Monday, May 1, 2017

Post 6: Raster Modeling

Background
As both the final exercise of the semester and the final part of the continued assessment of sand frac mining in Wisconsin, students were tasked with generating a number of rasters reflecting information on various surface features and utilizing these rasters to develop an index for the suitability and impact, both environmental and community, of potential mine locations in southern Trempealeau County, Wisconsin. The decision was made to only utilize the southern half of the county in order to speed up processing time. In addition, a viewshed would be utilized to see if any of the highly suitable areas are visible from the Black River, located in southern Trempealeau county. This stretch of river is a popular recreational canoe trail, where most individuals who utilize it would prefer to not see a frac sand mine. Thus, this viewshed would allow for the mapping of desirable mine locations visible from the river.
Datasets and Sources

Methods
First, the Trempealeau county shapefile was cut to the size of the study area by utilizing the clip tool and the Watershed Raster. Then, the newly shaped county shapefile was set as the mask in the environmental setting and the cell size was set to 30 to match the resolution of the raster data.
In order to properly generate suitability and impact indexes, each generated raster was eventually reclassified in levels of suitability and impact, with the higher the rank being the greater the impact or suitability. As rasters were generated and reclassified, their classifications were added to a table for more effective reference (Table 1).
Model
Factor or Feature
Assigned Suitability or Impact Value
Variable
Reason for Classification
Suitability
Geology
1
Ew, Ej
Jorden and Wonewoc formations are the most suitable for frac sand mining.
0
Ee, Et, Op, Oa, Em
Other types of geological formations are not suitable for frac sand mining.
Land Cover
0
Open Water, Developed, Low/Medium/High
Largely impossible to construct a sand mine due to water or development.
1
Emergent Herbaceous Wetlands, Woody Wetlands
The water element of wetlands make building a mine difficult.
2
Mixed Forest, Deciduous Forest, Evergreen Forest
A forest takes some effort to clear, but not as much as drying up a wetland.
3
Barren Land, Developed Open, Shrub/Scrub, Hay/Pasture, Cultivated Crops
These areas are either largely clear of vegetation and features or the vegetation is easily cleared.
Distance to Rail Terminal (m)
3
0 - 12198.032364
Closest distance to tail terminals as determined by natural breaks (jenks). Ideal distance.
2
12198.032364 - 20453.266388
Medium distance to rail terminal as determined by natural breaks (jenks). Manageable distance.
1
20453.266388 - 31542.386
Farthest distance to rail terminal as determined by natural breaks (jenks). Unideal distance.
Slope
1
Steep
Steep slope. Greater than a 13.638795% gradient.
2
Shallow
Shallow slope. Less than 13.638795% gradient.
Groundwater Depth (m)
3
187.790696 - 227.867689
Mines require access to water to operate. Thus, the least distance to the water table is the most ideal
2
227.867689 - 263.086866
Medium distance to the water table. Operable, but not ideal.
1
263.086866 - 343.240854
Farthest distance to the water table. Almost nonfuntioonal
Table 1: A table of each factor used for the suitability index and each rank assigned in each raster reclassify. In addition, a brief explanation is given for the assigned value
First, the TMP_geology feature class was clipped to the county size and re-projected into the NAD 1983 State Plane Wisconsin Central FIPS 4802 projected coordinate system, the chosen coordinate system for this project due to its area of focus and use of linear meters as a unit of measurement. As these steps were performed on most rasters during this exercise, these two processes will be abbreviated to "clipped and re-projected" to limit redundancy. It was then converted into a raster and reclassed into 1 and 0, with 1 being the desired geologic formations of Ew and Ej, with 0 being everything else. Based on prior research and the use of a reference map, the Jorden and Wonewoc formations (Ew and Ej) were determined to be the most favorable formations for frac mining
Utilizing the NLCD, a land-cover raster was created for the study which was clipped and re-projected. In addition, the raster was reclassified into two separate rasters, one that ranked each type of landcover from 0 to 3 depending on how suited it was for clearing for mining, and another that ranked each landcover as 0 or 1, with 0 denoting landcover that was virtually impossible for sand mining. This included thing like open water and populated areas where people likely wouldn't leave for a mine.
Utilizing the rail terminals selection created previously in another exercise, a euclidean distance raster was generated showing the the distance areas were from available rail terminals. In order to utilize the mines located outside the county, the extent was adjusted to a 15 mile buffer around the county study area. The euclidean distance raster was reclassified into three classes ranging from 1 to 3 based on how close an area was to a rail terminal. The division of these classes was based on a three part natural break classification, as it was felt that this best divided the county study area.
The DEM of the the county was first clipped and re-projected to match the study area. Then a slope tool was utilized to generated gradient % data for the county. To avoid a salt and pepper effect on the data, this raster was put through an average tool to average out the values based on a 3 by 3 cell filter. This was then reclassed into a two classes, steep and shallow, as determined by natural breaks. The shallow slopes were reclassed as 2, as mines require a shallow slope to properly operate, while the steep slopes were reclassed as 1, to show how they were unideal for use in mining.
Additionally, a water table contour map was utilized to generate a water table elevation raster which was then reclassified into three classes, ranked 1, 2, and 3, with the shallowest distance to the water table being assigned 3 and the furthest distances being assigned 1. This is because mines require easy
access to the water table to operate. Thus, the areas with the shallowest depth to a water table are the most ideal and received the highest ranking. This reclassification was based on a natural breaks classification.
Figure 1: A model showing the rasters and tools used
to create the final suitability raster model for frac sand
mines in Trempealeau county.
Once this was complete, the geology, landcover suitability, rail terminal euclidean distance, slope, and groundwater depth rasters were added together via raster calculator to create a final suitability raster, which was multiplied by the landcover exclusions raster to assign the unusable areas a value of 0 (Figure 1).Once this was completed, the individual recalculated rasters were used to create a map showing the suitability of Trempealeau county based on each factor and the final total suitability (Figure 2).







Figure 2: A map showing the reclassified suitability for sand frac mines in southern Trempealeau county based on geology, landcover, landcover  exclusions, distance to rail terminals, gradient, water table depth, and combined suitability. 






























Once this was complete, a second series of rasters needed to be completed in order to generate an environmental and community impact index which would be utilized in the results. Similarly to the suitability index, factors would be converted to raster and reclassified as either 1, 2, or 3 depending on their potential impact to the surrounding area (Table 2). Differently  from the suitability index, a rating of a 3 indicates a high impact and an area ill-suited for a mine.
Model
Factor or Feature
Assigned Suitability or Impact Value
Variable
Reason for Classification
Impact
Distance to Perennial Stream (m)
1
1922.031878 - 4712.674316
The farthest distance from perennial streams with the least impact.
2
628.356576 - 1922.031878
Medium distance to perennial streams. Moderate impact.
3
0 - 628.356576
Closest distance to perennial streams with the greatest impact.
Distance from Prime Farmlands (m)
1
1187.046272 - 3519.730225
Farthest distance from prime farmland with the smallest possible impact, as a result. Determined by natural breaks.
2
331.268727 - 1187.046272
Moderate distance from prime farmland. Determined by natural breaks
3
0 - 331.268727
Closest distance from prime farmland with the greatest impact as a result. Determined by natural breaks.
Distance from Residential Zone (m)
1
7713.82886 - 17254.617188
Areas farthest away from residential zones. Will have the smallest impact on these areas,
2
640 - 7713.82886
Area just outside the minimum distance a mine can be from a residential area.
3
0 - 640
Within minimum distance a mine cannot be from a residential area based on zoning law.
Distance from Schools (m)
1
6309.463722 - 13296.8037
Farthest distance from school owned parcels with the smallest possible impact.
2
3441.525666 - 6309.463722
Medium distance to school parcels with moderate impact
3
0 - 3441.525666
Closest distance to school parcels with the potential for the greatest impact.
Distance from Wildlife Areas (m)
1
10639.357384 - 21704.289063
Farthest possible distance from wildlife areas with the smallest possible impact on said areas.
2
4766.432108 - 10639.357384
Moderate distance from Wildlife Areas with a mid-level impact.
3
0 - 4766.432108
Locations close to wildlife areas
Table 2: A table of each factor used for the impact index and each rank assigned in each raster reclassify. In addition, a brief explanation is given for the assigned value.
First, a streams feature class was clipped and re-projected fro the project, then the perennial streams were queried out and exported as their own feature class. Perennial streams were determined to be the most critical to the study as they are running year round and would be most greatly impacted by sand frac mines. With the perennial selection feature class, a euclidean distance raster was generated and reclassed into three classes as determined by natural breaks to show how close areas in Trempealeau county were to perennial streams. The closest areas were reclassified as 3 as mines in these areas would have the greatest impact on perennial streams.
A second euclidean distance raster was generated for the prime farmland of Trempealeau county. This feature class was originally created by querying out the areas of prime farmland from a farmland feature classes. This euclidean distance raster was similarly reclassified as the perennial streams raster by reclassifying areas as 1, 2, or 3 based on how close they were to prime farmland, as decided by natural breaks and with 3 being the closest.
In order to generate a euclidean distance raster for, it first had to be decided on where and how to retrieve and interpret residential data. It was eventually decided to utilize a query and the zoning districts feature class to generate a feature class of the zoning districts listed as residential. However, one important note is that the zoning districts classifies many wide areas in the county as the same thing, so pockets of residential areas may have been missed. A euclidean distance raster was generated for the residential zoning districts and reclassified as three classes listed as 1, 2, or 3, depending on how close these areas were to zoning districts and utilizing a natural breaks classification.
A euclidean distance raster was generated for school owned parcels by utilizing a school parcels feature class which had been clipped, queried, and re-projected from an original Trempealeau county parcels feature class. The raster was then reclassified into three classes assigned 1, 2, or 3 based o n distance from the residential areas. However, the classes were partially determined manually and partially by natural breaks. This is because zoning law forbids mines from being built within 640 meters of a residential area. Thus, the closest location (0-640 meters) were labeled as 3 and the remaining two classes were determined by natural break and were labeled 2 and 1, respectively.
The last euclidean distance needed for the risk and impact index was based on a factor of personally choice. It was decided to use wildlife areas, as these are often vastly impacted by nearby mining.
Additionally, habitat restoration is a consideration of mining companies after mining is complete. A
Figure 2: A model utilized to generate the risk assessment and 
impact index for Trempealeau county.
euclidean distance raster was generated from the clipped and re-projected wildlife areas feature class. It was then reclassified into three classes, ranked 1, 2, and 3 based on how close potential mine sites were to wildlife areas, with the three classes being determined by natural breaks.
These impact rasters were added together utilizing a raster calculator in order to build an impact and risk assessment index (Figure 3), which was then displayed as a map along with each individual risk or impact factor (Figure 4).
Figure 4: A map showing the risk assessment and impact index of potential frac sand mines based of several factors and the total impact index of these factors in Trempealeau County.


Results
With the impact and suitability indexes complete, these two rasters were combined into a single raster
Figure 5: The model utilized to combine and reclassify the
suitability and risk assessment indexes into a single raster and
reclassify the result into five classes as based on a equal interval
classification.
by utilizing the raster calculator tool, subtracting the impact from the suitability, and reclassifying the resulting raster to create a final suitability raster (Figure 5). This reclassification was based on an equal interval classification to create five equal classes based on the resulting final suitability ranks assigned to each location. By subtracting the impact from the suitability, the resulting values with the highest scores are considered to have high suitability and low impact, with the lowest values indicating high impact and low suitability. Once reclassification was complete, the final suitability raster was utilized to generate a map of the best and worst locations to build a sand mine, with the five classes being renamed from the resulting values to levels of suitability (ex: high suitability, low impact) (Figure 6).
Figure 6: A map indicating the final suitability and impact of potential areas for sand frac mining in Trempealeau county.
Based on this data, the areas with the highest suitability for frac sand mining are within the the western, northwestern, and southwestern portions of the county area. In contrast, areas of average suitability are within the eastern half of the study. The areas of high suitability can be largely explained by the fact that they are closest to rail terminal, have the shallowest water table, and half the shallowest slope with the greatest distance from wildlife areas. In contrast, landcover type, slope, geology, impact on schools, impact on streams, and impact on farmland seem to have the least noticeable effect on the final suitability model, as these factors seem prevalent and dispersed through the county. What is interesting to note is that areas with poor suitability are largely the same as those areas marked as exclusions due to unusable land in the original suitability and landcover models. This is appropriate, as mines cannot exist in these locations.
In addition, a viewshed raster was generated for the Black River located in southeastern Trempealeau county. This is a popular canoeing river, and many people who recreationally use this rive would not like to see a frac sand mine along or near its banks. This raster was reclassed, with the visible areas being assigned a 1 and the non-visible areas being assigned a 0. Similarly, the final suitability raster was reclassed with the two most suitable classes of areas being assigned a 1 and the three remaining ill-suited areas being assigned a 0. These two rasters were multiplied together to visualize and display the areas suitable for mining which are visible from the Black River (Figure 7).
Figure 7: A map displaying the areas visible from the Black River in Trempealeau county and the suitable frac sand mine locations visible from the Black River. Black River is a popular canoeing river in Trempealeau county, and most residents would be against a mine located near the river.











While a great deal of Trempealeau county is visible from the Black River, much of it is not considered as prime location for frac sand mining. Indeed, the best locations for mining are located farther to the west of the river.
Conclusion
As the final assignment for this course, this exercise sought to generate suitability models for frac sand mining in Trempealeau county, as the final portion of the semester long exploration of frac sand mining in Wisconsin. From these models, it can determined that the primary areas to focus on mining in Trempealeau county are in the western portions of the state and mostly influenced by distance to rail terminal, water table depth, slope, and distance from wildlife areas. Other factors either impact most of the county and thus are negligible to the total model or rule out portions of the county where mining cannot occur, either developed areas with housing or areas of open water. This data was further analysed in regard to visibility from the Black River, as public opinion on popular recreational areas can influence the decision to build a mine. In addition, this data could be sterilized in further risk assessment and suitability modeling by further altering the data and how it is classified. Furthermore, these models can also be utilized to generate weighted models which emphasize a single factor by utilizing further raster calculations or python scripting.
Sources
Hupy, C. (2017). EX 8: Raster Modeling Part I Raster Analysis. Eau Claire, WI.

National Land Cover Database 2001 (NLCD2001). In Multiresolution Land Characteristics Consortium (MRLC). Retrieved on May 15, 2017 from https://www.mrlc.gov/nlcd01_leg.php

Wisconsin State Cartographer's Office (2013). Viewshed Analysis of Lower Wisconsin. Retrieved from http://sco.wisc.edu/images/stories/download/Final_Portfolio_20130411_Revised.pdf

NTAD railway data from the US Department of Transportation – Bureau of Transportation website 

NLCD 2011 data from the USGS National Map Viewer 

DEMs of Trempealeau country from the same USGS National Map Viewer

Cropland Data Layer from theUSDA Geospatial Data Gateway

Trempealeau County file geodatabase from the Trempealeau County Land Records divisional website

Rail terminals from the Wisconsin DNR
Posted by at

No comments:

Post a Comment

Subscribe to: Post Comments (Atom)