Spatial Analyst: Multiple Criteria Evaluation Material adapted from FOR 4114 developed by Forestry Associate Professor Steve Prisley Section 1: Data In this exercise we will be working with several types of raster data for the Eggleston quad in Giles County, Virginia. We will be doing a raster analysis in this area to find locations meeting a combination of criteria including elevation, slope, aspect, and National Forest ownership. Your data include: NFOwn 37080C5 Egg30 Eggleston Shapefile of USFS ownership boundaries (from GW-Jeff National Forest) USGS DRG of Eggleston Quad, Virginia 30m DEM of Eggleston, Virginia 10m DEM of Eggleston, Virginia Section 2: Overview We are going to use raster data to find areas that meet a combination of criteria: 1. Within National Forest Ownership; 2. Elevations greater than 600 m 3. Slopes greater than 30% 4. Southern and southwestern aspects. To do this, we will create slope and aspect rasters and convert a vector layer to raster format. Prior to doing all this, we will become familiar with raster data and learn to enhance the display of topographic raster data. Along the way, we will examine the impact of spatial resolution (grid cell size) on our output products. Section 3: Examining properties of raster data First, we will examine the characteristics of two DEM datasets that cover the same area. All of the digital elevation models we are using were downloaded from the www.gisdatadepot.com web site. 1. Open ArcMap to a new map document and add the Eggleston and Egg30 DEM raster layers. Zoom to the full extent of the map. The Eggleston and Egg30 rasters cover the same geographic area at two different resolutions. 2. Make sure the Spatial Analyst Extension is turned on: from the Customize menu, select Extensions and make sure Spatial Analyst is checked. If the Spatial Analyst toolbar is not displayed, activate it by going to the Spatial Analyst: MCE Page 1
Tools menu, Customize option, and checking Spatial Analyst on the Toolbars tab. 3. Record the following information (Layer properties, Source Tab): Cellsize (X, Y) Pixel Type Pixel Depth (bits) Number of Columns Number of Rows Egg30 Eggleston 4. As in our work with images, raster file size is a function of data depth (how many bits or bytes per cell or pixel) and image size (numbers of rows and columns). If we estimate raster file size as: (Number of rows) x (Number of columns) x (Data depth in bytes) Then estimate raster file size for each of these (important: recall that there are 8 bits in a byte). Egg30 Eggleston Raster file size (bytes) Key Point: As with images, DEM s involve a tradeoff between spatial resolution (cell size) and file size. High resolution DEM s have smaller cells but many more of them. With smaller cells we may get a more precise sense of landform. But with higher file size comes greater storage requirements, longer processing times, etc. 5. There is another significant difference between the Egg30 and Eggleston DEM s. Using the layer properties again, complete the following: Minimum Elev. Value Maximum Elev. Value Egg30 Eggleston Spatial Analyst: MCE Page 2
6. What is going on here? Because these datasets cover the same geographic area, one might expect them to have approximately the same elevations. Using the Egg30.txt and Eggleston.txt metadata files, find and explain these elevation differences [Hint: elevation values are often referred to as Z values, as in a three-dimensional system of coordinate axes, where X is east-west, Y is north-south, and Z is elevation]. Key Point: Unfortunately, this is exactly the way this data came from USGS through GISDataDepot. There is no standard for elevation units for DEM s; they may come in either feet or meters. NEVER ASSUME that all data files will be consistent. What would be the implication of calculating slope (rise/run) if you ignored units and the run (X & Y) units were meters and the rise (Z) units were feet? Would the calculated slope be correct? 7. Before doing further analysis with these datasets, we want to make sure the elevation units for both of these are in meters. We can use Raster Calculator to convert a grid with values in feet to values in meters: a. In the Search Window, type in Raster Calculator b. Select Raster Calculator (Spatial Analyst) tool c. Double-click on the name of the layer that has elevations in feet; it will appear in the expression area in brackets. d. Complete the expression to divide by 3.2808 (feet per meter): layername /3.2808 e. Select your output location and name the file Egg30m (since it will now be in meters). f. Click OK. g. Examine the layer properties of this new layer and compare the data range to the numbers you recorded in step 5; record for the new layer (to the nearest 0.1 m): i. Minimum Elevation value: ii. Maximum Elevation value: 8. Important: remove the Egg30 DEM from your map; in all further work in this lab, use the Egg30m raster in place of Egg30. Key Point: Many raster GIS operations involve calculations based on the cell values. For example, dividing an elevation in feet by a constant conversion factor can produce elevations in meters. Other times, we will perform mathematical operations involving multiple raster datasets as if they were variables in an equation; this is what the raster calculator is all about! Spatial Analyst: MCE Page 3
Section 4. Vector to Raster Conversion We now want to use a vector layer of national forest ownership to create a raster file for our study area. We will use this later to determine what areas within the Eggleston quadrangle are owned by the US Forest Service. When we convert vector data to raster format, we must specify things like the extent (what area should be covered by the raster), the cell size, the coordinate system, etc. 9. Add the NFOwn.shp shapefile to your map. This data contains National Forest ownership boundaries for the George Washington-Jefferson National Forest (downloaded from http://www.fs.fed.us/gwjnf/gisdatadictionary.html). 10. Zoom in to the extent of the NFOwn layer; note that it goes well beyond the extent of our study area. Do we want to create a 10m raster that large? 11. Zoom back to the extent of Egg30m. 12. We need to specify some options for our vector-raster conversion: a. From the Geoprocessing menu, select "Environments". b. Specify a workspace: this will be the location where all temporary raster files will be created unless you dictate otherwise; your analyses will operate faster if this is on the computer's hard drive and not on an external drive! Make sure that the location you specify does not have a space in any folder name (such as Documents and Settings )! c. Specify that you want your analysis output in the same coordinate system as the data frame: Output Coordinates > Same as Input d. Within Processing Extent set the Extent to "Same as layer Egg30m" this limits the processing area to just the area of Egg30m, no data will be processed outside of the raster. e. Within Raster Analysis : i. Set your Cell Size to "Same as layer Egg30m" ii. Set your Mask to be the "Same as layer Egg30m". f. Click OK to close the options dialog. Now, any new raster file you create should match your Egg30m layer by default. 13. In the Search Window, type in Feature to Raster, select "Feature to Raster (Conversion)". a. The input features should be the NFOwn shapefile. b. The field to use for values in the output should be the "NFOWN_ID" code. (This is an ID number for each unique polygon in the ownership shapefile). c. Specify an output raster with name NFOwn30 (indicating 30m resolution). d. Output cell size should be set to Egg30m (you may have to Browse to the data set and select it). e. Click OK to perform the conversion. Spatial Analyst: MCE Page 4
14. Examine your results: a. Set transparency on your vector layer (NFOwn) to 60%, and zoom way in to some ownership edges to see how closely the raster and vector boundaries are. b. Open the attribute table for NFOwn30 and note the numbers in the Count column. These are counts of the numbers of grid cells for each attribute value (in this case, the attribute value is the ID number of the polygons; you specified that NFOWN_ID was to be used as the attribute value). Note that using the cell counts, you can determine area (each cell is 30m by 30m or 900 m 2, correct?). Record your counts and compute area: Attribute Value: Count Area (m 2 ) Area (ha) Total: c. If this raster matches the extent and cell size of the Egg30m raster, why is the total number of cells recorded above much less than the total number of cells you observed in steps 3 and 4? d. Use the Identify tool and click in your map to obtain values from the raster ownership layer. What happens when you click outside a forest boundary? What value is displayed? e. Go to the Symbology tab of the layer properties for the raster ownership layer. Note that you can change the colors displayed for the forest categories. Note also that you can set a color to be displayed where the layer has "No Data" (see the button next to "Display NoData as" in the lower right corner). Try it, setting a color for "NoData". 15. Now you see that grids (rasters) can contain areas without data; voids. These areas are treated differently in many calculations and operations; you should be aware of them!! 16. Now, repeat steps 12-14 using the Eggleston raster as a template (matching the Processing Extent and cell size of Eggleston instead of Egg30m). Call the output NFOwn10. We might expect forest area to vary slightly if we use different grid cells sizes; the boundary will not be the same as when we used 30m cells. Record the total National Forest cell count and area from this new 10m raster. (Remember that these cells are no longer 900m 2, right?) a. Total count: b. Area (hectares): Spatial Analyst: MCE Page 5
How well does the total hectares of National Forest from your 10m grid match the total from your 30m grid found in 14.b.? 17. You may remove the vector ownership layer (NFOwn), the NFOwn30 raster ownership layer created above, and turn off the display of (uncheck) the 10m raster ownership layer; you won't be needing it for a while. Section 5. Creating a shaded-relief map Now, we will create a shaded-relied map to better visualize the topography. 18. In the Search Window type in Hillshade, select Hillshade (Spatial Analyst). Create a Hillshade of both of the DEM s (Egg30m and Eggleston), using default values for azimuth, altitude, etc. Each time you create a hillshade, click on the Environments button and make sure your cell size (within Raster Analysis) and Processing Extent matches the input surface ; this field usually defaults to the last raster created. a. You should now have two hillshade rasters covering the same area. Zoom in to a scale of 1:20,000. By turning on and off the top layer, you should be able to go back and forth viewing the two versions. b. Which is more detailed? 19. Next, experiment with creating additional hillshade rasters of Eggleston, using different values for azimuth and altitude. Which one looks correct to you? Why? 20. Now we will experiment with color symbology for the elevation layer: a. Open the Symbology dialog for Eggleston (layer properties) b. Under Show, set to Classified. c. Right-click on the color ramp, and uncheck the Graphic View selection. This will now display names of the color ramps. d. Scroll down and select the color ramp named Elevation #1. This is a common color ramp showing lower elevations in blue, then greens, and higher elevations in orange, red, brown and finally white. e. Click the Classify button to specify a classification using 10 equal interval classes. f. Click OK to close the Symbology dialog. g. Now we have a graduated color map of elevations. h. Experiment also with the same color ramp applied to a Stretched rather than Classified symbology. Which provides more detail? 21. Note that Spatial Analyst provides a quick way to get a histogram of your data: a. In the Spatial Analyst toolbar, set the Layer to be Eggleston. b. At the right end of the toolbar, click the Histogram icon. c. A histogram of the values contained in the raster is displayed. You may need to make the histogram window bigger to see it all. Spatial Analyst: MCE Page 6
d. Now, in the symbology dialog, note that the default stretch type is Standard Deviations. Try different stretch types, each time creating a histogram of the values. Note that the histogram changes when you change the stretch type. 22. Now we can add a shaded-relief effect. i. Create a hillshade raster of the Eggleston layer, using default values for azimuth and altitude, and an output cell size of 10m. ii. Move this hillshade raster down in the Table of Contents until it is beneath your Eggleston layer. iii. Bring up the layer properties of your Eggleston layer. iv. Select the Display tab and set transparency to 60%. v. Now, you have the color elevation gradient from one layer and the shaded relief effect from another. Key Point: Shaded relief maps are a very effective means of visualizing topography. A shaded relief map can be created using an elevation layer, with graduated color symbology (using any color ramp you wish; it doesn t have to be Elevation #1 ), and a hillshade layer. 23. Finally, we will overlay detail on the map from the DRG s of this area. a. Add the Eggleston DRG to your map (37080c5.tif). i. The "color table" (list of values and associated colors) in this type of image can be manipulated in ArcGIS to change the colors displayed. b. We can make use of this capability to enable DRG's to be overlaid onto other images, lending annotation and context as well as being able to see the underlying shaded relief map: i. Turn on one of the DOQQ images and zoom to its extent. ii. Make sure the DRG image is higher in the table of contents (displayed over top of the DOQQ). iii. Open the Symbology tab of the layer properties dialog for the DRG layer, it should be on Colormap under Show. iv. Now, double click the color symbol associated with each of the following values and set it to "No Color": a. 1 (white) b. 5 (green) c. 9 (pink) d. 10 (lavender) e. 11 (gray) f. 12 (light brown) g. 255 (white) c. This DRG overlay is not very helpful at small scales. Note that at small scales (numbers like 1:100,000), all detail from the DRG is lost. At extremely large scales (like 1:1,000) the individual image pixels from the DRG become more distracting than helpful. d. Determine, in your own judgment, for what range of scales the DRG overlay seems helpful: Spatial Analyst: MCE Page 7
i. Largest scale: 1:. ii. Smallest scale: 1:. e. Now, set the display range for the DRG so that it will only display when the map scale is within this range: i. From the layer properties for the DRG, go to the General tab. ii. Click the radio button for Don t show layer when zoomed: iii. Specify the minimum and maximum scales you selected above. Key Point: Note the checkbox in the Table of Contents. It has changed to show that the display of this layer is now dependent on the map scale. 24. Use the Go To XY tool to zoom in to the area around the Cascades (set units to Meters; UTM Coordinates X: 537550, Y: 4135000). Zoom to a scale such that the Cascades location is visible, and the DRG is displayed over your shaded-relief map. Section 6. Creating slope and aspect rasters and reclassifying data Now, we will use Spatial Analyst to derive rasters of terrain slope gradient and slope direction (aspect). These variables are used in a wide variety of analyses in areas with topographic relief. For example, in the northern hemisphere, southern and western aspects receive sunlight later in the day and are consequently much warmer and dryer; this often results in different vegetation and physical conditions. 25. In the Search Window type Slope, select Slope (Spatial Analyst). a. We want to create a percent_rise slope raster based on the eggleston dem (make sure your environment settings match your inputs!) b. Name the raster Slope. 26. Now search for Aspect in the Search Window, select Aspect (Spatial Analyst). a. We want to create an aspect raster from the eggleston dem. b. Name the raster Aspect. 27. We need to reclass our rasters of slope class and aspect class for further analysis. We will use the class definitions in the box below. (When you reclassify, you will assign class numbers. The meanings are simply how we will refer to these classes by a name). Spatial Analyst: MCE Page 8
Slope Classes: Aspect Classes: Slope Value (%) Class Number Class Meaning 337.5 N 22.5 0 15 1 Low 15 30 2 Moderate 292.5 W Xeric 247.5 67.5 Mesic E 112.5 30 45 3 Steep 45 + 4 Very Steep 202.5 S 157.5 Aspect Classes: Aspect value ( ) Class Number Class Meaning -1-0 1 Flat 0 112.5 2 Mesic 112.5-157.5 3 Intermediate 157.5-292.5 4 Xeric 292.5-337.5 3 Intermediate 337.5-360 2 Mesic 28. To reclassify your Slope and Aspect rasters, type Reclassify into the Search Window and select Reclassify (Spatial Analyst) a. Your input will either be your slope or aspect raster b. The reclass field should automatically be set to the Value field c. Click on the Classify button, change the Method to Equal Interval, set the Class number to match either the Slope or Aspect Classes listed in Step 27 Spatial Analyst: MCE Page 9
d. Manual type in the Break Values on the right, each break value is the highest value for a class (e.g. the break value for Aspect class 1 is 0) e. Click OK f. Change the New Values in the Reclassification to match Step 27 values for Class Number g. Name your output rasters either AspClass10 or SlpClass10 29. When you are done, complete the following table; enter the count of cells in each class for each raster. Then, multiply the count times the cell size to get area in each class; record area in hectares to the nearest tenth (1 m 2 = 0.0001 ha). Slope Aspect Class Count Hectares Class Count Hectares Low Flat Moderate Mesic Steep Intermediate Very Steep Xeric Section 7. Putting it together: raster selection (query) Now, we have developed raster datasets that correspond to the criteria we are interested in: elevation, slope, aspect, and Forest Service ownership. We can combine these layers to find areas of interest. For the sake of example, we'll assume we're looking for likely areas in which to find the mythical endangered fritillated lousewort (Pedicularis fictionalis). It grows on dry, steep slopes at high elevations. We want to know how many hectares of this type of land are within our study area, and how much of that is on national forest land. 30. Locate the Raster Calculator by searching for it in the Search Window. It is best to use the buttons provided on Raster Calculator versus typing in an expression as it is very particular about syntax. Use the raster calculator to locate areas within the Eggleston quadrangle that meet the following criteria: a. Elevation must be at least (>=) 600m. b. Slope class is Steep or Very Steep. c. Aspect class is Xeric. Spatial Analyst: MCE Page 10
Caution: Review your Spatial Analyst options- we want 10m cells! An example expression for Raster Calculator to locate all elevation over 600m is: eggleston >= 600 *Remember to use the raster values not their meanings in your expressions in raster calculator. 31. The result should be a raster containing just 0 and 1 values; 0 means the cell did not meet the criteria, 1 means it did. The most effective display of this type or result may be to set the 0 values to "No Color" and set the 1 values to a bright, contrasting color and display it over your shaded relief map with no other layers displayed. Then, the bright colored areas represent likely locations for sites favorable for the fritillated lousewort. 32. Now, turn on the Forest Service ownership raster in the map so that you can visually locate areas outside of forest service ownership that might be favorable to the lousewort. If large areas favorable to the lousewort are located within Forest Service ownership, the agency can manage the land to promote the species. If large areas exist outside federal ownership, it might be useful to pursue conservation easements or buy land for protection of the lousewort. To determine areas within national forests favorable to lousewort, you will need to perform another selection in raster calculator, using the previous selection (calculation) and the national forest ownership raster. Complete the following table summarizing areas (report areas in hectares to the nearest hectare): Within NF Outside NF Total: Favorable Unfavorable Total: Hint: You may encounter difficulty summarizing areas based on the national forest ownership, due to the large areas in that raster with NoData. You can reclassify a raster such that NoData cells are assigned a numeric value. Key Point: The Raster Calculator is an immensely powerful tool. You can use it to query a single raster layer, multiple raster layers, to perform arithmetic expressions on combinations of rasters, etc. You should get familiar with the Raster Calculator functionality if you anticipate doing much raster analysis in ArcGIS. Spatial Analyst: MCE Page 11
Example Shaded relief map with DRG overlay