Monday, December 15, 2014

Sand Mining Suitability and Impact, Raster Analysis

- Introduction -

The final exercise in this semester long analysis of frac sand mining in western Wisconsin is to make a suitability and impact model for a subset of Trempealeau County and create a final location index based off these models. Multiple raster layers will be used including geology, land cover, and a digital elevation model (DEM). Reclassifying Euclidean distance away from certain vector features (e.g. streams, wildlife areas, residential areas) allows for further raster interpretation in carrying out this analysis. Raster calculator is used to quickly analysis multiple layers of data, creating index values that relate meaning information on both mining suitability and the inherent environmental/community impact of sand mining activity. The suitability and impact models will create raster data that will then be used to create a final mine location index showing the areas were suitability outweighs impact and vice versa.

- Methods -

 Suitability Model

Five variables were taken into consideration for the suitability model:
1) Geology
2) Land cover
3) Proximity to rail terminals
4) Slope
5) Water table depth

Each variable needed to be in raster format and given an index value ranging between 1-3. For suitability, 1 indicates low suitability and 3 indicates high suitability. Each raster created was given a scale of 30 meters. Table 1 gives each layer, its attributes, ranks, and reasons for the ranks for both the suitability and impact models. Model 1 shows the workflow for the suitability model.

Model 1: Workflow of the suitability model

Image 1: Geology bedrock vector polygons on the left. The
raster created from the vector data in the middle. The final
 reclassified and clipped raster used in analysis on the right.
The bedrock geology data was obtained as vector polygons and converted into a raster preserving the geologic names. This geology raster was then reclassified with a value of 3 given to the most desirable geologic formations and a value of 1 given to the rest.

Using a National Land Cover Dataset, the land covers with the least overburden removal were given a ranking of 3 and those with more were given rankings of 2 and 1. Some land covers were excluded from the suitability analysis including urban, agriculture, wetland, and water because mining would most likely not happen in these areas or the land cover is just unsuitable. The excluded land covers were given a value of 0 in a separate raster to remove the areas from the final product.

A reclassification of Euclidean distance away from rail terminals gave a raster of suitability dependent on how far away a location would be from a sand unloading point for trucking. The ranks were decided based on an equal interval classification that covered the entire study area with areas closer to the rail terminal having a higher rank of suitability (less distance for trucks to drive).

Slope was calculated from a DEM of the area, making sure the x, y, and z coordinates all had the same units. The slope raster was then averaged with a 3x3 window to remove the salt-and-pepper effect throughout the image. The ranks were chosen based on the assumption that mines would prefer to locate in areas with low slope rather than steep slope.

Water table contours were used to create a raster of water table depth using the 'topo to raster' tool. Several sand mining processes involve the use of water and ranks were assigned on an equal interval basis with lower water table depth receiving a higher rank.

With the five variables converted to raster data and ranked by suitability, they were added together with 'raster calculator' and then multiplied by the land cover exclusion raster to create a final suitability raster. This suitability raster shows areas of varying suitability throughout the study area as well as areas of exclusion based on the land use/land cover distinction made earlier. The values of the final suitability raster after calculation ranged from 5 to 14 and were reclassified to 1 - 10, with 0 being excluded land.

Impact Model

Five variables were taken into consideration for the impact model:
1) Prime Farmland
2) Proximity to streams
3) Proximity to populated areas
4) Proximity to schools
5) Proximity to wildlife areas

Each variable needed to be in raster format and given an index value ranging between 1-3. For environmental and community impact, 1 indicates low impact and 3 indicates high impact. Each raster created was given a scale of 30 meters. Table 1 gives each layer, its attributes, ranks, and reasons for the ranks for both the suitability and impact models. Model 2 shows the workflow for the impact model.

Model 2: Workflow of the impact model

Prime farmland was obtained as vector polygons and converted to a raster. The ranks were determined by an attribute field loosely indicating 'prime', 'not prime', or 'prime if ...'. Using this attribute, prime farmland was given a ranking of 3, prime if drained was given a ranking of 2, and not prime or prime if drained and other additional steps were given a ranking of 1.

Image X: Euclidean distance raster on the left and the
reclassification on the right, with schools data shown.
For each proximity variable, a distance of 2,100 feet or about 640 meters was used for each rank as this distance was designated a noise/dust shed for this analysis. Sand mines will often create noise and dust while in operation and these factors are often a source of complaint with local residents. Areas that fell within 2,100ft were given a ranking of 3 for likely direct noise or dust impact, areas that fell outside of 4,200ft were given a ranking of 1 for unlikely direct noise or dust impact, and areas in-between were given a ranking of 2.

The streams data available for this analysis covered the entire study area and therefore only streams with the four largest stream orders were chosen as they are the largest and potentially the most meaningful. Residential, commercial, and institutional zoning data was used to designated populated areas.

With each variable converted into raster format, 'raster calculator' was used to add the layers together and create a final impact raster. This final impact raster shows areas were the environmental and/or community impact is high or low. The values of the final impact raster ranged from 5 to 13 and were reclassified to 1 - 9.

Table 1: The layers, ranks, and reasoning behind the suitability
and impact models.


Mine Location Index

A mine location raster was then created by subtracting the impact raster from the suitability raster with 'raster calculator.' The values of this mine location index range from -6 to 9. Higher positive values indicate areas were the overall suitability of sand mining outweighs the potential impact. Higher negative values indicated areas were the overall potential impact outweighs the suitability of the land for sand mining.

Model 3: The workflow of combining the results of each model
and creating the final mine location index


- Results -


Model 4: The overall workflow of the entire process

Map 1: The suitability, impact, and mine location index rasters

- Discussion -

Ideally, to have the least environmental and community impact, new sand mining operations would locate in areas of positive value in the mine location index. As stated earlier, positive values indicate areas were the suitability of the land is greater than the impact. When interpreting the index, green colors are the best, yellows are OK, and orange/reds are the worst. Based on the variables chosen, it seems most of the best area (green) is located in the northwestern portion of the study area. The areas in orange and red located in the south and west of the study area are usually around urban, residential, or commercial areas. Any mine that begins operation in an area of orange/yellow, orange/red, or red is running the risk of negatively impacting their immediate surroundings which could cause distress to local residents and unwanted negative attention for the mine itself.

Most of the desirable area for a sand mining operation seems to fall in less populated areas which makes sense given the variables chosen for the impact analysis. Even if suitability is high and impact is low (index is high, in the green) this does not mean that no negative impact will come from the mining activity. It really means that less people are likely to witness the impacts. Three of the five variables for impact directly dealt with humans. Although, distance to wildlife areas and streams were taken into consideration so the model does account for some environmental impact regardless of human conciseness.

- Conclusion -

 To analysis sand mining in western Wisconsin, both mining suitability and environmental/community impact was explored. Suitability involved using geology, land cover, rail terminals, slope, and water table depth data to locate the best areas for a sand mining operation. Impact involved using streams, farmland, residential/commercial zoning, school locations, and wildlife area data to estimate potential impacts a sand mining operation could have on its surroundings. Combining these analyses together created a final mine location index which shows which areas should be used as mining locations based on maximum mining efficiency and minimal environmental and community impact.

- Sources -

The Office of the Assistant Secretary for Research and Technology/Bureau of Transportation Statistics (BTS) National Transportation Atlas Databases (NTAD) 2014. Rail terminals feature class. Accessed via web: http://www.rita.dot.gov/bts/sites/rita.dot.gov.bts/files/subject_areas/geographic_information_services/index.html

Trempealeau County Land Records. Geodatabase. Accessed via web: http://www.tremplocounty.com/landrecords/

United States Geological Survey, National Map Viewer. National Land Cover Database (NLCD) raster and DEM. Accessed via web: http://nationalmap.gov/viewer.html

Wisconsin Geological and Natural History Survey (WGNHS). Water table contours. Accessed via web: http://wgnhs.uwex.edu/maps-data/gis-data/

Wisconsin Geological and Natural History Survey (WGNHS). B.A. Brown, 1988. Bedrock Geology of Wisconsin, West-Central Sheet, WGNHS Map 104. Digitizing by Beatriz Viseu Linhares, University of Wisconsin Eau Claire (UWEC). Accessed through UWEC Geography and Anthropology Departmental resources.