4. GROUND SHAPE BEFORE THE LANDSLIDE 
Ground surveys provide a DEM in the National 
cartographic system. 
In order to compare the ground relief before and after the 
landslides we used a map on a 1:5000 scale made twenty- 
five years previously. 
This map was commissioned by a regional Agency of the 
Cassa del Mezzogiorno and it is indicated by the 
abbreviation CdM. As said, it was not possible to have 
the photograms of that area. 
The areas affected by the landslides are at very different 
heights with different gradients. Between the niche zone 
and the foot of the landslide there are differences in 
height above sea level of 800 - 1000 meters. 
The contour lines are very close together. Manual 
digitisation is difficult and requires considerable time. 
So it's better to follow another procedure: 
The graphical CdM map was scanned and oriented 
through parametric grid marked on the map relating 
to the National system. 
A semi-automatic vectorization of the 25 m contour 
lines was carried out on the raster image by using the 
software MGE of Intergraph. Operator often must 
intervene owing to the interruption due to captions 
and symbols. 
The contour lines vectorized using raster were then 
transformed into a TIN and then onto a GRID. 
For one municipality - Bracigliano - we were able to use 
more recent data: an aerophotogrammetric survey of this 
municipality was made a few days before the catastrophe. 
Unfortunately, an accurate supporting topographic survey 
had not been requested. 
Cartographic material is not available, but .dxf files of the 
contour lines were provided from which it is possible to 
derive a DEM. Lines at a 5 meter level were used and a 
GRID was then generated. Even isolated marked height- 
points were inserted on the DEM. 
5. COMPARISON AMONG INTERPOLATORS 
Once the DEM of the ground relief before and after the 
landslides had been obtained, it was possible to compare 
and estimate the volumes. 
Unfortunately, the maps do not seem to be very reliable 
from the altimetric point of view, also because of the 
absence in the area of a good national leveling network. 
The lines of geometrical leveling of the main network 
pass at a distance of 40 to 50 km from this area. 
Problems regarding volume evaluation are the next: 
the possible lack of precision of the measurements 
made. Sometimes the post-event relief was not accurate 
enough for the use of low precision instruments, as in 
case of using MDL where the instrumental error increases 
with distance and the result is given from the ground 
level at a wide solid angle; or in case of motorised GPS 
where there is a large margin of error when the antenna is 
placed at ground level; 
the references to the cartographic heights and to the 
new surveys probably are different; 
the transformation and interpolation methods of the 
DEM could affect the comparison. 
As far as the effect of the interpolation method, some 
algorithms provided in a well known software (SURFER) 
were used on the Bracigliano landslide, for which the 
three different data input for DEMs production were 
available. 
In a evaluation of the best method, particular emphasis 
was placed on the following: 
the residuals of the interpolation; for some 
algorithms the standard deviation ranges from 32 to 43 
cm, others produce values that are too high. 
the comparison between the original contour lines 
and those reconstructed using the DEMs. 
Three algorithms were at last considered: Kriging, 
Inverse Distance to a power, Triangulation, in order to 
elaborate the DEMs on a grid step of both 1 and 2 m. 
In table 4 are reported the residuals of the DEMs 
interpolations of data set provided by the surveys made 
on the ground, both by the digitisation of CdM 
cartography and by one of Bracigliano municipality one. 
In the upper side of the table (input-data set consists in 
sparse points surveyed on the ground) you can note that: 
the more thick grid step causes lower values of 
standard deviation for all the methods and for the 
Triangulation the range of the residuals also results 
sensitively lower; naturally the time of elaboration is 
quadrupled; 
the times of elaboration are contained, and a great 
deal low for the Triangulation: 10” for the grid to 1 m and 
3” to 2 m. 
As far as it concerns the data from CdM (input-data set 
consists in TIN drawn by the vectorization of contour 
lines taken by cartography in raster form) you see from 
the table like the values results very lower than in the 
post-event case; those of the standard deviations minus 
then 14 cm. 
The maximum values of the residuals are smaller then 2 
m, sort exception for the method Triangulation, with grid 
step of 2 m, that also in this case it is revealed the worse 
one, to front of very short times of elaboration (10”). 
In the case finally of the DEMs interpolated by the data 
of the contour lines plotted from the flight of April ’98, 
the values of the standard deviation are noticed to be 
higher then in the preceding case, above all for the first 
method, and above all a notable growth of the ranges of 
the residuals; such values are however more contained in 
comparison to the ante-event data. 
To the usual one, the method that has furnished the worst 
residuals is the Triangulation. 
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. 
We should bear in mind the large number of points, tens 
of thousands, surveyed on the ground and the connection 
between partial surveys made with different techniques.