-
Notifications
You must be signed in to change notification settings - Fork 2
Home
This wiki is obviously under construction
The inter-regional model focuses upon long-distance freight flows within, to or from, and through the study area. We typically focus on trucks, although all modes of transport could be simulated. The two data sources easiest and most frequently used to describe such flows are the FHWA Freight Analysis Framework (FAF) and the IHS Global Insight Transearch data. The former are open data maintained by the federal government, while the latter is a proprietary offering. They are comparable in scope and coverage, although use somewhat different data, methods, and definitions. Both provide commodity flow estimates, measured in dollars and tons by mode of transport and commodity, between regions of the USA. The process for converting them into the daily flow estimates used in this modeling system are similar:
- The FAF annual flow database contains flows between all FAF regions within the USA. A subset of these flows that move within and through the study area must be extracted, and linked to gateways at the edge of the study area. For example, a model of Oregon would likely include flows with one or both trip ends within Oregon, as well as flows between California and Washington. Note that this step is only required when working with the FAF data, for with Transearch you specify the geography you want your model to work at. They take care of the rest for you.
- The annual tonnage flows by truck must be converted into truckload equivalents. This can be accomplished using a process devised by Battelle for FAF Version 3 or by using average payload weight data from other sources. Again, this conversion has already been carried out in the Transearch data. In either case the flow records must be converted into discrete annual trucks during this step, with the total value and tonnage by commodity split between the number of annual trucks flowing between each pair of regions.
- Empty truckloads must be accounted for. They are explicitly generated in the Battelle process, but must be accounted for by the user if using other means to map the tonnage flows to truckload equivalents. Note that the Transearch data does not include empty trucks, necessitating this post-processing step to add them.
- Most analyses will focus on weekly or daily traffic levels, which requires sampling of the annual truckload equivalents. Weekly analyses are often useful, for many long-distance truck trips take place over several days. Moreover, even large firms often have cyclical shipping patterns, making daily averages an obscure measure. Weekly estimates are easy to sample, for most commodities flow year-round. Thus, the probability of an annual flow occurring during any given week is about 0.019 (1/52). Daily trucks can be sampled from these weekly estimates by assigning a day it is likely to be within the modeled area, and extracting trucks for that particular day. Truck counts and/or establishment survey data can be used to estimate the probabilities for any given day of the week.
- The internal trip end(s) are allocated from FAF region or BEA analysis area (often used with Transearch) to traffic analysis zones. A simple allocation based on employment by zone can be used if desired, but input-output coefficients are typically used in this system to probabilistically assign commodities to the firms that produce or consume them.
The output from these runs is a list of discrete daily or weekly truck trips between traffic analysis zones (to include external gateways) by type of vehicle and commodity. This synthetic population of vehicles – usually restricted to trucks – should be compared to counts at the external gateways. It is highly unlikely they will match within acceptable tolerances, due to a variety of factors:
- The FAF flows, and most likely the Transearch as well, are based on the Commodity Flow Survey (CFS). The CFS is collected from a sample of firms as part of the Economic Census. The sampling frame for the CFS has been successively reduced over its past several cycles due to funding constraints, limiting the number of observations from which to build inter-regional flow estimates. Many of the reported estimates have high variances associated with them, owing to how thinly sliced the data are when tabulated by origin and destination region, commodity, and mode of transport.
- The CFS only covers certain industry groups (manufacturing, as well as some wholesale and mining industries). It is thought that much of the long-distance freight is generated by these industries, but it misses flows generated by shippers outside of those sectors.
- Some commodities have seasonal peaks, and weather conditions along the northern states result in different patterns throughout the year. If these are not taken into account the averages might not represent any given month well, or only a small number of them.
- Many long-distance trucks make multiple stops. The commodity flow data depict inter-regional flows, not truck tours. The incidence of multi-stop long-distance tours is likely less than for trucks operating within an urban area, but the likelihood of multiple stops at the destination end and increasing importance of distribution centers makes the direct mapping of commodity flows to truck tours tenuous.
- The flows measured at external gateways may include local truck flows. The incidence of such flows can be minimized by selecting external gateways in rural areas. This is often more difficult to achieve in the eastern part of the continent, where smaller states and higher density of places makes such buffers harder to define. Intercept surveys conducted at or near external gateways with a mix of flows can help separate local from long-distance flows.
Many analysts develop adjustment factors to account of these differences. In most cases they combine in different ways for each market interchange, necessitating unique factors for each external gateway. The weekly or daily truck estimates are typically factored up or down by these factors. Individual trucks can easily be removed when the factor is less than unity, or additional samples can be taken from the weekly or annual flows when additional trucks are required.
After the final factoring has been done (or omitted) the inter-regional flows should match the observed or projected volumes at the external gateways, minus known or asserted local truck flows across them.