241 
4. ANTE EVENT INPUT DATA FOR DEMS 
As said, to get the ante-event DEM we used the data 
deriving from the digitization of the contour lines of the 
cartographic elements of the CdM, at scale 1:5000. 
The numerical data produced have been furnished in 
input to the software Surfer to effect the various 
interpolations. 
It is to say that the different spatial disposition of the 
points in comparison to those surveyed for the post-event 
DEM, besides the greatest smooth of the surface, doesn't 
allow to export to this new source of the data the 
conclusions of the preceding tests. 
The co-ordinates of the vertices of the rectangular area to 
interpolate are obviously the same of those of the 
previous tests, as the step of the grid equal to 2 m. 
Tests described in the next have been made only on SA20 
landslide by using the default parameters. 
In table 6 are reported the residuals of interpolation and 
their statistic parameters. 
SA20 
ID. 
KR. 
M.C. \ 
N.N. 
! P.R. 
R.B.F. 
SH. 
TR. 
# of points 
50508 
50508 
50508 
50508 
1 50508 
50508 
50508 
50508 
# of cells 
9912 
9912 
9912 
9912 
1 9912 
9912 
9912 
9912 
Average 
[cm] 
-0.2 
0.1 
-0.1 
0 
1 -3318.4 
-0.1 
-0.3 
0.0 
St. Dev. 
[cm] 
18.3 
12.5 
21.3 
0 
7818.2 
3.7 
40.6 
13.6 
Minimum 
[cm] 
-236.4 
-170.4 
-181.2 
0 
-15336.5 
-53.4 
-386.5 
-145.3 
Maximum 
[cm] 
172.0 
178.6 
175.4 
0 
1 17200.7 
56.7 “ 
242.9 
169.9 
Table 6 - DEMs from cartographic data set. Default options. 
Observing the table 6 we note that almost all the methods 
introduce maximum values of the residuals and statistic 
parameters fairly contained with values of deviation 
standard of the order of decimetres. 
Also the least and maximum values of the residuals are 
rather small, of the order to the maximum one of a few 
meters. 
It constitutes exception the Polynomial Regression 
method that also with the new set of data introduce values 
of the residuals and statistic parameters absolutely 
unacceptable, that confirm its inadequacy to describe the 
surface, already underlined in the preceding tests. 
The Nearest Neighbor method obviously does not 
produce residuals, being already the input data on a grid, 
of multiple step of the interpolated one (10 m instead of 2 
m). 
You can besides note like the method that has furnished 
the worse results in the interpolation of the post-event 
DEMs, the Shepard's method, also in this case is the 
worse one from the statistic point of view. 
A test performed with the Inverse Distance method, 
assuming a value of the power equal to 3 rather than 2, 
have furnished the followings results, still better then the 
precedent ones: 
Average = 0 cm 
Standard Deviation = 2 cm 
Minimum residual = -27 cm 
Maximum residual = 21 cm 
The worse statistic behaviour of the interpolators on the 
post-event data can be explained partly from the greatest 
smoothing of the surfaces digitized from cartography and 
anyway, it is to consider that some out-liers can probably 
be present in the ground surveys. 
In figures 7 the contour lines maps elaborated beginning 
from the produced DEMs are represented. 
First of all, you can note at a glance as some methods are 
much more similar with this set of data rather then with 
the ones provided by means of ground survey. 
The visual analysis of the elaborate graphic leads to 
divide the methods of interpolation in the two followings 
groups: 
Acceptable: 
Inverse Distance to Power, Kriging, Minimum Curvature, 
Radial Basis function and Triangulation. 
Unacceptable: 
Nearest Neighbor, Polynomial Regression and Shepard's 
method. 
This confirms integrally almost the numerical results, 
except that for the Shepard's Method that, being the 
worse one, gave also acceptable statistic results. 
In this case we also consider opportune to use, for the 
following operations of comparison and estimation of the 
volumes, the four select methods already chosen for the 
interpolation of the post-event data: Inverse Distance to 
Power (with power equal to 3), Kriging, Minimum 
Curvature and Triangulation. 
That is because they given good results, likewise is to the 
tests on the preceding data and also because we 
considered opportune to perform the following operations 
on homogeneous DEMs for the method of interpolation 
used. 
The method Radial Basis Functions has not been instead 
taken in consideration for the respect of the volumes.