Class | Limits Percentage
1 0m x Ahl « 10m 39.1296
2 10 m € IAhl « 20 m 27.11%
3 20 m € IAhl « 50 m 28.10%
4 50 m € IAhl « 100m 5.1196
5 100 m € IAhl « 150m 0.48%
6 IAhl 2 150 m 0.08%
Table 4. Classes of elevation differences between Switzerland
and Piedmont and their percentages
=
e
c
=
150m
£ = =
> © ©
u» C =
Distribution of the elevation differences (absolute values)
CN
4661
465r
464r
463r
461r
e LD wo
<r <r <r
[IY
Figure 3. Spatial distribution of the elevation differences
between Switzerland and Lombardy DTMs.
3.2 Cross-validation of high and low resolution DTMs in
Lombardy
High and low resolution DTMS have been compared by
implementing new commands in GRASS GIS 7.0. The
procedures that have been applied for the cross-validation in the
Lombardy region are here shortly described.
The LR DTM has been transformed to the ETRF89 reference
frame. The result is a list of 3D points, that clearly are almost
regularly spaced but no more on a regular grid. For each point, a
square window of 20 x 20 m? has been considered. All the HR
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B4, 2012
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia
DTM nodes contained in the window are averaged and their
mean is compared with the LR height.
Note that the HR DTM covers only the hydrographic main
basins of the region, for a total coverage of about 3810 Km?
(which corresponds to the 16% of the total Lombardy region).
The following global statistics have been obtained on the
common area:
e number of points: 4048660
e mean, u(Ah) =0.5m
e standard deviation, o(Ah) = 6.6m
e maximum, max(Ah) = 204 m
e minimum, min(Ah) =-138m
Distribution of the elevation differences (absolute values)
|
8
[.]Y
Figure 4. Spatial distribution of the elevation differences
between Switzerland and Piedmont DTMs.
465
6 4
463r
46.1
To verify the existence of global biases, the differences have
been clustered in eight classes: the results are reported in Table
5. Figure 5 depicts the spatial distribution of the differences.
Moreover we have started to investigate a phenomenon that, at
first sight, seems quite strange. As example, Figure 6 and Table
6 report the differences relevant to the Valley of S.Giacomo that
covers all the B1 Lombardy grid.
The mean difference is 0.5 meters, which is not significant with
respect to the LR DTM accuracy. However, in the West slope
(the lower part of the figure) a negative bias between HR and
LR DTMs appears. In the other slope an opposite bias is
present. Such a behavior could be justified by a translation of
