Abstract
Microtopography is essential for reconstructing water flows resulting from tsunami events and for estimating the resultant damages. In order to construct the microtopography efficiently, it is essential to (1) develop a high-resolution digital surface model (DSM) of the flooded region on the basis of stereo-pair images and (2) quantify the accuracy of the generated model. However, using the stereo-pairs of aerial photographs may not be feasible because tsunamis generally affect coastal regions to a broader spatial extent. Moreover, in many cases, the archives of stereo-pair satellite images of tsunami-affected areas are not satisfactory because such images are captured only when they are requested.
In this study, we attempt to construct a DSM by using two single images captured by the satellites IKONOS and QuickBird for the Nam Khem plain located along the south-western coastline of Thailand, which was affected by the December 26, 2004, Sumatra-Andaman tsunami. These satellites have different satellite azimuths and elevation angles.
The DSM is generated using the stereo pattern matching algorithm. We experiment with template sizes of 11, 15, 19, and 23 pixels, and from the observation of stereoscopic satellite imagery, it is found that the template size of 19 pixels is the most accurate DSM.
For the DSM of the Nan Khem plain, the root mean square error (RMSE) of the elevation is calculated to be 1.32m.
