APPLICATION OF SURFACE MATCHING FOR IMPROVED MEASUREMENTS OF
HISTORIC GLACIER VOLUME CHANGE IN THE ANTARCTIC PENINSULA
M. Kunz**, J. P. Mills?, P. E. Miller”, M. A. King“, A. J. Fox’, S. Marsh®
School of Civil Engineering and Geosciences, Newcastle University, UK (*matthias.kunz@ncl.ac.uk)
®British Antarctic Survey, Cambridge, UK
“British Geological Survey, Keyworth, UK
Commission VIII, WG VIII/10
KEY WORDS: Glaciology, Glacier Change, DEM/DTM, Photogrammetry, Surface Matching
ABSTRACT:
Climate records show that the Antarctic Peninsula is rapidly warming. Dramatic changes in ice shelf and glacier extent have been
recorded over the last few decades. Mapping recent changes in the Antarctic Peninsula is relatively straightforward, as increased
amounts of earth observation data become readily available for scientific purposes. However, long term measurements of volumetric
changes within the region are rare and less is known of changes which have occurred over the second half of the 20th Century.
Nonetheless, historical observations are available in the form of archival aerial stereo-photography. However, extracting information
from historical data, i.e. to compare it to recent data, is not trivial, and to-date, this data source remains largely untapped, despite its
rich potential. Often such imagery is stored in non-digital format and may have degraded over time. Other problems relate to
insufficient metadata or ground control. Typically these difficulties result in poor registration of multi-temporal DEMs, which
degrade subsequent measurements of surface change. This is one of the fundamental limitations of accessing archival datasets. In this
research a least squares surface matching technique is introduced to overcome these challenges and achieve reliable registration of
multi-temporal DEMs. Historical imagery acquired in the 1960s for two Antarctic Peninsula glaciers is processed to extract DEMs,
which are subsequently compared to DEMs derived from modern ASTER satellite data and aerial photography. Through the surface
matching approach, it is shown that the registration accuracy of the historical and modern-day datasets can be improved significantly.
This enables precise quantification of glacier elevation changes on a multi-decadal time scale. Frontal glacier surface lowering of up
to 50 m was observed over the last ~4 decades. Results of this study allow a better understanding of historical volumetric glacier
changes of the Antarctic Peninsula and provide an efficient and automated method for improved DEM co-registration.
1. INTRODUCTION
The Antarctic Peninsula (AP) is one the fastest warming regions
of the planet, and has been shown to be particularly susceptible
to the effects of global warming. Increased air and sea surface
temperatures have led to dramatic changes in glacier and ice-
shelf extent in recent decades (Meredith and King, 2005; Cook
and Vaughan, 2010). Over the last 50 years more than 85 % of
244 observed glaciers showed retreat (Cook et al., 2005). An
assessment of ice mass changes in the AP is very important to
derive estimates of the region's recent and future contribution to
sea-level rise. However, records of long-term glacier mass
balance of the AP are rare, and are generally excluded from
global projections of sea-level rise (Lemke et al., 2007).
Repeated direct glaciological measurements are difficult given
the harsh polar environment and mountainous terrain. However,
airborne and satellite remote sensing offers an alternative tool
for mapping of inaccessible areas. In particular, it has been
noted that high-resolution satellite imagery, together with digital
elevation models (DEM) derived from such imagery are well
suited for the monitoring of glaciers (Oerlemans, 1998). In
order to facilitate change analysis over extended time periods, a
combination of historical and present-day stereo-imagery is
ideal.
The British Antarctic Survey (BAS) and the United States
Geological Survey (USGS) hold large archives of historical
aerial stereo-photography of Antarctica, which have been
acquired since the 1920s, and which include extensive portions
of the AP. The full potential of these historical datasets,
spanning several decades, has not yet been explored, despite the
demand for multi-decadal observations. This is primarily due to
the substantial challenge of reconciling multi-source, multi-
temporal and multi-resolution datasets in a manner which
enables reliable change avoids the
introduction of biases. Uncertainties, which may be introduced
measurement and
due to lack of ground control and other sources, will propagate
into subsequent elevation measurements, leading to decreased
accuracy in the assessment of elevation change.
Such difficulties can largely be overcome through rigorous
DEM co-registration, provided a suitable source of control can
be derived. The dynamic nature of glaciated terrain and the
difficulties regions making direct
measurement of conventional ground control points (GCPs)
near-impossible, and consequently, an alternative registration
approach must be considered. The technique of least squares
surface matching has been widely applied for registration of
photogrammetrically-derived DEMs in the context of a number
of accessing such
