provides near global coverage, to + 82° latitude which makes
ASTER particularly attractive for polar studies. With a spatial
resolution of up to 15 m, ASTER also offers relatively detailed
analysis of surface processes. Utilising the along-track nadir
viewing (band 3N) and backward viewing (band 3B, 27.7° off-
nadir) imagery in the near-infrared portion of the spectrum,
DEMs can be generated and ortho-images directly extracted.
The ASTER data was downloaded from NASA's Land
Processes Distributed Active Archive Center (LP DAAC). For
Nemo glacier present-day (February 2005) aerial stereo-
photography with an accompanying 2 m DEM was available
and provided by BAS. For this data, full camera calibration
information was available.
3. DEM EXTRACTION
The ASTER data was processed in /TT ENVI 4.6.1 and the
DEM was generated from Level 1B data (Band 3N & 3B) with
the additional DEM Extraction Module (Version 4.7). A
minimum of 50 tie points between the stereo image pair was
automatically generated. The generated tie-points were
examined individually and manually corrected or removed if
necessary. The output resolution was set to 15 m. Given that no
ground control data was available, relative DEMSs, based on the
sensor attitude and ephemeris data were produced and projected
into UTM coordinates based on the WGS-84 ellipsoid.
Add
DEN
strai
For
abso
SOC
SOC
imag
sens
intei
fidu
foca
radi:
cam
Spri
fran
each
for «
CS2
intei
prin
calc
redu
coo1
fidu
not
wor
othe
Dor
crea
pixe
Displacement [mm]
For
poll
autc
rest
dele
wer
feat
stat
orie
of 1
froi