E DAM- 
“TED BY 
erial pho- 
y the aerial 
om the out- 
eas, and its 
erated. 8] 
and slight 
leck points 
nified im- 
S the heavy 
slight. The 
eck points 
g the aerial 
a discrimi- 
the check 
sat TM data 
Among 40 
eavy dam- 
oints were 
amage,and 
oints as no 
raphs. The 
percentage. 
ated as the 
"M data,22 
surveyed 
nita Prefec- 
wn in Fig.l 
iparities by 
the ground. 
syed by the 
the evalua- 
; computed 
naged areas 
ad areas for 
  
each municipality surveyed by the Local De- 000 
  
velopment Office and estimated by Landsat (ha) | 
TM data,and Fig. 6 shows a relationship be- 
o 
© 
© 
eo 
tween the both data. Total data including the 
heavy and slight were used as the damaged 
| RIN 
areas estimated by Landsat TM data.It can be 
ine 
4000 
  
y = 0.3379x + 584.41 Li 
seen from Table 4 that the areas of the heavy 
R° = 0.79 
   
     
damage estimated by Landsat TM data are 2000 
Damaged areas surveyed by the 
Local Development Office 
smaller than those surveyed by the Local De- 
velopment Office,but including the slight e 
  
  
  
damage,the damaged areas by Landsat TM pie j i 
| 0 2000 4000 6000 8000 (ha) | 
become larger. As trees are partly fallen | 
a 5 par Damaged areas estimated by landsat TM data 
in the slight damage,it is considered that the | | 
: Fig. 6. Relationship between the damaged areas surveyed 
grass of fl lici damogc may Re estimated bri e Local Development Office and estimated by Landsat 
larger than the realities by Landsat TM data data 
including no damage area. The damaged ar- 
eas surveyed by the Local Development Office may count ‚but the points discriminated as the slight damage include 
only the areas of fallen trees where fallen trees and no dam- not a little the heavy damage (34%) . Putting together the 
age trees are mixed. The regression analysis between the — heavy and slight damage, the damaged areas can be ex- 
both data in Fig. 6 shows a high correlation coefficient of tracted with an accuracy of about 90%. Deciduous trees or 
0.893 but not a little rms error of about 1700 ha. pine trees are some times discriminated as the damaged ar- 
It can be said that damaged areas can be extracted by eas. 
Landsat TM data but the estimation of their absolute areas (4) The damaged area for each municipality within the test 
may be difficult. site was computed from the the extracted image of the dam- 
aged areas by Landsat TM data, and were evaluated using 
7 . CONCLUSION the damage area for each municipality surveyed by the Lo- 
cal Development Office through a regression analysis. The 
Damaged areas of fallen trees by typhoons using two tem- correlation coefficient between the both data was very high 
poral Landsat TM data acquired before and after the dam- | (0.893) ‘but the rms error was considerably large (1700ha) 
age were extracted. The aerial photographs taken immedi- . The damaged areas can be extracted by Landsat TM 
ately after the damage were used to examine the change  data,but the estimation of their absolute area may be diffi- 
characteristics of bands 1~5,7 and VI due to the cult. 
damage. The damaged areas extracted by Landsat TM data 
were evaluated using the aerial photographs and data on REFERENCE 
the damage surveyed by a Local Development Office. The 
following knowledges were obtained from this study. [1] G.Takao,Detection of the Windfall Damage to For- 
(I) TM band 5 and 7 show a tendency to increase more ests Caused by the Typhoons 9117 and 9119,Proceedings 
than other bands and Vegetaion Index (VI) shows aten- of the 13th Conference on Remote Sensing,pp.A-2-8-1~ 
dency to decrease due to the damage. A-2-8-6,1992 
(2) The damaged areas can be extracted classified into two 
groups of heavy and slight by a supervised maximum like- 
lihood classification method using a registered image of 
band 5,7 and VI of two temporal data. 
(3) . The points discriminated as the heavy damage by 
Landsat TM data were almost the heavy damage (7596) 
515 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996