Building the 4Ds into Travel Demand Models

1
Building the 4Ds intoTravel Demand Models
Don Hubbard, AICP, PE Senior Associate
2
Topics Covered

• What are the 4Ds and why are they important?
• Insensitivity of conventional models to changes
  in 4D characteristics
• Adjusting a model to reflect the 4Ds a case
  study from Sacramento
• Potential improvements to the methodology
• Conclusions What does it all mean?

3
What are the 4Ds and Why are They Important?
4
What are the 4Ds?
National research has found that certain
characteristics of the built environment tend
affect travel behavior in predictable ways.
These characteristics are

• Density in terms of dwelling units or jobs per
  acre
• Diversity of land uses within any given area
• Design of the pedestrian and bicycling
  environment
• Destinations proximity to regional activity
  centers

5
Why Are They Important?
Because they affect per-capita auto use
Environmental Characteristic Elasticity VT Per Capita Elasticity VMT per Capita
Density 4 to 12 1 to 17
Diversity 1 to 11 1 to 13
Design 2 to 5 2 to 13
Destinations 5 to 29 20 to 51
Sources 4D National Syntheses, Twin Cities,
Sacramento, Location Efficiency
6
and (on a different level) because they have
entered public debate

• Growing consensus that highway construction does
  not solve congestion problems in the long term
• Prominent groups (APA, FHWA) are supporting Smart
  Growth as a way to reduce the demand for road
  space as an alternative to increased supply
• Transportation planners are being asked to
  analyze these potential reductions

7
Insensitivity of Conventional Models to Changes
in 4D Characteristics
8
Blind Spots in Conventional Models Walking
Trips

• Walking trips must use road links, and only
  roads big enough to be in the traffic model
• Sidewalk completeness and other aspects of
  sidewalk condition (shade, aesthetics, etc.) are
  ignored
• Intra-zonal and adjacent-TAZ trips (the most
  important for walk mode) are handled very
  abstractly

9
Blind Spots in Conventional Models Land Use

• No consideration is given to the distances
  between land uses within a given TAZ
• Interactions between different non-residential
  land uses (e.g. offices and restaurants) not well
  represented
• Density is ignored (a TAZ with a dense
  development in one corner is treated the same as
  a TAZ with the same population spread evenly
  throughout its area)

10
Consequences of theBlind Spots

• Model tests of Smart Growth policies usually
  understate benefits things that cannot be
  measured tend to be ignored
• Planners are frustrated by model results that do
  not match field experience
• But if they reject the models results, then land
  use planning and transportation planning proceed
  on separate tracks

Blind spots in models affect real-life choices
correcting the models will reduce the distortions
in the project mix
11
Adjusting a Model to Reflect the 4Ds A Case
Study in Sacramento
12
4D Applications

• Already Done
• Atlanta (Atlantic Steel Site)
• Twin Cities
• East Bay (TBart)
• On-Going Projects
• Humboldt County
• San Luis Obispo COG

13
Background

• SACOG initiated a public visioning process for
  the long-term future of the Sacramento Region
• Smart Growth policies were prominently featured
  in the debate
• However, the regional model (SACMET) was
  insensitive to 4D characteristics
• The model needed to be augmented to enable
  quantitative forecasts of the effects of smart
  growth policies in different scenarios

14
Approach Used

• 4D adjustments were computed as elasticities
  (each change in neighborhood characteristics
  resulted in a certain change in travel
  behavior)
• changes based on differences from a Base Case
• These adjustments were applied to outputs from
  the SACMET model

15
Adjustment Methodology
The 4D adjustment component is shown in blue
16
Data Sources

• VT VMT data came from a large (4,000 HH) travel
  diary survey
• Households, jobs, and developed acres came from a
  parcel database (400,000 parcels)
• Sidewalk coverage and route directness came from
  aerial photographs

17
Regression Analysis
Different formulas were used for different trip
purposes
Some values were not statistically significant
NHB proved to be the most elastic
HBW was the least elastic
18
Problems Encountered

• Some of the relationships were not statistically
  significant
• Drop any below 90 significance
• Elasticity approach does not work well at the
  extremes (vacant-to-non-vacant)
• Use floor and ceiling values to keep results
  within actual experience in region
• Adjustment process very labor-intensive
• Automate data processing

19
PLACE3S Used for Land Use Data
Parcel-level data collected for all of Sac County

• Four land use scenarios
• Current Trends
• Increased Density
• Density with Smart Growth
• Land Use Balance

Two transportation networks (current trends and
Smart Growth)
20
Scenario Themes

• Base Case Continuation of current trends
• Density (only) Build to highest densities
  allowed under existing general plans for
  residential uses
• Density with Smart Growth - Most regional growth
  goes to compact, centrally-located transit- and
  bike-friendly communities. A widespread BRT
  system replaces limited LRT line extensions
• Land Use Balance - Sac County broken into ten
  areas, each with a good balance of residential,
  retail, and non-retail land uses

In each case regional population will
approximately double (from 1.2 million to 2.4
million)
21
Comparison of Transportation Results
22
Results Regional Auto Use
Change from Existing
Scenario Total VT/Day Total VMT/Day
Current Trends 140 120
Density Only 114 89
Dense Smart Growth 91 62
Land Use Balance 111 74
When population doubles, there will be a big
increase in auto use under any scenario
But 4D model shows smart growth policies could
reduce the growth significantly
23
Results Transit Mode
Scenario Down- town Sac County Total Region
Existing 5.5 1.1 0.9
Current Trends 7.5 1.1 0.8
Density Only 14.1 2.6 1.6
Dense Smart Growth 16.3 5.4 3.4
4D characteristics (especially density) make a
big difference in transit use that a conventional
model might miss
24
Results Non-Motorized Modes
Scenario Sac County Total Region
Existing 6.6 6.4
Current Trends 5.1 4.8
Density Only 11.6 8.9
Dense Smart Growth 23.5 18.0
Land Use Balance 13.9 10.6
Again, density and walkable design have major
impacts on the walking mode that would not be
detectable using a conventional model
25
Mode Split for Sac County
Scenario Auto Transit Non- Motorized
Existing 92.2 1.1 6.6
Current Trends 93.8 1.1 5.1
Density Only 84.9 2.4 12.5
Dense Smart Growth 71.1 5.4 23.5
Land Use Balance 83.0 3.0 13.9
4D model does not forecast the demise of the auto
mode, even under the most aggressive scenario.
But it does suggest that a more balanced mode
split is achievable in Sacramento
26
Potential Improvements to the Methodology
27

• Base the regressions on fixed radii rather
  than TAZs

TAZ boundaries tend to separate land use types
Fixed radii more accurately reflect field
conditions
Each TAZs land use seems to be poorly balanced
Also, less arbitrary and more transferable
28

• Shift the Adjustment to an Earlier Stage of
  the Model

Upstream adjustment means that the assigned
volumes reflect the adjustments
Post-processing means that link volumes cannot
reflect 4D affects
29

• Develop Elasticities for the Attraction End of
  the Trip

Conditions at both ends affect travel behavior
30

• Develop a 5th D Related to Transit
  Accessibility

• The Sacramento Case had a very simple mode split
  component
• Reductions in VT were assumed to shift to transit
  or walk/bike
• Split between transit and walk/bike based on
  split in SACMET model (which differs for each
  TAZ)
• Next generation will use elasticities based on
  characteristics of the neighborhoods around
  stations

31

• Develop Elasticities for Other Cities

• Sacramento had the first elasticities by trip
  purpose difficult to gauge transferability
• Regional differences may be significant
  (especially the effect of winter on walking)
• City size may also be significant
• Ultimately would like to have values usable (as a
  first estimation) without local surveys

32
Conclusions
33
What Does It All Mean?

• Conventional traffic models were not intended to
  measure the issues inherent in Smart Growth
• Tests using conventional models tend to show that
  Smart Growth policies would have few benefits
  but field experience suggests otherwise
• We now have a method to make forecasts more
  sensitive to Smart Growth policies
• So cities are now able to have more enlightened
  analysis of Smart Growth

34
Any Questions?Comments?Offers of Data Sets?