acts, to which we 
nsity with stable 
)r one season or 
>ne or more 
>arcels 
3 managed, 
an one 
especially in the 
much more 
3l. 
suggesting the 
d an economic 
Itural censuses, 
ively managed 
Land (tow 
:terised by: 
intensity of use 
here are usually 
f use which 
fragmentation which is characteristic of a substantial 
proportion of these intensively used rural areas high 
spatial resolution data are needed. By high resolution 
we mean at least 30 metres and to be certain of acquiring 
pure pixels and not mixels a resolution of 10 metres 
would be appropriate. Since such high spatial 
resolutions put a great strain on data transmission, 
archiving and processing systems there is an urgent 
need to determine the extent to which economies can be 
made in data capture and handling procedures. 
3 APPROPRIATE ECONOMIES IN DATA PROVISION 
Spatial data can be classified according to a number of 
criteria and for the purposes of the argument here we 
shall adopt the folowing: 
Vector data (point, line and area information) 
Raster data (information by individual grid squares) 
Static or relatively static spatial data - changes after 5 
years 
Dynamic spatial data - changes weekly 
The place of data on renewable natural resources falls in 
the position shown below in the following matrix: 
tract with a limited range of crops and substrate, namely 
two types of grassland and bare soil. The date of the 
imagery was 4 February 1983. 
The data were analysed to determine the effect of the 
level of sampling on the spectral information as such a 
measure would seem to indicate the extent to which 
spectral discrimination would be possible as well as to 
reveal the effectiveness of sampled data. The 30 metre 
spatial resolution of Landsat TM provided data which 
showed a large number of pixels in this particular study 
area where the fields were all a number of hectares in 
extent. 
Field S1 S2 G1 G2 L1 L2 
Area (ha) 7.2 11.16 3.69 8.82 10.26 10.98 
S1 and S2 were bare soil 
G1 and G2 were under grass 
L1 and L2 were under another grassland type 
The samples from the indivdual parcels were built up 
from five randomly selected pixels within the parcel. 
Boundary pixels (mixels) were not used in the samples. 
A total sample for each parcel of thirty pixels was built up 
by taking five further sets of five random pixels. The 
results were as follows: 
ick usually 
crop and 
e regions where 
d no economic 
ensive 
3mote sensing 
d types of land 
sted that these 
)ercent or more 
i managed 
o of land falls 
ns of 
ills outside the 
3 sensing. The 
e of cover which 
r. 
is associated 
nonitoring, 
stified if it is 
ihanges in land 
.te that only low 
II be appropriate 
obably in 
onniques. 
r that remote 
vhere at the 
dhave the 
jces at least 
p and livestock 
)mprises only 
ngeland cover, 
suit of the 
ecause of the 
Vector Raster 
Static Topographic 
mapping 
Dynamic — Renewable 
natural 
resource data 
All of the above types of data have an important role in 
renewable resource surveys but they can only be utilised 
if there is some means of merging them. The following 
section deals with a method by which this could be 
achieved. 
4 SAMPLING PRINCIPLES FOR RENEWABLE 
RESOURCES STUDIES 
It is unfortunate that the types of data relevant to the 
detection and mapping of renewable natural resources 
surveys fall into different spatial data domains, that is the 
vector data concerning the posiitons of the relatively 
static parcel boundaries and the raster data relating to 
the land cover. Yet It should be possible to merge the 
two types of data to maximise the effectiveness of the 
raster data which contain the dynamic land cover 
information but which are not effective in locating the 
precise position of linear features. It is the static 
information on the extent of parcels which is essential as 
comroi on to which some or all of the dynamic data can 
be registered. 
In the introduction it was emphasised that there was a 
need to economise in the volume of data handled 
especially as increased temporal resolution would be 
needed in future for reliable discrimination. The 
question should therefore be asked how few spatial data 
can suffice in discriminating parcels of particular crops or 
cover. In order to answer this question remotely sensed 
TM data of an area in southern England were used of a 
Grassland type 1 - mean & standard deviations 
of dns for TM bands 3 and 4 
Sample Spec, 
size band(TM) 
5 4 mean 
3 
4 s.dev 
3 
10 4 m 
3 
4 sd 
3 
15 4 m 
3 
4 sd 
3 
20 4 m 
3 
4 sd 
3 
25 4 m 
3 
4 sd 
3 
30 4 m 
3 
4 sd 
3 
Parcel G1 Parcel G2 
38.47 
35.07 
18.83 
16.03 
2.46 
1.87 
0.25 
0.31 
38.43 
35.87 
18.91 
16.26 
2.79 
2.10 
0.26 
0.39 
39.54 
36.92 
19.8 
19.15 
3.0 
2.38 
0.31 
0.44 
38.23 
36.08 
18.71 
16.76 
3.37 
2.66 
0.35 
0.51 
38.57 
37.11 
19.21 
17.24 
3.71 
3.20 
0.40 
0.57 
39.12 
37.24 
18.65 
17.30 
3.86 
3.72 
0.44 
0.65 
A number of features are evident in the above tables. 
First the three types of land cover are easily 
distinguishable in the spectral data, even the two very 
similar types of grass cover. Secondly the cover types 
can be as easily distinguished on the basis of five pixels 
as on the basis of 30 pixels or total cover. 
The total areas of the parcels ranged from 3.69 hectares 
to 11.16 hectares. In the former the number of TM pixels 
which could fall in the parcel would be 41 and in the 
latter 124. The results indicate that in areas where the 
parcels size is as large as 10 hectares it is possible to