Presentation - 15th TRB National Transportation Planning

advertisement

Enhancing MOVES Transportation and Air Quality

Analysis by Integrating with Simulation-Based

Dynamic Traffic Assignment

Yi-Chang Chiu, University of Arizona

Jane Lin, University of Illinois Chicago

Suriya Vallamsundar, University of Illinois Chicago

Song Bai, Sonoma Technology, Inc.

TRB Planning Application Conference, Reno, NV

May 9, 2011

Objectives

• To present, through a case study, an integrated modeling framework of MOVES and simulationbased dynamic traffic assignment (SBDTA) model, i.e., DynusT, especially for project level emission analyses

• To share our experience specifically in

– How to integrate a SBDTA model and MOVES

– How to properly run and extract traffic activity outputs from a SBDTA model

– Project level emission estimation in MOVES

– Differences in using MOVES default drive schedule (i.e., specifying only link average speed) versus local specific operating mode distribution input

2

Motivations of Our Study

• MOVES is the new EPA regulatory mobile emissions models for transportation conformity analyses.

• MOVES is capable of much finer spatial and temporal emission modeling than its predecessor MOBILE6

• Few research efforts exist in integrating MOVES with transportation models

3

Literature Review

• Most popular integration of traffic simulation and emission models in the U.S is between the VISSIM and

CMEM (Comprehensive Modal Emissions Model)

– Nam, E.K., C.A. Gierczak and J.W. Butler. 2003; Stathopoulos,

F.G. and Noland, R.B. 2003; Noland, R.B. and Quddus, M.A.

2006; Chen, K. and L. Yu., 2007.

• Integrations between CMEM and other traffic simulation models

– Barth, M. C. Malcolm, 2001; Malcolm, C., Score, G and Barth, M.

2001; Tate, J. E., Bell, M. C and Liu, R. 2005

• Integration between MOVES and traffic simulation models is very limited due to the fact that MOVES is new

– Integration between TRANSIMS and MOVES by FHWA

4

Simulation-Based Dynamic Traffic Assignment

• Iterations between

– Mesoscopic traffic simulation

– Dynamic user equilibrium (vehicles departing at the same time between same OD pair has the same experienced travel time)

• SBDTA retains advantages of:

Macro models – large-scale assignment (but with more realistic congestion patterns)

Micro models – high-fidelity traffic flow dynamics (but

1000+ times faster simulation)

• Improved temporal and spatial resolutions at low computational cost

5

Why Using Dynamic Traffic Assignment to

Support MOVES?

• Assignment is the linchpin between travel demand model and Mobile6/EMFAC

– Capture travelers’ route choice learning network changes.

• This linkage remains crucial when linking MOVES with traffic simulation models

– Without which, vehicles may be at wrong locations at wrong time – misleading VMT and VHT.

– One-shot micro simulation (no assignment) is not consistent with assignment/learning and likely to produce inaccurate and/or counterintuitive results.

– Micro models extracted from TDM sub-area cut may gridlock – OD in TDM not roadway capacity constrained

6

Modeling Demand/Supply Interactions in

Simulation-Based DTA

• Four fundamental transportation system elements

Infrastructure

• Geometries

Traffic flows

• Speed, density, flow, shockwaves, queue

Control systems

• Signals, ramp meters

Information

• Traveler information, message sings

7

Integrated Framework Component I :

( Dyn amic u rban s ystems for

T ransportation)

• Mesoscopic Dynamic Traffic Assignment (DTA)

• Developed since 2002, supported by FHWA, used in

20+ regions since 2005 (Univ. of Arizona)

– SCAG, PAG, MAG, DRCOG, PSRC, SFCTA, HGAC, Las

Vegas, ELP, NC Triangle, Guam, Florida, SEMCOG,

Toronto, SACOG, Mississippi, North Virginia, I-95, US36,

New York, Bay Area)

– 50+ agency/firm/university users internationally

• Open Source in 2011 (http://www.dynust.net)

8

Integrated Framework Component II: MOVES

• EPA’s Next Generation Emission Model

• “Modal based approach” for emission factor estimation

– Four major functions - Total activity generator, Source bin distribution generator, Operating mode distribution generator and Emission calculator

• Data driven model

– Data are stored and managed in MySQL database

• Outputs total emission inventories and composite emission rates

• Three scales of analysis

– National

– County

– Project

9

MOVES Modal Approach

• Associates emission rates with vehicle specific power

(VSP) and speed

• VSP – power placed on vehicle under various driving modes

• Distributes activities using several temporal resolutions

(e.g., hours of day, weekday vs. weekend)

• Classifies vehicles consistent with HPMS data

10

MOVES – Total Emission Estimation

11

MOVES Input Data

• National

– National default database and use of allocation factors

• County

– Use of default data and regional user specific data

• Project level

– Detailed local specific data

Data sources for MOVES project-level application

Travel models

Link characteristics

Driving Pattern

Vehicle Operating Modes

Vehicle Fleet Characteristics

Local source

Meteorological info

Fuel supply

Inspection/ Maintenance

Program

14

MOVES Activity Data from Transportation Models

• Key travel model outputs for emissions modeling

– Volume (or VMT)

– Speed (average for each roadway link)

– Fleet mix (cars vs. trucks)

• MOVES requires data at higher resolution than that is provided by traditional travel demand models

• Literature shows using processed traditional travel modeling data introduces noticeable discrepancies in vehicle emissions estimates

• Activity based travel demand models and simulation based DTA – suited to bridge travel activities and MOVES

Integration: Data Flow from DynusT to MOVES

Data Item Description

Link Roadway link characteristics

(Length, grade, average speed)

Link Drive Schedule

Operating Mode

Distribution

Speed/ time trace second by second

Operating mode distribution defined jointly by speed, VSP

(a)roadway links – optional

(b) off-network link - required

Link Source Type Hour Vehicle fleet composition/ link

Possible Source

User Defined

DTA models

DTA models

DTA models

16

Implementation of Integration (I)

• Two stages are involved in integrating the two components for project level analysis

First Stage

Modifying DynusT to output traffic data as required by MOVES

• Network Parameters

• Fleet Characteristics

• Driving Pattern – Operating Mode versus Drive Schedule Link

• Operating modes - “modes” of vehicle activity with distinct emission rates.

– Running activity has modes distinguished by their VSP and instantaneous speed

– Start activity has modes distinguished by soak time

17

Proposed Integrated Framework

Simulation based

Dynamic Traffic

Assignment Model

Built-in Converter to Link by Link

Operating Mode

Distribution

MOVES

18

Modification to DynusT Traffic Activity Output: Built in

Converter to Link by Link Operating Mode Distribution at Converged Iteration

Move-switch on and output interval in

Parameter.dat

moves_input.dat

At time t , for each vehicle n with prevailing speed V t and previous speed V t-1

Compute

 acceleration/deceleration = ( V t -V t-1 )/SimInterval

Operating mode bin count ++1

Total Count ++1

No t = t + 1

End of Sim?

MovesOut_Links_Hour_1

MovesOut_LinkSourceTypes_Hour_1.csv

MovesOut_opmodedistribution_Hour_1.csv

MovesOut_offNetwork_Hour_1.csv

Yes

MovesOut_Links_Hour_2

MovesOut_LinkSourceTypes_Hour_2.csv

MovesOut_opmodedistribution_Hour_2.csv

MovesOut_offNetwork_Hour_2.csv

Yes

…..

MovesOut_Links_Hour_n

MovesOut_LinkSourceTypes_Hour_n.csv

MovesOut_opmodedistribution_Hour_n.csv

MovesOut_offNetwork_Hour_n.csv

MOVES Excel Input File

Links opmodedistribution

LinkSourceType

OffNetwork

MOVES Excel Input File

Links opmodedistribution

LinkSourceType

OffNetwork

MOVES Excel Input File

Links opmodedistribution

LinkSourceType

OffNetwork

19

Implementation of Integration (II)

Data Item

Second Stage

Identifying sources for and preparing local data

Description Possible Sources

Source Type Age

Distribution

Off- Network

Meteorology

Fuel Supply

Inspection/

Maintenance Program

Vehicle age distribution

Off-network represents TAZs to model start emissions

• Local vehicle registration

• Converted from MOBILE

• MOVES default data

• DTA models/activity based models

Local specific temperature and humidity information

Fuel supply parameters

I/M program parameters

• Local specific

• Converted from MOBILE

• MOVES default data

• Local specific

• MOVES default data

• Local specific

• MOVES default data

20

Summary Features of the Integrated Framework

• Integrated framework: DynusT (DTA) +

MOVES – advantages of DTA over static traffic assignment and one-shot simulation

• Run Time integration with built in converters of traffic activity output from traffic simulation model to MOVES required operating mode distribution format

21

6. Sacramento Case Study (Parts 1 and 2)

• Part 1: improvement vs. baseline

• Part 2: local data vs. MOVES default

22

Case Study Setup: Baseline

• Emission analyses focus on CO

2 from on-road traffic

– Time period: 6-10 AM in a weekday, February 2009

• Downtown Sacramento area

– 437 nodes, 768 links,

– 66,150 vehicles (hourly demand variation: 23/22/18/37%)

– Fleet mix: 90% passenger vehicles and 10% heavy-duty vehicles

– Westbound congestion significant

Source: Google Map

Source: DynusT simulation 23

Case Study Part 1: Improvement Scenario

• Improving freeway interchange to relieve congestion

– Increase off-ramp and downstream interchange capacity

– Signal re-timing for higher off-ramp traffic throughput

Source: Google Map

24

Improvement vs. Baseline : Traffic Activities

VHT (hrs)

VMT (miles)

Total Stop Time (hrs)

Baseline

3,569

148,076

550

Improvement

3,130

141,775

338

% Change

12.3%

4.3%

38.5%

• Both VHT and VMT were reduced (12.3% and 4.3%) due to interchange improvement

• Total stop time was reduced by 38.5% (directly related to changes in operating mode distributions)

25

Improvement vs. Baseline : Traffic Activities

Speed improvement on Business Loop I-80 main lanes

Baseline Improvement

26

Improvement vs. Baseline : Operating Mode

58%

Baseline

19%

23%

Low-speed

Medium-speed

High-speed

62%

Improvement

19%

Low-speed

19%

Medium-speed

High-speed

27

Hour by Hour Comparison

6:00 - 6:59 AM Source Type VMT (mile) CO2e (kg) VMT (mile) CO2e (kg) VMT

LDV 45,309 13,936 44,558 12,983 -1.7%

LDT

HDT

4,553

445

1,909

730

4,596

428

1,877

621

0.9%

-3.8%

Total

Baseline

50,307 16,575

Improvement

49,581 15,481

% change:

Impr. v.s. base

-1.4%

CO2e

-6.8%

-1.7%

-14.9%

-6.6%

7:00 - 7:59 AM

Source Type VMT (mile) CO2e (kg) VMT (mile) CO2e (kg) VMT

LDV

LDT

HDT

86,849

8,954

726

26,031

3,657

1,199

84,392

9,056

851

23,644

3,593

1,309

-2.8%

1.1%

17.2%

Total 96,528 30,887 94,299 28,545 -2.3%

8:00 - 8:59 AM Source Type VMT (mile) CO2e (kg) VMT (mile) CO2e (kg) VMT

LDV 125,784 36,263 121,689 33,532 -3.3%

LDT

HDT

12,825

1,120

5,077

1,719

13,378

1,180

5,098

1,649

4.3%

5.4%

Total 139,730 43,058 136,247 40,279 -2.5%

CO2e

-9.2%

-1.7%

9.2%

-7.6%

CO2e

-7.5%

0.4%

-4.1%

-6.5%

28

Case Study Part 1: Conclusion

9:00 - 9:59 Am

Baseline Improvement % change:

Imp vs. Base

Source Type VMT (mile) CO2e (kg) VMT (mile) CO2e (kg) VMT

LDV 194,152 108,055 190,550 92,848 -1.9%

CO2e

-14.1%

LDT

HDT

20,453

1,802

15,190

5,025

20,346

1,945

13,368

4,758

-0.5%

7.9%

-12.0%

-5.3%

Total 216,407 128,270 212,842 110,974 -2% -13%

• Variation in VMT and CO

2 emissions (total and by source type) are consistent over the fourhour period

• CO

2 emissions benefit in the improvement scenario is related to:

– VMT reductions

– shift in operating mode distributions (reduced stop time and improved travel speed)

29

Case Study Part 2: Local vs. Default Data

• MOVES default drive schedule vs. user-supplied operating mode distribution

– How much difference in emissions estimates?

• Use of MOVES default drive schedule

– Easy to implement in practice

– Potential limitations

• Use of project-level operating mode distribution

– Requires data preparation and conversion

– Presumably more appropriate for emissions modeling

30

Comparison Scenarios Setup

• Using the same baseline scenario as presented previously for the Sacramento case study

• Running MOVES in separate runs with

1. Link average speeds, i.e., using MOVES default drive schedules, to replace user supplied operating mode distribution

2. User-supplied operating mode distribution, i.e., the baseline scenario in the previous case study

31

Comparison Results

6:00 - 6:59

Baseline

(Op. Mode Distribution)

Baseline

(MOVES default Drive Schedule)

Source Type VMT (mile) CO2e (kg) VMT (mile)

LDV 45,309 13,936 45,309

LDT

HDT

4,553

445

1,909

730

4,553

445

Total 50,307 16,575 50,307

CO2e (kg)

16,359

2,401

941

19,701

% change:

Default vs. Op Mode

CO2e

17.4%

25.8%

28.9%

18.9%

7:00 - 7:59

Source Type VMT (mile) CO2e (kg) VMT (mile)

LDV

LDT

HDT

86,849

8,954

726

26,031

3,657

1,199

86,849

8,954

726

Total 96,528 30,887 96,528

CO2e (kg)

30,821

4,649

1,543

37,013

8:00 - 8:59 Source Type VMT (mile) CO2e (kg) VMT (mile)

LDV 125,784 36,263 125,784

LDT

HDT

12,825

1,120

5,077

1,719

12,825

1,120

Total 139,730 43,058 139,730

CO2e (kg)

43,816

6,566

2,353

52,736

CO2e

18.4%

27.1%

28.7%

19.8%

CO2e

20.8%

29.3%

36.9%

22.5%

32

Comparison Results (cont’d)

9:00 - 9:59

Baseline

(Op. Mode Distribution)

Baseline

(MOVES default Drive Schedule)

Source Type VMT (mile) CO2e (kg) VMT (mile)

LDV 194,152 108,055 194,152

LDT

HDT

20,453

1,802

15,190

5,025

20,453

1,802

Total 216,407 128,270 216,407

CO2e (kg)

101,146

15,086

3,994

120,226

% change:

Default vs. Op Mode

CO2e

-6.4%

-0.7%

-20.5%

-6%

• Q/A check: VMT by source type remains the same;

• Results for the first 3 hours: using MOVES default drive schedules yields much higher CO

2 emissions;

• Results for hour 4: pattern is opposite.

33

Using MOVES Default Drive Schedules

Source: User Guide for MOVES2010a (EPA, 2010), pp 66.

34

Part 2: Conclusion (Local vs. Default Data)

• In this case (especially hour 4 results), for links with speed below 5.8 mph, MOVES does not provide HDV emissions if default drive schedules were used.

• Similar situation for LDV emissions (speed < 2.5 mph)

• The missed emissions associated with low-speed links contributed to underestimation in MOVES when using default drive schedules.

• Using local-specific data under a highly congested condition seems important to produce more consistent results than using default drive schedules.

35

Overall Summary and Next Steps

• An integrated modeling framework of DynusT and

MOVES - connecting and automating the modeling process from DTA to MOVES project-scale applications

• Advantages of the integrated model in policy analysis

• Using local-specific traffic activity inputs and operating mode distributions is important

• MOVES default drive schedules are convenient to use but may become questionable when modeling highly congested traffic; further investigation is needed.

36

Future Research

• Use DynusT project-specific drive schedules in

MOVES modeling

• Compare static traffic assignment with dynamic traffic assignment for emissions modeling

• Conduct a series of sensitivity analyses with selected traffic and MOVES parameters

37

Acknowledgments

• This research is part of the TRB SHRP C10 project led by Cambridge Systematics, Inc.

• This study is a joint effort among:

Dr. Song Bai, Sonoma Technology, Inc. sbai@sonomatech.com

Dr. Yi-Chang Chiu, University of Arizona chiu@email.arizona.edu

Dr. Jane Lin, University of Illinois at Chicago janelin@uic.edu

Ms. Suriya Vallamsundar, University of Illinois at Chicago svalla2@uic.edu

38

Download