Geo 327G Semester Project Landslide Suitability Assessment of Olympic National Park, WA Fall 2011 Shane Lewis 1
I. Problem Landslides cause millions of dollars of damage nationally every year, and are very prevalent in Washington. There are several causes for landslides, but they are often triggered by seismic activity. Nearby earthquakes or volcanic eruptions can cause unstable soils to shake, thus reducing pore space and increasing pressure. The material then acts as a fluid and consequently gives way, causing a landslide. Geologic hazards such as these have become a very real concern for residents and those looking to build in the state. Liquefaction susceptibility is now an important part of surveys in order to produce an assessment that will help to locate these hazards. The use of GIS software to process data can be very helpful in creating an accurate assessment for areas of concern. This project will focus on the areas of Olympic National Park that appear to be most prone to landslides based on slope, geology, and soil liquefaction. This will involve the collection of DEM raster data, vector data such as geologic contacts, units, faulting, and liquefaction susceptibility. 2
I. Data Collection The data used in this project came from a few different sources for GIS data. The digital elevation model (DEM) for the project, which covers most of the Olympic peninsula and is the 1 arc second NED shaded relief, was collected using tools from The National Map Seamless Server at http://seamless.usgs.gov/website/seamless/viewer.htm. Several datasets including geology, and previously landslide information was gathered from the Washington State Department of Natural Resources site at http://www.dnr.wa.gov/researchscience/topics/geosciencesdata/pages/gis_data.aspx. These files came as their own geodatabase sets containing related information, for example, the surface_geology geodatabase contains the contacts, dikes, faults, folds, and unit polygon feature classes. One of the other datasets containing several different formats of features was 3
taken from the Integrated Resource Management Applications (IRMA) Portal at https://irma.nps.gov/. Many of the datasets collected from these sources came with metadata that include definitions, ages, and spatial data. One example of this can be seen in Figure 1 below. Figure 1: Metadata information for part of the DEM II. Data Preprocessing All of the GIS data collected from these online sources were saved into compressed (zip) files. Before I could work with any of them, I had to extract the files from the compressed folder and save them in the formats that would be readable in ArcGIS (shapefiles, geodatabase feature classes). Since they were still saved in readable formats, I did not have to do any conversions to view most of them. Some of the data, for example, the files that came from the IRMA Portal 4
were not spatially defined, so I had to make a guess about what coordinate system to put them in so that the data would still fit correctly with the other layers in the data frame. A couple of the shapefiles seemed to work best in NAD83 UTM Zone 10. III. ArcGIS Processing After all of the data was converted into readable formats and defined properly, I began processing by adding the four DEM rasters that I obtained from the Seamless Server. Because they are technically four different rasters at this point, they assign elevation values to each cell differently based on the elevations present in each one. This is visibly noticeable and the seams between them are quite obvious as shown in Figure 2 below. Figure 2: The raster elevation values do not match up with each other 5
To fix this problem, I had to combine the rasters into a single one, so that the values of the cells are on the same scale. First, I had to navigate to the mosaic tool by going through ArcToolbox > Data Management Tools > Raster > Raster Dataset > Mosaic To New Raster. Once the tool opened up, I selected all four existing rasters for the input rasters, and my data folder as the output location. The pixel type was 32_BIT_FLOAT because that was the type of the original rasters. The number of bands was set to 1, and the new raster was created, showing no seams. With the new raster in the table of contents, I was able to delete the original rasters and start adding the other data. From the surface_geology_250k geodatabase, I added the geologic_unit_poly_250k shapefile. At first, it appeared as a single color for all units, so to fix this problem; I went into the layer properties symbology tab and displayed the categories by unique values. Once in this menu, I set the value field to GEOLOGIC_UNIT_LABEL and clicked on the Add All Values button to get a display like the Figure 3. Figure 3: Categorizing the layer values by the geologic unit label given to each unit 6
Now we have a map showing the digital elevation model, and a matching layer of geologic unit polygons. The next step was to add a polygon of Olympic National Park in order to define an area of interest. The file I added was the park_polygon shapefile from my olym_wqgis folder downloaded from the IRMA Portal. Now the map contains the park boundaries, the geologic units, and a visible DEM raster underneath after setting the transparency of the unit layer to 40%. The resulting map display is shown in Figure 4 below. Figure 4: Map showing the boundaries of Olympic National Park, geologic units, and the digital elevation model beneath Since we are only interested in the area within the park boundaries, we will want to clip the geologic units layer to the park_polygon layer. Since we are trying to clip vector data rather 7
than a raster, we will use a clipping tool. To do this, I navigated to the Clip tool in ArcToolbox by going through Analysis Tools > Extract > Clip. Once the menu box opens for the tool, I selected the geologic units layer as the input feature and the park_polygon layer as the clip feature. The resulting output was my geounit_clip layer. After setting the symbology to show the values from the lithology field, the map now displays the areas of the park with different types of lithology, rather than individual units. This makes it easier to distinguish areas that are more susceptible to landslides. Figure 5: Areas of different lithology within Olympic National Park It is possible to see the areas of different elevation with the DEM and the corresponding lithology; however it is much easier to analyze slopes with the addition of a hillshade. To do this, I went through ArcToolbox to Spatial Analyst Tools > Surface > Hillshade to bring up the menu box for the tool. From there all I had to do was to set the input to the DEM raster 8
and name the output file and location to generate a hillshade image that acts as a shaded relief raster. It can then be placed underneath the DEM raster already present, and will be most useful when the DEM transparency is set to 40%. The result is displayed in Figure 6. Figure 6: Map generated after the addition of a hillshade raster Now that the hillshade layer is in place it is easier to visualize the conditions where different lithologies are likely to be found. For example, the dark areas on the map are valleys or depressed areas, some of them containing rivers and streams. From the lithology display that is clipped to the park boundary, we can see that the lithology commonly found in these valleys is unconsolidated sediments (dark pink). It can be assumed that areas with unconsolidated 9
sediments will be more prone to landslides that areas with harder rock due to the lack of stability, but the slope angle must be taken into consideration as well. I was able to find data on ground response at the Washington State Department of Natural Resources site, and added that to the map next. The liquefaction_susceptibility shapefile gives values based on the degree of susceptibility. It ranks areas from very low to high susceptibility and also displays those areas that are not susceptible (bedrock, ice, peat, water). As expected, the areas I identified as unconsolidated sediments earlier are now appearing as areas of moderate to high susceptibility (Orange below). Figure 7: Liquefaction Susceptibility Now that I have several pieces of data to use, I can start to form my own suitability raster. First, I needed to make a raster based on slope values alone. I did this by going through Spatial 10
Analyst Tools > Surface > Slope in ArcToolbox. For the input raster, I selected the DEM, and the output raster was named slope. I selected the output measurement as degrees because the slope measurements are degress. The conversion factor is 0.000009. The resulting slope raster shows several different divisions of values, but the objective is to create a raster that has suitability rankings of 1 to 5. To fix this, I went into the property settings of the new slope raster, and under Classification I changed the number of classes to 5. Next I clicked the Classify button, and changed the new break values to 10, 20, 40, 60, and 90 degrees. This allowed me to separate the slope values into varying degrees of landslide susceptibility. Figure 8: Classifying the break values so that there are five divisions 11
Next, I needed to make each of the degree increments correspond to a rank of 1 5. I used the reclassify tool by going through Spatial Analyst Tools > Reclass > Reclassify in ArcToolbox. This tool allowed me to assign new values to represent the ranges of degrees I classified before. Figure 9: Reclassification of degree values to represent rank Figure 10: Reclassified slope raster with green representing the shallowest slopes and red showing the steepest 12
Now the slope raster portion of the suitability assessment is complete. Next, I had to make a raster of the geology to be able to add to the slope raster. The lithology in the geounit_clip layer was not ranked for suitability, so I went into the attribute table and added a new field named rank. I then assigned each type of lithology a number based on how landslide prone it is. The ranking system I used is as follows: 1 Glaciers and Snowfields, Water 2 Intrusive Rocks, Volcanic Deposits and Rocks 3 Mixed Volcanic and Sedimentary Rocks 4 Sedimentary Deposits and Rocks 5 Unconsolidated Sediments At this point, the geology is in the form of vector data. To convert the geology polygons into a raster, I went through Conversion Tools > To Raster > Polygon to Raster in ArcToolbox. I then filled in the box to make it look like the one in Figure 11. Note that the cell size must be the same as the slope raster. This will make it possible to add the rasters together and make the cells match up when I try to form the suitability raster. Once the tool has finished, the geology is displayed in the five ranks that I assigned in the previous step. A picture of the result is shown in Figure 12 on the next page. It is important that the rasters and the park boundary polygon are in the same coordinate system before we add everything together. To convert the slope raster from GCS NAD83 into UTM Zone 10, I used the Project Raster tool found in the Projections and Transformations section of ArcToolbox. I then set the output coordinate system to UTM Zone 10, the resampling technique to bilinear, and the output cell size to 30 13
meters. Now the slope raster is in UTM coordinates and I just need to clip it to the park boundary. Figure 11: Polygon to Raster Tool Figure 12: Map of lithology ranked from high susceptibility (red-orange) to low susceptibility (dark blue) 14
To clip the new slope raster to the park boundary, I used the Extract by Mask tool in the Spatial Analyst Tools section, and set the input raster to the slopeutm raster while the feature mask was the park_polygon. Now both the slope and the geology rasters are clipped to the park boundary, and I need to convert the geology raster to UTM coordinates. I followed the same procedure as the slopeutm conversion to do this. In order to make the assessment as accurate as possible, I made a third raster with rankings from the liquefaction_susceptibility layer. I started by clipping the liquefaction polygons to the park boundary using the clip tool again. I then used the same procedure to add an attribute field and assign ranks. The ranking system used for liquefaction is as follows: 1 Bedrock, Ice, Water 2 Very Low, Very Low to Low 3 Low 4 Low to Moderate 5 Moderate to High Once the ranks had been assigned I proceeded to convert the polygon to a raster using the same tool used to convert the geology layer. With the new liquefaction raster in a different coordinate system as the other rasters, it was again necessary to use the Project Raster tool to convert it to UTM Zone 10. An image of the Project Raster menu is shown in Figure 13. Now that all of the rasters have the same coordinate systems and are ranked 1-5 for landslide susceptibility, they are ready to add together. To add them, I went through Spatial Analysis Tools > Map Algebra > Raster Calculator and made the map algebra expression shown in the Figure 14 on the next page. 15
Figure 13: Project Raster Tool Figure 14: Raster Calculator tool used to perform map algebra 16
The Raster Calculator tool took the rank values from each of the three rasters and added them together to get a new value. For example, a cell with a value of 3 in the liquutm raster, a value of 4 in the geoutm raster, and a value of 3 in the slopeutm raster, would now have the value of 10 in the new landslide susceptibility raster. The new raster ranking system goes from a minimum of 3 to a maximum of 15. The result of the raster calculator after changing the coordinate system of the whole data frame to UTM Zone 10 is shown in Figures 15 and 16. Figure 15: Landslide susceptibility based on slope, geology, and liquefaction data. Red represents highest susceptibility and blue represents lowest 17
IV. Conclusion With the final landslide suitability raster displayed, I can start to analyze the data and draw conclusions about the areas which appear to have the most concern. As we can see from the map, areas that display in blue colors are those that are least prone. The most obvious of these are bodies of water, ice, and snowfields which show in the darkest shades of blue. The least prone cells have the lowest rank in each set of conditions (geology, liquefaction, < 10 slope). At the other extreme, we can see some areas where a darker orange and even red color is displayed, which represent the conditions that landslides are most likely to occur. The areas that appear in red have the set of conditions in which they are unconsolidated sediments, > 60 slope, and have a moderate to high liquefaction rating. Figure 16: Blue represents lowest landslide susceptibility, red represents highest 18
Most of the area within the park falls within an intermediate value on the scale because a lot of it is bedrock, but is exposed on slopes at higher angles. This accounts for much of the light green and yellow colored areas on the map. Another factor we can look at in deciding areas of concern, are previous landslides and other geologic hazards that have been recorded in the area already. The figure below shows mapped landslides and earthquakes above magnitude 3 that have occurred in the past. The gray spots within the park are small landslide polygons, and the yellow dots represent earthquakes. Figure 17: Past Landslides and Earthquakes Other factors other than seismic triggers are erosion, rainfall and human activity. In the case of rainfall, the water gets into the soil and rock which drives up the fluid pressure and makes it less stable. These may also have an impact on the susceptibility of the park. 19
Landslide Susceptibility Assessment of Olympic National Park, Washington Suitability Assessment Based On Slope, Geology, and Liquefaction ¹ Legend Park Boundary Landslide Susceptibility 3 Very Low 4 5 6 7 8 9 10 11 12 13 14 Very High GCS North American 1983 NAD 1983 UTM Zone 10N 1:600,000 0 10 20 30 40 Kilometers