positional quality; 
attribute quality; 
lineage; 
completeness; and, 
logical consistency. 
NS CN rm 
€ 9 We ats 
The two first refer to data describing INDIVIDUAL 
real world entities, with 4) and 5) referring to 
SETS of real world entities. The concept of 
"lineage" includes but also goes beyond "temporal 
data" (or "when data") to include information on 
how data were generated and what processing they 
As already mentioned, an entity’s attributes may 
be recorded by continuous variables or by 
discontinuous variables. A quality parameter 
associated with a continuous variable is, of 
course, Standard Deviation, while quality 
parameters associated with discontinuous variables 
are certainty statistics such as Probability or 
Certainty Factor. Influenced by ideas presented in 
1989 [GUPTILL, 1989] we propose that all 
attributes should have a quality parameter stored 
with them, as shown in TABLE 1. Such an approach 
is extremely flexible, as such database tables are 
have undergone (or "how data"), and also "by whom 
data". A "lineage" could apply to an individual 
real world entity as represented in the database, 
but more usefully a set of such entities. 
Concerning "completeness", if a GIS purports, 
e.g., to record all storm-drain inspection covers 
already accessed in the information generation 
procedures used in many GISs. 
Continuous variables record position and the 
quality parameter for such variables is Standard 
Deviation. Although each position can have x and y 
  
  
  
  
  
  
igation (and z) standard deviations it is likely that 
Jverlay 
system 
7. The 
apie POLYNR PASD PNSD VALUEl1 DIS1 QUAL1 UNIT1 VALUE2 DIS2 QUAL2 UNIT2 DESCRIPT 
w Land 
ty 12547 25 16 Has T 0.76 1.55 Up 0.10 4 MEMO 
1255 25 . 16 Hn33 T 0.65 1.90 F 0.10 M memo 
lality, 1256 25 16 —Hn31 "T 0.82 1.70 F 0.10 M memo 
1257 25-16 . E235 T 0.87 2.00 F 0:10 M memo 
1258 25 16 7835 T 0.76 1.390 — T 0:10 M memo 
1259 25 16 Hn35 T 0.58 1.55 EF 0.10 M memo 
1260: 25 16 "2335 T 0.76 1.60 ^F 0.10 # memo 
es to The ’MEMO’ (or Data Dictionary) associated with the SOILPOLS relation is: 
uality : 
POLYNR unique identifier for the soil polygon 
PASD standard deviation of arc coordinates, x and y 
patial PNSD standard deviation of node coordinates, x and y 
dn a VALUE1 Dutch Soil Classification System soil class 
DIS1 T to indicate a discontinuous variable (discontinuous is true) 
tities QUAL1 probability associated with the soil class being correct 
UNIT1 no units for soil class 
VALUE2 soil rooting depth 
data; DIS2 F to indicate a continuous variable (discontinuous is false) 
QUAL2 Standard Deviation of soil rooting depth measurement 
UNIT2 units for soil rooting depth (meters) 
eters, 
(or, 
Table 1 - The relation (or database table) soilpols 
a real 
ed. In 
E. by in a city - completeness indicates the percentage certain types of points (e.g. node points for 
ed by of these covers actually recorded. Concerning roads) within a large data set could have common x 
lso be "logical consistency", specific data processing and y standard deviations associated with them, 
or tasks (e.g. route selection, or a land-parcel and others (e.g. arc points for rivers) different 
mporal ownership query) assume certain characteristics of x and y standard deviations. This proposal is 
3 were the data (e.g. that all connected road segments easily implemented in an ‘off-the-shelf’ GIS, as 
meet at common nodes, that all parcels have a such „Standard deviations can also be stored in a 
: unique identifier), and it should be known to the relation such as TABLE 1 above. There is no need 
ytical GIS user to which extent these characteristics are to ‘break into’ the sometimes proprietary 
GIS to met: Structure of coordinate files - at least in the 
to be preliminary stages of developing a quality 
gical 2.1 Storage of position and attribute quality subsystem. Temporal data could also be stored in 
pulate | parameters TABLE 1 - although we have not done so, believing 
(e.g. that this can be handled from the Lineage Report 
[nous We have proposed methods for storing positional [RAMLAL, 1991]. 
itude) and attribute quality by data item (see 
ta and [RAMLAL,1991]  [DRUMMOND,1991]) and Lineage, 
Completeness, and Logical Consistency Reports by 
: data set. We have also proposed a Processing Model 
It is Quality Report [RAMLAL,1991], but as will be 
AGHAN, indicated in section 2.2 this may not always be 
ts of useful. 
357