Module 1.3: Data Quality - Assessment

 

Figure 1: Percentage difference in road network completeness between Tiger Road and Street Centerlines, separated by grid cell to show spatial variation network completeness.

The goal of this lab was to evaluate the horizontal accuracy and completeness of two road network dataset, county Street Centerlines and Tiger Roads, using a grid‑based comparison approach. By measuring how much road length each dataset contained within every grid cell, we were able to identify where the networks aligned closely and where one was significantly more complete than the other. This type of assessment is important for understanding data reliability, especially when road networks are used for routing, emergency response, or spatial modeling.

To complete the analysis, I first clipped both road datasets to the county boundary and intersected them with the grid so that each road segment was split and assigned to the correct grid polygon. After calculating segment lengths in kilometers, I summarized total road length per grid using the grid code field as the unique identifier. These per‑grid totals allowed me to compute percentage differences between the two networks and classify which dataset was more complete in each cell. Finally, I created a graduated‑color map using to visualize spatial patterns in completeness across the county.

Module 1.2: Data Quality - Standards

 

Figure 1: Overall Data Points

To complete the NSSDA accuracy assessment, I started by preparing three point layers: the ABQ city test points, the StreetMapUSA test points, and a set of reference points digitized directly from the orthophotos. I made sure all three point layers were in the same projected coordinate system and verified that each set of points aligned correctly on the map. 

After digitizing the reference intersections, I added XY coordinates to all three layers and exported the attribute tables into Excel. In Excel, I calculated the change in X and Y coordinates for each intersection, squared those values, and computed Error², the average Error², RMSEr, and finally the NSSDA 95% accuracy value for both datasets. 

I initially ran into issues with a few mismatched rows, but once corrected, the results stabilized and produced realistic accuracy values. Based on the final calculations, the ABQ city streets tested at 20.31 feet horizontal accuracy, and StreetMapUSA tested at 154.38 feet horizontal accuracy, both at the 95% confidence level according to NSSDA standards. Thus, the ABQ city map was much more accurate than StreetMapUSA.

Module 1.1 Lab: Calculating Metrics for Spatial Data Quality

The horizontal precision of 68% was 4.5 m.

The distance between the average waypoint location and reference point was 3.2 m; horizontal accuracy.

 

The estimated horizontal accuracy was 3.2 m, based on the distance between the average waypoint location and the reference point. The 68% horizontal precision was 4.5 m. Therefore, the horizontal accuracy error was 1.3 m smaller than the 68% precision estimate. The average GPS position was closer to the reference location than the distance representing the spread of 68% of the repeated observations. Based on these results, the difference between horizontal accuracy and precision does not appear substantial.

 

Vertical Accuracy at 68% was 5.96 m.

The vertical precision is 5.9 m.

The difference between these two values is only approximately 0.07 m, indicating that vertical accuracy error and vertical precision are very similar.

Overall, the results do not provide strong evidence of significant horizontal bias because the 3.2 m displacement of the average waypoint from the reference point is smaller than the 4.5 m horizontal precision estimate.

Vertically, the average GPS elevation was 5.96 m higher than the reference elevation, which indicates a positive vertical offset. However, because the 68% vertical precision was also approximately 5.9 m, the magnitude of this offset is similar to the variability observed among the elevation measurements. Therefore, the results should be interpreted cautiously rather than as evidence of a strong systematic bias.