1. Introduction
Transportation networks and their facilities (e.g. tollgates) play a vitalrole in the development of nations by providing accessibility services to the citizens and merchandisers. However, traffic emissions(e.g. carbon monoxide (CO) emissions) on roadways and highways are the major cause of air pollution. Traffic CO is mainly due to the large number of cars and smokeproducing motorbikes and the geometry of road and traffic. Various physical and psychological consequences may occur when traffic settlement level is inappropriate (Tobollik, 2016). CO emissions from vehicular exhausts have various negative effects on human health and comfort, such as cardiovascular diseases, respiratory effects, preterm births and cancer diseases (Garshick et al., 2003; Delfino et al., 2010; Crouse et al., 2010).
Collection of traffic CO concentrations on crowded and high-speed highways can be costly and time consuming and can create contingency. Therefore, traffic CO models are highly relevant and important. Traffic CO cannot be measured during the designing stage of highways.Traffic pollutionmodels, particularly traffic CO models, must be considered at the initial stage of new highway design projects for comfort in different industrial, commercial and residential environments. Therefore, traffic CO modelling can be effectively used as a decision support toolfortrafficCO simulation and prediction on highways. Consequently, friendly environmental roads can be properly planned (Van Woensel, 2001).
Many studies have been conducted using different techniques, such as traditional approaches based on simple geostatistical interpolation, dispersion models, machine learning models, land use regression (LUR) models and hybrid models (Singh et al., 2016; Kim et al., 2015; Schneider et al., 2017; Suleiman et al., 2016; Wang et al., 2015). These models provide prediction results that can be categorised into statistical tables, graphs and maps (Abdulkareem et al., 2018a, 2018b; Abdullahi et al., 2017;Althuwaynee et al., 2012, 2014; Bui et al., 2017a, 2017b; Chen et al., 2017; Nampak et al., 2014; Neshat and Pradhan, 2015; Neshat et al., 2014, 2015; Oh and Pradhan, 2011; Pradhan, 2013; Pradhan and Lee, 2010; Umar et al., 2014). Amongst these results, spatial prediction based on maps has attracted attention from authors. Thus, in the current study, we developed a spatial prediction model based on geographic information system (GIS) techniques.
The main purpose of the study is to simulate and predict trafficCO emissions by considering the tollgate locations along with some parameters as indicated in the literature.The primary and significant contributions in thisstudy lie in constructing accurate predictivemaps and providing an explanation of the variation of CO emissions at traffic tollgate and road corridors. Other remarkable advantages might be the easy implementation of the developed model using the geospatial approaches in which the users can employ the developed model for rapid assessments of CO emissions by producing the microscale as well as mesoscale prediction maps. In addition, a user can develop GIS based models as perthe requirements.We understood that the integrated approach of SVR and GIS algorithms could able to improve in the previously developed forecasting methods for prediction of CO emissions on traffic tollgate areas.
2. Related works
Many air pollution-related traffic emission studies were reviewed by Singh et al. (2016), Behera et al. (2015) and Johnson et al. (2010). The first GIS-based model was implemented on the basis of traditional methodsthat integrated data sampling and geostatistical interpolation or gridding analysisto produce prediction maps (Abbey, 1991). The traditional implementation of geostatistical interpolation can be a useful solution to reduce data collection cost and time by estimating pollutant values at non-sampled locations. Potoglou et al. (2005) developed a model based on data sampling and geostatistical interpolation to estimate traffic CO in Hamilton, Canada. The CO emissions collected using MOBILE5C emission model were interpolated on the basis of kriging geostatisticmethod.The findings showed that the maximum value of CO is 4.5 ppm. In a separate study, Behera et al. (2015) presented an approach based on an integration of data samples collected from air quality monitoring stations and geostatistical interpolation (kriging method) to predict nitrogen dioxide (NO2) in two main Indian cities, namely, Delhi and Kanpur. Their prediction maps showed higher NO2 concentrations in industrial sites than in traffic locations. Their results showed that the prediction maps can sufficiently represent pollutant behaviour and are valid for other types of pollutants.
Other important models are dispersion models, which are used forspatial prediction oftraffic emissions based on traffic parameters and dispersion algorithms. Kho et al.(2007) presented a methodology based on an air dispersion model (CAL3QHC) and meteorological data and traffic flow information to conduct a comparison in terms of traffic CO concentrations between 2004 and 2014 in Sabah, Malaysia. The final findingsrecorded that the maximum CO concentration is 9.3 ppm, whereas the minimum value is 0 ppm. This previous study recommended monitoring the CO concentration from road traffic in a sustainable environment and transportation planning. In the meantime, Borrego et al. (2016) developed a model based on dispersion model (Gaussian algorithm) and traffic information for traffic emission prediction within a small city in Portugal. The results observed 78% within a factor of two of the hourly average concentration, and this value increases to 94% when daily averages are considered. The main advantage of this approach isthe agreement betweenmodel and local measurement results, which show the potentiality of instantaneous traffic CO emissions data in details with urban-level values extracted from the filtered monitoring results. The model can also be applied to provide accurate data at the urban scale.
Recently, LUR models are attracting attention from authors because of their ability to predict air pollutants atspecific location and time and their ability to be integrated with GIS environment. Arain (2007) developed a LUR model based on traffic data, land use and wind speed and direction to predict NO2 concentrationsinToronto,Canada.The resultsshowed a significant variation in NO2 along the study area. Wind flow effects on the NO2 emissions were also evaluated. The results indicated that ultra-highresolution wind models are essential to capture localscale atmospheric patterns. By contrast, regional-scale models GEM-HiMAP used for weather forecast to predict wind fields cannot resolve the issue of wind flow patterns at small scales for the livelihood and health system. Bertazzon (2015) presented a modified approach to addressspatial, non-stationary connection in LUR models based on geographic, traffic and meteorology data for predicting NO2 in Calgary, Canada.The resultsindicated higher NO2 concentration in summerthan in winter, and the maps obtained by the model illustrated the spatial prediction of NO2 within the study area.The developed wind LURmodelsin this previous research offer a simple, significant and effective method to addressspatial autocorrelation and non-stationarity in a single model and reduced errors due to the spatial effects.
The newest models are designed on the basis of artificial intelligence models (e.g. machine learning or deep learning). These models are widely applied in prediction issues due to their ability to produce accurate results. Cai et al. (2009) developed a model based on neural network model and field data that contained traffic data, meteorology, proximity to roads and road direction to predict air pollutants every hour near roadways. This model indicated that the highest correlation coefficient for the CO prediction is 0.879. Shakerkhatibi et al. (2015) presented a hybrid model by combining neural network and evolutionary polynomial regression model to predict CO concentrations in Tabriz, Iran. Their results indicated that the neural network model is more reliable than evolutionary polynomial regression. The developed model is an excellent tool for air quality estimation. In the meantime, Moazami et al. (2016) proposed a methodology based on support vectorregression (SVR) model and different datasetsthat involved weather data and pollutant background to predict and simulate the next day’s CO concentrations. Their results indicated that the SVR model has less uncertainty value than neural networkmodel.These previous authors concluded that their methodology can be used in many other fields. However, the above-mentioned models are designed to efficiently consider modelling, uncertainty, multifactor and nonlinearity. Many researchers have tried to solve these issues. Nevertheless, they have mainly concentrated on modelling at large areas and big data that require effort, cost and high level of data processors (e.g. programming tools).
In the current research, we proposed an integrated approach between SVR and GIS models to simulate and predict traffic CO emissions at and near tollgate locations.This modelling with small data can relatively provide high prediction accuracy and explain the variation in traffic CO emission at tollgates, where workers are affected seriously. This developed model can be implemented in any GIS software in a short time period for the assessment of micro-scale vehicular emission modelling and mapping.
3. Methods and dataset
1) Study area
The area islocated in a densely populated place near Jalan Duta toll plaza, Kuala Lumpur, Malaysia (Fig. 1). Jalan Duta toll plaza is situated in Utara–Selatan Expressway–NKVE. The PLUS Official Website indicated that the NKVE is a route that is heavily utilised by residents of Kuala Lumpur, Petaling Jaya, Subang, Damansara, Sungai Buloh and Klang. The speed limit is mostly limited to 110 km/h (68 mph) on Bukit Raja–Bukit Lanjan stretch, Utara–Selatan and is 90 km/h (55 mph) onBukit Lanjan–Jalan Duta stretch. A set of data from NKVE was selected to execute the proper analysis of objectives in the current study. The site hasresidential, industrial and commercial areas and is thus polluted and suitable for traffic emission analysis.
Fig. 1. Location map of the study area.
2) Proposed methodology
This part describes the proposed methodology, which was designed on the basis of several steps. The proposed model was developed to estimate traffic CO emissions atspecific time and location.In other words, prediction maps at different times a day could be produced. The first step is the field surveying of collecting the samples of traffic CO by using CO detector and data logger for temperature and humidity information. The traffic data were collected using a digital camera. Other parameters (i.e. building height and proximity to roads) were derived from light detection and ranging (LiDAR) and GIS data. The second step is the data modelling based on SVR. The third step is the spatial modelling based on an integration of regression equation and GIS model. The results were validated on the basis of root mean square error (RMSE). The last step is the production of the prediction maps at different times a day. Fig. 2 illustrates the overall methodology.
Fig. 2. Overall methodology
(1) Field survey
Field survey was conducted involving several steps inApril 2018.The firststep isthe selection ofsamples. The samples were selected following the approach presented by Ragettli et al. (2016). They applied their method using GIS spatial analysis for traffic CO with some random point locations developed for data collection to ensure spatial balancing. The model generation of CO samples from the studied region ensuredoptimaldensityandadequatenumberofsamples. Traffic CO emissions were continuously measured in every 15 min interval (15 min averages) using a gas detector microclip5 (1 ppm resolution).TrafficCO was measured four times a day during weekends and weekdays. Every day, traffic CO concentrations were measured during morning (AM peak) (around 6.30 am–8.30 am), afternoon (around 11.30 am–1.30 pm), evening (PM peak) (around 6.30 pm–8.30 pm) and night (around 11 pm–12 midnight). In the meantime, traffic flow information (i.e. numbers of cars, heavy vehicles and motorbikes) and meteorological factors (i.e. temperature,relative humidity and wind speed and direction) were collected with samplessimultaneously.
(2) Proposed method
① Parameters of the traffic CO prediction model
As seen in Fig. 3, the proposed model aimed to predict and represent the spatial prediction oftrafficCO in the study area on the basis of several parameters as listed in literature (Azeez, 2018). In this model, the dependent factor was the concentration of traffic CO every 15 min (Azeez, 2018). Independent factors were selected on the basis of the literature presented and the nature of the study area. The selected factors were number of cars, number of heavy vehicles, number of motorbikes, temperature,relative humidity, wind speed, wind direction, building height and proximity to roads. Table 1 showsthe summary ofstatistics ofthe predictors of the model.
Table 1. Summary statistics of CO emission predictors
Fig. 3. Location of traffic CO measurement of highway section.
② SVR model
SVR is a machine learning algorithm used for nonlinearregression issues and can be used as universal approximates of multivariate task at any level of accuracy (Bishop, 2006). SVR model is applied to predict dependent variable ‘y’ depending on group of independent variables ‘x’, as presented in Eq. 1:
\(\mathrm{y}=\left(\mathrm{w}^{\mathrm{T}} \cdot \Phi(\mathrm{x})+\mathrm{b}\right)+\text { noise }\) (1)
Where the noise of the model is represented by the tolerance of error (ε). w represents the vector of coefficient, whereas b represents a constant value. The kernelfunction isrepresented by Φ and used to convert the data values to high-dimensional feature space to make them more separable than the originalspace.The task is to find a functional form for wT · Φ(x) + b. This form can be obtained by tuning the model. Then, the error function reduction is used to derive w and b using Eqs. 2 and 3 (Goel, 2009):
\(\frac{1}{2} w^{T} \cdot w+C \sum_{i=1}^{N} \xi_{i}+C \sum_{i=1}^{N} \xi \cdot i\) (2)
\(w^{T} \cdot \Phi\left(x_{i}\right)+b-y_{i} \leq \varepsilon+\xi \cdot i\) (3)
\(\xi \cdot i, \xi \cdot i \geq 0, i=1, \ldots, N\) (4)
where C indicates a positive constant that determines the degree of penalised loss as calibration error occurs; Nisthe sample size; and ξ ·i and ξ ·i are slack variables specifying the upper and lower calibration errors subject to ε, respectively (Goel, 2009).
③ GIS model
The proposed GIS model wasimplemented to create predictionmaps ofthe trafficCO emissionsin the study area (Fig. 4).The structure of GIS model was designed on the basis of the integration between the parameters of the model in GIS environment and the regression equation from the SVR model. The parameters of the model were converted to raster format using geostatistical interpolation (kriging). In the meantime, data containing the heights of the buildings were extracted from LiDAR. The proximity to road raster was generated using spatial analysistechniques. Spatial prediction results were presented on the basis of a very-high-resolution grid (5 × 5) m2 (Azeez, 2018). Thus, any grid has a CO value based on the integration of parameter values that will show the variation in traffic CO concentrations and their spatial distribution in the study area.
Fig. 4. GIS model.
4. Results and Discussion
1) Results of the traffic CO prediction model
The SVR model was trained and validated using the free source application (WEKA) based on nine parameters to develop the regression equation. Traffic CO samples were collected and used in the proposed model to calculate the correlation between the predictive parameters and the CO levels for training and the importance for validation. The developed model was used to create coefficients ofthe parameters ofthe model collected during weekdays and weekends at different times a day (morning, afternoon, evening and night). Table 2 shows the prediction results. As shown in the table, the correlation coefficient was 0.97, the mean absolute error(MAE) was 1.401 ppm and the RMSE was 2.45 ppm. After the SVR model was trained using the field collected data ofCO, a regression model was developed, as shown below:
\(\begin{array}{lc} Predicted(\mathrm{CO})=0.0049 \times \mathrm{Number}\ of \ cars -\\ 0.0024 \times Number \ of \ heavy\ vehicles+\\ 0.0016 \times Number \ of \ motorbikes - \\ 0.0027 \times Temperature -0.0118 \times\\ Wind \ speed -0.0029 \times\\ Building\ height +0.158 \end{array}\) (5)
Table 2. Prediction results based on SVR model
2) Results of GIS model
The GIS model was utilised on the basis of an integration of SVR equation and model predictorsthat were converted to GIS form and multiplied by their coefficients. The results were presented on the basis of a very-high-resolution grid (5 × 5) m2 that presented prediction maps at different times a day (Fig. 5). The final findings indicated that the traffic CO concentrations were very high at weekdays because of traffic congestions at peak hoursin the morning.These concentrations were higher than those recorded at daytime in a normal traffic flow. The minimum traffic CO concentration was 0 ppm. By contrast, the maximum value of 38.27 ppm was detected during morning at weekdays. The traffic flow data indicated that the highest numbers of cars (610), heavy vehicles (24) and motorcycles (41) were obtained during evening. The maximum average temperature of 33°C was recorded during afternoon in the study area. The minimum value was 27.4°C. In the meantime, the maximum average relative humidity of 90.1% was obtained during evening, whereas the minimum value was 60.2%. The maximum average wind speed of 7 km/h was detected in the afternoon. The minimum value was 2.3 km/h. The prediction maps (Fig. 6) showed that the maximum traffic CO concentrations were concentrated near tollgate areas due to heavy traffic congestion, whereastheminimumconcentrations were located near residential areas.
Fig. 5. Prediction maps during weekends.
Fig. 6. Prediction maps during weekdays.
3) Validation of the model
The model was validated on the basis of RMSE, prediction maps and testing data. The predicted traffic CO values were compared with the observed traffic CO. Table 3 shows the validation results at different times a day based on RMSE. The lowest mean of RMSE of 0.04 ppm was obtained during evening at weekends, whereasthe highest mean ofRMSE of 0.83 ppm was derived during morning at weekdays.
Table 3. Validation results at different times a day
4) Comparison with other models
The proposed model was compared with a simple linearregression (LR) model that does notrequire high level of skills for its implementation. The LR model was applied using the same parameters asthe proposed model. The LR model is shown in Eq. 6.
\(Vehicular\ \mathrm{CO}=0.14 * Number\ of\ heavy\ vehicles+5.62\) (6)
The effective comparison between the proposed method and the LR model indicated that the SVR model performed better than the LR model. The correlation coefficient based on the LR model was 0.892, the MAE was 3.23 ppm and the RMSE was 3.902 ppm. The relative absolute error was 54.68%, and the root relative square error was 39.79%. Fig. 7 shows the comparison amongst three models (the proposed SVR model, traditional model and simple LR model).
Fig. 7. Comparison between SVR and LR models.
5. Conclusion
Traffic CO emissions from dense vehicular environments (e.g. highways, road networks and tollgates) are major sources of air pollution in most urban areas. Prediction and geospatial models are generally applied to evaluate the impacts of traffic emissions emitted from different kinds of vehicles on the environment and human health.In the currentstudy, a model was developed on the basis of a combination of SVR and GIS models. The results showed that the accuracy of the proposed model was 74.4%, and the model obtained a lowest mean of RMSE of 0.1 ppm. The model had nine parameters, namely, number of cars, number of motorbikes, number of heavy vehicles, temperature, wind speed, relative humidity, wind direction, building height and proximity to roads.
The GIS model wasimplemented on the basis ofthe GIS layers extracted from interpolation analysis and LiDAR data and overlaid depending on regression equation to provide prediction maps at different times a day during weekends and weekdays. The prediction maps illustrated the spatial variation of traffic CO along the study area. Specifically, the maximum traffic CO concentration was detected near tollgate areas. By contrast, the minimum value was detected near residential areas
The integration of traffic CO prediction model and GIS modelling is an important tool for transportation assessment and planning. These models can be efficiently implemented as decision-making tools to find a suitable solution for mitigating traffic jams near tollgates, highways and road networks.
Acknowledgements
The authors acknowledge and appreciate the provision of airborne laser scanning data (LiDAR), satellite images and logistic support by the PLUS Berhad. In addition, the second author, Biswajeet Pradhan, gratefully acknowledges the financial support from the PLUS industry project grant and the UTS under grant number 321740.2232335 and 321740.2232357. For Hyung-Sup Jung, thisstudy was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (No. NRF-2018M1A3A3A02066008).
References
- Abbey, D. E., P. K. Mills, F. F. Petersen, and W. L. Beeson, 1991. Long-term ambient concentrations of total suspended particulates and oxidants as related to incidence of chronic disease in California Seventh-Day Adventists, Environmental Health Perspectives, 94: 43-50. https://doi.org/10.2307/3431291
- Abdulkareem, J. H., W. N. A. Sulaiman, B. Pradhan, and N. R. Jamil, 2018a. Long-term hydrologic impact assessment of nonpoint source pollution measured through land use/land cover (LULC) changes in a tropical complex catchment, Earth Systems and Environment, 2(1): 67-84, https://doi.org/10.1007/s41748-018-0042-1.
- Abdulkareem, J. H., B. Pradhan, W. N. A. Sulaiman, and N. R. Jamil, 2018b. Quantification of runoff as influenced by morphometric characteristics in a rural complex catchment, Earth Systems and Environment, 2(1): 145-162, https://doi.org/10.1007/s41748-018-0043-0.
- Abdullahi, S., B. Pradhan, and H. Mojaddadi, 2018. City compactness: assessing the influence of the growth of residential land use, Journal of Urban Technology, 25(1): 21-46, https://doi.org/10.1080/10630732.2017.1390299.
- Althuwaynee, O. F., B. Pradhan, and S. Lee, 2012. Application of an evidential belief function model in landslide susceptibility mapping, Computers & Geosciences, 44: 120-135. https://doi.org/10.1016/j.cageo.2012.03.003
- Althuwaynee, O. F., B. Pradhan, H. J. Park, and J. H. Lee, 2014. A novel ensemble bivariate statistical evidential belief function with knowledge-based analytical hierarchy process and multivariate statistical logistic regression for landslide susceptibility mapping, Catena, 114: 21-36. https://doi.org/10.1016/j.catena.2013.10.011
- Arain, M. A., R. Blair, N. Finkelstein, J. R. Brook, T. Sahsuvaroglu, B. Beckerman, and M. Jerrett, 2007. The use of wind fields in a land use regression model to predict air pollution concentrations for health exposure studies, Atmospheric Environment, 41(16): 3453-3464. https://doi.org/10.1016/j.atmosenv.2006.11.063
- Azeez, O., B. Pradhan, and H. Shafri, 2018. Vehicular CO Emission Prediction Using Support Vector Regression Model and GIS, Sustainability, 10(10): 3434. https://doi.org/10.3390/su10103434
- Azeez, O. S., B. Pradhan, H. Z. M. Shafri, N. Shukla, C. W. Lee, and H. M. Rizeei, 2019. Modeling of CO Emissions from Traffic Vehicles Using Artificial Neural Networks, Applied Sciences, 9(2): 313, doi:10.3390/app9020313.
- Behera, S. N., M. Sharma, P. K. Mishra, P. Nayak, B. Damez-Fontaine, and R. Tahon, 2015. Passive measurement of NO2 and application of GIS to generate spatially-distributed air monitoring network in urban environment, Urban Climate, 14: 396-413. https://doi.org/10.1016/j.uclim.2014.12.003
- Bishop, C.M., 2006. Pattern Recognition and Machine Learning, Springer, NewYork, USA.
- Borrego, C. J., H. Amorim, O. Tchepel, D. Dias, S. Rafael, E. Sa, C. Pimentel, T. Fontes, P. Fernandes, S.R. Pereira, and J. M. Bandeira, 2016. Urban scale air quality modelling using detailed traffic emissions estimates, Atmospheric Environment, 131: 341-351. https://doi.org/10.1016/j.atmosenv.2016.02.017
- Bui, D. T., Q.-T. Bui, Q.-P. Nguyen, B. Pradhan, H. Nampak, and P. T. Trinh, 2017a. A hybrid artificial intelligence approach using GIS-based neural-fuzzy inference system and particle swarm optimization for forest fire susceptibility modeling at a tropical area, Agricultural and Forest Meteorology, 233: 32-44. https://doi.org/10.1016/j.agrformet.2016.11.002
- Bui, D. T., T. A. Tuan, N. D. Hoang, N. Q. Thanh, D. B. Nguyen, N. Van Liem, and B. Pradhan. 2017b. Spatial prediction of rainfall-induced landslides for the Lao Cai area (Vietnam) using a hybrid intelligent approach of least squares support vector machines inference model and artificial bee colony optimization, Landslides, 14(2): 447-458. https://doi.org/10.1007/s10346-016-0711-9
- Cai, M., Y. Yin, and M. Xie, 2009. Prediction of hourly air pollutant concentrations near urban arterials using artificial neural network approach, Transportation Research Part D: Transport and Environment, 14(1): 32-41. https://doi.org/10.1016/j.trd.2008.10.004
- Chen, W., X. Xie, J. Wang, B. Pradhan, H. Hong, D. T. Bui, Z. Duan, and J. Ma, 2017. A comparative study of logistic model tree, random forest, and classification and regression tree models for spatial prediction of landslide susceptibility, Catena, 151: 147-160. https://doi.org/10.1016/j.catena.2016.11.032
- Crouse, D. L., M. S. Goldberg, N. A. Ross, H. Chen, and F. Labreche, 2010. Postmenopausal breast cancer is associated with exposure to trafficrelated air pollution in Montreal, Canada: a case-control study, Environmental Health Perspectives, 118(11): 1578. https://doi.org/10.1289/ehp.1002221
- Delfino, R. J., T. Tjoa, D. L. Gillen, N. Staimer, A. Polidori, M. Arhami, and J. Longhurst, 2010. Traffic-related air pollution and blood pressure in elderly subjects with coronary artery disease, Epidemiology (Cambridge, Mass.), 21(3).
- Garshick, E., F. Laden, J. E. Hart, and A. Caron, 2003. Residence near a major road and respiratory symptoms in US veterans, Epidemiology (Cambridge, Mass.), 14(6): 728. https://doi.org/10.1097/01.ede.0000082045.50073.66
- Goel, A. and M. Pal, 2009. Application of support vector machines in scour prediction on gradecontrol structures, Engineering Applications of Artificial Intelligence, 22(2): 216-223. https://doi.org/10.1016/j.engappai.2008.05.008
- Kho, F. W. L., P. L. Law, S. H. Ibrahim, and J. Sentian, 2007. Carbon monoxide levels along roadway, International Journal of Environmental Science & Technology, 4(1): 27-34. https://doi.org/10.1007/BF03325958
- Markey, J., V. Isakov, J.S. Touma, S. Mukerjee, and H. Ozkaynak, 2010. Evaluation of land-use regression models used to predict air quality concentrations in an urban area, Atmospheric Environment, 44(30): 3660-3668. https://doi.org/10.1016/j.atmosenv.2010.06.041
- Nampak, H., B. Pradhan, and M. A. Manap, 2014. Application of GIS based data driven evidential belief function model to predict groundwater potential zonation, Journal of Hydrology, 513: 283-300. https://doi.org/10.1016/j.jhydrol.2014.02.053
- Neshat, A. and B. Pradhan, 2015. An integrated DRASTIC model using frequency ratio and two new hybrid methods for groundwater vulnerability assessment, Natural Hazards, 76(1): 543-563. https://doi.org/10.1007/s11069-014-1503-y
- Neshat, A., B. Pradhan, and S. Javadi, 2015. Risk assessment of groundwater pollution using Monte Carlo approach in an agricultural region: an example from Kerman Plain, Iran, Computers, Environment and Urban Systems, 50: 66-73. https://doi.org/10.1016/j.compenvurbsys.2014.11.004
- Neshat, A., B. Pradhan, S. Pirasteh, and H. Z. M. Shafri, 2014. Estimating groundwater vulnerability to pollution using a modified DRASTIC model in the Kerman agricultural area, Iran, Environmental Earth Sciences, 71(7): 3119-3131. https://doi.org/10.1007/s12665-013-2690-7
- Oh, H. J. and B. Pradhan, 2011. Application of a neuro-fuzzy model to landslide-susceptibility mapping for shallow landslides in a tropical hilly area, Computers & Geosciences, 37(9): 1264-1276. https://doi.org/10.1016/j.cageo.2010.10.012
- Potoglou, D. and P. S. Kanaroglou, 2005. Carbon monoxide emissions from passenger vehicles: predictive mapping with an application to Hamilton, Canada, Transportation Research Part D: Transport and Environment, 10(2): 97-109. https://doi.org/10.1016/j.trd.2004.11.003
- Pradhan, B., 2013. A comparative study on the predictive ability of the decision tree, support vector machine and neuro-fuzzy models in landslide susceptibility mapping using GIS, Computers & Geosciences, 51: 350-365. https://doi.org/10.1016/j.cageo.2012.08.023
- Pradhan, B. and S. Lee, 2010. Landslide susceptibility assessment and factor effect analysis: backpropagation artificial neural networks and their comparison with frequency ratio and bivariate logistic regression modelling, Environmental Modelling & Software, 25(6): 747-759. https://doi.org/10.1016/j.envsoft.2009.10.016
- Ragettli, M. S., S. Goudreau, C. Plante, M. Fournier, M. Hatzopoulou, S. Perron, and A. Smargiassi, 2016. Statistical modeling of the spatial variability of environmental noise levels in Montreal, Canada, using noise measurements and land use characteristics, Journal of Exposure Science and Environmental Epidemiology, 26(6): 597. https://doi.org/10.1038/jes.2015.82
- Schneider, P., N. Castell, M. Vogt, F.R. Dauge, W.A. Lahoz, and A. Bartonova, 2017. Mapping urban air quality in near real-time using observations from low-cost sensors and model information, Environment international, 106: 234-247. https://doi.org/10.1016/j.envint.2017.05.005
- Shakerkhatibi, M., N. Mohammadi, K. Zoroufchi Benis, A. Behrooz Sarand, E. Fatehifar, and A. Asl Hashemi, 2015. Using ANN and EPR models to predict carbon monoxide concentrations in urban area of Tabriz, Environmental Health Engineering and Management Journal, 2(3): 117-122.
- Singh, D., K. Amit, K. Krishan, and B. Singh, 2016. Statistical modeling of O3, NOx, CO, PM2.5, VOCs and z levels in commercial complex and associated health risk assessment in an academic institution, Science of The Total Environment, 572: 586-594. https://doi.org/10.1016/j.scitotenv.2016.08.086
- Singh, D., S. P. Nigam, V. P. Agrawal, and M. Kumar, 2016. Traffic noise prediction using soft computing approach, Journal of Environmental Management, 183: 59-66. https://doi.org/10.1016/j.jenvman.2016.08.053
- Stefania, B., M. Johnson, K. Eccles, and G. G. Kaplan, 2015. Accounting for spatial effects in land use regression for urban air pollution modeling, Spatial and Spatio-temporal Epidemiology, 14: 9-21. https://doi.org/10.1016/j.sste.2015.06.002
- Suleiman, A., M.R. Tight, and A.D. Quinn, 2016. Assessment and prediction of the impact of road transport on ambient concentrations of particulate matter PM10, Transportation Research Part D: Transport and Environment, 49: 301-312. https://doi.org/10.1016/j.trd.2016.10.010
- Tobollik, M., M. Keuken, C. Sabel, H. Cowie, J. Tuomisto, D. Sarigiannis, D. Sarigiannis, N. Kunzli, L. Perez, and P. Mudu, 2016. Health impact assessment of transport policies in Rotterdam: Decrease of total traffic and increase of electric car use, Environmental Research, 146: 350-358. https://doi.org/10.1016/j.envres.2016.01.014
- Umar, Z., B. Pradhan, A. Ahmad, M. N. Jebur, and M. S. Tehrany, 2014. Earthquake induced landslide susceptibility mapping using an integrated ensemble frequency ratio and logistic regression models in West Sumatera Province, Indonesia, Catena, 118: 124-135. https://doi.org/10.1016/j.catena.2014.02.005
- Van Woensel, T., R. Creten, and N. Vandaele, 2001. Managing the environmental externalities of traffic logistics: The issue of emissions, Production and Operations Management, 10(2): 207-223. https://doi.org/10.1111/j.1937-5956.2001.tb00079.x
- Wang, P., Y. Liu, Z. Qin, and G. Zhang, 2015. A novel hybrid-forecasting model for PM10 and SO2 daily concentrations, Science of the Total Environment, 505: 1202-1212. https://doi.org/10.1016/j.scitotenv.2014.10.078
- Youngkook, K. and J.M. Guldmann, 2015. Land-use regression panel models of NO2 concentrations in Seoul, Korea, Atmospheric Environment, 107: 364-373. https://doi.org/10.1016/j.atmosenv.2015.02.053
Cited by
- Seasonal Water Change Assessment at Mahanadi River, India using Multi-temporal Data in Google Earth Engine vol.36, pp.1, 2019, https://doi.org/10.7780/kjrs.2020.36.1.1
- Methodology for determining ecological benefits of advanced tolling systems vol.258, 2020, https://doi.org/10.1016/j.jenvman.2019.110007