Geocoding Sand Mines for Hydraulic Fracturing in Western Wisconsin

Goal

In this lab, we were to acquire raw data of current or prospective sand mines in Western Wisconsin, find the location of each mine given the incomplete data, compare your location to locations other students chose, and examine the error.

Methods

In order for the mines to be geocoded in ArcGIS, they would each need to have a complete address, which was not the case for every mine site.  If addresses were not provided, I used imagery from Google Earth and other internet searches to try to find the exact address of each sand mine that is active, in process, or proposed.  The main components of the location that are needed are the facility address, community (city), zip code, and the state.  If this information is not provided, and it cannot be found any other way, a tool can be used in ArcGIS that allows you to pick an address from the map, where you can click your mouse on a map, and it will give an address for you, which is found in the geocoding toolbar.

Figure 1:  Table of the mines I was assigned after the addresses were found.

Once all the information is filled in for each mine, you can match the information to real space in the program.  A window something like the one below will show how closely matched your data is.

Figure 2:  A simplified example of the geocoding table used to match proximity of your data to real space.  

Once there is a 100% match, the geocoding process should be successful.  After this point, each student made a shapefile of all their mines (each student had about 14 mines), making sure that their unique field ID has been preserved through their geocoding process.  From here, all the shapefiles were merged together, not including the shapefile of the mines that I geocoded.  A query can be done to select only the mines that matched the unique field ID of the mines I was assigned.  Once those selected mines are made into their own feature class, the point distance tool can be run the see how close my location for each mine was compared to locations other students chose for the same mine.

Results

Figure 3:  A small scale map of my mine locations compared to other student locations of the same mine.

The point distance tool creates a table that calculates the distance between one of my mines, which is shown as a single point, and the nearest mine that was located by another student.  Each mine is given an input_FID in the table that shows the distances from my mine to everyone else's mine with the same ID.  The highlighted area in the table below represents the distance from my mine location to other students' mine locations.  The last column shows this distance in kilometers.

Figure 3:  This is part of the table that is created from the point distance tool.  The highlighted area shows the distance of my location for that mine compared to where my peers put the same mine.

When looking through the data, I looked within each input_FID for the shortest distance calculated and compiled those in a separate table to see the variety in distances between my mine locations and my peers' mine locations.

Figure 4:  This table shows a summary of the shortest distance between my mine locations and the location that a peer chose for the same mine.

The shortest distance found between two points of the same mine was about 9 kilometers, and the greatest distance between two points of the same mine was about 75 kilometers.

Conclusions

I think the largest source of error here deals with inherent errors of how people geocoded their mines.  They maybe found an address that was incorrect or for a different mine.  If they had to pick an address from the map, it is possible that they were zoomed in to the wrong area, or found the wrong location for that particular mine.  Since there are so many sand mines that are being implemented in Western Wisconsin, it could be easy enough to choose the wrong one if you do not have exact address information (data automation and compilation, Table 4.1, Lo, Chapter 4).  The only operational error I could see happening would be with the point distance tool.  The program only calculates a distance between two points that are closest to each other.  The user does not have any way of knowing which two points it used for this calculation.  In this case, the tool could have calculate a distance between two different mines.  In order to figure this out, you could do a quick draw between the two points that you want to know the difference between, and see if the calculation was close or not.

Citations
Lo (2001).  Data Quality and Data Standards, Ch. 4, 105-108.