Calculating Spatial Data Quality: Measuring GPS Accuracy and Precision

This week’s lab focused on evaluating the accuracy and precision of handheld GPS measurements using repeated waypoints. The dataset included 50 waypoints collected with a Garmin GPSMAP 76 unit at the same physical location. Since the points were scattered around the area, it wasn’t clear where the exact location was just by looking at the raw data. To get a better idea, I calculated an “average” waypoint by aggregating all the points.

After adding the waypoints to ArcGIS Pro, I created circular buffers showing 50%, 68%, and 95% of the points fall. These buffers illustrate how the GPS points cluster around the average location.

Figure 1. Map layout with projected waypoints, average point, and precision buffers (1m, 2m, 5m shown).

Horizontal accuracy, measured as the distance between the average GPS point and the true reference, was about 3.25 meters. Meanwhile, horizontal precision, which shows how spread out the individual points are around that average, was around 4.5 meters. This tells us the GPS readings were fairly accurate overall, but the individual measurements had a bit more variation.

To get a deeper understanding of the error distribution, I worked with a larger dataset and plotted a Cumulative Distribution Function (CDF).

Figure 2. CDF plot showing cumulative error percentage for all GPS points.

The CDF curve climbs steeply up to around 70%, then rises more gradually, showing that most points are pretty close to the true location, but a few outliers have larger errors. This matches the 68th percentile at about 10.09 meters. After that, the curve flattens out, meaning fewer points fall in those higher error ranges. The curve starts at the smallest error (0.02 m) and reaches 100% at the largest error (48.37 m). The median (5.98 m) sits about halfway up the curve, which fits what you’d expect. The RMSE (3.06 m) isn’t visible on the graph since it’s more of a summary stat. Overall, the CDF gives a clearer visual picture of how the GPS errors are spread out, way better than just relying on summary numbers.