Skip to content U.S. Department of Transportation/Federal Highway AdministrationU.S. Department of Transportation/Federal Highway Administration

Office of Planning, Environment, & Realty (HEP)
Planning · Environment · Real Estate

HEP Events Guidance Publications Glossary Awards Contacts

A Freight Analysis and Planning Model

2. Task 1 - Linking Argos to TransCAD

Task One of this project is to link the workflow ("Argos") to the TransCAD model. This chapter summarizes the Task One work and results. Linking the Argos model required replicating some steps of the workflow in TransCAD. Unless otherwise stated, all raw data used in the replication processes are the same as those detailed in (Giuliano et al. 2008). Those data are called base-case data hereafter.

We describe how TransCAD and other software packages can be used to complete the following modeling steps regarding regional traffic flow analysis:

This chapter is organized as follows. Section 1 shows how freight trips in the Argos workflows can be categorized to facilitate the replication. Sections 2 and 3 detail how different categories of freight trips are estimated in the Argos workflow and how those trips can be read into TransCAD. Section 4 explains how to assemble freight trips of different categories into a master matrix file in TransCAD. Section 5 describes how the matrix file for freight trips can be combined with its counterpart for passenger trips in TransCAD. Section 6 reports how freight and passenger trips are assigned to the network in TransCAD. Section 7 compares the assignment results obtained from TransCAD with actual ground counts for the base case. The main conclusions of this document are that:

2.1 Categorize Freight Trips in Argos: Base Case

Argos workflows and the original freight transportation model on which they are based use different data, procedures, methods and formulae to estimate freight trips with origins and destinations that are inside or outside the region in question (Pan 2003). Our case study area is the Los Angeles region, and the base case data is from 2001. The Los Angeles region is 35,000 square miles and includes the largest container port complex in the US and the third largest air cargo airport. Its 2000 population and employment are 16 million and 7 million respectively. Transportation modeling requires highly disaggregate data; we use traffic analysis zones (TAZs). There are 3,191 TAZs in the region.

The first step in generating the freight flow data is to enumerate all flows (trips) within and through the region. These trips can be grouped into four separate categories based on origins and destinations as shown in Table 2-1.

Table 2-1: Freight Trip Categories

Category 1: Internal to Internal

Trips between 3,191 TAZs within the region

Category 2: Internal to External

Trips from 3,191 TAZs within the region to 12 highway entry-exit points of the region

Category 3: External to Internal

Trips from 12 entry-exit points of the region to 3,191 TAZs within the region

Category 4: External to External

Trips between 12 highway entry-exit points of the region

To obtain these trips, one first needs freight trip productions and attractions for the different TAZs and entry-exit points. The 3,191 TAZs include 10 TAZs containing major seaports, airports and rail yards; in addition there are 12 highway entry-exit points. The Argos workflow automatically estimates freight trip productions or attractions measured in dollars and tons for all TAZs and entry-exit points, using data from multiple sources and computations developed by the Argos research team (Giuliano et al, 2008; Ambite and Kapoor 2007; Ambite and Kapoor 2007). After the test empirical application described in Chapter 1, the dollars to tons to PCEs conversion was added to the Argos planner.

It is important to note that PCE trips generated by Argos are the final product of multiple modeling steps. These steps start with the estimation of freight flows in dollars by commodity based on regional input-output data and point employment data. Flows are converted from dollars into tons by commodity based on regional freight flow data drawn from the Commodity Flow Survey (CFS). Tonnage flows are allocated to modes (air, water, rail, truck), also using CFS data. PCE values for the truck mode are obtained using the estimated ton values and load factors by commodity (Pan 2003; Giuliano et al. 2008).

2.2 Estimate Intraregional Trips in TransCAD

2.2.1 Generate Input Data for Trip Distribution

The Argos workflow originally produced the attractions and productions in two separate .csv files. These files store the numbers of trips in a long-table format. Figure 2-1 provides an example of table format in Microsoft Access. Column #1 is the commodity ID assigned by the USC industry sector coding system (see (Pan 2003). Column #2 is the TAZ ID. Column #3 shows the factor shares produced ("Export") or attracted ("Import") by TAZ by industry sector. The unit of measurement in this example is dollars.

Figure 2-1: Argos-generated .csv files by Commodity by TAZ

This figure is a screenshot of the Argos-generated .csv file. This file has three columns, Column 1 contains USC commodity codes; Column 2 is TAZ ID; Column 3 is the import or export value in thousand dollars.

There are 29 commodity types in the USC industry sector coding system.[2] Given 29 commodity types and 3,191 TAZs, there should be 92,539 rows (records) altogether in either .csv file. However, to save storage space, the original Argos workflow omits rows in sequence if numbers of productions or attractions are zero. The final .csv files generated by the original Argos workflows have 73,926 and 90,016 rows for attractions and productions respectively.

TransCAD uses .bin files for productions and attractions, and they are structured differently than the Argos .cvs files. An example is illustrated in Figure 2-2. There are three differences between them. First, .bin files do not distinguish commodity types. Second, .bin files do not omit TAZs even when there are zero productions or attractions. The third difference is that .bin files store attractions and productions in one file while .csv files store the attractions and productions in two separate files. By default, TransCAD uses .bin files as input for trip distribution. Therefore the original Argos .cvs files had to be converted to compatible .bin files.

Figure 2-2: TransCAD .bin Files

This figuresis a screenshot of the TransCAD .bin file. It has three columns that respectively contain TAZ ID, import value, and export value.

Initially the conversion into a format suitable for TransCAD was done outside of Argos in three steps described below. Now, an updated Argos workflow directly creates a file in the required format. This is implemented as a new step in the updated Argos workflow, which performs a join of the import and export tables along with a list of all TAZs so that there are no omitted rows.

Step 1: Generate a complete index list of trips by TAZ by commodity

As mentioned above, if visualized as a long table, a complete index list of trips by TAZ by commodity should have three columns and 92,539 rows, as shown in Table 2-2. Column #1 is used to store commodity type IDs 1, 2, 3, …,27, 28,29, and Column #2 to store TAZ IDs 1, 2, 3,…,3189, 3190, 3191. Column #3 is the concatenation of the values in Columns #1 and #2, which generates 92, 539 unique IDs for PCE trips produced or attracted by commodity type for each TAZ. We use a Visual Basic program in MS Excel 2007 and the embedded "concatenate" function to generate the complete index list as above.

Table 2-2: Complete Index List of Trips by TAZ by Commodity

Commodity ID TAZ ID Concatenated ID by Commodity by TAZ
1 1 1,1
1 2 1,2
1 3 1,3
…,…
1 3189 1,3189
1 3190 1,3190
1 3191 1,3191
2 1 2,1
2 2 2,2
2 3 2,3
…,…
2 3189 2,3189
2 3190 2,3190
2 3191 2,3191
…,…
29 1 29,1
29 2 29,2
29 3 29,3
…,…
29 3189 29,3189
29 3190 29,3190
29 3191 29,3191

Step 2: Generate unique IDs for Argos .csv files

The unique IDs are generated by concatenating Columns #1 and #2 in the .csv files. Again, the "concatenate" function and a VBA program in MS Excel 2007 were used. This step generates .csv files with unique IDs for each record.

Step 3: Link .csv files with unique IDs and the complete index list

Having the complete index list and .csv files with unique IDs, we use MS Access 2003 to link the list and the files together, using the unique IDs in the list and in the files as a common key. MS Access 2003 allows users to keep all records that the list has. All the omitted records that have zero trips in Argos .csv files are now padded, as shown in Table 2-3.

Table 2-3: Padded Argos .csv Files

Commodity ID

TAZ ID

Concatenated ID by Commodity by TAZ

IMPORTS

EXPORTS

1

1

1,1

0

0

1

2

1,2

10.45

10.00

1

3

1,3

10.00

20.00

…,…

1

3189

1,3189

0

0

1

3190

1,3190

10.45

10.00

1

3191

1,3191

10.00

20.00

2

1

2,1

0

0

2

2

2,2

10.45

10.00

2

3

2,3

10.00

20.00

…,…

2

3189

2,3189

0

0

2

3190

2,3190

10.45

10.00

2

3191

2,3191

10.00

20.00

…,…

0

0

29

1

29,1

10.45

10.00

29

2

29,2

10.00

20.00

29

3

29,3

…,…

0

0

29

3189

29,3189

10.45

10.00

29

3190

29,3190

10.00

20.00

29

3191

29,3191

0

0

Depending on whether different constant parameters are used to distribute trips by commodity type in TransCAD, the padded Argos .csv files can be exported to 29 separate .csv files with each file having 3,191 records, or to one .csv file with 92, 539 records. In our case, the same constant parameters are used to distribute trips of all commodity types, so only one .csv file is exported with 3,191 records by aggregating across all commodity types. The resulting file is now compatible with the TransCAD .bin file structure, and hence can be read into TransCAD.

2.2.2 Balance Trips Produced and Attracted by TAZ

Depending on whether total attractions equal total productions for all TAZs, an intermediate procedure called "trip balancing" is required before TransCAD is used to distribute attractions and productions. In travel demand models, trip balancing ensures that total attractions equals to total productions in the study area. Technical details of trip balancing can be found in (Caliper 2005). In our case, trip balancing is not performed since Argos workflows generate two .csv files that are constrained to have the same total intra-regional attractions and productions for all TAZs.

2.2.3 Distribute Argos-generated Intraregional Trips in TransCAD

The balanced trips produced and attracted by TAZ can be distributed with the embedded gravity model application in TransCAD. The gravity model is the most widely used trip distribution model. TransCAD also needs an input file of friction factors between TAZs. The factors represent the minimum generalized costs that trip makers bear when traveling between TAZs. The costs can be minimum monetary costs, travel distances, or time costs. We use shortest-path travel distances between TAZs. The distances are calculated in TransCAD using the 2000 Southern California Association of Government (SCAG) Regional Travel Model road network files in ArcView format, a TransCAD .net file, and TransCAD's embedded "multiple paths function". Technical details of generating .net file and finding shortest-path travel distances between TAZs (centroids) can be found in (Caliper 2006). With the balanced trip file and the friction factor file, one can now distribute the trips in TransCAD. Figure 2-3 is the screenshot of the user-interface for trip distribution in TransCAD.

The outcome of intraregional trip distribution is an N x N Origin-Destination (O-D) matrix (table) with N = number of TAZs. Each element in the matrix holds the number of trips between an origin and a destination within the study region. The row index value of the element is the ID of the origin (TAZ), and the column index value is the ID of the destination (TAZ). We do not include trips within the same TAZ in the traffic assignment, because there are so few such trips that travel on major roads.

Figure 2-3: Trip Distribution Interface in TransCAD

This figure s is a screenshot of user interface one would see when performing trip distribution in TransCAD. It shows a pop-up box where user instructions are entered, a portion of the trip production and attraction file, and a portion the shortest path travel time file.

KEEP 2.3 Distribute Interregional Trips

We now turn to Category 2, 3 and 4 trips, those with at least one end exiting or entering the region. Interregional trips travel by air, rail, water, or truck. Since we are interested in flows on the highway network, we must identify the portion of interregional trips that use the truck mode. The Argos workflow generates matrices of interregional flows by mode and commodity. The truck portion flows comprise the interregional trips used to generate the origin-destination matrix discussed here.

2.3.1 Balance trips attracted and produced at regional entry-exit points

The total number of freight trips attracted and produced at regional entry-exit points for modes of surface transportation (including 12 highway entry-exit points, 2 major seaports, 5 airports and 3 rail yards) generated by Argos do not equal each other, because attractions are regional imports and productions are regional exports. However, each import or export has a corresponding trip end within the region, so the total attractions and productions for imports are equal, as are the total attractions and productions for exports. In the case of through trips, the total number of attractions at entry/exit points equals the total number of productions at these points. Therefore no trip balancing is required for using Argos outputs.

2.3.2 Distribute trips between TAZs and regional entry-exit points

Freight trips between entry/exit points and TAZs within the region are assigned in proportion to commodity specific intra-regional attractions or productions using the following (Pan 2003):

An equation where All inbound freight trips from entry-exit point o to zone d are equal to the inbound trips at entry-exit point o for a given commodity type, times the ratio of internal trip attractions for that commodity type for zone d to the sum of the internal trip attractions for that commodity type for all zones, summed over all commodity types (1)

An equation where All outbound freight trips from zone o to entry-exit point d are equal to the outbound trips at entry-exit point d for a given commodity type, times the ratio of internal trip productions for that commodity type for zone o to the sum of the internal trip productions for that commodity type for all zones, summed over all commodity types (2),where

Symbol #1 (F^oe,d) representing inbound freight trips from entry-exit point o to internal TAZ d= inbound freight trips from entry-exit point o to internal TAZ d;

Symbol #2 (F^o,ed) representing outbound freight trips from internal TAZ o to entry-exit point d= outbound freight trips from internal TAZ o to entry-exit point d;

Symbol #3 (Inb(sub i)^Eo) representing inbound commodity i at entry-exit point o= inbound commodity i at entry-exit point o;

Symbol #4 (Outb(sub i)^Ed) representing outbound commodity i at entry-exit point d= outbound commodity i at entry-exit point d;

Symbol #5 (A(sub i)^d) representing internal trip attraction of commodity type i at internal TAZ d= internal trip attraction of commodity type i at internal TAZ d;

Symbol #6 (P(sub i)^o) representing internal trip production of commodity type i at internal TAZ o= internal trip production of commodity type i at internal TAZ o;

Symbol #7 (Σ(sub d) A(sub i)^d) representing sum of attracted trips of commodity type i= sum of attracted trips of commodity type i;

Symbol #8 (Σ(sub o) P(sub i)^o) representing sum of trip productions of commodity type i= sum of trip productions of commodity type i.

The original Argos workflow generated values forSymbol #3 from legend above,Symbol #4 from legend above, Symbol #5 from legend above, and Symbol #6 from legend above, and the values of Symbol #7 from legend aboveand Symbol #8 from legend above could be calculated based on the values ofSymbol #5 from legend above, andSymbol #6 from legend above. Thus, all variables in equations (1) and (2) were known and hence Symbol #1 from legend aboveandSymbol #2 from legend above could be calculated. However, manual calculations are extremely tedious and time-consuming. For instance, there are 12 highway entry-exit points and 3,191 internal TAZs. Similar calculations must be repeated 3,191*12 * 2=76,584 times to get all values for Symbol #1 from legend aboveandSymbol #2 from legend above. We have modified the Argos workflow so that this assignment of external trips takes place within the workflow.

With the values of Symbol #1 from legend aboveandSymbol #2 from legend abovecalculated, only through trips for the study region remain to be considered. Through trips include the imports and exports that simply flow through the region. Examples include imports arriving at the ports and flowing to destinations outside the region, or exports from outside the region arriving at LAX for transport to international destinations. A large portion of through trips does not travel by truck, and hence are not relevant for highway network forecasting and analysis. As described in Chapter 1, we factor out the air, water and rail trips, leaving truck through trips to be distributed.

Through trips are allocated to the 12 highway entry/exit points based on interregional truck flow data for the State of California. That is, the through trips are assigned entry/exit nodes based on observed truck flows at those locations. Note that "through trips" include a truck trip from Northern California to Mexico, a trip from the ports to Arizona, etc. This step is calculated by a SQL query outside Argos planner.

2.4 Aggregate Freight Trips of All Categories

The process described above result in two separate data files that are either .bin files or .csv files. Before these files can be used as input for traffic assignment, they have to be aggregated into a single matrix file in TransCAD. In TransCAD, the O-D matrix file has to be associated with a standard geographic file to be used as input for traffic assignment. TransCAD uses standard geographic files to store spatial and attribute information of the transportation system.

In our case, the standard geographic file used to assign trips is based on the 2000 SCAG Regional Travel Model's road network files, which were created in ArcView. The SCAG network files do not have our 12 highway entry-exit points, so we add them to the network, yielding a total of 3,203 origins/destinations. The ArcView file is converted into a standard geographic file in TransCAD. TransCAD then can be used to generate a blank matrix file that has 3,203 origins or destinations. Step-by-step guides regarding how to generate a blank matrix file in TransCAD can be found in (Caliper, 2006). Figure 2-4 is a screenshot of the user-interface for generating a blank matrix file in TransCAD.

Figure 2-4: Generate Matrix File in TransCAD

This figure is a screenshot of user interface one would see when using TransCAD to generate an OD matrix file. It shows a pop-up box where user instructions are enterend, and a blank matrix file.

Data from the .bin files or .csv files mentioned above can be imported into the blank matrix file, using TransCAD's embedded Matrix-Import Tool. Figure 2-5 shows the user-interface in TransCAD that is seen when reading .bin or .csv files into a matrix file.

Figure 2-5: Import .csv and .bin Files into TransCAD

This figure is a screenshot of what user would see when importing .csv or .bin files into TransCAD. It shows a pop-up box for user instructions, portions of the .bin file, and portions of the .csv file.

The first 3 columns of Figure 2-5 is a .bin file, and the remaining columns are in a matrix file, to which data have already been imported. Both files are not completely shown because both have thousands of rows or columns. Files are imported to the matrix based on TAZ ID. In TransCAD, values in corresponding cells in matrix files that are associated with the same geographic file can be added, multiplied, divided, or subtracted using the Matrix Operation tool. This tool is used to aggregate freight trips of different categories after they have been imported from .bin or .csv files into appropriate matrix files. The aggregation generates a new matrix file, whose cells hold the sum of freight trips of different categories. At this point, freight trips are aggregated and are ready to be combined with corresponding passenger trips.

2.5 Combine Freight Trips and Passenger Trips

Freight trips are only a small portion of all trips using a region's road network. To perform traffic assignment of all trips in a study area, one needs origin-destination data for both freight and passenger trips for the same time period.

2.5.1 Prepare the Matrix File for Passenger Trips

Argos does not address passenger trips. Thus the origin-destination matrix for passenger trips is generated using different data and procedures. In our case study, we use the base case equilibrium traffic assignment (AM peak, 6 to 9 AM) produced by SCAG (Southern California Association of Governments, 2003). These data are originally in .pcu format, which minimizes file sizes by using low-level binary format to store information. Before TransCAD can use them as input for traffic assignment, the .pcu data have to be converted into a compatible matrix file. The conversion involves the following steps:

Step 1. Convert the data in .pcu format into data in .txt format

A short program called "bintotxt.exe" is used to convert the .pcu file into a .txt file. The new .txt file is an N x N table, where N = total number of origins/destinations.

Step 2. Reshape a wide table of .txt format into a long table of .csv format

Although the .txt file has the same dimensions of the TransCAD matrix file, it cannot be directly read into TransCAD. Rather, the .txt table must be converted to a long table (e.g. in .csv format, which has N rows and three columns. We used SPSS 15's Restructure Data Wizard to produce the long table as a SPSS .sav file. A .sav file can be directly exported to a .csv file. A portion of the .sav file is shown in Figure 2-6.

Figure 2-6: Example of a .sav File

This figure is a screenshot of a .txt wide table after being imported into SPSS and saved as a long .sav table.

Step 3. Import data from a .csv file into a matrix file

We import the .csv passenger trip data (file) into a blank matrix file, producing the passenger O-D matrix in TransCAD.

2.5.2 Combine Freight Trips and Passenger Trips

We now have two O-D matrix files, one for freight and one for passengers. The"Matrix-Pack" tool in TransCAD is used to merge the two files. Figure 2-7 is the screenshot showing both matrix files and the menu for Matrix-Pack.

Figure 2-7: Matrix-Pack Tool in TransCAD

This figure is a screenshot of how one can find the matrix-pack tool from the drop-down menu of TransCAD.

2.5.3 Allocate Freight Trips to a Given Time Period

Argos generates freight trips for a whole year. To get the freight trips for a given time period from the annual data i, the following formula is used

fij=Fij/365 * α (3),

where fij is the freight trips measured in PCEs between origin i and destination j in the time period of interest (in the base case, AM-peak period was studied), Fij is the annual freight trips measured in PCEs between origin i and destination j in the matrix file containing annual freight trip data, α. is the estimated ratio of freight trips occurring in the time period of interest. In the base case, a value of 0.1537028 was used for α. This value was estimated and given in (Pan 2003). In other cases, the values for α can be adjusted, depending on the temporal distribution of freight trip and the time period to be simulated. In all cases where the value of α is known, TransCAD allows the modeler to automatically implement (3) by using the "Fill-Matrix" tool. Figure 2-8 illustrates the Fill-Matrix user interface.

Figure 2-8: Fill-Matrix Tool in TransCAD

This figure is a screenshot of the pop-up box for the fill-matrix tool in TransCAD.

2.5.4 Aggregate Freight and Passenger Trips

TransCAD has a "Quicksum" function to automatically aggregate trips for the same time period in corresponding cells in matrix files that are linked together. It also generates a new matrix file that holds the aggregate trips. Figure 2-9 shows the screenshot of the Quicksum function. Once the aggregated file is created, traffic assignment can be performed.

Figure 2-9: Quicksum Function in TransCAD

This figure is a screenshot of how one can find the Quicksum function from the drop-down menu of TransCAD.

2.6 Assign Freight Trips and Passenger Trips

2.6.1 Create a .net File

TransCAD needs three files to perform a traffic assignment: 1) a standard geographic file that represents the actual road network, 2) a matrix file that contains trips between all origins and destinations within the network, and 3) a .net file that is created based on the standard geographic network. A .net file contains algorithms that TransCAD uses to calculate the minimum-impedance paths between any two points in a standard geographic file. In TransCAD, a .net file is not visible and its source codes are locked to conventional users for proprietary reasons. An appropriate .net file has to be created and added before TransCAD can perform a traffic assignment. Where a standard geographic file is available, TransCAD allows the modeler to create the .net file by telling TransCAD which nodes in the file are treated as origins and destinations in traffic assignment and by executing a function called "Create Network". Figure 2-10 is the screenshot of the menu/function with which the modeler can create a .net file when a standard geographic file is opened in TransCAD. The .net file for our base case was created using the 2000 SCAG Regional Travel Model's geographic file for the road network.

Figure 2-10: Create a Network File in TransCAD

This figure is a screenshot of the pop-up box for “Create Network” in TransCAD.

2.6.2 Assign Trips

Various types of assignments are available in Trans CAD, including all-or-nothing, STOCH, capacity constraints, user equilibrium, system equilibrium, etc. In addition, the user controls the convergence criteria. Figure 2-11 is a typical user-interface for traffic assignment. Beneath the interface are a standard geographic file and corresponding matrix file.

Figure 2-11: User-Interface for Traffic Assignment in TransCAD

This figure is a screenshot of the user interface of “Traffic Assignment” in TransCAD. It shows the instruction pop-up box , the road network file, and a portion of the matrix file.

We selected user equilibrium for our base case traffic assignment demonstration. A user equilibrium assignment is deemed reached when the following conditions are satisfied:

(1-β) Fli-1less than or equal to Fli less than or equal to(1-β)Fli+1, (4)

where Fli is the traffic volume at any given segment in the road network at the ith iteration, and β is a pre-assigned parameter that may range from 0 to 1. Values of β typically equal 0.05 or 0.01. In our case, β=0.05.

Once an equilibrium assignment is reached, results are automatically stored in a .bin file that contains the following information:

One advantage of TransCAD over the traffic assignment software used in our previous research is the graphical user interface that provides illustrations of the assignment results. Figure 2-12 provides an example. It shows the actual traffic assignment results for the base case. The red color shows presence of traffic on the link, and the width of each segment shows relative traffic volumes.

Figure 2-12: Traffic Volumes by Road Segment at Equilibrium for the Base Case

This figure is a graphical representation of the equilibrium assignment on the road network for the base case. Volumes are shown by width of each road segment.

2.7 Evaluation of Results

In order to evaluate the results of our linked model, we compare the base case traffic assignment with ground counts, and with the previous results from the earlier version of our model. We use exactly the same data inputs as in our previous application (Giuliano et al, 2008).[3] However, we do not expect exactly the same traffic assignment results for the following reasons: 1) in reviewing the Argos workflow, we discovered one computational error, and therefore results would be somewhat different even if we had used our original method; 2) the highway network used in TransCAD is similar but not the same as the highway network used previously; 3) the assignment algorithms are different from those used previously.

Figure 2-13 illustrates differences in the coded network for a small portion of the region. The SCAG network is based more closely on the street network. The SCAG network is quite large and complex; it has 92,559 links (including centroid connectors) and 35,360 nodes, compared with 89,356 links and 30,385 nodes in the network used in the earlier work. We have no information on possible differences in the assignment algorithm, because TransCAD assignment methods are proprietary and the algorithms are hidden.

Figure 2-13: Comparison of Coded Networks

This figure is a graphical representation that compares two versions of SCAG’s road network files. The older version is show in red and the newer one is shown in grey.

We follow the same process as described in Giuliano et al, 2008. In the earlier application, we assigned truck PCEs to the network already loaded with the equilibrium passenger assignment. We conducted a 3 hour AM peak assignment, and then factored up to a 24 hour assignment in order to compare results with 24 hour ground counts. The two stage assignment allowed us to identify truck PCEs as the screenline count differences before and after adding the truck PCEs. TransCAD has a multiclass assignment function that allows distinguishing PCEs associated with different vehicle classes. However, it requires separate generalized cost functions for each class. We do not generate different generalized costs functions. We therefore used the same process as before, first assigning all passenger PCEs, then adding the truck PCEs to the loaded network.

We used the same screenlines and ground counts as in the previous application. The 18 screenlines are shown in Figure 2-14. The ground count data were provided by SCAG and were obtained from field surveys and highway sensor data for 2001[4]. We conducted two runs, one with PCE as a weighted average based on region level vehicle classification data (Comparison 1); the other with PCEs unique to each screenline and generated by SCAG (Comparison 2).

Figure 2-14: SCAG Screenlines in 2003

This figure shows a map of the SCAG region with the freeway system and major arterials. Screenlines are indicated by numbers.

Source: Adapted from Meyer & Mohaddes Associates, VRPA Technologies, and WILTEC 2004.

Results are given in Table 2-4 and Figures 2-15 and 2-16. Table 2-4 shows that using TransCAD somewhat reduces the fit of our results with the screenline counts, whether using actual PCEs or the SCAG generated screenline specific PCEs. However, given that we do not calibrate the model in any way, these results are still reasonable. As noted above, we would not expect the same results as in our previous work, but do not necessarily expect less robust results. The most likely reason for less robust results is the assignment process itself. In our earlier work, the assignment model allowed for adjustments in response to congestion on the system; we do not know the extent which such adjustments take place in TransCAD.

Having demonstrated that Argos results can be applied to TransCAD, the next step is to update Argos and apply it using more recent data.

Table 2-4: Comparison of Results, Prior Application vs TransCAD

  Comparison 1 (actual PCE) Comparison 2 (SCAG PCE)
  Prior results TransCAD Prior results TransCAD
Ave % difference 36.5 70.5 -5.3 3.61
Min % difference 0.8 1.7 7.5 -7.4
Max % difference 206.8 288.8 134.0 112.1
Ave weighted % diff 20.0 31.6 17.0 -22.4
Weighted % mean sq error 17.8 21.3 10.0 38.2
Regression R2 0.80 0.73 0.67 0.58

Figure 2-15: TransCAD results, Comparison 1, actual PCEs

This figure shows the plot of estimated HDT (y axis) and actual HDT (x axis) in PCEs. It shows the regression line and the estimated coefficients: Estimated = 0.84(actual) + 60188.

Figure 2-16: TransCAD results, SCAG PCEs

This figure shows the plot of estimated HDT (y axis) and actual HDT (x axis) in PCE. It shows the regression line and estimated coefficients: estimated = 0.56(actual) + 63635.

Updated: 10/06/2011
HEP Home Planning Environment Real Estate
Federal Highway Administration | 1200 New Jersey Avenue, SE | Washington, DC 20590 | 202-366-4000