Kentucky State Model Users Group

1
Kentucky State Model Users Group

• Network Development
• May 27, 2004

2
Subjects of Discussion

• Line Sources
• Centroids and centroid connectors
• Use of node layer
• Attributes
• From HIS
• From other sources
• Capacities
• Speeds
• Attaching network data
• Count data

3
Subjects of Discussion

• Network Review
• Node Density
• Quality control for link attributes
• Aerials
• Connectivity checks
• Directionality one way couplets, reversible
  lanes
• Network comparisons and future networks

4
Line Sources
5
Line Sources

• Generally, we inherit a network
• Could be a MINUTP or TRANPLAN stick network
• Sometimes, converted to TransCAD
• Sometimes conflated sometimes not
• In Lexington, conflated but lots of tunnel links,
  connectivity issues and directionality issues
• Clients never want to start from scratch which
  would sometimes be easier

6
Line Sources

• When adding new network or conflating, we
• Generally, conform to Tiger lines for conformity
  to Census geography
• Sometimes, local government GIS is more accurate,
  when available but then doesnt always conform
  perfectly to Census geography
• Always trace to avoid connectivity concerns
• Aerials consulted, when available

7
Line Sources

• Centroids and centroid connectors
• Consult aerials where possible for centroid
  placement
• Almost always load centroids as mid-link
  connectors
• Reasonable exceptions
• centroid connectors to model the third or fourth
  leg of a rural intersection
• occasionally, for strip commercial or mall TAZs
  where the centroid connector can double as a
  centralized main entrance
• We do not locate centroids as nodes on a link

8
Line Sources

• BLA has developed GIS/DK tool for automating
  centroid connections. Logic includes
• User-specified desired number of connections/TAZ
• Looks for the closest road to the centroid
• Prohibits connection to a freeway or expressway
• User-specified tolerance to avoid connections to
  facilities beyond the zone
• From first connection, move 90º or 180º depending
  on number of connectors allowed
• Prohibits connections to the same street, if the
  facility curves

9
Node Layer

• Lexington in the network but not used. Just
  designates connectivity
• At times, have used nodes to designate
• Centroid or not
• Presence or absence of a signal
• Knoxville
• Indiana Statewide Model (v.4)
• Lexington network does have signals, but treated
  strictly as a link attribute

10
Node Layer

• Nodes as signals
• Model logic only considers signals as a link
  attribute (by approach) created by GIS/DK tool
  with logic that looks at
• Nodes
• Other link attributes
• Node attribute says whether or not there is a
  signal here
• Then, double-checks link functional class or
  facility type to make sure it is not a freeway
• Many networks represent interchanges as a node

11
Node Layer

• Nodes as signals
• Why bother?
• Allows analyst to test treatments that are
  otherwise difficult in travel demand modeling
• Effect of signal removal or installation on
• volumes
• travel times
• capacities
• Allows for limited testing of certain TSM
  improvements (e.g., signal timing progression)
• Important in todays climate of evaluating new
  construction versus upgrades

12
Attributes
13
Attributes

• HIS for basic attributes
• HIS extracts to ArcView shape files available for
  the following layers
• I.D. Route, suffix, name, county, HIS_ID
• Geometric and Functional Data functional class,
  number of lanes, lane widths, median widths when
  present, left and right shoulder widths
• Count station and variety of counts, source, year
• Posted Speeds In Lexington, manually entered
  from HIS database
• Strong recommendation Make posted speeds
  available as an HIS extract layer

14
Attributes

• Non-HIS
• In Lexington Model
• Area type
• Flag field that identifies reversible lanes
  Nicholasville Road
• Number of lanes in each direction by time-of-day
• Bike lane flag field (for informational / mapping
  purposes)
• Signal or 4-way stop node layers available from
  Lexington-Fayette County GIS. Cobbled together
  for Jessamine County
• Multi-signal code number of upstream signals
  that affect arrival type. Worked with MPO and
  Engineers Office
• Signal priority High, equal, or low in
  comparison to cross street

15
Attributes

• Non-HIS
• Sometimes,
• Ramp code Off/on, system ramp, minor-to-major
  freeway, cross-over link
• Volume-delay function parameters
• sometimes, coded by facility type in the job
  stream and is overridden when link values are
  entered

16
Capacities and Speeds

• Strong belief in calculated speeds and capacities
• Why?
• No table is comprehensive enough to reflect
  variable conditions on the ground
• Permits testing geometric and even operational
  alternatives that speed-cap tables cannot
  accommodate

17
Capacities and Speeds
Facility Type f (number of lanes, access
control, presence or absence of median,
directionality, area type)
Free flow Speed f (facility type, posted speed,
delays on interrupted facilities)
Capacity Maximum Service Flow Adjustment
Factor Adj. Factor f (facility type, lane
width, shoulder width, median width, heavy
vehicles, directional K-factor, delay on
interrupted facility
18
Capacities and Speeds

• Delays on interrupted facilities based on HCM
  2000 methods
• Delay/Vehicle uniform delay PF incremental
  delay initial queue delay
• where, PF progression factor f (arrival
  type, g/C)
• uniform delay
• Varying g/Cs and PFs by signal priority and
  multiple signals
• PF Isolated signals Arrival Type 4
  (Favorable progression)
• Multiple signals Arrival Type 5
  (Highly favorable progression)
• Default cycle lengths 90 sec.

0.5C(1-g/C)2 1-min(1,v/c) g/C
19
Attaching Network Data

• Always hard work
• One way GIS/DK Tool If separate layers of data
  are available, creates a node in the network
  layer conforming to beginning and ending points
  of separate data layers. Then, use GIS tagging
  function to populate the attributes very
  carefully
• Conceivably, could code HIS into a TransCAD
  routing system and export data to network.
  However, lots of work creating a routing system

20
Count Data

• Should year of last count be included?
• Usually, doesnt matter since normal count
  variability is 20
• Sometimes multiple, recent counts are available.
  Which to use?
• If normal variability among counts, use the mean
• If not, pay attention to the model. Can often
  help to identify which to throw away.

21
Network Review
22
Minimizing Node Density?

• Can be helpful if there are truly redundant
  nodes.
• However, high node density may be a good thing
  just means a rich data base
• Typically, not an issue for us.

23
Reviewing Link Attributes

• Use TransCAD selection set functionality and
  check for logic of speeds, capacities, etc.
• TransCAD has built-in tool for generating basic
  statistics
• Averages
• Min
• Max
• For model output variables, compute weighted
  means and check for reasonableness
• VMT and VHT by functional class
• Weighted congested speeds by functional class

24
Checking against Aerials

• Bring up as a raster image whenever available
• Great for checking
• Appropriateness of splitting zones Shouldnt be
  afraid to.
• Locating centroids
• Locating centroid connectors

25
Connectivity Checking

• Use TransCADs Check Line Layer Connectivity
  Function
• It checks to see if there are stray nodes and
  flags suspicious connectivity Mainly used for
  catching snapping errors
• Functionality that looks at nodes within a
  user-specified distance of one another. Flags
  nodes inside distance into three categories
• Nodes with only one connection. Easy check to see
  if they are all centroids
• Nodes in only two connected lines
• Nodes with three or more connections

26
Directionality Checking

• Fire the guy who doesnt do this! Just a standard
  step in the validation process
• Run random paths across the study area cordon
  line to cordon line (Fire the guy who doesnt do
  this, too!)
• Use TransCADs Style arrow heads functionality
• Any backward arrows in the stream? Especially
  helpful in CBDs and complex interchanges

27
Network Comparisons EC and Future Alternatives

• The ultimate answer to questions about
  consistency among networks Have only one
  network
• TransCAD allows for multi-scenario networks
  We call it Multi-Net
• Comprised of two basic features
• Repetitive blocks of link attributes for each
  scenario
• New links specific to a given scenario
• Lexington has six scenarios 2000, EC, 2010,
  2020, 2030 and 2004

28
Network Comparisons EC and Future Alternatives

• In Lexington, BLA built script into the master
  program stream
• Allows user to extract from the Master Network
• the block of link attributes needed for the
  scenario
• the unique link records needed
• User must provide
• an _number for the scenario
• a flag field which identifies the records to be
  used

29
Network Comparisons EC and Future Alternatives

• BLA has since developed a user interface with
  drop-down lists for the scenario number and
  flag field

30
Network Comparisons EC and Future Alternatives

• Advantages of Multi-Net over the normal
  proliferation of networks
• Can copy a project from one scenario to another
  choose selection set, then copy from Scenario A
  to Scenario B
• Useful for adding committed project to Base or
  LRP scenarios
• Fewer questions about network consistency
• Assurance that the same project is always coded
  the same way

31
thank you!
32
Kentucky State Model Users Group

• Trip Generation
• May 27, 2004

33
Subjects of Discussion

• Trip Purposes
• How to choose them
• University trip purpose
• Special Purposes
• Predictor Variables
• Preferred variables
• Changes over time
• Procedures and data sources
• Sources of Trip Rates
• Deviation of Trip Rates
• Geographic or socioeconomic stratification
• Special generators

34
Subjects of Discussion

• External Trips
• Procedures and Preferences
• Together or Separated from Internal Trips
• Model Calculations

35
Trip Purposes
36
Trip Purposes

• Is there a preferred breakdown?
• No one taxonomy is right for every situation.
  Depends on
• Size of modeled area
• Are there major special generators? Especially,
  universities?
• Is modeled area multi-nucleated?
• The question of the number and type of trip
  purposes should not be divorced from larger model
  design questions
• Will there be time-of-day modeling?
• Do trip lengths or departure times differ even
  in the same purpose?
• Will there be a formal mode choice model?

37
Trip Purposes

• Given a consistent model design, it is possible
  to classify cities within a state into broad
  categories and apply consistency within those
  categories
• A good breakdown applicable to most cities

• Visitor Trips
• Internal Non-Freight Trucks
• E-I Freight Trucks
• E-I Auto Work Trips
• E-I Auto Non-Work Trips
• E-E Auto Trips
• E-E Truck Freight Trips

• Home-Based Work
• Home-Based School
• Home-Based Shop
• Home-Based Other
• Non-Home-Based Work
• Non-Home-Based Other

38
College Towns

• Is a special trip purpose needed?
• Answer Is the campus a special generator?
  Another way to ask the question Are students
  generating lots of vehicle trips?
• BLA models with large universities
• All have significant off-campus housing, except
  South Bend and Terre Haute
• All have a special college trip purpose, except
  South Bend and Terre Haute

• Muncie, IN
• Terre Haute, IN
• Evansville, IN

• Lexington, KY
• Knoxville, TN
• Bowling Green, KY
• Lafayette, IN

39
College Towns

• Most colleges very good about working with MPOs
  to share information
• Can usually get all off-campus housing addresses
  and address-match to TAZ
• 1999 travel survey of I.U. student households
  conducted by campus e-mail Great response rate
  very useful

40
College Towns

• TRIP ENDS (all purposes and locations)
• 56 originate on-campus
• 58 destined for campus
• TRIP ENDS BY PURPOSE (all locations)
• Home-Based Class Person-Trips / Off-Campus
  Household 2.05
• Home-Based Class/Work 53.0
• Home-Based Other 30.8
• Non-Home Based 16.2
• TRAVEL MODE (all purposes and locations)
• Auto 56
• Bus 11
• Bike 4
• Walk 29

41
Other Special Purposes?

• Military bases, large retirement developments,
  resort towns
• Look for large residential populations that
  generate off-campus trips
• In our experience, large medical centers usually
  dont meet this condition
• Can be handled with special generator rates on
  the attraction end
• Special events large concerts, sporting events
  ?
• Use a special purpose and rely on local knowledge
  versus ITE rates
• Pros/Cons Special purposes offer more
  advantages than disadvantages
• Can isolate the generator and maintain better
  control in model calibration / validation

42
Predictor Variables
43
Predictor Variables

• Cross classification always preferred over
  regressions for production generation
• No assumed linearities
• Greater temporal stability
• No aggregation biases
• Also, better accuracy relying on home-based cross
  class for Non-Home-Based regional control totals
  vs. zonal regressions used alone
• Better accuracy if flexible in the choice of
  household predictor variables vs. using same two
  for all trip purposes
• Workers x Autos preferred to Persons x Autos for
  HBW
• School-Aged Children preferred to Persons x Autos
  for HBSch
• Workers x Autos preferred to Persons x Autos for
  NHBW

44
Predictor Variables

• Rule-of-thumb Autos better predictor than income
• Are trip rates (and trips lengths) changing over
  time? Yes!
• Increase in National Person-Trips/Person
  (77-01) 49.2
• Annualized 1.68
• Increase in National Person-Miles/Person Trip
  (77-01) 10.4
• Annualized 0.41
• Recent trends Still growing, albeit slower
• A good case for occasional household travel
  surveys as part of model update development
• Source NHTS and NPTS

45
Procedures / Data Sources

• Best Practice
• Conduct household travel survey
• Non-parametric correlation analysis to select
  best predictor variables
• ANOVA to determine appropriate cell compression
  scheme
• Curve smoothing to eliminate small cell sample
  irregularities

Little Rock, Arkansas
46
Procedures / Data Sources

• Best Practice (cont)
• Curve fitting to estimate household
  stratification model
• Geo-coding of non-home-based trip ends as part of
  household travel survey
• Regression analysis and coefficient expansion for
  trip attractions
• Trip production rate expansion for under-reporting

47
Sources of Trip Rates
48
Sources of Trip Rates

• Growing interest in the use of GPS as independent
  checks on under-reporting in household travel
  surveys. Ex Ohio GPS Check found huge
  under-reporting. Multipliers
• NCHRP 187 too dated
• NCHRP 365 valuable if used selectively
• NPTS standard urban area samples too small
  helpful only if add-on sample is available
• Preferred Conduct survey / adjust or borrow
  rates from similar urban area

HBW 1.22 HBO 1.6 HHBW 2.32 NHBO 2.18
Trips lt 7.5 minutes 2.57 Trips gt 7.5 minutes
1.41 Overall expansion 1.70
49
Deviation of Trip Rates
50
Deviation of Trip Rates

• Our experience trip rates usually do vary in
  large multi-nucleated regions
• Rural households not too different from urban any
  more
• Use of special generation rates from ITE Trip
  Generation (7th edition) can be helpful, but no
  substitute for experimentation in the model
  validation process

51
Deviation of Trip Rates
Proposed Ohio Small Mid-Sized MPO Standard
Household Model Disaggregation into Income
Quartiles in each TAZ based on Public Use
Microdata Sample (PUMS)
Productions Separate rates by Four Area Type
of TAZ
HBW Workers x Autos
HBSch Children x Autos
HBShp HHSize x Autos
HBO HHSize x Autos
NHBW HHSize x Autos
NHBO HHSize x Autos
Area Types CBD, Urban, Suburban, rural
6 Purpose x 4 Income Quartiles Visitor Trips
25 Internal Trip Purposes
52
External Trips
53
External Trips

• E-E vs. E-I Splits at External Stations
• No substitute for a cordon line survey,
  regardless of survey method used
• If survey not feasible
• Critical link analysis at cordon line stations to
  estimate through movements in Statewide Model
• NCHRP 365 Modlins Equations for urban areas lt
  100,000 population
• E-Ei f (Population, functional classi ,
  heavy trucksi light trucksi , ADTi)

54
External Trips

• E-I Internal purposes or a single trip purpose?
• Simpler to treat as one, but often not very
  accurate. Works best when
• No time-of-day modeling
• Cordon line is far removed from urban activity
  destination choices have little to do with travel
  time
• Reasonably simple alternative E-I trips broken
  into trip purpose s based on NCHRP 365
  recommendations, then apply separate directional
  splits if TOD modeling needed (Holland MI Model)

55
External Trips

• E-I Internal purposes or a single trip purpose?
  (cont.)
• Ohio State Standard
• Cordon line survey for each MPO
• Auto trips stratified into work and non-work
• Friction factor based accessibility measures in
  the E-I regressions
• Sneak preview of Trip Distribution Uses a
  combination of travel time and angle of
  deflection at the cordon line as a composite
  impedance
• Problem with many external models the presence
  of an island of development just inside or
  outside the cordon line. Requires customized
  solution.

56
Model Calculations
57
Model Calculations

• TransCAD script or external spreadsheet?
• Advantages of script
• Once developed, easier to use than a spreadsheet
• Therefore, encourages use of the model
• Facilitates customization
• External cordon line problem can be hard-coded
• Opportunities for soft-coding to allow for new
  special generators and balancing methods
• Disadvantages of script
• Slightly more up-front development cost

58
thank you!