Yaron A. Felus
MULTI-SOURCE DEM EVALUATION AND INTEGRATION AT THE ANTARCTICA TRANSANTARCTIC
MOUNTAINS PROJECT
Yaron A. Felus and Beata Csatho
Byrd Polar Research Center
Scott Hall Room 108, 1090 Carmack Road
The Ohio State University
Columbus, Ohio 43210-1002, USA
felus.2@osu.edu ; csatho@ohglas.mps.ohio-state.edu
Working Group IV/2
KEY WORDS: Data fusion, DTM/DEM/DSM, Geophysics, Mathematical models.
ABSTRACT
Digital elevation models are essential tools in many glaciological studies and especially for mass balance studies,
structural geology modeling and advance remote sensing and geophysical processing. However, due to the hostile
climate and inaccessible environment of the Antarctic continent, there are insufficient elevation databases and their
quality is poor. In this paper, we analysis the spatial distribution of error in the different DEMs that exists at the
Antarctica Transantarctic Mountains. Based on this analysis, we investigate the various methods to combine elevation
models with different properties (resolution, horizontal and vertical accuracy). There are five major data sets in the
project area: The USGS 1:50000 maps which, covers the north west part of the project area and have 50 meter contour
line interval, USGS 1:250000, taken from the Antarctic Digital Database, which, covers all our project area and have
200 meter contour line interval; satellite radar altimetry data derived from ERS-1 with 5 km resolution; airborne Radio-
Echo Sounding profile data at the north east part of the project collected by Scott Polar Research Institute and field
surveying control points collected by USGS.
Our final goal was to compile all those elevation models into one uniform grid elevation model with the highest
accuracy and resolution that can be obtained. Many techniques and algorithm’s exists for integrating database, some are
based on interpolation methods in the boundary zone, other techniques perform simple data merging and apply various
filtering functions to make the transition smoother. We review those procedures and compare their properties and apply
some of them in our study. Last, we propose a method to combine the different DEM into one set using universal
Kriging concept. In this process, we compute a covariance matrix for every data set individually and a cross covariance
of the individual data set in the predication computation.
1 INTRODUCTION
11 The Tamara Project
The Tamara project is an international research aimed at integrating new aeromagnetic data, acquired by a cooperative
U.S.-German field campaign, with satellite imagery, geological and structural mapping, and existing ground-based,
airborne and marine geophysical data. With this comprehensive database we hope to answer outstanding questions
about the evolution of the Transantarctic Mountains (TAM) - West Antarctic rift- in southern Victoria Land. The
foundation of this database is a Digital elevation model (DEM) which is an essential tools in many glaciological studies
and especially for magnetic and gravity modeling. It is important to use a data set which will have the best accuracy and
with the highest resolution. However, due to the hostile climate and inaccessibility environment of the Antarctic
continent, there are insufficient elevation databases and their quality is poor. Consequently, we need to apply methods
to combine and integrate the different DEM's which were acquired from different sources with different spatial
properties.
l2 Review of data fusion methods
Many methods have been proposed for integrating multiple data sources. For a comprehensive review on data fusion we
refer the interested reader to Abidy and Gonzales (1992). Here we mention only a few methods that are important for
understanding the procedures described in this paper. Rapp (1984) examines various techniques; that can be used to
combine satellite gravity field information with terrestrial gravimetry. He is using spherical harmonic expansions
(Fourier analysis) to interpolate the data and weighted least squares to solve the augmented observation equations and to
compute the combined interpolation function coefficients. Hahn and Samadzadegan (1999) transform the data using
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part Bl. Amsterdam 2000. 117