The Census Bureau's Geographic
Support System Initiative – An Update
Council of Professional Associations on Federal Statistics
September 21, 2012
Tim Trainor
Chief, Geography Division
U.S. Census Bureau
Census Geographic Support –
Major Initiatives Over Time
For the 1990 Census – Introduced TIGER
For the 2000 Census – Introduced the Master Address File
For the 2010 Census – Realigned the street network
through the MAF/TIGER Enhancement Program
For the 2020 Census – The GSS Initiative
A Change in Methodology
In Taking a Census
Prior to
1960
• Door to door
enumeration
1960
1970
1980
1990
• First mail-out • Census
• ~95% of the • First use of
census
created an
U.S.
TIGER
address
population
is
• USPS
• Address list
register for
now included
delivered a
created from
in the mailquestionnaire densely
the ground-up
populated
out/mail-back
to every
USPS routes
census
household on
their routes
• First mail• Address list
out/mail-back
created from
• Enumerators
census
the ground-up
collected the
completed
• Urban areas
forms
mailed back
their forms;
rural area
forms were
collected by
enumerators
2000
2010
2020
• Birth of the
• Continuous
• Introduction
MAF
update of the
of Targeted
MAF to
• MAF/TIGER
Address
Enhancement support the
Canvassing
ACS
Project
and the
• 1990 Address • Address
Geographic
canvassing
list was used
Support
covered the
as a starting
System
entirety of the Initiative
point
U.S. prior to
(GSS-I)
• Began
Census day
receiving the
DSF from the
USPS
Improving Data Quality
1: Establish quantitative
measures of
address and spatial data
quality
Existing
MAF/TIGER
Data
New
incoming
data
3: Monitor
and Improve
the quality
of the:
2: Assign Quality
Indicators to
MAF/TIGER data
IT processes
for updating
the
MAF/TIGER
System
Geographic
products
output from
the
MAF/TIGER
System
Improved Partnerships
TIGERweb
Enhanced collaboration
Community TIGER
Expand Existing
Partnerships
Crowd Sourcing
Web-based Address Tools
New
Tools
Partners
Volunteered Geographic
Information (VGI)
Utilize new tools
and programs to acquire
address and spatial data in
the most efficient and least
intrusive ways
New and
Enhanced
Programs
Enhanced
Feedback
Engage New
Partners
Address Feedback
adhering to Title 13
confidentiality laws
Build on and Expand
Feedback for Spatial
Features
Research Activities
• GSS-I Working Groups
• Address Summit
– Address Pilots
• External Expert Reports
• Research Project Examples
–
–
–
–
iSimple
GSS Lab Data Viewer
Quality Indicators
Census 2010 Road Update Operations
Evaluation
– Targeted Address Canvassing Continuum
Project/Contract
Management
Quality
Assessments
FY2011 10 GSS-I
Working Groups
Policy
Research and
Development
To date, 11 IPTs formed
Highway Median
“Flag”
Improvements
Parcel Data and
Centroid Use
Feature Coverage
and Sources
Address Coverage
and Sources
Partnerships
MAF/TIGER
Integration/ Linkage
Geocoding
Global Positioning
Systems (GPS)
Problem Capture
Tool
Quality Indicators
Improving Group
Quarters Data
iSIMPLE
Metadata
Improvements
Features Source
Evaluation
The CATT
Better Meeting
MAF “Facility” Data
User Needs
MTAG
Census Address Summit Goals
• Educate our partners about the Geographic Support
System Initiative (GSS-I) and the benefits of targeted
address canvassing
• Gain a common understanding regarding the definition
of an address
• Learn how our partners are collecting, using, and
maintaining address data
Address Summit Participants
Participants - All Levels of
Government
Observations
•
•
•
•
Continuous partnerships are needed and welcome
Public safety is a driving factor for local governments
Urban and rural areas will pose different challenges
Address coverage varies and is sometimes not known or
quantifiable
• Communication and engagement are key
Results of the Address Summit
• Five Pilot Projects
• Address Authority Outreach and Support for Data
Sharing Efforts
• FGDC Address Standard and Implementation
• Federal/State/Tribal/Local Address Management
Coordination
• Data Sharing – Local/State/USPS/Census
• Hidden/Hard to Capture Addresses
2012 Address Pilot Schedule
Moving Forward
These pilots will provide:
• The Census Bureau with a testing ground for future
geographic partnership programs
• The Census Bureau with an opportunity to identify the best
methods for the continual update of the MAF/TIGER
System
• www.census.gov/geo/www/gss/address_summit/
15
Benefits of Establishing an Census
Address Ontology
• Establishing an Ontology allows for
– Effective communication
– Common language
– Ease the burden of data sharing
– Explicit terminology, concepts, and relationships
Expert Research at Census
• Five reports created by outside experts:
– The State and Anticipated Future of Addresses and Addressing
– Identifying the Current State and Anticipated Future Direction of Potentially Useful
Developing Technologies
– Measuring Data Quality
– Use of Handheld Computers and the Display/Capture of Geospatial Data
– Researching Address and Spatial Data Digital Exchange
– http://www.census.gov/geo/www/gss/reports.html
• Summer at Census:
– Steve Guptill; USGS Chief Scientist (Retired)
•
Quantifying the Quality of the MAF/TIGER Database
– David Cowen; Distinguished Professor Emeritus
•
Use of Parcel Data to Update and Enhance Census Bureau Geospatial Data
– http://www.census.gov/geo/www/gss/qaewg.html
• 2 In-Progress Reports
– Change Detection
– Master Address File (MAF) Evaluation
Analysis of the MAF/TIGER System
• iSIMPLE
– Evaluation of road features in TIGER
– Is TIGER consistent with imagery?
– 852,090 grid cells reviewed
• 94% had NO missing features
• 5% had 4 or less missing features
• 70% had NO misaligned features
• 26% had 4 or less misaligned features
– First web service based review
– Research will assist with targeting efforts
18
iSIMPLE Missing Road Features
GSS Lab Data Viewer
• An on-line, interactive mapping tool to facilitate visualization of data
and information
• Examples include:
– 2010 Census Data
• Address Canvassing adds
• Type A adds
• Undeliverable as Addressed
– Delivery Sequence File Statistics
– Natural Disaster Information
20
21
22
Quality Indicators
• Evaluating the current quality of the MTDB
-
Addresses
Features
Geographic areas
Geocodes
• And only evaluate MTDB
• Unit of work is the (current) census tract
23
Address Indicators
• Overall Address QIs
-
Address consistency
Mailability
Deliverability
Locatability
Geocode accuracy
Tests for ‘other’
24
Feature Indicators
• Overall Feature QIs
-
Spatial accuracy
Feature naming
Address ranges
Feature classification
25
Geographic Area Indicators
• For each Geographic Area, four major
tests or sub-indicators
-
Local review/approval of areas
Regional review/approval of areas
Program review/approval of areas
Independent subject matter review/approval
of areas
26
Geographic Area Indicators
• Additional tests for statistical criteria,
attributes, type of submission, contiguity,
etc…
• Also tests for geographic interaction
(slivers), and block size and shape
27
Geocode Indicators
• Combines specific sub-indicators from
each other category
- Locatability and geocode accuracy (Address)
- Spatial accuracy & address ranges (Feature)
- Block size & shape (Geography)
28
Overall indicators & weighting
• Addresses, Features, Geographic Areas,
and Geocodes QIs are then aggregated
according to subject matter formulas
• Each census tract will receive a single
overall score, and category scores where
relevant
• History and tendency will be tracked
29
External sources
• Quality Indicators are MTDB only
• In the future, external sources may also
help determine MTDB quality, such as:
- Population estimates
- Building permits (new development)
- Comparison to Imagery
• Additional tests to check for completeness
of MTDB (omission/commission)
30
Tract profiles
• Additional ability to adjust Quality
Indicators based upon profile elements of
the tract, such as:
-
Natural disaster
Unique address types
Rapidly changing development
Special land use areas
31
Rapid Landscape Change: Picher, OK
• Census 2000:
708 housing units
– 621 occupied
– 87 vacant
• 2010 Census:
30 housing units
– 10 occupied
– 20 vacant
The Result
• All census tracts will be tested and ranked
• Work and updates can then be targeted to
specific areas most in need of update
- Prioritization of internal work
- Prioritization of partner contact and file
ingestion
- Improved resource allocation
33
2010 Road Update Operations Evaluation
Project Scope:
The project evaluated the spatial accuracy of new road
edges added to the MAF/TIGER database (MTDB) by
2010 Decennial Update Operations.
The Decennial Operations in the Study:
•
•
•
•
•
•
Address Canvassing
Update Leave
Update Enumerate
Enumeration at Transitory Locations
Group Quarters Enumeration
Group Quarters Validation
2010 Road Update Operations Evaluation
Hypothesis (1): By using imagery to systematically assess
the spatial accuracy of road edges added by different
operations, we can choose update methods that consistently
produce higher quality linear features.
2010 Road Update Operations Evaluation
Hypothesis (2): Road updates made with GPS were more
spatially accurate than paper-based road updates.
2010 Road Update Operations Evaluation
Project Phases:
1. SQL Metrics – Queried MTDB for counts of new road edges added
during 2010 operations, by county.
2. Sample Design – Worked with DSSD, to design a sample of counties,
as random as possible, that would include all Operations and all
Regions.
3. Spatial Evaluation – Assessed selected edges, overlaid on imagery.
Tested spatial accuracy of the imagery to a CE95 of 5 meters or less.
4. Data Analysis – With DSSD, obtained metrics from observations.
5. Conclusions
2010 Road Update Operations Evaluation
We looked at over 42,000 edges… in 72 counties…….
Conclusions
• Road updates made with GPS were more spatially accurate than paperbased road updates.
•
•
An estimated 90% of road edges added with GPS were spatially accurate.
An estimated 67% of road edges digitized from paper-based operations were
spatially accurate.
• By using imagery to systematically assess the spatial accuracy of road
edges added by different operations, we can choose update methods that
consistently produce higher quality linear features.
Suggestions for Further Study
•
Find other ways to glean what contributes to spatial quality using the data
obtained in this review.
•
Are edges with SMIDs (Spatial Metadata IDs) more likely to be spatially
accurate than edges without SMIDs?
•
Are roads that were named more likely to be spatially accurate than those
not named?
•
Why was the incidence of roads with no name information 38%? Were
road names not collected?
•
Is there a correlation between the use of NAIP imagery and the number of
edges not visible in the imagery because NAIP is collected leaf-on?
•
Is it possible to operationalize or automate this review so that it may be
applied at a larger scale?
•
What other operations that add linear features would benefit from the use
of imagery for quality control?
Targeted Address Canvassing
Continuum
Targeted Address Canvassing
Continuum
Targeted Address Canvassing
Continuum
Targeted Address
Canvassing
Continuum Scores,
Census Tract 6069.04,
Howard County,
Maryland
2010 Base
Overall Score = 93.7
2010 Base: Category
Score
Ratio of 2010 HU counts to 2010 MAF
units
81.5
Percentage of area governments
participating in LUCA (one local
government)
100.0
Type A non-ID adds as percent of total
housing units (Score = 100 – Percent
Type A non-ID adds)
99.5
Mail back rate
81.2
No successful CQR cases (no cases =
score of 100)
100.0
Undeliverable as Addressed (UAA) as a
percentage of total housing units (Score
= 100 – Percent UAA)
91.4
DSF Stability Index
97.3
Ratio of Spring 2010 DSF to 2010 Census
housing unit count
98.5
Total Points
Overall Score
749.4
93.7
Current State: Category
Score
Quality Indicator Score
Percent City Style Addresses
Lack of/presence of hidden units
Targeted Address
Canvassing
Continuum Scores,
Census Tract
6069.04, Howard
County, Maryland
Current State
Overall Score = 96.9
Lack of/presence of informal or unique housing
situations
Lack of/presence of seasonal housing
(Score = 100 = pct seasonal vacant HUs)
Conversion from single to multi-unit or multi-unit to
single (Score = 100 – conversions as pct of all housing
units)
100.0
99.0
100.0
99.5
100.0
Area not known to subdivide single housing units into
multi-unit structures
Lack of/presence of hard to count populations
99.0
Percent MAF TIGER agreement on geocodes
98.0
Percent MAF address confirmation rate (matching rate)
with admin records
Percentage of city-style address MAF units preferred
MSPs
Area classified/not classified as "needs to be canvassed"
in field survey staff feedback
76.3
100.0
DSF Stability Index
GEO Change detection processes indicate no changes
have occurred
97.3
Overall Score
96.9
High Stability Census Tracts
Tract 3406, Harris County, TX
Tract 4302.03, Fairfax County, VA
Category
00-10 DSF Stability
Ratio Fall 09 DSF to 2010
Census HU Count
Value
1.0
1.0
Category
Value
00-10 DSF Stability
1.0
Ratio Fall 09 DSF to 2010
Census HU Count
1.0
Ratio Spring 09 DSF to
2010 Census HU Count
1.0
Ratio S09 DSF to 2010
Census HU Count
1.0
Type A adds
UAA
2010 HU
DSF Spring 11
DSF Fall 10
Census 2000 HU
Ad Can True Adds
Ad Can Deletes
3
7
988
988
988
989
0
3
Type A adds
5
UAA
14
2010 HU
910
DSF spring 11
910
DSF fall 10
910
Census 2000 HU
911
Ad Can True Adds
1
Ad Can Deletes
17
• Questions?