Nation and state level freight generation models

Date of Completion

January 2010


Engineering, Civil




Freight transportation is a critical component of socioeconomic activities. However, freight planning models are not as well developed as passenger traffic models There is an urgent need for freight planning at both the state and national level. The objective of this dissertation is to address several crucial issues related to freight modeling. ^ The availability and quality of freight and truck data are major obstacles for state level planning. The first part of this dissertation contributes to freight planning practice by quantitatively evaluating the merits of currently available data sources and developing a novel state level truck generation model. The study indicates a significant discrepancy among major sources, the Commodity Flow Survey, the TRANSEARCH database, and highway monitoring data for the State of Connecticut. A synthetic OD method was developed to integrate information from various sources to improve the truck origin-destination estimation. This approach provides a novel framework to improve state level truck traffic planning utilizing multiple data sources. ^ A second focus of this dissertation is to model the national level freight generation at county level. Advance statistical models were used to overcome the lack-of-fit problem with ordinary linear models and incorporate the spatial dependency of freight generation. The spatial dependency of freight generation is reflected in (1) the spatial characteristics of the zone, such as distance to transportation hubs; and (2) the fact that zones in close spatial proximity tend to have similar freight production and economic activity. We incorporated the spatial dependency by introducing spatial variables to represent the characteristics of individual zones as well as using a spatial autoregressive model to account for the spatial correlations. The results indicate models that incorporate spatial dependency are a substantial improvement over non-spatial models. ^ The zone system for freight data collection and publishing has significant impacts on freight analyses and planning. However, the majority of existing freight zone systems are provided in an ad hoc way. There is an urgent need for creation of a national level optimal freight zone system. In the third part of the dissertation, a novel model-based optimal zone design method was proposed to design freight zones for the continental United States. This method seeks to directly optimize zone configurations with two components: (1) the optimal criterion which quantitatively represents the desired properties of an optimal zone system, and (2) the shape constraints governing the shape, size, and continuity of individual zones. The method was applied to the continental US with an inter-zonal distance weighted by freight flows as the optimization criterion. Several alternative optimal zone systems for nation-level freight transportation were developed. The comparison of the alternative freight zone systems indicates that a 300-zone system provides an optimal balance between number of zones and the optimization criterion. The methodology and results from this study addressed several challenges related to zone design methodology issues and provide an important reference for national level freight data collection and publishing efforts. ^