Geocoding

Goals and Objectives

The goal of this exercise was to learn the process of obtaining data in the form of a spreadsheet and geocoding that data in ArcMap. We were to download an excel file that contained the locations of frac sand mines in Wisconsin from wisconsinwatch.org and display the addresses on our map.

The data was not able to be geocoded right away. The addresses were in different forms; some were street addresses, some PLSS locations, and some just had directions like "west of Augusta." Our objective was to map these locations as accuratly as we could.

Methods


Figure 1- Data table prior to being normalized.

Because of the inconsistancy of the data, I first needed to normalize the table so that each location had a street address, city, county, and state in seperate fields. Figure 1 shows how the data appeared before I worked on normalizing it. When running the geocoding tool using this table, almost no locations were matched. The ones that did match were mostly wrong.

Figure 2 shows a sample of my data after normalization. By using a combination of aerial photo interpretation, websites referencing specific mines, Google Earth, and mapquest.com, I was able to come up with a street address for each location.

Figure 2 - Normalized addresses.

When I ran the geocoding tool using this new normalized table, I had much greater success. All 14 of my mines were "matched," however some of them were in the wrong locations. For those, I simply used the "pick address from map" option and moved the point.



Figure 3 - Map showing my geocoded mines and those done by classmates

Results


Each mine had its own "unique ID" so that we could keep track of them. Classmates of mine had geocoded some of the same mines that I had. By merging all of the classes geocoded mine shapefiles together, then running a query and selecting those mines that had the same unique IDs that I had used, I created a layer that was made up of mines that should have been the same addresses as the ones I had geocoded. Just by glancing at figure 3, it's easy to see that simple errors can cause drastically different outcomes.

Figure 4 - Distance (in meters) between points.

Figure 4 is a table that shows the distance between my mines and the closest mine to it. The problem with this is that there is no way to get the tool to calculate the distance between my mine and the ones that have the same unique ID. We have to assume that the distance shown in the table to the closest mine is the same mine. Usually, this is probably the case. In some situations, however, a mine could have been geocoded to as far away as a different county.

Discussion

Some of these errors could be inherent, such as issues when changing the projection of a shapefile, but most of the discrepancies are operational errors. Operational errors occur when managing and processing the data. Inputting the wrong address, misinterpreting data, or selecting the wrong location from aerial photography are all situations that could lead to the sizable distance between some of these mines (Lo and Yeung, 2006). The only way in which we can know for sure what points are correct would be to contact the owner or manager of the mine, or visit the facility personally.

Conclusion

During this lab, I have learned the different ways to geocode, as well as the importance of normalization. I also realized how big of a difference seemingly small mistakes can make. Calculating point-distance is an interesting tool that will come in useful in the future, which I did not even know existed before doing this lab.

References

Lo, C.P., Yeung, A.K.W., (2003), Concepts and Techniques in Geographic Information Systems. (pp. 107-108). Pearson Prentice Hall.