The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008 
1425 
determination. The GLAS instrument was designed with three 
lasers, but only two are being used. One laser will operate at a 
time with the shot repetition rate of 40 Hz. Every shot is about 
70 m and each footprint is separated along-track by 172 m 
intervals. In the best conditions the accuracy of ICESat derived 
elevations is sub-decimetres (Fricker et al. 2005). 
Figure 2: Profiles ICESat over CDED 
The GLAS instrument, based on the principle of lidar, measures 
accurately how long it takes for photons from laser to pass 
through the atmosphere to the surface or clouds and return 
through the atmosphere. 
size. The vegetation over of the study area is dominated by 
more than 50% of coniferous trees. 
3. RESULTS 
3.1 The image’s difference and Interpretation 
Figure 3: The difference image 
This is the reference CDED level 1 minus SRTM model. The 
error (elevation difference: CDED level 1 minus SRTM model) 
per grid point is computed with the raster calculator tool of the 
ArcMap 9.2 tools bar menu. The error range varies from - 142 
to 78 m (Figure 3). The following spatial patterns are 
interpreted in the error image: 
The ICESat data (Figure 2) used in our study is extracted with 
the NSIDC GLAS Altimetry elevation extractor (NGAT) which 
is provided on the GLAS homepage http://nsidc.org/data/icesat/. 
The NGAT tool extracts elevation and geoid data from GLAS 
altimetry products. Among the 15 GLAS data product, we used 
GLAS/ICESat L2 Global Land Surface Altimetry data, 
specifically GLA14. From the same tool, we obtained outputs 
latitude, longitude, elevation and geoid in ASCII columns. The 
GLAM (version 26) is from laser 3A which provided the best 
accuracy among the measurements (personal communication 
with David Korn, GLAS team). For the fact that ICESat 
elevation data are referenced according to the 
TOPEX/Poseidon-Jason ellipsoid (Schütz et al. 2005), and for 
the purpose of comparisons with CDED level 1 and SRTM data, 
these data have to be transformed into orthometric heights 
according to Canadian Geodetic Vertical Datum of 1928 
(CGVD28) with the NAD83 UTM zone 19. The data used are 
those of the laser L3A corresponding to the period from 03 
October 2004 to 08 November 2004. We used bilinear 
interpolation method to make that each point ICESat will be 
coinciding with the corresponding CDED level 1 location. 
Because of the existence of false elevation resulting from clouds 
or valley fog, ICESat points have to be filtered. We rejected all 
value of elevation showing a difference between the 
interpolated CDED elevation and the ICESat elevation above 50 
m. 
2.5 EOSD data 
The Earth Observation for Sustainable Development (EOSD) of 
forest is a joint program between Canadian Forest Service (CFS) 
and Canadian Space Agency to develop a forest monitoring 
system for Canada. Land Cover are mapped for the forested 
area of Canada based on Landsat-7 Enhanced Thematic Mapper 
(ETM+) data acquired by the Centre for Topographic 
Information (CIT). The EOSD data was obtained from 
http://www2.saforah.org:7700/. EOSD data and products are 
freely available to the public and accessible. They are already 
referenced to the NAD83 UTM zone 19 and the resolution is 25 
m in raster format. This spatial resolution has been brought to 
the one of SRTM model through an aggregated pixel resample 
(i) From the CDED (Figure 2), high frequency errors location 
(residual anomalies) are evident on mountain features while the 
phenomenon is less evident in plane area; (ii) Mountains are 
located in the SE, S and SW and the highest error values are 
concentrated on those regions; (iii) General mean is -1.2 m and 
standard deviation is 15.6 m. (iv) The error’s histogram 
indicated that anomalies values are less. For statistical purpose, 
those values have been filtered. 
The strategy for terrain segmentation is justified by the fact that 
high frequency errors are located on the mountain feature. The 
study area is therefore divided into three slope classes: (a) Plane 
regions for slope < 5°, (b) the medium sloping regions where 
slope is > 5° and < 15° and (c) the highest sloping areas with 
slope > 15° 
3.2 Terrain segmentation 
Slope is a calculation of the maximum rate of change across the 
surface, either from cell to cell in the gridded surface like in our 
study or of a triangle in a TIN (Maune. F et al. 2001). If the 
partial derivatives of elevation (Z) along the East (x) and the 
North (y) directions are known then slope and the slope pointing 
orientation (aspect) are computed from the Eqs. (2) and (3) 
(Burrough, 1987 and Miliaresis. G et al. 2005). 
Slope = 
JlXyJI 
ay 
Aspect 
f 
arctan 
V 
\ 
J 
(2) 
(3) 
Slope is often calculated as either percent or degree of slope. In 
our study, slope was expressed in degrees while aspect was 
standardized to the eight geographical directions (N, NE, E, SE, 
SW, W, NW) defined in raster/grid representations. Aspect 
identifies the steepest downslope across a surface. Dymond et 
al. (1995) defined aspect as elementary terrain units composed 
by adjacent pixels with the same aspect pointing direction. The 
software ArcMap gives the measures clockwise in degree from