considered the most general form of base data, then 
came the IGN aerial photography, and, as the most 
detailed level, ground data. 
The circular zone was first divided into 340 
squares called "primary sample units", or PSU's. 
This division was done at the most general level 
with a scale of 1/200,000, and each square 
represented 100 square kilometers on the ground. 
From this population a total of 25 of the PSU's were 
selected using information interpreted from the 
S/V/LF base maps in a list sample procedure. These 
PSU's were marked on the aerial photos, either at 
the 1/60,000 or 1/70,000 scale, and divided 
again into 100 smaller zones, called "secondary 
sample units", or SSU's, each one square kilometer. 
Three of these SSU's were selected using more 
detailed information interpreted from the IGN aerial 
photography, and again the application of a list 
sample procedure. The final stage is to define the 
exact ground locations of the sites. Within each 
selected SSU a point is identified, completely at 
random, representing the location of the site at 
which field data is to be collected. At the PSU 
level an area totaling 2,500 square kilometers is 
addressed, which is only eight percent of the total 
area of a zone. At the SSU level, three percent of 
the total PSU area is selected (only 0.24 percent of 
a total zone). And at the final level of field 
work, data is collected in a manner to represent 200 
square meters. The 15,000 square meters sampled on 
the ground is a very small proportion of the area at 
the SSU level, and extremely small in comparison to 
the size of the the initial zone. 
At this point it is necessary to discuss the types 
of information interpreted at each level for use in 
the list sample procedure. At the PSU level the 
S/V/LF base map of the zone (Section 3.1) was 
interpreted to calculate the surface area of each 
specific ground condition, the TU's, for each of the 
340 PSU's. If the cumulative list, required in the 
list sample procedure, is developed with only this 
information the eventual placement of the sites 
would be a function totally of the relative surface 
area of each TU in the zone, ie., those having a 
relatively large surface area will have a 
correspondingly higher chance of having field 
samples in them. Evaluation of the S/V/LF base maps 
showed that generally, primary forestry lands occupy 
relatively small proportions of the area of a zone. 
If the random sample selection were to be done, 
weighted only by relative surface area, the 
procedure would clearly over-sample the agricultural 
and marginal forestry types, and under-sample the 
higher value forestry value types. This would not be 
an efficient use of the limited sampling resources 
available to the RIM section and would not maximize 
the value of the information presented to the G0N. 
It was, therefore, necessary to incorporate in the 
list sample a means of biasing the selection towards 
those TU's with relatively high forestry values. 
Figure 3 shows an evaluation sheet which was 
developed to assign these relative forestry values 
to each TU. Examination of the sheet shows that 
these values were determined on the basis of 
characteristics presented in the TU descriptions. 
However, with an average of 35 different TU's being 
observed in an urban zone, and 340 PSU's, it would 
be very time consuming to calculate a list value for 
each PSU, using each individual TU surface area and 
forestry value. Therefore, an average forestry value 
was calculated for each of the cartographic units 
mapped in the zone. This was done by multiplying the 
forestry value of each TU observed in a particular 
CU (see Figure 2 which shows the information 
presented on the S/V/LF base map) by it's relative 
surface area in the CU, adding the contributions of 
each TU, and then dividing the total by the number 
of square kilometers in the CU. This figure 
represents the average forestry value of each square 
kilometer in the CU. In order to arrive at a total 
value for each PSU the S/V/LF base map was 
interpreted to estimate the proportion occupied by 
each CU observed, multiplying this percentage by 
it's corresponding forestry value, and then adding 
all the CU observations to yield a "weighted" total. 
In this manner PSU's which have a high proportion of 
TU's with high forestry values will have a greater 
chance of being one of the 25 PSU's selected. A 
cumulative list is then created using the 340 
"weighted" values of each PSU in their sequential 
order. Twenty-five random numbers were generated 
using the random number algorithm on a Hewlett- 
Packard 11c hand calculator, and transformed to the 
interval of the cumulative list. These numbers fell 
into intervals which indicated which of the PSU's 
had been selected. This was the completion of the 
sample selection at the initial stage. 
Action in the second stage was initiated by 
identifying and assembling the aerial photos which 
gave stereo coverage for each of the 25 selected 
PSU's. The limits of the selected PSU's were marked 
on the appropriate aerial photos, the exact 
placement supported by interpretation of the LANDSAT 
images and the IGN topographic maps. On a clear 
plastic overlay material a grid was drafted which 
divided the area of the PSU into the 100 SSU's. This 
overlay, with it's grid at the scale of the specific 
aerial photos being interpreted, was placed on the 
IGN aerial photos in the position of the PSU. Using 
a mirror stereoscope, each SSU was interpreted to 
identify the terrain units present, their surface 
area percentage, and an estimation of their 
vegetation cover percentage category (1 to 5). At 
this SSU level the cumulative list contained 100 
weighted values which were developed by multiplying 
the relative surface area of a TU by: 1) it's TU 
forestry value, and 2 ) it's cover percentage, and 
then adding the contributions of all the TU's 
observed to get the weighted total. Therefore, at 
this level the selection process is biased by three 
Taxonoxic Unit Date Evaluator 
Figure 3. The data sheet used to calculate the forestry value for each of the terrain units. This value was 
used to bias the list sample selection towards those units which had relatively high forestry value because 
of constraints imposed on the collection and processing of field data. The values were calculated using 
information contained in the individual terrain unit descriptions.