FSUTMS Models Feedback Survey Results
Xiao Cui
Shi Chiang Li
Makarand Gawade
March 3rd 2016
Background
The objective of this survey was to understand how the FSUTMS models are beingapplied, their effectiveness and what areas need improvement. The questions includeidentifying concerns, potential improvements of models and usage of models fordifferent purposes.
The survey link was open from February 2nd, 2016 to February 19th, 2016.
There were 21 responses with 71.43% of respondents indicating they are ‘ModelDevelopers’, 66.67% of respondents indicating they are ‘Model Practitioners’ and23.81% of respondents indicating they are ‘Model Consumers’.
73.68 % of the respondents indicated that they have experience of developing DesignTraffic for highway projects using model.
85.71% of respondents indicated that they have performed subarea model validation
2
What have you been using the models for?
3
0
2
4
6
8
10
12
14
16
18
20
Developing LongRange Transportation
Plans (LRTP)
Forecasting trafficdemands for newhighway facilities
Estimating corridortransit ridership for
Federal TransitAdministration (FTA)
grants
Estimating revenuefor toll facilities
Applying model forland use amendment,growth management
or site impactanalysis
Planning Freight andGoods Facilities
Planning stage oftransit corridors thatmay be submitted to
FTA for funding.
Nu
mb
er
of
res
po
ns
es
Use of Models
In your opinion, FSUTMS models are effective tool:
4
0
2
4
6
8
10
12
14
16
18
20
For identifying new and/orwidening roadway projects
For estimating and planningsystemwide transit needs
For estimating and planning ITSrelated facilities
For testing alternative land usescenarios
For planning freight and goodsfacilities
Nu
mb
er
of
resp
on
ses
Strongly Disagree
Disagree
N/A
Strongly Agree
Neutral
Agree
What are the modeling issues related to applications of models for developing LRTP?
Model Validation for O-D Trip patterns
Suitability of ABM models and DTA.
Model noise between alternatives / scenarios
Region wide calibration not sufficient for corridor-level forecasts
To develop tools that can be used by all TPO to reduce the cost
Future networks and accurate forecasts of zonal data
Better definition of what is a good model and the best practice for how to use themodel for LRTP.
Include other modes5
What model improvements should be implemented for LRTP developments?
Incorporate features like MOVES, AV/CV procedure, ELToD, toll modeling
More standardized reporting of model outputs.
Consistent analysis across the State for ITS improvements, toll/managed lanefacilities, transit facilities (rail) and commodity flows.
Build user's confidence that the model could provide reasonable projection.
Alternative fuels, vehicle choice, integrated land use are types of policies to beaddressed.
Standard truck model that includes delivery, short haul and long haul trucksStandard ITS application (FITSEVAL), Standard Managed Lanes model.
6
Experience of developing Design Traffic for highway projects using model?
86% of the respondents indicated that they believe FSUTMS models are capable ofprojecting traffic for new roadways/bridge or roadway widening projects.
21.43 % of respondents indicated that they have never used other post modelrefinement process, such as incorporating Analyst ODME process for design traffic,whereas 42.86% indicated that they have used it rarely.
7
How do you apply models to develop Design Traffic?
8
0
2
4
6
8
10
12
Use model output directly Develop growth rates from base and horizon yearmodels and apply which on existing traffic count
Develop composite growth rates from zonal dataand model link volumes
Num
ber
of re
sponses
Always Never Rarely Sometimes
What are the modeling issues you have encountered while applying the models for Design Traffic?
Inconsistent for O-D traffic patterns compared to surveyed O-D trips, average travel distance,and average travel time
Study area level of validation
Depending on the project and time schedule, model refinements cannot be made for thecorridor or subarea. Traffic smoothing must be applied and is not always a preferred method.
Issues with socio-economic data, trip rates, trip distribution, mode choice and assignment beinginaccurate.
Over forecasting or under forecasting traffic
Useful for system-wide interactions of land use and major new capacity only.
Conversion of PSWADT to AADT has been automated in models and should be standardized
Standard "Design Hour" procedures do not work well for the larger urban areas9
In your opinion, what model improvements should be implemented for Design Traffic?
Revise parameter values in trip distribution models or check impedance values for highway path-building.
Develop a model that will produce design hour traffic directly.
Off model techniques. Procedure to implement Cube Analyst (already being developed).
Refine socio economic data (parcel size), refine network by considering all streets, refine mode choicewhich includes bicycle-walk, refine time period as least hourly and move away from gravity distributionmodel.
More outreach teaching public agency to build confidence and the best practice.
Better post-model tools that appropriately take model output into account in the design analysis.
Have every model produce AADT as well as the standard PSWADT which can be accomplished throughthe use of a lookup table of MOCFs and the coding of the MOCF on each link (NWFRPM does thisalready).
10
Besides zonal and network descriptions, what model parameters you have modified for validation
11
Friction Factors Capacity
Trip rates Vfactor (BPR curve)
Convergence criteria Ctoll
SPDCAP Time penalty
GEN Turn prohibition
SE Special generation rates
Speed Behavioral assumption
Zonal data Coefficient of utility functions in
mode choice model
Toll parameters Number of iterations
Checks for network impedances Auto occupancy factors
EE / EI trip tables Trip length
With your knowledge, what new observed information/data is available to improve modeling process?
12
O-D survey, speed using
Bluetooth or navigator
2009 NHTS national and state
Add-on, New household survey
was completed in Martin county
Airsage data Info-group
StreetLytics Urban Engines
Aerial and other GIS data Cell phone GPS data (with
concerns of accuracy and
assumptions)
Socio-economic data at parcel
level
Additional modes like bicycle
and walking
Visualization tools Other crowd sourced data
(better understanding needed)
How can the current model structure or process be streamlined to make FSUTMS models more agile?
Needs both enhanced transit model and standard process for preparation of inputhighway network based on GIS true shape, free flow speed, junctions, capacity and soon. Also, requires a validation of highway network.
For ABMs, procedures should be identified so that users know when to run differentmodel components
No fundamental model structure error.
Districting in larger regional models for decreased run times in assignment
Highway Only options in areas where Transit mode share is less than say 2%.
Eliminate components of the model that may not be necessary or cumbersome.
13
In your opinion, what practical model enhancements can be implemented in the next three years? (1)
Transit modeling along with mode choice modeling, and toll studies for both fixed and dynamic tolls.
Standard Florida trip production rates, friction factor curves, auto occupancy factors, and time of day factors based on the 2009 Florida NHTS Add-On survey.
A revamping of standard Florida network coding procedures is also overdue now.
Implementation of DTA for Managed Lane and NCHRP 765 Process for design traffic process.
Standardized ABM deployment, rather than every district proceeding on their own.
Re-structure socio-economic input variables. Use non-dbf inputs.
14
In your opinion, what practical model enhancements can be implemented in the next three years? (2)
Activity based models depending on area. Adding of bicycle and walk modes. Implementation of standardized process for transit, ITS, toll/managed lanes, freight/commodity analysis.
Faster run time. Running under Cube Cluster will NOT be able to duplicate model results running in different computers.
Enhance the model improvement based on best practice guideline from FSUTMS.
Improvement of Highway Travel Time Estimation.
Integration of statewide freight and passenger model with the new travel demand modeling systems : Time of Day Modeling, trip chaining, junction-based modeling, and dynamic traffic assignment.
A plan for the establishment of information technology resources for travel survey data collection and analysis
15
Please rate key concerns of the models (rating 1 has lowest priority and rating 10 has highest priority)
16
0 1 2 3 4 5 6 7 8 9
complexity of network descriptions
over specification of mode split structure
complexity of zonal data
documentation -- lack of users guides
documentation -- lack of model concept/algorithm
difficulty of trouble shooting
accuracy of truck percentages on links
accuracy of mimicking CTPP journey to work
model run time
accuracy of transit ridership at route level
accuracy of free-flow and congested speeds
accuracy of trip lengths
accuracy of link volumes (compared to counts)
Please rate types of model training (rating 1 has lowest priority and rating 10 has highest priority)
17
0 1 2 3 4 5 6 7 8 9 10
comprehensive training for four-step model
growth management, site impact, land use amendment
long range transportation plan
transit ridership
fundamental of statistics and model estimation and calibration
modeling freight
comprehensive training for Activity-Based model
highway forecasts
toll road and managed lanes
guidelines/workshop for model application, post model processing and forecasts
Please rate types of model projections you would need to have in the next five years (rating 1 has lowest priority and rating 10 has highest priority)
18
0 1 2 3 4 5 6 7 8 9 10
resilience of climate change
travel demand changes due to internet services (e.g., on-line banking) and shopping
travel demand changes due to fluctuation of gasoline price
travel demand changes due to increase of flexible working hours
travel demand changes due to transportation network services providers (e.g., Uber)
travel demand changes due to millennium generations of urban living style preference
transit ridership for light rail
travel demand changes due to autonomous car
freight/turck trips for inland logistic centers
transit ridership for commuter rail
transit ridership for BRT
highway project needs assessment
Please provide any additional comments or suggestions
More FDOT resources need to be allocated to survey data collection
Model improvements need to be sensitive to the agency and priorities. Thereare some obvious trends ahead of us that we all need to understand but Iwould stress that a complex model addressing these trends in detail may notbe appropriate for all agencies.
19
Thank you
20