Raster Analysis and Suitability Modeling for Frac Sand Mining in Trempealeau County, West-Central Wisconsin
Introduction
Raster datasets provide an opportunity that is quite unique and can be a very powerful tool for spatial analysis. If properly trained, the functionality of rasters allow the user to reclassify, calculate distance buffers, add multiple raster datasets, and use the product to produce a suitability model. In this exercise, our objective was to develop a number of suitability models, based on a given set of criteria, and combine them in the end to develop a final model for most suitable land. A Python script was then written to assign a weighted value to what I deemed the most prominent variable. The result was a number of suitability maps that showed the most suitable location for frac sand mines in the southern half of Trempealeau County.
Methods
The initial suitability model examined bedrock geology, national land cover data (NLCD), distance from a rail terminal, slope percent, and water table elevation. Most of these were in the form of vector data (points, lines, and polygons). Most of the data was acquired from the Trempealeau County GIS geodatabase (tremplocounty.com), with the exception of the water table data, which was acquired from the Wisconsin Geological and Natural History GIS database, and the NCLD data, which was acquired from the US National map viewer database. The vector data for bedrock geology, NLCD, rail terminal, and surface elevation were clipped to our study area, projected, and converted from vector to raster. The next step was to assign a euclidean distance to the rail terminal. This produced the raster equivalence of a vector buffer, though where the buffer only extends as far as the user input extends, the raster euclidean distance tool extends to the processing extent of the model. This tool produced a new raster of distance from the terminal over the entire study area. The next step was to reclassify all the data and assign it a value of one, two, or three. The most suitable land was given a designation of "3," while the least suitable, or unsuitable land, was given a designation of 1 (Figure 1). In the case of what we will call exclusion rasters, raster used to completely exclude certain areas based upon unsuitable criteria, a value of zero or one was assigned. A value of 0 meant that the area was not useable, such as open water, flooded zones, or developed areas, while a value of 1 meant that the area was suitable for use. An exclusion raster was developed for the NLCD data, in this section, and for areas containing protected wildlife habitat in the next section.
 |
| Figure 1. This figure shows the description of the values used in the reclassified rasters. The (2) variable is missing from Bedrock Geology. I chose not to use it because the bedrock units are either suitable or not suitable, but they need to retain the same weight that the other rasters have. |
After the rasters were all reclassified, they were able to be combined. To do this, the Raster Calculator tool was used to create an equation. The rasters for bedrock geology, NLCD, distance from terminal, percent slope, and water table elevation were added and then multiplied by the NLCD (Excluded) raster. This returned results of either a zero, where the land was unsuitable, or the previous added number where the land was suitable. A data flow model was created in Modelbuilder (Figure 2).
 |
| Figure 2. This figure shows the data flow model created in Modelbuilder for analyzing and combining the bedrock geology, NLCD, distance from rail terminal, percent slope, and water table elevation data. The water table data had to be imported, then converted using the Topo to Raster tool before it could be reclassified. |
The second portion of the exercise required mostly the same workflow to be used to analyze streams, prime farmland, proximity to residential areas, proximity to schools, and proximity to wildlife areas. The data was clipped, projected, converted to raster, and then reclassified based on a set of determined criteria (Figures 3 and 4). In this dataset, streams, farmland, residential proximity, and school proximity were given a one, two, or three designation, while proximity to wildlife areas was assigned a zero or one value. Only a certain type of streams were selected, to determine the areas of the most prominent flow. Streams of higher stream order are generally larger streams in a watershed and therefore are more likely to be polluted by the mines. Using the Euclidean Distance tool, these streams were given a distance of 500 meters around them before they were given their reclassified designations. This was a number I decided upon to create a buffer zone between potential mines and streams. The data for residential and school proximity was based upon a distance provided to us of 640 meters that were required to be between a residential zone or school and a mine. A distance within 640 meters was least desirable, while a distance of double that, 1280 meters, was the next least desirable, while anything over 1280 meters was desirable. Wildlife Areas within the county was a variable that I wanted to examine. I wanted to exclude these areas from potential mine locations, as I considered them a valuable resource for promoting ecological health. I decided that any distance within 500 meters of a wildlife area was not healthy for the wildlife area, thus it was to be entirely excluded from the search.
 |
| Figure 3. This figure shows the initial portion of the data flow model. An iterator tool was used to automate workflow for all included variables. In this case, it was used to clip all of the feature classes to the study area. Another data flow model was created and another iterator was used to project all of the feature classes. No more than one iterator can be used in a data flow model. |
 |
| Figure 4. This figure shows the description of the values used in the reclassified rasters. The reclassified raster for wildlife areas was determined to be an exclusionary raster, as was the viewshed visibility raster. |
The next objective was to determine a viewshed for a certain feature in Trempealeau County deemed a prime recreational area. I chose to work with a Mississippi River recreational trail. The Viewshed tool takes an input feature, a digital elevation model, and determines any areas visible from that location. Any area that was visible from the trail was considered to detract from the recreational potential and was to be omitted from the search. For this reason, a value of zero or one was assigned to the areas of the viewshed analysis. If any land was visible from the trail it was to be omitted.
All of this data was able to be combined using the raster calculator tool. First, the streams, farmland, residential proximity, and school proximity were added and multiplied by the wildlife areas to omit any area that did not meet the criteria of being suitable, but within the 500 meter zone near the wildlife areas. This raster was then multiplied by the viewshed raster to determine a final model for this section. A data flow model was created to illustrate this workflow (Figure 5).
 |
| Figure 5. This figure shows the data flow model created in Modelbuilder for analyzing and combining the streams, farmland, residential proximity, school proximity, wildlife areas, and viewshed. |
The raster that showed suitability for land necessity and rail terminal proximity was then combined with the raster that showed suitability for environmental and community impact (Figure 6).
 |
| Figure 6. This model generated the final suitability model that combined the other two suitability rasters. |
A python script was written to assign a weighted value to a raster that I felt was required more value than the other rasters. I chose to add a weight to the residential zones, as I felt they were the most likely to be affected by having a mine in close proximity to them (Figure 7).
 |
| Figure 7. This figure shows the python script required to add a weighted value to a certain raster. I chose to assign weight to the residential raster, as it was the most likely to be affected by the proximity of a mine. |
Results
It should be noted that all of the data in the following section is hypothetical and in no way calculates usable results other than those determined for the sole purpose of this exercise. The results gathered were a number of maps showing suitability. Figure 8 shows the suitability raster for land necessity and rail terminal proximity. Figure 9 shows the suitability raster for environmental and community risks. Figure 10 shows the viewshed visibility raster. Figure 11 shows the combined raster for the viewshed and the environmental and community risks. Figure 12 shows the combined raster for land necessity and rail terminal proximity and environmental, community, and viewshed suitability. Finally, Figure 13 shows the result of the weighted model created from the Python script.
 |
| Figure 8. This map shows the suitability results for the bedrock geology, NLCD, distance from rail terminal, percent slope, and water table elevation data. The most suitable location in this map is in the west central section of the study area. This is the location of the one rail terminal in the study area. |
 |
| Figure 9. This map shows the suitability results for the streams, farmland, residential proximity, school proximity, and wildlife areas. The most suitable location in this map is dispersed throughout the study area. The area with the highest suitability appears to be in the southern portion of the map, in a small dark green cluster, however it is well dispersed throughout the map. |
 |
| Figure 10. This map shows the viewshed suitability raster. Suitability was determined for a recreational trail along the Mississippi River in the southwestern portion of the study area. Black signifies areas that are not visible from the trail, and therefore areas suitable for a frac sand mine. The light gray signifies areas visible from the trail, therefore not suitable for a frac sand mine. |
 |
| FIgure 11. This map shows the suitability results for streams, farmland, residential proximity, school proximity, wildlife areas, and viewshed. Much of the area in the southwestern portion is red when the viewshed is taken into account. |
 |
| Figure 12. This map shows the suitability of the rasters from Figure 8 and Figure 11. The most suitable area is in the west-central portion, where there is the highest concentration of dark green, signifying the most suitable area. |
 |
| Figure 13. This map shows a comparison between the environmental and community suitability before and after the weight was applied to the residential zones. The left map shows the suitability raster before the weight was applied. The right map shows the suitability raster after the weight was applied. |
Examining the various criteria created a number of very interesting results. Perhaps the most interesting to me was examining the results derived from the community and environmental risk factors, as I feel this did the best job of denoting areas of unsuitability based upon their impact on desirable areas for community or environmental health. The original suitability raster developed shows the most suitable location based upon the criteria necessary for an economically productive frac sand mine. The correct geologic formations, proper land classification, gentle slopes, and a short distance to transport produced the most suitable location for a mine. In fact, it seems like the variable with the greatest effect on the data was that of the distance from the railroad terminal located in the west-central portion of the study area. In examining the final model before the weighted value was applied, it seems apparent that the area in the central portion of the study area, nearest the rail terminal, is the most suitable location for a rail terminal.
Conclusion
The functionality inherent in raster analysis tools opens a wide array of advanced analytical functions. Not only are we able to reclassify data based upon desired criteria, but we are also able to use the Raster Calculator function to compare and add rasters together. When determining suitability, these tools provide a toolset that vector analysis cannot provide.
Works Cited
Trempealeau County Land Records. 2014. [cited 13 December 2014]. Available from http://www.tremplocounty.com/landrecords/.Wisconsin Geological and Natural History Survey. 2014. [cited 13 December 2014]. Available from http://wgnhs.uwex.edu/maps-data/gis-data/.
No comments:
Post a Comment