Cost comparisons at this stage are also invalid because the degree of 
agreement between the Landsat based methods and those based on aerial 
photointerpretation, which have been accepted for some years, has not been 
evaluated. It is possible to say, however, that if current methods of 
digital analysis provide a map of acceptable standard, then the cost will 
be of the order of 25% of that involved in using photointerpretation 
systems. This estimate includes attributing the cost of the C.C.T. to the 
fire mapping in spite of the fact that the tape would likely be of value to 
other users. The aerial photographs on the other hand would normally only 
cover the fire affected area and would be of little value to other users. 
Pur ther Work Planned 
The enhanced visual imagery already available will be subjected to an 
accuracy evaluation by manually overlaying the result and the photointerpreted 
map. The COMTAL colour image display device will be used to determine the 
most appropriate density stretching function and specifications for further 
enhanced imagery will be prepared. This imagery will also be compared to 
the photointerpreted map by manual overlay. 
Working on the original Landsat data as well as the classification, 
further work will be directed to the image registration problem although 
the basic difference between picture element arrays and areas defined by 
generalised photointerpreted boundaries will remain a problem. 
Attempts will also be made to relate the signatures of the unburnt 
forest types to variations in the vegetation types as they occur on the 
ground. All work to date has concentrated on the burnt types and problems 
of image registration. 
Other forest fires were imaged by Landsat during the 1977-78 fire 
season and this opens up the possibility of examining the changes in 
reflectance characteristics before and after the fire since earlier images 
of the same areas exist. In the case of the Mt Buffalo fire, imagery 
before the fire was not available and the study has been limited to an 
examination of single date imagery. 
Conclusion 
Based on the results achieved it is too early to draw any meaningful 
conclusions. The degree of visual match obtained between the computer 
classifications and the photointerpreted map is encouraging but more 
precise accuracy evaluations are required before an operational system 
for digitally mapping forest fires from Landsat is likely to be accepted 
by users. The enhanced imagery also offers potential but it too requires 
precise evaluation and the development of practical mapping techniques and 
the problem of interpretation by skilled interpreters still exists. 
In the knowledge that Australia is to receive regular Landsat data 
after 1980 and that comprehensive systems for image enhancement and digital 
analysis will be readily available, the operational potential for the 
application of Landsat data to forest fire mapping is high.