GIS5935 Module 1.1: Fundamentals & Spatial Data Quality

With the dawn of the next season, I'm hoping life will be calmer this Fall semester and that I won't be juggling too much on top of classes. We began the Special Topics in GIS course in mid-August, jumping right in with the first lab of Module 1.1: Fundamentals.

This first module's lab covered Spatial Data Quality in two parts: precision versus accuracy and how to calculate them, as well as root-mean-square error (RMSE) and the cumulative distribution function (CDF). Precision and accuracy are foundational concepts crucial for distinguishing and properly interpreting data. Refreshers like this one are always welcome in my book. Speaking of books, as described by Bolstad and Manson (2022, p. 612), precision is the degree to which values are uniform and tightly clustered around an average; accuracy, on the other hand, is how close the value(s) are to the true benchmark value.

In the first part of the lab, Part A, we used data points near Tampa, Florida, consisting of 50 waypoints (white points) collected from a Garmin GPSMAP 76 unit, and calculated the average waypoint (yellow star) based on the waypoints' different horizontal distances. To represent the 50th, 68th, and 95th percentiles, buffers were created at each of the respective precision estimates. I thoroughly enjoyed making the layout for this map. Despite how simple it could have been, I took a swing at experimenting with this one, letting the map frame circle guide my creative process. It's always great to find new and fun ways to communicate the data! You can see the result of the waypoints' horizontal findings below.

On Target: GPS Waypoint Analysis. An Evaluation of Spatial Accuracy & Precision.
Featured are the 50 waypoints collected from the Garmin GPSMAP 76 device. The white points are the waypoints, the yellow star is the average, and the differently shaded red buffers represent their respective target percentiles.

The horizontal accuracy error, which is the distance from the average waypoint (yellow star) to the true reference point (blue point), was found to be at 3.24m. Meanwhile, the calculated estimated horizontal precision at the 68^th percentile was 4.5m. Because the horizontal accuracy error is below the 4.5m radius, the difference is not significant. Therefore, the true reference point falls well within the primary data cluster, indicating that the waypoint averaging minimized random positional errors.

Distance between the true reference (blue point, northwest) and the average waypoint (yellow star, center).

As for the vertical results, the average elevation (28.54m) compared to the provided true elevation (22.58m) yielded a vertical accuracy error of 5.96m, which is nearly identical to the 68^th percentile vertical precision estimate of 5.9m. This small difference is significant because the vertical absolute error exceeds the 68th percentile estimate of vertical precision. This breach suggests that elevation tracking is less reliable than horizontal tracking.

Regarding evidence of bias in my results, systematic error is clearly observable across both the horizontal and vertical planes. As detailed by Bolstad and Manson (2022, p. 612), positional bias occurs when coordinate data are uniformly shifted relative to the true reference position. This uniform shift is visible on the horizontal plane, where the waypoints and the average tended to cluster most in the eastern and southeastern areas relative to the true reference point. Furthermore, a stronger vertical bias was present, with the GPS unit consistently overestimating the true altitude by 5.96 meters. Despite its high precision, systematic bias is still evident with the GPS unit.

References:

Bolstad, P., & Manson, S. (2022). GIS fundamentals: A first text on geographic information systems (7th ed.). Eider Press.