127 
<P 
X 
Hobs 
DTMJd 
Hest 
Date 
Preprocessig 
methodjd 
statusjd 
45.1235 
12.1724 
1236 
3 
80 
05-02-1996 
2 
0 
47.3456 
19.4578 
3 
5 
349 
12-02-1999 
2 
1 
Table 3 - DTM outliers relation 
In Table 3 we have that: 
9, X are the geographical coordinates of the outlier; 
Hobs is the observed height (in meters); 
DTMJd describes the DTM to which the outliers 
belongs, by means of a DTM code; 
Hest is the estimated (by means of a preprocessing 
method) height in meters; 
Date is the time reference of the preprocessing; 
Preprocessing methodjd points out the method used 
to identify and eventually remove the outlier; 
Statusjd indicates if in the DTM (given by the DTM 
code) the observed original value was substituted by 
the estimated one. 
The other tables we need to describe the height 
observations in which we have the anomalous values are 
the “DTM source code “ relation (Table 4), the “DTM 
preprocessing method” relation (Table 5) and the “Status 
for DTM outlier” relation (Table 6): 
DTM-id 
DTM source 
description 
0 
ITALIAN 
1 
ITALIAN 
(with corrections) 
2 
ITALIAN 
(lakes - surfaces) 
3 
ITALIAN 
(lakes -condensed) 
4 
BATHYMETRY 
5 
FRENCH 
6 
SWISS 
7 
AUSTRIAN 
8 
GERMAN 
9 
ETOP05U 
10 
GTOPO30 
Table 4 - DTM source code relation 
Preprocessing 
methodjd 
preprocessig description 
1 
Neighbours moving average3x3 
2 
Neighbours moving average 5x5 
Table 5 - DTM preprocessing method relation 
Statusjd 
Status description 
0 
Observed value was not substituted 
1 
Observed value was substituted 
Table 6 - Status for DTM outlier relation 
3. DATABASE STRUCTURES EXTENSIONS FOR 
SPATIAL DATA 
Up to now we have considered the possibility to link a 
usual RDBMS to our geographical information system. 
But, as we will see, this solution is not efficient when we 
have to process large amounts of spatial data, due to the 
fact that this kind of data poses distinctly problems as 
compared to traditional “business” ones. 
The main characteristics of spatial data are that: 
• the geographical natural or artificial entities require, 
for a correct and efficient description, to introduce in 
the structure new abstract data type, not previously 
considered in traditional RDBMS; 
• in most cases the database access is driven by the 
proximity relations to spatial features in the Euclidean 
space (e.g.: find all the outliers identified by the 
neighbours moving average 3x3 method in a given 
region of the space). 
As an example of the first point, let us introduce the 
information layer describing the different source data 
concurrent into a DTM. In this case we are reasoning 
having in mind the Italian DTM used for the terrain 
correction in the ITALGE095 project (Barzaghi et al, 
1996). This DTM was built by different original source 
files, as it is shown in Fig. 2. 
In the original dataset the information related to the 
source of the data was contained in a grid DTM file, which 
has two attributes for each cell: an integer value 
representing the observed height and a character value 
encoding the DTM source.