Zahra Pourabdollahi

Joint Model of Freight Mode and Shipment Size Choice

Mode and shipment size choice are among the most critical logistics decisions but are typically studied separately in freight demand studies despite their strong correlations. The study described in this paper used an innovative copula-based framework to model freight mode and shipment size choice simultaneously as a joint decision-making problem. The study used a copula-based joint multinomial logit–multinomial logit model in which both mode choice and shipment size were modeled as discrete choices in a multinomial logit structure. The proposed copula-based model was intended to capture the potential effects of observed and unobserved factors that jointly affect both choices. The data used in this study were gathered through a large-scale establishment survey conducted in 2011 in the United States. The survey provided detailed information on more than 1,840 individual shipments that were used to develop the disaggregate models. Results of the estimated model underline the importance of the interrelationship between freight mode choice and shipment size and suggest that common unobserved factors influencing mode and shipment size choices exist. The model provides a better way to capture the effects of observed and unobserved factors that affect both choices simultaneously.

An activity-based freight mode choice microsimulation model

Abstract This study embarked upon the development of a nationwide freight activity microsimulation as an acceptable analysis tool for policy assessments. Mode choice component of a large-scale behavioral microsimulation framework, named Freight Activity Microsimulation Estimator (FAME) was developed and validated in this study. The results of a nationwide establishment survey, discussed in earlier studies, are used to develop the mode choice model for the entire US. Despite many previous freight mode choice models, the proposed model works at the disaggregate level of firm-to-firm. A new concept for firm-types is implemented in FAME to keep the computational burden at a reasonable level and to diminish the need for highly disaggregated data. A total of 45 206 firm-types were synthesized in the US, among which more than 14·8 billion tonnes of domestic shipments were simulated. Total tonnage, value, and tonne-mile of commodities for each mode were obtained as the final output, which showed a satisfactory match with public freight data in the US.

A behavioral freight transportation modeling system: an operational and proposed framework

This paper outlines a new conceptual framework for freight transportation modeling by incorporating more detailed logistics choices into an operational large-scale freight transportation modeling system named FAME (Freight Activity Micro-simulation Estimator). FAME is a micro-simulation model for freight transportation in the U. S. Despite many previous freight transportation models the model simulates commodity movements at the disaggregate level of firm-to-firm, but it also makes certain simplifying assumptions concerning logistics choices such as the use of intermediate handling facilities. This paper deals with issues related to incorporating the new logistics elements in the FAME framework and proposing a new conceptual framework named FAME II.

Shipping Chain Choices in Long-Distance Supply Chains

Shipping chain configuration is one of the key logistics choices that have been ignored or treated insufficiently in the current freight transportation models. This study focuses on the shipping chain configuration and the modeling of its relevant logistics choices, including number of stops and stop types. A brief descriptive analysis of the results of an online establishment survey to explore different shipping chain configurations used by decision makers (business establishments) was done. Also, a system of hierarchical database models of shipping chain choice for freight transportation was presented. A system of decision tree models was proposed to determine the shipping chain configuration of freight transportation by identifying number of stops and type of stops per chain. The results of the survey were used for analysis and model estimation. The proposed decision tree models were developed by using 80% of the observations in the data, and the remaining 20% were used to validate the models. The results of model estimation indicated that shipments’ attributes were the most significant variables in predicting the configuration of the shipping chain. The results of the validation showed that the estimated trees could predict the shipping chain configuration for the test data with an acceptable precision, which confirmed that decision tree models were powerful tools that could be used to predict shipping chain configurations of freight flows.

GPS and Driver Log-Based Survey of Grocery Trucks in Chicago, Illinois

Freight demand modeling lags in comparison with passenger demand modeling, largely because of the limited selection of available data. Commercial firms tend to approach inquiries into the operations and finances of their businesses with suspicion or at least impatience. The risk of losing company time or a competitive edge limits potential cooperation. Thus with each advance in data collection, valuable insights into the inner decision making of capitalist industry can be gained. This paper introduces a GPS freight survey conducted in the Chicago, Illinois, metropolitan area in the spring of 2012. The GPS data were augmented by driver diaries and warehouse and distribution center data records that were designed and used for minimizing respondent burden. Data collection focused on and was made possible through the cooperation of a major grocery chain in the region. In total, 108 trip days of GPS traces were obtained and allowed for the examination of lengthy tours between warehouses, distribution centers, and stores. A descriptive analysis of these tours is included with a focus on the spatial and temporal distributions of activities. From these data, approximately 89% of all activities were major work-based activities, and the number of activities per trip had a local maximum of three, by which 32% of all tours abided. The information gathered and presented will be used to supplement future disaggregate modeling exercises.