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

844 
Table 2. Terrain and land cover characteristics of the five study areas (Source: U.S. Department 
of the Interior, 1970, The National Atlas of the United States of America). 
Area 
Mean 
heiqht(m) 
Land-surface 
form 
1. St. Joseph, Mo. 
346 
irregular plains 
2. Mobile, Al. 
67 
irregular plains 
3. Tallulah, La. 
30 
flat plains 
4. Louisville, Ky. 
247 
open hills 
5. Starling, Co. 
1,369 
irregular plains 
Vegetation 
Land 
Soil tvoe 
cover 
use 
Mollisols 
(Udolls) 
Oak-hickory 
mostly 
cropland 
Ultisols 
pine 
forests and 
woodland grazed 
Inceptisols 
oak-gum- 
cypress 
cropland with 
pasture 
Alfisols 
Oak-hickory 
woodland with 
some cropland 
and pasture 
Mollisols 
(Ustolls) 
pine 
grassland and 
grazing land 
Table 3. Population estimation using area as input to a linear regression 
model and an allometric growth model 
Linear: P -- 
Area 
a + bA 
r* 
a 
b 
Allometric 
log P = log a 
r* Ioga 
+ b log A 
b 
(1) 
St. Joseph, Mo. 
0.648 
-3921.41 
27.78 
0.896 
-0.5854 
1.5940 
(2) 
Mobile, Al. 
0.973 
-2516.13 
116.87 
0.828 
1.1595 
1.2620 
(3) 
Tallulah, La. 
0.706 
291.65 
14.82 
0.742 
1.3861 
0.9004 
(4) 
Louisville, Ky. 
0.728 
3431.46 
8.05 
0.867 
2.3190 
0.5781 
(5) 
Sterling, Co. 
0.985 
-254.99 
14.87 
0.814 
0.2057 
1.2724 
*r = correlation coefficient,- all significant at a level of 5 percent 
or below 
helps to explain why in the case of the North 
China Plain even very small villages can still be 
detected (Fig. 2). The high degree of compactness 
of these formerly walled Chinese settlements made 
them good corner reflectors to radar signals. 
Another observation is that the detectability of 
the settlements appeared to be affected also by 
the nature of the geographic region (Table 1). 
The Gulf Atlantic Coastal Plain region came out to 
be the worst of all four regions while the Great 
Plains region was the best. To assist further in 
understanding the effect of terrain characteris 
tics and land cover types on the detectability of 
settlements. Table 2 was compiled. It appeared 
that mean terrain heights, soil types, and land 
use were important factors. High terrain, 
irregular plains, Mollisols (soils with nearly 
black, organic-rich surface horizon) with grass 
land and grazing land of the Sterling, Colorado 
strip in the Great Plains (Fig. 7) seemed to 
provide favorable conditions for the detection of 
settlements. On the other hand, the low, forest 
covered terrain and the lowlying alluvial plain of 
the Mississippi river covered with cropland on 
Inceptisols (wet soils with weakly differentiated 
horizons) were unfavorable (Fig. 5). These 
environmental conditions have probably affected 
the settlement-background contrast, thus making 
the detection difficult. 
4.2 Accuracy of settlement area measurement and 
population estimation 
An important application of the SIR-A data is to 
determine the area and population size of the 
human settlements detected. A commonly employed 
method is to measure the areas of these settle 
ments and then input them into a mathematical 
model linking area (A) with population (?). A 
popular model is the allometric growth model in 
the form of log P = log a + b log A (Lo and Welch, 
1977). In the present research, the area of each 
settlement was first measured with 1-mm square 
grid's directly- from the SIR-A images and then from 
the 1:250,000 scale, topographic map. It was found 
that the image area and map area of the individual 
settlements exhibited a very strong correlation of 
0.92 at 0.01 per cent level of significance. 
However, it was observed that all the measured 
image areas were exaggerated by a factor of 1.5X 
from the actual map areas. This may be caused by 
some human errors in measurement, but careful 
inspection revealed that more significantly the 
strong radar backscatter had produced a glare 
which tended to exaggerate the size of the 
settlement. It was fortunate that this 
exaggeration appeared to be constant and could be 
easily corrected. 
Despite some discrepancy in time, the 1980 
population figures of these settlements in 
different regions were correlated with measured 
image areas first in the form of a linear 
regression model and then in the form of the 
allometric growth model mentioned above. The 
results (Table 3) indicated overall strong 
relationship between population and area in the 
allometric growth model for all regions. It is 
noteworthy, however, that in Mobile, Al. (Fig. 4) 
and Sterling, Co. (Fig. 7) regions much stronger 
relationship existed with the linear regression 
model than the allometric growth model, a 
suggestion that the rate of settlement growth 
might have been faster in these two regions than 
in the others. These results indicated that 
settlement population estimation using settlement 
area as an independent variable could produce 
reasonably accurate results. 
4.3 Shaj 
It is ol 
settlemei 
correct 
is part: 
settlemei 
the den: 
the fav< 
is the 
distinct: 
some sel 
azimuth 
result. 
identify 
the fori 
the lati 
smooth i 
(Fig. 4) 
5 CONCLUi 
This re: 
exists : 
settlemei 
strong 
detectab] 
viewing 
settlemei 
observatj 
orlentinf 
the sel 
capabilit 
Culturt 
detectabj 
settlemei 
walls ii 
distinguj 
American 
Within 
detectabi 
because 
because 
type, v« 
that hif 
and Mo3 
landform 
image-ba< 
detectabj 
emerges 
settlemei 
environm« 
azimuth 
settlemei 
having i 
to detecl 
An imj 
very sti 
areas ai 
image ai 
areas oi 
usefully 
for popi 
regions. 
are not 
images, 
other ft 
when th< 
achieved 
To coi 
optically 
two-dimei 
resolutic 
settlemei 
other i 
yielding 
adverse 
have be« 
data to 
settlemei
	        
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.