In this week's lab, we were required to perform a completeness assessment on two road network shapefiles. Our goal was to discern which road network shapefile bore better representation for their real world counterpart. We would do this buy measuring which road network is longer (more complete) and comparing each network in a measured grid to determine the percent difference of completeness between the two. We would then display this with a choropleth map communicating the more complete network per each grid cell.
The comparative study was to be executed for Jackson County, Oregon. We were provided with a shapefile of the TIGER 2000 census lines the Street Centerlines developed by the county.
to complete the analysis we were also provided with a grid shapefile of 297 grids each measuring roughly 1km x 1km.
A completeness analysis is a relatively simple comparative analysis which determines how complete a certain data set is relative to a known reference data set. This form of analysis can also be performed between two disparate set of data measuring the same features.
A simple form of this analysis can be done by simply finding the sum of all road features for each shapefile and then finding the difference between them. For this lab the Street Centerlines totaled to 10,671.1 km and the TIGER lines to 11,253.4km.
We could further distill the completeness of each network over a large region by breaking the area down into a uniform grid and assessing which network is more complete in each grid cell. This would give us a visual distribution of locations where one grid network is more complete than the other and the corresponding percent differential. Per each grid cell, we would be able to find the distance in length between each shapefile and then subsequently each percent difference between the two. Once laid out in a choropleth map, its easy do discern areas where one network may suffer in its completeness relative to the other. Additionally, it tells us which networks are more reliable in each gridded area, allowing us to discern which road network would be best suited for use in a specific area.
Using ArcGis Pro, I trimmed the road networks to remove any lines features that resided outside of the grid cell. With the use of several geoprocessing tools, I was able to isolate the road networks that exist within each of the grids and then join the attributes residing in the polylines to the attributes of the grid file. Each of the grids were numbered with a grid code so I would be able to ensure that the attribute data for each polyline was added to to correct spatially corresponding grid cell. I then created two new fields in the merged attribute table. In the first field, I calculated the difference in length between the two shapefiles. In the second, I converted this number into a percent using the sum of the Street Centerlines polylines in each grid cell as the base.
I was then able to take all of these percent differentials and display them on a choropleth map. When the TIGER lines were longer in a grid cell, our percent value would be a negative number. When the Street Centerlines were longer, our percent was a positive value. When the values are displayed with a diverging color scheme, we are able to display and communicate which road network is more complete in each grid cell and by how much more (via percent) it is. You can also see the dispersal of completeness in a larger scale. In my results below, I classified the values in 8 classes in a manner where grids with little variance between the two can be easily identified due to its one designated hue (within 5%). Due to the range of values being much higher with the TIGER lines, I had to adjust the scale to go to over 100%. Because of this, I increased the percentile range as the data ascended in class. This helped polarize areas where there is a tremendous percent difference.
Below is the final map:
Comments
Post a Comment