2. THE 1:250,000 MAP SERIES AS A DATA SOURCE
FOR A NATIONAL TOPOGRAPHIC DATABASE
The SOI 1:250,000 maps are printed in five colours
from the folloving stable transparent film
separates:
1. black colour separate, containing point symbols
(trig. points etc.) and line symbols (power lines,
boundaries of water bodies, centrelines of small
permanent and intermittent streams, graticule,
etc. );
2. black name colour separate, containing
alphabetic and numeric character strings;
3. blue colour separate, containing point symbols
(wells etc.) and line symbols (centrelines of
canals etc.);
4. blue colour tint separate, providing the infill
for water bodies (lakes, reservoirs, major rivers,
seas, etc.);
contour
5. brown colour
lines;
separate, containing
6. green colour separate, containing point symbols
for different classes of trees;
7. green colour tint
for vegetated areas;
separate, providing infill
8. red colour separate, containing point symbols
for major buildings, temples, shrines, etc. and
line symbols for major road casements, centrelines
of minor roads and outlines of settlements, etc.;
and,
9. red colour tint separate, providing infill for
settlements.
The separates and the relevant printed maps were
considered to be the most likely sources from
which data could be captured to build a National
Topographic Database. Projection grid coordinates
were also available for the plotted trig. points,
and these have potential as control points in data
capture as an alternative to printed projection
grid intersection points.
2.1 The Quality of the SOI 1:250,000 Topographic
Series and its associated National Topographic
Database
The SPATIAL quality of real world entities which
are well defined on the printed SOI 1:250,000 map
series is represented by ‘a ‘Standard Error’
(equivalent to Root Mean Square Error or RMSE) of
0.25mm, at map scale (Goel, 1992). Although the
map printing process contributes to this RMSE and
it can be estimated that the RMSE associated with
the individual production separates is less than
0.25mm, in this investigation it was nevertheless
assumed that the RMSE of the production separates
was 0.25mm.
Tests carried out by the OEEPE (Thompson,1984)
indicate that various digitizing procedures have
Standard Deviations (SDs) of 0.08mm to 0.13mm
associated with them.
If it is assumed: 1) that RMSE provides an
acceptable estimation of the magnitude of the SD
of a procedure; and, 2) that an acceptable
estimation of the SD of an overall procedure to
198
which two separate procedures contribute is the
squareroot of the sum of the squares of the SDs of
the two contributing procedures, then, the overall
procedure (digitizing a production separate) which
combines two procedures (producing a production
separate and digitizing) will produce data having
an SD of 0.28mm, at map scale
(SORT((0.25*x2)+(0.13**2))). In the SOI an
estimate of maximum error is typically 3«SD, or in
this example 0.84mm. This value compares
unfavourably with the value of 0.50mm accepted in
certain professional environments (Drummond et
al.,1990) - but its determination through an
approximate process of 'pre-analysis' served as a
warning that the tested data capture procedures
might not produce data of adequate spatial
quality. Another aspect of quality which is
important to the general GIS user is attribute
accuracy. This was not considered in this project,
because information was not available on user
requirements regarding the nature and quality of
attributes, nor was time available to determine
these or to test even hypothetical requirements.
Logical Consistency requirements of a database
depend on the applications in which a database
will be used. At this stage the only definite
application of a National Topographic Database is
the production of (updated) 1:250,000 maps. As
automated cartography experiences have shown
‘junction cleaning’ to be a most time consuming
interactive task, it was decided that in this
investigation only the topology of' network
features would be checked in determining Logical
Consistency.
Another
database
important aspect of the quality of a
concerns its completeness. At the scale
1:250,000 many real-world features are
unrepresented. The SOI requirement on permitted
omissions at the 1:250,000 map scale is
(Goel, 1992):
"unimportant villages, hamlets, tanks and hills;
markets, dispensaries, footbridges, unimportant
fords and ferries; benchmarks except primary
protected benchmarks; relative heights;
embankments and cuttings under three meters;
shallow depressions (especially when dry); minor
streams and distribution canals; unimportant
springs and wells; unimportant temples, mosques,
pagodas etc.; unimportant footpaths and tracks,
camping grounds and deserted sites; and, reserved
forests of 10 sq.kilometeres or less in area"
As the specific definition of what — is
‘unimportant’ or ’minor’ in these instances would
have been a major additional task, in this project
we merely accepted. that every feature appearing on
the test document (with the exception of text)
must be captured for the database. This is of
course unsatisfactory for the future, when any
National Topographic Database will have to undergo
maintenance.
A final database characteristic often dealt vith
at the same time as the various aspects of quality
is ‘lineage’. At the moment a generally accepted
description of this does not appear to. exist. Many
mapping organisations (including SOI) have
produced ‘history sheets’, or similar, of their
traditional products showing how, when and by whom
a map was made. A database has the potential for
storing this ‘how, when and by whom’ information
for every attribute of its every object. It is not
known whether any existing spatial database
actually does this, or even how such information
might
consi
beyon
sheet
As
inves
lines
const
in p
the
expan
regio
3.1
I
This
avail
and
Micro
Calco
Delta
betwe
store
0.20m
Subse
using
This .
could
3.2 |
I
This
the S
Carto
SysSc
Vecto:
and
editi
softw
This
and
(prob
3.3
V
At S
requi:
cost
AGFA
Macin
hoste
perfo
addit
conve
scann
(EPSF
forma
Becau:
vecto:
this
Low-c
Ameri:
used,
produ
the
