Transferability of Household Travel Survey Data in small areas using the National Household Travel Survey data and the census data

Liang Long, PhD candidate, Advisor: Dr. Lin

Liang-LongHousehold travel surveys play a key but expensive role in transportation planning, with about $200 per household or half million dollars for just a 2,000 household sample. My research objective is to find an alternative solution for scant sample data issues especially in small areas and subpopulation groups (e.g., Hispanic, urban low-income group). The general idea is to borrow data from auxiliary sources such as-national household travel surveys (NHTS) and Census Transportation Planning Package (CTPP) for local areas.
In the study, the statistical clustering approach coupled with GIS spatial analysis is firstly applied to characterize neighborhood lifestyles using sixty-four features extracted from the Census Transportation Planning Package (CTPP) 2000 data. The resulting ten clusters reveal residential location preference as a result of individual or household socio-economic status such as income, occupation, age and ethnics. Travel characteristics of each cluster using the 2001 National Household Travel Survey (NHTS) travel data suggest five factors influencing household travel, socio-economic status, residential location and land use, household life cycle, activity type, and ethnics. Each neighborhood type is distinctively defined and reasonably homogenous in terms of socio-economic and travel characteristics.

Then the feasibility to transfer travel characteristics, i.e. trip rate, mode share, vehicle miles traveled etc. across geographic areas is tested by proposing hierarchical-random effect models. Although there have been similar studies, this is the first to test transferability at the disaggregated level by associating household level and neighborhood level characteristics with travel behavior, rather than the simple, conventional approach of comparing means. Equally important, it has practical values particularly to small metropolitan areas.

In the last step, small area estimation methods are applied to generate totals or means of travel attributes for target areas by using NHTS and CTPP. Three different methods are examined: Generalized regression estimator, synthetic estimator and empirical linear unbiased predictor.

Travel Time Prediction on Signalized Urban Streets Using AVL Buses as Probes

Wenjing Pu, PhD student, Advisor: Dr. Lin

Wenjing-PuProviding travelers travel time and related information is a crucial part of the advanced traveler information system (ATIS). Most metropolitan areas in the U.S., however, do not provide travel time information for urban streets because they are suffering from data unavailability. Utilizing automatic vehicle location (AVL) equipped buses as probe vehicles could be a cost-effective approach to advance ATIS on urban streets. The objective of this research is to develop prediction algorithms for urban street travel time in City of Chicago by utilizing the existing AVL buses as probes so that additional costs invested in traffic monitoring (e.g., loop detectors) on urban streets can be avoided.
The major challenges are:
1) characterizing the relationship between bus travel time and general traffic stream travel time; and
2) predicting bus arrival time and general vehicle travel time.
These problems become even more challenging under congested traffic conditions (both reccurring and non-reccurring). Sufficient network coverage and sample size in term of number of bus probes per link are also important issues in determining the reliability of travel time information. The achievements of this research will benefit both public and private transportation users, increase bus ridership and reduce vehicles waiting time and emissions.

Bus Route Schedule Adherence Assessment Using Automatic Vehicle Location (AVL) Data (MS THESIS)

Peng Wang, MS Candidate, Advisor: Dr. Lin

The thesis demonstrates an optimization method to develop a composite performance index of bus route schedule adherence by combining two elementary metrics: running time adherence and headway regularity together. The optimization method is build based on the data envelopment analysis (DEA) model and it incorporates the use of historical AVL data, which are used to generate the afore-mentioned elementary metrics. The method is applied to assess the service reliability performance of 48 bus route-directions selected from the bus network of Chicago Transit Authority (CTA) and the obtained performance scores are discussed and analyzed.
A research paper is being written with an emphasis of conveying the idea of performance index development using AVL data and DEA methodology, as well as exploring the practical implications.Peng-Wang

Improving the CTA Capital Improvement Program Management (CIPM) Asset Management (AM) System Prototype

Peng is also working as a graduate research assistant at the Urban Transportation Center (UTC) of College of Urban Planning and Public Affairs (CUPPA). At UTC, his current project is making improvements to the CTA CIPM AM System Prototype, which was initially developed by Dong Zhai, a former MS student in CME in 2003. The CTA CIPM AM system is a tool to facilitate and enhance CTA CIPM’s data management capacities. It links with CTA’s asset management database and Geographic Information System (GIS) software. Users can use the system to query data, generate geographic presentations, and conduct spatial analysis.

Modeling Land Use, Bus Ridership and Air Quality: A Case Study of Chicago Bus Service

Minyan Ruan, PhD student, Advisor: Dr. Lin

Minyan-RuanTransit ridership varies at the route, route segment and bus stop level. Previous study on the route level has shown that higher population in non-Hispanic city poor for unit route length will decrease the bus ridership, while longer road length in urban elite for unit route length will increase the bus ridership. The number of stops within the route, which represents the transit service, has the potential to increase the total bus ridership in that route. Also, ridership displays a seasonal trend in the model, with significantly lower passengers in holiday seasons like December and August.

However, the assumption of the route level model is that the land use and demographic characteristics are homogenous, which is apparently not right in most cases. I am now working on modeling the bus ridership and land use in Chicago area in stop level, trying to find out the impact of land use on CTA bus stops. The automatic passenger counter (APC) is now installed in a number of the CTA bus, which covers most of the routes and thus provides a good source of ridership data in the stop level. The variation of ridership across stops depends on the stop-specific variables within the stop service area, which will include pedestrian roadway environment, stop amenities, sociodemographics in land use types. Besides, the transfer passengers from automobile, bicycle, train and other bus stops, will be taken into account by adding variables like the auto parking availability, biking availability, train accessibility and the other bus stop accessibility. Adjacent routes can be complementary routes that will increase the ridership, or competing routes that will increase or reduce the ridership.

A mixed effect model with both spatial and temporal random effect will be built, considering the autocorrelations between stops within the same route and between different time periods within the same route.