719 
aerial 
igery the best 
ird 4,5,7 band 
:o blue, green 
and Laplacian 
nel. 
illy discerned 
sily so in 
small field 
:hickly wooded 
il contrast is 
is and rivers 
using Landsat 
i the case of 
id contextual 
study area and 
he Landsat TM 
ater bodies as 
spen moorland, 
ctares is more 
itifying small 
This compares 
dsat MSS and a 
for SPOT in 
metre- spatial 
1986) . Using 
itrast between 
features was 
hectares and 
m area of lake 
ected 
0.6 ha 
2.4 ha 
<0.5 ha 
<0.2 ha 
isier landcover 
SS imagery. In 
eas of forest, 
rban areas can 
d.Using aerial 
:ures enable 
jor landcover 
types to their 
important for 
TM imagery was 
imagery for 
o analyse this 
n as more than 
iater ease of 
ge areas this 
prohibitively 
tion was found 
lor necessary 
is and drainage 
W Photos Map 
thel:50000 scale Ordnance Survey map. It should be 
borne in mind that the drainage details on the 
1: 50000 map are taken from 1:10000 scale maps of 
the area. On the face of it these figures look 
very poor, however on closer examination of the 
Landsat TM details and comparison with field work 
readings the situation is much more promising (see 
Table 3) . The reason for this improvement is due 
to over half the drainage network consisting of 
streams less than 3 metres in width. 
Table 3. Details of streams identified from 
Landsat TM imagery. 
Number No.on TM % Correctly 
in field imagery identified 
Streams over 5 28 26 92 
metres wide 
Streams over 3 43 36 83 
metres wide 
The main drainage channels were easily determined 
on the enhanced Landsat TM imagery (see Figure 3). 
The detail of minor streams enables greater 
accuracy in delineating the catchment boundaries, 
than is possible using Landsat MSS imagery. Many 
smaller streams, some only 0.5 metres wide could 
also be identified, though there was great 
inconsistency of results at these widths, 
identifiable streams frequently being associated 
with minor tributary valleys a few metres deep. 
Landsat MSS enhancements were only consistent in 
delineating stream widths of 10 metres or more 
(of which there are few in this area) . Some 
streams as narrow as 3 or 4 metres appear 
sporadically. 
is 
km 
21.50 272.24 
0.52 1.15 
Streams delineated on 1:50000 map 
Streams delineated on Landsat TM imagery 
Figure 3. Comparison of stream delineation. 
channel length 
all three types 
extracted from 
When streams were ordered according to Horton's 
scheme of stream ordering (Horton, 1945) , the 
following results were obtained: 
Table 4. Stream orders from Landsat TM, 
photos and 1:50000 map. 
Stream Order 
1 
2 
3 
4 
5 
1:50000 map 
156 
36 
15 
3 
1 
Landsat TM 
33 
6 
2 
1 
Aerial photos 
72 
16 
5 
1 
Table 4 shows a variation in number of stream 
orders between the 1:50000 map and the satellite 
and aerial imagery. The fifth order stream on the 
map corresponds to the fourth order stream on the 
satellite and aerial imagery. It is at the first 
order where the discrepancies occur and these have 
an effect on subsequent stream orders. The 
drainage network on the 1:50000 map is derived 
from 1:10000 maps and includes many first order 
streams only 0.5 or one metre wide. Some of these 
can be delineated from the aerial photography but 
few from the Landsat TM imagery. 
Figure 4. Number of streams v Stream order. 
Under normal basin conditions the gradient of 
the graphs should fall within the range 3-5 
(Smart,1972). All three graphs lie within this 
range, their values being 3.51 for the 1:50000 
map, 3.16 for the Landsat TM imagery and 4.03 for 
the aerial photography. 
Using the Landsat TM imagery catchment 
boundaries can be defined with reasonable 
accuracy, though other catchments need to be 
examined before this accuracy can be realistically 
quantified. Landsat MSS imagery cannot be used 
for delineation of catchment boundaries in this 
area, due its delineation of only the largest 
streams in the area. Over larger catchment areas 
it has proved more useful- see Drayton and 
Chidley, 1985. 
One noticeable feature revealed during the period 
of this research was the variation in brightness 
on the colour slides when similar image 
enhancements had been carried out. This is due to 
one of four factors, namely the different image 
processors used, the level of illumination in the 
image processing area, the different matrix