105 
INVESTIGATION OF ELEVATION BIAS OF THE SRTM C- AND X-BAND 
DIGITAL ELEVATION MODELS 
K. Becek 
UBD, Geography Department, Jalan Tungku Link, Gadong, BE1410, Brunei Darussalam - kbecek@fass.ubd.edu.bn 
KEY WORDS: Remote Sensing, Vegetation, Retrieval, Multifrequency, DEM/DTM, SAR, Land Cover 
ABSTRACT: 
The paper presents results of a comparative study of the vegetation-caused elevation bias of the space shuttle topographic mission 
data product, both C- and X-band (SRTM.C/X). The SRTM.C/X bands data were compared against a high-resolution digital terrain 
model. Pixel-based differences in SRTM.X minus SRTM.C were correlated with land cover (‘agriculture’, ‘house’, ‘tree’, ‘water’). 
Findings of the investigations include that the SRTM.X does not represent a canopy top of vegetation and that the X-band penetrates 
deeper vegetation cover than the C-band. As a test site, an area of about 159 km 2 on the Gold Coast, Queensland, Australia, was 
selected. The area of interest is about 57% covered by vegetation varying from grassland and shrubs to forest. The study method 
allowed the development of a statistical model relating the elevation bias to the percentage of the vegetation cover of a given land 
parcel. This model, once verified on varieties of vegetation types, could be utilised to estimate and eliminate the elevation bias from 
the InSAR elevation model. This model could also be utilised for estimating biomass quantities and their variations. It is hoped that 
the results will also stimulate investigations towards developing a multi-frequency InSAR system for collecting both terrain 
elevation data and attributes of biomass. 
1. INTRODUCTION 
The shuttle radar topography mission elevation data product 
(SRTM) is one of the most valuable global resources of 
topographic data to date (Rabus, et al., 2003). It was 
developed for about 80% of the global landmasses using C- 
band (X = 5.3 cm or f = 5.7 GHz) InSAR (interferometry 
synthetic aperture radar) technology. The vertical (absolute) 
accuracy is quoted at ±3.3 m - 7.3 m (Rodriguez, et. al., 2005). 
Pixel size for the U.S. territories is one arc-second (~30 m), and 
for the remainder of the globe three arc-second (~90 m). The 
second instrument flown during the same Endeavour mission 
(February, 2000) used the X-band electromagnetic spectrum (X 
= 3.1 cm or f = 9.7 GHz), but without the so-called scan mode 
which provided the data for selected areas of the globe only. 
These data are available at one arc-second pixel size (~30 m) at, 
a cost of €400 for 15’ by 15’ tile. Both elevation data products 
(note neither DTM nor DEM) exhibit elevation bias due to the 
partial penetration of vegetation by the electromagnetic waves 
at those bands (C and X). The effect will be referred to as 
impenetrability of vegetation cover. The level of 
impenetrability is the layer of vegetation above the ground 
which is not penetrable by the C- or X-band. The magnitude of 
the impenetrability depends on many factors related to complex 
interactions between electromagnetic waves and the 3D 
object—the vegetation cover. The current scientific position is 
that the C-band penetrates the vegetation (forest) to about 50% 
of its thickness, whereas the X-band is reflected from the top of 
the tree canopy (Carabajal et al., 2006; Balzter et al., 2007). 
This position seems to be contradicted by at least a few authors 
who demonstrated that C/X-band impenetrability may be very 
similar (Werner et al., 2005; Simard et al., 2006). Moreover, it 
will be shown that the X-band impenetrability can be smaller 
than the C-band. These inconsistencies are most likely caused 
by the size, shape, and orientation of scatters within the tree 
canopy, and can theoretically be used for quantitative 
assessment of vegetation or biomass. Quantitative assessments 
of vegetation structure using SRTM data were studied by 
(Kellndorfer et. al., 2003), who considered determination of 
vegetation height, using for reference the ‘bald-Earth’ elevation 
and SRTM error mitigation procedure. Walker et al., 2007, 
considered the quality of the SRTM C- and X-band elevation 
data and their suitability for retrieval of the vegetation canopy 
height. Balzter et al., 2007, experimented with an airborne L- 
band and derived and smoothed the ‘bald-Earth’ elevation 
model and X-band canopy height. The reported accuracy of that 
experiment is remarkably high probably because of the 
relatively flat test site. 
In this paper, the impenetrability of vegetation cover is 
investigated. The test site, located in Australia, consists of 
various types of land cover, varying between grassland, urban, 
and tree cover with various densities. The main goal of the 
investigation is to document the existence of situations where 
X-band impenetrability is less than that of C-band, or in other 
words, when X-band microwaves penetrate the vegetation cover 
deeper than C-band waves. An attempt is also made to relate the 
difference in impenetrabilities to the height, density, and type of 
vegetation cover. Answers to those questions can lead to the 
development of a multifrequency (C/X/L-band) microwave 
system for mapping the vertical structure of vegetation cover. 
2. MATERIAL AND METHODS 
2.1 Test site 
The test site comprises the northern part of Gold Coast City, 
Queensland, Australia (top left: 153.25 E, 23.75 S, and bottom 
right: 153.5 E, 28 S). All water bodies were masked out from 
further consideration. The area of the considered terrain was 
about 153.9 km 2 . A sun-shadowed view of the test site is shown