Full text: Remote sensing for resources development and environmental management (Volume 2)

710 
BAND 4/BAND 6 
BIOMASS INDEX 
LAND USE CLASSIFICATION 
3 BAND COMPOSITE 
PRINCIPAL COMPONENTS 
Figure 1. Thematic partitions of the Taff River basin 
RO = f (RF) A'. 
c c 
where the suffix c refers to the whole catchment. 
We examined hydrological records for this area and 
found that the runoff function was a simple linear 
relationship, and hence the yield in millimetres over 
the catchment, 
Yield = — = a RF - b (mm) 
c 
Therefore, at the simplest level, our satellite 
classifications will be useful if they can be used to 
predict the values of proportional runoff, a , and 
losses, b , for any catchment. 
Our approach was to derive parameter values based 
on the conventional image processing techniques of 
density slicing and classification. Thus we could 
slice a Near Infra Red image and estimate the total 
area in the catchment in each of say 5 levels. The 
same process could be applied to Biomass Index, 
Principal Component or any band ratio. Alternatively 
a classification could be made, and the areas falling 
within various classifications measured. This applied 
to supervised or unsupervised classifications and any 
colour composites. 
Thus we could derive a set of variables such as 
LEVELAI, LEVELA2...LEVELA5, LEVELB1...LEVELB5, 
CLASSAI, CLASSA2, CLASSA5, etc., where LEVELA 
refers to a density slice in NIR, CLASSA refers to 
an unsupervised classification etc. The values of 
constants a and b could then be evaluated by multiple 
regression. 
Clearly a very large correlation analysis could 
ensue, so some preparatory work was needed. Only 
those parameters which exhibited a large variance 
and showed a strong correlation with streamflow, and 
showed little correlation with other parameters were 
considered. Once this weeding was done, then a 
multiple regression could be carried out. 
3 METHOD 
A suitable set of 'satellite' catchment character 
istics was chosen. Parameter values were extracted 
for the study catchments. Annual rainfall and runoff 
values were acquired, and for each catchment the 
constants a and b were calculated for the equation 
RO = a RF - b 
The constants a and b were regressed in turn on the 
set of catchment characteristics, and the optimal 
regression equations were found. The equations were 
tested by estimating values of annual runoff not 
previously used. 
4 DISCUSSION AND RESULTS 
The first impression was that the various slices and 
classifications enabled many different partitionings 
of the catc 
when the pr 
calculated, 
classificat 
as soon as 
became simi 
ent. 
Secondly, 
ations occu: 
negative co: 
proportion ( 
tion of LEVI 
correlation! 
FOREST corn 
The const, 
reduced set 
correlation! 
result, it : 
and points 1 
The under! 
the satelli- 
of the regrt 
e.g. CLASSA 
eneous. The 
evaluate th< 
case the ch; 
ion of sense 
catchment. 1 
should lie \ 
Landsat MSS, 
SPOT. This i 
there is in« 
and environi 
Thus, the de 
variance to 
To achieve 
include in t 
more dissimi 
similarity v 
remote from 
preting more 
will inevitc 
and differer 
that we shoe 
ics, and use 
consistent i 
and time to 
argument is 
to: either 
regression n 
Apart from 
clearly lost 
catchment ch 
the two main 
we have lost 
Although t 
unsuccessful 
for average 
annual flood 
summary stat 
sions were f 
place of con 
and supports 
ful if the r 
5 CONCLUSIO 
The hydrolog 
seen to depe 
characterist 
an hydrologi 
teristics as 
because the 
istics from 
was too smal 
would have t 
problems of 
different im 
insuperable.
	        
Waiting...

Note to user

Dear user,

In response to current developments in the web technology used by the Goobi viewer, the software no longer supports your browser.

Please use one of the following browsers to display this page correctly.

Thank you.