Post 6: Raster Analysis

Goals and Objectives: 

The goal of this exercise was to use raster geoprocessing tools to create models for frac sand mining suitability and environmental and cultural risk in Trempealeau County. The results could then be overlayed to find the most desirable locations for frac sand mining with the lest impact.

Background:

Raster analysis is valuable for multiple reasons. Calculating rasters is the same as intersecting suitable vector layers. The difference with using rasters though is that you can add rasters to determine a ranking of suitability. This can be more helpful than a simple yes or no answer received from intersecting vector files. In addition, using rasters for distance analysis is more efficient than using vectors because rasters can store continuous distance information, while vectors only store discrete information.

Data Sources: 

• Wisconsin Geological & Natural History Survey
• U.S Department of Transportation
• USGS National Map
• Landcover
• Digital Elevation Model
• U.S. Department of Agriculture
• Trempealeau County Land Records
• USDA NRCS Web Soil Survey

Methods:

The entire area of Trempealeau County was too big for the raster analysis, so the southern portion of the county was used as the study area. A mask of the study area was set as the processing extent in the environmental settings.

Raster tools used include Raster Calculator, Reclassify, Euclidean Distance, Slope, Viewshed, and Focal Statistics. The equal interval classification was used for many reclasses.

Data flow models were used to create all rasters (Figures 1-3). The models are shown below:

Figure 1: The data flow model for the suitability model. Individual factors were created and classified. The factors were then combined together to create an overall suitability model.

Figure 2: The data flow model for the risk model. Individual factors were created and classified. The factors were then combined together to create an overall risk model.

Figure 3: The data flow model for creating the combined suitability/risk model. The risk model raster was multiplied by -1 so it would properly combine with the positive suitability risk raster. Both a regular and weighted suitability/risk model were created in the model. Also note the viewshed tool was run in this model to determine how visible a mine would be to a bike trial running through Perrot State Park.

Suitability rankings were assigned on a 1-3 scale, with 1=Least Suitable and 3=Most Suitable. Reasons for the rankings are discussed in the Excel table below (Figure 4). Risk rankings were assigned on a 1-3 scale, with 1=Least Risk and 3=Most Risk. Reasons for the rankings are discussed in the Excel table below (Figure 4).

Figure 4: The table shows how factors were classified and given ranks for both the suitability and risk models.

Sand Mining Suitability Model:

Criteria used in the suitability model included:

• Geology 
• Land Use Land Cover
• Distance to Rail Terminals
• Slope
• Water Table Elevation (feet)

Specific methods for objectives in the suitability model are listed below:

Objective 4: The Slope tool was used to calculate slope based on percent rise. Raster Calculator was used to convert the DEM, which was in feet, to meters by multiplying all cells by 0.3048. The slope values were averaged using Focal Statistics, and then reclassed using the Reclassify tool based on the following parameters:

Objective 5: Depth to water table data was accessed online from the Wisconsin Geological & Natural History Survey (Wisconsin Geological & Natural History Survey, 2015).

Objectives 6 and 7: Raster Calculator was used to add all suitability classes together. This resulting in combined rankings of all factors. The factors were multiplied by the unsuitable NLCD feature class to exclude land unsuitable for mining.

Objective 14: The viewshed tool was used to determine which locations in the county were visible from a bike trail that runs through Perrort State Park. Results were ranked added to the suitability model.

Sand Mining Risk Model:

Criteria used in the risk model included: 

• Impact to Streams
• Impact to Prime Farmland
• Impact to Residential Areas (Noise shed/Dust shed)
• Impact to Schools
• Impact to Variable of your Choice (Wildlife Areas)

Specific methods for objectives in the risk model are listed below:

Objective 8: I used streams with a ranking between 3 and 6 because they included 23% of all streams and their primary designation was Primary Flow Over Land Perennial, which means it is a river that flows more than two years. Streams ranked 2-6 included streams designations of Primary Flow Over Land Intermittent, which meant the rivers existed only some of the time. Therefore, streams ranked 2 or lower were excluded.

Objective 10: The Zoning feature class was chosen to calculate distance from residential areas. Advantages of this feature class include it had land use types, which were helpful in distinguishing were people lived. Census data would have been hard to determine congregated areas where people live. A disadvantage of using the zoning feature class was that sometimes it was hard to determine which classes were residential. For example, I used my best educated guess and determined the Incorporated class could be considered residential.

Objective 12: Wildlife Areas were chosen as risk factor because mines should not encroach upon wildlife areas.

Python: Python scripting was used to weight the most important risk factor, impact to residential areas, by 1.5 (Figure 5). The results were added to the risk model, and then used to create the overall suitability/risk model (weighted).

Figure 5: The python script used raster calculator to multiply the most important mining risk factor by a weight of 1.5. The result was added to the other risk factors to create a newly weighted risk model for frac sand mining in Trempealeau County.

Overall Suitability/Risk Model: 

Raster calculator was used to combine the Mine Suitability and Mine Risk models for an overall mining suitability model. Both rasters were composed of other rasters ranked on a 1-3 scale. Raster calculator was used to convert 1-3 values of the Mine Risk model to make the values negative. When raster calculator was used to combine the models. the negative risk numbers would be added to the positive suitability numbers. This resulted in a correct Mining Suitability/Risk model. The same process was used to create a weighted mining suitability/risk model from the weighted risk model.

Results: 

Suitability Model:

Figure 6: The map shows were suitable geological formations (Jordan and Wonewoc formations), are located. All other formations are unsuitable for frac sand mining.

Figure 7: The map shows suitable land covers for mining. The more suitable locations are undeveloped and have little vegetation in the way. It should be not unsuitable land cover is listed on the map, which indicated areas populated by humans.

Figure 8: The map shows which areas of land are closer to the one rail terminal in the study area. Frac sand mines need to be close to rail terminals to ship the sand.

Figure 9: Percent slope was mapped to find areas with the lowest slope, which is most desirable for mining.

Figure 10: Water table elevations (feet) were mapped to determine where the water table was closet to the surface (most desirable).

Figure 11: All suitability factors were combined to create an overall suitability model for frac sand mining. Higher rankings are more suitable for mining.

Risk Model:

Figure 12: The map shows locations's proximity to streams. Mines closer to streams have a greater impact. 

Figure 13: The map shows where prime farmland is located. Mines on prime farmland would have a big impact.

Figure 14: The map shows proximity to residential areas. Mines closet to residential areas would have the greatest impact. Additionally, mines cannot be located within 640m of residential areas, so NoData was designated for any areas closer than 640m.

Figure 15: The map shows proximity to schools. Mines closer to schools have the greatest impact. Additionally, mines cannot be located within 640m of schools, so NoData was designated for any areas closer than 640m.

Figure 16: The map shows proximity to wildlife areas. Mines located closest to these ares would have the greatest impact.

Figure 17: All risk factors were combined to create an overall risk model for frac sand mining. Higher rankings are pose greater risk for mining.

Figure 18: The most important risk factor, impact to residential areas, was weighted by 1.5 using PyScripter. Higher rankings are areas where mines would have a greater impact.

Viewshed: Bike Trail:

Figure 19: The map shows visibility from a bike trail that runs through Perrot State Park. Establishing a mine in blue ares would keep mines out of site of the bike trail, while mining in green or brown areas would be highly visible.

Figure 20: The Bike Viewshed factor was added to the suitability model. Higher ranks are more suitable for mining (18 being the highest rank),

Overall Suitability/Risk Model:

Figure 21: The suitability model and risk model (multiplied by -1) were added to create an overall suitability/risk model for frac sand mining in Trempealeau County. Higher rankings (positive) are places more suited for mining.

Figure 22: This suitability/risk model for frac sand mining incorporated the weighted risk model in its creation. Higher rankings (positive) are places more suited for mining.

Discussion:

Large gaps in the final models can be seen (Figures 21 and 22). These are because mines cannot be within 640m of residential areas and schools, and therefore the areas were assigned NoData values.

Looking at the non-weighted suitability model, the best locations to establish frac sand mines would be in the middle of the study area indicated by spotted areas of teal (Figure 21). These areas are located further away from residential areas and schools, located on suitable geologic formations (Jordan and Wonewoc), and are located closet to the one rail terminal in the study area.

The weighted suitability model shows the bets locations to establish frac sand mines would be in the northwestern part of the study area (Figure 22). This is most likely because the risk factor "impact on residential areas" was weighted by 1.5. This shifted suitable mining areas from the middle of the study area to the northwestern area.

Overall, I believe the best location to establish a frac sand mine in this part of Trempealeau County study area would be in the northwestern part. It is close to the rail terminal, located on suitable geologic formations, and located away from residential and school areas.

Conclusion:

Raster analysis is an important tool for analyzing geographic questions. This exercise utilized many raster analysis tools to create a suitability model, risk model, and combined suitability/risk model for frac sand mining in a portion of Trempealeau County. Results indicate establishing a mine in the northwestern portion of the study area would be best. This knowledge can be used by land management officials to make informed decisions on where to mine frac sand. Knowledge of raster analysis will be very beneficial for my career in the geospatial workforce.

References:

Wisconsin Geological & Natural History Survey (2015). Generalized Water-Table Elevation Map of Trempealeau County, Wisconsin [coverage file]. Retrieved from: https://wgnhs.uwex.edu/pubs/000444/