In: Wagner W., Szekely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Vol. XXXVIII, Part 7B 
555 
observations are then used to evaluate water fractions derived 
from MODIS observations. The scatter plot is shown in Figure 
5, the evaluation results show correlation between MODIS and 
TM water fractions is 0.966 with bias of 2.16%, standard 
deviation of 3.89%, and rms of 4.45%, for total sample number 
of 50423. 
90‘W 
89 W 
31‘N 
30*N 
29* N 
31'N 
30“N 
29” N 
31*N 
(c) Aug. 27, 2005 
30'N 
29* N 
92‘W 5 
9TW 
0 10 20 30 40 50 60 70 80 90 100 
Figure 3. Water fraction map on August 31 (a), 30 (b), and 27 
(c), 2008.Figure 3. Water fraction map on August 31 (a), 30 (b), 
and 27 (c), 2008. 
5. SUMMARY 
In this study, the Regression Tree technique is applied to water 
body and flood identification with the EOS MODIS data. 
MODIS data has the advantage of global coverage, and so can 
be available worldwide. MODIS surface reflectance with the 
matched surface percent water data before flooding are used for 
training with the RT method. MODIS surface reflectance data 
at 500m resolution are aggregated to the same 1/6 degree 
resolution as the percent water data. When we test the rules and 
regression models obtained from the training to “predict” or 
model future flood, in order to get the geolocation information, 
we use the Level IB swath 1km calibrated reflectances at the 
TOA with the matched geolocation fields. 
31‘N 
30‘N 
29* N 
3fb 
92‘W 91‘W 90‘W 89‘W 
, j pipil % 
0 10 20 30 40 50 60 70 80 90 100 
Figure 4. Flood map on August 31 (a) and 30 (b), 2005 shown 
as the water fraction difference after and before flooding 
(August 27). 
Figure 5. Scatter plot of water fractions of MODIS and TM on 
08/27/2005 using regression-tree algorithm. 
The time series of water fraction maps are generated, 
monitoring area changes of inundation. The flood maps are 
derived by calculating the difference in water fraction before 
and after flooding, and show promising results. The successful 
applications of MODIS observations to water body and flood 
identification demonstrate the effectiveness of the RT approach. 
REFERENCES 
Ali, A., D. A. Quadir, O. K. Huh, 1989. Study of river flood 
hydrology in Bangladesh with AVHRR data. International 
Journal of Remote Sensing, 10(12), pp. 1873-1891.