Introduction, Goals, and Objectives:
Sand mining in the state of Wisconsin is a very profitable enterprise, with the sand becoming more valuable as new fracking operations expand throughout the United States. As such, mines will continue to be established to meet this overwhelming demand. While the establishment and continued operation of these sand mines means an economic boom in the area where these mines are located, it also means that sometimes these mines go up without taking consideration how they will affect the surrounding area. A mine has many components that can irritate the space around it, and if these are not taken into consideration, mining sand in that area will likely cause more harm than good. Mining operations have present a number of threats to the environment, such as contamination of nearby streams, and more generally, the water table. Additionally, the mining process will likely clear some land of other uses, such as eliminating the growing capacity of prime farmland. If built too close to residential neighborhoods, recreation facilities, or schools, the mining can at the very least provide a noise nuisance to the nearby people. Frac sand mining in close proximity to these areas is especially hazardous, as the sand has the potential to drift into these areas, resulting in the people breathing in the dust, which can result in serious health issues, which were discussed in the first post in this series.
To assist in planning for these issues, a suitability and risk model can be built using ArcGIS software, which can take into account these factors already discussed, and many more. The final product will be a map of Trempealeau County that will rank the best places to put a sand mine.
The data that was used in this exercise was principally gathered from the geodatabase that was built in the beginning section of this activity, largely including the Trempealeau County geodatabase features, like streams, land, zoning areas, parcel data, recreation facilities, and wild area. A
specific projection provides the best results, so NAD83 UTM 15N was
selected, giving the data the best possible representation. To determine suitable land cover and elevation for the study area was obtained through the
USGS National Map Viewer,
which provided DEM raster files that required mosaicking them together
to get a uniform image to be used for the model. The rail terminal locations were accessed by using the feature class that was provided for the network analysis activity. One new piece of data that we did not have from prior activities water table depth, which was obtained through the
Wisconsin Geological Survey.
Methodology:
A significant portion of this exercise involved using raster datasets to display the desired fields for both the suitability and risk models. By incorporating data in different extents, like the DEM, rail terminal locations, and land use/land cover, this requires setting a mask for the output so that the results of the analyses run would fit to a specified area. For this activity, the county boundary for Trempealeau County was used. With this done, the raster manipulation can occur. This involved a fair amount of reclassication of the original files and ranking the data on suitability or risk factors, with 3 being high, 2 medium, and 1 low. We were left a good deal of freedom as to how we wanted to rank each of the categories, but required justification on each.
Creating the Suitability Model for Trempealeau County
The first step of this activity was to create a suitability raster to demonstrate areas where building a sand mine would be the best, based on a number of factors including: geologic formation, land use/land cover, distance from rail terminals, slope, and water table depth.
Suitable Elevation for Sand Mine Establishment
To determine desired elevation for a sand mine, a number of processes needed to be completed. The most desirable geologic formations for sand mining in the state are the Jordan and Wonewoc formations. To identify these features, a georeferenced bedrock geology map was used as a base layer, and contour lines were generated through the Digital Elevation Model. The elevation of the two formations were identified, and then the DEM could be reclassified to take this information into account. Figure 1 below shows the results of the ranking process with elevations greater than 360 meters found to contain the Jordan and Wonewoc formations. Using these classification values, a model was produced to reclassify the raster into these three elevations, shown in Figure 2. After the reclassification, a map was then produced, Figure 3, again, with 3 being the favorable rank, and 1 being undesirable.
 |
| Figure 1: Table showing the ranking classification of the elevation criteria for the suitability model |
|
 |
| Figure 2: Model used to produce the reclassified elevation values for favorable geologic formations |
 |
| Figure 3: Reclassified raster demonstrating the suitable geologic formations in Trempealeau County |
Suitable Land Based on Land Use/Land Cover Data
By using satellite imagery to understand the type of land use/land cover of Trempealeau County, we were then asked to determine which types of land would be best to build new sand mines on. Land use/land cover types are broken up into the band that they reflect back to the satellite when collected. Each land cover type has a band of the possible 255 bands associated with it. For this activity, it was determined that 81, or pasture/hay, would be ranked the highest, because there is the least amount of work having to be done with the land to prepare it for sand mining, and it also does not disrupt crop growing. The second criteria was 71, or grassland/herbaceous. Again, the logic was similar to the first criterion, with grassland being relatively easy to clear, and no disruption to crop growing would occur. Lastly, the third criteria, 81, was cultivated crop land. This area would be easy to begin digging for sand, as the land has been used for growing food and already has a layer plowed. The downside here is that food could not be grown for the foreseeable future on this land. Figure 4 shows the table of classification values used in this section of the model, Figure 5 showing the associated model, with Figure 6 demonstrating the raster produced by this process.
 |
| Figure 4: The three land types chosen to be suitable for sand mining |
 |
| Figure 5: Model used to reclassify the land use/land cover raster of Trempealeau County |
 |
| Figure 6: Reclassified raster showing the three suitable land use/land cover types for frac sand mining in Trempealeau County |
In addition to suitable land use/land cover in Trempealeau, we were also asked to produce an exclusion raster to be used in the final risk assessment model. This process simply required a flipping of the values gathered in the first land use/land cover map, with values that were suitable given a 0, and values that were not the top three categories given a 1. Figure 7 shows the model used to produce Figure 8 below, which demonstrates the produced map.
 |
| Figure 7: Model used to produce the unfavorable land use/land cover types |
 |
| Figure 8: Binary raster showing the unsuitable land use/land cover for Trempealeau county. 1 means that the areas is unsuitable, and 0 means that it is suitable. |
Finding Land Suitable to Proximity to Railroads Criteria
A significant factor in the placement of frac sand mines is the mine's proximity to rail terminals. The sand is in such high demand that whatever gives the mine's owner the ability to distribute the sand to buyer quickly is ultimately a principle factor in the positioning a new mine. The less cost an owner has to incur due to transportation fees, such as leasing a truck, paying for gas, and other expenses, the more profit is available. Unfortunately, there are no rail terminals in Trempealeau County, with the closest ones being in Eau Claire and La Crosse Counties, to the north and south, respectively, of Trempealeau County. Consequently, mines located in the northern and southern sections of the county are more desirable from a transportation cost standpoint. To determine where the best places to locate these mines are, Euclidean Distance was run to determine the distance from the rail terminals. Then, the results of the tool were reclassified into three categories. The state was essentially divided into three equal sections, with the middle of the state, the furthest from the rail terminals, given the worst rank, the next sections out given a 2, and the sections closest to the northern and southern borders being the most desirable and given a 3. The model that was run in this process can be seen in Figure 9 below, as well as the table that produced the reclassified map, in Figure 10. The Euclidean Distance results and reclassified map can be seen in Figures 11 and 12, respectively. These two maps look like there were circles imposed over the county, and that is because ArcMap is drawing circles representing the distance from the rail terminals in all directions.
 |
| Figure 9: Model run to determine distance from rail criteria |
 |
| Figure 10: Reclassified values for the distance from rail terminals criteria |
 |
| Figure 11: Euclidean Distance from the rail terminals, in their relation to Trempealeau County |
 |
| Figure 12: Suitability raster showing the distance from rail terminals for Trempealeau County |
Determining Suitable Slope Criteria for Frac Sand Mines
Finding a suitable slope for construction of any mine, especially a frac sand mine, is an incredibly important factor. Building on a level surface, or as close as possible to one, makes the mining process easier, maintains the integrity of the environment, and allows for the greatest amount of sand to be excavated with minimal costs associated. To identify the appropriate slope in this category, a number of steps needed to be completed. First, a digital elevation model of Trempealeau County needed to have slope computed on it, to determine the percent rise. The output of this tool can be seen in Figure 13. Notice the somewhat grainy appearance of the results. To fix this, Block Statistics was run on the slope, giving a neighborhood setting to height and width of 3 cells. This produces a smoother appearing image that can be seen in Figure 14. Finally, by using the smoother image, a reclassification can then be done on the raster. It was determined that the more level surface would be the most desirable, so the three rankings put 0-10 percent rise at 3, 10-25 percent rise at 2, and 25 to 103.56 percent rise at 1. The ranking table can be seen in Figure 15. Now that the values were determined, the model can be run, seen in Figure 16, and the final raster for the slope criteria can then be generated, seen in Figure 17.
 |
| Figure 13: Percent rise of slope in Trempealeau County |
 |
| Figure 14: Block statistics ran on the slope values computed in Figure 13 |
 |
| Figure 15: Reclassification values of the slope block statistics |
 |
| Figure 16: Model run to determine the slope criteria for the suitability model |
 |
| Figure 17: Reclassified slope values for the suitability model |
Determining Suitable Water Table Depth Criteria
Depth from the water table is another factor in a suitability model that needs to be taken into consideration. Due to the mining process, there is the possibility of ground contamination that is taken into place, be it from oil or gasoline spills from the trucks, or mining fluids seeping into the ground. Water table depth is crucial, because it determines how far down the water is. The closer it is to the surface of the ground, the greater risk that water will be contaminated. To do this, water table contours were downloaded from the Wisconsin Geological Survey in a coverage (.a00) file type. This then needed to be uploaded to a geodatabase and converted in order to read in ArcMap, with the lines shown in Figure 18. In order to run raster analysis on the contour lines, the Topo to Raster Tool was run, converting these line features in to raster format, with the transformation visible in Figure 19. From this point, the raster was able to be reclassified based on the criteria found in Figure 20. The greater the distance of the water table depth would be the most desirable, with 290-342.72699 being ranked the best, 250-290 ranking 2, and 222.468-250 being the worst. These figures were arrived at by simply employing the Natural Breaks classification within ArcMap, which examines the input data, and proscribes breaks based on the natural distribution of values. The model run for the process can be seen in Figure 21, with the final reclassified raster seen in Figure 22.
 |
| Figure 18: Contour lines showing the depth of the water table in Trempealeau County |
 |
| Figure 19: Raster conversion of the contour lines showing water table depth in Trempealeau County |
 |
| Figure 20: Reclassification values for the raster generated in Figure 19 |
 |
| Figure 21: Model run to reclassify the water table depth raster |
 |
| Figure 22: Raster demonstrating the reclassified values for the depth from water table criteria |
Creation of the Risk Model for Trempealeau County
The next step in the process was to create a series of rasters that could be used to determine areas of risk in the county due to proximity to certain areas. These at-risk categories included proximity to streams, prime farmland, schools, and residential areas, visibility from prime recreation areas, such as walking trails and parks, and finally, a factor of our choice, which was distance from wildlife areas.
Proximity to Streams
The distance from streams is a crucial factor to consider when examining risk from a mine. The runoff from the mine has the potential to creep into the river system, contaminating everything that is inside of it. This seems like it could be an easy task, simply done by selecting all the streams in the county and ensuring that there is nothing that could be dangerous in their proximity. The problem arises when the number of waterways are taken into account. By including all running bodies of water into the model, there would be no areas in Trempealeau County suitable for a mine. So a decision needed to be made as to which streams would be included, and which would be removed. After careful consideration, it was decided that primary and secondary perennial, over land flow, streams would be the ones included in the model. These streams run years round and over land, making the risk associated with their contamination that much more serious. These streams were selected from the more general streams feature class found in the Trempealeau County geodatabase, and then had Euclidean Distance run on them, seen in Figure 23. From here, the distances were reclassified based on the values in Figure 24. Close proximity, from 0-2000 meters from a stream, were given a 3, medium distances 2000-6000 meters given a 2, and fairly far away distance of 6000-18090.498 meters given a 1. The model was then run, with the specified parameters visible in Figure 25, and the reclassified hazard index for streams can be seen in Figure 26.
 |
| Figure 23: Euclidean distance from selected streams in Trempealeau County |
 |
| Figure 24: Reclassification values to be used in the final stream risk model |
 |
| Figure 25: Model used to compute the hazard to streams criteria |
 |
| Figure 26: Reclassified raster showing areas of risk to the selected stream criteria in Trempealeau County |
Proximity to Prime Farmland
The distance to areas of prime farmland is a second factor that needs to be accounted for when determining the environmental effects of mining. The economic benefits of mining are wanted, but the side effects, including possible contamination of farmland, needs to be controlled. This was found in the Land feature class of the Trempealeau County geodatabase, and then selected out. Euclidean Distance was run to determine the distance from the farmland, which can be seen in the map in Figure 27. Reclassification values were then computed with the table in Figure 28. 500 meters was deemed to be very close to the farmland, so that had a rank of 3. 500-100 was ranked 2, and 1000-11530.81 got the lowest rank of 1. The model was then run, as seen in Figure 29, and the final hazard raster was then produced, as seen in Figure 30.
 |
| Figure 27: Euclidean Distance of prime farmland within Trempealeau County |
 |
| Figure 28: Reclassification values to be used in the proximity to prime farmland model |
 |
| Figure 29: Model used to compute the prime farmland risk raster |
 |
| Figure 30: Risk assessment of area of prime farmland within Trempealeau County |
Proximity to Residential Areas
Another factor that needs to be taken into account is the proximity of the sand mine to residential zones. Mining operations can be very loud, and cause a nuisance to anybody living in the immediate vicinity, so in order for mines to take this into consideration, they need to be a set distance away from these high population residential zones. To do this, a noise shed, or a range where the noise of the mine would be heard in normal circumstances, needs to be obtained. The minimum distance from a mine where noise cannot be heard is 640 meters, so that will be the highest in the risk classification. To determine residential zones, the zoning feature class of the Trempealeau County geodatabase can be imported into ArcMap, and the residential zones removed. For these purposes, Residential Zones 8 and 20 were selected to form the basis for these kinds of neighborhoods, and the resulting map can be seen in Figure 30. Euclidean Distance was then run on the zones, which can be seen on the map in Figure 31. From here, the areas were reclassified based on the criteria found in Figure 32, with 0-640 meters being the highest risk, receiving a score of 3, 640-1500 meters receiving a score of 2, and 1500-51223.742. Now that these values are determined, the model can be run, which can be seen in Figure 33, with the final result raster being visible in Figure 34.
 |
| Figure 30: Residential zones 8 and 20, the criteria for the proximity to residential zone criteria |
 |
| Figure 31: Euclidean distance of residential zones within Trempealeau County |
 |
| Figure 32: Reclassification values for the risk assessment raster of residential zones |
 |
| Figure 33: Model used to reclassify the residential zones in Trempealeau County |
 |
| Figure 34: Reclassified values of residential zones within Trempealeau County |
Impact to Schools Criteria
Related to the residential zones criteria, the proximity to schools is another crucial component to assessing the potential risk that a frac sand mining operation can have on a community. The distance from a mine to a school can have several consequences on the children and faculty, including, but not limited to, excessive noise, increased traffic, and dust from the mine blowing to the school. For those reasons, this is another important factor in the risk model. This is not a simple task, however, because there is no schools feature class in the Trempealeau County geodatabase. This meant that the plats feature class would need to be searched for land owned by a school. This involved writing a query statement to parse out the appropriate items. The land in the county owned by schools can be seen in the map in Figure 35. Euclidean Distance can then be run on the school owned land, visible on the map in Figure 36. Reclassification values were the same as the residential zones, because factors like the noise shed. 0-640 meters were ranked 3, 640-1500 meters got a 2, and 1500-14080 received a 1, as visible in the table in Figure 37. The model that ran these operations can be seen in Figure 38, and the final raster risk map is demonstrated in Figure 39.
 |
| Figure 35: School owned land in Trempealeau County |
 |
| Figure 36: Euclidean distance of land owned by schools within the county |
 |
| Figure 37: Reclassification values of the proximity to schools criteria |
 |
| Figure 38: Model used to compute the risk model for proximity to schools |
 |
| Figure 39: Reclassified raster showing the proximity of school-owned land criteria |
Visibility from Prime Recreational Areas
A factor that plays a factor in the risk assessment is the impact of the mines on recreation facilities. In the case of this model, this includes parks and trails, both of which are renowned in Trempealeau County, and can be seen in Figures 40 and 41, respectively. If these were impacted by the sand mines, the county could lose a lot of money in tourism funds. The main issue here is if the sand mines would be visible from the parks and trails. People use these facilities to get back to nature, and if there is a sand mine in the immediate vicinity, people will be less likely to visit these places. To calculate if the mines are visible from these areas, a viewshed was run on the parks and trails feature classes. Viewsheds compute the distance visible from the input feature, to produce a raster showing if an area can be seen from the input feature. To do this, a line or point feature class is combined with a digital elevation model to determine visibility from the areas.While this is convenient, the parks feature class in the Trempealeau County geodatabase is in a polygon format. So in order to run this took, the polygon needed to be converted to a point feature class, with the point being put in the middle of the polygon, seen in Figure 42. Viewsheds could then be computed on both the park and trails features classes by combining them with the digital elevation models, with those results visible in Figures 43 and 44, respectively. The viewsheds are then reclassified, based on the values in Figure 45, which were generated by simply using the Natural Breaks classification method. The model was run, demonstrated in Figure 46, and the final maps for parks and trails are seen in Figures 47 and 48, respectively.
 |
| Figure 40: Map showing the locations of parks within Trempealeau County |
 |
| Figure 41: Trails found within Trempealeau County |
 |
| Figure 42: Parks feature class converted to a point feature class for viewshed operations |
 |
| Figure 43: Viewshed run from the parks within Trempealeau County |
 |
| Figure 44: Viewshed run from the trails within Trempealeau County |
 |
| Figure 45: Reclassification values for both parks and trails within Trempealeau County |
 |
| Figure 46: Model run to compute the rasters for both the trails and parks criteria |
 |
| Figure 47: Reclassified raster of the parks hazard within Trempealeau County |
 |
| Figure 48: Reclassified raster of the trails hazard within Trempealeau County |
Proximity to Wildlife Areas
For the last factor in the risk model, we were given the freedom to choose our own input feature class that we wanted to include in the model. I decided to incorporate distance from wildlife areas in my model because of the obvious significance of protecting these areas from mining operations. These protected lands can be viewed on the map in Figure 49. The wildlife areas had Euclidean Distance ran on them, producing the values seen in Figure 50. From here, the raster was reclassified based on the values found in Figure 51. The 0-10,000 meter and 10,000-25,000 meter breaks were relatively arbitrary, but were selected to give a wide area to the wildlife areas, so as to not disturb the animals. The process was run based on the model seen in Figure 52, and the final raster, seen in Figure 54, was produced.
 |
| Figure 49: Wildlife areas found within Trempealeau County |
 |
| Figure 50: Euclidean distance from wildlife areas within Trempealeau County |
 |
| Figure 51: Reclassification values for the proximity to wildlife areas criteria |
 |
| Figure 52: Model used to compute the proximity to wildlife area criteria |
 |
| Figure 53: Reclassified rasters showing the risk assessment of the proximity to wildlife area criteria |
Discussion:
Creating the Suitability Model
Once the reclassified rasters were
created, the final suitability model can be compiled. This process
involves using a tool called Raster Calculator to add the rasters
together, giving an ultimate value from 3 to 15, with 3 being the worst
possible rank and 15 being the best rank. These value are arrived at by
adding up the three ranking classes for each model, and then a final
value can be arrived at. The exclusion raster was also subtracted from
the raster, which would eliminate the unsuitable land use/land cover
areas from the suitability raster. The model of the entire process,
which is quite extensive, can be seen in Figure 54, and the final
suitability raster can be viewed in Figure 55.
 |
| Figure 54: Model run to compute the suitability model raster |
 |
| Figure 55: Raster model showing areas of suitability within Trempealeau County: The middle of the state appears to be the most unsuitable for the developmetn of new mines, mainly because of its steep slope and its distance to the nearest rail terminals. |
Creating the Risk Model
Once these seven processes were run, raster calculator was then run to produce a raster map demonstrating the output of the seven risk maps added together. The model, which is quite detailed, can be seen in Figure 56, produced the map detailing the risk assessment of Trempealeau County seen in Figure 57. The values range from 7 to 21, with 7 being the lowest risk, and 21 being the highest risk.
 |
| Figure 56: Model used to compute the risk assessment of Trempealeau County |
 |
| Figure 57: Raster model showing areas of risk within Trempealeau County. The map appears not as uniform as the suitability model because of the extensive Euclidean Distance processes run on it, which produce a number of circles as outputs, which explains the presence of a large number of those shapes on this particular map. |
As visible from the two maps, there is quite a dissimilarity between the criteria of the suitability and risk models. The suitability models indicate that there is a great potential for sand mining in the northern part of Trempealeau County, while the risk model indicates a high number of hazards in that same part of the county. The safest part of the county appears to be in the middle of the state, where the grouping of hazards is the least, but this area is also the furthest away from rail terminals, a critical factor in the placement of new sand mines. It is clear that there will have to be some sort of compromise between the two criteria, with some less favorable areas being selected in the effort of getting the best possible sand mine, and eliminating the greatest amount of risk associated with opening a frac sand mine.
Conclusion:
This activity was an incredibly valuable learning experience. We were given a surprising amount of freedom as to how to execute the functions needed for each of the rasters. Instructions were given, but they were as bare as they could be. This was the first time in which we were essentially on our own for an assignment, and that amount of freedom was refreshing. It was empowering knowing that it was up to our own accumulated knowledge in the course to finish the raster models. There were difficulties along the way, principally the significant amount of time that was needed to run each of the operations in the process. Specifically, the viewshed for the trails was an extremely time consuming tool, taking several hours to complete. Being required to use Model Builder to construct our operations forced us to organize our thoughts prior to carrying out the tasks, which greatly assisted in logically planning out what we would do for each step. Having to work step by step through each function assisted in becoming proficient in the Spatial Analyst extension of ArcMap, giving us the skills to utilize this powerful section of the software for use in future assignments and operations in the professional world.
