International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B7, 2012
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia
SOURCES OF ARTEFACTS IN
SYNTHETIC APERTURE RADAR INTERFEROMETRY
DATA SETS
K. Becek ® *, A. Borkowski"
® Universiti Brunei Darussalam, Faculty of Arts and Social Sciences, Brunei Darussalam — kazimierz.becek@ubd.edu.bn
® Wroclaw University of Environmental and Life Sciences, Poland — andrzej.borkowski@igig.up.wroc.pl
Commission VII, WG VII/2
KEY WORDS: Geodesy, Estimation, Simplification, Landscape, DEM/DTM, Modelling
ABSTRACT:
In recent years, much attention has been devoted to digital elevation models (DEMs) produced using Synthetic Aperture Radar
Interferometry (InSAR). This has been triggered by the relative novelty of the InNSAR method and its world-famous product—the
Shuttle Radar Topography Mission (SRTM) DEM. However, much less attention, if at all, has been paid to sources of artefacts in
SRTM. In this work, we focus not on the missing pixels (null pixels) due to shadows or the layover effect, but rather on outliers that
were undetected by the SRTM validation process. The aim of this study is to identify some of the causes of the elevation outliers in
SRTM. Such knowledge may be helpful to mitigate similar problems in future InSAR DEMSs, notably the ones currently being
developed from data acquired by the TanDEM-X mission. We analysed many cross-sections derived from SRTM. These cross-
sections were extracted over the elevation test areas, which are available from the Global Elevation Data Testing Facility (GEDTF)
whose database contains about 8,500 runways with known vertical profiles. Whenever a significant discrepancy between the known
runway profile and the SRTM cross-section was detected, a visual interpretation of the high-resolution satellite image was carried
out to identify the objects causing the irregularities. A distance and a bearing from the outlier to the object were recorded. Moreover,
we considered the SRTM look direction parameter. A comprehensive analysis of the acquired data allows us to establish that large
metallic structures, such as hangars or car parking lots, are causing the outliers. Water areas or plain wet terrains may also cause an
InSAR outlier. The look direction and the depression angle of the InSAR system in relation to the suspected objects influence the
magnitude of the outliers. We hope that these findings will be helpful in designing the error detection routines of future InSAR or, in
fact, any microwave aerial- or space-based survey. The presence of outliers in SRTM was first reported in Becek,
K. (2008). Investigating error structure of shuttle radar topography mission elevation data product, Geophys. Res. Lett., 35, L15403.
1. INTRODUCTION outliers in SRTM. The major aim of this work is to provide
circumstantial evidence that the metallic structures and large
A robust and fully automatic DEM extraction method that ^ and smooth surfaces are the cause of outliers in the SRTM. The
would deliver elevations contaminated with only random errors objectives leading to this aim are:
of known statistical characteristics has yet to be developed. This
is valid for both stereoscopy-based and InSAR-based methods, a) to analyse SRTM data over large anthropogenic
but less so for the LIDAR method. A possible timeframe for that structures, in this case airports, and
to happen is impossible to estimate, assuming that it will b) to link the location of the outliers to the look angle
happen at all. Meanwhile, DEMs such as ASTER GDEM and look direction of the SRTM data takes and the
(Advanced Spaceborne Thermal Emission and airports’ infrastructure.
Reflection radiometer) and SRTM (Shuttle Radar Topography
Mission) are provided with a certain number of pixels whose The Global Elevation Data Testing Facility (GEDTF) has been
elevations deviate significantly from true elevations, or void or used to identify the location of airports and the required
“no data” pixels. Void pixels may be easily isolated, and an reference elevation data. Our findings recommend that the
appropriate mitigation procedure taken. These voids are formed original interferometry data sets be audited in an attempt to
because of lack of correlation between corresponding parts of a quantify the mechanism responsible for the creation of the
stereopair, or, in the case of InSAR, shadowing effect. A far outliers and working out a new interferometry data processing
more difficult case is the detection and correction of outliers, procedure to suppress this type of error in InSAR data products.
e.g. pixels having wrong elevations. Despite deployment of
sophisticated algorithms designed to trap outliers, the above- 2. METHOD AND DATA
mentioned DEMs still contain erroneous pixels. As far as we
know, there are no published attempts addressing the issue of — 2.1 Error Structure of DEM
outliers in the automatically derived DEMs. Rather, researchers
are focused on assessments of systematic and/or random errors After Becek (2008), we adopt the following error structure of
in DEMs. In this contribution we present our findings regarding the SRTM data:
* Corresponding author.
29
