Exercise 6: Geocoding Addresses

Goals and objectives
During this assignment I practiced geocoding addresses gathered by my professor from the WDNR. I was personally given 21 sand mine addresses. During this lab, I practiced the steps that are typically taking when geocoding addresses. For example, I was required to normalize my data table of addresses, geocode them using ArcMap, and assessed the error between my geocoded addresses, my classmates geocoded address, and the addresses actual location. The precise methods, and results of my lab are documented below.

Methods
I began this lab by locating a document of sand mine addresses produced my the DNR. Of these, I was designated 21 of the address. In order to geocode the addresses, I was required created a normalized excel data table (Figure 1). I created new columns for the following information: Unique Mine ID, PLSS code, Street Address, City, State, Zip Code, and any additional addresses. This process creates a standard for how the data is entered and organized.

After my table was normalized, I imported it into ArcMap. From here I turned on the geocoding tool. With the geocoding tool, I used two different methods to locate my addresses.

If an address was provided in the table, the geocoding tool located the correct location based on the address, city, state, and zip code provided. I verified these locations using the ERSI base imagery and Google maps imagery. Majority of this points were initial accurate, however a few required minor adjustments to place the address point at the entrance along the main road.

If an address was not provided in the table, I was required to locate the address point myself. When this occurred, I used the PLSS code to locate the address. The geocoding tool automatically defaulted these points to be located at the city's center, which was a highly inaccurate location. The PLSS is a four number code that signifies the Quarter Plot, Section, Township, and Region of the address, respectively. To find the location, I brought in each of these data layers from the WDNR database. The code is read from right to left, therefore I first located the Region of the code, followed by the Township, Section, and Quarter Plot. After I narrowed the address to the Quarter plot section, similar to the other method, I placed the address point at the entrance along the main road.

Once I plotted all addresses, I located my classmates' geocoded addresses and brought the shapefiles into my ArcMap document. From here I merged all of my classmates' shapefiles together. The output resulted in a feature class that contained all of their geocoded mines. From here, I ran the Near Proximity ArcTool to find the distance between my geocoded addresses and the my classmates geocoded addresses that shared the same Unique_ Mine_ID, therefore being the same mine. This output table result can be seen in Figure 2. This Near Proximity ArcTool method was repeated for the actual location of the address points, provided by the WDNR. Once again I compared my address locations with WDNR's geocoded addresses that shared the same Unique_ Mine_ID. This output table result can be seen in Figure 3.

Results

Below are my outputs and tables created throughout the lab.

Figure 1: My final normalized table including 21 sand mine addresses provided by the DNR



Figure 2: This is the Near Proximity ArcTool output, displaying the distance between my geocoded mines and my classmates' geocoded mines of the same Mine_Unique ID. This was ran to assess error between to two points, based on their distance apart.

Figure 3: This is the Near Proximity ArcTool output, displaying the distance between my geocoded mines and WDNR actual location of mines with the same Mine_Unique ID. This was ran to assess error between to two points, based on their distance apart.

Figure 4: Below is my map displaying my geocoded addresses, their actual locations, and my classmates geocoded mines. These points represent the location of 21 mines within Wisconsin.

Discussion
There are several types of errors that can occur during geocoding. According to Lo Chapter 4, there are three sources where errors may originate. These are from the original map, during data automation, or during the data processing and analysis stage. Within these, errors my begin at the earliest stage of setting your map projection and scale, or during image selection. However, the possibly of accruing errors continues during processes such as attribute input. Additionally, errors can arise while interpolation the data, during data classification, or but incorrectly rounding number values. The complete list of errors discussed by Lo can be seen below in figure 5.

Figure 5: Lo chapter 4 table that lists errors that may arise while analyzing geographic data.

Knowing there are an abundance of errors that may arise throughout the geocoding process makes it difficult to determine which of your geocoded points are actuate or not. In this lab, we practiced ways of eliminating error and assessing error while geocoding. For example, data that is normalized before it is geocoded should theoretically be more accurate than if than data that is not initially normalized because it would lacked an initial process of standardization that causes data format errors. Another way to increase a data point's accuracy is to compare the geocoded locations in ArcMap with their location through other sources such as Google maps. For example in our lab, mines that were visible in Google maps were most likely geocoded more accurately than mines that weren't visible because their address was able to be reaffirmed.  However, even using Google Maps to located the mine, it is possible the geocoded location are still inaccurate. In order to compare which points are most accurate, an error assessment can be ran. In this lab, it was used to compare distances between my personal geocoded addresses and the same address geocoded by a creditable source, such as the DNR. For example, as seen in figure 3, the Near Point Tool will show which points had the smallest distances to the DNR's location. In this case, the smaller distances to the DNR's mine suggests a smaller error margin exists, and therefore we assume these points are more accurate.

Conclusion 
This lab shows there is a great chance that error and variation will occur between the Ersi Geocoding tool users. This is due to the fact that errors mar arise through out the geocoding process, such as during data entry, placement of address point, or during error analysis. Also, it is likely the users are basing the address locations on entities that are subject to change, such as roads and entrances since within images. Setting data entry and overall geocoding standizations will help to minimize this error rate.