large area application and how these were overcome in the analysis. Training tüo ne 
and technology transfer were an integral part of the project and major data: 
contributors to the project's success. Results available at time of writing d 
; : manipul 
are given in the fifth section. Information on how technology transfer was cloud « 
incorporated into the project is discussed in the sixth section. 
THE GENERAL APPROACH tho ni 
|  agricul 
The general approach to the use of Landsat for agricultural and other | Correci 
studies in Canada has been described in detail previously (Brown et al., 1982; DICS ds 
Ryerson et  al., 1982b). The procedure consists of several closely the Ste 
interrelated stages - user contact, background spectroscopic research, method Similal 
development, and method application in demonstration projects. The progression 
to a next stage always depends on successful results obtained in the previous c 
stage. in gen 
proces: 
The approach involves the user or resource manager in all stages of the from a 
project from initial planning to method application. This involvement ensures Brunsw: 
that the methods are understandable to the user and that they meet his needs areast 
with respect to timeliness, cost and ease of verification. Landsal 
Methods based on software available on the CCRS Image Analysis System i 
(CIAS) (Goodenough, 1979) meet these general criteria, as well as those of compar: 
accuracy and reliability. The parallelepiped classification on the CIAS is the 1L 
useful since the user can see what is happening. The classification is very frames 
fast and conceptually easy to understand, while the results are displayed Studie: 
pictorally. There is no fear of some intermediate "black box" affecting before 
results, Since the system's design allows viewing of the video data and of study 
tentative classifications, the user is in a position to continually monitor the prints, 
results before they are accepted. 
Because of the user's continuous involvement from problem definition jme 
through ground data collection, image analysis and display (including visual Seem T 
interpretation and classification acceptance), to final use and publication of New Br 
results, the technology transfer and training take place over the course of the and ir 
project. Learning by doing has proven to be much more effective than short DICS € 
intensive workshops alone would be. simply 
LANDSAT DATA FOR CANOLA/RAPESEED CUM 
AREA ESTIMATION hence 
most sS 
Introduction to Large Area Methods: The procedure for using Landsat data for have s 
Canola/rapeseed estimates evolved from that developed for potatoes (Ryerson et howeve: 
al, 1981). It was recognized that modifications would be required as a result as raj 
of differences in the nature and scope of the problem in the Peace River méssur. 
District in northern Alberta and British Columbia (Table 1). 
In the Peace River District, agricultural fields are larger, but are Peace 
often less internally homogeneous than those in New Bruswick. The resulting durin 
rapeseed signatures are expected to be highly variable. In 1981 highly unusual IE 
weather conditions at the time of planting and germination added another source time 1 
of variability. In addition, there are fewer ground segments than in New Statis 
Brunswick. In 1981, there were five ground segments for training and not be 
verification for each New Brunswick 512 x 512 pixel subscene and only one for invest 
three subscenes in the Peace River District (Table 1). Although there are | 
fewer ground segments in the Peace, they conform to the generally easily ) 
recognizable, regular north-south/east-west survey pattern. Another feature of easily 
importance is the continually changing agricultural base in the Peace River Occurr 
District: unlike in New Brunswick, land clearing is occuring at a pace rapid sheet 
enough to outdate even one year old satellite coverage. 
388 
DE SEE CL u "e n M RR EN. PT ES 
s EMEN RE 0 dn > a — 
uum Enc UE