Data Quality Standards

 This week, we had to test the accuracy of road maps using methods as laid out by the National Standard for Spatial Data Accuracy (NSSDA). We were given 2 different road layers as well as the ortho imagery of the area. We used the imagery to create a layer of references points to test the accuracy.

I began by determining which intersection to use for testing. This involved creating the reference points layer, examining the imagery and road file to find suitable locations that were included in the study area as well as all three layers, and placing points at the center of intersections based on the imagery. I also had to ensure the even spacing of the points per NSSDA guidelines Next I digitized the intersections for each road layer. I computed the xy coordinates for each layer using the Add XY Coordinates tool and exported these to excel files. I assembled all of this data into the NSSDA worksheet and did some calculating.

These are the 25 points I selected to test for accuracy for the two road layers.

At first pass, the error distance was way too high - in the hundreds of thousands of feet. I knew from visual inspection that this was not correct. I did a lot of checking of calculations, searching the internet for clues, and chatting with classmates. I finally realized that I had not sorted the reference point data correctly so it was not calculating the point data for the same points. I both love and hate when it's something so dumb. 

And so I present my horizontal data accuracy statements:

City of Albuquerque data: Tested 18 feet horizontal accuracy at 95% confidence level.

Street Map USA data: Tested 355 feet horizontal accuracy at 95% confidence level.

Calculating Metrics for Spatial Data Quality

This week we leaned how to calculate the precision and accuracy for various points. We specifically looked at precision, accuracy, and error. In part A, we calculated the precision and accuracy for GPS points. In part B, we calculated the root-mean-square error (RMSE) and cumulative distribution function (CDF) of another set of point data.

Precision refers to the consistency of the measuring implement - does the GPS collect the point in the same spot each time. In determining precision, we looked at how closely the points fell within an average location. The horizontal precision was visualized below with buffers based on where 50%, 68%, and 95% of points fell. The measurement of the 68 percentile was used to represent the both the horizontal and vertical precision. The horizontal precision is 4.5m and the vertical precision is .9m.

Meanwhile, accuracy refers to how closely a point reflects the true location - does the GPS collect the point in the spot where you want it. To determine the accuracy of the points, we measured the distance between the average recorded point location and the know location of the reference point. The horizontal accuracy is 3.7m and the vertical accuracy is 6m.

GIS Internship

This semester, I am working on an internship to gather real-world experience. I was fortunate enough to take on an additional project at work to provide this experience. I will be creating a geodatabase for our Gullah/Geechee Burial Ground Reporting Project. This in an ongoing project to document cultural resources in eastern Nassau county significant to the African American and Gullah/Geechee communities. While I've been working on aspects of this project for several years, the skills I've learned through my GIS coursework will allow me to take a different role this fall - not just collecting data in the field, but actually compiling, analyzing, and creating geospatial deliverables. 

My new efforts in the project will include gathering historic imagery to help map the past landscape for the area, conducting in the field surveys to document resources including collecting and creating shapefiles and maps for each site, and conducting GPR surveys to help located buried features including unmarked burials. The GPR surveys include collecting geographic location information for each grid, creating a geodatabase for each project that includes georeferenced raster files for each grid slice (imagery every 10 cm) and a feature layer of potential areas of interest, and in many cases, taking this data back into the field via an online map in Field Maps to mark the locations back on the ground. 

I've also joined the Florida URISA user group to help me link into the GIS community in Florida. I've chosen this user group because they are looking at issues in Florida and providing webinars and other resources that could help me. 

Corridor Analysis

In this analysis, we had to create a model to suggest an area for a wildlife corridor for black bears in between two NPS parcels. We had to consider the landcover, elevation, and distance to roads as criteria for where the bears would most likely want to go. 

I prepare the road map by converting it to a raster using the Euclidean Distance Tool. Then I used the reclassify tool for all three data sets (roads, land cover and elevation) as per the guidance provided in the lab instructions. Next I added all three of these layers to the Weighted Overlay tool, set the scale from 1-10, and weight the layers as follows: 60% land cover, 20% roads, 20% elevation. I used the Reclassify tool to create a cost surface by changing the values to reflect the suitability of each cell. I calculated each value by subtracting it from 10. Then I ran the Cost Distance tool twice, using the previously created cost surface layer with Coronado1 and then Coronado2 as the source. Finally, I ran the Corridor tool using the two Cost Distance files. 

To create the final product, I had to examine the corridor and change symbology to reflect a reasonably-sized corridor so it didn't include the entire landmass between the two parks. This took a bit of trial and error. In the end, I used the suggested categories from earlier in the lab to create the corridors.

 

Rating Locations in Rasters

In this analysis, we used classified rasters to rank areas most suitable for development. The criteria considered included original land cover, soils, slope, proximity to rivers, and proximity to existing roads. The data for land cover and slope were already in raster format, so it was a simple process of reclassifying these cells into the suitability ratings. The data for soils were included as polygons, so I had to convert them into a raster and then reclassify based on suitability ratings. The rivers and roadways were polylines, so I used the Euclidean Distance tool to create a raster and then reclassifying those based on distances from the original polylines.

To combine all the data and find the areas most suitable, I used the Weighted Overlay Tool. I ran the tool twice, once with all criteria weighted equally and once giving more weight to the slope and less to river and road distance. I used the Zonal Geometry as Table tool to calculate the areas for each suitability rating for both models. Using the evenly weighted scenario found less land with the most suited rating, and equal amounts of land rated 3 or 4. The other alternative scenario found more land with the most suited rating, but far less with a rating of 4, and most land rated at a 3.