GIS CASE STUDY Application of GIS for earthquake hazard and risk assessment: Kathmandu, Nepal Part 2: Data preparation Cees van Westen (E-mail : westen@itc.nl) Siefko Slob (E-mail: Slob@itc.nl) Lorena Montoya de Horn (E-mail: montoya@itc.nl) Luc Boerboom (E-mail: boerboom@itc.nl) International Institute for Geo-Information Science and Earth Observation, ITC P.O. Box 6, 7500 AA Enschede, The Netherlands Although this exercise doesn t deal directly with specific hazard or risk related topics, it is a good introduction to the dataset, and to some of the functionalities of GIS (ILWIS in this case) that are useful in disaster management. We will concentrate on the following aspects: - Generation of an orthophoto for the city of Kathmandu - Generation of a steropair of airphotos or satellite images for visual 3-D digital interpretation - Fusion of a multi-spectral image and a higher resolution panchromatic image for better interpretation. Disclaimer The material in this exercise is for training purposes only. ATC-13 vulnerability functions are used. These functions are based on historical damage data and should not be applied without verification or calibration. The results should not be used in actual planning of the city of Kathmandu as ITC does not guarantee the accuracy and precision of the input data and adequacy of ATC-13 data in relation to the Nepali building stock. The GIS software that will be used in this exercise is the Integrated Land and Water Information System (ILWIS), version 3.12, which contains useful tools for digital stereo pair generation. 2-1
2.1 Georeferencing base data and preparation of digital stereopair This exercise will first deal with the application of a number of image processing tools for georeferencing the topomap, the aerial photographs and for developing a digital stereopair, which you will later interpret directly on the screen. Time requirement: 4 hours Objectives: Georeference the 1:25.000 scale topographic map. Ortho-georeference the central airphoto of Kathmandu Create an epipolar image from two digital airphotos Digitally interpret the stereo image Input data: Topomap of Kathmandu: Topomap Kathmandu (raster map) Digital road map: Road type (segment map) Aerial photographs of Kathmandu: 36_23 (left), 36_22 (middle) and 36_21 (right) Detailed aerial photographs of the northern part of Kathmandu: 8902_51, 8902_52 and 8902_53 (raster maps) 2.1.1 Georeferencing the topographic map Time requirement: 1 hour In this first exercise you will georeference the scanned topomap. Open the raster map Topomap Kathmandu. In the File menu, select Create Georeference. Name the georeference Topomap and make sure Tiepoints is selected. Next click the OK button. Once the tiepoint editor opens, go to the main window and open the Roadtype vector map (note that this map is in another folder). Resize the maps so that you can place them side-by-side. Search for a road junction visible both in the topomap and in the roadtype map. Zoom into this point in both maps. Select the arrow icon and click the first the point in the topomap. Then go to the roadtype map, select the arrow and click on the corresponding point. Repeat this procedure at least 10 times. Make sure that you select points in different parts of the topomap; in other words, avoid points which are clustered together. On the topomap, go the Add layer from the Layers menu and select roadtype. Zoom in to observe the results of your georeferencing procedure. If the fit is good and the error is within acceptable limits, close the editor (special button). If you are not satisfied, entered more tiepoints. When you are finished, select Exit Editor from the File menu. 2-2
2.1.2 Generating Georeference Orthophoto Case study Kathmandu: GIS for earthquake loss estimation Time required: 1 hour When creating a georeferenced orthophoto (differential rectification), you first have to reference the fiducial marks which appear on the scanned aerial photograph and you have to specify the principal distance of the photographic camera used. In photogrammetry, this process is known as the inner orientation. Open the raster map 36_22. In the File menu, go to Create Georeference. Name the georeference 36_22 and select GeoRef Ortho Photo. Select the raster map Elevation as the DTM (note that this map is in another folder). The Locate Fiducial Marks dialog box is displayed. In this dialog box you should indicate the fiducial marks of the photograph. The principal distance is the distance between the photography plane (the negative) and the projection center of the lens. This principal distance is generally slightly greater than the focal length, and should appear in the calibration report of your photograph. The principal distance is generally a value between 90 mm and 300 mm; many aerial photographs have a principal distance in the order of 152 mm or 153 mm. An aerial photograph usually has 4 fiducial marks in the corners of the photo and/or 4 fiducial marks in the middle along the edges of the photo. However, due to printing and/or digital scanning, it is possible not all fiducial marks may be visible. You must make a choice to either use the fiducial marks in the corners or the ones in the middle along the edges of the photo. It is best if you can locate all 4 fiducial marks. In case your scanned photo contains only 3 fiducial marks, use those 3. In case you can see only 2 fiducial marks on your photo, these must be located on opposite sides of the photo. If you have the photograph available in paper form, measure the distance between the fiducial marks as precisely as possible. In the dialog box, you then have to fill in photo-coordinates in mm for the fiducial marks. In photogrammetry, it is customary that the center of the photo (principal point) is located at 0, 0. If the size of your aerial photograph is 23 x 23 cm and you are using corner fiducial marks, the photo-coordinates will be on the order of the following values: 2-3
-159.3-159.1 (Lower, Left) -159.1 159.3 (Upper, Left) 159.8 159.4 (Upper, Right) 159.5-159.6 (Lower, Right) You should make sure that the photo-coordinates entered match the fiducial marks. Zoom in on the fiducial mark (circle) in the upper left part of the photo On the Locate Fiducial Marks screen, tick fiducial mark 1. On the map, select the arrow and click on the fiducial mark. Enter -159.3 and 159.3 as photo mm. Next zoom in on the upper right fiducial mark. On the Locate Fiducial Marks screen, tick fiducial mark 2. On the map, select the arrow and click on the fiducial mark. Enter 159.3 and 159.3 as photo mm. Zoom in on the lower left fiducial mark. Tick fiducial mark 3. With the arrow, click on the fiducial mark. Enter -159.3 and -159.3 as photo mm Zoom in on the lower right fiducial mark. Tick fiducial mark 4. With the arrow, click on the fiducial mark. Enter 159.3 and -159.3 as photo mm. Now you will see the scan resolution in dots per inch (dpi) and the sigma (error in mm/pixel). You will use 150 mm as the principal distance as we don t know the exact camera details. Click the OK button. Note: Our result will not be as precise as it could be as we lack have the camera calibration report and the information about the principal distance (also because to save time, we haven t actually measure the distances to the fiducial marks). Once all the fiducial marks are registered, the Tiepoint editor will be opened. In the Tiepoint editor, you can insert tiepoints, also known as ground control points, which establish the relationships between pixels in your digital photo (row,col) and real world XYZ-coordinates. If you know Z-coordinates for your control points, you can enter these, otherwise height values from the DTM are used. In photogrammetry, this process is known as the outer orientation. For best results you should use high precision GPS points measured in the field as control points. Since we don t have high precision GPS points, we will get use points from another GIS map. Open the road map Roadtype in a separate map window. Search for a road junction that you can see both in the aerial photograph and in the road map. Zoom into this point in both maps. On the aerial photograph 36_22, select the arrow icon and click on the point. Next go to the Roadtype, select the arrow icon and click on the corresponding point. Repeat this for at least 10 control points. On the aerial photograph, go the Add layer from the Layers menu and 2-4
select roadtype. Zoom in to observe the results of your georeferencing procedure. If the fit is good and the error is within acceptable limits, close the editor (special button). If you are not satisfied, entered more points. When you are finished, select Exit Editor from the File menu. The aerial photograph 36_22 is now ortho-georeferenced. It is not yet an orthophoto, because it still needs to be resampled (orthorectified) to the digital elevation model, or to the georeference of another map. We will not do that, as we will use the photograph with the georeference ortho in the next step: creating a stereopair. 2.1.3 Creating a digital stereopair. Time requirement: 1 hour (for one set, 1.45 hours if two sets are done) A stereo pair allows you to view raster maps, scanned photographs or images in stereo, using a stereoscope mounted onto your monitor or red-green or red-blue glasses (anaglyph). A stereo pair can be calculated: with the Epipolar Stereo Pair operation which requires two raster images with overlap as input, for instance two scanned aerial photographs with overlap. In the output stereo pair, you will be able to view the area of overlap in stereo or, with the Stereo Pair From DTM operation with a single raster map as input, for instance a scanned photograph or an image, and a Digital Terrain Model (DTM). In the output stereo pair, you will be able to view the whole area of the input map displayed over the DTM in stereo. A stereo pair is automatically calculated when it is opened for display. The stereo pair contains: two resampled output raster maps, where each raster map uses a new georeference which retains original coordinates. A stereo pair can be displayed: in a stereoscope window (a stereoscope is required) or as an anaglyph in a map window (red-green or red-blue glasses is required). If you carried out the exercise of the previous section, remove the georeference of the aerial photography 36_22. On the main window, select the 36_22 raster map and click the right button of the mouse. Select properties and select None as the Georeference. From the main menu select Operations, Image Processing, Epipolar StereoPair. Select: 36_23 as left image and 36_22 as right image. Name the output stereopair Kathmandu_new. Click the OK button. The following window is opened. 2-5
Now you need to fill in the fiducial points for both photos. Digitize the four fiducial marks (FM) in both photos. Note that you can switch from one photo to the other with the indicator in the button bar. If you have done this correctly and zoom out again, you can notice that both photographs now indicate the location of the principal point. The following step is to indicate the transferred principal point in the other photograph. You have to zoom in on the principal point of one photo and zoom in on the same area in the other photograph. Also enter two control points (scaling points) that are identical in both photos, one located more to the top of the photo and one located to the bottom. If everything is correct, select the button Exit and Show (color monitor button). Both images now have to be resampled; this procedure may take some time depending on your computer s characteristics. After the photos are displayed, close the window. Next you will visualize the resulting stereopair Kathmandu_new as an anaglyph image. On the main window, select the stereopair Kathmandu_new and click the right button of your mouse. Select Visualization, as Anaglyph. Next choose red-green or red-blue depending on the type of eyeglasses that you have. Click the OK button. You should now be able to see the image in stereo. Depending on time, you could also apply the same procedure for the more detailed photographs: 8902_52 (left) and 8902-51 (right). 2-6
2.2 Fusion of Ikonos pan and Quickbird MS satellite images You have been provided with satellite images of a portion of the city of Kathmandu in Nepal. The Quickbird false color composite has a resolution of 2.8 meters and the Ikonos panchromatic has a resolution of 1 meter. The Quickbird image was taken in 2002 while the Ikonos was taken late in 2001 which therefore makes not much difference from a temporal point of view. Obviously you would rather have the panchromatic band of the Quickbird as well but it is not available to you. To make the best out of your data, you can sharpen the Quickbird on the basis of the Ikonos panchromatic image. This procedure is involves the fusion of the two images. Time requirement: 40 minutes Objectives: To carry out a color separation of a false color composite To resample an image to the georeference of the other one To fuse both images in order to sharpen the multi-spectral image on the basis of a higher resolution panchromatic image. Input data: 2.2.1 Color separation Quickbird multispectral false color satellite image: Quickbird (raster map) Ikonos panchromatic satellite image: Ikonos_pan (raster map) In this first exercise you will carry out a color separation of the false colour composite that you have been provided with. You need to extract the hue and intensity from the color composite into different raster layers. Therefore you need some basic knowledge on the meaning of these terms. Colors can be defined using the Hue, Saturation, and Intensity (HIS) system. Hue is the basic component of a color and is the dominant or average wavelength. Saturation describes the color purity or the purity of color relative to Gray. The relative amount of Red, Green and Blue determines the Hue and the Saturation. Intensity describes the brightness of a color. The total amount of Red+Green+Blue determines the brightness. Time requirement: 10 minutes On the main window, select the Quickbird raster file. Make a click on the right button of your mouse and select Image Processing and Color Separation. Select Hue and name the output raster map Quickbird_Hue. Repeat this procedure, selecting Saturation and naming the output raster map Quickbird_Sat. 2-7
2.2.2 Resampling procedure Since the images have different pixel sizes, before you can fuse them you have to resample the Quickbird raster images (2.8 meter pixel size) to the pixel size of the Ikonos image (1 meter). Time requirement: 15 minutes On the main window, select the Quickbird_Hue raster file. Make a click on the right button of your mouse and select Image Processing and Resample. Select the georeference Ikonos and name the output raster map Quickbird_Hue_1m. Repeat this procedure, selecting Quickbird_Sat and naming the output raster map Quickbird_Sat_1m. 2.2.3 Colour composite The last step involves using the colour composite operation to combine the hue, saturation and intensity maps into a single image. The intensity is provided by the panchromatic Ikonos image. In ILWIS, colour composites can be created in various ways. For this exercise, we need to apply the method called 24 Bit HSI in which input values are interpreted as hue, saturation and intensity. The different methods of creating a colour composite are merely a matter of scaling the input values over the output colours. Time requirement: 15 minutes On the main window, choose Image Processing, Color Composite. Make sure you tick the HSI box and not the RGB box. Under Hue select the raster map Quickbird_Hue_1m. Under Saturation select the raster map Quickbird_Sat_1m. Under Intensity, select the raster map Ikonos_pan. Name the resulting raster map Quickbird_sharp. Make sure under Description you described what this image really is (e.g. Quickbird MS image sharpened by Ikonos panchromatic image). Open both the sharpened image and the unsharpened one. Re-size the windows so you can view both side by size. Observe the improvement achieved by the fusion of both images. 2-8