Special Topics in GIS - Module 2 - Data Quality - Standards

The second module in Special Topics in GIS focused on data quality standards with an exercise on determining the horizontal positional accuracy of two road networks in the city of Albuquerque, New Mexico. Our findings were to be reported in accordance with the National Standard for Spatial Data Accuracy (NSSDA). We were provided two polyline shapefiles that represented road centerlines from the city of Albuquerque and StreetMap USA as well as a mosaic of orthophotos of the study area.

The NSSDA standard requires at least 20 test points within our study area, with each point separated by more than one-tenth of the area's horizontal distance. The NSSDA value is an accuracy measurement of our Root Mean Square Error at the 95% confidence interval. To help achieve the requirements I used the split tool to divide the study area into four quadrants and then bookmarked each quadrant which reduced the need to zoom in and out to check the spacing of my points.

Next step was to begin finding "good" intersections that contained data from both the Albuquerque and Street Maps data sets.  After identifying my "good" intersections it was then time to determine what I thought were the "true" reference points at the intersections using the orthophotos. Below is a screenshot showing my reference or "true" locations of the intersections according to the orthophotos.

Next step was to use the Add XY Tool to determine X and Y coordinates for each of the points for all three datasets. I then exported the three attribute tables to excel files using the Table To Excel tool. Following the NSSDA horizontal accuracy statistic worksheet, the independent (true) points from the orthophotos X and Y coordinates were compared to the test points datasets. Calculations were then completed to determine the accuracy statistics for the two test datasets.  Below are the results from my worksheet for the StreetMap USA NSSDA value calculations: 

The final column of the table shows the calculated squared error distance. The values are summed and then averaged. The NSSDA horizontal accuracy is calculated by multiplying the Root Mean Square Error (RMSE) by 1.7308. Below are my final accuracy statements for each of the two datasets.

Tested 13.01 ft (3.96 m) horizontal accuracy at 95% confidence level for the Albuquerque Streets data set.

Using the National Standard for Spatial Data Accuracy, the Albuquerque Streets data set tested to 13.01 (3.96 m) feet horizontal accuracy at 95% confidence level.

Tested 312.95 ft (95.38 m) horizontal accuracy at 95% confidence level for the Street Map USA data set.

Using the National Standard for Spatial Data Accuracy, the Street Map USA data set tested to 312.95 ft (95.38) feet horizontal accuracy at 95% confidence level.

Special Topics in GIS - Module 1 - Calculating Metrics for Spatial Data Quality

The first module for Special Topics in GIS covered aspects of spatial data quality with focus on defining and understanding the difference between precision and accuracy. According to the International Organization for Standardization's (ISO) document 3534-1, accuracy can be defined as the "closeness of agreement between a test result and the accepted reference value". This document also defines precision as the "closeness of agreement between independent test results obtained under stipulated conditions" (ISO, 2007). 

In Part A of the lab assignment, the precision and accuracy metrics of provided data were determined. When determining precision, a distance (in meters) that accounts for 68% of the repeated observations was calculated.  When determining accuracy, the average waypoint was measured from an accepted reference point. Below is my map product showing projected waypoints, the average location, and circular buffers corresponding to 50%, 68%, and 95% precision estimates. A "true" reference point was later added to determine a horizontal distance to the established average waypoint location.

Horizontal accuracy refers to how close a measured GPS position (or the mean of many positions) is to the true location on the ground. It is typically reported as the distance between the GPS-derived position and a known reference point. 

Horizontal precision, on the other hand, describes how tightly repeated GPS measurements cluster together, regardless of whether they are centered on the true location. Precision is often expressed as the radius within which a certain percentage of positions (e.g., 68% or 95%) fall.

My horizontal precision (68%) was 4.5 m and my horizontal accuracy of 3.25 m produced a difference of 1.25 m. I would say that this would not be a significant difference because it sits within the 68% precision radius. My results for vertical accuracy were as follows with my mean waypoint elevation coming in at 28.54 and the mean elevation for the "true" reference point being 22.58. This is roughly a 5.96 m difference which I would think is significant at least in some cases. 

In Part B of the lab assignment, the RMSE metric was calculated, along with a cumulative distribution function (CDF). The CDF describes the probability of a random variable taking on a given variable or less, showing a more complete error distribution instead of selected metrics. For this portion we were provided another dataset where we used Excel for the analysis. Here we calculated minimum, maximum, mean, median, root square mean, and the 68th, 90th, and 95th percentiles. The final portion of the lab consisted of displaying the dataset using a cumulative distribution function (CDF) graph which is displayed below.

Overall, I really learned a lot in this lab and had the opportunity to brush up on my Excel skills which I have not utilized for a while.  I am looking forward to building upon what I learned in this module.