You are on page 1of 27

How far have we come? Where are we going?

Mark Bradley FUTURA , November 1, 2011

Out of 25 largest US MPOs


7 already use AB models 11 have AB models in development 7 not yet

Out of 19 California MPOs


10 have AB models in use or in development Includes the 5 largest, plus 5 smaller MPOs SB 375 requires land use

Units

Trips >> Tours >> Person-days >> Household-days >> Longer term

of decisions

choices

Method

Top-down aggregate shares >> Bottom-up micro-simulation

of predicting choices

Amount

of detail that can be accommodated

Socio-Demographic: A few segmentations >> Many variables Temporal: Broad time periods >> Hours or half-hours Spatial: Zones >> Parcels or points

Less

now)

familiar to many potential users (for


3

Households by population segment / residence TAZ

< Trip generation rates > Trips by population segment / trip purpose / residence TAZ < Trip distribution model > Trips by pop. segment / trip purpose / O-D TAZ pair < Time of day factors or model > Trips by segment / trip purpose / O-D pair / time period < Mode choice model > Trips by segment / trip purpose / O-D pair / time period / mode >>>>> Many millions of numbers, mostly small fractions of trips
4

Apply a hierarchical series of models to predict

behavior at several different levels for each representative household and person in the regional population: Work and school locations Auto ownership Household- and person-day activity schedules Tours Trips Origin and destination locations Departure time and arrival time Mode used >>>> Millions of simulated individual trips >> ADD THEM UP
5

URBAN FORECASTING AND PLANNING


Transport Networks Population Attributes

Zonal Attributes

Accessibility

Resident trips
--generation --distribution --mode split

Other trips
--airport --externals --commercial

OD Matrices

Static assignment

Network performance (skims)

TRANSPORT MODEL SYSTEM

URBAN FORECASTING AND PLANNING


Transport Networks Parcel and Zonal Attributes Population Attributes

Accessibility

Population and Long Term Choice Simulator HH/Person day-tour-trip list

Person Day Simulator

AB HOUSEHOLD TRAVEL DEMAND SIMULATOR


Trip Aggregator

Other trips
--airport -externals --commercial

OD Matrices

Static assignment

Network performance (skims)

TRANSPORT MODEL SYSTEM

CT-RAMP

(Vovsha, Freedman, Parsons Brinckerhoff)


DAYSIM

(Bowman, Bradley, Resource Systems Group)


CEMDAP

(Bhat

(U Texas), Goulias (U California), Pendyala (Arizona State U.))

Modeling

response to pricing and Value of Time distributions parking choice behavior

Simulating Modeling

non-motorized travel with land use models with dynamic traffic assignment

Integration Integration Methods

to use finer level spatial detail


9

URBAN FORECASTING AND PLANNING


Transport Networks Parcel and Zonal Attributes Population Attributes

Accessibility

Population and Long Term Choice Simulator HH/Person day-tour-trip list

Person Day Simulator

Other trips
--airport -externals --commercial

AB HOUSEHOLD TRAVEL DEMAND SIMULATOR

Trip Disaggregator

Trip List

Dynamic assignment

Network performance (skims)

TRANSPORT MODEL SYSTEM

Most

regional models were designed to plan major roadways and transit lines that use fine-level spatial detail have been used for localized projects policy environment and regulations (SB 375) require regional models to deal with the same types of localized effects

Models

New

Keep The The

the inputs disaggregate (parcel or block level) and aggregate the outputs

idea is not to use the forecasts at the finest level of detail idea is to model the behavior at its natural scale (avoid aggregation bias) important is to provide reasonable sensitivity to key variables that require spatial detail

Most

Level
Number per region Best used for.

Zones
1,000 4,000 Highway matrices Transit line-haul matrices

Census Blocks
30,000120,000

Parcels
300,0001,200,000 Everything else Bufferbased measures

Density of development

Destinations accessible by foot, bike or transit Diversity Design


Above land uses & parks and other open space Mixed use: entropy-type measures

Density of different land uses in terms of jobs, floor space, households, student enrolment, etc.

Distance to transit

Transit stops for various transit modes Number of 3+ node intersections Number of dead-ends, cul de sacs Fraction of street length with sidewalks, bike lanes Parking supply and price

15

Logistic weight = (1+exp(0.9*-0.5)) / (1+exp(0.9*(Dist/0.25 - 0.5))) Exponential weight = exp(-2.5205*Dist)


1.000 0.900 Cumulative fraction of trips / Buffer weight 0.800 0.700 0.600 0.500 0.400 0.300 0.200 0.100 0.000 Walk Trips Logistic Curve Exponential curve

X-Y straight line distance (miles)

Logistic weight =(1+exp(0.9*-0.5)) / (1+exp(0.9*(Dist/1.25 - 0.5))) Exponential weight = exp(-0.4365*Dist)


1.000 0.900 Cumulative fraction of trips / Buffer weight 0.800 0.700 0.600 0.500 0.400 0.300 0.200 0.100 0.000 .00 .30 .60 .90 1.20 1.50 1.80 2.10 2.40 2.70 3.00 3.30 3.60 3.90 4.20 4.50 4.80 5.10 5.40 5.70 6.00 6.30 6.60 6.90 7.20 7.50 7.80 8.10 8.40 8.70 9.00 9.30 9.60 9.90 X-Y straight line distance (miles) Bike Trips Logistic Curve Exponential curve

More

accurate placement of specific trip types near specific attraction types accurate measures of walk access distance to transit (more complex transit networks require more complex approach) accurate impedance measures for short trips (intra-zonal and nearby zones) for walk, bike and auto modes

More

More

TAZ-based Distance For

values are not very good unless zones are very small from each parcel to the nearest stop is good for simple transit systems more detailed transit networks, use Stop-Stop instead of TAZ-TAZ for line-haul matrices (San Diego, Philidelphia, ), and combine with walk time from each parcel to the stops within walk radius

20

21

NW

NE

SW

SE

Level
Number per region Best used for.

Zones
1,000 4,000 Highway matrices Transit line-haul matrices

Census Blocks
30,000120,000

Parcels
300,0001,200,000 Everything else Bufferbased measures

Does

not require good parcel data

Census

blocks retain a good deal of spatial detail, and are consistent across regions use Census SF1 and LEHD data for HH and employment, control to TAZ totals in San Diego, Los Angeles, elsewhere

Can

Used

Density of development

Destinations accessible by foot, bike or transit Diversity Design


Above land uses & parks and other open space Mixed use: entropy-type measures

Density of different land uses in terms of jobs, floor space, households, student enrolment, etc.

Distance to transit

Transit stops for various transit modes Number of 3+ node intersections Number of dead-ends, cul de sacs Fraction of street length with sidewalks, bike lanes Parking supply and price

You might also like