Multiple Representations of Geospatial Data: A Cartographic Search for the Holy Grail Barbara P. Buttenfield University of Colorado, Boulder babs@colorado.edu Collaboration with Cindy Brewer, Penn State University Funding from ESRI B6345-7
The Holy Grail? A single, detailed cartographic database for base mapping at multiple scales and for multiple purposes Reference feature types Standard compilation scales Base Carto Focus on what s in the database
1:24,000 DLG data Vancouver, Wash
1:100,000 DLG data Vancouver, Wash
1:2,000,000 DLG data Vancouver, Wash
Multiple map purposes Terrain analysis Traditional topographic views
Multiple map purposes Town planning Infrastructure development
Multiple map purposes Land cover analysis Site suitability
Where are the gaps? 1:100,000
The Holy Grail? A single, detailed cartographic database for base mapping at multiple scales and for multiple purposes Regardless of continuing debate on validity of a finest grain solution No single data compilation at ultimately fine resolution is currently available Gaps in current multi-scale data sources
The Holy Grail? Don t disparage the data producer Data are compiled at standard scales for specific reasons and conventions Bottom line A single unified database schema does not span continuous scale range Challenges seamless multi-scale map series and data production
Challenges Current databases mix derived representations with original compilations through life cycle Terrain Hydrography Image Base Roads Cultural (Land Cover) Smaller Scale Boundaries PLSS Place Names Finer capture use Resolution capture use capture use Hard to distinguish data as captured from data as processed
Additional Challenges Database features that persist across compilation scales vary in geometry, dimensionality, attribution and singularity (city as a point or a compound object) Inconsistent database semantics problems managing data and integrating data across scales among data providers
A Relational Database Approach Currently, most GIS and map production systems offer only minimal software support for linked multiscale data management Linkages only on simple attributes Object IDs, ID descriptions or timestamps Insert a database from another source agency, compiled in the missing scale range Integrate data schemas Link feature types
A Relational Database Approach Study and compare existing federal agency map series databases Identify scale gaps for DLG this occurs ~ 1:250,000 Insert a 1:250k dataset into the DLG schema Explore discrepancies omissions, comissions, changes to semantic hierarchy Follow European cartographic work practice Capture data within Digital Landscape Models (DLM) Derive data to Digital Cartographic Models (DCM) Targeted scales and purposes
Digital Landscape Model Cartographic Abstraction Digital Cartographic Model Map / Atlas Products DLM 1km resolution DCM_10Million 10M World Wall Map DCM_250K 250K Road Map DCM_100K 100K Recreation Map DLM 25m resolution DCM_50K 50K Wall Map DCM_Campus25K 30K Campus DCM_Topo24K 24K DOQ 24K Hillshade 24K USGS Topo DLM 5m resolution DCM_10K 10K OS Topo
Smaller Scale DLM DCM data production workflow R8 R7 Key to Representations = DLM = DCM R4 R3 = Derived R5 R6 R2 R1 Many Uses Single Use
Smaller Scale DLM DCM data production workflow R7 R8 R3 R4 R5 R2 R1 R6 Many Uses Single Use
Datasets DLG Digital Line Graph (USGS) - (reference data schema) 1:24,000 1:100,000 1: 2 million VMAP - Vector Map 1:250,000 Compliant with DIGEST Standard Same schema as DCW - Digital Chart of the World Derived from hardcopy maps, many pre-1992 Developed by NIMA - Maintained by NGA Available at http://geoengine.nga.mil
Match feature types for ~ 90 feature classes Intention automate as much as possible
Extract Unique Feature Types Manual Process Open each VMAP feature class.dbf file in Excel Advanced filter isolates unique records First pass Collect combinations for ALL feature codes Compile into single spreadsheet for each feature dataset
VMAP Feature Type Tables Up to 5 feature code descriptions Data are clearly not in normal form
Schema Crosswalk Based on DLG schema at 1:24,000 Geographic rationale Cartography Feature Type ID Match feature types using CFTID, DLG description and VMAP feature descriptions Import to DLG data schema Python script attaches CFTID codes Collect unmatched feature types Revisit unmatched schema rows individually
CFTID table after Matching Feature Types Water course lines before import to CFTID table Water course lines after import to CFTID table
Unmatched Schema Rows
Summary of schema discrepancies Either VMAP DLG Three types (at least) New feature geometries Emerging feature types Changes in schema hierarchy
New Feature Geometries Existing feature type represented with different dimension
Emerging Feature Types Existing feature type and geometry, but VMAP feature codes more detailed than DLG schema
Case Study: Airport table DLG VMAP
Hierarchy Changes 14 April 2005 Golledge Lecture
Residual counts Unresolved schema entries
Attaching symbol libraries ESRI DLG standard symbol library GEOSYM VMAP standard symbol library Other sources
DLG 1:2 million source S. California Mapped at 1:2 million
DLG 1:2 million source San Bernadino Mapped at 1:250,000 Mapped at 1:2 million
DLG 1:100K source San Bernadino Mapped at 1:250,000
VMAP 1:250K source San Bernadino Mapped at 1:250,000
VMAP 1:250K source San Bernadino Mapped at 1:100,000 Mapped at 1:250,000
Where to go from here? Alternative symbologies e.g., other NMA symbol libraries sociopolitical exercise Multiple symbol libraries in single schema table How far can a data schema be pushed? Symbol change alone Geoprocessing alone Some combination of the two Revisit the ScaleMaster fill in the gaps
Mock-up for data ftp site Choose map scale 1:45,000 Find the best mix of data source scales and tell us what you see PAN
Summary and Discussion Extracting unique features is reasonable undertaking with limited dataset, but more automation is needed Crosswalk provides insights into diverse ontologies New geometries, emerging feature types Change in hierarchy
Scaleless database? 1:100,000 RF refers to a range of map scales, not a single point Scale sensitivity varies by layer and by scale range (and this bears further scrutiny)