Chester G. Wilmot

Sequential Logit Dynamic Travel Demand Model for Hurricane Evacuation

Little attention has been given to estimating dynamic travel demand in transportation planning in the past. However, when factors that influence travel are changing significantly over time, such as with an approaching hurricane, dynamic demand and the resulting variation in traffic flow on the network become important. The decision to evacuate in the face of an oncoming hurricane is considered as a series of binary choices over time. A sequential binary logit model is developed to model the probability that the members of a household will evacuate at each time period before hurricane landfall as a function of the households socioeconomic characteristics, the characteristics of the hurricane, and policy decisions made by authorities as the storm approaches. Data collected in southwest Louisiana after Hurricane Andrew were used to estimate a model that produced dynamic travel demand estimates of hurricane evacuation. On the basis of the results, a sequential logit model appears to be capable of modeling dynamic evacuation demand satisfactorily.

Household Travel, Household Characteristics, and Land Use: An Empirical Study from the 1994 Portland Activity-Based Travel Survey

How a household’s travel behavior is influenced by its socioeconomic and land use factors has been a subject of interest for the development of travel demand forecasting models. This study investigates the relative importance of these factors based on the number of household daily trips and vehicle miles traveled (VMT). The travel data used in the study come from the 1994 Portland Activity-Based Travel Survey. In addition to income, vehicle ownership, and household size, other significant factors in household travel have been identified, such as the presence of car phones, dwelling type, home ownership, and even the length of resident’s time in the current home. Most important, this study has qualitatively revealed that land use makes a big difference in household VMT, whereas its impact on the number of daily trips is rather limited. After controlling for the land use variables, such as density and land development balance, it appears that there is little difference in household income distribution among three different land use areas. The household life stage/lifestyle appears to be more relevant to the residence location. And the land use development of the residence location imposes the greatest impact on the household daily VMT. The results from this study provide some empirical evidence to the development of travel forecasting models. Especially by examining the relationship between land use and household travel, the results shed light on how to incorporate land use factors into comprehensive travel demand models that can be used by policy makers in evaluation of alternative land use policies. This study serves as a step toward more comprehensive studies on transportation and land use. The results presented represent a preliminary analysis of an extensive data set; considerable additional analysis is already in process.

Modeling the Hurricane Evacuation Response Curve

The objective of this study is to develop a hurricane evacuation response curve that is sensitive to the characteristics of the hurricane, the time of day, and the type and timing of evacuation notice issued. Two data sets from past hurricanes, Floyd in South Carolina and Andrew in southeastern Louisiana, were used for model development and testing. A model developed on the Hurricane Floyd data set produced plausible results when it was tested with a series of storm scenarios and different evacuation notice policies. The same model predicted evacuation response behavior for Hurricane Andrew that was similar to observed behavior.

Survival Analysis-Based Dynamic Travel Demand Models for Hurricane Evacuation

Two dynamic travel demand models for hurricane evacuation based on survival analysis are presented: a Cox proportional hazards model and a piecewise exponential model. These models were used to estimate the probability of a households evacuating within discrete time intervals before hurricane landfall as a function of the households socioeconomic characteristics, the characteristics of the hurricane, and management practice in terms of the issuance of an evacuation order. Data from southeast Louisiana collected following Hurricane Andrew were applied. Both models reproduced the evacuation behavior satisfactorily, although the piecewise exponential model was slightly more accurate. Both appear capable of modeling dynamic hurricane evacuation travel demand. However, the Cox model cannot accommodate certain time-dependent variables because of its structure. The piecewise exponential model does not have such a limitation. The piecewise exponential model also has the advantage of estimating the baseline haza...

Methodology to Establish Hurricane Evacuation Zones

A review of practice in hurricane evacuation modeling reveals that the criteria determining the delineation of hurricane evacuation zones have not been clearly defined. In addition, there is no recommended procedure with which to establish hurricane evacuation zones once criteria have been accepted. A set of criteria has been adopted in this paper to design a procedure that mechanically establishes a recommended set of hurricane evacuation zones for an area. The procedure, which is based on a geographic information systems platform, is described, and its use is demonstrated for establishing hurricane evacuation zones for the northern part of the New Orleans, Louisiana, metropolitan area on the north shore of Lake Pontchartrain. The procedure can be applied to any area, and although it is specifically directed at identifying evacuation zones for hurricanes, it could be used for any emergency in which flooding is the major hazard, or it could be adapted to other emergency situations for which evacuation is ...

Comparison of Time-Dependent Sequential Logit and Nested Logit for Modeling Hurricane Evacuation Demand

Models that predict hurricane evacuation demand can play a crucial role in developing and evaluating alternative evacuation policies and plans. However, to evaluate alternative policies effectively, evacuation demand models should be sensitive to time varying characteristics of a storm and the contextual conditions surrounding an evacuee. The time-dependent sequential logit is one such model, but it makes use of restrictive assumptions about the dynamic choices made by evacuees. A new model, a time-dependent nested logit model, relaxes those assumptions. It was formulated and derived in this study, and its performance was then compared with that of the time-dependent sequential logit model by applying both models to data from Hurricane Gustav. The results indicated that the time-dependent nested logit model has better predictive capability than the time-dependent sequential logit model.

Modeling risk attitudes in evacuation departure choices

The decision of whether and when to evacuate can be characterized as decision making under risk. Presently, most models assume linear utility functions through which it is impossible to disentangle factors that influence risk attitudes and other factors that affect decision making under risk. There is a need to disentangle and study factors that affect risk attitudes from factors that affect an evacuees preparation time. The aim in doing so is to provide planners and practitioners with an ability to measure a persons risk attitude and develop appropriate strategies that could motivate people to evacuate. This study is expected to connect the theory of risk developed in economic theory with behavior under threat. The paper uses the Hurricane Andrew response data in conjunction with time-dependent data on the probability of a hurricane strike and the category of the hurricane data to develop a model for evacuation departure choice. A constant relative risk aversion specification is used to model risk attitudes. The process of an evacuation is abstracted as an individual being given a choice between two lotteries: either to stay or leave. The results show that the model is able to predict the total number of evacuees and the time varying evacuation rates with reasonable accuracy. Factors such as time of day, length of time spent in a region, and whether a mandatory evacuation order was issued affected risk attitudes. The presence of children affected the amount of time spent preparing if the family decided to stay.

Household Travel Surveys: Proposed Standards and Guidelines

Clearly some sort of guidance needs to be given in relation to design phases of travel surveys, travel survey instrument design, conducting travel surveys, and the coding and assessment of the data obtained. The benefits of these standards far outweigh the costs involved in implementation. In addition, these issues may apply equally to surveys in general; thus, demonstrating the usefulness of these standards. In this chapter, forty items in travel surveys were described. Recommendations for each item were listed along with justification for the recommendations. Standardization in other areas in survey research still needs to be conducted, thus further research efforts are required. It is also recognized that a number of the items discussed in this chapter are applicable to surveys in the United States and are not necessarily directly applicable in other countries or contexts. However, one of the challenges to the profession is to determine what of these standards can be adopted in other countries, and what needs adaptation to specific circumstances and contexts. Overall, standardization will not only make travel survey results comparable, but will enable the collection of higher quality data, by developing better survey instruments. Good survey design should lead to a reduction in the number of nonresponses, a problematic issue across all fields in social science and behavioral research.

Sequential Logit Dynamic Travel Demand Model and Its Transferability

A dynamic hurricane evacuation travel demand model was estimated by using sequential logit with the data from Hurricane Floyd in South Carolina. The model was estimated on a random sample of 75% of the observations and applied to the remaining 25% as a test. In the test data, a total of 241 evacuations were predicted when 246 were observed, and the model estimated the number of evacuations in each 2-h period over 4 days with a root-mean-square error of 2.79 evacuations. Evacuation orders were modeled as a time-dependent variable. This significantly enhanced model performance over that achieved with evacuation orders as a stationary variable in previous work and provided the capability to analyze the impact of the type and timing of evacuation orders. That capability permits analyzing staged evacuation, in which areas are directed to evacuate in a sequence that optimizes network use. A model estimated on Hurricane Floyd evacuation data was transferred to Southeast Louisiana; its predictions were similar to evacuation behavior observed during Hurricane Andrew. With updating of the alternative specific constant of the transferred model to ensure the correct prediction of the total number of evacuations, the model predicted evacuation with a root-mean-square error of 4.53 evacuations per 6-h period. It was discovered that applying the same distance function to the two different hurricanes was a major source of error in model transfer. The representation of distance and its interactions with other variables need to be investigated further. The procedures and the information needed for model update warrant further study.

TRANSFERABILITY OF TRANSPORTATION PLANNING DATA

The transferability of transportation planning data used to estimate travel-demand models is investigated. It is noted that several traveldemand models can be estimated by using aggregate data, such as trip rates, mode shares, and trip-length frequency distributions. Aggregate data of this type were extracted from the National Personal Transportation Survey of 1995 (NPTS 95) and the North Central Texas Council of Governments Survey of 1996 and compared with those from the Baton Rouge Personal Transportation Survey conducted in 1997. The measures were similar in certain cases and dissimilar in others. Bayesian updating was used to update values from the NPTS 95 survey with values from a small survey conducted in Baton Rouge in 1998. Transferability was marginally improved, although it is expected that the improvement would have been more pronounced if a larger sample of approximately 500 households were available. A panel of 500 households surveyed on an annual basis and used to update transferred data from a variety of sources appears to present a more cost-effective solution to the provision of urban transportation planning data for model estimation than a periodic full travel survey.