High Resolution Remote Sensing Data & GIS Techniques in Updation of Infrastructure Details for Flood Damage Assessment - A Case Study
2. Study Area
Marigoan district of Assam State is taken up for this study as this district is severely affected by the floods almost every year. It is covered in 7 topographic maps on 1:50,000 scale and has an area of 1522 Sq. km. The general climate of the district is hot and moist except in the months of December and January. The average temperature of the district is 350 C with an average humidity of 70% - 80% but in peak summer temperature rises to 370 - 400 C during May-June. The average annual rainfall is about 200 cm.
3. Data Used
- IRS -1C, LISS -III data path / row no. 111/53 of 19 February &18 December, 1998
- IRS -1C PAN data path / row no. 111 /53 B & D of 28 March, 1998
- IRS -1C WiFS data of 8 September, 1998
- Topographic maps on 1:50,000 scale
- Village boundary map
4. Details of Investigations
In the present study digital image processing has been carried out using ERDAS IMAGINE and EASI/ PACE image processing software. The former is used for creation of digital map database, rectification of PAN, LISS -III, WiFS data, while the later is used for classification of LISS - III data and fusion of LISS -III+ PAN data. Digitization of infrastructure layers, natural features and integration of thematic layer with satellite data is carried out using ARC/ INFO software.
4.1 Preparation of Digital Map Data Base
A digital map database is created on 1:50,000 scale topomaps covering Marigoan district using 200 Dots Per Inch (DPI) with corresponding resolution on maps 0.127 mm which is equivalent to 6.35 m on the ground. Though at least 400 DPI is recommended but keeping in mined the computer memory space requirement with respect to 400 DPI and the utility of digital map database, 200 DPI has been accepted as a trade-off. Moreover, 200 DPI corresponds to a spatial resolution of 6.35 in the ground (Das RK et al 1999). This database fulfill the requirement of compatibility with IRS -1C PAN data of 5.8 m. All digital maps are geo corrected and mosaiced and one uniform base is prepared for district extraction and digitization of infrastructure and other details for updation and change detection.
4.2 Geometric Rectification of IRS -1C PAN and LISS- III Data
Pixels of raw / unrectified satellite image does not have appropriate geometry and uniform scale hence it cannot be used for mapping purpose. After rectification process is performed, each pixel in the image can be referenced not only by its row and column position but also in degree, feet or meter in a standard map projection system. There are several ways and techniques to convert raw image into map but where micro or sub micro level mapping has to be carried out using high-resolution images, it is advantageous to rectify satellite image using Ground Control Points (GCPs) from large-scale topomaps as base for rectification.
Digital map database has been used for the geometric rectification of IRS -1C PAN data. A second degree polynomial model is utilized to establish the transformation between the image and map leading to Root Mean Square Error (RMSE) of about 35 m (which is 15 -25 m with respect to GCP). Each GCP is taken on cultural features, because cultural features are unlikely to change in course of time. Two IRS -1C PAN scenes are used in this district which are geometrically rectified and then mosaiced for district image extraction and updation. These rectified PAN scenes are utilized as base for rectification of LISS - III data.
4.3 Fusion of IRS -1C PAN + LISS - III Data
The advancement in digital image processing has provided additional tools such as data fusion or data merging. IHS to RGB transformation method is used for the fusion of two different sensor data. This method delivers best color output by merging high-spatial resolution single band black & white data with low-resolution multispectral data. Here, in this study IRS -1C PAN (electromagnetic spectrum range 0.50 - 0.75) single band black & white data with 5.8 m spatial resolution and IRS -1C LISS - III multispectral data with 23.5 m resolution are used for fusion. Spatial resolution of both PAN and LISS - III data are rounded off to 6 m and 23 m respectively for calculation of scale factor. LISS - III sensor (electromagnetic spectrum range 0.52 - 1.70 micrometer) have 4 bands, i.e., band 1 - Green, 2 - Red, 3 - Near Infrared and band 4 - Short wave Infrared. As short wave infrared band with very coarse spatial resolution is not useful for mapping of infrastructure details hence band 1,2,3 are taken for fusing with PAN data. The fused/ merged output is in 6 m resolution with three RGB viewing image channels (Figure-1).
4.4 Digitization of Infrastructure/ Natural Feature Details
Infrastructure details which are required to be updated such as settlements, road network (unmetal & metal), railways, embankments/ dykes, etc., are digitized in ARC / INFO software environment using digital map database. Other than infrastructure natural features such as reserve forest boundaries, drainage, water bodies, river islands, etc., are also digitized to study the changes occurred in last 30 to 35 years. Separate layer is prepared for each feature with assigning proper attributes.
4.5 Updation of Infrastructure/ Natural Feature Details
All the layers digitized using digital map database are overlaid on fused image with 6 m resolution (FCC) and changes are incorporated in respective layers (Figure- 5). During the process of updation, it has been experienced that some features that are linear in nature can be updated using single band PAN data. But complexity arises when features are non-linear and there is possibility of mixing with other features. PAN data has various tones of black & white and some times it is very difficult to fix the boundaries between one object and the other. Whereas fused data offers more shades of color and enable the interpreter to separate out subtle differences in objects because of change in color tone. It may be noted that human eye can distinguish easily more shades of color than shades of gray tone in black & white. Therefore, it can be safely stated here that fused data is always more useful than black & white PAN data for mapping of such details if multispectral data of high-resolution is not available.
