Status of the SEMCOG E6 Travel Model SEMCOG TMIP Peer Review Panel Meeting December 12, 2011 presented by Liyang Feng, SEMCOG Thomas Rossi, Cambridge Systematics.

Status of the SEMCOG E6 Travel Model

SEMCOG TMIP Peer Review Panel Meeting

December 12, 2011

presented byLiyang Feng, SEMCOGThomas Rossi, Cambridge Systematics

presented to

Objectives

Improve key modeling components as needed to analyze key projects and policies

Reflect the most recent available data

Implement 2004 TMIP peer review recommendations

2

E6 Model Components

Data Sources

2004-2005 household survey (SEMCOG, MI Travel Counts)

2010-2011 transit on-board survey

SEMCOG traffic count database

Information from transit providers (ridership counts, schedules)

4

Work to Be Done for All Model Components

Model estimation

Application programming (TransCAD)

Validation at component level

5

E6 Status

6

Model Component

Objectives Status

Trip generation

Rates reflect most recent survey, explain differences in subareas

Awaiting final validation

Trip distribution

Parameters reflect most recent survey, test destination choice formulation

Gravity model complete

Time of day Consistency with most recent survey and analysis needs for highway and transit

Awaiting final validation

Transit model

Consistency of parameters throughout process, reflect recent transit survey

Awaiting final validation

Mode choice Appropriate for New Starts analysis, capability to analyze proposed new transit services

Estimation nearly complete

Commercial vehicle

Better reflect current commercial vehicle/truck movements in region

Estimation nearly complete

Trip Generation

Identified ways to improve the trip generation rates

Home based university trip purpose added

Parameters updated using household survey data

Factors used to separate non-motorized travel

Air passenger model updated

7

Trip Generation

Income segmentation by quartiles for HBW, HBShop, HBO – primarily for environmental justice analysis

HBSchool not sensitive to income – persons x children

HBU – Trip rates/person to 25 largest colleges by type by distance

Attractions – reclassified employment types (Basic, TCUW, Retail, Service, Educational, and Government)

HBU Attractions – based on total enrollment minus group quarters population

8

Trip Generation Validation

Initial validation showed trip productions in Monroe and Livingston Counties substantially overpredicted

Calibrated area type adjustment factor (rural/non-rural)

Further adjustments regarding external travel to be performed during system calibration

9

Trip Distribution

Gravity model parameters recalibrated by trip purpose (income segmentation for HBW, HBShop, HBO)

Logit destination choice model to be estimated» Using the most recent data, test whether destination choice

model produces better results» If so, implement and validate logit destination choice model» If not, revalidate existing gravity model using recent data

10

Time of Day

New time periods defined…» Definitions useful for both highway and transit analysis

Factors reestimated using household survey data

11

Period Definition

AM 6:30-9:00 a.m.

MD 9:00 a.m.-3:00 p.m.

PM 3:00-6:30 p.m.

Evening 6:30-10:00 p.m.

Overnight 10:00 p.m.-6:30 a.m.

Time of Day Factors

12

AM MD PM Evening Overnight1HBW From home 29.9% 11.0% 3.4% 1.4% 8.9%1HBW To home 1.2% 7.4% 25.5% 6.8% 4.5%2HBO From home 13.7% 18.5% 13.7% 7.2% 1.4%2HBO To home 3.9% 12.0% 14.2% 12.8% 2.8%

3HBSH From home 3.2% 19.7% 10.7% 6.6% 1.0%3HBSH To home 0.7% 22.6% 20.7% 13.5% 1.3%

4HBSCH From home 54.2% 3.1% 0.7% 0.3% 0.5%4HBSCH To home 0.1% 16.3% 23.9% 0.8% 0.1%

5HBU From home 18.2% 21.3% 9.7% 1.9% 0.8%5HBU To home 0.0% 16.4% 13.1% 16.9% 1.8%

6NHBW From work 2.1% 30.3% 26.6% 4.2% 1.3%6NHBW To work 12.2% 18.2% 3.2% 0.8% 1.1%7NHBO All 7.8% 40.9% 33.1% 16.3% 1.9%

Transit Model

Focus on transit network parameters and path building processes

Parameters:

13

Used new on-board survey data» Compared paths between survey and model» Adjusted path building settings to improve match

Transit Model Speed Definition

Using 2010 data, SEMCOG did a comparison between model auto time and scheduled bus time for 145 routes for AATA, DDOT, and SMART

Initial analysis adjusted to account for systemic differences

Stop (dwell) time adjustments by operator

14

Scheduled_bus_time = 0.917 * (Model_autotime) + .318 * (Model_stops)

Transit Walk Access Time

E5 model – Walk access capped at 18 minutes

Examined on-board survey data

Recommended increase to 36 minutes (about 90% of observations after eliminating outliers)

15

Transit Network and Path-Building Procedure Checks

Reviewed survey data boardings and determine prevalence of reported multipath transit use

Checked that all transit routes have non-zero flow

Constructed aggregate prediction success table of the reported boardings per passenger trip with modeled boardings of paths (prediction success rate = 73%)

Modified path building parameters to improve the path building prediction success outcome

Recommended allowing park-and-ride in off-peak to better balance daily O-D

16

Mode Choice

Existing mode choice model needs to be evaluated:» Range of current and potential transit services » FTA New Starts analysis» Project impacts on population segments» Incorporation of transit model improvements» Use of recent data (counts, surveys)» Efficiency of model structure and procedures» Validity of results

Recommendations for structure, parameters of mode choice model to be implemented

Reestimate/revalidate

17

Mode ChoiceNesting Structure Tests

18

Mode ChoiceNesting Structure Tests (continued)

19

Handling New Modes in Mode Choice Application

Arterial Rapid Transit (ART)

Bus Rapid Transit (BRT)

Light Rail (LRT), including on Woodward

Commuter rail (CRT) from Detroit to Ann Arbor

20

Commercial Vehicle Model

Three-step model – generation, distribution, assignment

Prepared vehicle classification count data – adjusted for growth/decline in region

Adjusting parameters to reflect current data

Adjustments to reflect changes in external station volumes

Revalidating

21

System Calibration

Validate individual components as they are developed

Use recent data to see “what has changed”» Enhance short-term forecast capability

Get the “big picture” correct

Examine “trouble spots” from previous model versions

Make sure forecasts make sense

Expected completion – March 2012

22