Professional Documents
Culture Documents
February 2010
Notice
This report was produced by Atkins limited for High Speed Two Limited for the specific purpose of High
Speed Two Modelling Framework Development.
This report may not be used by any person other than High Speed Two Limited without High Speed Two
Limited’s express permission. In any event, Atkins accepts no liability for any costs, liabilities or losses
arising as a result of the use of or reliance upon the contents of this report by any person other than High
Speed Two Limited.
Document History
Matt Michael
1 Draft 1 Steve Miller Steve Miller 01/02/10
Carlson Hayes
Contents
Section Page
1. Introduction 5
1.1 Background 5
1.2 Purpose of Report 5
1.3 Organisation of Report 5
2. Overview 6
2.1 Project Scope and Requirements 6
2.2 HS2 Forecasting Framework Specification 9
3. HS2 Forecasting Framework 12
3.1 Framework Description 12
3.2 Framework Operation 14
4. PLANET Long Distance Model 16
4.1 Introduction to the PLANET Strategic Model 16
4.2 Key Features of PLANET Strategic Model 16
4.3 Key Elements of Existing PSM 17
4.4 PSM Update to PLD 21
5. PLANET Long Distance Rail Model 24
5.1 Overview 24
5.2 Demand Matrices 24
5.3 Rail Fare Matrices 33
5.4 Rail Network 34
5.5 Assignment Process 38
5.6 Station Choice 43
5.7 Interfaces 48
6. PLANET Long Distance Highway Model 51
6.1 Overview 51
6.2 Supply Update - Highway Schemes 51
6.3 Demand Data Update 51
6.4 Highway Assignment Process and Parameters 52
6.5 Interfaces 53
7. PLANET Long Distance Air Model 54
7.1 Overview 54
7.2 Demand Data Update 54
7.3 Network Update 56
7.4 Air Assignment Process and Parameters 57
8. PLANET Long Distance Mode Choice Model 58
8.1 Overview 58
8.2 Summary of PSM Mode Choice Model 58
8.3 Review of Existing Mode Choice Model 60
8.4 Recalibrated Mode Choice Model 62
8.5 Distributed Value of Time (DVoT) 64
8.6 Summary of Approach 71
List of Tables
Table 2.1 - Modelling Options 10
Table 4.1 - Treatment of Premium Fares 22
Table 5.1 - Postal Sectors in Solihull 28
Table 5.2 - Treatment of Ticket Types in LENNON 31
Table 5.3 - Journey purpose splits by ticket category 31
Table 5.4 - Approaches to Rail Fare Representation 34
Table 5.5 - PLD Network Bottlenecks 38
Table 5.6 - Assignment Types 39
Table 5.7 - Assignment Parameters 40
Table 5.8 - Stations in PLD Station Choice Model 44
Table 5.9 - Greater London Station Choice Zones 47
Table 5.10 - West Midlands Station Choice Zones 47
Table 5.11 - Example Station Choice Output 47
Table 5.12 - Demand Interface between PLD and PS 49
Table 6.1 - Highway Demand Totals 52
Table 6.2 - Highway Volume Delay Functions 53
Table 7.1 - Annual Domestic End-to-End Air Demand 54
Table 7.2 - PLD 2007 Daily Air Demand 55
Table 7.3 - Percentage of Demand to SE England 55
Table 7.4 - Air Passenger Data & Modelled Flows at Heathrow Airport 56
Table 7.5 - Air Assignment Parameters 57
Table 8.1 - PSM Mode Choice Parameters (2002 Prices and Values) 59
Table 8.2 - Comparison of VoTs (p/min ,2002 prices and values) 65
Table 8.3 - Mu (μ) and Sigma (σ) by purpose 66
Table 8.4 - Business Fares (Average Pounds, one-way) 68
Table 8.5 - Leisure Fares (Average Pounds, one-way) 68
Table 8.6 - Business Demand 69
Table 8.7 - Leisure Demand 70
List of Figures
Figure 3.1 - PLANET Long Distance Model Framework 12
Figure 3.2 - Base Case Run 15
Figure 3.3 - Test Run Extension 15
Figure 4.1 - PSM Network Coverage 17
Figure 4.2 - PSM Model Elements and Sequence 18
Figure 5.1 - PLD 235 Zone System 25
Figure 5.2 - Postal Sectors in the West Midlands [NRTS leisure trip-rates] 27
Figure 5.3 - Concentration of districts providing journeys to/from particular stations 29
Figure 5.4 - Percentage of car available trips by originating station (NRTS) 29
Figure 5.5 - PLD Matrix 'holes' 33
Figure 5.6 - PLD Rail Network 35
Figure 5.7 - PDFH Adjustment Output 41
Figure 5.8 - Relationship between PLD and LHR / PS / PM 48
Figure 8.1 - Mode Choice Model Structure 58
Figure 8.2 - Fitted High Speed Proportion by Journey Time from EU data on major corridors 64
Figure 8.3 - Chart showing the fit of lognormal distributions of income to actual data 65
Figure 8.4 - Spreadsheet Model 67
Figure 9.1 - PS Matrix 'Island' 77
Figure 10.1 - PM Matrix 'Cordon' 85
Figure 11.1 - Heathrow surface access/egress 2007 (CAA) 89
Figure 11.2 - Heathrow Model Catchment Areas 90
Figure 11.3 - UK Business Mode Choice Hierarchy 91
Figure 11.4 - Foreign Business Mode Choice Hierarchy 91
Figure 11.5 - Leisure Mode Choice Hierarchy 92
Figure 11.6 - Heathrow Demand Model Structure 93
Appendices
Appendix A - Transit Line Parameters 97
A.1 Transit Line Validation 97
Appendix B - Highway Parameters 99
B.1 Volume Delay Functions 99
Appendix C - Air Parameters 100
C.1 LHR Model Matrices 100
Appendix D – Model Outputs 101
D.1 Rail Demand Elasticities w.r.t Cost Component Changes, 2021 102
D.2 Air Demand Elasticities w.r.t Cost Component Changes, 2021 103
D.3 Highway Demand Elasticities w.r.t Cost Component Changes, 2021 104
D.4 Rail Demand Elasticities w.r.t Cost Component Changes, 2031 105
D.5 Air Demand Elasticities w.r.t Cost Component Changes, 2031 106
D.6 Highway Demand Elasticities w.r.t Cost Component Changes, 2031 107
1. Introduction
1.1 Background
In April 2009, Atkins was appointed by High Speed Two (HS2) to produce a forecasting framework
to support development, testing and appraisal of options for high speed rail between London and
Birmingham, as well as a wider high speed rail network across the UK. Atkins was supported by
Arup and Sinclair Knight Merz (SKM) in development of the framework and its constituent
components.
The forecasting framework was developed in agreement with HS2, Department for Transport
(DfT) and an external challenge panel of demand forecasting experts over a period of three
months during Summer 2009, with further refinements during Autumn 2009.
2. Overview
2.1 Project Scope and Requirements
2.1.1 Background to High Speed Two Project
In January 2009, the government announced the creation of the company High Speed 2 (HS2) to
develop proposals – and case for – a first new line between London and Birmingham. This new
line could potentially be extended further north, and serve Heathrow Airport either directly or
indirectly.
The work of HS2 built on several previous studies into the case for a north-south high speed rail
line in the UK. In 2001, the Strategic Rail Authority commissioned a study into the case for high
speed rail between London and Scotland. The principal driver of this work was the need to provide
additional capacity on north-south routes. This study reported in 2003 that there was an outline
case for several different routes between London, the North of England and Scotland;
In 2005, DfT commissioned further work to look at the case for high speed rail as part of the
Eddington Study and development of the subsequent 2007 Rail White Paper. The White Paper
noted that there may be a case for high speed rail between London and Birmingham in the future,
once small and medium scale capacity enhancements had exhausted available capacity.
After the completion of High Speed One extension to St Pancras in November 2007, political
interest in high speed rail increased, with the Conservative Party, engineering consultancy Arup
and lobby group Greengauge 21 producing proposals for high speed rail links.
As a result, DfT set out a remit for HS2 to produce proposals for high speed rail between London
and Birmingham, and to set out the case for building the line and potentially a wider a high-speed
rail network. The case would consist of a number of elements:
A business case, demonstrating that the financial and economic benefits produced by the
reduced journey times and reduced congestion across all modes are significant enough to
attract public and private sector investment;
An environmental case, setting out the environmental benefits associated with demand
shifting to new high speed rail and enhanced conventional rail services away from
“carbon-hungry” modes such as road and air; and
A transport case, including how a new high speed line could ameliorate the capacity
problems, otherwise expected on the rail network, and which the scheme is able to
address.
It is important to note that the work goes beyond the outline business case work undertaken in the
2003 High Speed Line Study and includes the analysis needed to build up more detailed
proposals, such as: whether a new line should be completely separate or integrated with the
existing network and services; locations of stations and alignments, particularly relating to
parkway stations and serving Heathrow Airport; re-use of the capacity released on the existing rail
network; operating service patterns and speed of services.
These detailed design questions inevitably involve trade-offs between engineering constraints of
construction and operation, and the economic, financial and environmental benefits of the scheme
as a whole. Option development requires parallel understanding of all of these aspects.
2.1.2 Analytical Challenges
In modelling terms, the challenges faced by HS2 meant that a wide range of different forecasting
and economic questions needed to be addressed, to assist HS2 in making trade-offs between
engineering, environmental and economic costs and benefits.
The remit for the framework included being able to address a wide range of analytical challenges,
such as:
How many passengers would use a high speed rail link between London and Birmingham,
based on different journey times and service patterns? What level of economic benefits
would accrue to passengers for each option, including the amount of crowding relief
provided on existing West Coast Main Line services? How should capacity released on
the existing network be used to maximise benefits from high speed rail?
Should services on the first section of line be integrated with the existing network to
provide through services to Scotland? Would the environmental and economic benefits
outweigh the costs of providing new high speed rolling stock which would only operate at
maximum speed for a small proportion of the total journey?
Do the benefits of removing longer-distance services from the network and allowing more
local and regional services to operate outweigh the costs of providing new dedicated
alignments and terminal capacity in London and Birmingham for new high speed
services?
Can high speed rail be partly financed through premium fares? Is it susceptible to
competition from lower cost and speed services offered by open access operators on the
existing network, and how could it integrate with existing services on the West Coast Main
Line?
What is the most cost-efficient way of serving Heathrow Airport, whether through a high
speed station on the airport campus, a hub station close to the airport but located on and
connecting with the Great Western Main Line, connecting services from either Old Oak
Common or a London terminal station via Crossrail, or other dedicated rail services?
Would there be knock-on benefits to passengers in London and the South East in
accessing the airport?
Would the development of parkway-style stations on the M25 and M42 reduce or increase
the environmental benefits of high speed rail? Would they encourage more people to drive
to high speed rail stations, or less people to drive all the way to their destinations?
Where should stations be sited in London and Birmingham, in order to increase
accessibility for passengers either by road, public transport or both? What would be the
knock-on impacts on local transport networks on accessing the station, such as
congestion on underground services in London?
Does high speed rail provide significant mode shift to rail from air and highway? Could it
support wider carbon emissions reductions objectives? How much does the case for high
speed rail depend on expansion of Heathrow Airport?
How should a wider high speed rail network serving the whole of the UK be developed?
The potential range of analytical questions, and depth of associated analysis, was limited by the
timescales available for reporting. Options needed to be narrowed down over a very short
timescale, with detailed option testing commencing in June 2009. The preferred option needed to
be developed and appraised by the end of October 2009, with final reporting completed by the
end of December 2009.
Despite the short timescales, options had to be developed to a level resilient to scrutiny by a wide
range of rail industry, private sector and government stakeholders with a similarly wide range of
views on which option – if any – should be pursued.
2.1.3 Modelling and Forecasting Challenges
The forecasting approach needs to take into account different behaviours associated with
introduction of high speed services on long-distance demand, mode and destination choice and
trip-making frequency. These are difficult issues which are still not well understood – indeed they
are the subject of ongoing research by DfT. Airport access and international trips display different
behaviours and need to be considered separately, although they interact heavily with other
passengers potentially using the same services and stations.
Careful treatment of the trade-off between high speed rail and conventional rail is also needed.
This included understanding values of time for different demand segments as well as their
valuation of reliability and any “modal constant” associated with passenger perceptions of the
comfort or other non-defined benefits of travelling by high speed rail as a mode.
Station location is an important aspect of the development of a preferred option, and the
attractiveness of different locations by different access modes. Within London, where short taxi or
public transport access are critical in rail competing with highway and air for business trips, the
location of any terminal station and its accessibility by all modes from different areas of London
and the South East was assessed and fed into the wider assessment of the effectiveness of high
speed rail. Again, this issue is magnified when there is the choice between new high speed and
conventional rail services operating from different terminal stations.
There is also the critical issue of the interaction between long and short-distance passengers on
current long-distance services, especially on the Coventry – Birmingham corridor and the Milton
Keynes – London section of the West Coast Main Line. These issues also extend to other areas
not under immediate consideration, including Stoke / Crewe to Manchester and the Manchester –
Leeds lines.
The issue is magnified on the strategic highway network, where local and long-distance
movements (both passenger and freight) share network capacity: reducing the volume of one type
of movement may increase the attractiveness of road as an option for other types of movement.
Analysis of these issues is made difficult by the lack of up-to-date information on highway
movements on the strategic highway networks, and for long-distance trips in general.
2.1.4 Key Requirements for Forecasting Framework
Based on the analytical challenges set out above, the following key functionality requirements
were identified for a forecasting framework to support HS2 option development, testing and
appraisal:
The ability to develop passenger demand forecasts for a variety of high speed rail options
serving different destinations with a range of journey times and service frequencies. The
demand impacts should take into account shift from existing rail services as well as mode
shift from air and highway modes and impacts on trip frequency and generation;
Specific representation of the interaction between local and long-distance passengers on
long-distance services where they serve both markets, particularly on Wolverhampton /
Birmingham to London services which also serve the local Coventry to Birmingham and
Milton Keynes to London markets;
Understanding the impacts of passengers across the existing rail network re-routeing to
take advantage of faster journey times on HS2 services for part of their journey;
Service options for re-using capacity created on the existing network, including enhanced
London / Birmingham suburban and inter-regional services;
Demand and economic impacts of different high speed rail station locations, including
provision of parkway stations in the Birmingham and London areas and impact on local
transport networks through access trips to the high speed rail stations; and
Specific examination of the market for high speed rail access to Heathrow Airport, taking
into account the different behaviour of people making airport access trips.
Above and beyond the framework functionality required above, there were more general
requirements for the framework to meet HS2’s programme delivery, including:
The need to examine multiple options over a short time period requires fast turn-around
times, certainly within an overnight period;
Ensuring the underlying model data is the most accurate and up-to-date available,
maximising the forecasting accuracy of the framework;
Transparency of forecasting approach, where forecasting results can be easily analysed
and an “audit trail” produced to support model outputs. This is especially important to
maintain confidence in forecasting results; and
The ability to provide a wide range of outputs to support option development and
appraisal in areas ranging from environmental assessment to operating cost assessment.
Timescales
Inevitably, the level of any framework development is determined by the time available. For the
HS2 programme, only three months of forecasting framework development was available from
April 2009, to enable option testing to commence in June 2009. Further incremental development
was undertaken between July 2009 and October 2009 to support preferred scheme appraisal.
A summary of the assessment of how each of the broad modelling approaches matched HS2’s
requirements is shown in the table and discussed further below.
In order to meet the technical requirements for a High Speed Rail business case it was imperative
that a network-based model was used.
This is because the new line options to be considered will be non-incremental changes to the
existing rail network. The options will create many new routeing opportunities and changes in
travel times and costs for so many different journeys that the only way to feasibly and robustly
calculate routes and associated travel times and costs is using a network model. Network models
can do this automatically using algorithms based on travel behaviour, whereas spreadsheet-based
models will require pre-coded information or, at best, very simplistic route choice representations.
This is vitally important for examining issues associated with high speed rail: for example, choices
between travelling on a direct, conventional speed service between, say, London and Stafford or
using high speed services from London to Birmingham and connecting regional services to a final
station. A network model was also needed where options are looking at yet-to-be-defined
corridors other than London-West Midlands, e.g. London-Leeds. Similar issues also arise with
station access where a network model enables station choice to be represented explicitly.
A network model is also required to provide times and costs for alternative modes – particularly
highway since a network model aids the calculation of highway congestion relief. Network-based
models also robustly calculate travel times and costs on an origin-destination basis. This is
essential for the economic evaluation of different options.
This over-arching technical requirement has a major bearing on the overall modelling approach
and effectively rules out Option 3:
A spreadsheet-only approach does not meet HS2 modelling and appraisal technical
requirements. Even though, at first sight, a spreadsheet approach does have some
advantages - models can be quicker to develop and run than network-based models and
can, potentially, offer greater transparency – they do not meet the flexibility and detail
requirements of HS2;
PSM offered the best overall forecasting functionality available of any national network
model available. DfT’s Long Distance Model was still under development; DfT’s Network
Modelling Framework only represented travel by rail, and did not have sufficient
representation of the interaction between local and long-distance trips; similarly DfT’s
National Transport Model lacked sufficient detailed representation of the rail network and
had unsuitably long option coding and run times. Other regional models, such as the East
of England Regional Transport Model (EERTM) did not provide sufficient coverage for the
London to West Midlands corridor, let alone the rest of the UK; and
The HS2 programme required a working modelling framework by June 2009, thus
precluding development of a new network-based model to replace PSM – to follow such
an approach would be an extremely high risk exercise with no realistic possibility of
supporting HS2’s wider programme of delivery by December 2009.
2.2.2 Development of PSM for HS2 Forecasting Framework
While the analysis above set out development of PSM as the only realistic alternative for an HS2
forecasting framework, there are also other positive reasons why PSM was used:
PSM is an accepted, modularised, well understood and recently used model that can form
the platform for updating and enhancement. It is a well constructed, documented and
tested model that provides the network supply and demand elements that are critical for
examining HS2 options;
It has a wide geographic scope, meaning that it can flexibly examine other corridors if
required. The networks can be readily updated with new information; and
Though certain elements of the demand model require enhancement the basic building
blocks of the model (such as network cost skimming, generalised cost calculations, logit
model applications, links to evaluation modules, model output modules) all exist and are
understood.
In particular, existing links to evaluation modules allowed use of established robust, audited
interfaces and methods to derive economic and environmental impacts from PSM consistent with
DfT guidance.
However, in order to make the existing PSM meet HS2’s forecasting requirements more closely,
further development was required:
Updating and enhancement of the functionality of the existing PSM, to provide a new
PLANET Long Distance (PLD) model. This includes:
New rail, air and highway demand matrices representing 2007/08, reviewing the
corresponding model networks for each mode;
Updating the demand model to reflect expert views as to traveller behaviour for
long-distance trips; and
Providing new functionality on aspects of station choice, critical for assessing the
implications of alternative city centre and parkway station sites;
Linking PLD to the recently updated PLANET South and PLANET Midlands models to
enable impacts of new lines on classic rail services to by fully represented. PLANET
South and Midlands provide a much better representation of local rail movements, in the
London & South East and West Midlands respectively. Interaction of long-distance and
local demand on inter-regional services is critical to understanding capacity issues on the
rail network; and
Supplementing the PLANET models with an integrated spreadsheet model to address
specific issues related to passenger access to Heathrow Airport.
Section 3 describes the integrated framework in more detail, with subsequent sections describing
the development of the individual model components.
Similarly, demand associated with access to Heathrow Airport is also imported into PLD from a
separate Airport Access spreadsheet model, although this demand is input on a matrix origin-
destination basis and allowed to re-route to take advantage of quicker routes or reduced crowding.
While the existing PSM model included a high level station choice model for London and
Birmingham (and three other cities), the station choice model did not work at a sufficient level of
detail to capture accessibility of different station sites within the Greater London and West
Midlands areas. As a result, the station choice procedures were updated to take into account
forecast station access times from Railplan (for Greater London) and PRISM (for the Birmingham
urban area) models. It is important to note that the inputs on station accessibility are static, i.e.
they do not include any feedback or “knock-on” effects of increased London Underground or
highway congestion resulting from accessing alternative station locations.
3.1.2 PLANET South (PS)
The existing PLANET South (PS) model has been in use for several years for modelling forecast
crowding on the London & South East rail network and associated impacts on London
Underground lines. The model represents morning peak period (0700-1000) rail movements within
an area which covers the former Network South East area, with less detailed representation
beyond those areas.
PS was adapted for this framework to include only trips within London, South East and South
West areas, to eliminate the previous overlap with PSM. The model includes both local and long
distance services into London. However, long-distance demand – i.e. rail demand between inside
and outside London & South East / South West areas – is included in the model by means of
“wormholes”.
These “wormholes” are effectively dummy zones that feed origin-destination demand information
from PLD into PS into nodes at the edge of the PS network area, having been scaled from all-day
to peak-only levels. This demand is then assigned onto long-distance services, with the PS
assignment routines allowing the demand to reach their final destinations within the PS area using
any other rail or underground services. The model works in a similar way for trips from the London
& South East area to the rest of the country.
It is important to note that while long-distance movements are included in PS demand and are
assigned in a way that allows them to find quickest and least crowded routes to their final
destination, they are not subject to any demand response in PS. By contrast, local trips within the
London & South East area are responsive to any changes in crowding as a result of changes to
long-distance services and demand – expected to be a significant effect.
3.1.3 PLANET Midlands (PM)
The PLANET Midlands (PM) model was developed during early 2009 for DfT. The model is
generally similar to PLANET South but covers the West Midlands and East Midlands rail networks
and demand. It also has slightly different functionality from PLANET South in the way that it deals
with station choice: while PLANET South uses default EMME/2 assignment processes to assign
demand from zones to stations, PLANET Midlands uses detailed accessibility information for each
station from each zone.
PLANET Midlands was adapted for the framework by cutting back its area of scope to the
immediate greater Birmingham area. Both local and long-distance services are included in the
model. However, in contrast to PLANET South, long-distance demand is fed from PLD by a pre-
load approach, identifying the level of long-distance passengers on long-distance service groups,
and adjusted to reflect the morning peak demand levels.
This altered approach reflects the much higher levels of car access for longer-distance trips
to/from Birmingham, and the relatively lower importance of knock-on dispersal effects of crowding
on local rail services feeding long-distance rail passengers from the outskirts of Birmingham into
central Birmingham to catch long-distance services.
1.1 PLD
Surface Long Long
access distance distance
costs matrix pre‐loads
2.1 PLD
Surface Long Long
access distance distance
costs matrix pre‐loads
3.1 PLD
3.1 PLD
Surface Long Long
access distance distance
costs matrix pre‐loads
4.1 PLD
Response to congestion on both highway and rail modes; highway congestion fully
interacts with other modes, in that mode shift away from highways results in marginally
shorter journey times by highway, which can feed into shorter car access times to stations
for car available trips; and
Response to changes in fare levels, although road tolls are not included.
The model provides a range of outputs to inform scheme development and decision-making,
including:
Statistics on passenger flows by air, rail and car – such as number of people, travel time
and distance travelled;
Passenger flows on the strategic routes;
Levels of road traffic congestion and train passenger crowding anticipated;
Fare revenue by operator group; and
Economic benefits 1
1
As a result of the use of PSM, it is possible to assess the economic worth of the proposed intervention, by
comparison of travel times and costs between the two scenarios, and allowing for the operating and capital
costs involved; this uses conventional Cost Benefit Analysis and Discounted Cash Flow techniques.
5082342/5082342 Model Development Report (26-02-2010).doc 17
Model Development Report
Base Run Test Run
Highway
•Business, Leisure, Commuting Highway
•Business, Leisure, Commuting
Assignment Assignment
Rail
•Business, Leisure, Commuting Rail •Business, Leisure, Commuting
Assignment Assignment
Air
•Business, Leisure Air •Business, Leisure
Assignment Assignment
•Base Skims •Test Skims
Skims Skims
Mode
•Recalculate Mode Shares
Choice
2
For instance, the majority of the London, Birmingham, Manchester or Glasgow rail commuter networks are
excluded from the model, but those services sharing tracks with the various strategic routes would be
included.
5082342/5082342 Model Development Report (26-02-2010).doc 18
Model Development Report
Some of these features had been found to not be ideal, and were the subject of revision in the
development of the replacement PLD model: see Section 4.4.2 below.
PSM rail demand was originally constructed from the DfT's National Rail Passenger Matrices, built
from CAPRI data (from a period prior to the Hatfield disruption in October 2000, to ensure the
disruption was not included in the matrices). The demand was segmented into business, leisure
and commuting purposes, with a further split car availability.
For the update to PLD, the rail industry's LENNON data for 2007/08 was used, (LENNON
replaced CAPRI in 2003).
The LENNON data was collected at station to station level, and distributed to ultimate origins and
destination zone using National Rail Travel Survey data. The production of the current PLD rail
matrices is discussed in detail in Section 5.2.
4.3.4 Highway Model
The highway model represents the UK strategic road network, with notional access links from
model zones to the highway network. Generally, the model includes motorways and the primary
route network, with infill in certain areas where the primary route network is sparse. Key features
of the highway model include:
Strategic route choice for car trips across mainland Britain;
All day representation of demand, converted to hourly demand to be compatible with
hourly speed / flow relationships, based upon COBA (The Cost Benefit Analysis program
developed by DfT/HA for road schemes) data;
Strategic demand only, supplemented by local pre-loads;
No representation of junction delays;
Three trip purposes (for strategic trips only), plus a single pre-load for local trips
The highway network is also used for air / rail access trips; and
Demand data from multimodal studies (but updated), as explained below:
PSM highway demand was originally built by converting demand data from the then currently
available multimodal models into the PSM zone system, and combining the data in such a way as
to remove multiple observations of the same trips. Short distance trips were removed to avoid the
strategic-only highway network from being swamped by local traffic. Data was segmented into
business, leisure and commuting segments, either by combining more numerous segments (in all
but one case) or by estimating a purpose split by trip length.
Vehicle occupancy by purpose was accounted for when converting from vehicles to person trips.
Pre-loads were attached to links to adequately reflect local traffic, to allow the strategic trips to
adequately react to background loading levels. These pre-loads are calculated to be the difference
between count data and the assigned flows from the demand matrices. The pre-loads solution is a
good substitute for a fully national highway model.
It is important to note that local road congestion is not included in the model (as the nodes and
links and detailed zone structure do not exist), though this is not an issue as the model is
interested in change in journey times on the strategic leg of the highway journeys. Of the included
strategic roads, junction delays are not modelled due to the relatively low proportion of delay
attributable to junctions for long distance trips.
Since the mode choice model is incremental, the main function of the highway network is to
provide robust strategic journey times. In many ways the absence of local networks, with local
trips and the associated junction delays, enhances the stability of the model.
4.3.5 Air model
The air model covers most domestic air services in mainland Britain. Key features of the air model
are:
Highway Demand
For the update to PLD, new demand data was sourced from PRISM (a detailed multi-modal model
for the West Midlands) and from NoTAM (the North Thames Highway Assignment Model
developed for the Highways Agency) and incorporated. The demand was then rebased to 2008 by
balancing the matrices to new totals derived using factors from TEMPRO (the national trip-end
model and presentation tool provided by DfT). The pre-loads were updated to represent 2008, and
replaced with TRADS (the Traffic and Accidents Database maintained by the Highways Agency)
data for the London to Birmingham corridor. The production of the current PLD highway matrices
is discussed in detail in Section 6.3.
Mode Choice and Premium Fares
One of the important policy questions for development of high speed rail options was the relative
benefits of introducing premium fares for high speed rail services, so that passengers benefitting
from faster journeys contribute more towards the construction of the line. While the central case
for HS2 assumes no premium fare, some tests are required to understand the impact of premium
fares on demand, revenue and economic benefits.
Premium fares for the high speed services are implemented at the matrix level as inputs to the
logit mode choice process. This is in preference to modelling fares (and especially premium fares)
by assignment, which can run the risk of very 'lumpy' assignments to the high speed (such as 0%
or 100%) depending upon the level of fare and difference in journey times. The logit model trades
off fares and journey times in much more detail than can be achieved through assignment
processes.
Since the mode choice model is incremental, it pivots around the base mode shares. Shares for
the high speed mode, as a new mode, have to be forecast by an absolute process. The following
options were investigated:
A new mode with mode constants;
A new mode using the Distributed Value of Time (DVoT) approach (which uses different
values of time for proportions of the population, and hence has a better representation of
people’s responses) requiring no mode constants; or
Modelling high speed as the same mode as conventional rail with an assignment bias.
Each of these approaches has strengths and weaknesses, as summarised in Table 4.1 below.
Table 4.1 - Treatment of Premium Fares
It was concluded that “DVoT” was preferred for premium fares, while “No New Mode” was
preferred where no premium fare was to be assumed.
However, to retain a level of flexibility, different versions of the model were developed to cover
each of the three approaches and to enable them to be tested and applied.
Section 8 explains these issues in more detail.
Miscellaneous Updates
Finally, the opportunity was taken to improve the overall operation of model by:
Volume-averaging in the assignments;
Updates to Values of Time; and
More general improvements to reduce model run-times.
Also, the development of output macros was undertaken, to automate the production of:
A standard output Excel workbook;
Economics outputs and an analytical spreadsheet; and
Mode choice diagnostics spreadsheets.
all the purposes, the leisure traveller is likely to be most able to schedule the trip to avoid the most
congested times, or to take advantage of cheaper advance purchase tickets.
PLD is concerned with strategic movements greater than 100km between major centres, The
zone system is designed such that strategic trips have a reasonable geographical disaggregation,
however, many local trips are intrazonal as a result.
PLD is geographically disaggregated into 235 zones, as shown in Figure 5.1 below. These zones
are equivalent to districts or aggregations of districts. For example, the 32 boroughs of Greater
London are aggregated into 7 geographical sector zones (plus Heathrow as an explicit zone).
Rural Cumbria on the other hand retains its constituent districts. This is done to group zones into
patterns of similar access and egress (such as Camden and Islington London Boroughs in north
London), while acknowledging that east and west Cumbria may have very different access and
egress, despite the far smaller population and trip activity in each district.
Figure 5.2 - Postal Sectors in the West Midlands [NRTS leisure trip-rates]
PLD zones - which are local authority districts or groups of districts - are considerably larger
than postal sectors. GIS methods were used to assign each postal sector to its appropriate
PLD zone. For example, Solihull (PLD zone 181) comprises the following 23 postal sectors
(Table 5.1):
Table 5.1 - Postal Sectors in Solihull
Within the NRTS data, the top 15 districts producing originating rail trips for each station were
identified. For all but the largest stations, this threshold accounted for all originating journeys (see
Figure 5.3 below). A lower threshold of 10 was applied to outward egress to the ultimate
destination, as distances tend to be shorter with use of the household car precluded.
During the production of the matrices, an issue with NRTS data was brought to HS2’s attention. A
small minority of observations to/from London have ultimate origins and destinations transposed,
such that an out-and-back rail trip from Manchester Piccadilly to Euston might be shown as
produced in Westminster and attracted to Salford. To remove such cases from the analysis of
access and egress zonal distribution, a distance cut-off of 80 miles was imposed. This value was
chosen as it prevents transposition of districts/zones in the key West Midlands to London market.
With the latter observations removed, Figure 5.3 shows the distribution across stations in the
number of districts reported within NRTS as ultimate origins and destinations. Almost all stations
appear at least once in the NRTS in each of the 4 market segments, allowing all demand to be
channelled to/from the associated district. (Absent stations will tend be the least important in terms
of passenger volumes.)
Over 1,500 stations are accessed from least 2 districts by car-owning passengers, compared to a
figure of around 750 for non-car available access. At the access threshold, 109 (mostly major)
stations are accessed by car owners from at least 15 districts, falling to 31 stations for non-car
owning households.
At Euston, the top 15 districts account for 80% and 86% of car available and non-car available
access respectively 3 . At Birmingham New Street the corresponding figures are 97% and 96%,
whilst at Manchester Piccadilly there is no residual access, with all demand caught by the top 15
districts. For Euston egress, the top 10 zones account for 92% and 85% of trip attraction amongst
car available and non-car available rail travellers respectively. At New Street and Piccadilly, the
largest residual is just 2.4% - found amongst non-car owners egressing from the Manchester
station.
3
The 15th ranked district is Ealing, with a car available market share of just 1.2%.
5082342/5082342 Model Development Report (26-02-2010).doc 28
Model Development Report
1500
1000
500
0
At least 1
At least 2
At least 3
At least 4
At least 5
At least 6
At least 7
At least 8
At least 9
At least 10
At least 11
At least 12
At least 13
At least 14
At least 15
Figure 5.3 - Concentration of districts providing journeys to/from particular stations
Using SPSS software, the car available and non-car available access and egress demand shares,
and their associated districts, were merged onto the origin (access) and destination (egress)
stations in the 2007/8 LENNON dataset. This allowed distribution of station to station demand
between districts of ultimate origin and destination.
The LENNON data were divided between car available and non-car available journeys on the
basis of NRTS data aggregated for each origin (i.e. rail trip producing) station. Further
disaggregation of car availability (e.g. by journey purpose, ticket type, or destination) was rejected
in order to maximise sample sizes.
Figure 5.4 below shows the distribution of car availability rates for outward journeys across NRTS
stations. The large spike above 1.0 reflects the fact that at smaller stations with limited NRTS
samples, all respondents report ownership of a household car.
300
250
200
150
100
50
0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Station-to-station journeys within a given market segment (i.e. ticket category plus car availability)
were then divided between the 150 (15 access * 10 egress) combinations of trip-producing and
trip-attracting districts. Finally, conversion of districts to the 235 PLD zones allowed SPSS to be
used to aggregate journeys across zone-to-zone pairings 4 .
An early/pre-validation sense-check of the PLD LENNON data was provided by aggregating
districts to regions, and comparing against the 2006/7 regional demand flows shown in ORR’s
National Rail Trends (2007/8 edition, Section 7). This revealed a significant shortfall in journeys
from the North-West to London. Further investigation revealed that processing of the PM
LENNON data had removed southbound trips having ‘Manchester stns’ (joint code) recorded as
origin together with ‘London Terminals’ (joint code) as destination. Although this is a trivial issue
for a model focussed on commuting into Birmingham (PM), it required an additional stage of
processing to correct the PLD matrices.
Deannualisation of LENNON for modelling of typical weekday demand
PLD estimates origin to destination travel on a typical weekday, summed across outward and
return journey legs. PM and PS further restrict their modelled travel period to the AM weekday
peak, dominated by the outward leg of return journeys (from the point of trip production to the
point of attraction).
With modelling undertaken for travel choices/patterns within a single weekday or AM peak, the
production of PLANET demand matrices requires the 2007/8 LENNON journeys database to be
‘deannualised’. That is, a methodology is needed to remove travel at weekends, and, in the case
of PM and PS, to estimate the proportion of weekday trips occurring in the AM peak. PS
deannualisation is undertaken within an SQL server with the factors dependent on flow distance
and based on ORCATS (Operating Revenue Computer AllocaTion System, a data source
maintained for ATOC as part of the work of Rail Settlement Plan) assumptions for Season tickets,
or a bespoke (weeklong) LENNON download undertaken as part of SDG’s update work in 2004/5.
As ORCATS factors are aging (and thought for example to underestimate travel at weekends),
and as SDG’s deannualisation factors for the PS area may be ill-suited to PM and PLD, it was
decided to deannualise LENNON using up-to-date assumptions for the incidence of weekend
travel by ticket category.
For Full-fare returns, deannualisation requires a distance dimension. On the shortest flows, where
first class is typically unavailable, no full fare returns will be purchased at the weekend. On longer
distance flows, first class ticketing may be purchased at weekends to take advantage of the
associated exclusiveness and greater comfort. Moreover, in the case of travel to London, full fares
may be bought to avoid evening restrictions on the use of Savers and CDRs (now both ‘off-peak’)
affecting the return journey leg. Finally, full-fare returns are valid for a month on longer O-D flows,
but only on the day of issue where Savers are unavailable. This means that some long distance
returns will be bought at the weekend to allow the return leg to be undertaken during the Monday
peak (when Savers are invalid).
4
Analysis proceeded as far as possible using districts rather than PLD zones because of the possibility that
PLD zoning might be reviewed for some reason.
5082342/5082342 Model Development Report (26-02-2010).doc 30
Model Development Report
The details of the deannualisation used to estimate weekday one-way travel from producer zone
to attractor zone are shown below in Table 5.2.
Table 5.2 - Treatment of Ticket Types in LENNON
The final output of the HS2 LENNON-NRTS analysis was 6 PLD input matrices representing
outward travel on a 2007/8 weekday from the zone of trip production to the trip attracting zone. As
PLD is an all-day matrix, the P-A matrix was transposed in EMME to produce an A-P matrix for
return travel 5 .
5.2.3 Strengths and weaknesses of Rail Demand Data
Strengths
The purpose splits allow the analysis of a reasonably distinct set of travel behaviours
(business, leisure and commuting). This is of paramount importance in order to
understand the behaviour of different types of travellers. The three purposes provide a
reasonable segmentation at the all-day level; and
The car availability allows a reasonable, if synthetic, split between 'car available from' and
'car available to'. This prevents a passenger, with a car available, from using that car for
both access to, and egress from, a rail journey.
Weaknesses
The zoning system can still allow large numbers of local trips, such as Bolton to
Manchester. This could potentially swamp the strategic trips in the matrices with local
trips, though this is mitigated by the inclusion of local services, which has to be done on
an ad-hoc basis. In PLD, this is mitigated by removal of the major commuting demand
associated with the West Midlands and South East, where use is made of the PS and PM
models instead.
5.2.4 Modification for use in PLD
To enable their use within PLD in conjunction with PS and PM, the PLD matrices have 'holes' cut
in them to avoid overlap with PLD, in the PLANET Midlands and South areas, as in Figure 5.5
below:
5
This approach suggests that journeys on full-fare Single tickets should also have been divided by two
during deannualisation. However, as PLD is a long distance model where use of Full-Fare singles is
relatively rare, this inconsistency is not thought to introduce significant bias to the matrices, and validation
tends to confirm this view.
5082342/5082342 Model Development Report (26-02-2010).doc 32
Model Development Report
On balance, the change to handling fares at the matrix level is the more appropriate thing to do,
as the purpose of the model is to forecast the trade-off between rail and competing modes, rather
than within the rail mode.
As a result of this, fares are now removed from the assignment. Fares are used within the mode-
choice model, in section 8 below.
5.3.2 Rail Fare Matrix Production
Rail fares are based on EDGE 6 outputs for revenue and journeys, using a simple average yield
calculation (revenue / journeys).
It might be noted that for the business segment in particular, an increase in distance may not be
associated with an increase in yield. For example, use of Full fares from Glasgow to London is
extremely limited, due to the fact that the Standard Class Saver product was, until recently,
unrestricted for Anglo-Scottish travel.
A specific exception to this approach of fare production is that of Heathrow. Production of these
fares is discussed in section 11 below.
6
EDGE (Exogenous Demand Growth Estimator) is envisaged as the replacement for RIFF (Rail Industry
Forecasting Framework). EDGE provides PDFH compliant forecasts of rail journeys and revenue.
5082342/5082342 Model Development Report (26-02-2010).doc 34
Model Development Report
Market Harborough.
Rail Links
New links were added to PLD to allow new services to be added to reflect current operations.
These were:
Wigan to Salford Crescent;
Stratford to Tottenham Hale;
Kettering to Oakham;
5.4.2 Rail Service Updates
The national rail services were replaced with the December 2007-May 2008 timetable, with
Wednesday 13th February 2008 taken as the nominal date. A Network Rail CIF 7 file was provided
by the DfT to give service definitions, such as journey time and stopping pattern. An
accompanying file was provided with stock type, together with the seated and standing capacity,
for each train operated.
Methodology
A Perl 8 software script was developed to extract data from the CIF file and convert into EMME
format in the relevant node numbering system. The approach was as follows:
Extract subset of ‘relevant’ trains from CIF;
For ‘relevant’ trains extract the subset of ‘relevant’ nodes required;
Look up ‘relevant’ nodes dependent on direction and TOC;
Calculate the journey time between ‘relevant’ nodes;
Aggregation of identical lines;
Allocate a PLANET service code to each line;
Allocate an appropriate Vehicle Type;
Export in EMME format; and
Import to EMME and interpolate stop to stop times.
This allowed a consistent and automated approach.
Vehicle Type and Capacities
In general, rail vehicles are given vehicle types of 888, which is a generic vehicle type. The seated
and total capacities are held within the ut1 and ut2 fields in each transit line or service, and hence
the train capacity is service specific. These values are used in crowding calculations.
Train capacity is fundamentally important when working with crowded assignments. This is to
ensure that load factors, as the main building block of a crowding function, are calculated
correctly. In principle, seated and total data is available for every rail service (as extracted from
the DfT's Network Modelling Framework), though this does not mean every train service is
included. This is for several reasons, mainly that as a frequency-based all-day model, PLD does
not model every train throughout the day individually, as there is no concept of 'when' during the
day the trains run. This is also true of the demand, which represents the entire day, thus ensuring
that the network and demand data are consistent. This makes it impossible to differentiate
between peak and off-peak trains which may differ in terms of train length (such as 4, 8 or 12
coach local services) and thus capacity.
7
Common Interface File.
8
Practical Extraction and Reporting Language. See http://www.perl.org
5082342/5082342 Model Development Report (26-02-2010).doc 36
Model Development Report
It should be noted that there is a danger of causing assignment instability by modelling mixed
length trains, as each will be assigned to in proportion to frequency, and not in proportion to the
number of available seats. As such, the most robust way to code services in order to model the
all-day level is to aggregate services according to their stopping patterns, and attach the average
(service specific) capacity to each set of aggregated trains. This situation is most likely to occur for
local services, as the longest distance services from London tend to be fixed formation of 8+
coaches, and not typically divisible.
Since this is less problematic for longer distance services, those services for TOCs in the London
to West Midlands corridor, or other long distance routes, are given more explicit treatment of
vehicle types in PLD. This allows easier identification of vehicle types for TOCs in the model
scope area. These TOCs are as follows:
East Coast (Including Hull Trains, Grand Central)
West Coast
Midland Main Line
Great Western
Cross Country
Chiltern Trains
Wrexham and Shropshire
These TOCs use specific vehicle types.
Network Capacity Bottlenecks
The PLD rail model has large zones with the demand created from a full representation of
passenger movements, except for those trips completely within the South East of England or the
West Midlands. As a result, it has a large number of intra-zonal trips and shorter trips, often
representing travel between stations not explicitly modelled. In addition, it does not include a
representation of all local services.
As a result, there are localised situations where the model exhibits extreme overloading, which
over estimates the crowding penalties imposed on the longer distance passenger.
This situation can potentially escalate such that capacity is overwhelmed by demand, causing
unrealistically high loadings, thus crowding penalties. This can cause major re-assignments
between iterations, making convergence more difficult to achieve. This can then feed into the
model skim matrices as an unrealistically high origin-destination skim, and can have severe
impacts on the demand model, causing significant (and undesired) shifts away from the rail mode.
One possible solution would be to have more separate models in those problematic local areas,
such as Manchester or Leeds. However, neither area is currently available in PLANET, so a
network amendment method was undertaken instead.
As part of the model checking process, localised area of high crowding were identified by plotting
the ratio of demand (passengers carried) to supply (seated capacity). These were inspected and
network amendments made, usually by inserting additional local services where these had been
excluded during the network build process.
As capacity bottlenecks are more marked in the future year networks, where higher demand levels
are forecasted, this checking process was mainly undertaken at that stage in the model
development and application programme. However, such capacity enhancements were included
in all future year tests (Base and Test) to ensure consistency.
Table 5.5 below summarises the amendments made to address such bottlenecks.
Vicinity Amendments
WCML: Preston area Add local services on Wigan - Manchester corridor to reflect
current service pattern;
Add connector to serve Liverpool;
Extend modelled services from Wigan to serve Bolton
Swansea - Fishguard Add dummy service between Swansea and Fishguard to reflect
those serving Milford Haven and Carmarthen
Leeds - Halifax Add dummy service to represent local services in the corridor
Nottingham Amend Lincoln – Leicester service
Glasgow - Motherwell Add local service to represent local services between Glasgow
and Motherwell
On balance, for testing strategic schemes several years into the future, a frequency-based
assignment is more appropriate, and has been retained for PLD.
5.5.2 Generalised Time
PLANET models, in common with many assignment based models, use the concept of
generalised time as a measure of the disutility of travel. The model seeks routes that minimise this
measure for each passenger then assigns the demand to those routes.
Generalised time comprises the weighted sum of the elements of the journey, as far as these can
be determined from the available data. These are then weighted to give an aggregate perceived
value in equivalent minutes. The components of generalised time can include:
Access and egress time;
In-vehicle time (derived from the time table data)
Reliability and Ambience;
Crowding penalty;
Waiting time;
Boarding penalty;
The weights used in PLD, which are the same as those developed for PSM, are included in
Table 5.7 below, and further discussed, where appropriate, in the following sections.
Parameter Value
Board Penalty (minutes) 30
Wait Time Factor 0.4 9
Wait Time Weight 2.0
Access/Egress Time Weight 4.0
Board Time Weight 1.0
9
0.4 is used to represent increased timetable knowledge for strategic rail trips.
5082342/5082342 Model Development Report (26-02-2010).doc 40
Model Development Report
In PLD, the crowding penalty is applied as a weight on the in-vehicle time for each segment of the
train journey:
Perceived Time = Actual Time * (Crowding Penalty)
The level of crowding is converted to a penalty, by reference to the relationships given in the
Passenger Demand Forecasting Handbook (PDFH). This quantifies the time penalty factor
perceived by passengers for enduring the conditions for various levels of crowding, for each trip
purpose 10 , for both seated and standing passengers. These parameters come from various
sources since 1987, and are referenced in full in PDFH (chapter B5).
These factors are multiplied by actual travel time to yield a perceived travel time.
A process is in place to use observed means and standard deviations of train loadings to adjust
from the individual train PDFH crowding response to reach a time period adjusted crowding
response.
( x x )2 ( x x ) 2 ( x x ) 2
1 r ,p 1 1 1
C r ,p ( x )
2 r
2
e 2 2 r dx
2 r
2 r ,p
( Ar , p Br , p x )e 2 2 r dx
2 r
2 (C
1
r ,p Dr , p x )e 2 2 r dx
Where:
x is the average level of vehicle occupancy;
r is the observed standard deviation of vehicle occupancy to TOC group r;
r,p is the lower threshold of crowding for TOC group r and journey purpose p;
Ar,p and Br,p are the lower crowding function parameters for TOC group r and journey purpose p;
Cr,p and Dr,p are the higher crowding function parameters for TOC group r and journey purpose p.
The effect of this adjustment is shown in Figure 5.7 below. In this Figure, “Capacity utilisation” is
the ratio of passenger load to seated capacity.
2.80
2.60
2.40
Crowding factor on IVT
2.20
2.00
Actual
1.80
Adjusted
1.60
1.40
1.20
1.00
0.80
0.50
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
1.00
1.05
1.10
1.15
1.20
Capacity utilisation
10
Trip purpose is no longer distinguished in PDFH 5
5082342/5082342 Model Development Report (26-02-2010).doc 41
Model Development Report
The adjusted crowding factor (upper line) represents the extent to which passenger demand for
rail services is unequal throughout the day. For example, a train service with demand of 300
passengers and supply of 500 seats would have a loading factor of 60% across the entire period.
This would imply no crowding penalty based upon the table above, However, since the demand is
not likely to be equally distributed across the modelled time period (whether an AM peak or all
day), but peaked in profile: such peaks needs to be taken account of in the crowding function. For
this reason, the above adjustment is undertaken. This implies that the 300 demand and 500
supply will yield a crowding penalty of around 1.1 in the above example.
The adjustment uses different normalised standard deviations for different time periods and
service groups. The current model uses the following groups:
Inter-Urban London services;
Sub-Urban London services;
Non London services: and
All Day (07:00 to 22:59)
The PLD model retains the original method of using seated capacity only, when calculating
crowding ratios.
5.5.6 Assignment Process
The PLD rail assignment has 3 trip purpose segments (business, leisure and commuting) and 3
car availability segments (non car-available, car-available from, and car available to). This gives 9
assignment classes in a base run, and 18 classes in a high speed test run (as the 9 segments are
further subdivided into classic and high speed). Equilibration of supply and demand is by the
Method of Successive Averages (MSA), undertaken for 10 iterations. Most of the procedure is
carried over from the 2002 Atkins HSL model and the 2004 audited PSM model. The
enhancements are explained in this section.
This iterative process is used to achieve convergence in the crowded assignment,
n1
This ensures that each iteration is combined with all previous iterations in a way which
encourages convergence at the link flow level. However, it is important to stress that convergence
at the link flow level does not guarantee convergence at the demand model level.
Previous versions of PSM relied upon the final iteration for the resulting model flows (although
based upon a crowding penalty which was constructed using an MSA method). While this was
mostly adequate, it sometimes caused large changes in flows between two similar scenarios. This
was found to be due to the variability of flows resulting from different iterations, no matter how well
converged the assignment model. This has been revised for PLD, by replacing the final iteration
flow with a suitably converged value, using MSA. This is essential to ensure that model stability at
the link level is retained.
There are two variants of the rail assignment model: one for conventional rail only, and another for
future year assignments where high speed is treated as a separate mode.
unrealistic domination of a single station), However, this was not sufficient for detailed station
choice, such as individual locations within central and west London, as only the main cities of
each conurbation were included, and detailed zoning and network data does not exist in PSM. To
remedy this for PLD, attention was focused upon Greater London (7 zones and 9 stations) and
West Midlands (5 zones and 6 stations). The accessibility of each zone to each possible station
was calculated within detailed London and Birmingham models (Railplan and PRISM) to allow the
generalised costs of accessing each station to be traded off in a logit model, producing demand
weighted shares for the zones in PLD.
As the focus is on London to Birmingham, these locations were expanded to provide more detail.
The existing PSM station choice routine was developed for one zone in each of central London,
Birmingham, Manchester, Bradford and Bristol, and ignored the other boroughs or districts within
those conurbations. The station choice model was originally designed to avoid the problem of “all
or nothing” assignments to stations within cities which had two or more viable alternative routes,
and to give a smoother assignment between each station. For example, Manchester to Liverpool
journeys can be made from either Manchester Victoria or Manchester Piccadilly and Bradford to
Leeds journeys can be made from either Bradford Forster Square or Bradford Interchange
stations.
It should be noted that the existing PSM includes the option, for car access legs, of using the
strategic road network to any rail station. Therefore options such as direct trains from Stafford to
London and driving to a Birmingham Parkway station to catch a high speed rail service to London
are already available in the model. This additional functionality is only required for areas closer to
potential HSR station sites where the large zone sizes and the strategic level of detail of the
highway network mean that the model cannot predict station choice with any reasonable degree of
accuracy.
The station choice model is updated for PLD in the following ways:
Removing Bradford, Bristol and Manchester;
Concentrating on London and Birmingham;
Using Railplan and PRISM journey time data to inform the station choice.
These changes are explained in this section.
The station choice process now handles the stations listed in Table 5.8 below.
Table 5.8 - Stations in PLD Station Choice Model
The fundamental approach is to use the more detailed Railplan and PRISM zoning systems as
subdivisions of the much larger PLD zones. Using databanks created outside the PLD zone
system, we use artificial Railplan and PRISM databanks to undertake “park and ride” type
calculations of the accessibility of each of the conventional and high speed rail stations. This
allows the demand from the zones representing Greater London (zones 117 to 123 inclusive) and
greater Birmingham (zones 5, 47, 176, 181 and 217) to be disaggregated into separate “station
specific” dummy zones. Average access times are also imported from the Railplan / PRISM
models into the zone connectors to allow the economics and mode choice aspects of the model to
operate correctly.
Railplan is the TfL model of public transport assignment in Greater London, and uses 1571 zones
with a single journey purpose. A set of public transport generalised journey time data is used to
inform station choice. This data was chosen as a good representation of public transport, despite
the fact station access is often by car / taxi in London. However, this would have added
unnecessary complication to the process, while public transport access is deemed more important
in London, particularly in the central area.
PRISM is the West Midlands multi-modal model, using 928 zones. It is segmented into several
purposes, but a single set of highway generalised journey time data to each station is used to
inform station choice. These highway data were chosen for use, despite the fact that some rail
access is via public transport (particularly in Birmingham itself), as there is more possibility of a
parkway station in the greater Birmingham area than in greater London.
This process is implemented as part of the existing “separate stations” and “combine stations”
processes. Demand from the geographic London and Birmingham zones is distributed to the
dummy station specific zones in “separate stations” for assignment and skimming purposes. Once
the assignment and skimming is complete, demand-weighted average skims and demand are put
back into the original geographic zones before mode choice calculations are undertaken (with
demand and skim matrices reset to zero for the station choice zones). Hence, the dummy station
specific zones are only altered during the assignment process and are reset to zero after
completion of that macro.
Because of the increase in number of dummy zones, the station choice mechanism is removed
from Manchester, Bradford and Bristol zones. If necessary, the functionality could be brought back
in at a later stage, but effort is focussed on the London to Birmingham corridor.
It should be noted that this mechanism is not designed for assessment of potential demand at an
intermediate station (somewhere between Oxford and Northampton). Assessment of demand at
these sites will be determined by PSM’s normal highway accessibility mechanisms, although we
believe that off-line calculations are likely to be required to give a more detailed understanding of
the relative accessibility of any new high speed rail station for commuting to London.
5.6.2 Operation of Station Choice Model
Within a 10-iteration crowded assignment loop, demand is separated out from the geographical
zones after the first uncrowded assignment iteration. At the end of the assignment loop, the skims
for the station zones are recombined as a weighted average back to the geographic zones. This
ensures the demand model operates on the geographical zones. The matrices are stored as
geographical zones, with temporary copies for station-choice demand.
The London and West Midlands sections run sequentially, such that London end Station Choice is
calculated first. The West Midlands station choice is calculated second, taking into account the
London end choices. This prevents 'unlikely' station choices, such as Euston to Snow Hill.
There are two basic aspects to the station choice operation: Separating demand before
assignment, and combining weighted skims post assignment. See below for the implementation of
Separate Stations and Combine Stations.
Separate Stations
This is a 3-step process to re-allocate demand to the expanded zone system:
Step 1: Export PLD data
Generalised journey times are exported from PLD for the subset of journeys from London to the
Non-London zones elsewhere in the model.
e gc KX
Share KX
e gc KX e gc Eus e gc Pad e gc StP ...
The spread parameter used (lambda) is 0.10.
In addition, average access times are calculated from Railplan data. Once the shares are
calculated, the journey direction is transposed and the calculations are repeated for the opposite
direction of travel.
Finally, the share and average access times are exported from Railplan.
Step 3: Import and Apply Station Shares
The shares and access times are imported into PLD. From these shares, the actual demand to
and from London stations is multiplied by the shares, and new matrices are created.
To ensure consistency, the main city zones are reset to zero, and the intrazonal trips are replaced
in the main city zones.
This in effect creates an 'exploded' matrix of 328*328 zones, with a total matching to the 235*235
original,
Combine Stations
At the end of a 10 iteration assignment, the following steps are required:
Add station access times for London / West Midlands; and
Combine demand-weighted skims for city stations.
This gives demand weighted skims to feed the mode-choice model, which then operates on the
geographic zones. At all times, copies are kept of the pre- and post- station-choice process.
As the station choice model processes London first then Birmingham, the London end of all trips is
given a free choice of station, then the West Midlands end station is calculated. This time, the
London choice of each station is already made, so the choice of West Midlands station depends
upon the London end. For example, if a proportion of trips chooses Euston at the London end,
then those same trips are likely to choose Birmingham New Street or International (rather than
Moor Street or Solihull).
5.6.3 Station choice zoning system
The PLD station choice zoning is provided in Table 5.9 for the London stations and in Table 5.10
for the West Midlands stations.
It can be seen that the seven London zones have nine potential station choices, making 63 new
zones in total; while the five West Midlands zones have six potential station choices, making 30
new zones. This gives 328 zones in total (235 + 63 + 30). These additional zones are used to
facilitate a logit choice of stations from each of the zones.
HS- HS-
Zone BHM BMO BHI SOL
CBirm PWBirm
BIRM 5 421 422 423 424 425 426
DUDL 47 431 432 433 434 435 436
SAND 176 441 442 443 444 445 446
SOLI 181 451 452 453 454 455 456
WALS 217 461 462 463 464 465 466
In this example, it can be seen that whilst many station pairs report shares of 0%, 87% of the trips
travel from Euston to Birmingham New Street (49%) or International (37%). This minimises the
phenomenon of 'unlikely trips'. As usual with a logit model, very small shares go to less attractive
options, though these outliers have the option of using a station walk link to find their way on to
the best assignment, so are not forced to board at those stations.
5.7 Interfaces
5.7.1 Overview
Section 3 shows the structure of the multiple models. The framework makes use of the best parts
of the different models to ensure that strategic trips are handled by PLD, while AM crowding in
London and the West Midlands is modelled in PS and PM respectively. To do this, the framework
runs the models iteratively, passing the required data from each model to the next. This is shown
in detail in Figure 5.8.
PLD
Surface Long Long
access distance distance
costs matrix pre‐loads
LHR PS PM
PLD
The different time periods of the models are handled by using corridor-specific factors to convert
select link matrices from 16 hours to 3 hours, or in terms of pre-load passengers per train per
hour.
This section describes the interactions between PLD and the other models in the framework. The
overall structure is set up to allow model results to be extracted from the most suitable
components, such as long distance trips from PLD, intra south east trips from PS. This ensures
that these trips are modelled and extracted once and once only from the framework of models.
Figure 5.8 above shows the relationships between the models.
5.7.2 Demand transfer to PS
16 hour demand matrices are extracted from PLD at cordon points to the south east. These are
converted to 3 hour matrices for import to PS as long distance demand, which was stripped from
PS.
Demand transfer from PLD to PS is undertaken by means of a series of “select link” matrices, to
enable the pattern of demand to be retained and assigned in PS. This process is undertaken for
each direction and for each journey purpose, although the demand by purpose is then aggregated
to a single matrix prior to export.
As the zone systems differ between PLD and PS, the demand is exported using zone groups
which can then be equivalenced and disaggregated in PS where appropriate. The trip end outside
the PS model area is given a unique zone group, as identified in the table below, which is
represented as a specific additional zone in PS. The trip end within the PS model area is exported
using a zone group “gs” where each PLD zone has a unique definition (gs01 to gs235). In addition
factors are applied to convert the daily demand exported from PLD to an equivalent AM Peak
period demand, as used in PS.
Details of the select links and zone groups used are provided in Table 5.12 below.
Table 5.12 - Demand Interface between PLD and PS
Sector 2007/08
Matrix Totals (cars
per day)
Business 1,340,083
Other 2,103,305
Commuter 1,335,254
Total 4,778,642
6.5 Interfaces
The highway speeds are used to produce Car Available rail access and egress times 11 on a
subnetwork between zones and stations. This ensures that highway congestion is represented in
the choice of railhead for rail passengers who have a car available. The same times are available
for air passengers 12 , who are assumed to have car/taxi available to connect to the relevant
airports.
11
At the home end of the trip only.
12
At both ends of the trip.
5082342/5082342 Model Development Report (26-02-2010).doc 53
Model Development Report
2004 2007
Business 7,455,111 7,261,279
Leisure 5,867,504 5,714,950
Total 13,322,615 12,976,229
The annual matrices were subsequently converted into daily matrices by dividing by 365, as
summarised in Table 7.2 below.
2007
Business 19,893
Leisure 15,657
Total 35,551
PLD allows airport trips to choose between available airports using the strategic highway network,
and this functionality is being retained. Associated sense-checks were undertaken, such as:
Checking that the “fare paid” is non-zero for airport available trips, using PLD skims (this
ensures that no trips from, say Berkshire to Buckinghamshire are contained within the air
matrix); and
Catchment areas by airport for trips to London and South East England.
Analysis indicated that more than 70% of the air demand to South East England is generated from
Scotland. Manchester and the North East of England generate the other reasonably significantly
demand. This is summarised below in Table 7.3.
Table 7.3 - Percentage of Demand to SE England
A sense check with the CAA end to end data at Heathrow airport suggests this data to be in the
right order with more than half of all end to end trips to Heathrow being from Scotland.
Manchester and Newcastle are the other largest generators of demand to Heathrow.
7.2.3 Air demand matrix checks
The following validation checks were undertaken on the updated air demand matrices.
Boarders by airport, checked against CAA data (via DfT’s Long Distance Model validation
report); and
Loading figures on all flights, to ensure reasonable levels of boardings per plane.
SKM provided end-to-end domestic passenger movements at Heathrow airport. These were used
to check that the air model within PLD was validating appropriately. The results of this validation
are summarised in Table 7.4 below.
Table 7.4 - Air Passenger Data & Modelled Flows at Heathrow Airport
As there is no capacity constraint or crowding on air travel within PLD, the loading figures for each
of the transit lines were checked to ensure no flights were carrying unrealistic numbers of
passengers.
Parameter Value
Board Penalty (mins.) 163
The air assignment uses the sum of the in-flight travel time and the average fare (converted to
equivalent minutes).
interpreted as implicitly including some element of trip substitution for shorter distance trips not
represented in the PSM trip matrices.
Separate mode choice parameters were derived for business (all), commuting (car available and
non car available) and other (car available and non car available). Since only limited information
was available for commuting trips, information from the stated preference surveys was
extrapolated by comparison with standard PDFH elasticities. The parameters used in the model
are shown in Table 8.1
Table 8.1 - PSM Mode Choice Parameters (2002 Prices and Values)
Commuting Commuting
CA NCA Other CA Other NCA Business
Values of Time (p/min)
The PSM forecast of induced rail travel in the corridor may be assumed in part to be
diverted from other corridors and some proportion could be assumed to come from
rail. To examine this, the levels of induced travel and the implied elasticities for key
movements (e.g London to Birmingham) were reviewed during the Steering Group
Meeting (July 2009)
Potentially, the Long Distance Travel Model can be used to provide some information
on the potential relative scale of the loss of revenue due to re-distribution arising from
HSL services; it seems likely that this could be done through an analysis of the test
runs already prepared. However, this would not be possible within HS2 model
development timescales.
8.3.2 Representation of the choice between High Speed and Conventional Rail
The choice between conventional and high speed rail has several key fundamental issues:
Whether the existing PSM absolute logit choice structure is applicable, and if so what
needs to be updated;
How to deal with premium fares;
David Ashley’s review considered the mode constants to be the least reliable output of stated
preference surveys. The mode constant in the original PSM model for high speed rail relative to
conventional rail are, approximately 25 -35 minutes for Business and 35-45 minutes for leisure.
13
TN4: Development of HS2 Demand Model, 5th June 2009 (Michael Hayes / David Ashley / John Bates)
5082342/5082342 Model Development Report (26-02-2010).doc 60
Model Development Report
The leisure values were insignificant in the original Stated Preference work, but once the value of
time was reduced to a more reasonable value, the mode constants increased (to compensate).
The other mode choice parameters were deemed suitable (see 8.3.3).
Two possible options to obtain more realistic high speed model parameters were: to obtain values
from other models or undertake additional stated preference surveys. The first option was
considered inappropriate, and the second would take too much time without any guarantee of
providing more reliable values.
A new approach was proposed by the Analytical Challenge Panel (July 2009) as a direct
replacement for the absolute choice between high speed and conventional rail. This approach
required a distribution of value of time to calculate the proportion of people willing to pay for high
speed rail’s journey time savings. The approach intrinsically takes into account a fare premium,
and trades this off against the generalised time saving. The advantage of the approach are that it
would give better estimates of the effects of fares, but as fare premiums are not the main focus of
the study would not necessarily be essential.
The distributed value of time approach (see section 8.5) was initially applied in a spreadsheet
model that allowed the process to be debugged and independently reviewed. Once the process
had been checked, it was transferred to PLD, and tested for a range of fare premium scenarios.
The performance of these tests was reviewed by the HS2 Analytical Challenge Panel
(November/December 2009).
As a result of the initial options, the appropriateness of mode constants within the mode choice
model was reviewed. The mode constants represent “inherent preferences” such as comfort that
may no longer be relevant between modern high speed and conventional trains. . The conclusion
reached was that these qualitative differences were of negligible importance compared to the
actual time savings that the services offer.
This led to a new approach that would remove the uncertainty of the high speed model
parameters, by removing the high speed sub-mode altogether; instead, these high speed services
would be considered as conventional services with improved reliability. As the demand using the
high speed services would be indistinguishable from the other conventional demand, the share of
high speed user would be determined by assignment rather than the matrix approach used within
the mode choice model.
Based on the decisions to remove the mode constant, and following the review of the distributed
value of time approach, the following decisions were made:
To remove the high speed mode from the central case (where no fare premium would
be present)
To undertake any fare premium analysis using the distributed value of time approach.
feasible, the extra uncertainties associated with the judgements needed to parameterise these
procedures would exceed the perceived forecasting benefits.
The use of the distributed value of time approach intrinsically considers value of time segments for
the rail market. One of the outstanding challenges is how best to segment this distribution within
the appraisal, whilst keeping it consistent with the central case (where only a single central value
of time is present).
Typically we might expect urban mode choice parameters to be around -0.03 to -0.06. We expect
values to reduce with distance, and the modelling reviewed confirm that for international high
speed lines the values are typically around -0.01 and -0.02.
Therefore, PSM values, which are mostly in the range of -0.02 to -0.04, seem credible,
recognising the range of distances covered. Commuting is questionable, but as this makes up
such a small part of long distance travel, the impact is likely to be small.
The implied elasticities are discussed in detail in section 8.4.2. After the changes to demand, and
value of time growth, the output elasticities were deemed adequate.
DT
1
DB
elasticity
CT 0.1 * C B
1
CB
5082342/5082342 Model Development Report (26-02-2010).doc 62
Model Development Report
Where:
Excluding short trips (0-50 miles), The elasticity of rail demand to rail fares is in-line
with guidance; and
The elasticity of highway demand to highway fuel cost is of a lower magnitude than the
expected value of -0.3. Since no option tests or sensitivities include changes to
highway fuel cost, this is unlikely to be an issue for HS2 work, but may need
addressing at a later stage.
The following chart (Figure 8.2) shows a fitted High Speed Share by journey time derived from
observed market shares on major corridors within the EU. The elasticities derived from this data
are (based on an average speed of 100mph): -0.55 for 200 miles, and -1.55 for 300miles, not
including generation effects.
Comparing these with PLD, estimated rail journey time elasticities are approximately -1.5 for 200
miles, and -3.0 for 300 miles in PLD, These are uniformly higher than those derived from the
observed data, which can be attributed, in part, to the generation response. Based on this and
other guidance, the elasticities up to 200 miles are at the upper end of the expected range.
However, it should be emphasised though – by contrast – that the level of demand response in
PLD at the lower end of the distance range (between 100 and 200 miles) is likely to underestimate
the demand impacts of improved rail services.
Inherently, logit based forecasting models give implied elasticities that increase with journey
length, contrasting with elasticity based forecasting models which maintain fixed elasticities
regardless of journey length. Although the original PSM mode choice calibration specifically
investigated a distance relationship, only a weak relationship was isolated. Further investigation
and development of parameter models may improve this relationship in the future.
1.0
0.9
0.8
0.7
HS Mode Share
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0 50 100 150 200 250 300 350 400 450 500
Journey Time (Minutes)
Figure 8.2 - Fitted High Speed Proportion by Journey Time from EU data on major corridors
(C H C S )
pr[ HSR / All Rail ] pr[v v* ] f (v) dv 1 F v *
(TH TS ) v*
where C and T are the cost and time components of GJT respectively, f(v) is the distribution of
VoT, and F(v) its cumulative distribution.
To model this effectively in EMME, a log-normal distribution of passenger Value of Time was
created from NRTS income data that was converted into Value of Time using some additional
guidance on the relationship between value of time, income and journey distance. A companion
spreadsheet model was built to independently verify the results produced from the EMME
distributed value of time macros.
CDF pax incomes: actual vs log normal: Business/Leisure [to/from LSE > 75 miles]
1.0
CA Lei Non-LSE Actual
0.9 CA Lei Non-LSE Log norm
CA Lei LSE Actual
0.8 CA Lei LSE Log norm
CA Biz Non-LSE Actual
0.7 CA Biz Non-LSE Log norm
CA Biz LSE Actual
0.6 CA Biz LSE Log norm
0.5
0.4
0.3
0.2
0.1
0.0
0.0 12.5 25.0 37.5 50.0 62.5 75.0 87.5 100.0
Figure 8.3 - Chart showing the fit of lognormal distributions of income to actual data
1 (ln x ) 2
f X (x; , ) exp , x0
x 2 2 2
Where μ and σ are the mean and standard deviation of the variable’s natural logarithm (by
definition, the variable’s logarithm is normally distributed).
Each distribution derived from the NRTS data had its own individual mu and sigma (μ,σ)
parameters that define how high speed usage will react to a change in cost. There is a significant
amount of variation between purposes that is reflected in the mean/ μ. A higher variation within the
distribution is reflected by the size of σ parameter.
μ σ
ln(v*) ln(v*) 2
U C * C H .C v TH .T
ln(v*) e ln(v*) 2
2
U T * TH .T C H .C
v
8.5.5 Implications for Appraisal
With the use of DVoT changes would need to be made to the standard appraisal approach. The
main reason for this relates to the central idea behind the approach: that value of time would not
be constant, and therefore the average high speed and classic rail value of time would be different
for each journey. Hence the standard constant values would no longer be appropriate. The
standard appraisal approach of outputting the calculations as times and applying the appraisal
values of time was modified so that the values coming out using DVoT already had the derived
values of time applied.
8.5.6 Spreadsheet Models
The DVoT approach has a lot of complicated components. These would be very difficult to check
step-by-step in EMME, so initially building and testing was undertaken in spreadsheets using data
from the PLD model. Sub-matrices of five sample origins (Birmingham, Bristol, London, Leeds and
Manchester) to all destinations were imported and the necessary calculations undertaken. This
approach allowed transparency and allowed the process to be de-bugged and independently
checked. When the checking was completed the spreadsheets were used to cross-check the
implementation within EMME and to see that what was happening within PLD was sensible.
0 10 20 30
0 10 20 30
0 10 20 30
The model produced the following results (Table 8.6 and Table 8.7 for Business and Leisure
respectively):
Table 8.6 - Business Demand
London to Birmingham
0 10 20 30
Base Classic Rail Demand 785 785 785 785
Base Air Demand 0 0 0 0
Base Highway Demand 469 469 469 469
Base Total Demand 1254 1254 1254 1254
Test Classic Rail Demand 0 0 1 9
Test Air Demand 0 0 0 0
Test Highway Demand 397 412 424 440
Test HS Rail Demand 981 935 898 849
Test Total Demand 1379 1347 1322 1297
London to Glasgow
0 10 20 30
Base Classic Rail Demand 114 114 114 114
Base Air Demand 776 776 776 776
Base Highway Demand 2 2 2 2
Base Total Demand 892 892 892 892
Test Classic Rail Demand 0 0 0 0
Test Air Demand 571 589 664 679
Test Highway Demand 2 2 2 2
Test HS Rail Demand 411 384 272 249
Test Total Demand 984 974 938 931
London to Liverpool
0 10 20 30
Base Classic Rail Demand 306 306 306 306
Base Air Demand 41 41 41 41
Base Highway Demand 0 0 0 0
Base Total Demand 348 348 348 348
Test Classic Rail Demand 0 0 0 0
Test Air Demand 15 16 25 27
Test Highway Demand 0 0 0 0
Test HS Rail Demand 467 454 383 371
Test Total Demand 482 470 408 398
London to Manchester
0 10 20 30
Base Classic Rail Demand 804 804 804 804
Base Air Demand 489 489 489 489
Base Highway Demand 298 298 298 298
Base Total Demand 1591 1591 1591 1591
Test Classic Rail Demand 0 0 0 4
Test Air Demand 254 271 379 401
Test Highway Demand 204 211 258 265
Test HS Rail Demand 1479 1416 1075 1015
Test Total Demand 1936 1898 1713 1685
Table 8.6 shows that business demand does not demonstrate a significant switch to Classic Rail
with increase in Fares. The noticeable reduction in High Speed Rail usage relates to a decrease in
generated travel.
As expected, Table 8.7 illustrates that leisure users respond with a greater shift to Classic Rail
with fare increase than Business.
50000 14000
12000
40000
10000
30000 8000
20000 6000
4000
10000 2000
0 0
0% 10% 20% 30% 0% 10% 20% 30%
30000 14000
25000 12000
10000
20000
8000
15000
6000
10000 4000
5000 2000
0 0
0% 10% 20% 30% 0% 10% 20% 30%
Table 8.8 provides an analysis of the revenue changes, and cost/demand levels reveal evidence
of the following effects occurring within the DVoT model:
Crowding effects - reductions in travel decreases the level of crowding
Interaction between purposes – with a fare increase purpose with lower value of time
will be quicker to transfer to alternatives. This will free up capacity for those with a
higher value of time.
Interaction between destinations – two destinations on the same route may impact
each other if they have differing levels of attractiveness.
Optimal fare levels – there are noticeable peaks in revenue, though this is different by
origin and destination pair.
The complex interaction between these factors makes it difficult to quantify the effects of fare
change on a global basis. The situation is made even more complex by the existence of different
fare structures based on advanced purchase and time-of-travel ticket restrictions and yield
management systems. A greater appreciation these combined effects is required before an
optimal fare strategy is reached.
High speed rail services often effectively substituting for existing long-distance
services on the WCML or providing an integrated service provision with residual
services on the route. Many passengers making intermediate journeys would either
see a like-for-like replacement journey or a mix of connecting high speed and
conventional rail services. The assignment model should reflect this situation;
However, in the situation where premium fares are considered, the DVoT approach provides initial
views as to the demand and revenue effects. Further work is required to understand appropriate
levels of premium fares which could be achieved, and the value for money of implementing such
policies.
14
Estimated from passenger kilometres
5082342/5082342 Model Development Report (26-02-2010).doc 73
Model Development Report
As PS includes all services operating wholly or partly within southern England, this includes those
services represented in the PLANET Long Distance model (e.g. trains from Birmingham to London
Euston).
The Assignment Process
The model identifies the route choice for passengers, based on their journey origin and
destination, taking into account access and egress time, boarding penalties, waiting time, in-
vehicle travel time and crowding, together with penalties where interchange is involved. These
elements are combined using weights to give a single measure referred to as a “generalised time”,
which the model minimises. The model can, where appropriate, “spread” demand for a particular
origin to destination journey between different services or even routes to ensure the generalised
time is minimised.
Waiting time is calculated on the basis of the average service frequency, assuming that the
passenger arrives randomly, i.e. passengers exhibit no prior knowledge of the service timetable.
The assignment process takes account of passenger crowding on the route chosen by applying a
time penalty to the published train journey times, based on research and advice embodied in the
Passenger Demand Forecasting Handbook (PDFH). Crowding is calculated based on the number
of passengers carried between each station served, compared to the seating and notional
standing capacities of the service concerned.
The model does not take into account the fare paid by the passenger. This is excluded as a large
proportion of passengers are fare insensitive, as they hold a season ticket or similar, or their route
choice is not affected by the fare paid.
Growth in Demand
The model is required to demonstrate two elements of demand growth:
Exogenous growth in demand for rail travel over time (driven by factors such as changes
in population and economic activity); and
Endogenous growth in demand, e.g. as a result of changes in rail supply or road
congestion.
As already noted, exogenous growth is derived externally to the model from other tools and
explicitly defined for PS. This approach ensures an explicit approach to underlying demand growth
is applied in PS. It also ensures the assumptions are consistent, both between model tests and
with other demand forecasting tools, such as DfT “TEMPRO”. In addition, as different growth
assumptions can have a marked effect on the passenger flows and user benefits, it facilitates a
flexible approach to testing alternative growth assumptions.
For endogenous growth, the model uses the concept of demand elasticity to forecast the demand
response to changes in the services provided, based on the change in generalised time as a
result of such changes. There is no capability to identify modal shift or the impact of changes in
competing modes, in particular the use of the private car, as there is no comparable
representation of the costs of car travel. To enable such changes in generalised time to be
calculated, the model is used to compare the change between two scenarios or cases: a “Base”
(or “Do Minimum”) and a “Test”.
route transfer more explicitly. These trips were replaced by transferring equivalent demand data
from the associated PLD model run.
This demand data interface took the form of a set of demand tables exported from PLD which
contain the pattern of travel on each entry corridor of the rail system into the PS model area.
There were 20 corridor points (10 southbound and 10 northbound) plus allowance for a further two
points associated with demand on the HS2 corridor. The relevant demand was factored down
from the PLD level of daily demand to the AM Peak Period modelled in PS, by means of corridor
specific factors. The trip end within the PS area was then disaggregated to the finer zone system
used in PS, whilst the corridor concerned was adopted as a dummy location for the remote trip
end. This demand table was then assigned as a fourth “user class” alongside the three purpose
based demand tables. This approach was adopted to ensure that the route used and impact of
changes in such longer distance demand were more accurately reflected, taking account of the
more refined route choice and service pattern available in PS.
The passenger loading interface was required to enable PLD to take account of the impact of local
passengers on crowding levels on longer distance services. This was calculated in PS in units of
passengers per train per hour, for selected (long distance) services only, and exported. This
approach enabled this level of demand to be “pre loaded” onto the equivalent services in the PLD
model, despite the different time periods and service frequencies used in the two models, and
then included in the calculation of crowding within the PLD model assignment process.
15
The motivation behind the recent National Rail Travel Survey (NRTS) was to provide national data of the type provided
by LATS in the LSE area.
5082342/5082342 Model Development Report (26-02-2010).doc 76
Model Development Report
In the HS2 Integrated Modelling Framework (IMF), an 'island' is cut in the PS demand matrices,
such that only demand internal to the wider South East is retained (i.e. below the blue screenline
from South Wales to the Wash in Figure 9.1 below). This is to avoid overlap with origin-destination
flows covered by PLD (e.g. Birmingham to London). These flows are replaced, in the HS2 model
framework, by the transfer of demand from PLD, using a series of select link analyses, and the
combination of these demand elements into a fourth “purpose” (long distance demand), which is
assigned in PS alongside the three conventional purpose segments.
n1
This ensures that each iteration is combined with all previous iterations in a way which
encourages convergence.
The assignment parameters used are carried forward from PS, and the main parameters are
summarised in Table 9.1 below.
Table 9.1 - PS Assignment Parameters
Parameter Value
Board Penalty (minutes) 3.5 16
Wait Time Factor 0.5
Wait Time Weight 2.0
Access/Egress Time Weight 2.0
Board Time Weight 1.0
The 4th journey “purpose” (long distance demand) is used only for crowding calculations, and is
not specifically used for economic evaluation. Additionally, cost skims are also combined with the
MSA process.
PS (and PM) uses a modified crowding calculation, where both seated and standing capacities
are taken into account when calculating capacity. This is more realistic in the case of AM peak
models with significant commuting flows.
16
Rail is generally 3.5, with exceptions on the Metropolitan Line.
5082342/5082342 Model Development Report (26-02-2010).doc 78
Model Development Report
PS AM Peak
Station Station
From Station To Station Zone Group Dummy Factors
Code Code
Zone
Newport NWP Cardiff CDF gc01 999101 0.2520
Newport NWP Hereford HFD gd01 999102 3.2089
Gloucester GCR Cheltenham Spa CNM gf01 999103 0.0435
Moreton in Marsh MIM Evesham EVE gg01 999104 0.0764
Banbury BAN Leamington Spa LMS gh01 999105 0.1625
Wolverton WOL Rugby RUG gj01 999106 0.1618
Wolverton WOL Northampton NMP gk01 999107 0.0887
Bedford BDM Wellingborough WEL gl01 999108 0.0988
Peterborough PBO Leicester LEI gq01 999109 0.2408
PS AM Peak
Station Station
From Station To Station Zone Group Dummy Factors
Code Code
Zone
Peterborough PBO Grantham GRA gr01 999110 0.1723
Cardiff CDF Newport NWP gc01 999101 0.1305
Hereford HFD Newport NWP gd01 999102 0.4138
Cheltenham Spa CNM Gloucester GCR gf01 999103 0.0576
Evesham EVE Moreton in Marsh MIM gg01 999104 3.6892
Leamington Spa LMS Banbury BAN gh01 999105 0.3285
Rugby RUG Wolverton WOL gj01 999106 0.2547
Northampton NMP Wolverton WOL gk01 999107 0.6139
Wellingborough WEL Bedford BDM gl01 999108 0.4553
Leicester LEI Peterborough PBO gq01 999109 0.0241
Grantham GRA Peterborough PBO gr01 999110 0.3404
HS2 outbound from SE gu01 999111 0.1250
HS2 inbound to SE gu01 999111 0.4000
The demand data for each cordon point is imported and disaggregated within the PS model area
and then accumulated into a single matrix for all the cordon points. The resulting demand matrix
can then be assigned as part of the amended PS assignment procedure. This additional long
distance demand in PS is not evaluated in the economics, as it is already included in PLD, and is
purely there to provide suitable crowding levels in the model, and to enable the patterns of long
distance demand dispersal to be better represented and understood.
9.7.3 Demand transfer from PS
The basic process is to identify the local demand in PS which has been assigned to long distance
services, and calculate the level of that local demand in units of passenger demand per train per
hour. These preloads are then exported from the PS model and imported into PLD.
A service is considered as long distance if it has as least one stop between the London terminus
and the “cordon point” and proceeds beyond the “cordon point”. This cordon is shown in Figure
9.1 above. The initial step in the process is to define a series of “dummy services” to represent the
corridors of interest, then relevant services in those corridors are “flagged”, and finally the local
demand preload values are calculated and exported.
passenger sensitivity to crowding and fare (as revenue estimation is an output from the model).
The three purposes used are:
Commuting;
Business; or
Other (leisure).
10.1.2 How PM works
Overview
PM essentially “assigns” or loads a pattern of passenger demand onto the available services,
allocating passengers to services and routes to ensure their overall generalised journey time (a
weighted combination of in-vehicle time, crowding penalty, waiting time, boarding penalty, access
and egress times) is kept to a minimum. It then calculates the level of crowding on each service,
which changes the crowding penalties, and hence the generalised journey times. The process is
repeated iteratively until the model is expected to “converge” or reaches a stable balance of
supply and demand. This process is further discussed below.
To operate, the model requires two major data elements:
Passenger demand by journey purpose (on an Origin to Destination basis); and
A network of public transport services available to those passengers.
These are further discussed further in this section.
The PM model was developed to represent 2007/8 patterns of demand and supply just prior to the
HS2 model framework development, and hence no major updating to the model’s underlying data
was required.
To forecast the impacts of changes to the rail system, the model is usually run for a “forecast
year”, or years, to take account of exogenous demand growth, and for a “Base” and a “Test”
scenario, to assess the impacts of the changes in services proposed. These aspects are further
discussed below.
Passenger Demand Data
Demand is based on a representation of the pattern of demand from 2007/8 ticket sales data,
provided from the DfT Rail “LENNON” database, but relates solely to travel during the AM Peak
Period (07:00 – 10:00).
Demand is disaggregated into one of three journey purpose, as this is a simple indicator of
passenger sensitivity to crowding and fare (as revenue estimation is an output from the model).
The three purposes used are:
Commuting;
Business; or
Other (Leisure).
As the demand data is based on all journeys into, out from or within the east or west Midlands, it
includes passengers that are also represented in PLANET Long Distance model (e.g. journeys
from Birmingham to central London).
PM differs from the other constituent models in the HS2 model framework in that it uses a different
approach to station access, station egress and station choice. It utilises recent research,
embodied in the national accessibility model, to estimate the accessibility of the nearest five
stations to both journey origin and destination, at a very detailed geographical level (known as
Census Output Areas). This analysis leads to a station-station choice between 25 possible
combinations. This is undertaken separately for Car Available (CA) and (Non Car Available (NCA)
passengers, as these experience very different levels of accessibility.
For endogenous growth, the model uses the concept of demand elasticity to forecast the demand
response to changes in the services provided, based on the change in generalised time as a
result of such changes. There is no capability to identify modal shift or the impact of changes in
competing modes, in particular the use of the private car, as there is no comparable
representation of the costs of car travel. To enable such changes in generalised time to be
calculated, the model is used to compare the change between two scenarios or cases: a “Base”
(or “Do Minimum”) and a “Test”.
route for a given zone-zone journey, following an intervention affecting timetables and/or
crowding 17 .
PM’s deannualisation applies the same assumptions as PLD to estimate and remove weekend
travel. To filter demand further for AM peak trips, it is assumed that weekday travel before 10:00 is
dominated by use of Full-fares and Season ticketing. Certain ticket types (e.g. standard class, full-
fare day returns) are purchased only because they are valid in the AM weekday peak. This implies
that all outward travel is undertaken in this period, such that if LENNON shows 1000 journeys
from station A to station B in 2007/8 (including return legs), and with 240 working days a year,
around 2 journeys will be produced by ‘A’ and attracted to ‘B’ in each AM peak.
Moreover, for other ticket types - e.g. (former) Cheap Day Returns valid after 09:30 - the NRTS
questionnaire asks about time of travel and hence allows the estimation of the shares of (outward)
weekday journeys occurring during the AM peak.
PM matrices are split into 2 further classes per trip purpose:
Car Available (CA);
Non-Car Available (NCA);
This makes 6 demand classes in total, though these are summed to 3 classes in Origin -
Destination format for assignment. The car availability dimension allows the station choice (logit)
model to apply a shorter access time (from demand zone to station zone) when passengers are
not limited to public transport or walking.
To avoid overlap with origin-destination flows covered by PLD (e.g. Birmingham – London), the
PM demand matrices are truncated for use in HS2 model framework, with trips going beyond a
West Midlands “cordon” removed (see Figure 10.1 below). Demand to and from outside the West
Midland is transferred from PLD as a pre-load.
17
Further information on how PM models the effects of local interventions in the Midlands is available in the
DfT’s PLANET Midlands Development Report.
5082342/5082342 Model Development Report (26-02-2010).doc 85
Model Development Report
n1
This ensures that each iteration is combined with all previous iterations in a way which
encourages convergence.
The assignment parameters used are carried forward from PS, and the main parameters are
summarised in Table 10.1 below.
Table 10.1 - PM Assignment Parameters
Value
Parameter
Board Penalty (minutes) 20
Wait Time Factor 0.5
Wait Time Weight 2.0
Access/Egress Time Weight 2.0
Board Time Weight 1.0
The assignment in PM remains a 3-class public transport assignment process, but takes account
of pre-loads provided from PLD in the calculation of passenger crowding. Additionally, cost skims
are also combined with the MSA process.
It is important to point out that the main difference in the HS2 model framework between PM and
PS is that PS handles long distance trips as a matrix, while PM handles them as pre-loads.
18
London Airports Surface Access Model, created for BAA. BAA has given permission for the use of
LASAM parameters for this project.
5082342/5082342 Model Development Report (26-02-2010).doc 88
Model Development Report
For comparison purposes, CAA data suggest that a total of 258k (annual onward air passengers)
travelled between Birmingham (PLD Zone 5) and Heathrow in 2008, where there is hardly any air
service provision. For Manchester (as defined by PLD Zone 130), the corresponding figure is
176k, of which around 60% are interlining passengers. Central London (PLD Zone 117) accounts
for 13.5m, with a further 7m from the remainder of Greater London.
The Heathrow model uses a spreadsheet model developed by SKM drawing upon the knowledge
gained during development of LASAM. The Heathrow spreadsheet model is used to estimate the
following effects:
diversion to HS2 of current Heathrow surface access trips in the HS2 corridor – excluding
trips from London; and
diversion to HS2 of ‘interlining’ air passengers taking a domestic flight to/from Heathrow
with a connecting international flight.
The Heathrow model spreadsheet is a standalone spreadsheet that was incorporated into PLD.
This section discusses the spreadsheet itself, and section 11.2 will discuss how this spreadsheet
was embedded within the framework. Full details of the specification of the model can be found in
SKM’s HS2 Airport Demand Model Methodology Report.
Remove Heathrow Express, Underground, RailAir Coach and Airport Transfers from this
sub model as main mode options, as they are only relevant to trips from London, which
are excluded from this implementation. However, the modes are still available as part of
longer rail journeys on conventional or high speed rail, the main focus of this study;
Conversion of time periods (AM Peak, Interpeak, PM Peak) to average weekday;
Use of one zone to represent Heathrow. The central terminal area is used as a reference
for level-of-service;
Segmentation: PLD Car Available Business and Leisure are the only purposes dealt with
by the Heathrow Model with each split into UK/Foreign sub-segments.
As most air passengers using Heathrow who originate in the catchment area will be travelling on
international rather than domestic flights from Heathrow, international model coefficients and
economic assumptions were adopted from LASAM rather than the domestic equivalents.
The final Heathrow model structure, including all inputs, is shown in Figure 11.6 below.
To ensure the Airport Demand Model is as compatible with PLD as possible, where available, cost
skims from PLD are used in preference to those from LASAM. Key data that was absent from the
PLD skims included the following, where the LASAM data was used:
Air Check-in times;
Taxi / Minicab fares;
Airport Parking Charges.
The components of generalised cost for each mode, segment and zone are combined using the
following generalised equations:
where D = Highway Distance, = 0.4 and N = Group Size. Ψ= 0 for Business trips, and the
Parking Costs PCost are replaced by Hire Costs for Foreign passengers
The generalised cost parameters used in the formulae are given below for both the base year and
forecast future year (2031).
Table 11.1 - Airport Mode Choice Model Parameters
Generalised Cost
Parameters Business 2008 2031
Business
Business
Business
Non-UK
Non-UK
Non-UK
Non-UK
Leisure
Leisure
Leisure
Leisure
UK
UK
UK
UK
Value of time
(Heathrow) p/min 73.60 27.01 64.77 26.97 110.9 37.52 97.63 37.46
Vehicle operating cost p/km 11.79 5.39 5.39 5.39 10.54 4.81 4.81 4.81
Time coefficient α(p) 0.18 0.20 0.22 0.25 0.18 0.20 0.22 0.25
Wait coefficient β 0.49 0.55 0.47 0.66 0.49 0.55 0.47 0.66
R_Walk coefficient δ 0.17 0.25 0.22 0.30 0.17 0.25 0.22 0.30
Access coefficient φ 0.55 0.96 0.93 1.17 0.55 0.96 0.93 1.17
Rail Interchange
coefficient 0.81 0.61 0.44 0.74 0.81 0.61 0.44 0.74
Bus Interchange
coefficient 1.63 0.90 0.44 1.09 1.63 0.90 0.44 1.09
K&F time coefficient 2 α(d) 0.13 0.22 0.02 0.10 0.13 0.22 0.02 0.10
Distance exponent θ 0.40 0.50 0.40 0.50 0.40 0.50 0.40 0.50
The demand forecasts to be output from the spreadsheet in a suitable text file format for
input back into the Framework (see Section 11.2.2).
Matrix calculations undertaken within the Framework to introduce the Heathrow demand
forecasts into the assignment and mode choice model where necessary.
11.2.1 Inputs
The Heathrow Model spreadsheet requires all cost components that build the generalised costs
for each mode, for all origins/destinations to/from Heathrow. The matrices output from PLD to the
Heathrow model are summarised in Appendix Appendix C. The modelled year and directory path
are also required so that the correct files and year specific model parameters are used. Each
spreadsheet input is created using EMME macros just before the spreadsheet is called. This data
is pre-processed by a spreadsheet using Visual Basic, which is then transferred to the Heathrow
model spreadsheet.
The Heathrow spreadsheet automatically detects whether the high speed mode=h is present: if
the high speed costs are empty, then the high speed mode share is zero. This allows the
Heathrow model to work with the standard model or the DVoT approach.
11.2.2 Outputs
Additional Visual Basic code was added to the end of the Heathrow model’s calculations to
process the output in text format for EMME. This required a set format to be adopted and specific
matrix numbers to be given to the demand by mode so that it would be input back into the
databanks successfully. The output process was checked so that the totals in the spreadsheet are
preserved throughout the model.
11.2.3 Assignment and Mode Choice interaction
Additional Macros modifications were made to the Framework for the following purposes:
To introduce the Heathrow Model demand into the assignments;
To freeze the Heathrow Demand within the main mode choice model.
Calculations have been inserted into the assignment macros to overwrite demand to/from
Heathrow with demand from the spreadsheet model, for all zones. This demand is then assigned
to its relevant network, allowing it to interact with other travellers and affect their journeys.
Heathrow demand only impacts Planet South and Midlands as a result of its impact on the PLD
assignment.
The Heathrow Demand is frozen from the main PLD mode choice model, as it is the best estimate
of mode share to/from Heathrow. The Heathrow demand only has indirect impacts within the main
mode choice model: this is as a result of the Heathrow demand within the assignment having an
impact on the costs that drive the main mode choice model.
Do-Min Test
% change
Air Car Rail Air Car Rail Rail
Scotland 264 1 3 264 1 4 34%
North East 119 2 10 119 2 10 -1%
North West 122 11 26 109 10 40 57%
Yorkshire and
Humberside 32 50 30 31 49 32 5%
Wales 10 108 64 10 108 65 2%
West
Midlands 11 158 87 10 156 90 3%
East Midlands 15 257 61 15 257 61 1%
South West 0 90 2 0 90 2 4%
Total 574 677 282 558 673 303 7%
Table 11.3 - 2021 “Day1” Leisure Demand by Mode From Heathrow (produced by Heathrow Model)
Do-Min Test
% change
Air Car Rail Air Car Rail Rail
Scotland 624 5 25 622 5 27 8%
North East 295 12 52 295 12 52 -1%
North West 351 37 45 337 36 62 38%
Yorkshire and
Humberside 104 203 120 104 201 123 2%
Wales 19 364 141 18 363 144 2%
West
Midlands 40 468 325 37 467 329 1%
East Midlands 32 525 202 31 525 203 0%
South West 0 104 18 0 104 19 2%
Total 1465 1718 928 1444 1713 956 3%
Table 11.2 and Table 11.3 show consistent trends between Business and Leisure. The results
illustrate the regions that benefit most are along the route of the HS2 scheme, with the North West
increasing by the largest proportion. The most significant mode shift effect occurring is from Air to
Rail which is most likely to relate to Air being the largest market.
Correct Routing
Issue: To confirm that the correct series of nodes are visited by the transit lines
Check: Sorting transit lines by number of segments in Emme 3 to identify lines with
suspiciously high numbers of segments. These often signify a transit line where
an incorrect node is specified, causing a large detour to be required.
Result: Transit line routings are sensible.
Issue: All lines are named according to their origin and/or destination. Any line failing to
be treated by the process was left with the original CIF identifier of a letter and 5
numbers.
Check: ‘Find in Multiple Files’ in TextPad was used with a regular expression with all
output files to find any transit line header starting with a letter followed by a
number:
^a ‘[A-Z][0-9]
Result: All transit lines have a name in the correct format of AA111A.
No import errors
Result: No errors.
Issue: The trains per hour by each TOC on each link needed to match a published
source (Railplanner software, paper timetable, PIXC counts at terminus stations)
Check: The number of trains per hour on a link was compared with the existing reference
case. These were largely consistent, and areas where matches were less
convincing were investigated in more detail.
Result: The transit frequencies matched the published data in spot checks.
Train Aggregation
Issue: The number of trains per hour stopping at nodes should be correct.
Check: The number of trains stopping was checked in Emme 3 by displaying proportional
circles fed by the following configurable attribute:
ca_nlinesstop_i == isIStop*60/hdw
In particular, locations were checked where ca_nlinesstop_i==0, i.e. no trains
stopped there.
Result: This was useful to visually identify areas where a station node had an incorrect
number of stopping trains.
Issue: The number of trains per hour stopping at London Termini should be correct.
Check: This was compared with Railplanner 19 software.
Result: This was found to match to a satisfactory level.
Redundant Services
19
Railplanner Rail Timetable Tool. See http://www.travelinfosystems.com
5082342/5082342 Model Development Report (26-02-2010).doc 98
Model Development Report
fd11 (((volau + ul1) / lanes .le. 1200) * ((length / (116 - (.006 * (volau + ul1) / lanes))) * 60)) + (((volau +
ul1) / lanes .gt. 1200 .and. (volau + ul1) / lanes .lt. 2520) * (length / (109.5 - (49 * (((volau + ul1) /
lanes - 1200) / 1000)))) * 60) + (((volau + ul1) / lanes .ge. 2520) * (length / (45 / (1 + 45 * ((volau +
ul1) / lanes - 2520) / (8 * length * 2520))) * 60))
fd12 (((volau + ul1) / lanes .le. 1200) * ((length / (112 - (.006 * (volau + ul1) / lanes))) * 60)) + (((volau +
ul1) / lanes .gt. 1200 .and. (volau + ul1) / lanes .lt. 2430) * ((length / (105.5 - (49 * (((volau + ul1) /
lanes - 1200) / 1000)))) * 60)) + (((volau + ul1) / lanes .ge. 2430) * (length / (45 / (1 + 45 * (( volau +
ul1) / lanes - 2430) / (8 * length * 2430))) * 60))
fd13 (((volau + ul1) / lanes .le. 1080) * ((length / (108.5 - (.006 * (volau + ul1) / lanes))) * 60)) + (((volau +
ul1) / lanes .gt. 1080 .and. (volau + ul1) / lanes .lt. 2260) * ((length / (102.5 - (49 * (((volau + ul1) /
lanes - 1080) / 1000)))) * 60)) + (((volau + ul1) / lanes .ge. 2260) * (length / (45 / (1 + 45 * (( volau +
ul1) / lanes - 2260) / (8 * length * 2260))) * 60))
fd14 (((volau + ul1) / lanes .le. 1080) * ((length / (104.5 - (.006 * (volau + ul1) / lanes))) * 60)) + (((volau +
ul1) / lanes .gt. 1080 .and. (volau + ul1) / lanes .lt. 2180) * ((length / (98.5 - (49 * (((volau + ul1) /
lanes - 1080) / 1000)))) * 60)) + (((volau + ul1) / lanes .ge. 2180) * (length / (45 / (1 + 45 * (( volau +
ul1) / lanes - 2180) / (8 * length * 2180))) * 60))
fd15 (((volau + ul1) / lanes .le. 1100) * ((length / (91 - (.0175 * ( volau + ul1) / lanes))) * 60)) + (((volau +
ul1) / lanes .gt. 1100 .and. (volau + ul1) / lanes .lt. 1860) * (length / (71.5 - (.035 * ((volau + ul1) /
lanes - 1100)))) * 60) + (((volau + ul1) / lanes .ge. 1860) * (length / (45 / (1 + 45 * ((volau + ul1) /
lanes - 1860) / (8 * length * 1860))) * 60))
fd16 (((volau + ul1) / lanes .le. 1200) * ((length / (96 - (.006 * ( volau + ul1) / lanes))) * 60)) + (((volau +
ul1) / lanes .gt. 1200 .and. (volau + ul1) / lanes .lt. 2520) * (length / (88 - ( 49 * (((volau + ul1) /
lanes - 1200) / 1000)))) * 60) + (((volau + ul1) / lanes .ge. 2520) * (length / (45 / (1 + 45 * ((volau +
ul1) / lanes - 2520) / (8 * length * 2520))) * 60))
fd17 (((volau + ul1) / lanes .le. 1050) * ((length / (61 - (.006 * ( volau + ul1) / lanes))) * 60)) + (((volau +
ul1) / lanes .gt. 1050 .and. (volau + ul1) / lanes .lt. 1631) * (length / (45 - ( 45 * (((volau + ul1) /
lanes - 1200) / 1000)))) * 60) + (((volau + ul1) / lanes .ge. 1631) * (length / (45 / (1 + 45 * ((volau +
ul1) / lanes - 1631) / (8 * length * 2520))) * 60))
High Speed mf225 mf228 From home In-vehicle times (Conventional Component)
High Speed mf255 mf258 From home In-vehicle times (High Speed Component)
High Speed mf256 mf259 To home In-vehicle times (High Speed Component)
High Speed mf275 mf278 From home Additional crowded time (Conventional Component)
High Speed mf276 mf279 To home Additional crowded time (Conventional Component)
High Speed mf285 mf288 From home Additional crowded time (High Speed Component)
High Speed mf286 mf289 To home Additional crowded time (High Speed Component)
High Speed mf295 mf298 From Home No. of rail only boardings
Rail Times Rail Fares Highway Times Highway Fuel Cost Air Times
Lond ‐ B'ham ‐0.83 ‐0.34 ‐0.30 ‐0.49 ‐0.48 0.00 ‐0.42 0.00 0.52 0.06 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Wolv - Lond ‐1.14 ‐0.49 0.00 ‐0.39 0.00 ‐0.54 ‐0.54 0.00 ‐0.43 0.00 0.62 0.13 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Lond ‐ Wolv ‐1.47 ‐0.50 0.00 ‐0.38 0.00 ‐0.56 ‐0.47 0.00 ‐0.36 0.00 0.97 0.15 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Cov - Lond ‐0.55 ‐0.27 0.00 ‐0.23 0.00 ‐0.35 ‐0.45 0.00 ‐0.38 0.00 0.33 0.07 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Lond ‐ Cov ‐0.46 ‐0.31 0.00 ‐0.23 0.00 ‐0.34 ‐0.56 0.00 ‐0.41 0.00 0.23 0.17 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
B'ham - LondNE ‐1.39 ‐0.86 ‐0.29 ‐0.73 ‐1.15 0.00 ‐0.38 0.00 1.36 0.88 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
LondNE ‐ B'ham ‐1.37 ‐0.72 ‐0.28 ‐0.73 ‐1.06 0.00 ‐0.41 0.00 1.39 0.73 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
0-50 miles ‐0.31 ‐0.32 ‐0.15 ‐0.08 ‐0.07 ‐0.09 ‐0.23 ‐0.06 ‐0.06 ‐0.03 0.19 0.02 0.11 0.00 0.00 ‐0.05 ‐0.13 0.00 0.00 0.00 ‐0.05 ‐0.13 0.00 0.00 0.00
50-100 miles ‐1.67 ‐1.23 ‐0.55 ‐0.37 ‐0.35 ‐0.71 ‐1.03 ‐0.26 ‐0.27 ‐0.16 0.25 ‐0.24 0.25 0.00 0.00 ‐0.45 ‐0.57 0.00 0.00 0.00 ‐0.45 ‐0.57 0.00 0.00 0.00
100-150 miles ‐1.63 ‐0.91 ‐0.76 ‐0.47 ‐0.49 ‐0.59 ‐0.76 ‐0.66 ‐0.42 ‐0.32 0.62 0.27 0.24 0.00 0.00 ‐0.15 ‐0.10 0.00 0.00 0.00 ‐0.14 ‐0.10 0.00 0.00 0.00
150-200 miles ‐1.66 ‐0.90 ‐1.86 ‐0.52 ‐0.79 ‐0.55 ‐0.88 ‐2.50 ‐0.58 ‐1.08 0.63 0.28 0.93 0.00 0.00 0.09 0.00 0.00 0.00 0.00 0.28 0.05 0.00 0.00 0.00
200-300 miles ‐3.07 ‐1.85 ‐1.05 ‐0.73 ‐0.93 ‐0.78 ‐1.34 ‐1.06 ‐0.61 ‐0.98 0.43 0.36 0.10 0.00 0.00 ‐0.12 ‐0.07 0.00 0.00 0.00 0.20 0.13 0.00 0.00 0.00
300+ miles ‐6.62 ‐3.98 ‐4.76 ‐0.98 ‐1.34 ‐0.87 ‐2.13 ‐3.16 ‐0.60 ‐0.98 0.43 0.26 2.65 0.00 0.00 0.11 0.04 0.00 0.00 0.00 1.10 0.79 0.00 0.00 0.00
Full matrix -0.85 -0.66 -0.17 -0.25 -0.09 -0.28 -0.52 -0.08 -0.21 -0.04 0.29 0.06 0.12 0.00 0.00 -0.09 -0.17 0.00 0.00 0.00 -0.05 -0.14 0.00 0.00 0.00
Rail Times Rail Fares Highway Times Highway Fuel Cost Air Times
Lond ‐ B'ham 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Wolv - Lond 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Lond ‐ Wolv 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Cov - Lond 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Lond ‐ Cov 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
B'ham - LondNE 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
LondNE ‐ B'ham 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
0-50 miles 0.01 0.00 0.00 0.00 0.00 0.08 0.04 0.00 0.00 0.00 0.02 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
50-100 miles 1.01 0.35 0.00 0.00 0.00 0.87 0.19 0.00 0.00 0.00 1.07 0.19 0.00 0.00 0.00 0.84 ‐0.01 0.00 0.00 0.00 0.84 ‐0.01 0.00 0.00 0.00
100-150 miles 0.58 0.61 0.00 0.00 0.00 0.22 0.49 0.00 0.00 0.00 0.64 0.40 0.00 0.00 0.00 0.05 0.20 0.00 0.00 0.00 ‐0.68 ‐0.21 0.00 0.00 0.00
150-200 miles 0.64 1.04 0.00 0.00 0.00 0.27 1.01 0.00 0.00 0.00 0.60 0.39 0.00 0.00 0.00 0.03 0.06 0.00 0.00 0.00 ‐0.77 ‐0.77 0.00 0.00 0.00
200-300 miles 0.41 0.70 0.00 0.00 0.00 0.10 0.48 0.00 0.00 0.00 0.58 0.36 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 ‐0.70 ‐0.63 0.00 0.00 0.00
300+ miles 0.33 0.67 0.00 0.00 0.00 0.05 0.37 0.00 0.00 0.00 0.30 0.38 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 ‐0.60 ‐0.60 0.00 0.00 0.00
Full matrix 0.42 0.72 0.00 0.00 0.00 0.11 0.48 0.00 0.00 0.00 0.44 0.37 0.00 0.00 0.00 0.01 0.01 0.00 0.00 0.00 -0.66 -0.62 0.00 0.00 0.00
Rail Times Rail Fares Highway Times Highway Fuel Cost Air Times
LonSW ‐ NYork 0.64 0.50 0.00 0.00 0.00 0.27 0.65 0.00 0.00 0.00 ‐2.78 ‐1.82 ‐0.96 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.06 0.01 0.00 0.00 0.00
B'ham - Lond 0.65 0.53 0.00 0.00 0.00 0.34 0.63 0.00 0.00 0.00 ‐1.68 ‐1.25 ‐0.45 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Lond ‐ B'ham 0.58 0.54 0.00 0.00 0.00 0.35 0.75 0.00 0.00 0.00 ‐1.60 ‐1.31 ‐0.46 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Wolv - Lond 0.68 0.64 0.00 0.00 0.00 0.32 0.70 0.00 0.00 0.00 ‐1.57 ‐1.29 ‐0.48 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Lond ‐ Wolv 0.38 0.62 0.00 0.00 0.00 0.15 0.59 0.00 0.00 0.00 ‐1.20 ‐1.31 ‐0.50 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Cov - Lond 0.51 0.41 0.00 0.00 0.00 0.34 0.66 0.00 0.00 0.00 ‐1.60 ‐1.16 ‐0.41 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Lond ‐ Cov 0.57 0.36 0.00 0.00 0.00 0.42 0.64 0.00 0.00 0.00 ‐1.92 ‐1.17 ‐0.44 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
B'ham - LondNE 0.02 0.03 0.00 0.00 0.00 0.01 0.04 0.00 0.00 0.00 ‐0.81 ‐0.53 ‐0.51 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
LondNE ‐ B'ham 0.01 0.13 0.00 0.00 0.00 0.01 0.20 0.00 0.00 0.00 ‐0.81 ‐0.73 ‐0.52 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
0-50 miles ‐0.16 ‐0.08 0.03 0.00 0.00 ‐0.17 ‐0.08 0.01 0.00 0.00 ‐0.57 ‐0.35 ‐0.25 0.00 0.00 ‐0.17 ‐0.09 0.00 0.00 0.00 ‐0.17 ‐0.09 0.00 0.00 0.00
50-100 miles ‐0.11 ‐0.13 0.05 0.00 0.00 ‐0.14 ‐0.14 0.03 0.00 0.00 ‐0.86 ‐0.65 ‐0.47 0.00 0.00 ‐0.15 ‐0.17 0.00 0.00 0.00 ‐0.15 ‐0.17 0.00 0.00 0.00
100-150 miles 0.21 0.25 0.00 0.00 0.00 0.15 0.22 0.00 0.00 0.00 ‐1.01 ‐0.62 ‐0.66 0.00 0.00 0.12 0.15 0.00 0.00 0.00 0.12 0.15 0.00 0.00 0.00
150-200 miles 0.25 0.27 0.01 0.00 0.00 0.17 0.23 0.01 0.00 0.00 ‐1.41 ‐0.86 ‐0.85 0.00 0.00 0.14 0.14 0.00 0.00 0.00 0.17 0.15 0.00 0.00 0.00
200-300 miles 0.08 0.27 0.00 0.00 0.00 0.01 0.23 0.00 0.00 0.00 ‐1.94 ‐1.10 ‐1.05 0.00 0.00 0.00 0.16 0.00 0.00 0.00 0.05 0.19 0.00 0.00 0.00
300+ miles 0.29 0.26 0.01 0.00 0.00 0.17 0.18 0.00 0.00 0.00 ‐4.29 ‐2.54 ‐1.78 0.00 0.00 0.15 0.12 0.00 0.00 0.00 0.43 0.31 0.00 0.00 0.00
Full matrix -0.11 -0.05 0.03 0.00 0.00 -0.13 -0.06 0.01 0.00 0.00 -0.70 -0.43 -0.28 0.00 0.00 -0.13 -0.07 0.00 0.00 0.00 -0.13 -0.07 0.00 0.00 0.00
Rail Times Rail Fares Highway Times Highway Fuel Cost Air Times
Lond ‐ B'ham ‐0.79 ‐0.34 ‐0.30 ‐0.42 ‐0.44 0.00 ‐0.40 0.00 0.47 0.06 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Wolv - Lond ‐1.02 ‐0.46 0.00 ‐0.39 0.00 ‐0.43 ‐0.48 0.00 ‐0.40 0.00 0.51 0.10 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Lond ‐ Wolv ‐1.40 ‐0.48 0.00 ‐0.39 0.00 ‐0.47 ‐0.42 0.00 ‐0.34 0.00 0.91 0.12 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Cov - Lond ‐0.51 ‐0.26 0.00 ‐0.23 0.00 ‐0.30 ‐0.41 0.00 ‐0.36 0.00 0.29 0.06 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Lond ‐ Cov ‐0.44 ‐0.29 0.00 ‐0.23 0.00 ‐0.29 ‐0.50 0.00 ‐0.38 0.00 0.20 0.15 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
B'ham - LondNE ‐1.38 ‐0.85 ‐0.29 ‐0.65 ‐1.06 0.00 ‐0.36 0.00 1.38 0.89 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
LondNE ‐ B'ham ‐1.38 ‐0.70 ‐0.28 ‐0.65 ‐0.95 0.00 ‐0.38 0.00 1.44 0.72 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
0-50 miles ‐0.36 ‐0.38 ‐0.16 ‐0.09 ‐0.08 ‐0.12 ‐0.29 ‐0.06 ‐0.05 ‐0.03 0.18 ‐0.03 0.13 0.00 0.00 ‐0.08 ‐0.19 0.00 0.00 0.00 ‐0.08 ‐0.19 0.00 0.00 0.00
50-100 miles ‐1.70 ‐1.31 ‐0.54 ‐0.36 ‐0.35 ‐0.77 ‐1.11 ‐0.24 ‐0.25 ‐0.15 0.14 ‐0.39 0.25 0.00 0.00 ‐0.53 ‐0.70 0.00 0.00 0.00 ‐0.53 ‐0.70 0.00 0.00 0.00
100-150 miles ‐1.51 ‐0.88 ‐0.71 ‐0.46 ‐0.47 ‐0.55 ‐0.73 ‐0.52 ‐0.40 ‐0.25 0.49 0.18 0.21 0.00 0.00 ‐0.17 ‐0.14 0.00 0.00 0.00 ‐0.16 ‐0.13 0.00 0.00 0.00
150-200 miles ‐1.52 ‐0.85 ‐1.99 ‐0.51 ‐0.79 ‐0.51 ‐0.81 ‐2.41 ‐0.55 ‐0.97 0.48 0.20 1.04 0.00 0.00 0.03 ‐0.03 0.00 0.00 0.00 0.19 0.01 0.00 0.00 0.00
200-300 miles ‐2.84 ‐1.74 ‐1.07 ‐0.72 ‐0.93 ‐0.71 ‐1.22 ‐0.97 ‐0.58 ‐0.89 0.33 0.28 0.13 0.00 0.00 ‐0.15 ‐0.10 0.00 0.00 0.00 0.14 0.07 0.00 0.00 0.00
300+ miles ‐6.30 ‐3.61 ‐4.82 ‐0.96 ‐1.34 ‐0.79 ‐1.84 ‐2.86 ‐0.56 ‐0.89 0.32 0.20 2.74 0.00 0.00 0.06 0.02 0.00 0.00 0.00 1.03 0.71 0.00 0.00 0.00
Full matrix -0.93 -0.73 -0.19 -0.27 -0.10 -0.32 -0.58 -0.07 -0.22 -0.04 0.25 -0.01 0.13 0.00 0.00 -0.13 -0.22 0.00 0.00 0.00 -0.09 -0.20 0.00 0.00 0.00
Rail Times Rail Fares Highway Times Highway Fuel Cost Air Times
Lond ‐ B'ham 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Wolv - Lond 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Lond ‐ Wolv 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Cov - Lond 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Lond ‐ Cov 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
B'ham - LondNE 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
LondNE ‐ B'ham 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
0-50 miles 0.01 0.00 0.00 0.00 0.00 0.08 0.04 0.00 0.00 0.00 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
50-100 miles 0.96 0.35 0.00 0.00 0.00 0.86 0.18 0.00 0.00 0.00 0.99 0.19 0.00 0.00 0.00 0.84 0.00 0.00 0.00 0.00 0.84 0.00 0.00 0.00 0.00
100-150 miles 0.59 0.58 0.00 0.00 0.00 0.20 0.45 0.00 0.00 0.00 0.63 0.37 0.00 0.00 0.00 0.04 0.18 0.00 0.00 0.00 ‐0.70 ‐0.22 0.00 0.00 0.00
150-200 miles 0.69 1.06 0.00 0.00 0.00 0.27 0.97 0.00 0.00 0.00 0.60 0.38 0.00 0.00 0.00 0.03 0.06 0.00 0.00 0.00 ‐0.79 ‐0.78 0.00 0.00 0.00
200-300 miles 0.43 0.70 0.00 0.00 0.00 0.09 0.45 0.00 0.00 0.00 0.57 0.35 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 ‐0.70 ‐0.62 0.00 0.00 0.00
300+ miles 0.38 0.69 0.00 0.00 0.00 0.05 0.36 0.00 0.00 0.00 0.29 0.36 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 ‐0.61 ‐0.61 0.00 0.00 0.00
Full matrix 0.46 0.74 0.00 0.00 0.00 0.11 0.46 0.00 0.00 0.00 0.43 0.36 0.00 0.00 0.00 0.01 0.01 0.00 0.00 0.00 -0.67 -0.63 0.00 0.00 0.00
Rail Times Rail Fares Highway Times Highway Fuel Cost Air Times
LonSW ‐ NYork 0.74 0.60 0.00 0.00 0.00 0.31 0.72 0.00 0.00 0.00 ‐3.28 ‐2.11 ‐0.99 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.05 0.01 0.00 0.00 0.00
B'ham - Lond 0.69 0.54 0.00 0.00 0.00 0.33 0.60 0.00 0.00 0.00 ‐1.79 ‐1.30 ‐0.46 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Lond ‐ B'ham 0.62 0.55 0.00 0.00 0.00 0.34 0.71 0.00 0.00 0.00 ‐1.74 ‐1.37 ‐0.48 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Wolv - Lond 0.77 0.66 0.00 0.00 0.00 0.33 0.68 0.00 0.00 0.00 ‐1.75 ‐1.36 ‐0.49 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Lond ‐ Wolv 0.48 0.66 0.00 0.00 0.00 0.16 0.58 0.00 0.00 0.00 ‐1.39 ‐1.42 ‐0.52 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Cov - Lond 0.54 0.41 0.00 0.00 0.00 0.32 0.63 0.00 0.00 0.00 ‐1.74 ‐1.23 ‐0.42 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Lond ‐ Cov 0.59 0.37 0.00 0.00 0.00 0.39 0.62 0.00 0.00 0.00 ‐2.07 ‐1.26 ‐0.47 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
B'ham - LondNE 0.03 0.03 0.00 0.00 0.00 0.02 0.04 0.00 0.00 0.00 ‐0.85 ‐0.56 ‐0.53 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
LondNE ‐ B'ham 0.02 0.16 0.00 0.00 0.00 0.01 0.22 0.00 0.00 0.00 ‐0.86 ‐0.82 ‐0.55 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
0-50 miles ‐0.20 ‐0.10 0.03 0.00 0.00 ‐0.22 ‐0.11 0.01 0.00 0.00 ‐0.63 ‐0.38 ‐0.26 0.00 0.00 ‐0.22 ‐0.12 0.00 0.00 0.00 ‐0.22 ‐0.12 0.00 0.00 0.00
50-100 miles ‐0.15 ‐0.16 0.06 0.00 0.00 ‐0.19 ‐0.18 0.03 0.00 0.00 ‐0.93 ‐0.70 ‐0.49 0.00 0.00 ‐0.20 ‐0.21 0.00 0.00 0.00 ‐0.20 ‐0.21 0.00 0.00 0.00
100-150 miles 0.26 0.31 0.00 0.00 0.00 0.19 0.28 0.00 0.00 0.00 ‐1.02 ‐0.59 ‐0.68 0.00 0.00 0.16 0.21 0.00 0.00 0.00 0.16 0.21 0.00 0.00 0.00
150-200 miles 0.31 0.32 0.00 0.00 0.00 0.21 0.29 0.01 0.00 0.00 ‐1.44 ‐0.85 ‐0.87 0.00 0.00 0.18 0.19 0.00 0.00 0.00 0.21 0.20 0.00 0.00 0.00
200-300 miles 0.08 0.33 0.00 0.00 0.00 0.01 0.27 0.00 0.00 0.00 ‐2.02 ‐1.09 ‐1.07 0.00 0.00 0.00 0.21 0.00 0.00 0.00 0.05 0.23 0.00 0.00 0.00
300+ miles 0.35 0.30 0.00 0.00 0.00 0.21 0.21 0.00 0.00 0.00 ‐4.47 ‐2.62 ‐1.83 0.00 0.00 0.20 0.16 0.00 0.00 0.00 0.49 0.36 0.00 0.00 0.00
Full matrix -0.14 -0.07 0.03 0.00 0.00 -0.16 -0.08 0.01 0.00 0.00 -0.76 -0.46 -0.29 0.00 0.00 -0.17 -0.09 0.00 0.00 0.00 -0.17 -0.09 0.00 0.00 0.00