48 
Interpolator 
Inverse Distance 
Eri gin g 
Triangulation 
Grid step 
[m] 
1 
2 
1 
2 
1 
2 
Post-event 
St Dev. 
[cm] 
17.9 
31.5 
18.5 
29.6 
20.2 
32.9 
data 
Range 
[cm] 
-314/220 
-317/265 
-245/171 
-299/241 
-220/246 
-339/411 
Elab. Time 
5’ 
1’ 20” 
13’ 
3’ 
10” 
3” 
Ante-event 
data from 
CdM 
St. dev. 
[cm] 
0.2 
1.9 
5.1 
9.0 
7.9 
14.3 
Range 
[cm] 
-13/6 
-138/64 
-90/55 
-156/86 
-93/63 
-257/114 
Elab. Time 
42’ 
10’ 
2h 40’ 
23’ 
30” 
10” 
Ante-event 
data from 
flight Apr98 
St. dev. 
[cm] 
1.8 
10.0 
6.4 
12.7 
7.6 
15.3 
Range 
[cm] 
-91/82 
-294/249 
-80/85 ; 
-221/229 
-221/161 
-353/339 
Elab. Time 
2h 35’ 
39’ 
n.r. 
1 
50’ 
20” 
10” 
Table 4 - DEMs comparison 
The residual histograms are shown in figure 5; they are 
referred to the three DEMs interpolated on grid of 1 m 
step with the select algorithms. 
The intervals of the classes are of 5 cm for the histograms 
of the post event DEM, of 1 cm for the histograms of the 
ante-event ones. The narrower the base of the histogram 
is, the closer the residual is to zero. The base of the 
histogram is wider in the case of the post-event DEM. 
You can also notice like the residuals of the tests 
performed with the method Inverse Distance is very more 
thickened around the average of those relative to the 
others two methods. 
6. EVALUATION OF LANDSLIDE VOLUME 
Once the interpolation algorithms and the GRID step 
have been defined, that must be the same for all the 
DEMs, it is possible to calculate the volumes. 
As previously said, the boundary line which defines the 
landslide area that is to calculate is the same for all the 
DEMs. 
The accuracy of the values of the computed volumes 
strictly depends on the rightness of the absolute height 
adopted in drawing the contour lines in the digitized 
cartography (the interval between digitized lines is 25 m). 
Obviously, any error in the absolute height definition 
involves a systematic error in the volume evaluation. 
In the table 6 are brought the calculated volumes using 
the DEMs produced with the three algorithms of 
interpolation chosen (both on grid step of 1 and 2 m), 
using the three methods of numeric integration foreseen 
by the software: Trapezium, of Simpson and of Simpson 
3/8.” 
Values concernins to the “cut volume”, relative to the 
zones interested from removal of material, and to the “fill 
volume”, relative to the zones where the accumulation 
has been verified are also brought; such values are 
referred to the trapezoidal method. 
As you see in the table, the value of the estimated volume 
highly results from the algorithm used; but it doesn't 
results from the integration method differences (relative 
variations are in the range of 10‘ 4 ) and by the grid step 
Differences between the values relative to the three 
interpolators arrive, as far as it concerns the cut volume, 
at 30 % in the case of the CdM and at 18 % in the case of 
the data from the flight of April ’98. 
Kriking and Triangulation give results practically equals 
while the first method over-estimates the values of the 
above mentioned quantity. 
Note that the cut volume is the same either using CdM 
data or April’98 ones, while the fill volume is strongly 
different in the two cases. 
In order to have only a reference value for the volume, 
we also make a weighted average of the values obtained 
by the three interpolation methods used. The inverse of 
the variance obtained in the interpolation process was 
assigned as weight in the calculus. This last values are 
also brought in table 6. 
Such kind of weighted average produces a value 
practically coincident with that obtained with Inverse 
Distance method, which has the lowest value of standard 
deviation among all. The average between the other two 
methods seems to be more significant. 
Moreover, note as the cut volume (that especially regards 
the high side of the landslide) is not equal to the fill 
volume (the low side to the plain). Indeed, the mud which 
hit the built-up area was taken away in the period soon 
after the event. 
-2,25