Jennifer Symoun
Good afternoon or good morning to those of you to the West. Welcome to the Talking Freight Seminar Series. My name is Jennifer Symoun and I will moderate today's seminar. Today's topic is Freight Analysis Framework 3. Please be advised that today's seminar is being recorded.
Before I go any further, I do want to let those of you who are calling into the teleconference for the audio know that you need to mute your computer speakers or else you will be hearing your audio over the computer as well. I'm also going to bring up a poll to get an idea of how many of you are using the voice over IP option versus how many are using the teleconference option.
Today we'll have three presenters, Mike Sprung of the FHWA Office of Freight Management and Operations, Diane Davidson of the Oak Ridge National Laboratory, and Maks Alam of Battelle.
Mike Sprung manages the Freight Analysis Framework program for the FHWA, Office of Freight Management of Operations. Mike's efforts are focused on the integration and analysis of freight transportation data from a variety of sources to estimate commodity flows and related transportation activity among states, regions, and international gateways. Prior to his work at the Federal Highway Administration, Mike managed the international trade and transportation data program at the Bureau of Transportation Statistics, and has also worked as a survey statistician on the Commodity Flow Survey at the U.S. Census Bureau.
Diane Davidson has 30 years of experience in executive leadership, transportation management, transportation planning and operations. She currently serves as the Director of the Center for Transportation Analysis (CTA) at Oak Ridge National Laboratory (ORNL). She previously served the state of Tennessee as Director of Rail, Transit and Waterways from 2004-2007. She founded and directed a regional non-profit organization, the Transportation Management Association Group (TMA Group), which became nationally recognized for excellence in transportation innovation, demand management and operations. She served on the Nashville Metropolitan Planning Organization (MPO) Technical Coordinating Committee from 1988-2004. While at TMA, she founded and served as General Manager of the Franklin Transit Authority. Diane received a Master of Science in Urban and Regional Planning degree from the University of Tennessee. She is the Project Manager for the national Freight Analysis Framework version 3, for the Federal Highway Administration. Other key areas of research expertise include rail and waterway freight; multi-modal planning; transit planning and operations; sustainability; demand management; Intelligent Transportation Systems; and, travel behavior.
Mohammed Maks Alam is a research leader for the Battelle Transportation Group and has more than 25 years of progressive experience in civil engineering practice, education, training and research in pavement management, traffic engineering and transportation planning, transport policy as it relates to freight modeling highway infrastructure management, construction and contract administration, traffic information systems, Geographical Information Systems (GIS) and project management in Canada, US and Asia. Mohammed has published a number of papers related to freight modeling, highway inventory and traffic data integration and management as it relates to decision making and GIS applications in highway and transportation engineering. Since its inception, Maks has been working for FAF as the principal investigator for the freight assignment and the capacity component of FAF.
I'd now like to go over a few logistical details prior to starting the seminar. Today's seminar will last 90 minutes, with 60 minutes allocated for the speakers, and the final 30 minutes for audience Question and Answer. If during the presentations you think of a question, you can type it into the chat area. Please make sure you send your question to "Everyone" and indicate which presenter your question is for. Presenters will be unable to answer your questions during their presentations, but I will start off the question and answer session with the questions typed into the chat box. Once we get through all of the questions that have been typed in, the Operator will give you instructions on how to ask a question over the phone. If you think of a question after the seminar, you can send it to the presenters directly, or I encourage you to use the Freight Planning LISTSERV. If you have not already joined the LISTSERV, the web address at which you can register is provided on the slide on your screen.
Finally, I would like to remind you that this session is being recorded. A file containing the audio and the visual portion of this seminar will be posted to the Talking Freight Web site within the next week. We encourage you to direct others in your office that may have not been able to attend this seminar to access the recorded seminar.
The PowerPoint presentations used during the seminar are available for download from the file download box in the lower right corner of your screen. The presentations will also be available online within the next week. I will notify all attendees of the availability of the PowerPoints, the recording, and a transcript of this seminar.
One final note: Talking Freight seminars are eligible for 1.5 certification maintenance credits for AICP members. In order to obtain credit for today's seminar, you must have logged in with your first and last name or if you are attending with a group of people you must type your first and last name into the chat box. I have included more detailed instructions in the file share box on how to obtain your credits after the seminar. Please also download the evaluation form from the file share box and submit this form to me after you have filled it out.
We're now going to go ahead and get started. Today's topic, for those of you who just joined us, is Freight Analysis Framework 3. Our first presenter will be Mike Sprung of the FHWA Office of Freight Management and Operations.
As a reminder, if you have questions during the presentation please type them into the chat box and they will be answered in the last 30 minutes of the seminar. Before we begin our presentations I will turn it over to Tony Furst, director of FHWA Office of Freight Management and Operations.
Tony Furst
Thank you, Jennifer. It is a pleasure to kick off this session of Talking Freight. As many of you know FAF is the most comprehensive data in the U.S. It covers our entire transportation network and all modes. The manner which we do all the work is open and available to anybody who wants to take a look at it. We're always open to suggestions on how to improve the services that FAF provides both at the state level, national level and local level. Before we get started I want to acknowledge all the hard work that Mike and his team of contractors have done. Everybody likes to take a look at the maps that you see currently on the slide which shows the freight flows throughout the country. That front end of the framework is what we use in a whole host of settings. Underneath all of those maps there is an inordinate amount of number crunching and data work that, I don't want to call Mike and his team contractors mechanics under the hood, but they are in the nuts and bolts of FAF working on how to get that data compiled, put together, analyzed and flowed over the network. There is a huge amount of work to get these maps published. I appreciate it and want to thank the team for the work they have done. Today's session will walk you through some of the nuts and bolts about how FAF3 has changed from FAF2. It continues to evolve and improve. With that I will turn it over to Mike. Again, thank you to Maks, Diane and all the others involved with this effort.
Mike Sprung
Thanks a lot Tony. I certainly appreciate being acknowledged for that and I also appreciate you acknowledging the excellent contractors that worked with me through this whole thing and that have made this possible. So let me start off by defining the FAF a little. Tony gave us a good intro. I may be a little repetitive in what I have to say, but the Freight Analysis Framework version three, FAF3, is a freight transportation data program designed to provide a comprehensive picture of freight movement in the United States. We frequently refer to it as a model because of the traffic assignment procedure which made those traffic flow maps of FAF1 and FAF2 so famous. As Tony pointed out, there is a lot more to FAF than just the model, although that is big piece of it. The database making that traffic assignment possible is really the result of an enormous data compilation that results is the only publicly available comprehensive data for basically the entire country.
FAF was designed to provide an understanding of the current and future pressures of freight movement on the transportation system. These are some examples of the type of issues that need to be understood: how freight movement affects congestion, infrastructure, safety, and the environment; how congestion, delay and cost affects freight movement; how shifts in the economy affect freight movement; how policies affect freight movement and what are the consequences of those. Where can investments be made to improve operations or increase capacity? Decision-makers need to understand these in order to guide investment policies that will address expected growth in the demand for freight as well as to keep the economy moving.
This understanding is built on answers to seemingly simple questions made very complicated by the size of our country and the transportation network. How much, what is moving, how it's getting there, and when does it move? With these answers, we can start to link freight to the benefits and consequences and ways to develop ways to support the positive and mitigate the negative. FAF3 answers the how, what and where questions for the previous slide at a national scale for 2007 and provides forecasts for every five years out to 2040. The Commodity Flow Survey done by the Census Bureau as part of the economic census is the largest source of input data and provides the majority of domestic freight movement data, but it's only conducted every five years so we also have estimates for the most recent years based on trends from specific sources, international trade, and economic data. We also use the origin destination matrix to conduct network assignment, regional flows, and that allows us to display trucks trips, tonnage, congestion and such on the National Highway network. Since it only deals with annual averages, we have to turn to other data sources for when and daily and seasonal variations in freight activity. One such source is the Freight Performance Program which is also managed by the Federal Highway Administration's Freight Office. They monitor truck activity continuously on interstate highways and border crossings on the National Network.
This shows all of the data elements included in FAF3 which includes the value and weight for all domestic shipments, exports, and imports. We have eight domestic modes: truck, rail, water, air, multiple modes and mail, pipeline, other, and unknown. The eighth one we added since the FAF3.0 data which we released in July, a no-domestic mode category. If there's a shipment of crude petroleum that does not move past the port of entry, we needed a way to handle that and the no-domestic mode category was our solution. We also have seven foreign modes that match the domestic modes minus the last one I just mentioned. There are 123 domestic regions taken from the regions used in the Commodity Flow Survey. We also have eight international regions, Canada, Mexico individually and six groupings based on U.N. definitions. We get the commodity classifications from the Commodity Flow Survey; there are 43 2-digit SCTG codes. This is a map of the regions that are included in FAF3. We end up with 131 regions which cover state portions and large metropolitan areas. There are 11 new metro areas added. If you used FAF2 in the past, we had to use additional international gateways that differed from the regions used in the Commodity Flow Survey. This time the CFS incorporated the 11 new metro areas, nine of which focused on international gateways that were selected for the volume of foreign goods moving to and from the U.S. Now our regions are directly in line with the Commodity Flow Survey. I should also note while I'm talking about international shipments, the FAF3 also identifies where international freight flows enter and leave the country, so we'll have a foreign region and a U.S. region where the flow enters or leaves and we'll have a domestic origin or destination as well.
The bottom line is that FAF provides the only comprehensive national picture of freight flows and major corridors. It shows states and localities where they fit in the national and global freight system. This is essential in most places where the majority of freight is inbound and outbound rather than local. While FAF does an excellent job of illustrating freight flows at this level, it does need to be supplemented with local data for planning of local projects. And with that I'm going to pass the microphone to Diane from Oak Ridge National Lab who will tell about how we assembled this origin-destination matrix. Diane, are you ready?
Jennifer Symoun
Diane, if you could hold on one second I think we're having audio problems. Hold on a second and let me see if I can reconnect. Give me one second. We'll see if this works, you can go ahead.
Diane Davidson
Can you bring up my presentation? Thank you. ORNL participated in FAF2 and FAF3. We are pleased to be on this panel today. Mike has covered a lot of the basics on purpose, mode and geography, so I will pick up from there. The origin-destination-commodity-mode (O-C-D-M) database is the most comprehensive database available to examine geographic relationships between freight movements by mode and commodity for tons and value. Great use can be made of this because it provides estimates of total volumes of freight that is moved into, out of and within the U.S., sub-state regions and international gateways. The 2007 FAF flow dimensions include 131 flow origination regions, 131 flow destination regions, 43 commodity classes at the SCTG two digit level and seven (7) modal classes (or eight in domestic, as Mike described earlier). When you put all this together, the full FAF3 flow matrix has almost 10 million data sets which need to be estimate or suppressed. That's a large job and that's the work of ORNL performs for FHWA.
The FAF3 O-D-C-M annual freight flows matrix is used to develop a number of additional FAF3 data products, including the assignment of spatially disaggregated flows to specific links on the transportation network, for the purpose of ton-mileage computations and infrastructure capacity analysis. Maks Alam will speak about the network assignment methodology later in the Seminar. Additional products are the long range freight flow forecasts, which have a future target year of 2040, with the projected truck flows being assigned to (i.e. routed over) a representation of the future year U.S. highway network; as well as the provisional Annual Update of the 2007 O-D-C-M freight flows matrix, which is based on the latest data available on mode and commodity specific economic activity and traffic growth. I do want to mention a new product that's coming your way which is an on-line, web-based tool for extracting data elements from FAF3. Mike will discuss this after Maks' presentation.
Now we will move on and talk about how this O-D-C-M annual freight flow matrix is constructed. I'm going to give a very high level overview of these major data modeling steps involved with generating the freight flow matrix. FAF draws from many data sources but the most important is the Commodity Flow Survey (CFS). As you know, the CFS is conducted every five years as part of the U.S. Economic Census, with major funding from the Bureau of Transportation Statistic (BTS). Data are collected on all shipments from the surveyed establishment for an entire week in each of the four quarters of the census year. It reaches a subset of all the American establishments and participation is mandatory. Notably, in 2007, about twice as many establishments were recorded as in 2002, so we had about 100,000 samples. Even with this larger sample, the CFS is missing about 30% of the total FAF in volume and 32% in tons. Therefore, the work starts with a very large and very sparse O-D-C-M table - only about 5% of the cells contain non-zero flow values. So we begin with the In-Scope Flow, meaning they are multi-modal commodity flows by weight and value that are contained in the CFS. Since the O-D-C-M CFS table is so sparse, we need an approach or methodology that will fill the gaps. The approach taken is to use a variety of missing value inferencing techniques, with the results of applying spatial interaction and other data models to a variety of carrier reported and economic activity data. That process yields an estimate of about 40% of the tonnage of the U.S. freight movements that are not captured by the CFS. If there are no observations in the CFS, the zero value cells are treated as zero and are constrained to be so. However, many CFS sales are reported as "suppressed" for reasons of confidentiality or statistical reliability, and therefore need to be estimated. So the estimation of the missing cell values occupies a great deal of time and effort. We use a three-step process to accomplish this process. Step 1 begins with a Log-Linear / Iterative Proportional Filling (IPF) routine that uses a variety of methods to identify sampling versus structural zero-valued cells. This maximum likelihood based LL/IPF procedure was developed by ORNL to fill gaps in the data set. The CFS geographic zoning system is adopted as the basis for the FAF3 zoning system. Then, Waterborne Commerce data (U.S. Corps of Engineers), Public Use Railcar Waybill data (Surface Transportation Board) and the 2002 CFS (BTS) are used as a reference to estimate the value of the suppressed cells in 2007. These data help to determine cells that are truly a zero; i.e. were there shipments that show up in other data sources? Next, the log-linear model fills in estimates of the suppressed data cells. In Step 3, iterative proportional fitting (IPF) routine is performed, which constrains the FAF3 flow matrix to be consistent with the CFS totals. The cells with non-suppressed data in the CFS are constrained to remain unchanged during this process. After conducting these three steps in an iterative process, which entails running the model several times, the enhanced CFS flow matrix is fully populated at an O-D-C-M four-dimensional level, in terms the annual tonnage and annual value of goods transported.
However, at this point it does not include the flows that are Out-of-Scope or undercounted in the CFS. So, the next step is to address the flows that are OOS to the CFS. These OOS commodities are: Farm-based agricultural shipments, Fishery, Crude Petroleum, Natural Gas, Municipal Solid Waste, Forestry, Construction, Retail, Services, Household and Businesses Moves, and Imports. Commodity and mode specific OD flows are assigned to these commodities using other methods, and as you would expect, many different data sources have to be accessed and a variety of different applications are used. Time prohibits going over all these treatments but I will give an example of the farm-based agriculture shipment OOS commodity flow for which we used the USDA 2007 Agriculture Census and applied estimation methods to impute missing values, based on related known information. CFS origin and shipping-distance distribution, based on VIUS, were used to identify possible destinations, by agricultural commodity group, to distribute the amount of products to each FAF destination zone. That's just one example of a process and each out-of-scope as its own unique treatment. The documentation that is available on the Federal Highway Administration FAF website will present all the various treatments of the out-of-scope flows.
Next, we estimate the OOS domestic truck flow, using four primary steps. The first step is to estimate national or regional shipment totals for each industry by the FAF commodity classes. VIUS, Census data, and National Input-Output, BEA Make and Use tables are applied here. Input-output data is used to convert OOS industrial sector inputs and outputs to FAF3 commodity inputs and outputs. In step two, national shipments are regionalized by ton and value at the U.S. level down to the county. We go to County Business Pattern Sales and Employment data for performing this kind of regionalization. Step three is to estimate O-D flows at the county level using spatial interaction modeling. Step four aggregates the O-D estimates from the counties back up to the regional zones. So at the end of this process we have FAF OOS region-to-region, origin-destination commodity mode flows by industry.
Because the CFS only captures outbound shipments from domestic firms, it does not include movements of foreign imports. So foreign trade data, including land crossings between the countries in North America have to be added to the CFS. Imported as well as exported freight flows in FAF3 are constructed from mode specific data sources, each of which must have its flows converted from HS commodity codes to FAF3's 2-digit SCTG codes, as well as have its flows either spatially aggregated or disaggregated to match FAF3 analysis zones. There are four major areas that require estimation.
When these four imported and exported OOS freight flow procedures are concluded, the commodity and the spatial zoning system crosswalks are applied to the various data sets to match the FAF commodity classes and the 131 zones. Data specific flow modeling and data gap filing procedures are performed to yield the FAF Foreign Origin (O)-Destination (D)-Commodity(C)-Mode (M) flow matrix by annual tons and 2007 dollars.. CFS in-scope flows, enhanced by log-linear and IPF procedures, are joined with the domestic out-of-scope treatments and are merged with the Foreign O-D-C-M matrix to create the full 2007 FAF3 O-D-C-M matrix.
I just want to say what a great improvement this PIERS data made in the ability to estimate the allocation of waterborne imports and exports to the FAF domestic zones of freight origination, for U.S. exports, and destinations, for U.S. exports. Other significant improvements to the commodity flow matrix over previous versions included the doubling of the U.S. CFS establishment sample. We incorporated several additional data sets in the enhanced FAF3 log-linear model and made greater use of the U.S. inter-industry input-output (‘use' and ‘make') coefficients and additional federal datasets in the development of the FAF3 out-of-scope (to the 2007 CFS) commodity flow estimates. In addition to the improvements to data methods and data sources, another improvement is the enhanced ability to access FAF in a user friendly web-based tool that allows the user to customize and download tables from the database.
In summary, the data products from the FAF include the following:
You can find all the data products as well as documentation and the data extraction tool at the following FHWA website: http://www.ops.fhwa.dot.gov/freight/freight_analysis/faf/index.htm
In closing, I want to thank the clever and talented team that worked on this project: Ho-Ling Hwang, Frank Southworth, Bruce Peterson, Dave Vogt, Shih-Miao Chin, and James Li, all made valuable contributions. They will be available during the question and answer period to address your questions I also want to acknowledge the contributions of the larger FAF team composed of MacroSys, Battelle and Global Insight. Last, the thought leadership on the part of Federal Highway Administration's Rolf Schmidt and Mike Sprung has been very instrumental in making FAF3 a great product. In particular, Mike's expertise in foreign trade data was a factor in the improvement of FAF3 over FAF2.
With the delivery of FAF3.1, the ORNL team has already begun to process the 2002 and 1997 FAF matrices to create a time series for the user community. Thank you so much for your attention.
Jennifer Symoun
Thank you. I'm going to go back one slide and I am going to type in the URL in the chat box for the FAF3 website so anyone can get to it. I will turn it over to Maks Alam. Maks, I'll bring up your presentation.
Maks Alam
Thank you Diane and FHWA. The way we started this project, the best flow methods we have to take the 131 by 131 OD matrix so that the capacity can be estimated. This slide just gives a basic overview that we had to apply. There are three-main components. One is to update the FAF network which initially was developed for FAF-1 and FAF-2 and then it's also enhanced for the FAF3. Step two was to update the link, national network link as well as other physical and operational constraints. And the task 3 was to use the tonnage, convert them equivalent truck traffic and finally assign that tonnage as equivalent truck traffic on the network system.
I'm going to describe briefly the overall process. This is a very comprehensive process. I'm not able to describe everything in detail, but you're welcome to send an email to Mike or Jennifer. The methodology will also be available on the website. I will go into a little bit of detail on the first task. The network was original developed from the National Network. We have to make sure that every link that is part of the network is also a part of the National Network. The National Network is a federally mandated truck route network. Also, every year, the National Highway System is updated by the Office of Intermodal. Another task was to make sure each freight network is updated to NHS data and intermodal connectors as well as any recent alignment that was done after we have done the FAF2. In addition, we wanted to make sure that when we assignment is done, trucks are not attracted to highway links that had restricted truck access so we had to update each link with restrictions.
So that gives us a kind of look at the geographical based network. Task 2 was essential not only for improving it but also the end of the day another part is to estimate the capacity and the condition of the higher network. So the idea was to bring HPMS database that every year collect both operational and physical information on the highway. Our job was to make it a part of the network. The assignment process we do not only assign the truck by itself, but we also add passenger volume as a pre-load. That's the typical approach for the assignment process. Then the problems happen when you take that and try to bring in part of the HPMS and part of the network. There's always a drop of flow at the boundary. I think this is because of jurisdictional reasons and the time when the database was collected. So we apply to make sure that truck volume which we normally use is consistent in the inbound and the outbound boundary. The OD data that we receive from Oak Ridge is in tonnage, but to estimate the capacity and the delay, it requires physical data. The challenge was to figure out how to convert those tonnage values to different operational values. I'll explain more detail at the end of the presentation about what are the truck types and what we used to convert this truck tonnage to equivalent truck traffic. As mentioned earlier, that since we have done step one and step two, we want to make sure that it is consistent across all three programs with respect to the information that we started with.
The next step is to take the 131 by 131 OD matrix and desegregate it to approximately 4,600 by 4,600 OD matrix so we can achieve a more realistic flow on the network. One of the differences between typical assignment processes that we do apply various processes and really different in the sense that we have to desegregate domestic movements separately of international movement. This is because of the nature of the movement and the controlling factor that goes to the desegregation methods. We used a proportional approach to get 4,600 metrics. This identified a matrix based on various components; one was the 18,000 distribution centers in the data. Then we used the volume that we derived from the HPMS as a pre-load. The reason we use this is because freight is typically done all or nothing or basic travel time driven assignments. However, because of the national assignment there were problems, for example near the Los Angeles I-5 and US-99. All or nothing would try to load most of the traffic on I-5, but when you used the constraints and tried to calibrate it, you would find that the travel distances were different than in the HPMS. And, obviously, when you do that you have to do the calibration. What we did, we did a multiple trial and error iterations to make sure that the overall VMT and the volume is comparable with the HPMS. Now it is very difficult to calibrate because of the nature of the data and the coverage of the geography, so we try to limit our final calibration of VMT distribution for the base year 2007.
So this slide shows how to take the tonnage and convert to equivalent truck traffic. What we did was to identify vehicle groups and select major body types. Then we tried to calculate percent distribution by vehicle group and body type. For each body type, we try to calculate the mean payload by vehicle group. We then try to use weigh-in-motion data and see if the data is comparable. Then we try to get the regional variability because obviously truck traffic in Washington and Montana is not the same as that of East Coast. We try to figure out is it possible to bring in different truck weight regulation as part of this. We decided to use the national averages when converting tonnage to truck traffic. These are the five vehicle groups we used. These are the major body types we used: dry van, platform, bulk, reefer, tank, logging, livestock, automobile, and other. Now at the end of the day after doing all these things, we came up with some like basic formula to compare each specific tonnage and the body type. This is not that complex, it's a three-dimensional algorithm. We have the loaded trucks because we use the tonnage. Then empty trucks are added to the matrix based on empty truck distribution.
Overall summary, this is an in between two and three-step process. We wanted to make sure that at the end of the day if we get it back to 131 by 131 OD matrix, the total volume will be the same. I like to call it a two and a half step process. So, anyway this diagram summarizes what this is about. When we summarize all of these things, you can see it's close to 4609 including domestic. One other thing you can see in Montana is that the state is one zone of the FAF. We have to desegregate those areas so it's comparable with the database. This is the tonnage flow. We'll also give the flow map as part of the report. It's very comparable with the flow pattern that's reported in the HPMS. That's the end of my discussion. If there are any comments, I will take questions. This is the preliminary.
Jennifer Symoun
Thank you. We'll turn it over to Mike now.
Mike Sprung
Thanks Maks and thanks again to Diane and Oak Ridge. I wanted to spend the last few minutes for the speak portion of this to go over what's actually a major improvement in my opinion. Aside from all the technical improvements that were made to make the actual FAF data better, it can be the best thing in the world, but if people can't access it or can't use it, maybe it's not worth doing. So this is how we've enhanced access to the data this time around by developing a web-based summary tool that allows you to extract specific sections of the FAF3 data without downloading an entire database last time which was broken up into pieces based on international modes for air, water and land. This makes it a lot easier to get at. So real quickly jump into this.
The main page gives the option to still download the entire database in .CSV format and in Microsoft Access. On the left you can see options to get to some descriptions of the different releases, summary statistics, the data extraction tool, and the documentation. By selecting the data extraction tool option, it brings you to a page where you can select total flows that would be domestic and international flows combined or you can select domestic alone, import alone or export alone. For the sample here we've just selected the total flows option. You'll notice on the screen shot, just the year 2007 is there because what's presently available on the website is our FAF3.0 data that we released in July and that was only for 2007. I'll go into more detail about what's going to be included in the release next week when I'm done with this, but you'll be able to select the year and that will include 2007, then the provisional year of 2009 and our forecast years every five years from 2015 through 2040. You have the option of selecting origin, either combining for the entire nation by state specific or down to the FAF geography level. That's the 131 regions that we've been talking about. You have the same option on the destination side. You can aggregate all the commodities together, you can select all individually, or you can select individuals or combinations of them. The same thing applies for the modes.
As simple as that, you pick your options, I think here I did Baltimore as an origin and mode of truck and all FAF regions as a destination and sorted this by total tonnage, by truck. Of course Baltimore to Baltimore within is certainly the biggest piece for truck but that picture would change if you selected air or rail as the mode. You can sort in descending order. You can sort this by value. If you want to download this you can as some of the tables can get big. You can view it on the screen and sort it to make it easy or you can download the table into a .CSV file. Basically when you open up a .CSV file, as long as you have Microsoft Excel or similar program, it will pull that data right in. So in summary, that's really how the data extraction tool works. It's certainly a lot easier than working with the 125 megabyte file with close to 500 million cells. If you just want to see what's going out of Baltimore by truck or what's flowing between your region and someplace else.
As I said, FAF 3.0 was released in July. That was the first version of the regional flow database for 2007, and the basic data summary and extraction tool on the website that you see here. Next week what we will be releasing is revisions to that 3.0 dataset. We'll also be adding to that the forecast that goes from 2015 to 2040 in five-year increments. The network assignment maps that Maks just discussed as well as the data for that and the highway link assignment data. Also, the provisional updates for 2009. Later, early next year we'll go back and reprocess 2002 and 1997 data. At the moment the FAF3 data is not comparable with FAF2 data. A lot of changes were made to the Commodity Flow Survey; changes were made to the methodology in FAF3 versus FAF2. Changes were made to the sources we use. Because of all of those things, the numbers are not directly comparable. If you do so, you do so at your own risk. We certainly don't advise it which is why we're going back and we're going to reprocess it using the same methodologies that we used in FAF3 to give better time series. Another potential future product for us will be an online mapping capability, but that will be beyond the reprocessing of the 2002 and 1997 FAF data.
So to wrap things up, I just want to say that we spent countless hours checking the results and data sources input to it, but we do rely on users to help us maintain the quality of the data we released. These files are enormous and the methodologies for each source are varied, so we need eyes that are familiar with the different commodities, the different regions to help us identify any unexpected results. With that, here is the website for access to the FAF data. The first one is the FHWA Freight website. Also, a bit shorter and easier to remember is the US DOT Freight website, www.freight.dot.gov. You can also get to the FAF data at the Oak Ridge National Lab website at the link that Jennifer posted earlier. That's it. I think we have a lot of questions so hopefully we'll have time to get to those.
Jennifer Symoun
We'll move on to the Question and Answer session. We will start with the questions that have been posted online. Actually, I'll bring back the slide with the web addresses. I'll start off with the top of the questions and go on down and we'll see how far we get. If we don't get through all of them, we'll have the presenters answer them via email. I don't know if you want to comment on this, there is no reference in all three to Pacific-Coast Mexican ports, roads, and KSR rail to Texas, bypassing Southern California.
Mike Sprung
The one thing I can say quickly about that is one of the new additions in FAF3 is the use of Canadian and Mexican trans-shipment data. That is trade coming into the United States from other countries but going through Canada and Mexico. That's actually a brand-new source that we were able to get through the Trade Division of the Census Bureau. Freight that is doing that is included in FAF. What you'll see are rail shipments to Southeast Asia going through Texas. That's exactly what that is. Those are shipments that are being trans-shipped through Canada and Mexico, and that is included.
J. Symoun
Also, there was no reference to modal shift of greater than 300 miles from truck to rail.
Maks Alam
There's no modal shift as a part of this assignment process.
M. Sprung
Maybe we can revisit that one at the end, move on to this one.
J. Symoun
Sure. Both UP and BNSF have or soon will have Near Dock Rail Loading in LA. Will trucks to rail transfers be counted as trucking or rail or maybe a new category, such as piggy-backing?
M. Sprung
For where we capture and estimate these rail to truck transfers, those will end up in what we call multiple modes and mail, but if things are being loaded from a vessel directly to rail, I'll also point out, as Diane did that the data had a feature where they did a matching with the railroad for individual containers. We captured a much larger percentage of the domestic leg of rail movements from international.
J. Symoun
The next question is for Diane. How was the 2040 forecast were developed and are forecasts available for the interim years like FAF2?
M. Sprung
The forecasts were similar to what was done in FAF2; were developed by Global Insight using their regional commodity specific forecast models. I'm not going to get into all the details now. I wish we time to cover that in this Talking Freight session. I will say the documentation on the procedures will be available on our website shortly after we release the data next week.
J. Symoun
And when will the commodity flows data be available?
M. Sprung
That will be next week. We have a press release scheduled for next Thursday. That's when all of the FAF3.1 data will be made public.
J. Symoun
What is included in the multiple modes and mail category? What kinds of intermodal modes are included?
M. Sprung
The multiple modes category includes anything that's moved by more than one mode for the domestic leg of the journey as well as mail shipments that were reported.
J. Symoun
Could you please elaborate a little bit more what is captured in the 'no domestic mode' category?
M. Sprung
The no domestic mode is imports, specifically imports of crude petroleum, coming in by a vessel that was off-loaded directly at a refinery. So there's no physical move, no domestic move of these, so they will show up as an import, but there is no domestic move because they didn't travel domestically. That was really the only way we could do it. Initially we had those by water, moving domestically by water, but it didn't make sense. We then put it by pipeline, but that skewed the pipe data. Wherever we put it skewed the tonnage numbers for what really is no domestic move. We ended up with a no domestic mode category.
J. Symoun
Can we get out-of-scope flows at county geography?
M. Spring
Not from the FAF. I guess I'll defer to the experts at Oak Ridge which put together the out-of-scope data. We have no plans to release any of the FAF data at a county level, but that's not to say that some have our source information isn't available at a county level and you could go to directly, but, of course, it would limit the comparability with FAF. I'll see if anyone at Oak Ridge has a comment on that.
Ho-Ling Hwang
It varies, but for example farm based estimate is based on USDA 2007 Agriculture Census data. That data is published at the state level as well as county level even though some county level data needs to be imputed. Other data sets could be just at state level like the Transborder data set that we have. They are at the destination of the state or for the export, the origin of the state. So from state going down to the FAF zone, what we have done is basically used data from the County Business Patterns for the particular application. So it really depends on the source. Like Mike said, on the FAF, there's no plan basically for releasing data at the county level.
M. Spring
Thanks, Elaine. I see that the next question is asking for flows at the zip code level and I think we can infer the zip code is far beyond anything we can do with FAF. I'll just reiterate, the FAF is intended to offer a national picture of regional flows and was never designed to provide that level of detail, the zip code geography.
J. Symoun
Has any comparison been performed between FAF3 and previous FAF2 forecasts?
M. Sprung
Certainly. That was a part of the review process, more so comparing the changes in the growth rates more so than the actual numbers. As I pointed out, the FAF2 numbers and FAF3 numbers are not directly comparable. Without the final report from Global Insight, I can't say in complete detail all of the changes, but it's obvious that there's been a big difference in the optimism of the forecast versus what was calculated for the 2002 data. We had the dip for the last two years that appears to be returning to growth for 2010, but the long range forecasts are a bit more conservative than the FAF2 forecasts were.
J. Symoun
Will the assignment methodology documentation be available online when everything is released next week?
M. Sprung
They will be available on the website but not for next week. The documentation has to go through the final week of compliance in preparation for the website. We'll be looking at early next year for posting on the website. If you ask specific questions, feel free to contact us and we can answer them on a case by case basis before that.
J. Symoun
Maks, in Task 2 Where did the state traffic flow and truck percentages at state boundaries come from?
Maks Alam
The main source HPMS. If there are any questions on the sources, we try to get the traffic flow maps from state DOT. This traffic flow data is available in some published books. We go through the books and try to solve the problem if the data is questionable from the HPMS. Mostly, the data comes from the HPMS database.
J. Symoun
What assumptions or adjustments were made to the model to account for the economic slowdown since 2007 in order to make future forecast into 2040?
M. Sprung
In some ways we were lucky that the data collection was done in 2007. It was largely undisrupted by the downturn in 2008 and 2009. Now, of course, that did need to be taken into account for the forecast, which has been accounted for in the regional commodity forecast models that Global Insight uses.
J. Symoun
Are the 4609 loading nodes and centroids identified in the FAF3 network? They were not identified in FAF2.
M. Alam
We didn't identify them all specifically. We have used demand responsive virtual disaggregation process. So there are no fixed locations.
J. Symoun
Diane, what local data sources would you suggest using to supplement the FAF3 data? I am especially curious about waste, service, and similar movements.
Diane Davidson
I think as Mike has explained, FAF is primarily a regional database although we may use a variety of national, state and regional data sets to construct the FAF. As far as Municipal Solid Waste data sources, there are two state level data sets which are helpful. One is published by Franklin Associates and the other is called the Biocycle Publication, which is a survey of state waste management agencies. I suggest that maybe you could go to some of these state sources and see if they drill down to specific levels for your particular need.
M. Sprung
I'll also add there is a project underway though the National Corporative Freight Research Program specifically to address this and the title of is close to if not exactly the Development of Local Commodity Flow Data. The work on that is underway right now, and short of saying that you may need to do your own data collection or rely on existing sources, it's kind of specific to what you need.
J. Symoun
I just want to mention that questions that are directed toward as specific region or area can be addressed offline. We want to focus on the broad sections. In the network assignment, do you include any empty truck movements?
M. Alam
Yes, the empty truck estimate is based on the 2002 VIUS estimates.
J. Symoun
Is the county to county OD matrix show on Alam's slide 10 produced using the same methods which were used in FAF2?
M. Alam
No, at this time we have not used county to county methods. We used virtual nodes and the 18,000 warehouses in the warehouse distribution warehouse database.
M. Sprung
And the procedures will be described in the documentation.
J. Symoun
Can you briefly touch on how courier such as FedEx and UPS are classified in terms of both mode and commodity?
M. Sprung
As far as the commodity flow survey goes, which I'll address it that way; the commodities are reported as what the commodity is. The mode choice is mail or courier that's given in the commodity survey. We pretty much stick with that. We put it together with the multiple modes in mail assuming that they do move by multiple modes. Some of it last time ended up in Other Unknown, but it seemed more appropriate to include with the multiple modes in mail.
J. Symoun
In the data extraction tool can you select a specific and port of entry?
M. Sprung
No. It's limited to the regions that we have in FAF3 and all imports and exports through individual ports are aggregated to those regions. If you're looking for import, and export data at a port level, this will not be specific. You would end up using some of our source data.
J. Symoun
Can we simply get flows of the trucks, with their origins and destinations?
M. Sprung
Yes. You can do each individual mode.
M. Alam
You can also get truck flow over links.
M. Sprung
Yes. That would be the assignment database.
J. Symoun
Imports, per se, are not covered by the CFS; however, the imported commodities may eventually get partially included in the CFS flows after an initial stop/destination within the U.S. For example, if the imported commodity gets sent to a warehouse/distribution center for repackaging, then it could get covered by CFS from that point. In the 2007 CFS, areas near and around border crossings and entry ports were oversampled in an attempt to better capture some of the import flows. As the FAF estimates imports, could this lead to some double coverage?
M. Sprung
Yeah, absolutely. Thanks. It certainly could. Unfortunately, that's where we spent a good portion of our time; trying to eliminate as much of this as possible. Where an import shipment is moving beyond the port to that distribution center, it will show up as an import going to the distribution center, and then covered by CFs from there domestically. Of course, that's in a perfect world. It may get reported from Laredo, and moving through Arkansas, but it actually goes to a distribution center in Texas and then Arkansas. That would be evidence of double counting, and we're holding a session at the annual TRB meeting to discuss that, among other issues, how to better address it. We've pretty much exhausted all sources possible to get this as right has possible this time, but thank you for bringing that up because that's certainly a possibility. I should mention we did use some control totals from the foreign trade data to limit the maximum amount of freight we show coming in and out. So that would have corrected some of that.
J. Symoun
This looks like a recommendation, and this feature may exist. It might be nice if you provide users the ability to save their selections from the web site to be able to run them again if you are going to update and continually add data.
M. Sprung
Thanks Nathan.
J. Symoun
Can we choose a select link and get a through tree of O-Ds?
M. Alam
No, you can only get it for the specific link, but you cannot do that, you need the virtual OD.
J. Symoun
There I think something will be sent through the Freight Planning LISTSERV. I realize I don't have the address but I can type in the link. I can also send something out to everybody who attended today. Is there a commodity to industry and industry to commodity table?
M. Sprung
Yeah.
Oak Ridge National Laboratory
It's based on the 2002 data and 2008 data, too, including the industries sending the commodities. One way you can get a better picture is to go through the model prepared by the Bureau of Economic Analysis so you can refer to which commodity is going to which industry. Right now there's no data published.
J. Symoun
Are you using or plan to use electronic freight manifest data as opposed to or compliment survey method to keep the model updated?
M. Sprung
I guess what I would say to that is at whatever point some of that information becomes available to us, we would look into incorporating it in our provisional updates. However, the likelihood of that happening before we get to a FAF4 may be unlikely. Obviously, an electronic manifest is available now, but it's getting access to it and the ability to publicly release it at the level that we want to, are issues that will be looked into but I couldn't say yes or no whether that's going to happen.
J. Symoun
Can you point us to more information and background on the allocation to the 4609 assignment points?
M. Alam
We started with all the border crossings. One border crossings at United States-Canada is one point, each domestic port is a point each, if an intermodal facility has a field connector is it a point, and the rest of it is coming from the warehouse distribution center. We did a special model to identify clusters which is within seven miles of the given center.
J. Symoun
Maks, can we get the truck flow and truck percent data at state boundaries for all the states?
M. Alam
Yes. That will be part of the link specific flow as Mike mentioned. It will be similar to FAF2.
J. Symoun
Can I find out the estimated commodity trips on a particular highway versus another highway in my state between two surrounding states or the total commodities per highway between those states and for particular commodities?
M. Alam
Yes, but we need to do a special analysis for this.
M. Sprung
I think actually what they're asking is analysis of specific highways, maybe within a state versus another state, and right now in FAF2 that can be done. That file will not be released next week, but we plan to early next year.
J. Symoun
We have four questions left. Let's see if we can move through them quickly. Is there any method to determine freight movement through state?
M. Sprung
Well, we are going to be doing an analysis of that. We'll call it FAF3.1 plus. We'll be doing a ton-mile analysis and releasing ton-mile tonnage and tonnage estimates to, from, within, and through totals for each of the states.
M. Alam
Yes and in a similar way it was done for FAF2. It is a special analysis and it takes time, so it's not coming into play until next year.
J. Symoun
Are the empty truck movement separately enumerated on the assigned network file?
M. Alam
No. The empty trucks are added to the loaded truck traffic and they are all loaded at the same time during the assignment.
J. Symoun
Was a specific point location file for warehousing and distribution centers created for use in FAF3, and is this file available?
M. Alam
No, we have an agreement with our information source that we cannot release specific locations.
J. Symoun
How are logistics accounted for in the assignment?
M. Alam
There is no specific logistic constraint used in this assignment process other than the capacity constraints.
J. Symoun
There was question; I think somebody wasn't able to hear the answer to the questions. The answers will be documented in the transcript. This will be the last question. Why can't we get vehicle OD flows between regions or in and out of regions rather than just link flow data?
M. Sprung
The OD matrix provides exactly that information. That's the 131 by 131 regions, and that's on the website.
M. Alam
I think the question was the truck, not the tonnage. Is the question about tonnage or trucks?
M. Sprung
He did say truck. That's certainly we'll consider to include in future versions.
J. Symoun
Okay. We're going to close out. We're a minute or two over. Before I do, I want to say we are soliciting topic ideas for the 2011 seminars. If you have ideas for topics or presenters, you can type them in the chat box or send an email. Thank you all for attending today's seminar. The recorded version of this event will be available within the next few weeks on the Talking Freight website.
As a reminder, if you are an AICP member and would like to receive 1.5 Certification Maintenance credits for attending this seminar, please make sure you were signed in today with your first and last name or type your first and last name into the chat box if you are attending with a group of people. Please download the evaluation form and email it to me after you have completed it. Please also download the CM Credit instructions if you are unsure of how to obtain your credits for today's seminar.
The next seminar will be held on November 17 and will be about the Cross Town Improvement Project. This seminar is not yet available for registration but I will send out additional information through the Freight Planning LISTSERV once registration is available.
If you haven't done so already, I encourage you to visit the Talking Freight Web Site and sign up for this seminar. The address is up on the slide on your screen. I also encourage you to join the Freight Planning LISTSERV if you have not already done so.
Enjoy the rest of your day!
