Broombridge Luas Broombridge St. Stephen s Green to Broombridge (Line BXD) Annex F Human Beings: Traffic Modelling Report Cabra Phibsborough Grangegorman Broadstone - DIT Parnell Dominick O Connell Upper Marlborough O Connell - GPO Westmoreland Trinity Dawson St. Stephen s Green
Railway Procurement Agency May 2010 LUAS Line BXD Traffic Modelling Report
Prepared by:... Shane Dunny Senior Consultant Approved by:... Alan O Brien Regional Director LUAS Line BXD Traffic Modelling Report Rev No Comments Checked by Approved Date by 0 Draft for Comment SD AOB June 09 1 Final Draft for Comment SD AOB July 09 2 Final Report SD AOB August 09 Ground Floor, Grand Canal House, Upper Grand Canal Street, Dublin 4, Republic of Ireland Telephone: +353 (0) 1 238 3100 Website: http://www.aecom.com Job No Traffic LUAS Line BX&D Reference RPT This document is confidential and the copyright of AECOM Limited. Any unauthorised reproduction or usage by any person other than the addressee is strictly prohibited. document2
Table of Contents 1 Introduction... 2 1.1 Project Appraisal Guidelines... 2 1.2 Project Description... 2 1.3 Report Outline... 3 1.4 Structure of Report... 3 2 Data Collection... 6 2.1 Introduction... 6 2.2 Traffic Survey Data... 6 2.3 Other Transport Models... 7 2.4 Video Surveys... 7 2.5 Alignment & Adjacent Uses Assessment... 7 3 Base Model Development... 10 3.1 Overview... 10 3.2 Cordon Selection... 10 3.3 Model Development... 12 3.4 Zone Structure... 12 3.5 Assignment Procedure... 13 4 Base Model Calibration & Validation... 16 4.1 Methodology... 16 4.2 Calibration... 16 4.3 Matrix Estimation... 20 4.4 Validation... 21 4.5 Summary... 23 5 Future Model Devleopment... 28 5.1 Overview... 28 5.2 Do Minimum Road Network Changes... 28 5.3 Do Something Road Network Changes... 32 5.4 Matrix Development... 37 5.5 Network Statistics... 39 Appendix A Model Network... 41 Appendix B Calibration Links... 42 Appendix C Validation Links... 43
Introduction
AECOM Traffic Modelling Report 2 1 Introduction 1.1 Project Appraisal Guidelines The NRA Project Appraisal Guidelines (PAG) (March 2008) set out the following deliverables required as part of the appraisal process for major road schemes: - Project Brief; - Traffic Modelling Report; - Cost Benefit Analysis; - Project Appraisal Balance Sheet; - Business Case; and - Post Project Review. The purpose of the Traffic Modelling Report (TMR) is to describe the traffic forecasting that has been undertaken. The report outlines the development of the base year traffic model and the methodology for forecasting future year demands. 1.2 Project Description Luas Line BX is the proposed cross city Luas Line that will provide a link between the Red and Green Lines and importantly, will allow an onward extension to Broombridge via Grangegorman and Broadstone. The project represents a major milestone in the development of at-grade Light Rail in Dublin. It is a project intrinsic to delivering an integrated public transport network, through connection of the existing Luas Tallaght and Sandyford Lines, and the Maynooth suburban rail line at Liffey Junction. The alignment will negotiate its way through some of the most heavily congested public spaces in the city, where significant concentrations of buses, private vehicles and pedestrians contend for use of the corridor each day. The preferred alignment is shown in Figure 1.1 below and is outlined in more detail below. Figure 1.1 Luas BXD Preferred Alignment
AECOM Traffic Modelling Report 3 The Preferred Alignment (PA) for Luas BXD runs from St. Stephens Green via Dawson Street to College Green where the line changes from double track to form a single track loop. A single track runs north along Westmoreland Street, over O Connell Bridge and along the west side of the O Connell Street median. North of the Henry Street junction the alignment diverts into the central median. At the northern end of O Connell Street the line turns east into Parnell Street and then south to run along Marlborough Street, crossing the River Liffey on a new bridge to be built by Dublin City Council, to continue along Hawkins Street and College Street and complete the single track loop joining up with the double track section of the line at College Green. North of O Connell Street a double track is proposed for the on-street section from O'Connell Street to Broadstone. The alignment will run along the northern side of Parnell Street on to Dominick Street Lower and Dominick Street Upper. It is proposed that the tracks will cross Constitution Hill at-grade at a signalised crossing. North of Broadstone Luas Line BXD will then follow the alignment of the disused railway cutting with stop locations, generally in the vicinity of Cabra Road and Fassaugh Road. An interchange stop with Iarnród Éireann Maynooth railway line will be provided at Broombridge. Provision will be made to enable a future possible extension to Finglas at a future date. The proposed stops on Luas Line BXD are Dawson St on the double track section and Westmoreland, O Connell Upper and Lower, Parnell, Marlborough and Trinity on the single track loop section, additional stops would be located at Dominick Street, Broadstone, Phibsborough, Cabra and Broombridge with a potential stop at Grangegorman. 1.3 Report Outline This report will outline the steps taken by AECOM in developing the traffic models for the Luas BXD project. We describe the type of traffic modelling software used and the background to the Dublin Office s model for the Greater Dublin Area, from which the Metro North and Luas BXD base year model were developed. This report will; - Explain the stages in the development of Luas BXD base year model, including the data sources used for the development of the matrices, road networks and validation process; - Report on the type of analysis required by the EIS, and how this model is an appropriate tool to test impacts of interest to the environmental study. - Outline technical details are then given on how the model was calibrated and the results of this calibration process. Calibration involves the correction of network and demand errors to reduce discrepancy between measured data and modelled outputs. For the purposes of forecasting it is assumed that the parameters changed during calibration remain constant over time. Validation tests the ability of the model to predict behaviour. Validation therefore should test some independent count data against flows obtained from the calibrated model. - Outline the sources on traffic model calibration/validation guidance which have been used to inform the model development process; - NRA Project Appraisal Guidelines (March 2008) - Department for Transport (UK) Advice on major scheme appraisal; - Dublin Office validation criteria; and - SATURN manual validation guidelines; 1.4 Structure of Report The structure of this report is outlined below; Chapter 2 discusses the principal data sources used in the model, including traffic counts, journey time surveys, on street activity and other transport models (DTO Model), together with the processing of this data where appropriate. This chapter also provides an overview of some of the fundamental characteristics of the model, notably the software, cordoning process, time periods
AECOM Traffic Modelling Report 4 modelled, zone structure, assignment procedure, convergence criteria, values adopted for key parameters, network coverage and coding conventions. Development of the network is described in detail in Chapter 3 which also includes the procedures for creating the trip matrices and developing the base model. Chapter 4 sets out the calibration methods employed to ensure the Luas BXD models are fit for purpose to be used to assess impacts of Luas. To calibrate the model, changes are made to the highway network and vehicle trip matrices to obtain a better fit between observed and modelled traffic flows. The key results of the model validation are also presented and discussed in Chapter 4, following which Chapter 5 outlines the method and processes involved in the creation of the future models for forecasting. All figures and tables have been embedded in the main body of the report where possible. However, some detailed information, which would otherwise interrupt the flow of the text, has been placed in appendices at the end of the report.
Data Collection
AECOM Traffic Modelling Report 6 2 Data Collection 2.1 Introduction In order to develop a Traffic Model, a significant level of traffic data is required to ensure that the model can replicate existing traffic patterns and volumes. This section of the Traffic Modelling Report describes the collation of traffic data for the construction of the Luas BXD Model. 2.2 Traffic Survey Data The following surveys were commissioned to fully appreciate the characteristics of the local road network and were utilised for validating/calibrating the traffic model. The junctions and survey type undertaken are listed below. - Junction 1 St Stephen s Green/Dawson Street; - Junction 2 Dawson Street/Nassau Street; - Junction 3 Suffolk Street/College Green; - Junction 4 Nassau Street/Kildare Street; - Junction 5 Westmoreland Street/College Green/Grafton Street; - Junction 6 Dame Street/South Great Georges Street; - Junction 7 Pearse Street/Westland Row/Lombard Street East; - Junction 8 Townsend Street/Lombard St/Hanover Street; - Junction 9 Pearse Street/Shaw Street; - Junction 10 Hawkins Street/Townsend Street; - Junction 11 Eden Quay/Marlborough Street; - Junction 12 Hawkins Street/Burgh Quay; - Junction 13 D Olier Street/Westmoreland Street/Aston Quay/ O Connell Bridge; - Junction 14 O Connell Bridge/Bachelors Walk/O Connell Street/Eden Quay; - Junction 15 Inn s Quay/Arran Quay/Church Street; - Junction 16 Essex Quay/Wellington Quay/Parliament Street; - Junction 17 Upper Ormond Quay/Capel Street; - Junction 18 Wood Quay/Winetavern; - Junction 19 Upper Ormond/Chancery Place; - Junction 20 O Connell Street/Abbey Street; - Junction 21 O Connell Street/Cathedral Street; - Junction 22 O Connell Street/Cathal Brugha Street; - Junction 23 Marlborough Street/Talbot Street; - Junction 24 Cathal Brugha Street/Marlborough Street; - Junction 25 Parnell Street/Marlborough Street; - Junction 26 Parnell Street/O Connell Street; - Junction 27 Parnell Square South/Parnell Square West; - Junction 28 Parnell Square North/Parnell Square West; - Junction 29 Parnell Square North/Parnell Square East; - Junction 30 Parnell Street/Lower Dominick Street; - Junction 31 Upper Dorset Street/Lower Dominick Street; - Junction 32 Upper Dorset Street/Granby Row; - Junction 33 Upper Dorset Street/Frederick Street/Blessington Street; - Junction 34 Bolton Street/Capel Street; - Junction 35 Church Street/King Street; - Junction 36 Mountjoy Street/Western Way/St Mary s Place; - Junction 37 Upper Dominick Street/Western Way; - Junction 38 Constitution Hill/Western Way; - Junction 39 Lower Dorset Street/North Circular Road; - Junction 40 North Circular Road/N3; - Junction 41 Prospect Road / North Circular Road / Phibsborough Road
AECOM Traffic Modelling Report 7 Overall 41 Manual Traffic Counts were commissioned. A map highlighting the locations of these surveys can be seen in Figures 2.1 2.3 below. Figure 2.1 Southern Surveys Figure 2.2 Central Surveys Figure 2.3 Northern Surveys 2.3 Other Transport Models SATURN models developed by MVA on behalf of the RPA for Metro North were provided and were used as a base from which the Luas BXD model was cordoned. Models were available for 2011, 2013 and 2028 future years. 2.4 Video Surveys As part of the traffic counts undertaken above videos of each of the junctions were provided to allow differentiation between cars and taxi s, private bus and Dublin bus etc. 2.5 Alignment & Adjacent Uses Assessment Prior to the beginning of the modelling process it is imperative to understand the impact of the tram on traffic along each alignment, and therefore we would undertake a route long links and nodes appraisal of each option, to take into account the feasibility in relation to parking, servicing, pedestrians, traffic flow, buses and a variety of other issues. As part of this stage detailed site visits were undertaken during the AM peak, PM Peak, in mid evening and at a weekend to; - Record all traffic management layouts at 1:1000 scale; - Record information on adjacent land use and in particular parking and service access facilities; - Record all visible transport infrastructure (e.g. bus stops/taxi ranks/signing); - Review the use of each junction and link; - Record information on signals; - Record information on pedestrian desire lines and relative volumes; - Record general site observations; and - Create a detailed digital photographic record approximately every 25m along each alignment (in both directions). - Undertake a desktop review of each junction and link to: - Understand available carriageway width and compare it to a series of design criteria for items such as parking bay widths, bus stop sizes, tram lane requirements and traffic lane widths. - Understand the use of carriageway by buses, in terms of routes and volumes of services;
AECOM Traffic Modelling Report 8 - Understand traffic flow information in terms of volumes of vehicles and turning movements at junctions, both now and for the years 2018 and 2033 through assessment of the SATURN models; - Where pedestrian count information is available use it to understand flows; - Consider signal timings given to each arm of a junction; - Consider accident records along the corridor, based on the Irish National Accident database which we currently hold; and - Review any data on on-street parking available along the route; - Consider submissions from the Public in respect of each route option. - Analysis of Accident Data to undertake a relative safety review of the alignment options. This stage considered which streets in the immediate vicinity would become alternatives for the range of functions that the highway currently serves and the ability of them to cope with changes in demand once Luas BXD is in place. This was based on local knowledge of available capacity and road layout on each of the side roads or affected approaches to nodes that emerge as still feasible. At this stage the following were considered; - Reasons for traffic to flow to/from side accesses; - How traffic re-routes when some road or turns are restricted; - Whether conversion of streets to one way or giving priority for certain modes (bus/cycle) is warranted; - How strategic routes intersecting the alignment can be maintained, while using the opportunity of the tram to reduce unnecessary movement through inappropriate surrounding streets; - The specific land uses and associated parking, servicing and maintenance; needs of each street; and - Current linkages between sets of signals and where consideration of timing changes would be needed both at individual junctions and on an area wide basis. The data outlined above was utilised to ensure a robust, fit for use model was developed to assess the impacts of the Luas BXD project.
Base Model Development
AECOM Traffic Modelling Report 10 3 Base Model Development 3.1 Overview In order to develop forecast traffic levels it is first necessary to develop a robust representation of current traffic patterns. This section of the report describes the development, calibration and validation of the Luas BXD Local Area Model (LAM). The DTO Model A comprehensive transportation demand forecasting model has been developed by the DTO to provide a useful planning tool for the Greater Dublin Area. The model was developed from an extensive survey of travel behaviour conducted across the Greater Dublin Area, and has been successfully validated as a forecasting tool. The model extends from Drogheda in the north to Arklow in the South, and Kildare in the west, with an increasing level of detail and accuracy as one travels further into Dublin City Centre. The basic structure of the DTO Model is as follows: Demand Matrix: Outlines the number of trips that will be made within the study time period (i.e. AM Peak period). Trip Distribution: Links the trip productions and attractions to determine origin to destination travel flows. Mode Split: Determines the travel mode for each origin to destination flow. A multinomial logit model is used to determine the breakdown by mode. Trip Assignment: Determines the trip route through the given transportation system for each origin to destination flow based on shortest path. In recognition of the interaction between the four components of travel behaviour, equilibration is achieved by iterating through the three stages of trip distribution, modal split and trip assignment until a reasonable level of conformity is achieved between assumed and actual highway travel speeds. Although inherent in the DTO model, the level of detail of existing models for the area along the Luas BXD corridor are not sufficiently refined for use as part of the current study, and hence some further enhancement was necessary. These enhancements are outlined in Section 3.6 below. 3.2 Cordon Selection The development of Local Area Models is well documented by DTO, in the Project Appraisal Guidelines and in the Design Manual for Roads and Bridges. Figure 3.1 below demonstrates a proposed cordon for a Local Area model that will allow the impacts of Luas BXD to be assessed. The extent of the SATURN traffic model is shown in Appendix A is this report.
AECOM Traffic Modelling Report 11 Figure 3.1 - Proposed Local Area Model Extents The previous Local Area Model for Luas BX covered an area from Parnell Square to Ranelagh. As such, the proposed extension would require additional detail to be coded along the northern fringe of the existing models to include the area up to the old Broadstone terminus. Whereas this would provide a relatively quick method for updating the Local Area Model, such an approach has been discounted following discussion with DTO. In such a discussion, it was felt that the coding of additional data would inhibit reassignment effects to be fully understood, and that the use of a Local Area model first developed in 2005 may not therefore be appropriate. We have continuously found that selection of an appropriate cordon is imperative at the outset, as any amendments to this cordon cannot be undertaken later. Below we have highlighted where the key effects of Luas BX/D are likely to occur which informs the definition of the proposed model cordon area. The most notable effects of Line BX are on Pearse Street, Fitzwilliam Square and St Stephens Green East. For Line D, the most notable effects would be expected in the area between Gardiner Street and Capel Street, with perhaps additional impacts on North Circular Road. We also propose extending the boundary further west than that used for the Luas BX Model, such that Winetavern Street is included. This will allow the potential impact of diversion away from Westmoreland Street onto Winetavern Street and Ormond Quay for northbound traffic to be better understood. We should point out that a key assumption of our model development is that there will be limited traffic impacts north of the Broadstone stop which will require assessment using the traffic model. Any impacts in this area arising from access to stops from overbridges can be dealt with in isolation as necessary.
AECOM Traffic Modelling Report 12 Finally, whilst Luas BXD is likely to reduce the level of car trips in the city, we propose maintaining a fixed trip matrix for the purpose of the highway assessment. Actual reductions in trips along roads through the area would be extremely limited and therefore would not warrant the additional labour of multi-modal modelling. In any case, the mode shift effects will be assessed by RPA in the development of the Economic case for the study. This is a robust approach for preparing the Railway Order documentation and can be supported at Oral Hearing. 3.3 Model Development Following preliminary discussions, it is deemed appropriate that the Metro North Traffic Model (MNTM) would be used as the basis for the traffic modelling. The MNTM is based on the DTO Model as outlined in Section 3.1 above. The MNTM will be cordoned as outlined in Section 3.3 below. The focus of the Luas BXD model calibration, in addition to subsequent application and testing will be on a central area of the city along the proposed alignment. On this basis, we propose the following approach: - Procure the Base Year (2005) and Construction Year Do-Minimum Metro North Traffic Model (2011 Year); - Review all roads/junctions within the assessment area and amend as appropriate to reflect current turning restrictions. In addition, check that the Dublin Port Tunnel and M50 upgrade are correctly coded into the Full Area Model (FAM); - Collect traffic count information (set out below) and incorporate this information into the modelling files for the AM and PM Peak; - Develop the 2008 Matrix by direct interpolation from the 2011 MNTM matrices; - Calibrate the model to a 2008 Base Year through further network refinements in the City Centre, and direct matrix estimation. The calibration criteria will be based on comparison of link flows within the Assessment Area outlined above (i.e. within the area where traffic count data is to be collected); - Calculate the difference in the demand matrix between the 2011, 2013 and 2029 Metro North Traffic Models. Linear interpolation was used to calculate the equivalent difference for 2008-2018 and 2008-2033. Apply these differences to the 2008 Luas BXD traffic model to obtain 2018 and 2033 Luas BXD Traffic Models. This process is discussed further in Chapter 5; Base Year and Time Periods The modelled Base Year represents an average weekday (Monday to Friday) in October 2008, and the network within the model has therefore been developed to reflect the infrastructure that was in place in October 2008. Furthermore, all traffic count information was collected during October 2008. The model has two distinct time periods the morning peak hour between 08:00 and 09:00 with preload from 07:00 to 08:00 to show passed on queuing and the evening peak hour between 17:00 and 18:00 with preload from 16:00 to 17:00 to show passed on queuing. Model Package The Luas BXD LAM utilises SATURN version 10.8.17. All assignments have used this version of the SATURN software. The network is coded in SATURN using both simulation and buffer networks. The simulation network models the delays and queues that occur at junctions and uses this information, in conjunction with link-based speed/flow curves, for the determination of route choice. The buffer network does not model junction delays as it gives a more simplistic representation of travel behaviour based entirely on speed/flow curves. 3.4 Zone Structure The zone structure for Luas BXD Traffic Model has been designed to ensure that the model loads trips onto the strategic road network at appropriate locations and that the routeing through the network is sensible. In addition, the sensitivity of the model to specific highway improvements (for example, widening a single section of the network or enhancing associated junctions) needs to be realistic, which implies that the zoning system needs sufficient detail to represent key alternative route choices. The need for detail is greater in the area where scheme-specific junction changes and transport interventions will be assessed.
AECOM Traffic Modelling Report 13 Figure 3.2 Zone Structure for Luas BXD Model From Figure 3.2, it can be observed that the size of the zone varies according to the road network and population density within the Region. In areas where the road network is sparse and population relatively low, the zone size is larger than areas with a dense road network and high population. The general expectation of the zoning was that there should be centroid connectors near key nodes of the strategic road network, together with representation of the demand accessing the network from intermediate locations. The total number of zones used within the Luas BXD model is 102. This zoning system has largely evolved from the DTO Model. The zone naming is as per above however when the above zones were cordoned to reflect the study area a number of external zones were formed as part of the cordoning process to reflect the through traffic from zones outside the study area. 3.5 Assignment Procedure The assignments are based on generalised cost, which is a function of both travel time and travel distance. The different value of time (VoT) for each user class ensures that route choices reflect the behaviour of different groups of road users. The highway assignment uses two principal components within the SATURN model, SATASS and SATSIM. Separate assignments are carried out for each time period. Initially, the matrices for each of the two user classes (Cars and HGV) are assigned to the appropriate network using the assignment program SATASS, based on Wardrop s Principle of Traffic Equilibrium, which states: Traffic arranges itself on congested networks such that the cost of travel on all routes used between each O-D pair is equal to the minimum cost of travel and all unused routes have equal or greater cost That is that each traveller chooses a least-cost route available to them. After completion of SATASS, the simulation program SATSIM is run, which calculates the delays within the simulation area based on the flows assigned in SATASS. SATASS is then run again, re assigning the trip matrix based on the delays calculated by SATSIM. This iterative procedure continues until convergence is achieved. The process can be seen diagrammatically in Figure 3.3, along with the inputs and outputs associated with the procedure. In the first instance, an assignment representing the previous hour (07:00-08:00 or 16:00-17:00), known as the Queue assignment, is undertaken, in order to add queues to the network.
AECOM Traffic Modelling Report 14 This assignment represents the hour immediately before the full assignment and ensures that, at the beginning of the main assignment, a representative level of delay is present within the network. In the main assignment, HGVs trips are assigned to the network first as a pre load, followed by the car user class assignment. The pre-loading of HGVs enables the HGV movements to be loaded onto the correct routes as the routes are significantly different to Car routings due to HGV bans at the canal cordon and other locations within the model e.g. due to low bridges. Figure 3.3: The SATURN Assignment Procedure In view of the iterative nature of the assignment, it is important to verify that convergence has been achieved, i.e. that a stable set of flows and delays has been modelled, which would not be substantially changed by further iterations. The NRA Project Appraisal Guidelines (March 2008) details best practice for transport planning procedures, including transport modelling. Accordingly, the guidance covers model development, assignments, convergence, calibration and validation criteria, as well as other related topics.
Base Model Calibration & Validation
AECOM Traffic Modelling Report 16 4 Base Model Calibration & Validation 4.1 Methodology The methodology adopted in calibrating and validating the LAM has been undertaken in line with the guidelines presented in the NRA Project Appraisal Guidelines (March 2008). This is an internationally recognised standard for traffic model development, and is in line with the protocols of the Dublin Transport Office (DTO). The process of calibration involved the following tasks: - An audit of the junctions coded to ensure that they were up to date with current configurations, link distances and signal timings; - A series of checks were made to ensure that suitable routes were being used between origins and destinations. - A first stage comparison of the modelled and observed link flows was undertaken to highlight areas where either the network or matrix was insufficient. Appropriate manual amendments were made based on this analysis; The calibration was undertaken with a view to improve the way that vehicles accessed and routed through the road network. Following the calibration process model validation was undertaken to ensure the modelled outputs were fit for use. The process of validation involved the manual adjustments to the trip matrices where a direct correlation with a traffic count could be made and further refinement of the road network and matrix estimation as outlined below. 4.2 Calibration As part of the calibration process an audit was undertaken of junction arrangements within the model to ensure that the model reflected the existing situation. Where turning counts were available the turning capacities were assessed to ensure the capacity within the model was sufficient to allow for surveyed volumes. Further refinement was undertaken based on knowledge of the local area, site visits, mapping, video surveys etc. The following parameters were altered as part of the calibration process to ensure the model best reflected existing conditions, Junction Type, No. arms at junction, no. lanes on each arm, link travel speed, link capacity, signal timings and phasing, turn capacity etc. The model calibration process is outlined in Figure 4.1 below.
AECOM Traffic Modelling Report 17 Assign base CAR and HGV matrixes to the model. Analyse output using Calibration Spreadsheet Review result, if global GEH statistic is <85% adjust matrix directly or use TFlow Fuzzy matrix estimation function. >85% <85% Import new estimated matrix Reassign the Model <85% Analyse output data, and repeat steps 1-5 until global GEH statistic > 85%. >85% CALIBRATED Figure 4.1 Model Calibration Process Following this process the modelled link flows were compared to surveyed link flows to ensure a reasonable comparison had been achieved. To further improve the calibration the matrix estimation process was used on the O-D matrix as follows. 4.2.1 Calibration Criteria The NRA Project Appraisal Guidelines specifies the acceptable values for modelled and observed flow comparisons and suggests how calibration should relate to the magnitude of the values being compared.
AECOM Traffic Modelling Report 18 A summary of these targets is shown in Table 4.1: Table 4.1 Model Calibration Criteria: Individual Flows Criteria and Measures Individual flows within 15% for flows 700 2700 vph Individual flows within 100 vph for flows <700 vph Individual flows within 400 vph for flows > 2700 Guideline > 85% of cases The standard method used to compare modelled values against observations on a link involves the calculation of the Geoff Havers (GEH) statistic (Chi-squared statistic), incorporating both relative and absolute errors. The GEH statistic is a measure of comparability that takes account of not only the difference between the observed and modelled flows, but also the significance of this difference with respect to the size of the observed flow. The GEH statistic is calculated as follows: Where M = Modelled Flow and O = Observed Flow. Guidance in the Project Appraisal Guidelines sets out the following criteria: Table 4.2 Model Calibration Criteria: GEH Values Criteria and Measures Guideline GEH statistic Individual flows: GEH < 5 > 85% of cases 4.2.2 Calibration Results The results of the calibration exercise are outlined below in Tables 4.3 and 4.4. The detailed summary tables are included in Appendix B. Table 4.3 Calibration Results: Individual Flows Time Periods % of Calibration Sites that meet the following criteria Total Traffic AM Peak Individual Flows within 15% for flows 700 2700 vph - 83% Individual flows within 100 vph for flows < 700 vph - 80% Individual flows within 400 vph for flows > 2700 vph - 100% PM Peak Individual Flows within 15% for flows 700 2700 vph - 79% Individual flows within 100 vph for flows < 700 vph - 100% Individual flows within 400 vph for flows > 2700 vph - 100% Guideline 85% 85%
AECOM Traffic Modelling Report 19 Table 4.4 Calibration Results: GEH Values Time Periods % of Calibration Sites with GEH < 5 Total Traffic Guideline AM Peak 85% 85% PM Peak 85% 85% The comparison of modelled and observed flows has identified that the AM and PM Peak period models meet the flow criteria for all user classes. Likewise, the GEH results show that the AM and PM Peak periods models also meet the criteria for all user classes. The results therefore confirm that the models are fit for use and have been calibrated to a standard compliant with the PAG criteria for all user classes and all time periods. Figure 4.2 Location of Counts used for AM Calibration
AECOM Traffic Modelling Report 20 Figure 4.3 Location of Counts used for PM Calibration As can be seen from Figure 4.2 and 4.3 above traffic counts along the boundary of the traffic survey locations were used to ensure trips coming in and out of the alignment area were calibrating. 4.3 Matrix Estimation The matrix estimation process compares the flow outputs initially forecast by the model to observed traffic count data, and adjusts the trip matrix to achieve an improved match between modelled and observed flows. Thus, the process allows the incorporation of traffic count data (which identifies volume, but has no information on origin, destination and purpose) in the matrix construction. Matrix estimation does assume that any discrepancies between observed and modelled flows are entirely due to shortfalls in the demand matrix and not in the way the model decides route choice. Therefore, it is essential to finalise network editing before commencing. Accordingly, a degree of network validation was undertaken prior to matrix estimation using logic checks and tree checking.
AECOM Traffic Modelling Report 21 The trip matrices (Car and HGV) were then fed into the SATURN matrix estimation process that compares modelled and surveyed flows and adjusts the trip matrix to improve this comparison. This process can be undertaken wholly within the SATURN suite of programs using the programs SATPIJA and SATME2. This process takes an old or prior trip matrix (in this case the AM matrix derived from the MNTM model) and current traffic counts to estimate the most likely trip matrix consistent with the information contained in the counts using the prior matrix as a starting point. In order to do this it requires a PIJA file, each element of which represents the proportion of trips between a particular origin-destination pair, which uses the counted link. The model uses an iterative procedure to find a set of balancing factors for each counted link to ensure that the assigned flows match the counts within certain user-defined limits. Certain zones were excluded ( frozen ) from the estimation process as their totals were already known to be correct. Surveyed flows used as part of the estimation process were not used for validation so as to maintain a level of independence with the validation. As there is no PM DTO model from which to obtain a PM matrix, one was created by transposing the AM matrix and factoring it using survey data and as part of the estimation process. 4.4 Validation The purpose of model validation is to assess the reliability of the information provided by the traffic model, in order to test its validity as a base for forecasting. Thus, validation is undertaken to ensure that the detailed network description, trip matrix and methods of assignment are sufficiently robust to facilitate replication of observed patterns for a given Base Year, in this case 2003. Where possible, it is carried out using independent data which have not been used in the development of the model. 4.4.1 Validation of Traffic Flows The observed and modelled flows were compared at each of the validation sites in accordance with the criteria above. The permissible difference was calculated for each value (based on the observed figure) and compared with that which had been modelled. Validation results are included in Appendix C and are summarised in Table 4.5 and 4.6 below. Table 4.5 Validation Results: Individual Flows Time Periods % of Validation Sites that meet the following criteria Total Traffic AM Peak Individual Flows within 15% for flows 700 2700 vph - 89% Individual flows within 100 vph for flows < 700 vph - 100% Individual flows within 400 vph for flows > 2700 vph - 100% PM Peak Individual Flows within 15% for flows 700 2700 vph - 81% Individual flows within 100 vph for flows < 700 vph - 94% Individual flows within 400 vph for flows > 2700 vph - 100% Guideline 85% 85% Table 4.6 Validation Results: GEH Values Time Periods % of Validation Sites with GEH < 5 Total Traffic Requirement AM Peak 85% 85% PM Peak 85% 85% The comparison against the validation counts shows that AM and PM Peak period models clearly meet the Project Appraisal Guidelines requirements for traffic flow on links and meet the GEH criteria of 85%. The results therefore demonstrate that the validation criteria are successfully met and the models are fit for use.
AECOM Traffic Modelling Report 22 The NRA Project Appraisal Guidelines state that the guidelines for a validated network are that 85% of the links should have a GEH statistic of less than 5. The results of the validation process show that this guideline was met, with 85% of links having a GEH of <5 for the AM and PM Peaks and this, along with the calibration checks undertaken, shows that the models are fit for use in forecasting mode. The link flows used in validation are included in Appendix B of this report. 4.4.2 Validation of Journey Times In addition to the link flow calibration/validation undertaken above journey times through the road network were also validated to ensure routes in the SATURN reflected the current situation. The journey time surveys were undertaken between 08:00-09:00 and 17:00-18:00 using QSTARZ GPS journey time recording software. The journey times routes used for validation are shown in Figures 4.4 and 4.5 below. Figure 4.4 AM Journey Time Route
AECOM Traffic Modelling Report 23 Figure 4.5 PM Journey Time Route The Project Appraisal guidelines set out a minimum requirement of model journey times to be +\- 15% when compared to actual journey times. Table 4.7 Comparison of Saturn and GPS Route Times Total Route Time Period Saturn Time GPS Time Difference % Difference AM Peak 01:18:06 01:16:34 00:01:32 +2% PM Peak 00:40:25 00:35:31 00:04:54 +14% As outlined in Table 4.7 above the SATURN model journey times validate well against actual journey times and meet the criteria set out in the Project Appraisal Guidelines of +/- 15%. 4.5 Summary To show the impact of the calibration and validation processes the modelled flows and observed flows were compared graphically. Figures 4.4 and 4.5 below show a plot of observed (survey) flows against modelling flows for both the AM and PM time period. The figures show that the plot was scattered with a weak correlation between observed and modelled flows prior to calibration/validation.
AECOM Traffic Modelling Report 24 Figure 4.6 - AM Pre Calibration/Validation Figure 4.7 - PM Pre Calibration/Validation Figures 4.6 and 4.7 above show that whilst the majority of points in graph are positioned adjacent to the straight line there is significant scatter in the position of the points. This highlights some significant differences between observed and modelled flows throughout the models. Calibration was then undertaken with a view to improve the way that vehicles accessed and routed through the road network. This improvement can be seen in Figures 4.8 and 4.9 with the modelled and observed flows becoming less scattered. Figure 4.8 - AM Post Calibration Figure 4.9 - PM Post Calibration The calibration process involved the refinement of speed flows curves along links, link distances, signal refinement etc, the plot of observed flows versus modelled flows became more correlated as per Figures 4.6 and 4.7 above. Following the calibration process model validation was undertaken to ensure the modelled outputs were fit for use. The process of validation involved the manual adjustments to the trip matrices where a direct correlation with a traffic count could be made and further refinement of the road network and matrix estimation. The effect of these adjustments is seen in Figures 4.8 and 4.9 with a strong correlation between observed and modelled traffic flows.
AECOM Traffic Modelling Report 25 Figure 4.10 - AM Post ME/Validation Figure 4.11 - PM Post ME/Validation The locations of the links used for validation are shown in Figures 4.12 and 4.13 below. Figure 4.12 Location and GEH range of links used for AM validation
AECOM Traffic Modelling Report 26 Figure 4.13 Location and GEH range of links used for PM validation
Future Model Development
AECOM Traffic Modelling Report 28 5 Future Model Development 5.1 Overview Opening and Design Years of 2018 and 2033 have been selected to assess the traffic impacts of Luas Line BXD. The Luas BXD 2018 and 2033 future models were developed based on future models prepared as part of the Metro North project. The growth per zone was ascertained from the Metro North models (which are a refinement of the DTO models) and applied to the validated 2008 models prepared by AECOM as part of the Luas BXD project. Models for both the future Do- Minimum and Do-Something road networks were prepared for both the 2018 and 2033 forecast years as outlined below. The Do-Minimum and Do-Something network changes are outlined in Sections 5.2 and 5.3 below. 5.2 Do Minimum Road Network Changes Following consultation with Dublin City Council and the RPA, a number of amendments to the existing road network have been identified which will be put in place regardless of the construction of Luas Line BXD. It is noted that the future years selected post date the opening of the proposed Metro North, and consequently do not include the full range of short term traffic management measures required for the construction period. The proposed infrastructural amendments to the Do Minimum network are listed below; Macken Street Bridge The construction of the Macken Street Bridge will provide another crossing of the Liffey between Guild Street and Cardiff Lane. It is anticipated that this will significantly reduce north south traffic movements in the city centre, particularly on Amiens Street. The bridge is currently under construction and is scheduled for completion in March 2010. A number of turning bans are proposed to accompany the new bridge which are listed below and illustrated in Figure 5.1: - Right turn ban from Macken St Bridge to North Wall Quay - Left turn ban from North Wall Quay to Macken Street Bridge - Right turn ban from Guild Street to North Quay - Right turn ban from Guild Street to Sheriff St Lower to prevent traffic accessing Port Tunnel from bridge - Right turn ban from Macken Street Bridge to City Quay
AECOM Traffic Modelling Report 29 Figure 5.1 Macken Street Bridge Proposals Public Gate on College Green Dublin City Council is progressing plans to implement a public transportation gate on College Green which will remove the majority of private vehicles from Westmoreland Street, College Green, Lower Grafton Street and Nassau Street (West) during the AM and PM peak periods. This measure will facilitate the Metro North works on O Connell Bridge and reinforce the city centre as a public transport friendly environment. The gate is scheduled for implementation on 27th July 2009. The restrictions on car traffic at this location are illustrated in Figure 5.2 and require the following road network changes: - College Green to Westmoreland Street is 2 bus-only lanes, flaring to 3 on the approach to the College Street junction. The third lane will not be designated as a bus only lane; - Taxi rank on College Green is to be retained but a u-turn facility provided for eastbound traffic who can no longer access Westmoreland Street; - D Olier Street approach to College Street is 3 lanes; the two western most lanes are for general traffic going to Westmoreland Street, the eastern lane is for buses, taxis and access only to Grafton Street (Lower); - College Street approach to Westmoreland Street is 2 lanes, both of which are for use by general traffic; - College Street approach to College Green is 2 lanes; both lanes are designated as bus lanes; and - College Green westbound is 2 lanes.
AECOM Traffic Modelling Report 30 Figure 5.2 Public Gate Proposals
AECOM Traffic Modelling Report 31 Marlborough Street Public Transport Bridge Dublin City Council is proceeding with plans to provide a new bridge over the Liffey between Marlborough Street and Hawkins Street. This new facility is intended to provide an additional river crossing point for buses in the city centre and will be particularly valuable during Metro North construction works when there will be significant traffic disruption on O Connell Street/O Connell Bridge. Figure 5.3 Marlborough Street Public Transport Bridge Proposals Right Turn Ban from Westmoreland Street to D Olier Street This measure is being proposed to facilitate the construction and operation of Metro North. It will be retained permanently on completion of the works. Left Turn Ban from Westmoreland Street to Fleet Street This measure is being proposed to facilitate the operation of Metro North and will be retained permanently on completion of these works. Closure of Glover s Alley It is proposed to close the Glovers Alley access from St. Stephen s Green West in order to facilitate a reduction in traffic levels during Metro North construction and operation. Alternative access to the multi-storey car parks will be available from Mercer Street (York Street). Right Turn Ban from O Connell Bridge to Eden Quay This measure is being proposed to facilitate the construction of Metro North. It will be retained permanently on completion of the works to facilitate pedestrian movements in the area.
AECOM Traffic Modelling Report 32 5.3 Do Something Road Network Changes This section examines each section of the Luas BXD alignment in turn to assess the required road network changes to facilitate the operation of Luas trams. It is noted that the changes identified are in addition to those required for the Do Minimum network as outlined above i.e. those amendments to the network which will occur in the absence of Luas BXD. The amendments noted in this section are in addition to traffic signal changes which will be required at each signalised junction along the alignment. The proposed network amendments are listed below; St. Stephen s Green College Street It is proposed to remove the eastbound traffic from St. Stephen s Green North when the Luas Line is in place which will result in only Luas trams being permitted to make a left turn from St. Stephen s Green North onto Dawson Street. In order to facilitate service arrangements Dawson Street will become 2-way between the junctions of Ann Street South and Duke Street. Left turns will be permitted from Dawson Lane at the junction of Dawson Lane/Duke St/Dawson St and right turns will be permitted from Duke Street onto Dawson Street. A left turn into Molesworth Street will then be permitted on the southbound lane of Dawson Street. Shared running will be permitted along this section. A right turn ban for southbound traffic from Dawson Street will be imposed at the junction of Dawson Street and Nassau Street. Due to the Luas line running along the eastern side of the road lane width along Dawson Street will be reduced. A twin track alignment is proposed on Nassau St, which will incorporate a two way shared running surface between the junction of Dawson St and Grafton Street. A one-way shared surface will be permitted between College Street and Grafton Street, existing bans on the left turn will remain unchanged. Figure 5.4 Proposed Network Changes - Dawson St /Nassau Street College St - O Connell Street Lower A single track northbound alignment will run along the eastern side of Westmoreland Street. The number of available traffic lanes on Westmoreland Street will therefore be reduced from three to two from the College Street junction to the Fleet Street junction. A northbound lane and a right turn only lane (to Fleet Street) will be provided.
AECOM Traffic Modelling Report 33 North of this junction two lanes will be provided (a dedicated bus only lane on the west side of the street and a traffic lane on the east), flaring to three lanes at the junction with Aston Quay where one left turn and two straight ahead lanes will be provided adjacent to the proposed alignment. The southbound Luas alignment will run over a proposed new public transport bridge over the Liffey between Eden Quay and Burgh Quay. The new bridge will be located at the junction of Eden Quay/Marlborough St and Burgh Quay/Hawkins Street. This bridge will be used by trams and buses. Light rail vehicles will continue south via Hawkins Street and College Street necessitating a reduction in the number of available lanes on these streets from two to one and from three to two respectively. The northbound Luas will run in a single line track alignment over O Connell Bridge and onto O Connell Street. Light rail vehicles will proceed northbound on O Connell Street Lower adjacent to the central median on the east side of the northbound carriageway. Shared running by trams and other vehicles will be facilitated from the Aston Quay junction, across O Connell Bridge and on O Connell Street Lower form the Bachelor s Walk junction to the Henry Street junction. To the north of Henry Street, on O Connell Street Upper the Luas alignment will divert into the central median between the Spire and the Cathedral Street junction. Figure 5.5 Proposed Network Changes Westmoreland St/O Connell Bridge O Connell Street Upper Parnell Street Marlborough Street The northbound Luas alignment will run within the central median on O Connell Street Upper. The right turn traffic movement from O Connell St to Cathal Brugha St will be banned to reduce the number of conflicting traffic movements. On Parnell Street (east of O Connell Street) the southbound tram alignment will follow the northern kerbside on the street. Due to the width of the existing traffic lanes it will be possible to retain the existing single lane in both directions on this stretch of the street albeit with reduced lane widths required to accommodate the proposed alignment and platform.
AECOM Traffic Modelling Report 34 The southbound Luas alignment will cross the junction of Parnell St/Marlborough Street. This junction will need to be signalised to accommodate this change and a right turn ban will be introduced from Parnell Street to Marlborough Street. The Luas will run down the eastern side of Marlborough St, and it is proposed to introduce/maintain one way flow, southbound, over its entire length as appropriate. Shared running will not be facilitated for the most part on Marlborough Street other than for access/egress as required. Cathedral Street is currently pedestrian only from the junction of Marlborough St to the junction of Thomas Lane. It is proposed to allow one-way westbound traffic on this street to facilitate access to Thomas Lane. At the junction of Marlborough St/North Earl St/Talbot St, a new signalised junction arrangement will be provided in order to facilitate the Luas alignment. Egress from North Earl Street to Marlborough Street is to be banned with all such traffic being rerouted via Earl Place and Sackville Place. The signal staging at the Abbey Street/O Connell Street junction will need to be revised to accommodate Luas BXD. Figure 5.6 Proposed Network Changes O Connell St/Parnell St/Marlborough Street
AECOM Traffic Modelling Report 35 Figure 5.7 Proposed Network Changes O Connell St/ Marlborough Street Parnell St - Dominick Street The Luas Line is proposed to run on the northern side of Parnell St between the junctions of Dominick St Lower and O Connell Street. A shared running surface will be provided for right turning traffic from Parnell St to Parnell Square West. Due to the Luas line running along Parnell Street there will be a reduction in lane width. At the junction of Moore St/Parnell St a new junction arrangement is required to reflect the location of the twin track alignment on the north side of the street however all existing traffic movements are to be retained. It is proposed to make Dominick St one way southbound therefore a revised junction arrangement and staging will be required. Traffic from Parnell St will be banned from turning into Dominick St.
AECOM Traffic Modelling Report 36 Figure 5.8 Proposed Network Changes Parnell Street Dominick Street - Constitution Hill The Luas will run in a twin track alignment on the southern side of Dominick St Lower, from the junction of Parnell St to Dorset Street. Dominick Street Lower will become a one way, southbound only street with the number of lanes being reduced to one. Figure 5.9 Proposed Network Changes Dominick St Two way shared traffic flow will be permitted on Dominick St Upper between Mountjoy Street and Dorset Street. One way southbound traffic flow will be permitted on Dominick St Upper from Palmerstown Place to Mountjoy Street. A right turn movement from Western Way to Mountjoy Street will be permitted to accommodate access to Dominick Street. No access will be allowed from Western Way to Dominick Street. Access to Temple Cottages from Dominick St is to be closed. An alternative access will be provided from Constitution Hill. An at-grade crossing of Constitution Hill has been assumed.
AECOM Traffic Modelling Report 37 Figure 5.10 Proposed Network Changes Western Way / Constitution Hill The provision of the Luas Line BXD will impact on junctions and roads along its route. The impacts vary along the route from simple signal timing changes to the banning of movements, loss of road capacity for other users and also revised junction arrangements. The details above give a summary of the impact the Luas Line BXD will have on the junctions and roads from St. Stephen s Green, where it will tie into the existing Luas Green Line, to Constitution Hill, where the alignment will go off road, and travel to Broombridge on the old railway alignment. 5.4 Matrix Development The future matrices for the Luas BXD were built by factoring the zonal trip ends (i.e. different factor for each zone in the model) based on factors ascertained from models prepared for the Metro North model. The Metro North models are a refined version of the DTO model and therefore fit for use to ascertain future traffic growth within the Luas BXD study area. Do-Something AM Metro North models (i.e. with Transport 21 schemes and Metro North in place) were available for 2013 and 2028. The process by which these models were used to create 2018 and 2033 Luas BXD future models is outlined in Figure 5.11 below.
AECOM Traffic Modelling Report 38 Figure 5.11 Future Matrix Development Process The difference between the 2013 Base MVA model and 2028 Models was calculated and the 2008, 2015 and 2030 MVA models developed using interpolation. The percentage difference in the 2008 2018 and 2008 2033 trip ends was calculated for both the origin and destination. This factor was then applied to the AECOM 2008 matrix to give Origin and Destination target trip ends. The Furness method then used to calculate 2018 and 2033 matrices.