Sand Mining in Western Wisconsin: Suitability Model

Goal and objectives

Mining and environmental sustainability have traditionally been at odd ends. The purpose of this exercise is to create a suitability model able to show both the best locations for sand mining and highlight areas that a sand mine would have negative effects on environmental health in the county of Trempealeau Wisconsin.

Data Sources

The data used for this project is publicly available and was collected in an earlier project (see Data Collection). Water table data was downloaded from the Wisconsin Geological Survey's Website.

In order to complete the objective for this project three models were created, each of which produced a unique raster image. Below is an image of each model with a description of the steps completed.

Sand Mine Suitability Model

Figure One. This Mine Suitability Model is designed to produce a raster which ranks areas most suitable for the construction of a frac mine based off of several variables. 

This model (Figure One) uses four different datasets to produce a final raster image highlighting the area that would work well for the construction of sand mines while identifying areas which a sand mine would not be cost effective or no resources would be present. Each data set was put through a series of manipulations and reclassified to create a suitability raster file based on a number of classes. Table One below outlines the vales for each class change. 

Table One. Reclassified values and class break locations for each data source. 

The first operation is identifying areas that have the correct types of bedrock for a sand mine. The TMP_geology vector file was used for this task. It was converted to a raster file, then reclassified to three rankings with the highest number being 3 for Jordan and Wonewoc formations while all other formations were assigned a one, symbolizing them as an area where sand mine construction would be unlikely. 

The second dataset used was the digital elevation model for the county. ESRI’s slope calculator tool was used to identify the slope of all areas within the county. The block Statistics tool was used to generalize this output before it was reclassified to put a high suitability number (3) to the most flat areas with a low number for areas with steep slopes. Classifications were selected using an equal interval break system; this classification method was used for all following operations unless specified in the classification table or writing.  

As freshly mined sand needs to be washed early on in its processing, a mine located in areas that have a high water table have a cost advantage over mines that would have to use an extensive pumping system to get water. Water table data was used find areas with a high water table. The data downloaded for this task would be seen by some as historical but for this project its temporal accuracy is not of a huge concern. The data was supplied as a topographic file meaning that it needed to be converted to a raster file. From here the same processes used on the earlier raster files were applied to reclassify each pixel to the most suitable water levels.  

Information pulled from the National Land Cover Database (NLCD) was used in two unique ways. As with the other rasters it was reclassified to display the areas most suitable for sand mining. It was also reclassified using the integers 1 and 0. This was so it can be used later to exclude unsuitable areas from the final output raster altogether. 

Figure Two. Euclidean Distance model used to create a distance raster from rail depots near Trempealeau County. 

The final dataset used was a point feature class vector file featuring the locations of rail depots around Trempealeau County. Due to software issues a separate model (Figure Two) was created to complete this operation. The output Euclidian distance file was then reclassified in the suitability model (Figure One) following the same classification conventions as the other vectors.

Figure Three. NLCD data was reclassified to determine which areas mine can and cannot be built. 

The final operation used to create a mine suitability raster was ESRI’s raster calculator tool. As its name suggests, this tool creates a new raster file from operations performed on input raster files. In this particular case the tool was used to add all coorisponding pixel values of the previous rasters together. The second NLCD raster (Figure Three) was multiplied to this equation. All pixels that fell within the correct land cover areas would retain their values (Multiplied by one) while pixels in unsuitable areas would be multiplied to zero, making them easy to identify. 

Sand Mine Impact Model

The second component needed to assess the viability of any sand mine location is to identify areas which a mine would have detrimental effects to surrounding and nearby areas. Much like the first suitability model, this impact model (Figure Four) uses many of the same processes and data sources to create a final output impact raster file.

Figure Four. Mine impact areas model. 

The impact model uses three separate vector data sets pulled from the Trempealeau County geodatabase and the NLCD raster file. Like the operations in the previous model the NLCD raster file was reclassified (to show areas with are populated) and had a Euclidean distance tool run to determine how far sand mines need to be from inhabited areas. Unlike the previous raster distance operations this reclassify operation was not divided into equal distance class breaks as a minimum distance of 640 meters has been prescribed to keep mine noise pollution from being overwhelming to people living near mines. The second class break exists at 2.5 kilometers with the third encompassing all remaining data. An inhabited area raster file containing only the 640 meter noise control zone was also created from this data set, as it was later used to exclude areas suitable for mines.

As mentioned before, the other three datasets used for this impact model are vector files. In order to even begin manipulating them several operations needed to be completed, the first of which was making sure they were properly projected. One functional advantage to databases is the ability to create a feature data set (Labeled as PRJ in Figure Four). This is essentially a subfolder stored within a geodatabase. All the files it contains are stored with the same user specified projection. Meaning that any file imported to this feature dataset will automatically be projected. The Feature Class to Feature Class tool was used to accomplish this task and more, as a key asset to this tools use is the built in SQL selection tool. This allows the output file to only contain the features that a user wants it to. An SQL statement was used for both Steams to select streams of a higher order than four and wetlands, as only year round wetlands and areas of open water are of interest to this project.  

From this point forward different operations were performed on all the vector files in the feature data set. ESRI’s feature to raster tool was used on Streams (this operation was not entirely necessary). This was followed by the Euclidean distance tool, extraction by mask and finally the reclassification tool to create a three category raster file symbolized to show High (3) Medium (2) and Low (1) risk for areas in Trempealeau County.

The wetlands dataset merely had to have the Euclidean distance tool and reclassify operations performed on it to produce a raster to be used in the final raster calculation tool. Three class equal interval classification method was used for the reclassify function. Prime farmland was reclassified to associate a high risk number with areas are able to yield a vast amount of agricultural product.

Once all these intermediate raster files were created a single raster calculator tool was inserted to the end of the model. In short the tool added up all the output raster files and multiplied by the 640 meter noise pollution exclusion zone around inhabited areas.

Suitability and impact raster overlay

Figure Five. Suitability and impact overlay model used to determine the best locations for sand mines while considering factors such as resource availability and human/environmental impact. 

To bring together the results from the previous two models a third model was created. There are two major process branches in this model which combine to create a final product. The first branch uses the raster calculator tool to simple subtract the impact raster from the suitability raster. This creates a decent representation of areas that are better suited for mining than others. Unfortunately, alone it does not account for areas where sand mining is absolutely not allowed. The second branch compensates for this calculating and reclassifying the two exclusion rasters from the previous two models to create a raster containing only areas which mines are allowed. The final operation in the model extracts these allowed areas form the first branch’s combined impact and suitability raster. The final output raster only includes areas where mining is allowed, which, in turn, effectively solves above mentioned problem. 

Results

Figure Six. Major output files from the suitability model. red indicates ares of low suitability while green indicates areas of high suitability. 

Figure Seven. Output raster from mine suitability model showing areas that would work well for mine construction (green) and highlighting areas to be avoided (red). 

Figure Eight. Major output files from mine impact model. Level three areas (red) indicate places which sand mine construction and operation are detrimental to human inhabitants and to specific environmental factors. Level one areas (green) represent areas which have a relatively small impact to these factors.   

Figure Nine. Final Output raster from impact model. Higher values represent areas were a sand mine would be rather detrimental.  

Figure Ten. Final sand mine ideal locations created by combining results from both the impact and suitability raster models. The higher values (green) represent areas that are best for sand mines. lower values (in red) are locations where the construction of a sand mine would not make sense or where they are not allowed to be built. 

Conclusion

From the suitability model it was determined that the areas nearing the centers of major watersheds are best suited for sand mine construction, not only are the proper resources (sand) in abundance but the flattish slopes, land cover types, and high water table levels associated with the areas make the construction of sand mines quite promising. 

The final raster output from the impact model happens to somewhat contradict the suitability model raster. as many of the areas which a sand mine would have the largest detrimental effects are in fact the same types of places which the suitability model suggests are best for a sand mine to exist. Watersheds collect waste laden water meaning that mines should not be located near high order streams. Wetlands are also susceptible to harmful sediments from mines. Mines are also not allowed to be constructed near areas of human habitation so as to prevent excessive noise pollution. These factor combine to really reduce the area which mines are best suited in the county. 

As was mentioned in the first paragraph of this study it seems as if factors contributing to environmental sustainability are in fact at ends with the sand mining industry. Luckily there do appear to be some areas in Trempealeau County in which sand mines are able to operate without many negative consequences. Most of these areas are still within major watersheds but due to negative impact factors a great deal of prime sand mining locations have been excluded.