238 
ID. 
1 KR. 
M.C. 
NN. 
P.R. 
R.B.F. 
SH. 
TR. 
# of points 
918 
918 
918 
oo 
0\ 
918 
918 
918 
918 
# of cells 
915 
915 
915 
915 
914 
915 
915 
901 
o 
Average 
JsmL 
0.5 
-0.4 
-3.5 
0.2 
-9.0 
-1.2 
-6.9 
-5.5 
< 
St. Dev. 
JsmL. 
37.9 
35.4 
48.4 
21.0 
3490.6 
8.9 
54.4 
43.7 
m 
Minimum 
JsmL 
-343.2 
-226.5 
-370.1 
-345.1 
-7765.3 
-67.7 
-444.0 
-402.7 
Maximum 
JsmL 
231.8 
289.5 
263.4 
130.2 
7173.0 
74.5 
249.6 
301.0 
Elab. Time 
1’14” 
2’21” 
0’1” 
O’ll” 
0T” 
6’16” 
1’ 14” 
0T” 
# of points 
997 
! 997 
997 
997 
997 
997 
997 
997 
# of cells 
993 
993 
993 
993 
993 
993 
993 
982 
© 
Average 
.JsmL 
1.3 
-1.2 
0.2 
0.0 
-18.5 
-1.9 
-9.0 
-6.4 
< 
St. Dev. 
JsmL. 
31.5 
31.6 
36.0 
8.8 
2336.4 
8.4 
46.5 
38.3 
in 
Minimum 
_JsmL. 
-123.4 
i -172.4 
-188.1 
-106.6 
-3754.3 
-51.9 
-193.1 
-366.3 
Maximum 
JsmL 
151.7 
! 125.8 
184.7 
82.3 
17295.8 
38.7 
237.3 
180.2 
Elab. Time 
1 ’46” 
3’ 15” 
0’2” 
0’7” 
0’1” 
6’28” 
1 ’48” 
0T” 
# of points 
24869 
24869 
24869 
24869 
24869 
24869 
24869 
24869 
# of cells 
24858 
! 24858 
24858 
24858 
24858 
24858 
24858 
24792 
© 
Average 
JsmL. 
0.8 
! L8 
1,3 
1.2 
0.5 
-32.4 
0.7 
1.4 
s© 
< 
St. Dev. 
JsmL 
68.3 
i 61.7 
60.8 
65.0 
2528.0 
2270.3 
95.5 
70.3 
m 
Minimum 
[cm] 
-909.3 
! -778.6 
-978.8 
-1072.8 
-4539.3 
-98721.2 
-1242.2 
-1096.7 
Maximum 
JsmL 
1166.2 
! 1173.8 
1151.1 
1281.1 
15264.6 
143185.4 
2659.7 
1199.0 
Elab. Time 
l h 10’Q0” 
1 l h 15’00” 
0’6” 
3’47” 
0’1” 
,l. h 09’QQ” 
0’1” 
Table 1 - DEMs from survey data set. Default options. 
Observing the statistic parameters brought in the table 1 it 
is clear that the results depend partly from the data of 
input and it is therefore not possible to give an absolute 
judgement of goodness of an algorithm putting aside 
from them. Besides, for SA60 all the parameters of 
interest have higher values. 
However, the method Polynomial Regression is revealed 
to be without doubt the worse one, with unacceptable 
values of the residuals and standard deviations for all the 
three landslides. 
All the other methods give acceptable values of the 
residuals and statistic parameters; particularly: 
the methods Inverse Distance to Power, Kriging, 
Minimum Curvature and Triangulation give values of 
standard deviations nearly from 30 to 50 cm for SA20 
and SA30 and from 60 and 70 cm for SA60. The values 
of maximum and least residual are of the order of the 
unity of meters for SA20 and SA30 and of the tens of 
meters for SA60. 
the methods Nearest Neighbor and Radial Basis 
Function give better results then the precedents for SA20 
and SA30. For SA60 the N.N. method furnishes 
parameters of the same order of greatness of the group of 
algorithms described previously while the second one 
furnishes elevated values - not acceptable - of the 
residuals and of standard deviations, due perhaps to the 
presence in the data of blunders. 
the Shepard's method, finally, produces worse results 
slightly of those of the first group for all the landslides, 
also always acceptable. 
Another aspect observable in the table is that relative to 
the long times necessary to some methods to interpolated 
the data, especially if they are very numerous. Since one 
of the free options of the used software of interpolation 
refers to the choice of the amount and distribution of data 
taken into account to the goals of the calculation of the 
co-ordinate Z of the knot (all data or only those ones 
which are in a certain around, defined from the users), 
some other tests were carried out varying such parameter 
for choice a more suitable search radius than the default 
one. For SA60 a search radius of 100 m does not produce 
appreciable differences in the values of the residuals in 
comparison to that of default equal to 3130 m, while the 
times of elaboration are been reduced sensitively. 
As far as it concerns the visual aspect instead, in figure 2 
are brought the contour lines maps relative to the DEMs 
produced from the interpolations with the various 
algorithms. Only the elaboration carried out for SA20 are 
presented but the considerations on the results are also 
valid for the others two landslides. The representation 
scale is 1:20.000 although the density of the data and the 
precision of the survey are compatible with 
representations to a scale of 1:1000. With the dots is 
marked the limit of the landslide.