255 
The following list demonstrates that there are four possible ways of cre 
ating TA and TB. They constitute different mapping strategies. 
1) TA = T2(p,r), TB = T3.6(r,g) . 
TA is a digital base map with region labeling of its pixels. TB is a 
table which assigns a graphic symbol code to each region. 
2) TA = T23(p,u), TB = T4.6(u,g) . 
TA is again a digital base map, but now with geographical unit in 
stead of map region labeling. The graphic symbol code table TB uses 
those unit numbers as keys accordingly. 
3) TA = T2.4(p,x), TB = T56(x,g) . 
TA is now a digital map in which region or unit labels have been re 
placed by attribute values, i.e., it is an attribute map. TB is a 
table which relates the graphic symbol codes to attribute values. 
4) TA = T2.5(p,c), TB = T6(c,g) . 
Here TA is a digital attribute map which carries class numbers rather 
than attribute values, and TB is organized accordingly as a class/ 
symbol code-table 
Solutions 1 and 2 are two variants involving a digital base map. We 
therefore call this approach the base map (BM) technique. Of course, 1 
and 2 are identical if T3 is an identity relation. Solutions 3 and 4 are 
two variants involving a digital attribute map, and we call this ap 
proach the attribute map (AM) technique (compare with STEINER 198lb). 
The latter has the advantage that the raster map can be treated in a way 
similar to the handling of a satellite image, for example. This means 
that many of the standard processing and display techniques common for 
remote sensing applications can be used. The major drawbacks of the AM 
technique are the large storage and computing time requirements, given 
that for each attribute a separate raster map must be produced. The BM 
technique avoids these difficulties in that only one raster map, the 
base map, is needed. Consequently, it can be implemented successfully 
even on small computers. If used in conjunction with a raster CRT dis 
play , a refresh memory and a video lookup table the BM technique opens 
up a variety of interactive possibilities (see following sections). 
Before concluding this section we may want to look at the mapping pro 
cess also in the following way. According to MORRISON (1978) we can re 
gard any mapping process as a multi-step transformation. The formulation 
of map production as in (5) suggests that, at a summary level, there are 
two essential steps to be distinguished. We may take any set of elements 
between the p's and the g's (use (l) as a reference) as a starting point. 
Then each element of this set must be associated with a number of map 
output locations m as expressed by Tl.TA = Tla(m,z) (z = r,u,x,c). This 
is a spatial allocation operation, call it A. Also, to each element of 
the same set a graphic symbol s must be assigned as expressed by TB.T7 = 
TB7(z,s) (z = r,u,x,c). This relation formulates a graphic transforma 
tion, say G. The complete mapping process can therefore be described in 
compact fashion by 
M = A . G . (6) 
From a conventional point of view one would probably always regard the 
attribute values x as the elements to start with. Since for technique 3 
above we have A = TlA(m,x) and G = TB7(x,s) we may look upon it as a 
conventional mapping solution.