Qualitative approaches to rapidly identify completely submerged rice due to tropical cyclone using satellite data
MetadataShow full item record
The objective of the present study is to identify completely submerged rice areas due to tropical cyclones using remotely sensed data. The Kendrapara district of Orissa state hit by a tropical cyclone on 30(th) October 1999 is considered as study area and for this area, pre event (October 11, 1999) visible-near IR image and pre (October 11, 1999) and post event (November 2,4999 and November 4, 1999) Radarsat images were procured. The pre event IRS ID LISS III (resolution = 22m) image of Kendrapara district was geometrically corrected and classified into several landuse and landcover classes. Supervised classification technique was used for landuse/landcover classification. This landuse/landcover map is assumed to be accurate and is used as a base map in the present study. Pre and post event Radarsat-SAR images were also geometrically corrected. Further preprocessing included speckle noise removal, data calibration and incidence angle adjustment. Based on literature, a threshold of -16.5db (DN value =100) was chosen to classify each pixel in pre Radarsat-1 SAR image as water or non-water. The landuse/landcover map was used to identify the rice regions in the pre and post-event Radarsat images. Application of the threshold allows for the determination of the submerged rice areas. To determine the validity of a single threshold, water pixels in pre event Radarsat-1 SAR images were extracted corresponding to the base map. A histogram of these values suggests that a single value threshold approach may not be fully accurate. To overcome these limitations, two alternative approaches, namely image histogram and change in db were formulated. For both approaches, the rice pixels in pre and post event Radarsat-SAR images were extracted corresponding to base map rice pixels. In case of image approach, a histogram was plotted for the DN values of the pre and post Radarsat-1 SAR rice pixels. This allows the qualitative identification of the submerged rice areas. Using change in db approach, pixel-to-pixel change in db in pre and post event radarsat-1 SAR images in rice pixels was calculated. Analysis of these values allows for the identification of different effects of submergence on the rice area. This type of analysis will help policy makers in determining the extent of submergence and could serve as a tool for rapid assessment of damage and help expedite release of relief funds and aid proper allocation of funds to the affected areas/people.
- Proceedings papers