Larry Stanislawski, Michael HowardCenter of Excellence for Geospatial Information Science (CEGIS), United States Geological Survey, Rolla MO
Marc-Olivier Briat, Edith PuntEsri, Inc., Redlands CA
Cynthia Brewer Department of Geography, Pennsylvania State University, Unversity Park PA
Barbara ButtenfieldDepartment of Geography, University of Colorado-Boulder, Boulder CO
Density-Stratified Thinning to Support Automated Generalization of
Transportation
15th ICA Generalisation Workshop, Istanbul, Turkey, September 13-14, 2012
Outline
• Esri Thin Road Network Tool
• Density-stratified Thinning
• Results of Stratified Thinning
• Summary Statements and Future Work
15th ICA Generalisation Workshop, Istanbul, Turkey, September 13-14, 2012
Cartography Geoprocessing Toolbox
• ArcGIS 10 introduced contextual generalization tools that consider relationships between features from multiple layers- Maintain representative patterns,
density, and character- Resolve conflicts between
symbolized features
Esri Thin Road Network tool
• Maintain pattern and density while retaining connectivity• Keep significant roads only
- Balanced by road classification- Retain specific features by locking
• Visibility controlled by attribute, easy to modify
Esri Thin Road Network Tool
15th ICA Generalisation Workshop, Istanbul, Turkey, September 13-14, 2012
Inputs:1. Road network features,2. Minimum length,3. Invisibility field, and4. Hierarchy field
Marks features for elimination to create a simplified pattern of roads that maintains connectivity, representative pattern, and density
Limitations of Thin Road Network Tool• Preprocessing• A single minimum length can homogenize local density
variations (more than expected)• Difficult to set tolerance values for tool
Preprocessing for Thin Network Tool
15th ICA Generalisation Workshop, Istanbul, Turkey, September 13-14, 2012
• Road network featureso Projected coordinate systemo Remove coincident featureso Transfer names to retained featureso Multi-part features to single-part featureso Ensure features are split at intersections
• Hierarchy fieldo Compute importance values based on
road class (and names where class is missing)
Test Data: Four Subbasins in Rural Iowa and Part of Atlanta Metropolitan Area
Part of Atlanta MSA• ~10,000 sq km• Dense urban area• Atlanta MSA pop. ~5.4M• Over 393,000 road features• Nearly 49,000 km of roads
15th ICA Generalisation Workshop, Istanbul, Turkey, September 13-14, 2012
Four subbasins in Iowa:• ~24,000 sq km• Rural midwest agricultural area • Des Moines ~580,000 persons• Over 109,000 road features• Nearly 36,000 km of roads
Road data from transportation layer of USGS Best Practices (BP)database
Test Methods1. Subdivide data into density classes
• Iowa: < 1.50 and > 1.50 km / sq km, min. polygon area = 45 sq km• Atlanta: < 2.50, 2.50 to 4.50, and > 4.50 km / sq km, min. polygon area = 45 sq km.
2. Determine 100K target density estimate for each class using Radical Law3. Run thin network tool multiple times to find which minimum length comes closest to
the 100K target density for each density class4. Extract visible lines for each class using the invisibility field5. Compare resulting 100K extracted lines with 100K DLG lines by subtracting the raster
road-density thinned Best Practices (BP) data from the 100K DLG roads (300m grids).
15th ICA Generalisation Workshop, Istanbul, Turkey, September 13-14, 2012
IowaResults
15th ICA Generalisation Workshop, Istanbul, Turkey, September 13-14, 2012
BP and 100K DLG RoadsStudy area in Iowa
15th ICA Generalisation Workshop, Istanbul, Turkey, September 13-14, 2012
DLGs compiled 1981 to 1985Compiled
2011
Tom Tom and 100K DLG RoadsStudy area in Iowa
15th ICA Generalisation Workshop, Istanbul, Turkey, September 13-14, 2012
Tom Tom Roads and Density Strata(class breaks: < 1.5 and more than 1.5 km per sq km)
15th ICA Generalisation Workshop, Istanbul, Turkey, September 13-14, 2012
Thinning Tom Tom to 100K Radical LawThinning entire data set Urban Rural Partitions
Density Class Rural Urban
Density (km/sq km)
Percent from Radical Law
Density (km/sq km)
Percent from Radical Law
Density Class Breakless than 1.50
more than 1.50
Average Density of class at 1:24,000 1.24 3.47Radical Law Density Estimate for 1:100,000 0.61 1.70Minimum Length 1 km 1.14 87.90 2.50 46.98Minimum Length 2 km 1.10 81.67 2.12 24.44Minimum Length 4 km 1.06 74.86 1.80 5.65Minimum Length 10 km 0.63 4.35 1.19 -29.98
15th ICA Generalisation Workshop, Istanbul, Turkey, September 13-14, 2012
Tom Tom Roads Thinned to Radical Law 100K
15th ICA Generalisation Workshop, Istanbul, Turkey, September 13-14, 2012
AtlantaResults
15th ICA Generalisation Workshop, Istanbul, Turkey, September 13-14, 2012
TomTom Roads and Density Strata(class breaks: < 2.5, 2.5 to 4.5, and more than 4.5 km per sq km)
15th ICA Generalisation Workshop, Istanbul, Turkey, September 13-14, 2012
Thinning Tom Tom to 100K Radical LawThinning entire data set Urban Rural Partitions
15th ICA Generalisation Workshop, Istanbul, Turkey, September 13-14, 2012
Density Partition Density (km/sq km)
Density Class Rural Suburban Urban
Density (km/sq km)
Percent from Radical Law
Density (km/sq km)
Percent from Radical Law
Density (km/sq km)
Percent from Radical Law
Density Class Breakless than 2.50
less than 4.50
more than 4.50
Average Density at 1:24,000 1.65 3.35 7.20Radical Law Density Estimate for 1:100,000 0.81 1.64 3.53Minimum Length 500 m 1.53 88.83 2.54 54.88 5.03 42.59Minimum Length 1000 m 1.32 63.04 1.94 18.06 3.74 5.89Minimum Length 1500 m 1.20 47.59 1.67 1.37 3.18 -9.71Minimum Length 2000 m 1.13 39.59 1.54 -6.26 2.88 -18.34Minimum Length 2500 m 1.09 34.49 1.47 -10.38 2.69 -23.85
Tom Tom Roads Thinned to Radical Law 100K
15th ICA Generalisation Workshop, Istanbul, Turkey, September 13-14, 2012
Left: All Atlanta TomTom Roads
Middle: TomTom Roads thinned using 2 km minimum length
Right: TomTom Roads thinned in three density-strata
10 km
• Density-stratification is a flexible approach to use Thin Network Tool that preserves local density variation better than using a single minimum length
• Future work:o Select formal density class breaks for stratifying the countryo Identify thinning requirements for 100K and smaller scaleso Automate selection of minimum length for each density classo Ensure feature connectivity at density-class boundarieso Test a complete work flow of the Esri transportation
generalization tools (include merge divided highway tool, remove road conflicts tool, etc.) [is this necessary to include…]
o Test and implement parallel processing to improve performance
o Test network navigation capabilities before and after thinning
15th ICA Generalisation Workshop, Istanbul, Turkey, September 13-14, 2012
Summary and Future Work