GIS 5935 Module 2

In this module, formally established methods of determining accuracy standards are reviewed, notably the National Standard for Spatial Data Accuracy (NSSDA). This is a national standard for determining the confidence of point positional accuracy. This method uses a sample of real and test points selected at random to generalize the accuracy of the entire data set. As an example of this practice, I tested the accuracy of two road maps of Albuquerque streets, one provided by the city of Albuquerque, and one a USA street map. Those these data layers are lines and not points, points are placed at intersections and used to test the accuracy of the road segments. Other methods for determining positional accuracy of line segments, such as the epsilon band method, are not formally standardized (Bolstad, 2016, Chpt 14). Using NSSDA guidelines and procedures, confidence of both data sets is determined and reported to the nearest foot.


Methods: 

1) Divide study area: Study are is divided into four quadrants using the measuring tool in ArcGIS Pro. 

2) Select and establish locations of true reference points: I selected 20 points randomly and by hand, 5 within each quadrant and ensuring they are all cross at 10 percent diagonally across the quadrant, using the distance between points divided by 10. This is achieved using the measuring tool in ArcGIS Pro. These points are recorded in a new feature class and represent the true locations of points. They are marked at the center of street intersections, as seen in aerial images, following best protocol for selection including the intersection being present in the both street data sets, the intersection is not a complicated intersection, and the intersection is at a right angle. 

3) Mark test points: In two new feature classes, one for each street layer, the intersections as they are represented by the street line data sets are marked using points. All points are assigned a unique point ID. 

4) Perform calculations: The NSSDA provides a table with formula guidelines to calculate the distance from the true and test points, and to determine the distance range at which 95% of points statistically should fall within of the true point location. The 3 tables are exported from ArcGIS Pro, that is the true points and each of the 2 test points, and opened in Excel. The values for each column in the table are assigned and calculations performed. These include the locations of each point, true and test, the difference between them, the squared difference between them, and the sum of the squared difference between them, performed for each of the 20 points. Then, a sum of those squared differences, an average of the squared differences, the root mean square error, and the NSSDA standard, which is the root mean square error multiplied by 1.7308. This last calculation yields to distance, in this case in feet, that 95% of the test points are likely to fall within from the true location. 

5) Report accuracy statement: The NSSDA provides 2 statement guidelines to quickly report the accuracy of the data. This structure is followed for both test data point sets. The USA streets were found to be much less accurate (NSSDA = 435.1972 ft) than the Albuquerque city streets (NSSDA = 16.6198 ft). Accuracy statements are reported below.

USA Streets: Compiled to meet 435.1972 feet horizontal accuracy at 95% confidence level. 

ABQ Streets: Compiled to meet 16.6198 feet horizontal accuracy at 95% confidence level. 

The above map shows the study area (red outline) with with 20 true points. 





Sources: 

        Bolstad, P. (2016). GIS Fundamentals: A First Text on Geographic Information Systems, Fifth Edition (5th edition). XanEdu Publishing Inc.

        Minnesota Planning Land Management Information Center. (1999, October). Positional Accuracy Handbook Using the National Standard for Spatial Data Accuracy to measure and report geographic data quality. Minnesota IT Services Geospatial Information Office. https://www.mngeo.state.mn.us/committee/standards/positional_accuracy/positional_accuracy_handbook_nssda.pdf


Comments

Post a Comment

Popular posts from this blog

GIS6005 - Module 1 San Francisco Bay Area, Point of Interest

Image
San Francisco Bay Area, Points of Interest Because this is a general reference map for public use, I opted for many preset font options available in ArcGIS Pro. This included default text options for water, sized appropriately for the significance of the body of water. The labels were placed on a bend to mimic the flow of the water and the shape of the shorelines. For land formations or parks, another preset font option was chose, the land formation option. Again, the text was placed on a curve to mimic the curve of the land feature being labeled. The font type for features that are populated places was also a font preset, the Populated Places option. Some resizing was done to make the label for San Francisco stand out from the other labels. The label for the bridge was also labeled using the populated places option, but sized smaller to fit along the narrow bridge. A couple callouts were used where the available land area was coastal and also too small to hold a label. The use o
Read more

GIS 5100 Module 2

Image
In this weeks module, LiDAR data is used and converted using several different methods to obtain information about tree height and canopy density in a grid section of Shenandoah, VA. Data transformation methods and tools include Minus, LAS to MultiPoint, IS NULL, Con, Plus, Divide, and data to raster tools for different data types. The input data is a LAS grid of Shenandoah obtained from Virginia LiDAR, and the outputs are raster datasets showing tree height, surface elevation, and canopy elevation. The following tools are used to determine tree height: The Minus tool is used to determine the difference in feature height from elevation and ground surface features which is useful in determining tree height.  The following tools are detailed as they are used to determine the biomass density of the grid patch.  LAS to Multipoint converts the LAS into a format more useful to analytics. This is performed on both the surface and vegetation datasets. IS NULL is used to identify t
Read more

GIS 5100 Module 6

Image
 This module was on conducting damage assessment on coastal flooding using ArcGIS Pro analysis tools. A case study of Hurricane Sandy is used. Using imagery and various analytic tools, a general assessment of structure damage caused by flooding and wind is conducted for an area in New Jersey.  Structures by structural damage level after Hurricane Sandy. Process summary: 1) 2 mosaic datasets are created, one for the study are before the storm and one for after, using aerial rasters. The mosaics are created within the gdb being used. The result are two projected aerial images with grids highlighting the study area. When turning on and off these two layers, damage such as flooding and missing buildings is apparent. 2) Within the gdb, new domains are created that will be used to categorize damage by type and severity. This is done by adding Domains under the gdb, and the categories for Inundation, Structure Damage, Wind Damage, and Structure Type are added alon with options for each domain
Read more