266 
Manual Photo Interpretation and Classification of Aerial Photography 
A two-level classification scheme was drawn up, based broadly on the 
level I and level II approach of the U.S. Geological Survey 
classification (Anderson et al., 1976). Seven level I and 32 level II 
classes were selected based on their probable occurrence in the Aberdeen 
area and on the detail interpretable from the photography (Table 1). 
The Parish of Old Machar, about 29 km 2 in the north east of Aberdeen 
City District, was selected as a pilot study area. This is one of a 
small number of suburban areas in which there have been significant 
on-shore developments associated with the exploitation of the North Sea 
oil and gas fields. One National Grid Square (NGS) with coordinates 
E393 and N810 was selected from the parish to use in the initial 
developing and testing of the method. (A National Grid system divides 
all of the UK into 1 km 2 NGS, geographical and statistical units which 
are identified by east and north coordinates.) 
Using 1966 aerial photography at 1:23,000 scale, a land cover 
classification of this NGS was transferred onto a 1:10,000 scale base 
map via a Bausch and Lomb zoom transíerscope. A transparent grid 
overlay with pixel size equivalent to a ground area of 40m x 50m was 
placed on the classified map and the class of each pixel was transferred 
to a coding form using a unique letter or number for each class. This 
procedure was repeated for a ground cell size of 20m x 25m. The 
resulting data sets revealed reasonably small differences in class areas 
(Table 2). 
Hectareage from Hectareage from 
40m x 50m cells 20m x 25m cells 
Class 
86.0 
2.6 
4.3 
86.6 
3.0 
4.6 
Cropland (A) 
Deciduous woodland (F) 
Quarry (Q) 
Table 2. Class areas in hectares for NGS E393N810 derived from 
two ground cell sizes. 
Since the smaller cell size produces four times as much data which 
requires significantly more handling time, only the larger ground cell 
size was used for subsequent coding. 
The above mentioned interpretation procedure was then used to generate 
classification data for 40-odd NGS. A land cover map was created from 
1:23,000 scale aerial photography of 1966. Aerial photography of 1981 
at 1:50,000 scale was fitted to the 1966 interpreted map using the zoom 
transferscope and the changes since 1966 were marked on the map in a 
different color. Fitting one year's data directly to another year's 
data eliminates the possibility of uncorrected registration problems. 
The 1966 and 1981 classifications were transferred to coding forms. 
Computer Processing of the Data 
To facilitate analysis of the data a software system was developed for 
use on the CBM PET microcomputer (model 3032) with matrix printer (model 
2023) and dual drive floppy disk (model 3040). The programs were 
written in the Basic programming language--specifically, PET Basic 3. 
The software has been designed with emphasis on a geodata base approach-