Although accuracy figures were based on improved cropland in 1986, half of which 
the numbers of polygons detected and was created after 1961 and mapped from 
classified, experience gained with the Thematic Mapper imagery. The comparisons 
interpretation and accuracy assessment show that land of poorer capability for 
suggests that higher error rates are agriculture is increasingly being used for 
concentrated among the smaller polygons of cropland. Approximately 40% of the land 
change. Although not measured for this with a high capability for wildlife use 
project, it is believed that accuracy has been developed for agriculture and 
levels based on the area of change would another 25% is threatened with 
be significantly higher than the accuracy development. Such analyses are only 
figures guoted. Therefore, the accuracy possible when recent detailed land use 
figures quoted here are likely a low information such as that obtained from 
estimate of what would be expected over a Thematic Mapper imagery is combined with 
reasonably large area. the analytical capabilities of GIS. 
SPATIAL ANALYSIS 
Input of Data 
The data input operation involved the use 
of both mainframe and micro computer GIS 
systems. Land use change overlays were 
prepared from the interpreted land use 
change manuscripts. Coding sheets were 
prepared linking the pre and post change 
class combination with the land use change 
polygons. Unclassified polygons and 
topological inconsistencies were detected 
and corrected during this preparation 
stage. The edges of adjoining map sheets 
were checked for omissions, mismatched 
polygon boundaries and classification 
inconsistencies. 
The change maps were converted into 
digital form using a drum scanner. The 
resulting raster files were then edited to 
remove any remaining topological 
inconsistencies. The Canada Geographic 
Information System (CGIS), a mainframe, 
was used to store the map files and 
assemble them into files for the three 
geographic zones of the project area. Map 
files containing all land use change 
polygons larger than a few hectares were 
downloaded to a PC-based GIS (SPANS by 
TYDAC Technologies Inc.) workstation from 
which Quadtree map files of changes were 
created for each zone. The coding sheets 
were compiled separately on the mainframe 
into land use change attribute files for 
the polygons in each zone. The attributes 
were transferred to the microcomputer 
workstation by converting them to dBase 
III Plus files. The polygons and their 
attributes were combined in SPANS for 
analysis. 
Data Analysis 
Digital Canada Land Inventory (CLI) map 
data on agriculture, forestry and wildlife 
land capability as well as the CLI 
"Present Land Use" files were downloaded 
from CGIS to SPANS. These data sets were 
analyzed along with the land use change 
information in order to determine the 
degree to which the land use in 1986 
compared with the capability of the land 
to support such uses. 
The methods used in this project for 
deriving current land use information from 
satellite imagery have proved to be an 
efficient means for objectively assessing 
the spatial dimensions of progress to 
sustainability for various resource 
options. For example, in the Peace River 
zone, there were almost 20 000 km2 of 
DISCUSSION 
Imagery 
The use of the boreal enhancement 
developed by CCRS (Ahern and Sirois, 1988) 
was a significant element contributing to 
the success of this project. The project 
benefitted the most from this enhancement 
in the Peace River and Central zones which 
are substantially forested. For scenes 
that contained a large amount of 
agricultural or grazing land, boreal 
enhancements tended to appear "washed out" 
with less spectral contrast between 
various cover types. However, because 
numerous images were interpreted for the 
same area, any problems encountered with 
one particular image were alleviated. 
Regarding seasonal aspects of the imagery, 
it was observed that images from August 
produced the best overall results, in 
particular because of the wide range of 
distinctive colours corresponding to the 
various maturing crops distinguishing them 
from other cover types. Images from May 
were useful for revealing pasture and 
grazing areas. The bare soil associated 
with improved cropland contrasted the high 
infrared reflectance associated with 
pasture and grazing lands. Winter images 
proved not only valuable for 
discriminating between forested and non 
forested field but also for providing 
within-field textural information that in 
some cases was associated with unimproved 
agricultural land use (K). 
Method 
A visual interpretation technique 
incorporating images and maps presented 
several advantages and also some 
limitations. One advantage of using 
existing land use maps was that it 
provided immediate calibration for the 
land use classes since a large proportion 
of each map area had not changed land use 
over the 25 year period. Once calibrated, 
the focus of the interpretation shifted to 
the portion of the Landsat scene which 
exhibited spectral and textural 
characteristics different from the 
unchanged portion. Additionally, the 
revised maps provided a up to date 
manuscript of all land use as of 1986. 
Probably the most severe limitation of the 
method was the amount of time and effort 
required to create a digital data base 
from the manuscript maps. A digital data 
base was very important in order that the 
362