437 
3t areas 
Lvation. 
workers in 
uding those 
and Lebon 
t included 
ilicable to 
rmation not 
material. 
the work in 
974)and the 
ence of the 
ssification 
asses. 
oblems with 
comprising 
quicaocally 
1 with the 
The finally 
than lOha 
s greater 
6. INTERPRETATION PROCEDURE 
6.1 It was not possible to make a complete survey 
of the whole area, the object of the project being 
to obtain information with the minimum of expense 
and effort. 
A sampling technique was therefore used 
consisting of a series of primary samples 
(quadrats) subdivided into secondary samples. The 
quadrat size was dictated by the coverage of the 
average stereomodel and formed a rectangle. Each 
rectangle contained 72 dots evenly distributed in 
12 rows and 6 columns. The dot separation was 
arranged in relation to the scale of the 
photograph so that each fell in the centre of an 
area of 2 hectares. This unit of area was chosen 
using two separate formulae developed by Brunt 
(1966) (1) and Bonnor (1975) (2) which estimate 
the error at 95% confidence level of a given 
sample according to its coverage and density: 
8. 
ANALYSIS PROCEDURE 
8.1 The sample data was transferred manually from 
the interpretation record sheet to the storage 
disk of the computer where it was held available 
for analysis. 
8.2 Analysis was undertaken by a series of 
programmes written for the purpose using as a 
guide the BASIC programmes published by Cooke et 
al (1982). These established the coverage of each 
class at the different epochs and showed how the 
frequency of distribution varied from place to 
place within the sample. 
9. LAND USE 
(2) E 
Class 
Year 
% 
Precision 
■k 
(1) 
(2) 
E = 
(100 - P) 384000 
4PN 
Settlement 
1950 
0.9 
14.6 
13.7 
1972 
2.0 
8.6 
9.2 
E 
153.1/(AD)°- 58 
Farmland 
1950 
12.5 
3.2 
3.2 
E = 
1/2 the % error of a 
class in a 
1972 
24.1 
2.1 
2.2 
sample at 95% confidence 
level. 
Grass/Scrub 
1950 
32.0 
1.8 
1.8 
P = 
% area of the class 
1972 
27.8 
2.0 
2.0 
N = 
total number of dots 
Woodland 
1950 
50.0 
1.2 
1.4 
1972 
40.7 
1.5 
1.6 
A = 
area in sq km of the class 
Barren 1950 
4.1 
5.9 
6.0 
D = 
dot density per sq km 
1972 
5.4 
5.2 
5.2 
95% confidence 
emnants of 
analysis, 
five larger 
34 
as 230mm x 
examined 
mirror 
cular high 
suitable for 
lotter was 
optics and 
iced by the 
Land 
use 
form 
and 
APPLE 
II 
and a 
dot 
6.2 According to these formulae it was decided 
to use a primary sample of 89 quadrats, one for 
each of the stereomodels on the 1972 photography, 
and a secondary sample of 72 dots, that is 6408 
points. Regarding 2 hectares as the basic unit, 
this gave a sample of about 17% of the total 
area. It also gave an acceptable precision for 
most of the classes encountered in the survey. 
The exception being settlement, which account for 
such a small surface area, and could not be 
estimated with a high degree of precision by this 
method. 
6.3 Samples were taken on both the 1950 and 
1972 photography in such a way that the points 
sampled were identical in position. This was done 
by a combination of measurement and visual 
inspection to ensure that the quadrat was located 
over the same images on both sets of photography. 
Fortunately, ground height in the test area is 
reasonably uniform and there are therefore few 
distortions due to relief. Those that are present 
are unimportant. The flying height did not vary 
greatly and it was therefore unnecessary to plot 
grids on a large number of different scales. 
7. FIELD CHECKING 
7.1 Before analysis, the sample from the 1972 
photography was subjected to field checking. 
About 10% of the sample ( 8 overlaps ) was 
compared in the field with the ground truth at the 
time of inspection. The validity of this checking 
was not very satisfactory however because of the 
changes in land use between the time of 
photography (1972) and field inspection (1983). 
(1) Brunt (2) Bonnor 
9.1 The table shows the proportions of the 
different classes. In 1950 half the area was 
under woodland, one third under grass/scrub and 12 
% or 94.5 sq.km was farmed. Settlement and 
barren land accounted for under 1% and just over 
4% respectively. Of the grass/scrub some will be 
land which was not actively farmed at that time 
but had been not very long before, i.e. early 
bush-fallow. Another part of the grass/scrub 
class will be fadama which actually was 
recognisable as a separate category occupying 
about 5-6% of the area, say 40 sq km but was not 
analysed separately because of difficulty in 
identifying it on the 1950 photography. 
9.2 By 1972 ten percent of the woodland, 7500 
hectares, had gone, the amount of grass/scrub had 
also fallen by 3200 hectares and most of this land 
had been occupied by the increase in farmland 
which almost doubled its area over the period 
examined. Of the other classes, settlement had 
doubled in area and barren land had also increased 
from 4.1% to 5.4%, an additional 983 hectares. 
However this may not be a true increase, but 
merely an artefact of the interpretation, and it 
illustrates the need for supporting evidence; 
remote sensing is not ground truth. 
10.DISTRIBUTION OF LAND USE 
lO.lThe classes of land use are not distributed 
evenly over the test area, responding as they do 
to physical features, soils, nearness of the road 
and railway, and the time it takes to travel to 
Kaduna. All these factors interact on the