In SAR images, brightness values will decrease when the earth's surface is flooded.

We utilized this fact to conduct a research to develop a method for detection of flooded area based upon satellite SAR image data.

For the Naka River flooding that occurred on August 28, 1998 we detected flooded area from the differences in brightness values of RADARSAT satellite SAR images taken before and after the flood, and evaluated the precision of the result of detection through comparison with the reference data taken from other sources.

1) Satellite data used: RADARSAT SAR images

Prior to flood: December 8, 1997
After flood: August 29, 1998

2) Summary of analysis

We created a digital orthophoto-map by combining a 1/25,000 topographic map with aerial photographs and 50m interval digital elevation data. We used this to make geometric corrections of SAR images captured on the two different dates. Finally, we made brightness corrections to make the scattering influence from phenomena other than the flooding as uniform as possible.

From these two images we created an image of the difference in the brightness from the time periods before and after the flooding, established threshold values, and detected the flooded area. We compared this to reference data created from sources such as field surveys, hearings and interpretations of photographs in order to verify the result.

1) The flooded area detection accuracy at a threshold value of 3dB (the proportion of flooded area in the reference data to regions judged as flooded by using the SAR data) was 86%, and the detection accuracy at a threshold value of 5dB was 70%.

2) The flooded area detection error rate at a threshold value of 3dB (the proportion of areas judged to be flooded area based on the SAR images that were not flooded area in the reference data) was 36%, and the detection error rate at a threshold of 5dB was 28%.

3) The SAR brightness is almost entirely independent of the floodwater depth.

We attained the specified results for detecting flooded area by using SAR images. In many instances it is impossible to make observations during flooding periods because the weather during the flooding period is rough and satellite optical sensors will be greatly affected by the weather. SAR images, however, can be expected to provide a means to understand the state of flooding at the flood peak as long as the timing for taking the images is suitable.