Introduction to Geographic Information Systems in Forest Resources

UW Home GIS@UW Search

Syllabus Schedule Class Meetings Assignments Course Data Internet Search Current Grades Contact Us CFR 590 Internet-only section Lab Locations

1. Open ArcView 3.x and start a new project
2. Add remotely sensed image themes to the view
3. Alter the display of a LandSAT TM image theme
4. Convert a LandSAT TM image to a series of grid themes
5. Create an NDVI grid
6. Create a Tasseled Cap greenness grid
7. Create a Tasseled Cap wetness grid
8. Select a training set for image classification
9. Select pixels matching the training set

1. Start ArcView 3.x.
   
   
2. Remember to set your working directory.
   
   
3. Enable the Spatial Analyst and IMAGINE Image Support Extensions.
   
   
4. Create a view called Pack Forest RS.
   
   
5. Add the Stands theme to the view.
   
   
6. Alter the fill pattern of the Stands theme so that it has no fill symbol, and red outlines.

1. Add the data source ortho_91.bil from the CD:\packgis\forest directory.
   This data source is a digital orthophotograph.
   
   
2. If you do not see the data source listed, make sure that the Data Source Types pulldown is set to Image Data Source.
   
   
3. Zoom in so that you can see the relationship between the stands vector theme and the Ortho_91.bil image theme.
   
   
4. Turn off the Ortho_91.bil theme.
   
   
5. Add the image data source pf_lan.img (also in CD:\packgis\forest) to the view. This is a subset of a 7-band LandSAT TM image from 1991.
   
   
6. When the theme is loaded, it will be displayed automatically with the bands 1, 2, and 3 displayed in red, green, and blue, respectively. This color combination is known as false color, in which features appear close to their actual color.
   
   

1. Make the Pf_lan.img theme active, and open the legend editor.
   
   
2. Alter the color mapping of the bands, so that the order of the band display is 4, 3, 2.
   
   
   
   
   Vegetation will appear in shades of red and pink.
   
   
3. Make the Stands theme active.
   
   
   
   
4. Using the Identify tool, look at the age of some of these stands. Remember that the image theme dates from 1991, whereas the Stands theme was most recently updated sometime in 1999. Do you notice any inconsistencies due to dates?
   
5. Alter the image legend to display bands 7, 4, and 3 as red, green, and blue. What patterns do you see here?
   
   
   
   
6. Feel free to experiment with different color combinations. Image processing experts can use different combinations of bands to identify properties of landscape features.
   
   
7. Alter the image legend to display only a single band. Experiment by looking at the spectral reflectances for each different band. Make sure to click the Default button after you change bands.
   
   
   
   
   
   
8. Revert back to Multi-Band display and click the Linear button. This opens the Linear Lookup dialog.
   
   
   
   
   Try altering the slope and position of the lines for each band, one at a time. Notice what happens when you alter the line and then click the Apply button. For more details on how to control the linear stretch for bands, press the <F1> key for help on the Linear Lookup dialog.

You have just altered the display of a LandSAT image. Altering the display is one of the best methods for picking out individual features.

1. Make the pf_lan.img theme active.
   
   
2. From the Theme menu, select Convert to Grid.
   
   
3. When asked, convert each band to a new grid.
   
   
   
   
4. Convert the first band to a grid called Lan1.
   
   
   
   
5. Continue converting, naming each new grid with the same naming convention (Lan2, Lan3, etc.).
   
6. When asked, add each grid theme to the view.
   
7. View each band in turn.

You have just created a new grid from each band of this 7-band LandSAT TM image. Each grid has cell values that represent reflectance intensity in specific wavelength ranges. These individual grids can be used with map calculations to perform multispectral analysis.

1. Invoke the Map Calculator (from the Analysis menu).
   
2. Format a calculation with the NDVI function. Feel free to cut-and-paste from the browser:
   
   ([Lan4] - [Lan3]).Float / ([Lan4] + [Lan3]).Float
   
   
   
   
3. When the calculation completes,
   1. Classify the new grid into 30 equal-interval classes.
   2. Use the Browns to Blue-greens dichromatic Color Ramp.
   3. Sort in the Value field in Descending order. Vigorous vegetation will appear in the darkest browns.
      
      

You have just performed the calculation of NDVI on a series of grids representing different bands from a LandSAT image. This results in a single grid whose value represents the NDVI value for each pixel.

Create a Tasseled Cap greenness grid

1. Open the Map Calculator and enter the Tasseled Cap greenness function:
   ( [Lan1] * -0.2848 + [Lan2] * -0.2435 + [Lan3] * -0.5436 + [Lan4] * 0.7243
   + [Lan5] * 0.0840 + [Lan7] * -0.1800)
   
   I suggest using cut-and-paste from the browser to your Map Calculation dialog. You can select Edit > Copy from the browsers's menu. Then place the cursor in the Map Calculation expression area and use the <CTRL-V> keystroke combination to paste.
   
   
   
   
2. When the new grid is added to the view, alter the legend to use the Chartreuse monochromatic Color Ramp.
   
   Now vegetation that is "greener" will actually appear a darker green in the view. You may also want to load the Boundary theme for reference.
   
   

You have just created a single grid that represents Tasseled Cap greenness of the original LandSAT image.

1. Using the Map Calculator, create a calculation for the Tasseled Cap wetness index:
   ([Lan1] * 0.1509 + [Lan2] * 0.1973 + [Lan3] * 0.3279 + [Lan4] * 0.3406 + [Lan5] * -0.7112 + [Lan7] * -0.4572)
   
   
   
   
2. When the new grid is added, change its legend to use the Cyan Monochromatic Color Ramp.
   
   
   
   Now pixels that with vegetation that is more "wet" are shaded a deeper cyan.

You have just created a single grid that represents Tasseled Cap wetness of the original LandSAT image.

Although the optional (at a cost) Image Analyst is an extension that provides many image analysis operations, ArcView 3.x 's basic functionality is really not made to do these sorts of things, so this will be a kludge at best.

1. Turn off all themes except the pf_lan.img image.
   
   
2. Make all the LandSAT grid themes in your view active.
   
   
3. To save Table of Contents real estate, select Theme > Hide Legend from the menu.
   
   
4. Add the arc feature type from the stands data source. Change the symbol so you can see stand boundaries.
   
   
   
   
5. Zoom in to an area that appears homogeneous in color.
   
   
   
   
6. Create a new polygon theme, and outline a polygon delineating a group of cells. These cells will be the training set.
   
   
   
   
7. From the menu, select Analysis > Summarize Zones. Use ID as the field that defines zones.
   
   
   
   
8. Choose Lan7 as the grid to summarize. Dismiss the dialog that asks which field to chart.
   This will result in a table containing one record, representing statistics for the Lan7 grid for the cells within the outlined polygon. These cell value statistics will be used in querying the image.
   
   
   
   
9. Repeat the last step for each Lan grid.

You have just gathered statistics for each of 7 bands for a particular polygonal area on the forest.

1. This may take some time, but create a Map Query that contains ranges of values for cells for each Lan grid, e.g.,
   
   ([Lan7] > 20) and ([Lan7] < 28) and
   ([Lan5] > 74) and ([Lan5] < 86) and
   ([Lan4] > 88) and ([Lan4] < 110)
   
   Here I have chosen a set of cells which are within an arbitrary range of the mean cell value for the grids Lan7, Lan5, and Lan4 within the training set.
   
   The results of the map query:
   
   
   
   This shows the series of pixels that have similar reflectance values to the training set in bands 4, 5, and 7.
   
   
2. Add the stands polygon theme and find out the polygon attributes for any stands that have a large proportion of pixels matching the training set. The stand within the training set is in age class 10-20. The other stands with large proportions of matching pixels are all young stands as well.
   
   You should beware that we are not selecting pixels containing young forest cover per se, but we are only selecting pixels whose reflectances match for a given set of bands and a given reflectance tolerance. It is essential, after performing these kinds of analysis, to ground truth the similar pixels.
   
   If we could verify that all pixels selected using this technique are within young stands at Pack Forest, then we could say with a certain amount of confidence that other pixels on the image are also located in young stands. What we cannot say is that this method picks up all young stands; it is obvious that there are pixels within young stands that were not selected according to these criteria. Better software and better methods are needed to deal with more sophisticated classifications.
   
   
   This is a very simple and unsophisticated analysis. But it does, in essence, show how supervised classifications are performed on multispectral imagery. If you are using image processing software, your analyses will be more sophisticated, but will essentially use the same process.
   
   1. Select pixels from locations with known land cover.
   2. Create a multispectral statistical "profile" of this area.
   3. Search for all pixels matching this profile.

Syllabus Schedule Class Meetings Assignments Course Data Internet Search Current Grades Contact Us CFR 590 Internet-only section Lab Locations

The University of Washington Spatial Technology, GIS, and Remote Sensing Page is provided by the College of Forest Resources and the College of Ocean and Fisheries Sciences through Unit-Specific UIF. Site administrator: Phil Hurvitz.

Copyright © Phil Hurvitz, 1998-2003