1,2 A.W. Reynolds & ¹ B.M. Broderick,

1Air Quality Research Group,² Dept. of Mechanical Engineering,

Dept. of Civil Engineering, National University of Ireland, Galway,

Trinity College, Dublin Galway

ABSTRACT

This paper describes the development of an integrated mobility/emissions model (IMEM). The detail and structure of the model allow more reliable calculations of mobile source emissions based on appropriate spatial- and time-resolved consideration of vehicle activity within a traffic network. The model is designed to incorporate the finer spatial-resolved vehicle activity data (i.e. street by street) and time-resolved modal emissions data, (i.e. data that are directly related to the vehicle operating modes, such as steady-state cruising, queuing, accelerating and decelerating). As a test case, the model is being applied to Dublin City. The emissions produced on a given link or road can be displayed on a geographic information system (GIS) of the city and can be identified on the basis of transportation planning data. Any subsequent changes in emission levels due to alterations in the road network, can then be automatically assessed by creating different input data sets for the various scenarios to be evaluated.

INTRODUCTION

The atmosphere of the world’s cities is increasingly affected by air pollution. Hence, deteriorating air quality is fast becoming one of the most pressing and unrelenting environmental problems of our times. Poor air quality in large metropolitan areas is not a new concern, but the nature and type of air pollution has changed over the last number of years. In the past, pollutants were emitted primarily from the industrial and domestic sectors. Now, emissions from road traffic are the principal source of air pollutants in all large conurbations throughout the world. These emissions can have substantial impacts on the local environment (e.g. degradation of buildings and health effects). In order to assess these impacts, the quantity of emissions produced must first be determined. There are many methods of calculating emissions from traffic; one study alone identified 30 calculation methods in use in Europe [Negrenti, 1995]. Many of these methods estimate emissions using traffic data from mobility models. However, mobility and emission models tend to be developed separately and this leads to conceptual and data difficulties when their results are combined. To avoid these problems, an integrated mobility/emissions model has been developed.

The layout of the paper is as follows. First, the mobility component of the IMEM is first described in detail. Sample results pertaining to the mobility module are then given. The interface between the emissions and mobility components are discussed. A summary of the results from the IMEM is shown. Due to the fact that the emissions module is heavily dependent on motor vehicle statistics, a sensitivity analysis is carried out on the primary input parameters. Finally, the scope for future work in the project is outlined.

Mobility ModULE

The mobility model used in this study comprises of SATURN (Simulation and Assignment of Traffic in Urban Road Networks) [Van Vliet, 1995] and SATCHMO (SATURN Travel CHoice MOdel) [Hall, 1996]. The mobility model provides a detailed representation of the greater Dublin transportation system including junctions (e.g. priority, signal and roundabout) for both average weekday morning peak and afternoon inter-peak hours. The model distinguishes between three vehicle types, namely light duty vehicles (LDVs), heavy goods vehicles (HDVs) and buses. The period from 8 AM to 9 AM is assumed to be representative of the peak hour while that from 2 PM to 3 PM is assumed to be representative of the inter-peak hour [Steer Davies Gleave, 1993]. For modelling purposes the city is divided into a number of zones.

The mobility model is divided into four primary modules: trip generation and attraction, trip distribution, modal split, and assignment. The units in which these modules operate are trips. A trip can be defined as a one-way movement from an origin to a destination, for each of which the traveller has a purpose.

The trip generation module predicts the number of trips generated in a zone as a function of socio-economic, socio-demographic, and behavioural characteristics of that zone. These characteristics include total population, working population, school aged population, employment places, school places, total households and average salary. The generation model can be described as a zonally based local model, where ‘local’ implies that zones are addressed individually rather than being subject to uniform pre-determined growth rates. The trip generation predictions are grouped according to trip purpose with the minimum distinction being work, school, business and other (e.g. social and recreational). Non-home based trips are modelled separately.

The trip attraction module predicts the number of trips attracted to that zone. There are separate attraction models for separate travel purposes. For example, there is an attraction model for home to work trips as a function of zonal employment and an attraction model for shopping trips as a function of retail floor-space. The attraction model does not include transportation variables.

In the trip distribution module, the forecasted number of trips generated by each origin zone attracted to each destination zone is modelled.

The modal split module determines the mode of transport that is used. The functional form employed is the ‘availability’ or ‘non-availability’ of a car to distinct categories of individuals or market-segments. The categories with a car-available at the time of their trip have a choice of using the car or public transport (i.e. bus or rail). On the other hand, the categories with no-car-available at the time of trip are limited to public transport. The choice methodology is arranged in a hierarchical configuration. The actual modal split is determined by employing ‘logit’ formulation and the application of generalised ‘cost’ parameters. These cost parameters are calculated from research methods (e.g. stated preference surveys) and the relative costs of private and public transport. Private transport costs includes such variables as vehicle operating cost and parking charges, while public transport costs include fares, in-vehicle time, transfer time, walk time and waiting time.

The assignment module uses two sub-models to assign trips between all zone pairs to links, based on the minimum time (or cost, or combinations of time and cost) path between zones and also on capacity constraints. The first sub-model employs cyclic flow profiles to represent the movement of platoons of motor vehicles over a road network. This sub-model takes account of the interaction of different flows at roundabouts, signal controlled and priority junctions. It requires data concerning the flow on each link in the network to determine queues, capacities and time delays. A separate sub-model employs a ‘trip-matrix’ to determine these flows. The module also assigns passengers to public transport networks. Mode choice and assignment in congested urban networks are co-dependant.

There are two generic inputs for the mobility model, namely the ‘trip-matrix’ and the ‘network’, which may be regarded as ‘demand’ and ‘supply’ inputs. The trip-matrix specifies the number of trips from zone to zone and the ‘network’ specifies the geographical location and topography of the carriageways upon which the trips take place. The primary source of raw data for the trips-matrices is a comprehensive set of kerbside surveys and public transport origin/destination interviews (i.e. on-board and bus-stop/rail-station interviews). Obviously, it is next to impossible to generate trip-matrices to cover all of Dublin from the above methods. Therefore, traffic count surveys and air passenger traffic were used to assist in the derivation of ‘most-likely’ trip-matrices. The latter trip-matrices can be used to ensure that the assigned flows match the observed flows within certain pre-defined limits. Individual matrix cells represent the number of trips from one zone to another. Total trips to and from a specific zone for a given period of time are determined from trip matrix column and row totals. Separate matrices are created for two time periods (peak and inter-peak) and four modes of transport (car, HGV, bus and rail).

The physical network comprises all motorways, main roads, secondary roads, and the more commonly used minor streets. The junctions are represented by a series of uniquely numbered four-digit nodes while the segments of carriageways are represented by links. The links are defined in terms of the nodes at either end. Traffic enters a link at one node and exits at another node. A ‘turn’ is represented by a sequence of three node numbers: start, end and next. The network in Dublin is divided into two distinct parts – ‘inner’ and ‘outer’. The inner network has a relatively high number of junctions in comparison to carriageways while the outer network mainly contains road data.

Both the trip-matrices and road network are employed in a ‘route choice’ model which allocates trips to ‘routes’ throughout the network, as a result of which total volume flows along links may be summed and the corresponding network ‘costs’ calculated.

Figures 1 to 3 shows the relative distribution of peak and inter-peak hour velocity (km/hr), time delay (seconds), and average queue (pcu) in the greater Dublin area. Figure 4 displays the percentage of peak and inter-peak hour kilometres travelled relative to velocity for all links in the study area. The variation of vehicle kilometres travelled (VKT) for various velocities are characterised by the majority of kilometres travelled to have occurred at velocities less than 50 km/hr. A greater amount of time is required to travel trips at lower velocities. Therefore, a large number of vehicle hours are necessary to travel the number of vehicle kilometres shown in Figure 4 for velocities less than 50 km/hr. Few kilometres are travelled for velocities between 70 and 90 km/hr. However, an astonishingly large number of VKT occur at velocities greater than 90 km/hr. This is due to the large amount of traffic travelling on the ‘C-Ring’ and other motorways. Figure 5 displays the large number of hours spent travelling at low velocities. For example, during peak hour, forty-five percent of vehicle hours travelled (VHT) are at velocities of 10 km/hr or less. Figures 4 and 5 represent the possible inefficiency within the road network in the greater Dublin area. The transportation time of the vehicles travelling under congestion conditions has numerous costs. These include costs extra time spent travelling, additional vehicle operating costs, and the additional pollutants emitted at low velocities.

Mobility/Emissions Model Interface

The mobility model provides static and dynamic data for each road link in the greater Dublin area. Static data is passed only once to the emissions model, while dynamic data is updated depending on the simulation (peak or inter-peak). Shared parameters between the mobility and emissions models are shown in Table 1.