Data Quality - Standards - Module 2 GIS5935

 This week's lab had us determine 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). The two polyline shapefiles that were provided represented measurements made the city of Albuquerque and StreetMap USA, a tele-atlas that is distributed by ESRI with ArcGIS, exclusively. Additionally, we were provided a mosaic of orthophotos of the study area, a shapefile perimeter of the study area, and grid codes for the orthophotos to align with.

The NSSDA standard requires that we place a minimum of 20 test points in our area of study and that they are greater than one tenth of the horizontal distance of the study area apart. Having 20 test points makes computations easier for the 95% confidence interval. Following this, the NSSDA value is an accuracy measurement of our Root Mean Square Error at the 95% confidence interval. When placing points within a rectangular area, such as our provided study area, having 20% or higher of the points in each quadrant of the rectangle, spaced accordingly, is considered a uniform spread.

To create test points, I first needed to set up corresponding points along the networks of each polyline shapefile. Finding areas like intersections or sharp turns in a roadways make matching the two corresponding points on each feature much easier. For this assessment, I primarily utilized intersections as they were easy to define. Additionally, I would later be adding reference points which need an area that can be identified as the 'true' location; finding the center of an intersection is far less taxing as defining other features. There are several ways to settle on the exact location of the intersecting lines. I simply selected Add Point on Line, once the desired line involved in the intersection was highlighted, and following the provided arrows for the direction of the network line, I chose to either add a point to the beginning or the end of the line (whichever side of the line was involved in the intersection). The diagonal of the study area was measured and the study area was separated into quadrants. Since there were four quadrants to place points within, each needing 20% or more points, I decided to give each quadrant 5 points spaced over one tenth of the horizontal distance of the study area apart. 

These are the points for the Albuquerque road network:

To ensure the 'truth' of our test points, orthophotos were placed behind the two shapefiles. This allowed me to tell that the crossing of vectors in my shapefile were actually street intersections when compared to their location on the orthophoto.

Once I had created the two test point shapefiles, it was time to make a reference shapefile. This was accomplished through a simple investigation of the orthophotos. Here, I would find the location where the crossed polylines of our road networks resided and then add a new point (to a new shapefile) on the orthophoto representing my best estimate at the true location. In most real-world settings, this would be done far more thorough through the use of surveying and GPS, however, for the interest of the lab and determining a general accuracy, this more than suffices. 

Once I had my three new point shapefiles, I used the Add XY tool to add coordinates to each attribute table. Due to my operating in New Mexico State Plane projection, my coordinates were populated in feet as their distance unit.

These attribute tables were then exported as .dbf files and uploaded to an Excel spreadsheet. The goal of this spreadsheet was to take the x and y coordinate of each point, subtract them to find the difference, square the difference, add the x and y squares together, find their average, and then take the square root of it. This gives me the Root Mean Square Error which is the accuracy determining measurement. I was curious about the accuracy for the city data, as it seemed quite accurate from mere visual observation. The StreetMap USA measurements, however, were so far off that at times it was quite difficult to find the corresponding intersection or sharp turn between the corresponding roadways. 

Once I found the RMSE of each road network, I then had to multiply that number by 1.7308 as this represents the standard error of the mean at the 95% confidence level. Essentially, since we had only 20 points, one could infer that only one point would reside outside of the accuracy distance provided by the NSSDA value. 

Here are my data tables with my NSSDA accuracy statement:

City of Albuquerque road map:

Using the National Standard for Spatial Data Accuracy, the data set tested 24.16 feet horizontal accuracy at a 95% confidence level.

StreetMap USA:

Using the National Standard for Spatial Data Accuracy, the data set tested 247.64 feet horizontal accuracy at a 95% confidence level.