English Arabic French German Hindi Italian Portuguese Russian Spanish

Development of interrupted flow traffic noise prediction model for Dhaka City

AddThis Social Bookmark Button

S. Tanvir & M.M. Rahman

Bangladesh University of Engineering and Technology, Dhaka, Bangladesh




In the context of urbanization process, the increasing urban transport demand, growing conges- tion, environmental implications, the large size of investments and the impacts of urban transport on the quality of life of people. With deteriorating level of mass transport services and increased use of personalized motor vehicles,  vehicular  noise  pollution  is  assuming serious  dimensions  in  most of  the  metropolitan  cities  in view of its associated health hazards.   Simulation of urban traffic noise along with a mathematical model of interrupted flow traffic noise for a road network in Dhaka city is represented in this paper. The analysis data consisted of traffic characteristics, geometrical dimensions of road sections and its noise levels at nine different  sites  of  Dhaka  city.  Characteristics  of  noise  from  different  types  of  vehicles  were  analyzed,  including CNG driven Autorickshaws and Motorcycles. Characteristics of traffic noise levels and data on other traffic noise parameters were used to analyze and subsequently construct the interrupted flow traffic noise simulation model. The resultant models, which are separated into acceleration and deceleration lane models and their sta- tistical goodness- of- fit tests, are presented. The collected data were used to check compatibility of different globally accepted traffic noise models e.g. FHWA model, Stop-and-Go model (Thailand, 1999), Regression model (India, 2003) and Acceleration and Deceleration Lane model (India, 2006) for Dhaka condition.   After comparison of result, it was observed that separated lane models of Stop-and-Go model and Acceleration and Deceleration Lane model are more appropriate for Dhaka and could give acceptable result. In addition, statis- tical analysis has been done on the data between measured and predicted values and good agreement was ob- tained. Also, noise level was compared to the standards and the study reveals that the central part of the city e.g. Bijoy Soroni More, Bangla Motor, Poribaag has been the most affected and noise abatement of these area is nearly impossible without efficient traffic management.




Noise is unacceptable level of sound that creates annoyance, hampers mental and physical peace, and may in- duce severe damage to the health. Along with the increasing degree of air and water pollution, noise pollution is also emerging as a new threat to the inhabitants of Dhaka City. Exposure to high level of noise may cause severe stress on the auditory and nervous system of the city dwellers, particularly the children. Although there are many sources of noise, which include industries, construction works and indiscriminate use of loud speakers, motorized traffic is the principal source of creating noise in urban areas (OECD, 1995). With the increase of in the number of motorized vehicles in the city, the hazard of noise pollution has increased and exceeded the level of tolerance. It is reported that the hearing ability of the inhabitants of the City has reduced during the  last  ten  years. About five to  seven  percent  of the  patients  admitted  to the  Bangabandhu  Seikh  Mujibur Rahman Medical University, Dhaka are suffering from permanent deafness due to noise pollution (Ahmed, 1998). Disturbances created by noise may cause hypertension, headache, indigestion, peptic ulcer, pharyngitis, atherosclosis, bradycardia and ectopic beat (Papacostas and Prevedouros, 1993; Kadiyali, 1997).


The major factors which, influence the generation of road traffic noise are traffic flow, traffic speed, pro- portion of heavy vehicles, gradient of the road and nature of the road surface. In addition, the following fac- tors influence the noise level at a point distant from the highway: attenuation of sound waves due to distance between source and receiver and also due to ground absorption, obstruction due to noise barriers ,obstruction of the sound waves due to restricted angle of view of the same line from the reception point, reflection effects (Agent & Zeeger, 1980). Traffic noise prediction models have been attempted in most of the developed countries. But the effort has been limited in free flow conditions mostly. Urban planners often have to rely on road traffic noise prediction models for their assessment. The more popular ones include the CRTN model in the UK, the FHWA model in the US, the RLS90 model in Germany, the OAL model in Austria, the StatensPlan- verk48 model in Scandinavia, the EMPA model in Switzerland, and the ASJ model in Japan. All these models are based on the uninterrupted traffic flow consideration. But the scenario in urban movement is very different than uninterrupted. Moving vehicles tend to accelerate and decelerate frequently due to interrupted flow occurring at closely  spaced  traffic  signals.  Recently  noise prediction  models  are  being  developed  for  inter- rupted conditions. Different characteristics that are apparent throughout the noise measurement period of in- terrupted traffic flow conditions in urban areas make formulating a theoretical noise prediction model difficult and complex.


In Bangladesh, researches on traffic noise pollution are limited when compared to other developed countries. Moreover, Bangladesh does not have its own traffic noise prediction model that encompasses the Bang- ladeshi traffic characteristics and prevailing environmental conditions. Most of the research work on noise is limited within to check the standards. Only attempt made by Alam et al. (2001) is noteworthy. This  model was developed using dataset from seven locations in Dhaka, but the main drawback of this model is the ab- sence of interrupted flow condition and use of noise equivalence that is not consistent with traffic-noise characteristics of Dhaka city.


To deal with these drawbacks our study had focused on four specific objectives:

i.  To develop traffic noise production equivalences for different vehicle types in comparison with automo- biles (car/ jeep/ van).

ii.  To check various noise generating factors for their correlation with noise level.

iii.  To produce interrupted flow traffic noise model that is suited for Dhaka city.

iv.  To evaluate various empirical and analytical models accepted globally for their suitability in the condi- tion of Dhaka.






2.1   Determination of acoustic equivalence for vehicles


At first, acoustic equivalence level of different vehicle types with respect to automobiles are to be generated. This  value  is  highly  needed  to  be  determined  for  calculation  of  noise  weighed  flow.  Equivalence  is  deter- mined using Equation 1 for each of the six vehicle classes.


Equivalence, E = 10^ [(Mean Leq of any vehicle - Mean Leq of Passanger Car)/10]                              (1)


2.2   Finding out suitable predictor


All the measured parameters which were assumed to be an effect on traffic noise production are correlated with noise level and also among themselves. This is done in two ways: plotting relational graphs of individual variables with noise (in MS- Excel) and evaluating Pearson’s Correlation and 1- tailed Significance Values in a form of matrix that is found from analysis.



2.3   Development of model


The mutually exclusive independent variables are then analyzed for model building. The analytic framework is Multiple Linear Regression Analysis (MLRA). Both forced (enter) and stepwise methods have been used. Forced entry regression model is intended to be more acceptable even if its goodness- of –fit is poor. This is because they intend to establish a relationship with previously modeled predictors globally accepted. Statistical software SPSS is used to perform this operation.


2.4   Comparison with other models


The suitability of some globally recognized empirical and analytical traffic noise prediction models has also been explored. They include: FHWA model, Stop- and- Go model (Thailand, 1999), Regression Model (In- dia, 2003), Acceleration and deceleration lane model (India, 2006). A comparison of the calculated values and observed  values  has  also  been  done  to  find  out  their  compatibility  by  finding  corresponding  R2  value  and paired t- test.





3.1   Data for noise equivalence


For convenience of measuring the effect of individual vehicle noise a long stretch of road with minimum traffic movement was needed. So, we have strategically selected Ashulia, a sub urban point of Dhaka city, which consisted of all types of vehicle that runs on Dhaka roads yet contain fairly long vehicle spacing. Traffic noise level was measured with an integrated precision noise level meter placed about 15 m away from the centerline of the roadway at a height of 1.2 m. Noise level was measured at A- weighted scale for time duration of 10 seconds. The corresponding vehicle type was also noted along with. Data for the analysis of the relationship between vehicle noise and speed of this figure consisted of 619 data sets, separated into 6 types of vehicle, which were collected from approximately a minimum of 40 vehicles/ type.



3.2   Data for the model


Study locations of this study include mainly the urban road network in the central part of Dhaka, the capital city  of  Bangladesh.  Data  collection  sites  were  Shahbaag  East,  Shahbaag  North,  Poribagh,  Banglamotor, Moghbazaar, Kakrail, Bijoy Shoroni More, Kakoli and Khilkhet. At the selected nine study locations 20 data- sets are obtained at both acceleration and deceleration sides of the roadway during a period of March 2009 to September 2009. A usual dataset contains volume and spot speed readings at both sides of the roadway, geo- metric features of the roadway cross section and noise level value at a definite distance and height from road- way curb ( at 1m distance, height 1.2 m).Categorized vehicle count and the counting period are noted for de- termination of equivalent hourly volume. We have used video-camera recordings to find out traffic volume data (Minimum 200 hours for each site). To calculate overall mean speed of the vehicle stream we have used laser gun technology by picking up spot speed of different vehicle in a random manner. The vehicle type is also noted along with their speed which was applied to find the acoustic adjustment of speed later on. Geometric measurements of roadway cross section are done with the help of an odometer. Noise level measurement was  done  simultaneously  with  volume  and  speed  recording  using  a  sound  level  meter  called  RION  NL-32(class 1). The equivalent noise level from the collected data is used and recorded for further analysis.



3.3   Calculation of noise equivalence


Using Equation 1, we get the noise equivalence for different vehicles as shown in Figure 1.



From these calculation, Equation 2 is derived that provides us noise equivalent traffic flow.

QE  =QCar/Jeep/Van + 7.161*QTru + 4.188* QBus + 2.742*QAuto + 1.811*QLCV/MB + 1.245*QTw   (2)


Where, QE= Equivalent Traffic Volume, QTru= Truck volume, QBus= Bus volume, QAuto = Autorickshaw volume, QLCV/MB = Light commercial vehicle /Minibus volume, QTw= Two wheeler volume.


3.4   Single lane model analysis


Single lane model is the simplest approach to catch the urban stop-and-go situation traffic noise. Here we use our noise measuring device at a specified height and distance from curbside in various distances from the in- tersection. Figure 2 describes parameters that can be used as predictors. Independent variables primarily con- sidered are Vn, Vf= Volume of traffic for near side and far side of observer in vehicles per hour, Sn, Sf=Mean speed of traffic for near side and far side of observer, Dg= Geometric mean of the road-side section= √(Df * Dn) in meters, and Dn and Df are the distance from the observer to the central line of the near and far-side roadway in meters, J= Distance to nearest intersection stop line in meter, W= Width of the carriageway in me- ters, L= Queue length in meters.




The single lane model has provided us with Equation 3. It has an adjusted R2 value of 0.895 and standard error of 0.75.


Leq (dBA) =27.977 + 9.664 log Vn+ 4.375 log Vf – 0.021 Sn+ 0.079 Sf – 0.046 Dg                         (3)


3.5   Separated lane model analysis


This  analysis  approach  acknowledges  the  difference  in traffic noise  characteristics  between  an  acceleration lane and deceleration lane on both sides of urban road when vehicles leave an intersection on a green traffic light and come to a stop on a red traffic light.


3.5.1   Acceleration lane model:


The acceleration lane model was built data generated from noise level meter placed on the sidewalk near the acceleration lane of the roadway when traffic leaves the intersection. Data for nearside and farside parameters from all the locations subject to the acceleration lane condition were applied to build the model. Every parameter, which was considered to generate potentially traffic noise under acceleration condition, was tested against observed traffic noise level for its correlation with generated Leq and to observe whether any colinearity existed among noise generating parameters. From this we get Equation 4 as our acceleration lane model that has an adjusted R2  value of 0.927 and standard error of 0.514.


Leq (dBA) = 60.667 + 3.755 log Vn- 3.408 log Vf + 0.418Sn- 0.054 Sf + 0.042 Dg                              (4)


3.5.2   Deceleration lane model


In case of deceleration lane traffic noise model, the same procedure as procedure as mentioned earlier was adopted.  The  noise  data  was  obtained  from  the  roadway  data  collection  stations,  except  that  this  time  the noise level meter was set up by the side of a deceleration lane where traffic enters into an intersection. Analysis provides Equation 5 as our deceleration lane model with adjusted R2 value of 0.977 and standard error of 0.396.


Leq=35.978 + 6.357 log Vn- 3.477 log Vf -0.032Sn + 0.165 Sf – 0.067 Dg + 2.601 log L                     (5)


3.5.3   Decisions on separated lane model

The highest value of R2  obtained from the two separated lane models of the acceleration and deceleration lane were  0.927  and  0.977  respectively.  These  R2   values  are  higher  than  the  R2   value  obtained  for  the  single model, which implied that the independent variables or noise generating parameters used in the two separated models  provided  a  better  explanation  of  the  dependent  variable  or  Leq  than  the  parameters  used  in  single model. The separated model thus provided a better result for forecasting traffic noise than single model. All the apparent coefficients in the acceleration and deceleration lane models also had logical meanings in refer- ence to traffic and traffic noise characteristics. In the analysis process of separated models, parameters of dis- tance from the observer to the nearest junction (J) together with the distances from observer to nearside and far side building facades were screened out because of the insignificant effects of these parameters to each of these two separated models.


3.6   Comparison with known models


In this study four different models are utilized for the prediction of noise level namely FHWA, Stop- and- Go Model (Thailand, 1999), Regression (India, 2003), Acceleration and Deceleration Lane model (India, 2006). FHWA  modeling  is  a  part  of  empirical  modeling  strategy  while  the  others  are  classified  as  an  analytical model. The predicted results are then analyzed by comparing them with the observed noise model. Table 1 summarizes comparison of different accepted noise model in accordance with our observed data.






3.7   Check of noise against ambient level


Noise level against all 20 data sets is compared to their ambient noise level corresponding to their zone. All these data points have exhibited higher level of noise than their standard level. This is described in Table 2.





Increasing use of motorized private vehicles and deterioration of mass transport facilities has lead to an enor- mous problem of noise pollution in urban and semi urban areas of Dhaka city. This problem has been multi- plied due to positioning of closely spaced intersection and accelerating- decelerating noise of motorized vehi- cles  resulting  from  them.  This  study  attempted  to  assess  the  impact  of  various  predictors  on  traffic  noise generations  and  develop  an  analytical  tool  to  be  used  for  the  purpose of  predicting  interrupted  traffic  flow noise. Equation 3, 4 and 5 consist of set of predictors that have been used to develop several globally accepted interrupted flow traffic noise models (TNM). Suggestion of the study is to use separated lane models rather than single lane. A comparison of some globally accepted noise models with prevailing noise data revealed that they have a very low level of predictability of the actual situation. Used all around the world the FHWA noise model, is an empirical model based on free flow condition , has shown a very poor correlation with the observed noise level and paired t- test revealed a very high t- statistics (negative value suggest this model under predicts the actual noise). As far as the analytical models are concerned Stop and Go Model and Accelera- tion and Deceleration Lane Model (India, 2006) had shown a good correlation with the observed noise level with the difference of observed and predicted noise being minimum for their acceleration lane models. The Regression Model (India, 2003) is a single lane model which seems to have almost no correlation with the ob- served noise values and decided unsuitable for use in Bangladeshi conditions. Moreover, the existing severity of noise pollution of Dhaka city has soared to a position that at some places it has become nearly impossible to  lessen  the  level  of  severity  within  acceptable  limit  with  only  structural  abatement  measures  (Table  2). Therefore it has become very urgent to apply various traffic management strategies to control this noise pollu- tion. Some important findings of our study are:


1.  At all other studies previously done noise levels have been found to have directly correlated with nearside and farside volumes and speeds of the roadway. But in our study we have found that with the increase of decelerating lane speed noise level tends to decrease. This phenomenon may be better explained by effect of horns because at interrupted flow condition vehicle drivers tend to use horns more frequently as the reg- ulations concerning use of horn are not properly enforced.

2.   For developing deceleration lane model a new parameter, Queue Length has been introduced than those were previously used in stop- and- go noise model. Logarithmic value of this predictor has shown a very good correlation with the deceleration lane noise level and therefore it is introduced into the forced regres- sion predictors (Equation 5).

3.  Observations include motorcycles, heavy vehicles (trucks and buses), and vehicles with faulty exhaust sys- tems tend to produce high noise levels. Older diesel engines tend to be the noisiest, followed by gasoline and natural gas, hybrid, and electric vehicles being quietest.

4.  It has been found studying the acceleration lane speeding that Lower speeds tend to produce less engine, wind and road noise. Aggressive driving, with faster acceleration and harder stopping, increases noise as engine noise is greatest when a vehicle is accelerating or climbing an incline.

5.  Effect of horn is by far the most notorious and complex one which has yet to be explained into the predic- tors of TNM. The study revealed that horn tends to increase the level of noise to an extent of 7 to 10 dBA.


All in all, we think result of this study will assist the urban planners and traffic mangers of Dhaka city in a great way by providing them an analytical tool to select proper noise abatement and control procedure.




Agent, K.R., Zeeger, C.V. 1980. Effect of interrupted flow on traffic noise. Res. Rep. No. 542, Kentucky Department of Transporta- tion, Frankfort, Ky.

Ahmed, K. 1998. A study of noise pollution in Dhaka City. Department of Environment (DOE), Bangladesh.

Alam, J. B., Rauf, A.F.M. A. and Ahmed, M.F. 2001.   Traffic Induced Noise Pollution in Dhaka city. Journal of Civil Engineering, The Institute of Civil Engineers,Bangladesh, Vol. CE 29 , No. 1, 2001

Alam, J. B. 2006. Risk assessment due to in discriminated pesticide use. Journal of Bangladesh Environ. Sci., 57 (4):57-64. Benchakulpitak, S., and Noppaird, P. 1993. Analysis of vehicle noise level in Thailand. Department of Civil Engineering, King

Mongkut Institute of Technology- Thonburi, Bangkok, Thailand.

FHWA 1995. Highway Traffic Noise Analysis and Abatement Policy and Guidance, U.S. Department of Transportation, Office of environment and Planning, Noise and Air Quality Branch, Washington, D.C., June 1995.

Harris, C. M.1979. Handbook of Noise Control. McGraw Hill, USA.

Jain, S. S. and Parida, M. 2001. Final Report on ‘Development of Comprehensive Highway Noise Model and Design of Noise Bar- rier, Sponsored by Ministry of Road Transport & Highways, Government of   India, New Delhi.

Jain, S. S., Parida, M., Mittal, N., & Bhattacharya, C. C. 2001. Development of comprehensive highway noise model for Indian conditions. Journal of the Indian Roads Congress, 62(3), 453–488.

Kadiyali, L. R. (6th Ed.) 1997. Traffic Engineering and Transportation Planning. Khanna Publishers. India.

Nirjar, R. S., Jain, S. S., Parida, M., & Katiyar, V. S.2003. Study of transport related noise pollution in Delhi. Journal of the Institute of Engineers (1) – Environment, 84,6–15.

OECD (Organization of Economic Cooperation and Development). 1995. Road side noise abatement. Road Transport Research, OECD, France.

Pamanikabud, P. Chairsi, T.1999. Modelling of Urban Area Stop- and-Go Traffic Noise, ASCE Journal of Transportation Engi- neering, vol 125, no 2, 1999.

Papacostas, C. S. and Prevedouros, P. D.(2nd  Ed.)1993. Transportation Engineering and Planning.   Prentice Hall Publishers, USA.


Source: iebconferences.info



Add comment

Security code

| + - | RTL - LTR