4: TIGER, Topology and Geocoding

ACCURACY IMPROVEMENT

Starting in 2002, in preparation for the 2010 census, the Census Bureau commissioned a six-year, $200 million MAF/TIGER Accuracy Improvement Project (MTAIP). One objective of the effort was to use GPS to capture accurate geographic coordinates for every household in the MAF. Another objective was to improve the accuracy of TIGER’s road/path features. The project aimed to adjust the geometry of street networks to align within 7.6 meters of street intersections observed in orthoimages or measured using GPS. The corrected streets are necessary not just for mapping, but for accurate geocoding. Because streets often form the boundaries of census areas, it is essential that accurate household locations are associated with accurate street networks.

MTAIP integrated over 2,000 source files submitted by state, tribal, county, and local governments. Contractors used survey-grade GPS to evaluate the accuracy of a random sample of street centerline intersections of the integrated source files. The evaluation confirmed that most but not all features in the spatial database equal or exceed the 7.6 meter target. Uniform accuracy wasn’t possible due to the diversity of local source materials used, though this accuracy is the standard in the “All Lines” Shapefile extracts. The geometric accuracy of particular feature classes included in particular shapefiles are documented in the metadata associated with that shapefile extract.

MTAIP was completed in 2008. In conjunction with the continuous American Community Survey and other census operations, corrections and updates are now ongoing. TIGER/Line Shapefile updates are now released annually.

PRACTICE QUIZ

Registered Penn State students should return now to the Chapter 4 folder in ANGEL (via the Resources menu to the left) to take a self-assessment quiz about MAF and TIGER.

You may take practice quizzes as many times as you wish. They are not scored and do not affect your grade in any way.

4.5. Shapefiles

Since 2007, TIGER/Line extracts from the MAF/TIGER database have been distributed in shapefile format. Esri introduced shapefiles in the early 1990s as the native digital vector data format of its ArcView software product. The shapefile format is proprietary, but open; its technical specifications are published and can be implemented and used freely. Largely as a result of ArcView’s popularity, shapefile has become a de facto standard for creation and interchange of vector geospatial data. The Census Bureau’s adoption of Shapefile as a distribution format is therefore consistent with its overall strategy of conformance with mainstream information technology practices.

4.6. Topology

Topology is different than topography. (You’d be surprised how often these terms get mixed up.) In Chapter 2 you read about the various ways that absolute positions of features can be specified in a coordinate system, and how those coordinates can be projected or otherwise transformed. Topology refers to the relative positions of spatial features. Topological relations among features—such as containment, connectivity, and adjacency—don’t change when a dataset is transformed. For example, if an isolated node (representing a household) is located inside a face (representing a congressional district) in the MAF/TIGER database, you can count on it remaining inside that face no matter how you might project, rubber-sheet, or otherwise transform the data. Topology is vitally important to the Census Bureau, whose constitutional mandate is to accurately associate population counts and characteristics with political districts and other geographic areas.

As David Galdi (2005) explains in his white paper “Spatial Data Storage and Topology in the Redesigned MAF/TIGER System,” the “TI” in TIGER stands for “Topologically Integrated.” This means that the various features represented in the MAF/TIGER database—such as streets, waterways, boundaries, and landmarks (but not elevation!)—are not encoded on separate “layers.” Instead, features are made up of a small set of geometric primitives—including 0-dimensional nodes and vertices, 1-dimensional edges, and 2-dimensional faces—without redundancy. That means that where a waterway coincides with a boundary, for instance, MAF/TIGER represents them both with one set of edges, nodes and vertices. The attributes associated with the geometric primitives allow database operators to retrieve feature sets efficiently with simple spatial queries. The separate feature-specific TIGER/Line Shapefiles published at the county level (such as point landmarks, hydrography, Census block boundaries, and, the “All Lines” file you are using in the multi-part “Try This”) were extracted from the MAF/TIGER database in that way. Notice, however, that when you examine a hydrography shapefile and a boundary shapefile, you will see redundant line segments where the features coincide. That fact confirms that TIGER/Line Shapefiles, unlike the MAF/TIGER database itself, are not topologically integrated. Desktop computers are now powerful enough to calculate topology “on the fly”from shapefiles or other non-topological data sets. However, the large batch processes performed by the Census Bureau still benefit from the MAF/TIGER database’s persistent topology.

MAF/TIGER’s topological data structure also benefits the Census Bureau by allowing it to automate error-checking processes. By definition, features in the TIGER/Line files conform to a set of topological rules (Galdi 2005):

1. Every edge must be bounded by two nodes (start and end nodes)..
2. Every edge has a left and right face.
3. Every face has a closed boundary consisting of an alternating sequence of nodes and edges.
4. There is an alternating closed sequence of edges and faces around every node.
5. Edges do not intersect each other, except at nodes.

Compliance with these topological rules is an aspect of data quality calledlogical consistency. In addition, the boundaries of geographic areas that are related hierarchically—such as blocks, block groups, tracts, and counties—are represented with common, non-redundant edges. Features that do not conform to the topological rules can be identified automatically, and corrected by the Census geographers who edit the database. Given that the MAF/TIGER database covers the entire U.S. and its territories, and includes many millions of primitives, the ability to identify errors in the database efficiently is crucial.

So how does topology help the Census Bureau assure the accuracy of population data needed for reapportionment and redistricting? To do so, the Bureau must aggregate counts and characteristics to various geographic areas, including blocks, tracts, and voting districts. This involves a process called “address matching” or “address geocoding” in which data collected by household is assigned a topologically-correct geographic location. The following pages explain how that works.

4.7. Geocoding

Geocoding is the process used to convert location codes, such as street addresses or postal codes, into geographic (or other) coordinates. The terms “address geocoding” and “address mapping” refer to the same process. Geocoding address-referenced population data is one of the Census Bureau’s key responsibilities. However, as you know, it’s also a very popular capability of online mapping and routing services. In addition, geocoding is an essential element of a suite of techniques that are becoming known as “business intelligence.” We’ll look at applications like these later in this chapter, but first let’s consider how the Census Bureau performs address geocoding.

4.8. Geocoding with TIGER/Line Shapefiles

4.9. Geocoding Online

No doubt you’re familiar with one or more popular online mapping services. How well do they do at geocoding the location of a postal address? You can try it out for yourself at several Web-based mapping services, including MapQuest.com, Microsoft’s Bing Maps, and Tele Atlas/TomTom’s Geocode.com. Tele Atlas, for example, is a leading manufacturer of digital street data for vehicle navigation systems. To accommodate the routing tasks that navigation systems are called upon to serve, the streets are encoded as vector features whose attributes include address ranges. (In order to submit an address for geocoding at Geocode.com you have to set up a trial account through their EZ-Locate Interactive web tool or download the EZ-Locate software).

Submitting an address to Tele Atlas’ Geocode.com service for geocoding. © 2013 TomTom North America, Inc. All rights reserved.

Shown above is the form by which you can geocode an address to a location in a Tele Atlas street database. The result is shown below.

Tele Atlas’ Geocode.com service estimates the location of the address relative to the address range attributes encoded in its database. © 2013 TomTom North America, Inc. All rights reserved.

Let’s compare the geocoding capabilities of MapQuest.com to locate the address on an actual map.

Address geocoded by MapQuest.com. © 2013 MapQuest.com, Inc. All rights reserved.

The MapQuest.com map from 2013 estimates the address is close to its actual location. Below is a similar MapQuest product created back in 1998, when this course was first being developed. On the older map the same address is plotted on the opposite side of the street. What do you suppose is wrong with the address range attribute in that case?

On the map from 1998, also note the shapes of the streets. The street shapes in the 2011 map have been improved. The 1998 product seems to have been generated from the 1990 version of the TIGER/Line files, which may have been all that was available for this relatively remote part of the country. Now MapQuest licenses street data from a business partner called NAVTEQ.

Same address geocoded by MapQuest.com in 1998. © 1998 MapQuest.com, Inc. (formerly GeoSystems Global Corp.) All rights reserved.

The point of this section is to show that geocoding with address ranges involves a process of estimation. The Census Bureau’s TIGER/Line Shapefiles, like the commercial street databases produced by Tele Atlas, Navigation Technologies, and other private firms, represent streets as vector line segments. The vector segments are associated with address range attributes, one for the left side of the street, one for the right side. The geocoding process takes a street address as input, finds the line segment that represents the specified street, checks the address ranges to determine the correct side of the street, then estimates a location at the appropriate point between the minimum and maximum address for that segment and assignes an estimated latitude/longitude coordinate to that location. For example, if the minimum address is 401, and the maximum is 421, a geocoding algorithm would locate address 411 at the midpoint of the street segment.

4.10. Applications beyond the Census Bureau

Two characteristics of MAF/TIGER data, address range attributes and explicit topology, make them, and derivative products, valuable in many contexts. Consequently, firms like NAVTEQ and Tele Atlas (now owned by TomTom) have emerged to provide data with similar characteristics as MAF/TIGER, but which are more up-to-date, more detailed and include additional feature classes. The purpose of the next section is to sketch some of the applications of data similar to MAF/TIGER data beyond the Census Bureau.

4.11. Geocoding Your Customers

Geocoded addresses allow governments and businesses to map where their constituents and customers live and work. Federal, state, and local government agencies know where their constituents live by virtue of censuses, as well as applications for licenses and registrations. Banks, credit card companies, and telecommunications firms are also rich in address-referenced customer data, including purchasing behaviors. Private businesses and services must be more resourceful.

Some retail operations, for example, request addresses or ZIP Codes from customers, or capture address data from checks. Discount and purchasing club cards allow retailers to directly match purchasing behaviors with addresses. Customer addresses can also be harvested from automobile license plates. Business owners pay to record license plate numbers of cars parked in their parking lots or in their competitors. Addresses of registered owners can be purchased from organizations that acquire motor vehicle records from state departments of transportation.

Businesses with access to address-referenced customer data, vector street data attributed with address ranges, and GIS software and expertise, can define and analyze the trade areas within which most of their customers live and work. Companies can also focus direct mail advertising campaigns on their own trade areas, or their competitors’. Furthermore, GIS can be used to analyze the socio-economic characteristics of the population within trade areas, enabling businesses to make sure that the products and services they offer meet the needs and preferences of target populations.

Politicians use the same tools to target appearances and campaign promotions.

4.12. Delivering Products and Services

Operations such as mail and package delivery, food and beverage distribution, and emergency medical services need to know not only where their customers are located, but how to deliver products and services to those locations as efficiently as possible. Geographic data products like TIGER/Line Shapefiles are valuable to analysts responsible for prescribing the most efficient delivery routes. The larger and more complex the service areas of such organizations, the more incentive they have to automate their routing procedures.

In its simplest form, routing involves finding the shortest path through a network from an origin to a destination. Although shortest path algorithms were originally implemented in raster frameworks, transportation networks are now typically represented with vector feature data, like TIGER/Line Shapefiles. Street segments are represented as digital line segments each formed by two points, a “start” node and an “end” node. If the nodes are specified within geographic or plane coordinate systems, the distance between them can be calculated readily. Routing procedures sum the lengths of every plausible sequence of line segments that begins and ends at the specified locations. The sequence of segments associated with the smallest sum represents the shortest route.

To compare various possible sequences of segments, the data must indicate which line segment follows immediately after another line segment. In other words, the procedure needs to know about the connectivity of features. As discussed earlier, connectivity is an example of a topological relationship. If topology is not encoded in the data product, it can be calculated by the GIS software in which the procedure is coded.

Input form for an early version of the MapQuest routing utility. © 1998 MapQuest.com, Inc. All rights reserved.

Several online travel planning services, including MapQuest.com and Google Maps, provide routing capabilities. Both take origin and destination addresses as input, and produce optimal routes as output. These services are based on vector feature databases in which street segments are attributed with address ranges, as well as with other data that describe the type and conditions of the roads they represent.

An early interface to MapQuest’s routing options. Different algorithms are required to calculate shortest and fastest routes. Specific attributes must be encoded in the database to provide the options to avoid limited access highways, toll roads, and ferry lanes. © 1998 MapQuest.com, Inc. All rights reserved.

The shortest route is not always the best. In the context of emergency medical services, for example, the fastest route is preferred, even if it entails longer distances than others. To determine fastest routes, additional attribute data must be encoded, such as speed limits, traffic volumes, one way streets, and other characteristics.

MapQuest routing solution. © 1998 MapQuest.com, Inc. All rights reserved.

Then there are routing problems that involve multiple destinations–a complex special case of routing called the traveling salesman problem. School bus dispatchers, mail and package delivery service managers, and food and beverage distributors all seek to minimize the transportation costs involved in servicing multiple, dispersed destinations. As the number of destinations and the costs of travel increase, the high cost of purchasing up-to-date, properly attributed network data becomes easier to justify.

4.13. Delineating Service Areas

The need to redraw voting district boundaries every ten years was one of the motivations that led the Census Bureau to create its MAF/TIGER database. Like voting districts, many other kinds of service area boundaries need to be revised periodically. School districts are a good example. The state of Massachusetts, for instance, has adopted school districting laws that are similar in effect to the constitutional criteria used to guide congressional redistricting. The Framingham (Massachusetts) School District’s Racial Balance Policy once stated that “each elementary and middle school shall enroll a student body that is racially balanced. … each student body shall include a percentage of minority student, which reflects the system-wide percentage of minority students, plus or minus ten percent. … The racial balance required by this policy shall be established by redrawing school enrollment areas” (Framingham Public Schools 1998). And bus routes must be redrawn as enrollment area boundaries change.

The Charlotte-Mecklenberg (North Carolina) public school district also used racial balance as a districting criterion (although its policy was subsequently challenged in court). Charlotte-Mecklenberg consists of 133 schools, attended by over 100,000 students, about one third of whom ride a bus to school every day. District managers are responsible for routing 3,600 bus routes, traveling a total of 82,000 daily miles. A staff of eight routinely uses GIS to manage these tasks. GIS could not be used unless up-to-date, appropriately attributed, and topologically encoded data were available.

Another example of service area analysis is provided by the City of Beaverton, Oregon. In 1997, Beaverton officials realized that 25 percent of the volume of solid waste that was hauled away to land fills consisted of yard waste, such as grass clippings and leaves. Beaverton decided to establish a yard waste recycling program, but it knew that the program would not be successful if residents found it inconvenient to participate. A GIS procedure called allocation was used to partition Beaverton’s street network into service areas that minimized the drive time from residents’ homes to recycling facilities. Allocation procedures require vector-format data that includes the features, attributes, and topology necessary to calculate travel times from all residences to the nearest facility.

Trade areas defined by 3 miles travel distance (blue) and 8 minutes travel time (yellow). (Francica n.d.). Used by permission.

Naturally, private businesses concerned with delivering products and services are keenly interested in service area delineation. The screen capture above shows two trade areas surrounding a retail store location (“Seattle Downtown”) in a network database.

Former student Saskia Cohick (Winter 2006), who was then GIS Director for Tioga County, Pennsylvania, contributed another service area problem: “This is a topic that local governments are starting to deal with … To become Phase 2 wireless capable (that is, capable of finding a cell phone location from a 911 call center within 200 feet of the actual location), county call centers must have a layer called ESZs (Emergency Service Zones). This layer will tell the dispatcher who to send to the emergency (police, fire, medical, etc). The larger problem is to reach agreement between four fire companies (for example) as to where they do or do not respond.”

4.14. Summary

To fulfill its mission of being the preeminent producer of attribute data about the population and economy of the United States, the U.S. Census Bureau also became an innovative producer of digital geographic data. The Bureau designed its MAF/TIGER database to support automatic geocoding of address-referenced census data, as well as automatic data quality control procedures. The key characteristics of TIGER/Line Shapefiles, including use of vector features to represent geographic entities, and address range attributes to enable address geocoding, are now common features of proprietary geographic databases used for trade area analysis, districting, routing, and allocation.

4.15. Bibliography

Charlotte-Mecklenberg Public Schools (n. d.). Retrieved July 19, 1999 from http://www.cms.k12.nc.us

Cooke, D. F. (1997). Topology and TIGER: The Census Bureau’s Contribution. In T. W. Foresman (Ed.), The history of geographic information systems: Perspectives from the pioneers. (pp. 47 – 57). Upper Saddle River, NJ: Prentice Hall.

Dangermond, J. (1982). A Classification of Software Components Commonly Used in Geographic Information Systems. In Proceedings of the U.S.—Australia Workshop on the Design and Implementation of Computer-Based Geographic Information Systems, Honolulu, HI, pp. 0-91. In Demers, M.N. (1997) Fundamentals of Geographic Information Systems. John Wiley & Sons, Inc.

Discreet Research (n.d.). Retrieved July 19, 1999 fromwww.dresearch.com

ESRI (1998) Shapefile Technical Description, An ESRI White paper. Environmental Systems Research Institute, Inc. Retrieved October 4, 2010, from http://www.esri.com/library/whitepapers/pdfs/shapefile.pdf

Federal Geographic Data Committee (April 2006). Retrieved July 19, 1999 from http://www.fgdc.gov

Framingham Public Schools (1998). Racial balance policy: Assignment of students to schools. Retrieved July 19, 1999 fromwww.framingham.k12.ma.us/update/0198rbp.html (since retired).

Francica, J. (n.d.). Geodezix Consulting. Retrieved July 19, 1999 fromwww.geodezix.com (since retired).

Galdi, D. (2005). Spatial Data Storage and Topology in the Redesigned MAF/TIGER System. Retrieved 19 October 2010 fromwww.census.gov/geo/mtep_obj2/topo_and_data_stor.html (since retired).

MapQuest (n.d. a). Retrieved July 19, 1998 fromhttp://www.mapquest.com

MapQuest (n.d. b). Retrieved January 15, 2013 fromhttp://www.mapquest.com

Marx, R. M. (Ed.). (1990). The Census Bureau’s TIGER system. [Special issue]. Cartography and Geographic Information Systems 17:1.

Navigation Technologies Inc. (2006). Welcome to NavTech. Retrieved July 19, 1999 from www.navtech.com

Rammage, S. and P. Woodsford (2002). The Benefits of Topoplogy in the Database. Retrieved October 6, 2010 fromhttp://spatialnews.geocomm.com/features/laserscan2/

TeleAtlas (2006). Welcome to TeleAtlas. Retrieved May 3, 2006 fromwww.teleatlas.com/Pub/Home (since retired).

Theobald, D. M. (2001). Understanding Topology and Shapefiles.ArcUser April-June 2001. Retrieved October 5, 2010 fromhttp://www.esri.com/news/arcuser/0401/topo.html

U.S. Census Bureau (1997). TIGER/Line Files (1997 Technical Documentation). Retrieved January 2, 1999 fromwww.census.gov/geo/tiger/TIGER97C.pdf (since retired).

U.S. Census Bureau (2003). TIGER/Line Files, 2003 (metadata). Retrieved February 3, 2008 fromwww.census.gov/geo/www/tlmetadata/tl2003meta.txt

U.S. Census Bureau (n. d.). 21st Century MAF/TIGER Enhancements. Retrieved February 3, 2008 fromwww.census.gov/geo/mod/overview.pdf (since retired).

U.S. Census Bureau (2004). MAF/TIGER Redesign Project Overview. Retrieved October 19, 2010 fromwww.census.gov/geo/mtep_obj2/obj2_issuepaper12_2004.pdf(since retired).

U.S. Census Bureau (2005). Geography division map gallery. Retrieved July 19, 1999 from www.census.gov/geo/www/mapGallery/

U.S. Census Bureau (2012). TIGER/Line Shapefiles Technical Documentation. Retrieved June, 2013 from of thewww.census.gov/geo/maps-data/data/pdfs/tiger/tgrshp2012/TGRSHP2012_TechDoc.pdf