it is not possible to arrive at more detailed conclusions 
regarding change in the Delta. 
IMPLICATIONS FOR THE STUDY AREA 
It is not possible to distinguish between all of the groups 
delineated at a satisfactory level. Confusion between 
classes may have occurred because of: 
o misclassification of pixels 
o large variation within some groups - This may be 
caused by inclusion of two sub-groups within one 
group (for example, grouping wet mudflats with 
dry mudflats, or clear water with muddy water). 
o too many classes were delineated - When the three 
vegetated groups were put together, classification 
accuracy improved significantly. 
o complexity of the study was too great - This 
method may be more suited to areas with fewer 
classes to be distinguished, and with larger, 
more homogeneous units. 
o geomorphological analysis is, in many ways, a 
discerning of pattern - To this extent, digital 
analysis would be less productive than visual 
analysis. Many of the important clues to unit 
classification are in the relationship between 
pixels . 
Therefore, use of this method to monitor change in the 
Athabasca Delta is not recommended unless major improvements 
can be made to overcome some of the these problems. 
IMPLICATIONS FOR DIGITAL ANALYSIS 
The lack of success in classifying the study area does not 
imply that the method could not be useful in other areas. 
Boyd (1981) has successfully applied this method in other 
parts of Alberta, particularly the parklands. Recommenda 
tions for future projects using this method are: 
o projects selected should ideally be concerned with 
units which are large and homogeneous 
o the units chosen, whether biophysical or geomorpho 
logical should be related to one cover class; if 
this is not possible, more than one subunit should 
be delineated and these subunits amalgamated in the 
production of maps - For example, clear water and 
turbid water may have to be classified separately, 
even if they are eventually to have the same classi- 
fication. 
o the number of classes should be as few as possible, 
notwithstanding the above point 
324