cost 24 man-days or 1 man-month. This makes updating of the existing
maps with oblique airphotos quite feasible, even in big cities).
4. The research questions
The interpretations can be analysed from three slightly different
angles. First, the thematic maps produced can be compared with each
other. This leads to the question:
— if the area is analysed by several interpreters, do they
indicate as newly built-up the same areas on the map?
Secondly, the maps can be compared with a standard interpretation, and
the following questions can be asked:
— how correct are the individual interpretations?
— how accurate is the information the interpreier provides to
the city administration?
These last two questions are not equal. The first is asking for all
errors that are made; the second is asking how many of these errors
cancel out each other.
5. Do the interpreters indicate the same areas as newly built-up?
Visual comparison of the maps (see fig. 1 no. 1-11) of the eleven
interpreters reveals a great similarity in the sizes and the spatial
distribution of the newly built-up areas. A more detailed search,
however, discloses a great number of minor differences. (Average 20
per interpreter). If all maps are placed on top of each other (sieve
mapping), one is able to find:
— all areas that are indicated by one or more interpreters as
newly built-up areas. This was 58 ha, 21% more than the real
48 ha. (See table 1, total area 57.8 ha - 100%);
— for each area, the number of interpreters that have indicated
an area as newly built-up. (See fig. 2, the sieve map);
— he total area identified by 11, 10, 9, ..., 1 interpreters.
(See table 1 and figures 3A and 3B).
In figure 2, the darkest screen indicates areas interpreted by all
interpreters as newly built-up. This 'hard core' covers 34 ha or
59% of all areas indicated. (Table 1, line 1).
The second darkest screen in figure 2 is used for areas indicated as
newly built-up by at least six interpreters. That means that one to
five interpreters have not been mapping these areas. This 'overlooking'
can be explained by three types of errors:
b