WP3 Network Modelling

1
WP3 Network Modelling

• Otto Anker Nielsen, oan_at_ctt.dtu.dk
• Professor, Ph.D.
• Technical University of Denmark (DTU)
• Centre for Traffic and Transport (CTT)

2
Presentation overview

• Data formats and interfaces
• Road Network
• Rail Network
• Air Network
• Other networks
• Zones and matrices

3
Road network

• Road network is now coherent and quality
  controlled
• Road charging data is not available at link level
• Assume on fixed links (bridges, tunnels) in
  countries with no general road tax
• Norway assumed to be Oslo and Trondheim only
• Assumed to be all motorways in other countries
• Border resistance
• Note has been received but not coded
• Ferries
• Have been coded and quality controlled
• Frequency, cost, and time remains to be estimated
  (based on standard function)

4
Network - Rail

• Source (2)
• UIC
• Received
• 28 April, 2005
• Coverage
• Primary and secondary network in Europe
• Some secondary links are in fact primary
• No. of records
• 38,400 rail segments
• Right projection

• Source (1)
• GISCO version II TEN-STAC
• Received
• 28 April, 2005
• Coverage
• Primary rail network in Europe
• Some primary links are in fact secondary
• No. of records
• 17,600 rail segments
• Wrong projection

5
Network - Rail

• Map (1)

• Map (2)

6
Completed validation and actions

• Validation of rail network
• Main rail network is spatially insufficient
• Secondary network does not include attributes
  except length
• Combination of the networks possible via IDs
• Strange values of variables e.g. travel time
  -31,767
• Projection in ArcGIS not correct
• Actions
• Combination of the two networks and estimation of
  attributes on secondary links
• Editing of attribute values
• Missing attributes Frequency, cost
• Using detailed network to correct for wrong
  projection in the other

7
Complete Rail network

• Key figures for the layer Rail_gisco_ctt
• 38,391 records
• Data on type of the railway line
• No relevant attributes
• The best network integrity
• The most complete network

8
Network Adding rail lines

• 16.000 km rail were added manually to the
  existing basis network (indicated by red)
• The aim was to provide the basis network with
  extra rail lines were necessary, to generate a
  realistic route choice

9
Network improvements

• At the same time a first hand quality check of
  the basis network was performed.
• Securing and repairing the network integrity
• Regenerating the FromTo and ToFrom node
  attributes

10
Estimating speeds

• All the manually added and many of the basis
  rail lines are not provided with any indication
  of speed
• Consequently different techniques for
  estimating travelling speeds, depending on
  available data are considered. The technique
  which delivers the best speed estimate is of
  course used as extensively as possible
• In theory using the passenger and freight time
  attributes should give the best results. Time
  attribute is based on the UIC network model and
  therefore requires a recalculation to fit our
  network. The speeds derived from this do on
  average seem sensible, but going into detail
  reveals that there is a huge variance in the
  speed on the same line
• On average the traveling time derived speed is
  70 of the top speed attribute. It is decided to
  use 70 of the top speed as traveling time for
  first choice traveling speed, since the speed
  variations in this way seem more credible

11
Estimating speed preference The procedure

• If the maximum speed available then 70 of the
  maximum speed
• Speed calculated on the basis of travelling time
  information (from UIC network model)
• Speed estimated on NUTS2 level from similar types
  of rail lines
• Speed estimated from similar types of rail lines
  in general
• Manually added lines are given a speed of 50 km/h
  for passenger and 40 Km/h for freight
• The few remaining are manually estimated

12
(No Transcript)
13
High-speed rail

• Not a special mode, but special link types with
  higher speeds
• Treated as part of the passenger assignment model
• NOT as a special mode in the mode choice model

14
Calculation procedure, rail

• Frequency is assumed fixed or estimated in a
  special loop in the assignment
• Frequencies are proposed to be dependent on
  railway type (major, minor, double or single
  track, electrified or not), and volumes
  (volume-frequency function)
• The assignment depend on cost and time and follow
  a stochastic loading procedure
• The frequencies are collected along t he chosen
  paths and averaged
• This is NOT used in the assignment, since no
  knowledge exist on lines and transfer points
• BUT the averaged frequencies are used in the
  passenger demand model

15
Air network

• Geocoded list of airports
• All international airports
• 290 airports in total
• Fairly thorough quality control revealed good
  quality
• Missing
• Former Yugoslavia (all states)
• Albania
• Russia, Belarus, Ukraine
• Keflavik (Iceland)
• Turkey?
• Greenland?

16
Air network

• Air network
• IWW Air network consisting of origin and
  destination airport, distance, time and 3 cost
  categories (Cost-B, P, H)
• 4,805 air connections
• Appears to have quite few connections per airport
  (16.5)
• gt led to a manual inspection of connections on a
  random sample

17
Air network major airports

• Connections found from airports homepages using
  Google combined with air specific search engines
• Copenhagen
• 125 connections in reality. 75 in Transtools
• Malaga
• 105 in reality
• 50 in Transtools
• Maybe OK, since the more important connections
  are present, BUT

18
Air network medium sized airports

• Seville (Spain)
• 30 In real life as of 16/1-2006
• 13 in Transtools (of which 9 are also existing
  according to home-page)
• Trondheim (Norway)
• 24 in real life
• 6 in Transtools (5 in common)

19
Air network minor airports

• Pampolona (Spain) PNA
• 26 connections as of 13/1-2006, hereof 4 to major
  hubs
• 1 (one) in Transtools, only to secondary hub
  (Barcelona)
• Brest (France), BES
• 20 connections as of 13/1-2006
• 7 in Transtools (of which only 4 are existing
  according the to homepage of the airport)
• Several connections to foreign cities are missing
• Sturup (Sweden)
• 25 connections according to homepage
• 3 in Transtools
• La Rochelle (France) LRH
• 8 as of 13/1
• 2 according to Transtools (one common with
  homepage)
• 6 international flights are missing

20
What to do?

• Identify major and national hubs (refer to
  discussion on air access models), i.e. 46 major
  hubs
• Add missing connections to/from all major hubs
• gt This would be likely to double the number of
  connections
• gt But most likely the ones with highest demand
  and frequency
• This will secure that medium and small airports
  are connected to all correct major hubs (as each
  link is bi-directional)
• Links between medium sized and small airports are
  not quality controlled, i.e. 244 airports are not
  checked
• Justification
• The quality control is quite time consuming
• Finding the airport homepage and double
  controlling with air scheduling engines
• I.e. the effort is almost identical per airport

21
Inland waterways and harbours a

• Inland waterways OK
• Harbours along inland waterways are assumed to be
  in all NUTS III they pass?
• What about non-connected inland waterways?
• Edgar has delivered two possible methods for
  calculations?
• Harbours
• Important harbours need to be coded (Ming
  Christian solve this)

22
Zones and matrices

• Zone structure need to be finally agreed upon
• Zonal data have to be sent to CTT
• Country level, NUTSII, NUTSIII
• Matrices should be delivered to CTT
• IWW needs to provide information on conversion
  from passengers to cars
• This is a pre-condition for CTT to be able to
  calculate LoS matrices as well as availability
  matrices (inland waterways, ships for TNO)