Printer Friendly Format

Page 1 of 1

Abstract

Typical rainfall, particularly during September in each year, causes floods with inundating damaging in Sukhothai province. With the advent of state of the art in radar RS technology, we may obtain the image, regardless of day or night or even cloud piercing period of time. This application gains more popularity recently. In this study, the author employs 3-multitemporal of SAR data and 2-multispectral TM images from RADARSAT and Landsat-5 respectively. The images had been acquired during the inundation of Sukhothai province from the period of 2002 up to 2006. The images were firstly georeferenced and then being subtracted, the SAR images being filtered, hence, yielded the extracted inundated area.

After the post classification processing, the overlays amongst the five-date classified images, were applied with the frequency of flooded and flood risk area had been successfully extracted. The outcome of the processing will enable both government and local authority in the decision making concerning flood disaster prevention and mitigation.

1. INTRODUCTION
Among the several types of disaster, flood disaster is worldwide natural phenomena especially in flood plain like the province of Sukhothai, the lower north of Thailand. The characteristics of land and the meteological statistics of the region are the main factors for this disaster. Such disaster is increased by the human activities on land, like construction, agriculture, deforestation. This area is flooded every year.

Remotely sensed data were intensively employed to monitor flooding and assess its damage during several years. NOAA AVHRR data successfully used to monitor the flood process by taking advantage of its high temporal resolution and Landsat MSS and TM to estimate the land use type in inundated area by its high spatial resolution and multiple bands. However the visible and infrared do not function effectively during the flood seasons due to the cloud cover. The data of SAR (Synthetic Aperture Radar) are the priority data for floods monitoring. For this study, 2 images of Landsat-5 TM and 3 images of RADARSAT acquired in the year 2002 up to 2006 during the inundation over Sukhothai province were selected. Then apply the technique of Geographic Information Systems to estimate and compare the inundated areas and to establish the flood risk map.

2. STUDY AREA
Sukhothai is located in the lower north of Thailand between latitude 16°41'33"N to 17°48'20"N and longitude 99°18'20"E to 100°07'43"E or approximately 440 kilometers in the north of Bangkok and cover an area of 6,596 square kilometers. The province is administratively divided into 9 districts. Its topography is composes of plains, mountains. The rivers in the central of the region. The highest mountain is approximately 1,200 meters from mean sea level. Its climate influenced by southwest monsoons and Northeast monsoons, the average temperature is 27.6 degrees Celsius. Average precipitation during all the year is approximately 1,208.8 millimeters.

3. DATA USED
The 3 images of RADARSAT and 2 images of Landsat-5 TM were used for this study. These images were acquired in 2002 up to 2006 during the floods of each year and another image of Landsat-5 TM was acquired in dry season. These RADARSAT images were acquired on three different dates. The images of September 12, 2002 and October 4, 2005 were acquired whereas the satellite moved towards the south (descending) and the image of September 23, 2003 was acquired on the ascending orbit.

The two images of Landsat-5 TM were acquired during the flood of October 11, 2004 and May 26, 2006. The last image was acquired on March 7, 2006 during the dry season. This image was used to extract the permanent water. The other data which are used in this study are the database of topographic map scale 1:50 000, the land use data and SRTM DEM.

4. METHODOLOGIES

4.1 Image processing

4.1.1 Georeference

This process was carried out with the ENVI 3.6 software. The first stage consisted with georeference the image of Landsat-5 acquired on March 7, 2006 in dry season with the vectors of the topographic map scale 1:50 000 which is carried out by Royal Thai Survey Department (RTSD). Then georeference the Landsat-5 images and RADARSAT images during the flood in the same way but by using the Landsat-5 image of March 7, 2006 as the reference image.

4.1.2 Radar image smoothing
The presence of speckle in SAR imagery reduces the ability to resolve fine details within the image. Excessive speckle can result in an uninterpretable image, so we have to trade resolution for speckle reduction by using a certain filter. Lee filter is selected to reduce the speckle effect by comparison of the filtered images from Lee filter, Frost filter and Kuan filter, on the conditions that a good filter should maintain edges and texture in origin image and filtered image should be more contrast, compared to the original image.

In addition, larger kernel window is suitable for homogeneity analysis. Different kernel window from 5*15 are also experimented, 11*11 kernel is best selected for texture analysis of water bodies on RADARSAT images.

4.2 Image Classification
Classification produces a natural regrouping of the pixels of the image which names “spectral regrouping” or “class”. Thus, it is supposed that the areas of the image having the same spectral signature have a similar type of land use. The analyst must then determine the identity of these spectral regroupings. In this study, the principal algorithm of classification K-means was applied. For work, it is necessary to create a mask around flooded zones to reduce the processes of post-classification. The classification of the flooded zones starting from image of SAR is based on the fact that calm water surface always causes mirror reflection, no echo signal will be found so the image will appears in dark tone.

It is difficult to automatically extract flooded zones from SAR image in mountain area because of mislabeling shade of mountain as flood. To solve this problem, we used the data of SRTM DEM with the function 3D Analyst of ArcGIS to correct the results of the classification of the RADARSAT images. The contribution of the 3D provides information interesting for the detection of the flooded areas. During the flood, there are many cloud cover areas. For this reason, it is difficult to extract the flooded areas in the optical image from Landsat. Consequently, the error of the flooded zones is corrected by visual interpretation.

From the Landsat-5 image of March 7, 2006, the permanent water body was extracted by the segmentation method. This process was done with band 4 of Landsat-5 TM data due to its characteristics which is covered the wavelength from 0.76 to 0.90 µm. (Near Infrared). This band is used to interpret the various types of vegetation, detecting moisture in the ground and for the delineation the borders of water and the ground. After having detected the flooded zones of the years 2002 up to 2006, the following stage consists in making the subtraction with the zones permanent water to obtain the actual flooded zones for each year.

Flooded area in the period of 2002upto2006Flood Map Flood Risk Management Flood Risk Map Filtering Georeference Image Extraction Image Classification RADARSAT-1 2003LANDSAT-5 2004RADARSAT-1 2005LANDSAT-5 2006Satellite Imageries Dry season Wet season RADARSAT-1 2002 Permanent water body Actual Flooded area


Figure 1: General methodology in this study.

5. RESULTS
5.1 Extraction of flooded area

The flooded zones were extracted successfully. The flood affected area for every year was also calculated and presented in the following table. The percentages of flooded zones were calculated by compared to the total surface of Sukhothai province (approximately 6,651 square kilometers).


Table 1: Total flooded area by year.

After overlaying of the actual flooded area of each year, the frequency of flood was obtained. Then we can find the zones of flood risk by the use of the frequency of flood.


Figure 2 : Flooded area in Sukhothai province from the year 2002 up to 2006.


Figure 3: Flooded risk map of Sukhothai province.

To mitigate flood disasters, the flood protection priority for the province was classified the priority of starting from the flood risk area. The very high priority is for the districts which have the maximum value in frequency of flood during 5 years and the low priority corresponds to the districts having the minimal frequency which have less frequency of flood.

5.2 Land use statistics in flooded area
Overlaying the flooded area with the land use map interpreted from Landsat TM image, the flood affected area was calculated and have shown in the following table. In this study, the most affected by flood was the paddy field and the most flooded zone was in 2006.


Table 2 : Area of land use in inundated zones of Sukhothai province.

6. CONCLUSION
As the flood disaster occurs regularly on the plain of Sukhothai province, effective and timely information of flood monitoring and their effect are essential for the flood prevention program. Being weather independent, microwave remote sensing can provide important information for flood study. By applying the remotely sensed data with the Geographic Information Systems, the actual flooded area, flood affected area from the period of 2002 up to 2006 had been successfully extracted. The use of the RADARSAT data combined with Landsat TM data for the analysis of the flooded zones is encouraging. The extracted information can be used for the flood management and flood monitoring.

References :
Campbell James, 1996, Introduction to Remote Sensing, Gilford Press, New York.
J. B. MacQueen, 1967, Some Methods for classification and Analysis of Multivariate Observations, Proceedings of 5-th Berkeley Symposium on Mathematical Statistics and Probability, Berkeley, University of California Press, pp. 281-297.
Kiyoshhi H., Francis X.J Cannisius, Bhuwneshwar P.S., 1997, Flood Monitoring in Central Plane of Thailand Using JERS-1 SAR Data, Asian Institute of Technology, Pathumthani, Thailand.
Ping Chen, Soo Chin Liew, Hock Lim, 1999, Flood detection using multitemporal Radarsat and ERS SAR data, National University of Singapore, Singapore.
Yang Cunjian, Zhou Chenghu, Wan Qing, 1999, Deciding the Flood Extent with RADARSAT SAR Data and Image Fusion, Chinese Academy of Science, Beijing, China.
Yang Xiaomei, Zhou Chenghu, 1998, Recognition of Flooded Area in Radar Image using Texture Feature Analysis, Chinese Academy of Science, Beijing, China.

Page 1 of 1