Module 1.1 - Fundamentals - Calculating Metrics for Spatial Data Quality Lab

If there is one piece of knowledge I have gained thus far in my GIS journey it is that spatial analysis is only as good as the quality of the data. In this module, I learned new ways to test data quality with accuracy and precision assessments. I also learned how to calculate error metrics from spatial data via two main methods. Each of these outcomes were exercised in two parts in ArcGIS Pro that I will go over in more detail.

Part A;

    Part A of the lab assignment dealt with horizontal and positional accuracy. Before dealing with any data, I first made sure to understand the true difference between precision and accuracy. Per the International Organization for Standardization's (ISO), accuracy refers to how close a test result is to an accepted reference value and precision refers to how close tested results are to each other. In this exercise I learned how to apply this to test for vertical and horizontal accuracy and precision. I was provided a shapefile of points taken from a GPS mapping unit and another shapefile with a point to serve as an accepted reference of actual location.  

    Horizontal and Vertical Precision

        I began by recording the average X and Y positional value of the test points. I created a point with the average X and Y value as its coordinates and added it to the map. I then projected them to a coordinate system with linear units instead of decimal degrees. Next, I created 3 buffers around the average test point at 1, 2, and 5 meters using the Multi-Ring Buffer tool in ArcGIS Pro. These buffers served to find a distance from the average test point that corresponded to a specific percentage of observed test points. The goal here was to find a buffer size that accounted for 50, 68, or 95% of the observations. To do so, I created a Spatial Join between the test points and the average location to measure the distance of each waypoint from the average. With this information, I used indexing to find the correct buffer distance for each percentile. The most common measure of precision is where 68% of the observations fall from the average, which was determined to be 5.4 meters. Generally speaking, the higher this number, the lower the precision. However, this is somewhat on par for consumer GPS devices. Vertical precision was found by first finding the average elevation from the spatially joined shapefile I had created earlier. I added a field to the attributes of this file and calculating the field for the absolute difference from the average for all of the test points. This was determined to be 5.71 meters.

    Horizontal and Vertical Accuracy

        

        This process was far simpler than it was for precision. With the reference point, I tested for horizontal accuracy by using the Measure tool to record the distance between the reference point and the average point created earlier. I calculated the absolute difference in elevation between these points for vertical accuracy. I recorded the result to be horizontally accurate within 6.4 meters and vertically accurate within 5.9 meters. Again, this value is relatively high while being somewhat consistent with the expectations from consumer GPS devices. 

    Layout of Project Used for Assessments

Part B: 

This portion of the lab assignment I learned how to easily calculate the RMSE and CDF statistics for a set of XY data. I used Excel to calculate these values as well as other common error metrics. It was good experience to make these calculations and observe their differences both in practice and in representation of the data.