transformation of DTM's (see Frederiksen, 1980). 
Faintich et. al. (1982) consider DIM's convolved with a 
yariety of filters. Various aircraft simulation scenarios 
tessellate DTM's into best-fit triangular patches. (See 
Bunker, 1974.) At best, these transformations are zero 
error in nature; i.e., the original data can be recónstructed 
without any loss of information. Simpson (1979) investi- 
gated a variety of compression techniques and found that 
for large areas (one degree square cells) zero error Com- 
pression was at least 4:1 and as large as 10:1 in certain 
regions. The 4:1 zero error compression ratio seems to 
be a common finding among other investigators. Many of the 
above mentioned transformations, however, are certainly not 
zero error in nature, and the loss of information can be 
statistically described, usually by maximum and standard 
deviation or by bit significance lost. 
Cartographic Versus Photogrammetric DTM's 
A further consideration must be the source of DTM informa- 
tion and the resultant impact upon geomorphological quality 
and numerical accuracy. 
Cartographically produced DTM's all suffer from the same 
problem.  Inherent in all cartographic DTM's is the fact 
that there is not any information collected between contour 
lines except for the knowledge that the terrain does not 
vary enough to produce another contour line. If the contour 
line data is to be used for other than automated chart pur- 
poses of contour line regeneration, then additional data 
points are usually required. Independent of all horizontal 
and vertical accuracy considerations, the production of 
terrain information between contour lines requires a model 
or interpolation algorithm to be supplied to the process. 
Although several advanced techniques have been developed 
(e.g., Clarke et. al. 1982, ond Davis et. al., 1982), they 
all suffer from the same paradox: the proper model to use 
for interpolation is the same one the algorithm is trying 
to produce, i.e., the actual terrain model, and any other 
model only approximates the actual model and will produce 
either inaccuracies or anomalies, or both. This should 
not lead to the conclusion that cartographic DTM's are 
not useful. It does point out, however, that DTM data 
should be collected using parameters and representations 
that match usage requirements as closely as possible in 
order to minimize interpolation between collected data 
points. 
The advantage of photogrammetric DTM's over cartographic 
DTM's is clear. The former relies upon rigorous mathe- 
matical computation based upon the geometry of the sensor 
and the surface, whereas the latter relies upon an