Goals and Objectives
In this lab, our goal was to show the suitability and risks of frac sand mining within a section of our study area Trempealeau County (Figure 1). This is accomplished through building suitability models using raster analysis. Using spatial analysis tools including reclassify, Euclidian distance, and raster calculator it is possible to assign rankings to a variety of data, and then combine them together into single maps that show a range of rankings that correspond to the best and worst places to build a new sand mine, based on a variety of different criteria.
 |
| Figure 1. Location of Trempealeau County in Wisconsin |
As the prevalence of frac sand mining increases, it becomes necessary to look at how negative impacts can be mitigated, reduced or avoided. Furthermore, as citizen concerns grow about environmental, economic, public health concerns associated with frac sand mining(Midwest Environmental Advocates, 2014), preemptive analysis and avoidance of high risk area is essential to the industry. Raster analysis provides a way to critically look at suitability and an understanding of raster analysis tools can be beneficial to have when looking at complex problems with many factors.
 |
| Figure 2. Photograph of a sand mine from (Midwest Environmental Activists, 2014) |
Data
A variety of publicly available data was utilized for making our suitability model. Most of the data came from feature classes in the Trempealeau county geodatabase and rasters that we had downloaded from various websites (Figure 3). A more detailed description of this data collection is available in a previous blog post. Additionally we downloaded coverages of a water elevation map for Trempealeau county. The sources for the data at the end of the blog.
 |
| Figure 3. Selection of rasters utilized for analysis in sand mining suitability model. |
Multiple characteristics were chosen as criteria to evaluate suitability and risk (Figure 4) The type of geologic formation, the current land use of the area, the distance to rail terminals, the slope of the land were all used to evaluate suitability of particular areas. The sand mine would ideally be in a place that has geologic formations where frac sand is occurring with a land use that could be altered (for example, it is unlikely a lake would be drained to build a sand mine). Additionally access to rail terminals for transportation of the sand, a slope that isn't too high, and a water table that isn't too low so that water can be accessed to process the sand.
The distance from streams, residential areas, schools, landfills, and the occurrence of prime farmland were all used to evaluate risks associated with sand mining in different areas. Ideally a new sand mine would be located away from schools and residential areas to reduce social impact, away from prime farmland to reduce economic impacts, and away from both streams and landfills to reduce environmental impacts that could be associated with both. I personally chose landfills to add to my raster analysis, because I think there would be pollution concerns if landfills needed to be moved to accommodate sand mines.
 |
| Figure 3. Visual representation of different criteria utilized in sand mining suitability analysis |
Methods
For each of the individual factors, rasters were made and reclassified so that they would give a rank of the data, which could that be added together using raster calculator. The ranking system that we had ranged from 1-3 with 1 being low suitability and 3 being high suitability. There are 10 factors that we are looking at, meaning that the maximum value for suitability that should be visible in the final raster in 30.
The majority of the processing was achieved with three spatial analysis tools, Reclassify, Euclidean distance, and Raster Calculator.
Reclassify
Reclassify allows you to change values within a raster. Functionally this can allow you to do many things, such as reduce the amount of distinct values you have in a raster or create an inverted version of a raster that you have. In this lab reclassify was frequently used to produce rasters with rankings.
Euclidean Distance
Euclidian distance calculates a number for each cell within the raster that corresponds to its distance from the closest input source feature. For example, if the input source feature was a road, Euclidian distance could generate a raster where each cell gives the distance from the road. This tool was useful when settings ranks for risk based on how close it is to a certain feature.
Raster Calculator
Raster calculator allows you to write equations for rasters that will be performed on every cell, generating an output raster based on your equation. Using this tool to add rasters together can give you a combined rating based on multiple rasters. This tool was useful when deriving our final suitability rasters.
Other tools such as, Topo to Raster, Polygon to Raster, Slope other spatial analysis tools were also utilized throughout the activity.
Ranking Determination
In the lab, each of the ten criteria that were used as part of the final model had to ranked from 3-1 with three being the highest suitability and 1 being the lowest suitability. The rankings for the datasets can be seen in Figure 4, and are explained below:
Geology
The Jordan and Wonewoc formations are the most sought after for purposes of sand mining (Dustman et. al., 2011). These formations were given a ranking of 3, and the other formations were given a formation of 1. The geology ranking raster was produced by converting a geology feature class into a raster, and then performing reclassify (Figure 6.)
Land Use
Land use was ranked for suitability based on what was most practical. For example, it is not practical to drain a lake, or move an apartment building in order to build a sand mine. The most impractical areas were removed by reclassifying them as NoData, and the other land uses were given rankings. Barren land, pasture, and grasslands were given the highest ranking due to practicality in terms of overburden removal. The land use ranking raster was produced by reclassifying the National Land Cover data raster (Figure 7.)
Railroads
The railroad ranking was based on how far away a given place was from a rail terminal. This is because it can be both expensive for the companies to truck the sand across roads, and can also have wear out the local roads (Hart, et. al., 2013). The railroad ranking raster was produced running the Euclidian distance tool with rail terminals in the state of Wisconsin as an input source, and then reclassifying this into a ranking. (Figure 8.)
Slope
The slope ranking was based on how steep the slope was, with lower slopes having the highest ranking. The slope ranking was produced by taking an digital elevation model from the USGS, using the slope tool, and then reclassifying this into a ranking (Figure 9)
Water Table
The water table ranking was is based on how deep the water is in the ground. Sand mines need large quantities of water for processing (Prengaman, 2013), so areas where water is closer to the surface received the highest ranking. The water table ranking was produced by taking water elevations contours, converting this into a raster, and then reclassifying this into a ranking (Figure 10.)
Streams
Streams located near sand mines may be negatively impacted by storm run off events, and degradation of riparian ecosystem can cause loss of wildlife(Wisconsin DNR, 2012), therefore areas that are farther from streams are given a higher rank than areas nearer to the streams (the rank corresponds to suitability to mining). The distance from streams ranking raster was produced by selecting streams that were greater than 2 in order, using them as a input source feature to run euclidean distance and then reclassifying this into a ranking (Figure 11.)
Prime Farmland
There are some studies that reclaimed sand mines put back into use as agriculture may still produce crop yields of similar volume (Orr and Krumenacher, 2015), but it is still better to find mine sites at places that are not prime farmland, therefore areas that were rated as prime farmland were ranked as low suitability. The farm ranking raster was produced by converting the farmland feature class into a raster, and then reclassifying into a ranking (Figure 12).
Residential Areas
Noise, and air pollution are both two reasons why it would be better to keep sand mines further away from residential areas ((Wisconsin DNR, 2012), therefore areas close to residential areas were given the lower ranking, and areas further away were given the higher ranking. The residential ranking raster was taken by reclassifying the land use raster for residential areas, and then running a euclidean distance and then reclassifying into a rank.
Schools
For similar reasons as the residential areas, it would be better to keep the sand mines further away from schools, therefore areas closer to schools were given the lower ranking and areas further away were given a higher ranking. The school distance ranking raster was made by selecting land parcels owned by school districts and then running a euclidean distance and then reclassifying into a rank (Figure 13)
Landfills
I chose this site because there are sometimes issues with contaminated runoff associated with frac sand mining. There could some potential for greater environmental harm if a sand mine was located near a landfill, not to mention the economic costs as properties that are both near a sand mine and a landfill may be very low in value. Therefore areas farther away from landfills I ranked as high, and areas near the landfill I ranked as low. The landfill ranking raster was made by running a euclidean distance with landfills as the source feature, and then reclassifying this into a rank (Figure 14)
 |
| Figure 4. Values associated with the ranks of various datasets |
 |
| Figure 5. Geologic suitabilility dataflow shown on left, and map shown on right |
 |
| Figure 6. Land use suitability dataflow shown on left, map shown on right |
 |
| Figure 7. Distance from rail terminal suitability data flow shown left, map shown on right |
 |
| Figure 8. Suitability of slope dataflow shown on left, map shon on right |
 |
| Figure 9. Suitability of water table dataflow shown on left, map shown on right |
 |
| Figure 10. Stream risk dataflow shown on left, map shown on right |
 |
| Figure 11. Farmland dataflow shown on left, map shown on right |
 |
| Figure 12. School risk dataflow shown on left, map on right |
 |
| Figure 13. Lanfill risk dataflow shown on left, map on right |
The data flow models were split into two sections, suitability and risk, shown in figures 14 and 15 respectively. Afterwards those models were run, the two outputs of each of the respective models were added together using raster calculator.
 |
| Figure 14. Mining suitability dataflow model |
 |
| Figure 15. Mining risk data flow model |
Results
The results of the mining suitability data flow model can be seen in Figure 16, and the results of the the mining risk data flow can be seen in Figure 17. The combined overlaid outputs of both suitability and risk can be seen in Figure 18. The final has a ranking of classification that goes. There was an issue with the raster being clipped to a smaller extent, but it still can be used to show the different areas of high and low suitability.
 |
| Figure 16. Output of the mining suitability data flow model |
 |
| Figure 17. Output of the mining risk data flow model |
 |
| Figure 18. Combined output of both models |
Discussions
Using the map, it is easy to see that different areas have different rankings in terms of risk and suitability for sand mining. The final map does offer some usefulness, but could be improved by making adjustments to the model. If this map was intended for actual use by a mining company, it would be of greater importance to do more detailed research on the qualities of land that make them suitable or risky places to mine. This would help set the rankings better reflect the reality of conditions. Additionally, in a more advanced model, it would not be assumed that all of these factors are of equal importance, and therefore different weightings could be assigned (This was shown in a Python Script).
Conclusions
In this lab, the suitability and risk of frac sand mining was evaluated for areas within our study region. This lab, and the previous lab on network analysis, can offer a critical evaluation of sand mining, and can be useful in several ways. It can be useful for the frac sand mining industry, as it helps search for appropriate places to build mines, and it could also be useful to policy makers looking for ways to regulate sand mining. It further shows the usefulness of GIS techniques in helping solve real world problems, and specifically how raster analysis can be useful. The tools of raster analysis can be used to design a wide array of suitability analyses which can help facilitate effective decision making.
Data Sources
Land use, National Land Cover Dataset
Railroad network, Department of Transportation
Slope, National Elevation Dataset
Water Table, Wisconsin Geological Survey
Stream, Trempealeau County Land Records
Prime Farmland, Trempealeau County Land Records
Residential areas, National Land Cover Dataset
Landfill, Trempealeau County Land Records
County boundaries, Trempealeau County Land Records
Information Sources
Dustman, J.E., Gulbranson, B., Bell, P., Gregg, W., 2011: Characteristics of high quality frac sand, and where to find it in the upper Midwest., Geological Society of America Abstracts with Programs, Vol. 43, N 5.
Hart, M. V., Adams, T., & Schwartz, A 2013. Transportation Impacts of Frac Sand Mining ;CFIRE
Prengamon K. 2013 New studies measure air, water impacts of frac sand mines, Wisconsinwatch.org
Wisconsin DNR, 2012 Silica Sand Mining in Wisconsin, dnr.wi.gov
Image Sources
Sepp, S. Fine grained sand (2015). Retrieved:
https://commons.wikimedia.org/wiki/File:Sand_from_Gobi_Desert.jpg
Nasa. Satellite photograph (2009). Retrieved:
https://en.wikipedia.org/wiki/Agriculture#/media/File:Precision_Farming_in_Minnesota_-_Natural_Colour.jpg
Sturmovik, Railroad photograph (2008). Retrieved:
https://en.wikipedia.org/wiki/Pennsylvania_Railroad#/media/File:NS_Buffalo-Line-Signal-304-3042-APPROACH.jpg
O’Neil,, A. Sloping hill photograph (2011). Retrieved:
http://www.geograph.org.uk/photo/2424923
Dilmen, N. Water texture (2013). Retrieved:
https://commons.wikimedia.org/wiki/File:Water_texture_1390020_Nevit.jpg
St. John, J. Sandy River photograph (2009). Retrieved:
https://www.flickr.com/photos/jsjgeology/20035367951
USDA NRCS, Corn in Residue with Healthy Soil (2013). Retrieved:
https://www.flickr.com/photos/nrcs_south_dakota/9314025147
Hodgson, A. Residential Area (2005). Retrieved:
https://commons.wikimedia.org/wiki/File:Residential_area_off_Tileshads_Lane,_Cleadon_Village_-_geograph.org.uk_-_6409.jpg
Public domain photo, second grade writing class (2010). Retrieved:
https://www.flickr.com/photos/wwworks/5074150006
