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 
(1) 
2 2 2 2 
OsRTM "OT * OT t Og 
where the first two variations, o2 and oi, represent 
. ; 2 
the target-induced and the instrumental component of C'srTMm - 
the SRTM (elevation) error variation. The target-induced error 
variation - c2 may be estimated using the following formula 
(ibid): 
1 
o2 = 
(d? tan? (s) Q) 
where d is the pixel size and s is the slope. 
The instrumental error variation - ef was estimated at 2.4 m, 
which is equivalent to its root square, 0, = +1.55m (ibid). 
The third component, c2 - other errors variance may be caused 
by factors including type of surface, look angle and look 
direction (ibid). In some cases, these factors may cause a 
significant variation in radar brightness (Rodriguez, et al, 
2005), which in turn leads to an erroneous elevation and even 
data voids. It the following we focus our attention on this 
component of the total SRTM error. 
2.2 Method 
A search for suitable objects to investigate the major aim of this 
research has resulted in the selection of large anthropogenic 
structures, which typically are big airports. Another requirement 
for the test sites was that they were topographically indifferent. 
Again, the airport sites are, in the majority of cases, relatively 
flat, at least in a certain radius from a centroid of the site. In 
order to maintain the assumed flatness of the site, the diameter 
of the buffer should be smaller than the length of the runway(s), 
which are in the range of 3,000 m at big airports. The following 
steps have been performed to achieve the aim and objectives of 
the investigations: 
1. A set of large airports was selected (further referred to as 
AOI [airports of interest]); the word ‘international’ in the 
airport's name was used as a qualitative indicator of the 
size of airport. 
2. Relevant data on AOIs have been downloaded from the 
aeronautical Web site. 
3. For each AOI, a corresponding SRTM tile has been 
downloaded. 
4. A circular buffer centred on the airport’s reference point 
was constructed. 
5. Further analysis was focused on the SRTM pixels found 
within the buffers. 
6. Pixels have been classified into three groups: voids, pixels 
having values within a range allowed by statistical 
considerations and other pixels considered as outliers. 
7. For each set of pixels, a semivariogram on the disparities 
in elevation between the reference elevation and the pixel 
value has been calculated. 
The semivariograms were related to the look angles and look 
directions (see the following section for definitions of these 
variables) of data takes of the SRTM mission. 
30 
2.3 Data 
In this study we have used the following data sources: 
1. The Global Elevation Data Testing Facility (GEDTF, 
2011); 
2. The Aeronautical Information Package (API, 2011); 
3. The SRTM Coverage Plotting Tool (JPL, 2008); and 
4. The SRTM downloading facility (NASA, 2001). 
Source 1) was used to extract location data and the physical 
characteristics of the runways among the airports of interest. 
Source 2) was used to extract the reference data on each 
airport’s infrastructure, including the coordinates of runways 
and the airport’s reference elevation. Source 3) was used to 
extract geometric variables of data takes over AOIs. These 
variables are the look direction (LD) and the look angle (LA). 
LD is the azimuth of the radar beam during the acquisition of 
data take. LA is the angle between the vertical at the SRTM 
instrument and the radar beam during the acquisition of data 
take. 
The SRTM tiles were downloaded from NASA’s data 
distribution centre (NASA, 2001). 
2.4 The SRTM dataset 
The Shuttle Radar Topography Mission (SRTM) took place in 
February 2000. The mission acquired the Synthetic Aperture 
Radar (SAR) data suitable for the interferometric processing 
and development of the digital elevation model of the Earth's 
surface. This mission was a collective effort of NASA and 
German/Italian space agencies. The acquisition of data had been 
carried out using the C-band (5.6 cm) and X-band (3.1 cm) 
microwaves. Hence, two distinct DEM models have been 
produced. In this project, we use the C-band SRTM DEM, 
which has the pixel size of 3 arc-seconds. This resolution was 
achieved by resampling of the original 1-arc-second DEM. Both 
DEMs over the United States are available for download free of 
charge (NASA, 2001). 
3. RESULTS AND DISCUSSION 
For the purpose of the study, a set of 64 major U.S. airports has 
been selected (API, 2011). The available data on airports in this 
source (ibid) include coordinates and the elevation of the 
airport's reference point, which constitutes an approximate 
geometric centre of the airport's area. A georeferenced chart of 
each airport is a part of the source (ibid). The coordinates are 
provided in the WGS84 datum. The elevations are provided in 
feet above mean sea level. A circular reference buffer (RB) with 
a radius of 1.5 km centred on the airport's reference point has 
been constructed for all chosen airports of interest (AOI). The 
RB covers mostly the elements of the airport's infrastructure 
including runway(s), terminal, hangars, taxiways, tarmac, and 
aprons. It is important to note that the topography of the area 
where an airport is constructed must be flat. Obviously, this is 
the case for the terrain located within the RB. Using RB, the 
three-arc-second SRTM pixels were extracted from the SRTM 
(NASA, 2001). There were approximately 824 SRTM pixels in 
each buffer. The total number of pixels extracted for all RBs 
was 52,753. The initial assessment has identified a number of 
void pixels (no data). A summary of the void pixels follows: 
1. No of void pixels: 970 (~1.8% of all pixels); 
2. No of airports with void pixels: 22 (~35% of all airports);