The availability of large samples of truck GPS data has presented a new, unprecedented source of information for understanding truck travel patterns and forecasting truck demand. The American Transportation Research Group (ATRI) now is now collecting roughly 1 billion GPS data points per week, representing several hundred thousand individual trucks out of the 2.4 million trucks registered in the US. This data is now being incorporated in the development of statewide models, most recently Tennessee’s. This presentation will share the process by which the raw GPS data is used to produce an expanded truck trip table for use in Tennessee’s new statewide model.
Given the size of ATRI’s database, it was necessary to draw a sub-sample of their data. The sub-sample for Tennessee was drawn from eight weeks in 2013, spread over all four quarters. This yielded data on over 234,000 individual trucks making over 6.5 million trips. Based on Tennessee DOT estimates of truck VMT, this represents roughly 11% of the multi-unit trucks on the road in Tennessee for 56 days.
Of critical importance in the development of an expanded trip table from this data is the development of expansion factors to expand the sample data to represent all truck movements. This is challenging, since the sample is not randomly drawn and therefore cannot be presumed to be representative. For instance, previous work in Iowa confirmed that short-haul movements, while present in the data, are under-represented. Without correcting for this, it is not possible to produce accurate information regarding average trip lengths, etc., or to accurately forecast future truck activity. This presentation builds on previous work in Iowa to understand the representativeness of the American Transportation Research Institute’s (ATRI) truck GPS data and develop methodologies to produce factors for expanding it to ensure it is representative of truck travel patterns in general. The new application in Tennessee both corroborates initial findings in Iowa and sheds new light on the variability of sampling rates geographically.