1. Ohio Traffic Forecasting Manual
Module 3: Travel Demand Modeling

2. Training Organization
Forecasting Training Modules
Ohio Traffic
Module 1: Traffic Forecasting Background
Policy, Administrative and Technical Overview
Module 2: Traffic Forecasting Methodologies
Data and Parameters, Step-by-Step Procedures, Examples
Module 3: Travel Demand Forecast Modeling
Model Selection, Checks and Refinements
Intended Audience
Forecast User
Traffic Forecaster/Modeler
Ohio Traffic Forecasting Manual

3. Outline – Part II Standard Model Checking Advanced Model Checking
Special Requirements
Ohio Traffic Forecasting Manual

4. Documentation and Coordination
Throughout this training, the following symbols indicate: Items that are required or suggested for documentation Items that are required or suggested for coordination with ODOT M&F
Ohio Traffic Forecasting Manual

5. Standard Model Checking

6. Using the Model for Projects
The Model of Record will generally need revision to provide the location-specific accuracy needed for project level forecasts.
This is accomplished with 3 processes:
Model Checking
Model Refinement
Model Adjustment
NOTE: The checks/refinements/adjustments covered in this training focus on the needs of highway projects, transit and policy type projects will have their own requirements outside the scope of this training.
Ohio Traffic Forecasting Manual

7. Model Checking
All model work should begin with the validated Model of Record.
Model checking is generally conducted at two levels:
For the entire regional model
Within the project study area
Project
Location
Study Area
Ohio Traffic Forecasting Manual

8. Regional Checking
Regional (system-wide) checks should be performed when:
The base year is different from the validation year.
Model adjustments are made specifically for a project.
The checks conducted at the regional level are simply a subset of the more extensive series of checks that will be conducted within the study area. The regional checks to be discussed include:
%RMSE Check
VMT Check
Screen line Check
V/G Ratio Check
Ohio Traffic Forecasting Manual

9. Study Area Checks
Study area checks should be conducted for all projects. The first four checks are the same as conducted regionally:
%RMSE (or Percent Difference)
VMT
Screen line
V/G Ratio
Additional study area checks include:
Network Coding
Base Land Use
Forecast Land Use
Growth Rates
Ohio Traffic Forecasting Manual

10. Isolating Study Area Statistics
With the standard MPO model, these can be checked on the study area by coding a unique project identifier in field PROJID on the network.
Ohio Traffic Forecasting Manual

11. Isolating Study Area Statistics
All assignment validation statistics are reported for both the region and all unique PROJID.
Ohio Traffic Forecasting Manual

12. Isolating Study Area Statistics
If you don’t have this capability inherent, you can add a link attribute (link-attribute-add) to accomplish the same thing (you’d need to add conditional logic to your reporting scripts as well).
Ohio Traffic Forecasting Manual

13. %RMSE Check
The root mean square error (RMSE) is a measure of the relative error of the assignment compared to actual traffic counts. When compared to the ground count, it is used to measure the relative accuracy of the assignment.
Ohio Traffic Forecasting Manual

14. Allowable %RMSE by Directional Link Volume Group
%RMSE Check
%RMSE is applied to volume groups rather than links.
Ignored for directional volumes < 500/day.
Allowable %RMSE by Directional Link Volume Group
%RMSE reported graphically or in tabular format showing achieved versus desired level by volume group.
Ohio Traffic Forecasting Manual

15. %RMSE Check %RMSE is contained in report 7 of the standard model:
RPT7MPOYRVALAA.DBF
This can be imported to spreadsheet and plotted versus the desired %RMSE curve.
Conversely, percent difference can be calculated directly on the links (possibly by saving the study links to DBF and importing to spreadsheet).
Ohio Traffic Forecasting Manual

16. %RMSE Check midpoint desired 250 200 1000 100 2000 62 3000 54 4000 48
5000 45
6250 42
7750 39
9250 36
11250 34
13750 31
16250 30
18750 28
22500 26
30000 24
45000 21
65000 18
97500 12
Ohio Traffic Forecasting Manual

17. VMT Check
The Vehicle Miles Traveled (VMT) check calculates the volume times the distance of the link to produce vehicle miles. VMT checking is somewhat less important at the study area level, particularly when there are few links in a category (in which case individual percent differences are more important). Still, it provides a benchmark on model performance, particularly for larger study areas. VMT Check Acceptance Criteria
Ohio Traffic Forecasting Manual

18. VMT Check VMT summary is contained in report 8 of the standard model:
RPT8MPOYRVALVMTAA.DBF
Ohio Traffic Forecasting Manual

19. Screen Line Check
A true screen line crosses the entire model area. A cut line (or auxiliary screen line) is a line which intersects several parallel roadways which form a corridor. A cordon line is a special case of the screen line which connects back upon itself, isolating all traffic movements into and out of an area.
Ohio Traffic Forecasting Manual

20. Screen Line Check
Screen lines, cut lines and cordon lines are collectively referred to as screen lines.
A screen line check consists of adding the total volume on a screen line and comparing to that of the corresponding ground counts.
The add a lane, drop a line criterion is used on the % difference.
Higher volume corridors are allowed less error.
Lower volume corridors are allowed more error.
Total number of lanes in the corridor should be correct.
Ohio Traffic Forecasting Manual

21. Screen Line Check
Screen line checking is less important at the study area level.
However, for some large study areas, some formalized screen lines might be defined for the study area to ensure certain important corridor flows are considered.
The screen line summary is contained in report 9 of the standard model:
RPT9MPOYRSCRNLNAA.DBF
Ohio Traffic Forecasting Manual

22. V/G Ratio Check
The V/G ratio check is an alternative way to express percent differences in map format: Volume / ground count Shown as a link map colored as follows:
Include link map, annotated with the directional assigned volume and count
Ohio Traffic Forecasting Manual

23. Other Checks
If model adjustments or recalibration are performed, additional checks should be conducted identical to the relevant checks report in the model’s validation report.
Ohio Traffic Forecasting Manual

24. Network Coding Check
Network checking is a comparison of the coded attributes with known conditions.
Networks should be thoroughly checked in the study area.
Independent sources such as ODOT Roadway Information Database, aerial photos and windshield surveys should be conducted to ensure the network data is correct in the study area.
Additional roadways might also be brought into the network during the refinement stage to provide greater network detail (all roadways imposing traffic control on modeled roads of interest should be included).
Ohio Traffic Forecasting Manual

25. Network Coding Check
Besides ensuring the proper roadways are coded and contain the correct directionality, the following attributes should be checked:
Ohio Traffic Forecasting Manual

26. Network Coding Check
These attributes are all contained at the front of the link attributes list for the standard models.
If time of day specific overrides exist, they are further down the attribute list and must be checked as well.
Note that TURNLANE and IXTYPE are intersection attributes coded on links, it is important to be careful to keep these associated with the correct intersection node.
Ohio Traffic Forecasting Manual

27. Base Year Land Use Check
Base year land use should be checked in the study area in three ways (the level of checking depending on the magnitude of the project):
Check consistency in a spreadsheet
Check accuracy versus independent data in GIS
Check trip generation versus ITE Trip Generation
Ohio Traffic Forecasting Manual

28. Land Use Consistency Check
Import the base year SE file (for the standard model this will be called MPOYRZDAA.DBF)
Calculate ratios between various variables such as:
HHPOP/HH
WRKER/HH
VEH/HH
POP18/POP
HOTELROOM/NAICS72
ENROLL/NAICS61
Graphs can also be useful here.
Ohio Traffic Forecasting Manual

29. Land Use Accuracy Check
Compare the variables to other sources such as:
Census
QCEW/InfoUSA
Previous Traffic Impact Studies
Windshield surveys/aerial photos
GIS is a useful tool to bring these data together for this analysis.
Ohio Traffic Forecasting Manual

30. Land Use Accuracy Check
Centroid connectors should be checked versus actual access points.
Actual
Old
New
Ohio Traffic Forecasting Manual

31. Forecast Year Land Use Check
Consistency checks similar to those for the base year should be conducted in a spreadsheet.
In addition, the forecast socio-economic data should be compared to the base year to ascertain whether the implied growth makes sense.
Sometimes, forecast year variables are updated with methodologies that are different from those used to develop base year variable, resulting in major inconsistencies.
Ohio Traffic Forecasting Manual

32. Growth Rate Check
At the most important project locations, the annual traffic growth rate implied by the model should be compared to independent sources such as:
Past traffic count trends
HPMS VMT growth rates
ODOT Congestion Management Process growth rates (TIMS)
ODOT annual adjustment factors report
Population/employment growth trends (ODOD projects etc.)
Ohio Traffic Forecasting Manual

33. Percent Difference Check
The Percent Difference check is the same as the %RMSE check conducted for the entire model. The percent deviation of the daily directional volume of individual links in the study area is compared to the Maximum Desirable Deviation Curve. (by direction!)
Ohio Traffic Forecasting Manual

34. Percent Difference Check
Ohio Traffic Forecasting Manual

35. Advanced Model Checking

36. Advanced Model Checking
There are additional study area checks that should be conducted for larger, more complex projects.
These include:
Travel Time
Paths
Select Links
Distribution
Modal
Generation
If “Refined Alternative Level Traffic” will be produced for operational level analysis, these checks are required!
Ohio Traffic Forecasting Manual

37. Travel Time Checks
When independent travel time estimates are available (such as those now made available by ODOT and FHWA), the network travel times/speeds obtained from the paths or on individual links can be checked.
SPDMOD field can then be used to adjust these if necessary.
Ohio Traffic Forecasting Manual

38. Path Checks
Paths taken through the network can be checked to insure they are logical. They are best checked by using the path file output from a Cube assignment.
Two possible examples:
Making sure a freeway-freeway movement stays on the freeway through the study area.
Making sure most through traffic follows a signed state route.
Ohio Traffic Forecasting Manual

39. Select Link Analysis
Select link analysis shows where the traffic using a given link comes from/goes to and all other links used. It is useful for diagnosing problems with specific links (similar to path checking and select zone maps shown previously) also critical for weave analysis.
Ohio Traffic Forecasting Manual

40. Select Link Analysis
When the path file isn’t too big, it is easiest to do the select link analysis on the fly using this file. Otherwise, the assignment script must be modified to do this.
Ohio Traffic Forecasting Manual

41. Select Link Analysis
To save the select link directly when performing an assignment, the script is modified as follows: This will save the select link assignment of cars and trucks in output volume fields 3 and 4 (1 and 2 contain the total cars and trucks). It is important to have a “V=“ statement in the ADJUST phase when doing this (standard model does).
Ohio Traffic Forecasting Manual

42. Distribution Checks
Screen line checks, as well as the desire line and select zone checks shown under trip rate checking, provide checks on the trip distribution of the model (how far and where trips from a given zone choose to go).
However, sometimes more detailed checking of the trip table (generally at a district level) is needed using independent data such as:
CTPP journey to work data
Trip table developed from surveys or purchased from vendors who process anonymous signal data (cell phones etc.) into OD tables
Ohio Traffic Forecasting Manual

43. Distribution Checks
When model ADJUSTMENTS (as detailed later) occur, it is usually necessary to recheck the region-wide distribution statistics just as is done during the base year validation:
Check average trip lengths by purpose
Percent of intrazonal trips by purpose
Trip length frequency distribution coincidence ratio by purpose
Ohio Traffic Forecasting Manual

44. Distribution Checks
File MPOYRTLFDAA.PRN from the standard model contains output TLFD’s that can be imported to a spreadsheet for analysis.
Report RPT4_DIST_TOD.PRN from the standard model contains the trip distribution summary report with average trip lengths and percent intrazonal statistics.
Ohio Traffic Forecasting Manual

45. Modal Checks
This training is focused on highway projects due to the many special circumstances involved with transit, however a few brief points:
The study area for transit studies will likely be defined by transit routes.
Route ridership is the primary thing to check and is generally available from the operator.
Other statistics ,such as transfer rates and user profiles, are often available for the entire system.
Special transit on-board surveys can provide other data, like stop specific boarding/alighting and route- specific OD patterns.
CTPP Journey to Work has modal OD patterns
Ohio Traffic Forecasting Manual

46. Generation Checks
Generation checks are especially useful to check trip rates when special generators are added, regional generation rates are changed, or TIS trips are being overlaid.
Ohio Traffic Forecasting Manual

47. Trip Rate Checks
Trips coming out of zones in the study area can be compared to ITE Trip Generation rates for non-residential land uses.
Additionally, previous Traffic Impact Studies should be consulted to ensure consistency with past assumptions.
Major discrepancies can be accommodated with special generators.
Ohio Traffic Forecasting Manual

48. Trip Rate Checks
There are various graphical ways to check trips from a zone in Cube or GIS.
Desire Line Map
Select Zone Map - Similar to select link, but selects a zone and shows the roadways that are carrying traffic to/from that zone.
Ohio Traffic Forecasting Manual

49. Scaled Symbol Theme/Chart Map
Trip Rate Checks
In Cube, open the trip table and the network.
With the network active, select Analysis > Link to Matrix.
Scaled Symbol Theme/Chart Map
Color Zone Map
Ohio Traffic Forecasting Manual

50. Special Requirements

51. Special Requirements
Models can produce various other traffic data besides 24- hour link volumes, including:
Turn Movement Volumes
Model Period (sub-daily) Volume
Directional Volumes
Truck Volumes
The use of these more detailed data requires additional checking. If checking results in refinements/adjustments to the model, then the 24-hour link volume checks must be repeated!
Ohio Traffic Forecasting Manual

52. Turn Movement Checks
Modeled turn movements are usually only used where necessary, i.e. when a given movement doesn’t exist in the base case.
It is possible, however, to check turn movements vs. turn counts using the same maximum allowable percent deviation curve used for link volumes (see %RMSE section).
Matrix estimation at the turn movement level can also be used to adjust turning movements.
Ohio Traffic Forecasting Manual

53. Turn Movement Checks
Turn movements are generally reviewed using turn movement diagrams. These are automatically saved during a standard model assignment (go to Intersections-Turn Volume File with the assigned network open to open the turn movements). Model period and 24 hour turn movements are available in files: MPOYRJTURNS24AA.DBF MPOYRJTURNSPMAA.DBF Etc. More detailed intersection operating characteristics (such as LOS) are available in the output junction files for the model periods only(no 24 hour). Go to Intersections-Output File with the assigned network open to open the intersection file): MPOYRJTURNSPMAA.INT
Ohio Traffic Forecasting Manual

54. Turn Movement Checks
Once opened, select the desired intersection node and pick the cross icon to open.
The DBF turn movement file can be further processed (say in a spreadsheet) to compare it to counts.
Ohio Traffic Forecasting Manual

55. Model Period Volume Checks
All of the models in Ohio now assign traffic at a sub-daily level, called model periods (typically Night-NT, AM Peak-AM, Mid Day-MD and PM Peak- PM).
These are added together to obtain daily traffic.
The sub-daily results should only be used directly if validated (in the study area).
Sub-daily counts can be coded in the appropriate standard network fields (COUNTAM, COUNTAMT, etc.).
In addition, there are period-specific overrides for most key input fields, allowing period level refinement and adjustment.
Ohio Traffic Forecasting Manual

56. Model Period Volume Checks
The period level volumes need to be subjected to the same validation tests as the daily volumes.
Period level validation
Model period percent difference by link check
Model period VMT check
Model period screen line check
Model periods proportion check (changes of >3 percentage points, should be flagged)
If these procedures are not followed, development of project level K should follow the traditional method.
Ohio Traffic Forecasting Manual

57. Model Period Volume Checks
The same maximum allowable percent deviation curve used in the %RMSE comparison is used to validate these volumes.
However, the target volumes listed under “Midpoint” on the desired error table shown previously should be multiplied by the percentage of traffic expected to occur in the model period in order to determine appropriate targets.
Ohio Traffic Forecasting Manual

58. Model Period Volume Checks
Example for period level validation.
2015 Urban Daily Traffic Distribution
Ohio Traffic Forecasting Manual

59. Model Period Volume Checks
Besides validating the base year period level volumes directly, the percent of the daily traffic occurring on a link in the forecast year should be compared to the percentage from the base year to ensure any difference is logical.
Ohio Traffic Forecasting Manual

60. Directional Split Check
The models are validated directionally, so a well-validated model should be able to give directional volumes.
This is typically only an issue when using sub-daily model results since the 24 hour results tend to be balanced.
Comparisons between count and model directional split and base year versus forecast year directional split are useful to ensure these make sense in the study area.
Ohio Traffic Forecasting Manual

61. Directional Split Check
If directional split information is used, the following checks should be conducted:
Map of model vs. count directional split in base year, coloring links with a difference of >3 percentage points.
Map of base year vs. forecast year directional split, coloring links with a difference of > 3 percentage points.
These checks are used to ensure that the directional splits are reasonable. Model refinements or adjustments should be made if necessary.
Ohio Traffic Forecasting Manual

62. Truck Volume Check
Most of the models have separate truck components and use multi-class assignments to provide both car and truck volumes with all assignments (trucks are class 2 in the standard model).
Truck volumes should go through all of the same validation checks independently, if they will be used (truck volumes were validated regionally for all of the standard models).
However, since the maximum allowable percent error curve is based in capacity, the curve must be adjusted to passenger car equivalents (PCE) prior to use.
For example, if PCE = 2, then the maximum allowable error of 100% would apply to truck volume = 500, instead of an AADT = 1000.
Ohio Traffic Forecasting Manual