TRAFFIC AND ENVIRONMENTAL STREET NETWORK MODELLING: BELGRADE CASE STUDY

e paper presents a procedure for modelling transport demand and analyses e#ects caused by changes in the characteristics of street network exploitation. Based on the e#ects of the measures predominantly connected with speed restrictions on characteristic sections, the Scenarios resulting in reduced congestion during peak hour and a decrease in harmful e#ects of tra$c were selected. A rough e#ectivity analysis of transport vehicle kilometres travelled and travel time on the network was carried out. Transportation modeling is the most e$cient way to analyze the e#ects of tra$c solutions. Once established, transport model brings only bene%ts in future projects. Speed reduction strategy was easily simulated using the transportation modeling procedure. Other sustainable transport and mobility management strategies considering sustainable transportation could also be simulated.


Introduction
e rst step in solving the problems encountered in the transport system is to establish a unique database with the aim to give the planners a good initial basis (a model of the existing state) for the quanti cation of data relevant to the procedures of modelling, predicting and testing variants for the future development of the transport system.
In addition to tra c criteria, modelling may be performed on the basis of environmental criteria related to the emission of noise and hazardous gases into the atmosphere.
According to methodology de ned in Transportation Planning Laboratory (PLASA) on Tra c and Transport Engineering Faculty in Belgrade, a large scale research was conducted during the period 2004-2007. Transport Model of Belgrade was formed as a part of previous scienti c and research e orts.
is idea was carried out through the process of creating Belgrade Transport Model applying it in the analysis of transport demand on Belgrade tra c network. e conducted analysis was made as a part of planning documentation (Jovic et al. 2007(Jovic et al. , 2008). e transport model of Belgrade implies a set of relevant data (numeric, graphic etc.), indicators, parameters and simulation models expressed in space and time. e general application purpose of the Transport Model is to serve as an o cial basis for calculations to authorities or organizations the activities of which are directly linked to planning, programming, managing and building the transport system and as a basis for various research and scienti c projects, appraisal processes, analysis of tra c regime optimal solutions, intelligent transport systems in the portion of dynamic tra c control etc. (Jovic 2003;Jovic et al. 2007Jovic et al. , 2008.
A lot of scienti c and research projects resulted from Transport Model of Belgrade. Project results and collected experience denote the importance of tra c modeling and environmental calculation in procedures of tra c solutions evaluation. Tested solutions are consistent with the concept of sustainable transport. Case Study presented in this paper came as a result of hypotheses and solution scenarios generation according to infrastructure capabilities and expert experience.
Sustainability is not about threat analysis; sustainability is about systems analysis. Speci cally, it is about how environmental, economic, and social systems interact to their mutual advantage or disadvantage at various space-based scales of operation (Toward a Sustainable Future 1997; Urban Environmental Management … 2009).
Sustainability and sustainable transportation are di cult to measure directly, so various performance in-dicators are used to evaluate them. e focus can be on a single impact, such as air pollution emissions, or a wider range of e ects (Litman 2010a(Litman , 2010bOnline TDM Encyclopedia 2009).
As road transport is a major source of air pollutant emissions in European cities an e ciency assessment of possible future measures to reduce air pollution is required for future tra c planning, regulatory and scal initiatives (Lumbreras et al. 2008).
Tra c speed reductions have mixed air emission impacts, depending on tra c conditions, driving conditions, vehicle type and which emissions are considered. e speed dependency of the pollutant emissions using available emissions data and functions are known in the literature (André and Hammarström 2000;New Directions … 2009;Török 2009;Kuprys and Kugelevičius 2009;Kovács and Török 2010).
Transportation modeling is the only way to forecast the need for transport in future and the behavior of the system participants, as well as to plan actions for the implementation of the future scenarios. Also, modeling is the only economical and su ciently reliable way to carry out a forward assessment of the impact of the innovations to be applied on the overall system (Ušpalytė-Vitkūnienė et al. 2006;Tanczos and Torok 2008;Daunoras et al. 2008;Kinderytė-Poškienė and Sokolovskij 2008;Bazaras et al. 2008;Burinskienė and Rudzkienė 2009;Burinskienė 2009;Mačiulis et al. 2009;Szűcs 2009;Filipović et al. 2009;Paslawski 2009). is paper presents the procedure of transport demand modeling and environmental analysis of the effects caused by changes in the street network exploitation characteristics applied to a Case study. e aim of the paper was to present a transportation modeling procedure with environmental impact calculation in order to quantify the e ects of tra c solutions according to principles of sustainability and mobility management strategies. is procedure would enable estimation of tra c pollution in the future in di erent time periods (hour, day, year). e procedure like this, that combines tra c modeling and environmental calculation, was not applied in countries of South-East Europe.

Case Study
Belgrade, the biggest city on Balkan Peninsula, is the capital of Serbia with approximately 1.5 million inhabitants. It is located on Corridor X and represents a signi cant tra c node in the region (Smartplan Beograda 2001). e total length of the street network in the existing state is 1686 km. About 67% of the primary urban street network is with a single lane per direction. e network consists of 5900 links and 2398 nodes. e elements required to generate the state in the future time horizons are added through the future urban street network (street network of the year 2021).
Transport demands are de ned through a number of spatial distribution matrices determined on di erent time periods, purposes and modes of movement. e matrices used in this paper relate to passenger car trips during the morning peak period. In the existing state, the speci c matrix has 60000 vehicles, whereas the number of passenger cars in the forecasted year amounts to 79530 passenger cars during a peak hour. e most important projects in Belgrade include the completion of Belgrade Bypass and construction of one bridge over the Sava River (Fig. 1). e mutual goal of these two projects is to provide signi cant unloading of Belgrade street network, particularly the urban highway relocating transit ows of heavy freight vehicles to the Bypass rst and providing an alternative with one more bridge over the Sava River within the Inner Semi-Ring Road second. e urban highway in the territory of Belgrade is a part of Corridor X ( Fig. 1) (Jovic et al. 2006;Vuchic 1999).
Due to such development of Belgrade infrastructure, the need to model and analyse tra c load in the zone of alternative crossings (the Old Gazela bridge and the New bridge over the Sava River) with several solutions and to perform basic environmental analysis have arisen which would help in envisaging and quantifying the impact these tra c infrastructure facilities would have on their surrounding from the environmental point of view. e basis for this modelling is the Transport Model of Belgrade 2021 with the road network (with the values of section characteristics -capacity, free ow speed, rank, road pro le -for the year 2021). e subject of this analysis was a portion of 2021 network comprising the following tra c routes: -a part of the urban highway passing through Belgrade with the Old Gazela bridge; -a planned urban highway alternative, the New bridge over the Sava River and the tunnel under the Topčider Hill; -transverse connections between speci ed directions at the beginning and end of the observed area and central transverse connection across Belgrade Fair. By changing the values of free ow speeds in the tunnel and on the urban highway, 5 network Scenarios were created and then tested in the Model.
Speeds in di erent Scenarios (Table) are estimated according to a survey on the present street network and measures to be applied in the future tra c management system.

Scenarios
Free ow speed in tunnel V 0t Free ow speed on urban highway V 0hw ese Scenarios were created on the basis of a system modi cation Scenario and the real possibility of applying management measures in order to obtain the desired e ects (speeds).

Modelling Procedure
Modelling transport demands and environmental indicators was conducted in several steps: Step I. Coding Belgrade tra c network primarily refers to entering new sections planned on Belgrade trafc network for the target year 2021. is speci c area of particular signi cance covers the New Bridge over the Sava River and the tunnel under the Topčider Hill. It was essential to de ne capacities, free ow speeds, the number of lanes and the way of intersecting the existing road network. As regards the networks reserved for the movement of freight vehicles, it was necessary to prohibit the movement of freight vehicles on the portions of a lower-rank network.
Step II. e de nition of VD function includes the legality of speed variation (from free ow speed to operating speed) related to the tra c ow that is the same for all Scenarios and is provided in the form of VD function (volume delay function). BPR function (U.S. Bureau of Public Roads) was used as a basis for de ning VD function that can be expressed in the following form: With reference to data about travel times on Belgrade network and exit routes, 9 variants of BPR function (including various parameters a, b and c) corresponding to road/street categories de ned for Belgrade tra c network were formed (Vukanovic et al. 2006;Vukanovic 1997).
Step III. Network loading is loaded with travel matrices during the morning peak hour (from 07:00 to 08:00 hour) for passenger cars and freight vehicles in the forecasted state (year 2021). During a peak hour in 2021, the passenger car movement matrix has 79528 PA/h. It was obtained on the basis of the existing matrix (year 2005) and the predicted change in socio-economic indicators (primarily number of inhabitants). e existing matrix was obtained on the basis of tra c surveys (household interviews) conducted in Belgrade during 2005 and made 57444 PC/h. Socio-economic indicators were obtained on the basis of 2002 census. e factors by which the matrix was calibrated to the target year were obtained on the basis of population growth forecast and the predicted change in the level of motorization. In this process, di erent factors were de ned for central and peripheral zones in Belgrade. e freight vehicle trip matrix was produced calibrating the matrix of the existing state obtained by counting vehicles on the outer cordon and an interview conducted at freight terminals. e matrix containing 2589 FV/h was processed, whereas it was taken into account that Belgrade Bypass should be completed, which will considerably impact the structure of tra c ow on the urban highway and trunk roads.
Step IV. Selecting indicators of tra c network volume. Tra c ows on a speci c tra c network were analysed and then compared with the de ned capacities to obtain capacity utilizations per sections. In addition, a drop in speed due to congestion was indicated through the values of free ow speeds and operating speeds on the network.
Step V. Selecting environmental parameters (Fig. 2). e same parameters related to noise and pollution transformation were used for all Scenarios in the analysis.
All parameters o ered by so ware were considered as: -noise (dB) -level of noise generated by vehicles in tra c ow; -NO x (g/km) -level of nitrogen oxide; -SO 2 (g/km) -level of sulphur-dioxide; -CO (kg/km) -level of carbon-monoxide; -HC (g/km) -level of hydrocarbon.

Modelling Results
e most loaded sections in all Scenarios are the sections on Corridor X, to be more speci c, the sections on the Old Gazela Bridge over the Sava River. Numbers 1 and 3 in Fig. 3 denote interchanges at which maximum load occurs. ese sections, normally, sustain the greatest load because they are the points where the urban highway intersects intensive urban trunk routes. Marks C1 to C3 denote transversal connections. Interchanges and transversal connections are numbered according to their signi cance. In the following text above, the presented marks will be used for the convenience of reference.
Tra c planning methodology presumes that intersections are treated as points. erefore, for tra c planning needs, when coding the network, it is necessary to simplify grade-separated interchanges which is the main reason for such high capacity utilisation values on interchange sections and that is why the values on the sections with the real tra c ow values are given the central position.
Scenario 1: e most loaded section on the network in Scenario S1 is the Old Gazela bridge section (Q c = 5.769PA) in the direction towards the city centre the ow of which exceeds the capacity (q/C = 111%) and the section downstream of the Mostar Interchange (mark 1 in Fig. 3) which is also congested (q/C = 125%). An increase in speed in the tunnel to 80 km/h has slightly relieved the Old Gazela Bridge compared to Scenario S2. Most of the urban highway sections are in a congested state as a result of urban highway attractiveness.
Scenario 2: e most loaded section on the network in Scenario S2 is the same section on the Old Gazela bridge (Q c = 5.778PC and q/C = 111%) and the section downstream of the Mostar Interchange which is also congested (q/ C=128%). A slight increase in the ow is resulting from the de ned free ow speed in the tunnel of 60 km/h which further diminishes the competitiveness of an alternative route. Irrespectively of such a considerable change in free ow speed in the tunnel, a change in the ow is rather small which indicates small sensitivity to change speed on the alternative route.
Scenario 3: e most loaded sections are the same as with the previous Scenarios and include the Old Gazela bridge (Q c = 5.146PC and q/C = 100%) and the section downstream of the Mostar Interchange (q/C = 112%). As a consequence of decreased free ow speed, the number of vehicles in speci c sections is smaller and the travellers more easily choose their alternative routes. is Scenario is operating at the limit of urban highway capacity. Although alternative routes provide satisfactory values of capacity utilization, it should be taken into account that higher ow was recorded on other bridges outside the zone of interest as a consequence of speeds de ned in this way (Bress Paradox) (Vukanovic 1997). e trap of Bress Paradox was avoided by a screen line analysis of the crossing over the Sava River.
Scenario 4: e same most loaded sections with the highest values of capacity utilization, the Old Gazela bridge (q/C = 113%) and the section downstream of the Mostar Interchange (q/C = 131%), are a consequence of even greater di erence between free ow speed in the tunnel (V 0 = 50) and free ow speed on the urban highway (V 0 = 80).
Scenario 5: e Old Gazela bridge section has boundary capacity utilization (Q c = 5.023PC and q/C = 97%) while the section downstream of the Mostar Interchange is slightly congested (q/C = 103%). By reducing free ow speed on the urban highway, the alternative over the New Sava bridge gains added signi cance and load distribution between alternative routes is the most favourable one.
A graphic presentation of tra c load and derived indicators is made for each Scenario like in Figs 4 and 5. Such presentation is very clear for those decision makers who are not experts in the eld of tra c.

Tra c load analysis
In Scenarios 1, 2 and 4 with free ow speed on the urban highway of V 0 = 80 km/h, the attractiveness of the urban highway is great. Hence, it dominantly in u- Fig. 3. Scope of analysis ences load distribution and attracts a great number of vehicles. e alternative route over the New Sava bridge introduced in this Scenario represents no valid alternative to the urban highway where ow exceeds its capacity, thus entering into the congestion state because of which exponential speed decreases from free ow speed to 35 km/h which is a considerable drop in speed and the greatest drop in the Scenario with the lowest free ow speed on the alternative route (50 km/h).
As regards freight vehicles, route selection cannot be in uenced because of an e ective decision regulating the routes on which the movement of freight vehicles is permitted. In this way, no great changes in freight tra c load occur. erefore, in the Scenarios, these loads can be treated as constant loads.
As a critical point on the separated network there appears the section connecting alternative routes (section C3 in Fig. 3). It is the section that leads tra c coming from the peripheral areas of the City into the city centre. is section considerably in uences speed on the alternative route over the new bridge and is detected as the network bottleneck.
When changing speeds in the tunnel, load on the transversal connection (section C1 in Fig. 3) does not change considerably. During a peak hour, this connection is at the limit of its capacity, and hence is unable to attract tra c.
e Scenarios predicting lower free ow speed on the urban highway (3 and 5) are favouring the use of the New Sava bridge. In this way, a balanced ow diagram on the entire observed network is obtained. Scenario 3 with equal speeds (60 on the urban highway and in the tunnel) only in uences a decrease in load on both routes, whereas relations (proportionally) remained unchanged as in the Scenario with a speed of 80 km/h.
In Scenario 5 with speed on the urban highway 60 km/h and 80 km/h in the tunnel, volume to capacity ratio on the urban highway is close to 1. On the other hand, volume to capacity ratio on the New Sava Bridge and in the tunnel increases. e basic idea in Scenarios 3 and 5 is a reduction in free ow speed that depends on the pro le. In order to achieve a free ow speed of 60 km/h, it is necessary to introduce speed restriction on the critical part of the urban highway. As this measure itself is not su cient, a better treatment for on-ramps on the urban highway is proposed. For example, during a peak hour, the ramp could have a highway lane reserved for merging which should surely reduce free ow speed because urban highway pro le should be reduced from the previous three lanes to two lanes. An interim solution is possible in which the use of the third lane would be allowed only for merging vehicles from the previous ramp as well.
Based on load analysis, Scenario 5 with a speed of 60 km/h on the urban highway and 80 km/h in the tunnel, was selected for further analysis. A comparison with Scenario 1 was made to quantify advantages and disadvantages. Such choice was made because the urban highway is the most important route in the City, particularly the Old Gazela Bridge de ned as Corridor X bottleneck on the maps of world logistics. e intention was to show the di erence in load as a result of changes in free ow speed on the urban highway.
Based on load analysis, data on noise level and air pollution parameters (NO x , CO, SO 2 and HC) can be also obtained. Data on the environmental e ects of these Scenarios can be compared with reference to which the generated Scenarios can be relatively ranked. Fig. 6 provides a group presentation of environmental indicators which may be generated from the models o ered by so ware. Relations contained in the models are mainly obtained employing empirical methods (PTV).
To be graphically presented, each environmental indicator may be independently treated, as shown in Fig. 7.
e economic e ects of the proposed solutions on the selected Scenarios S1 and S5 were considered through data on vehicle-kilometres and vehicle-hours (Transportation System Performance 2007).
In Scenario 1, there were 71182 vehicle-km and in Scenario 5 it made 68746 vehicle-km. Vehicle-km were reduced by 2436 vehicle-km during a peak hour which represents a decrease of 3.4% which is a consequence of shortening routes where vehicles move across the network accordingly to a change in speeds.
On the other hand, total travel time expressed in vehicle-hours in Scenario 1 is 1.630 and in Scenario 5 it makes 1.654. Travel time increases by approximately 24 vehicle-hours which represents an increase of about 1.5%. Travel time also increases as a consequence of a change in travel speed. Better load distribution is achieved on the network together with more balanced utilisation of capacity but lower travel speeds.

Conclusions
1. Tra c planning cannot be treated as an objective but is rather shaping the transport system so as to become compliant with the principles of sustainability in order to derive optimal functions from the system with de ned physical characteristics and having minimum utilization of resources in the operation phase. Energy e ciency at the time of energy shortage plays an important role for tra c and the list of tra c infrastructure priority (Saunders et al. 2008). New technical solutions and transport systems enable many options in the development of the system having to accommodate growing transport demands. 2. Transportation modelling is the most e cient way to analyze the e ects of tra c solutions. e bene ts of the approach are the possibility of working with a large number of developed scenarios and fast scenario modi cation producing and testing the new ones. 3. Once established, a transport model brings not only bene ts to the future projects, considering more economical project production and higher quality but also requires constant maintenance, calibration and validation expressing changes in the transport system. 4. Speed reduction strategy was easily simulated using the transportation modelling procedure the results of which were used to present clear and understandable output displaying graphics and tables. 5. Other sustainable transport and mobility management strategies considering sustainable transportation could also be simulated. 6. From the point of view of tra c, the best solution may not be optimal from an ecological point of view (Jovic and Djoric, 2009). 7. A lot has been achieved in Belgrade in the last few years. Several capital projects have been initiated and the most interesting among them are the City Bypass (about 50% of the project completed), the New Sava bridge (the nal phase, will be completed in 36 months) and the Inner Semi-Ring Road (preparatory phase). Extensive transport demand analyses considering various time horizons were made on all Projects using the Transport Model of Belgrade. 8. e previous analyses and assessments of the e ects of planning measures provide considerable savings related to the 'blind' implementation of measures. By modelling transport demands and transport system supply, it is possible to quantify the impact of the envisaged actions and loading of routes as well as a great number of derived tra c e ects (Djoric 2007(Djoric , 2008. 9. e e ects of time or a spatial dislocation of the problem are also avoided solving some speci c problems Comparative graphical presentation of NO x selected Scenarios this way. A problem of identical or higher intensity may appear in other period of time or on the other part of the tra c network which most easily can be perceived on the graphical presentations of the network. 10. e Transport Model of Belgrade was made applying so ware package VISION in the part of the program for tra c planning and macro-simulation (VISUM). e model is iteratively improved inserting data on a higher level of detail. e program is GIS-based, thus enabling referencing and ltering a great number of data. However, the basic advantage lies in the possibility of overlapping various transport systems, tra c modelling of the system and the events occurring in the systems. 11. By aggregating all data within the model, it is possible to create a complete database representing a basis for all analyses to be performed in the future which is vitally important because this way, the possibility of input data manipulation is eliminated. 12. us, the created model enables the presentation of tra c system load in a su ciently precise manner. A great number of system Scenarios are created simply by changing relevant transport system parameters. It is easy to perform the quanti cation of the obtained output indicators (both tra c and environmental) that are su cient indicators of the quality of the applied planning measures and changes on the network. e output from macro-simulation represents input data for micro simulation which processes the obtained results at a higher level of detail. 13. A graphic and tabular presentation of data provides a possibility of presenting all information from the database in the form enabling the demonstration of the obtained results in the most simplest and precise way. It is very important to provide a uniform presentation of the results for decision makers who are not very o en experts in the subject matter because development policy is going to be created on the basis of such results. Taking into account the absolute values of this data, it is used to present the measure of the realized improvements and worsening, whereas on the other hand, it serves as input into the process of evaluating tra c solutions. e possibility of presenting and quantifying environmental e ects is emphasised as an important segment.