SIMTRAP
Simulation of Traffic Induced Air Pollution
Using Parallel Computing in a Distributed Network

Keywords:Urban air quality modelling, traffic management, decision support system
Application site: Berlin, De; Maastricht, NL; Milan, I; Vienna, A
Internet: http://www.web.system.ptv.de or http://www.ess.co.at/simtrap

Context
Authorities in metropolitan areas throughout the world require support in managing and controlling road traffic to avoid exceeding pollution limits set by national and international governing bodies. For developing medium term strategies, t h e demands on integration between the traffic model and the air quality model are highest, whilst the time scale allows for more detail and dynamics in each of the components. Examples of such strategies at the urban level are demand control during adverse meteorological conditions (e. g . the odd / eve n number plate mechanism used in Paris in Summer 1997), the planning of signing and guidance during special events (City centre closures, sports events, but also medium term bridge or sewer repairs). Integrated dynamic tools are required to quantify the effects of alternative strategies, and to aid in choosing the most appropriate one.

Objectives
The project addressed primarily the re l a t i o n between traffic and ozone, because (1) there are particularly severe acceptance problems with traffic regulation measures, since the effect on ozone concentration is often counterintuitive – hence the need for an objective decision basis, and (2) the computational challenge is bigger than for other pollutants due to the necessary large study area and complex photochemistry.

Results
An integrated system for traffic flow simulation, air pollution modelling and decision support is developed in a distributed HPC (High Performance Computing) network. It builds around two well-established core components: the meso-scopic dynamic traffic simulation tool DYNEMO (Schwerdtfeger, 1984) and the 3D air quality model DYMOS (Sydow, 1994). The integration of both modules occurs in a remote HPCN environment – a parallel computer will enable the detailed simulation of an area of sufficient geographical extent (approx.100x100km).The interpretation and visualisation of results takes place in a local 3D GIS system. Communication is achieved by using existing computer networks and protocols.

Determination of standard interfaces for the communication between the three systems helps to run the various models on different HPC hardware platforms at different locations. Additional or substitutive models can easily be incorporated using the defined standard interfaces. The integrated simulation tool is meant as an optimisation and decision support tool for practitioners (transport planners and traffic engineers). It links planning options on the transportation side (changes to the infrastructure, changes to technology, changes to demand) to effects on key pollutants. Results are p resented as maps of spatially and temporally resolved pollutant concentrations which may be compared. There is limited support for scenarios for other emission sources (industry, household, land use patterns) so that their potential for air pollution improvement can be compared to the influence of traffic.

Demonstration of three test sites are listed below:

City of Milan, Italy:
The model applied on Milan municipal area, considers the air quality monitoring network of the Province of Milan with 52 monitoring sites, 10 of them located in Milan city.

The system was applied by the "real" end-users and not by the developers.

City of Berlin, Germany:
The Berlin network is used as a test-site during the development and the integration of the system components, as well as in providing data for different control measures to assess different traffic scenarios. Among the scenarios studied are gating strategies for the city centre and the impact of a proposed traffic management scheme for the area around the new exposition centre, as well as general trends in emission technology.

City of Vienna, Austria:
The system is applied to simulate the effect of different dynamic network control strategies (dynamic driver information or dynamic route guidance) on the pollution situation. Accuracy of the component simulation models was validated several times before the start of the project. As these measuring campaigns are very costly, these tests will not be repeated within the project. In fact, this was one of the compelling reasons for reusing existing traffic and air pollution models.

Technical characteristics
The basic architectural decision in SIMTRAP is therefore to separate user interface and visualisation from simulation and specify a flexible communication structure that can be implemented using a variety of protocols.

The SIMTRAP Client can be any of a wide variety of machines, as long as they support LAN/WAN connections and are capable of running a GIS system. For the demonstrator applications within the SIMTRAP project the prototype SIMTRAP Server will be implemented on a SUN Sparc workstation under UNIX (Solaris 2.5 or higher). The separation into a SIMTRAP client (graphical user interface, decision support system) and a MODEL server (simulation) is depicted in the figure below.

The SIMTRAP system provides an overall integration including a communication with the model server(s) and with the monitoring system (optional), and a management of all model input data. The interactive user interface, includes GIS and visualisation functions, model scenario editing and management and DSS functionality.

The best performance of the SIMTRAP SERVER is obtained with a UNIX workstation with a high-resolution graphics system (1280 by 1024), 256 simultaneous colours, (i.e., an 8 bit graphics frame buffer) with 64 MB RAM and, depending on the application,1 GB of (data) disk space or more. However, the operational use of the system foresees also the application on a Windows NT, or be even on the same hardware platform as the MODEL Server. These use a Java based client-interface that can be run from any Java-enabled browser program on a PC or Network computer (NC).

As Operating system the SIMTRAP SERVER is supported on SUN Sparc architectures, under SUN Solaris 2.4 or higher. Support for alternative architectures and operating systems (HP UX, IBM AIX, Intel Linux) can be made. The host of the SIMTRAP SERVER requires a local LAN connection or Internet access for some of the systems intended functionality, and in particular, the connection to the parallel HPCN models.To access remote computing and data resources, a minimum of 64Kb/s is required (e.g., through an ISDN phone line). The model server is going to be implemented either on a dedicated parallel machine at the GMD, or on a UNIX workstation cluster under MPI.

As external tools for data preparation, editing etc, the following tools are recommended: basic screen editor like vi, graphical editors and image conversion programs, including xv, xpaint,giftrans, a standard GIS for geographical data capture and editing, such as Arc/Info, IDRISI or GRASS.

The SIMTRAP MODEL Server
The parallel versions of the DYNEMO and DYMOS programs run on all hardware platforms. Highly parallel system e.g. Intel paragon, Cray T3D, Thinking Machines CM5, Parsytec Power GC or CC, IBM SP2, iPSC/860 are recommended for highest and fastest performances. However, also workstation clusters can be used e.g. UNIX platforms interconnected by a variety of networks (Ethernet, Token ring, FDDI,...) with TCP/IP. At least 2 computers with 64 MByte RAM each are necessary.

Producer	Model	Operating system
SUN	Sparc	Solaris1 (SunOS)
SUN	Sparc	Solaris2
INTEL	Intel x86	Linux,386BSD, NetBSD, BS DI
HP	PaRisc	HP-UX
IBM	RS/6000	AIX
SGI	R4000 Iris and up	Iris
DEC	Alpha	Digital Unix

PC Network
An alternative to UNIX workstations clusters are networks of the following WINDOWS platforms interconnected via TCP/IP. At least 2 computers with 64 MByte RAM each are necessary.

Producer	Model	Operating system
Intel	Intel x86	Window NT
Intel	Intel x86	Window 95
DEC	Alpha AXP	Window NT

Heterogeneous Networks
In addition to the above options, the SIMTRAP MODEL Server can be operational on heterogeneous networks of different platforms out of all listed computers above. In addition to the hardware requirements a FORTRAN and C compiler has to be available on the chosen platform. PVM3 (Parallel Virtual Machine) is used as the underlying message passing system. This tool allows a heterogeneous collection of workstations and supercomputers to function as a single high performance parallel machine.

Transferability
General feedback on the system from these initial applications was encouraging, and therefore the developer partners have taken steps to increase their marketing activities for the TRAQS product derived from SIMTRAP. TRAQS will build on the results of this project, yet add support for further models and will be available for a Windows platform which, compared to the current Unix platform, will remove a significant barrier. During an initial mailing action late in 1998,a number of seriously interested potential customers was identified. It is an important challenge to find an organisation for the consortium that will enable further marketing beyond the project end.

SIMTRAP is a tool for integrated transportation and environmental planning that can suit any region or city where traffic contributes significantly to air pollution.

To implement the project the users should gather model input data, install the SIMTRAP software, define scenarios, simulate, and interpret results. The costs involved are the cost of the SIMTRAP software (varying with configuration, starting at approx. 25000 ECU), suitable hardware (varying with configuration, starting at approx. 10000 ECU) and the cost of the data gathering (estimated around3 to 6 person-months).

The transferability in a broader sense, i. e. technology transfer to other applications, is summarised in the following points.

• SIMTRAP contains two reusable elements: the design, which matches computational resources to computational needs and the component simulation models. These elements are relevant outside of its current scope of implementation
• The models could be of interest to transportation and environmental planners, even if they do NOT follow an integrated approach. An example is the traffic flow model which can be used as a test bench for future telematics services (without re g a rd for air pollution)
• Although the project team currently feels that an in-house solution (end-user owns SIMTRAP server and client) would be preferred by the majority of customers, there may be scope for service providers offering computing time on a remotely accessed server. The potential service providers could form a target group outside authorities and consultants
• The exploitation plan for SIMTRAP contains options for supporting other applications through additional models/configurations. Whether these options are implemented depends on market response during the remainder of the project's lifetime. If implemented, these options will be available to cities/regions at a cost comparable to the original SIMTRAP product
• Re-use of the design pattern and communication technology offers an indirect benefit for customers, lowering the development cost of future dedicated applications
• Two of the developers (ESS, PTV) will offer complementing consultancy services for data gathering and setting up the system. This will also apply to the additional options mentioned. Development of custom-built systems for applications outside the scope of the SIMTRAP product, but incorporating SIMTRAP technology could be considered
• If the technology is reused in a new system, it will lower the development cost, but the bulk of the application will still have to be developed anew
• Parts of the user interface, e.g. the scenario management, which is tailored to the set of scenario parameters used in SIMTRAP are specific for the application.

---

 
Do you have a comment, reaction or opinion regarding what you've read? If so, please leave it at the online discussion board!

---

REC * ENWAP * VIRTUAL GALLERY

The Regional Environmental Center for Central and Eastern Europe