The differences noted between the two regions have resulted in changes to 
  
  
rainins the methodology in terms of organization; location of segments on the satellite 
writing data; selection of training data; classification and  post-classification 
or was manipulation; and handling of problems of missing or questionable ground data, 
cloud cover, haze, etc. 
Eighteen system corrected Landsat picture centers are required to cover 
the whole Peace River region — only part of which is agricultural. The 
| other | agricultural region is covered by 29 geometrically corrected Digital Image 
1982: Correction System (DICS) scenes (Guertin et al., 1979). The area covered by 
closely DICS data analyzed in the Peace River District is roughly equivalent in size to 
method the State of Iowa, a major agricultural state studied in the United States with 
similar methods (Hanuschak et al., 1979). 
ression 
Leviqus Since DICS data take over two weeks to produce, they are not as efficient 
in generating timely crop area estimates as system corrected data (typically 
of the processed in 48 hours) would be, provided that similar results are available 
ensures from each. Since the constraints which resulted in the use of DICS data in New 
: Brunswick -- small fields and many complex segments — do not exist in the 
> needs areas considered for Canola/rapeseed, it was decided to assess system corrected 
Landsat digital imagery as an alternate data source. 
au cet Where system corrected data were used an additional step was required 
IAS is compared to the DICS data analysis. Existing images may be consulted to assess 
is Very the likely coverage of each new image. When an area is covered by two or more 
splayed frames from two or more orbits one must ask which part of the area should be 
fectin studied from which frame. Together the maps and previous data can be used 
8 before analysis to select a likely boundary between the two frames. In this 
and of study it was a river and a road which were visible on 1:1 000 000 B&W paper 
tor the prints. 
{iit ion Segment Location for Training and Analysis: In the potato project, ground 
segments were located before the summer analysis season began to reduce the 
visual time required at the peak season. The segment location was time consuming in 
tion of New Brunswick as a result of the confusing survey pattern, small field sizes 
of the and irregular shapes of the areas to be bounded. These same factors also made 
| short DICS data attractive, since one could locate boundaries on "old" data and 
simply (but accurately) overlay them on the data from the current crop year -- 
regardless of complexity. In the Peace River District the segments were 
regularly shaped and easily located on the DICS data amid the more regular and 
hence clearer survey pattern during the crop area estimation. In addition,. 
most segments in New Brunswick contained at least some potatoes and thus had to 
ita for have areas of potatoes measured for them. In the Peace River District, 
Feon et however, many segments did not contain rapeseed, had nothing in them classified 
f SI as rapeseed (see below), and thus could be dealt with easily without any 
yor measurement or precise boundary location. 
dt^ are Taking into consideration the differences between New Brunswick and the 
sulting Peace River District, it was preferable to locate the segments in real time 
1 during the analysis of the Peace River DICS data. This did not add appreciably 
amis ya to the time required during the analysis phase and greatly reduced the total 
Source time required. It also will reduce the future cost of data preparation to 
Statistics Canada since extra copies of segment photos and location maps need 
dS d not be prepared, nor must historical data be purchased if new areas are to be 
one for investigated in western Canada. 
re are 
se The general location of segments was established through reference to 
rare 9 easily identifiable lakes or other geographic features. Where difficulties 
: ap occurred, it was also possible to measure the distance from the edge of a map 
sheet and relate this to pixel counts from the edge of the DICS image and its 
389 
ey S p 4 
- _—_ RB. T E : ie ‘ 
ve nat i m e. ERR a ee re t RR RE An 
——— MÀ ——— DLL SR