51

A GIS-based Air Pollution Modeling in Tehran PARANG S.(1) & SOURI A.H.(2) (1)

Geodesy MSc student, Dept. of Surveying and Geomatic Eng., College of Eng. University of Tehran, Tehran, Iran, [email protected] (2) Remote Sensing MSc student, Dept. of Surveying and Geomatic Eng., College of Eng. University of Tehran, Tehran, Iran, [email protected] Abstract. The population growth and development of the mega cities and the impacts on urban traffic is one of the most important problems of the mega cities. Increased traffic volume and air pollution lead to population health problem. In this research a prediction model has been proposed for air pollution prediction in 2004 using the data of 2002 and 2003 comparing the prediction results with the actual results of 2004. In addition, by using the method of local contribution to concentration in canyon streets, the concentration of both CO and NO at each month for six highways of Tehran and for each vehicle is calculated. The prediction model is a combination of CORINIER and Gualtieri-Tartaglia models. The proposal GIS-based model employs street geometry and vehicle numbers. Operation of CO and NO models Shows accuracy of 90% and 60%, accordingly. The evaluation done in this article demonstrates that Gualtieri-Tartaglia model. The results are appropriate for a one year prediction of CO, whereas for NO it is not appropriate and the innovation in this paper is that the results of the previous modeling (Gualtieri-Tartaglia) results are valid for about one day, however, in this paper by improving the model of Gualtieri-Tartaglia integrated with the CORINIER method, the model can be applied to estimate the air pollution within one year. The implemented data in this paper include the average monthly values of NO and CO in the second half-year of 2002 and whole years of 2003 and 2004.CO data used during 2005-2010 provided the similar results. Keywords: GIS, Air Pollution, Traffic, Climate, Canyon Streets, Modeling, Prediction Model. 1. Introduction. Air pollution is one of the most important problems of cities .Unfortunately in Tehran a few reaserchers focus on this problem. In this paper the emission model of two gases including CO and NO is discussed so far. Air pollution from motor vehicles is one of the most serious and rapidly growing problem in Tehran.The modelling provides the ability to assess the current and future air quality in order to enable informed policy decision making.Thus, air quality models play an important role in providing sufficent information for better and more efficient air quality management planning. The main purpose of this article is to predict pollution measurement caused by transport vehicle at 6 main highway in Tehran. For the prediction model average values of pollution measurement stations have been used.The model used for calculating the overall pollution at 6 highways in Tehran is CORINIER model which it’s outputs will be used as the inputs of Gualtieri and Tartaglia model.Gualtieri-Tartaglia developed and validate an urban street canyon model based on carbon-monoxide experimental data measured in Firenze. The canyon street models usually concentrate on inert pollutant whose dispersion process exclude chemical and photochemical reactions. Among the various techniques for simulating dispersion of pollutants due to road vehicles in street canyons, the semiempirical approach appears to be the most convenient one for practical use (Reference [8]). Street canyon model is usually focused on pollutants, whose dispersion processes do not involve chemical or photochemical reactions. The main objective of this paper is modelling CO and NO in 2002,2003 and 2004 and predicting CO and NO pollution in 2004 according to 2002 and 2003 data and comparing the prediction model with the actual data. The simulation model presented in this paper for evaluation of local air pollution consists of an air quality prediction. A Geospatial information system(GIS) is a computer-based tool related to mapping and analyzing spatialy distributed phenomena related to the earth. GIS technologyintegrates common databaseoperations with the unique vizualization. GIS have been set up according to diagram presented in Figure1. The advantage of GIS according to other information systems is the high power of analysing of spatial data and handling the large spatial databases. In air pollution there are a large amount of data that GIS can be usefull in handle them. Data that is used for modeling of air pollution are wind direction,wind speed, traffic flow, solar radiation , air temperature and mixing height. This method is based on methodology that developed by CORINAIR in order to calculate the emission/consumption factors from road traffic. However, other methods which based on emission/consumption factors, can simply be compatible with tools. According to CORINAIR method and emission/consumption factors per vehicle and per driving distance unit related to vehicle category and average driving speed.vehicles are categorized by vehicle category(passenger, car, bus, truck,...) engine type,motor size,fuel type. The highways that are studied in this research are Hemmat, Hakim, Jenah, Modarres, Zeinaldin and Basij. The programm and softwares used are Matlab and Arc-GIS.

International Conference of GIS-Users, Taza GIS-Days, May 23-24, 2012 Proceeding Book

52

Fig. 1. GIS Structure for Vehicular Pollution Modelling (Reference [10]). 2. Developing the problem. Among the various technics that exist for simulating the pollutant dispersion caused by road vehicles in canyon street, semi-empirical method seems the simplest way for practical use. In fact,this method allows the prediction dispersion of pollutant moving from the the knowledge of some easy-to-measure parameters such as traffic flow or average speed, fleet composition, wind speed, and site geometric features. Resulting model, often as street canyon models are able to forecast pollutant concentrations based on the simplified scheme of local isolated air circulation patterns that take place in the road sections surrounded by the buildings.coefficient of these types of models usually needed to consider based on local site’s charachtristics. In this article for evaluating the success percentage of agreement has been used. In fact this index shows that how well predicted values are error free (as seen in equation (1)).

d = 1−

N 2 ∑ (P − O ) i i i =1 N 2 ∑ [P −P + O −O] i i i =1

In which : Pi : predicted values Oi : values of observations : average of observations : average of predicted values

(1) In this equation, value of d is demonstrator of proximity value of predicted data to observations. This means that the more value of d is closer to 1 the better prediction model will be and if the value of d tend to zero, prediction will be inefficient. So as the above equation shows the value of d is a number between zero and one. According to equation if the predicted value is put to equal value of observations, value of d would be equal to zero. this means that if the value of prediction is only considered as the average of observations without any modeling, value of d would be equal to zero. So, the more value of d tends to 1, it can be said that the prediction model will better than predicted model with the help of average values accordingly. In order to estimate NO and CO emission, a suitable application of CORINAIR emission models have been done. The steps of the the modeling process is as follows: The first step is to calculate the stack's volumetric gas flow rate ( ) using Equation (2). (Reference [1])

. V(m 3 /min) = Gas Velocity (m/s) ×  π × (diameter of dipersion(m)) 2 /4 × 60(Sec/min)   Where

(2)

π = 3.1416

We then modified to correct the stack gas flow rate for the moisture content and standard conditions using Equation (3): (Reference [1])

. . 273.15 K Pactual Dry V (m 3 /min) = V (m 3 /min) × × × (1 − humidity) T 1atm actual Colloque International des Utilisateurs de SIG, Taza GIS-Days, 23-24 Mai 2012 Recueil de Proceeding

(3)

53

Where PActual: Actual Pressure and TActual: Actual Temperature Equations relevant to the emission/consumption and percentage of each vehicle category in comparision to all vehicle (extracted from statistical surveys) are stored in GIS database. This data, with the average speed of vehicles and traffic flow patterns supplied by the demand traffic model are used as inputs of models to forecast air pollution-emission and energy consumption. Results not only include total emission level of each pollutant and total consumption of each fuel type,but also inculde average by vehicle and by link. The next step is to convert PPM (volume) to mass emission rate (kg/h) as follows: (Reference [1])

. MW substance Er(kg h) = ppm × density of air at standard condition × Dry V× 60(min/h) × MW air

(4)

Where MWsubstance: Mass volume of Substance and MWair: Mass volume of Air Emission factor is a value that represented to relate the quantity of released pollutant and the activity that associated with the release of pollutant.emission factor is ratio between amount of pollutant per throughput of materials.(for example:pounds of NO per gallon of residual oil burned) .emission factors are founded on the assume that there is a linear relationship between the emission of air pollution and the activity level. Diffrent types of sources can use emission factors to estimate their emissions. The emission factor Ef is calculated using Equation (5): (Reference [1])

 Er  g hour   Ef =  Fuel feed rate  kg  hour 

(5)

The emission factor Ef due to road vehicles belonging to group g is expressed as the mass of pollutant per unit length as a function of the average travel speed Vm . Total pollutant emission, Q, produced by the traffic flow, f of N vehicular groups is computed using Equation (6): (Reference [10])

N cg Q= ∑ × Ef(Vm ) × f g =1 100

(6)

Where cg is the percentage of vehicular group g with respect to the vehicle fleet. Concentration is calculated by adding local and areal concentration. In this paper our objective is modeling the local concentration which is more effective than areal concentration. Local Contribution to concentration is calculated by different approaches i.e. Zannetti, (Reference [9]), DePaul and Sheih (Reference [5]). The model developed by Gualtieri-Tartaglia developed is more useful for calculating the local contribution to concentration because of considering more elements of meteo climatic variables. The climatic data period was between July 2002 and December 2004 for both CO and NO. This model is based on the calibration process and performed for CO and NO. The street canyon model used is based is based on Equation (7): (Reference [10])

 Q  C = a  × F + bT + cH + ∑ d i Rad i + k 0 l  U + 0.5  i

(7)

3

Where: Cl (µg/m ): modeled CO or NO2 concentration; Q (g/ms): mean emission rate; F (m-1): shape factor (FW, FL, FI), depending on the specific sector (Table.1); U (m/s): air wind speed; T (C): air temperature; H (m): mixing height; 2 ∑ Radi (W/m ): solar radiation; a, b, c, d, k0 : model linear coefficients to be calibrated. QS is calculated according to Equation (6). F is calculated with respect to Table 1. U, T, H, ∑ Radi are accessed from Iran Meteorology Organization. a, b, c, d, k0 are the calibration coefficients that are discussed in conclusion item. Shape factors values F as a function of sector geometrical features are shown in Table1

International Conference of GIS-Users, Taza GIS-Days, May 23-24, 2012 Proceeding Book

54

Sector

Shape factor

Windward

FW

Leeward

-1

Value(m )

7×

FL

h*w

1

7× x

Intermediate

h−z

2

+z

2

+ L0

(FW + FL )/2

FI

h: buildings mean height; w: street canyon width; x: distance of the receptor from street axis; z: height of the receptor; L0: Vehicles mean width, generally assumed to be 2; The flowchart of the modeling process is shown in Figure 2. Table 1.Shape factors values F as a function of

sector geometrical features. 3. Conclusions. As mentioned before the available data for 2002 was in second half of the year but in 2003 and 2004 we use whole of year data, so as can been seen the results of the calibration coefficients for CO in 2002 are different from the results in 2003,2004. So it is concluded that the canyon model is efficient when complete data through out year is used.The diagrams of the changing in calibration coefficients for taxi vehicle in 2002,2003,2004 are depicted in Figures 3, 4, 5, 6, 7, 8, 9, 10 and 11. Mean Speed of the

Volumetric Gas Flow

The Stack Gas Flow Rate fortheMoisture

Convert From PPM to Mass Emission

Calculation of Emission

Number of Vehicles Calculating Total Pollutant Emission Traffic Flow

Mean Emission

Meteo Climatic

Concentration

Geometric Data

Model of Air

Calculation of Calibration Modeling Air

Prediction

Fig. 2. Flowchart of the modeling process

Colloque International des Utilisateurs de SIG, Taza GIS-Days, 23-24 Mai 2012 Recueil de Proceeding

55

Fig. 3. Calibration coefficients for taxi in 2002 with the windward direction.

Fig.4 .Calibration coefficients for taxi in 2002 with the leeward direction.

Fig. 5 .Calibration coefficients for taxi in 2002 with the intermediate direction.

Fig. 6 .Calibration coefficients for taxi in 2003 with the windward direction.

Fig. 7 .Calibration coeffiecients for taxi in 2003 with the leeward direction.

Fig. 8 .Calibration coeffiecients for taxi in2003 with the intermediate direction.

Fig. 9 .Calibration coefficients for taxi in 2004 with the windward direction.

Fig. 10 .Calibration coefficients for taxi in2004 with the leeward direction.

International Conference of GIS-Users, Taza GIS-Days, May 23-24, 2012 Proceeding Book

56

Fig. 11 .Calibration coefficients for taxi in2004 with the intermediate direction.

Considering the results, the differences between the wind sector in 2003 and 2004 coefficients are negligible, of course if data of pervious years added, results may improve. For example the calibration coefficients for taxi in year 2004 and 2003 in windward direction approximately are the same. In the following the results of using 2002 and 2003 data for estimation and prediction of pollutant in 2004 are depicted. The following graph shows the actual data of 2002 and the estimated values of 2002 are shown using 2002 data. The horizontal axis is depicting six months for the six highways in Tehran and the vertical axis depicts the concentration of CO for each month and for each highway.please refers to Figure 12 and Figure 13. After using 2002 data for prediction of 2003 air pollution, in this section we use 2003 actual data which is for second half of the year, for predicting 2004 concentration. As Figure depicts, the results are acceptable and the difference is negligible. Please refer to Figure 14 and Figure 15.

Fig. 12 .Estimate of 2002 with 2002 coefficients for CO d = 97%.

Fig. 13 . Estimate of 2003 with 2003 coefficients for CO d =97%.

Colloque International des Utilisateurs de SIG, Taza GIS-Days, 23-24 Mai 2012 Recueil de Proceeding

57

Fig. 14 .Predict CO 2004 with 2003coefficients d=90%.

Fig. 15 .predict NO 2004 with 2003 coefficients d=65%. According to the yielded graphs, it become crystal clear that Gualtieri-Tartaglia model is also appropriate for a long term of prediction gas CO. The predicted data that are the outputs of Gualtieri-Tartaglia are entering to GIS after the analysis and investigation. GIS has been used in this research to analyze and depict emission and dispersion model of CO in six highways in Tehran. Innovation of research Gualtieri-Tartaglia model previously used for predict of gas CO, NO was valid for maximum term of one year. Where as in this article by developing this model and CORINIER model we could validate the prediction for one year.please refer to Figure 16 and Figure 17.

International Conference of GIS-Users, Taza GIS-Days, May 23-24, 2012 Proceeding Book

58

Fig. 16 . An example of pollutant emission mapping due to all traffic volumes in Tehran. Case study CO for January 2004 emission rate.

Fig. 17. An example of pollutant dispersion mapping in Tehran. Case study concerning CO concentrations in January 2004. 4. Acknowledgements This work is supported by the University of Tehran, College of Engineering, Department of Surveying and Geomatic Eng and by Iran Meteorological Organization,.We would like to thank Dr. Delavar for his support and giving us valuable advice. References [1] Environment Canada's National Pollutant Release Inventory (NPRI). [2] Bellasio, R.G. Lanzani, M. Tamponi and T. Tirabassi. Boundary layer parametrization for atmospheric diffusion models by meteorological measurements at ground level. IlNuovoCimento (1994). [3] Van Mierlo J., Simulation software for comparison and design of electric, hybrid electric and internal combustion vehicles with respect to energy, emissions and performances, PhD thesis, VrijeUniversiteitBrussel (2000). [4] Ntziachristos L., Samaras Z., Copert III: computer programme to calculate emissions from road transport, European Environment Agency (2000).

Colloque International des Utilisateurs de SIG, Taza GIS-Days, 23-24 Mai 2012 Recueil de Proceeding

59

[5] [6] [7] [8]

[9] [10] [11] [12] [13] [14]

[15] [16] [17]

DePaul, F.T. and C.M. Sheih. Measurements of wind velocities in a street canyon. Atmospheric Environment (1986). Florian, M. and S. Nguyen. An application and validation of equilibrium trip assignment methods.Transportation Science (1976). Hoydysh W. G., Dabberdt W. F., Kinematics and dispersion characteristics of flows in asymmetric street canyons, Atmospheric Environment (1988). Tartaglia M., A. Giannone, G. Gualtieri and A. Barbaro. Development and validation of an urban street canyon model based on carbon-monoxide experimental data measured in Firenze. In: Urban Transport and the Environment for the 21st Century (Sucharov L.J. (Ed.)). Computa-tional Mechanics Publications, Southampton (1995). Zannetti.Air pollution modeling, Computational Mechanism Pubblications, Southampon (1990). Gualtieri G., &Tartaglia M., Predicting urban traffic air pollution: a GIS framework. Transportation Research - D, Elsevier Science (1998). R. Mohanraj, P. A. Azeez. Urban development and particulate air pollution in Coimbatore city, India Urban Planning for Tehran, By Using Environmental Modeling and GIS/RS Dr. AlirezaGharagozlu, Manager of Public Relations and International Affairs of NCC Research Institute of National Cartographic Center of Iran (NCC) Gis based urban scale air pollution modeling within a german-ulgarian twinning project Helmut Lorentz, Jürgen Friebertshaeuser and Achim Lohmeyer. Munn, R.E (1979). Environmental Impact Assessment, Principles and Procedures. JohnWiley& Sons, New York. Mahoney, J.R. (1974) Meteorological aspects of air pollution. In Industrial Pollution (N.I. Sax, ed.), Section 15. Van Nostrand Reinhold, New York. The Pollution Control Department, Air and noise quality management division, State of Air and Noise Pollution Management 2003, He’s Co. Ltd, Bangkok, 2003, pages (pp. 8–17). Joumard, R., et al. Model of exhaust and noise emissions and fuel consumption of traffic in urban areas, INRETS, France (1992). De Paul, F.T., Sheih, C.M. Measurements of wind velocities in a street canyon, Atmospheric Environment (1986).

International Conference of GIS-Users, Taza GIS-Days, May 23-24, 2012 Proceeding Book