The International Archives uf the Photogrammetry. Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008 
1035 
Table 2: Comparison of Hapke parameters obtained in the inte 
grated approach with results of Jehl et al. (2006, 2008). 
Par. 
4 Parameter 
Model 
3 Parameter 
Model 
Jehl et al. 
(2006, 2008) 
w 
0,851 ±0,002 
0,869 ± 0,002 
0,78 . 
. 0,88 
b 
0,141 ±0,005 
0,294 ± 0,002 
0,20 . 
. 0,45 
c 
1,821 ±0,042 
1 
0,45 . 
. 0,65 
ё 
29,5° ± 0,4° 
32,3° ± 0,9° 
25° . 
. 30° 
Fig. 5: Phase function plot: c against b (Hapke, 1993). Typical 
values form an L-shaped area, which is shown in yellow 
with corresponding material properties. Gusev hills pa 
rameters as obtained here as well as those of Jehl et al. 
(2006, 2008) are marked in red - cp. Table 2. 
A second, three parameter model has been calculated by cons 
training c = 1, which is the theoretical maximum (pronounced 
backscattering) while in practice values c > 1 occur along with 
low b values. The resulting b-c-plot and its relation to surface 
particle properties are illustrated in Fig. 5. 
JEHL et al. (2006, 2008) have derived Hapke Model parameters 
from multiple HRSC orbits in the Gusev area. Although those 
studies did not account for the atmospheric optical depth, they 
are used for comparison - see Table 2 and Fig. 5. As a result, 
all parameters besides the already discussed c match very nice 
ly. However, standard deviations seem to optimistic. 
Radiometric analysis of the Gusev hills indicates that they are 
covered by bright particles (high albedo), which show pronoun 
ced backscattering due to a medium to high number of internal 
scatterers (cp. Fig. 5). The surface is comparatively rough. 
5.3 Small Craters 
Mars is characterized by a huge number of impact craters like 
those shown in Fig. 6 with sizes of approximately 700 m. This 
area has also been part of the HRSC DTM Test, based on the 
very same data set (orbit 894). One of the test results was that 
only 5% of craters below 1 km in diameter have been resolved, 
even in the DTM considered as “best overall result in terms of 
accuracy and fine detail” (Heipke et al., 2007). 
Based on the integration of matching and photoclinometry, a 
DTM with a post spacing of 50 m has been collected. Both cra 
ters have been successfully modeled, which is nicely illustrated 
in Fig. 6 by contour lines laid on top of the derived orthoimage 
as well as by two shaded relief views. The first one is calculated 
with the original illumination conditions; it appears almost like 
the orthoimage, because it basically “inverts” the photoclino 
metry part of the modeling algorithm. Most critical is not the 
direction of the illumination - and, depending on the reflectance 
model, viewing directions to a certain point - but the direction 
perpendicular to it: from equation (6), which is used for photo 
clinometry, follows that those facet tilts are less constrained (as 
they do not affect illumination angles). Thus, the respective 
shaded relief reveals a few local inaccuracies related to the il 
lumination direction. Nevertheless, geometric characteristics of 
this area, in particular crater shapes, are clearly recognizable. 
Absolute DTM accuracy can be checked by comparison with 
MOLA points, i.e., altimeter measurements with standard de 
viations of about 1 m in height for smooth terrain (Smith et al., 
2001). Heights at those point locations have been interpolated 
in the HRSC DTM. As a results, the mean height difference of 
the investigation area with respect to MOLA is -6.2 m. Devi 
ations in individual points are independent from terrain - see, 
e.g., the crater area in Figure 6; the maximum of -21.0 m occurs 
near the border of the investigation area, where the DTM deri 
vation is least constraint. The standard deviation of individual 
height differences is 7.1 m, which is well below the pixel size 
of 25 m for the orthoimage. (This value has been chosen in- 
between HRSC ground resolutions, cp. Table 1). 
6. DISCUSSION, CONCLUSION, AND OUTLOOK 
In conclusion, the combination of object space matching and 
photoclinometry, integrating geometric and radiometric mode 
ling of the surface along with atmospheric parameters, indicates 
Fig. 6, from left to right: Overlay of orthoimage and contour lines; shaded relief with original illumination conditions (sun vector 
marked by the red arrow); shaded relief illuminated perpendicular to the original sun vector; color-coded height differences 
between HRSC DTM, derived with the integrated approach, and MOLA height measurements at MOLA spots.