Enhanced transport-related air pollution prediction through a novel metamodel approach

This research proposes a novel approach to improve the ability to forecast low frequency extreme events of transport-related pollution in urban areas using a limited input data set. The approach is based on the idea of a self-managing model, able to adapt to unexpected changes in pollution level. In more detail, for a given combination of variables, it selects the most suitable prediction model within a set of alternative air quality models, estimated for a wider range of locations and conditions.In this study, the new approach is tested for the prediction of nitrogen dioxide concentration in the United Kingdom (UK), specifically in an air quality monitoring site of the Greater Manchester Area, by comparing it with a context-specific statistical model (ARIMAX). The analysis results show that the two methods are similar in terms of global covariance and difference between observed and simulated concentrations, however the performance of the new approach in the prediction of extreme air pollution events is up to 27% better than the standard statistical approach and up to 113% better than the artificial neural network method.     

A latent segmentation based multiple discrete continuous extreme value model

We examine an alternative method to incorporate potential presence of population heterogeneity within the Multiple Discrete Continuous Extreme Value (MDCEV) model structure. Towards this end, an endogenous segmentation approach is proposed that allocates decision makers probabilistically to various segments as a function of exogenous variables. Within each endogenously determined segment, a segment specific MDCEV model is estimated. This approach provides insights on the various population segments present while evaluating distinct choice regimes for each of these segments. The segmentation approach addresses two concerns: (1) ensures that the parameters are estimated employing the full sample for each segment while using all the population records for model estimation, and (2) provides valuable insights on how the exogenous variables affect segmentation. An Expectation–Maximization algorithm is proposed to address the challenges of estimating the resulting endogenous segmentation based econometric model. A prediction procedure to employ the estimated latent MDCEV models for forecasting is also developed. The proposed model is estimated using data from 2009 National Household Travel Survey (NHTS) for the New York region. The results of the model estimates and prediction exercises illustrate the benefits of employing an endogenous segmentation based MDCEV model. The challenges associated with the estimation of latent MDCEV models are also documented.     

Air itinerary shares estimation using multinomial logit models

ABSTRACT

The main goal of this study is the development of an aggregate air itinerary market share model. In order to achieve this, multinomial logit models are applied to distribute the city-pair passenger demand across the available itineraries. The models are developed at an aggregate level using open-source booking data for a large group of city-pairs within the US air transport system. Although there is a growing trend in the use of discrete choice models in the aviation industry, existing air itinerary share models are mostly focused on supporting carrier decision-making. Consequently, those studies define itineraries at a more disaggregate level using variables describing airlines and time preferences. In this study, we define itineraries at a more aggregate level, i.e. as a combination of flight segments between an origin and destination, without further insight into service preferences. Although results show some potential for this approach, there are challenges associated with prediction performance and computational intensity.     

Predicting short-term bus passenger demand using a pattern hybrid approach

This paper proposes an Interactive Multiple Model-based Pattern Hybrid (IMMPH) approach to predict short-term passenger demand. The approach maximizes the effective information content by assembling the knowledge from pattern models using historical data and optimizing the interaction between them using real-time observations. It can dynamically estimate the priori pattern models combination in advance for the next time interval. The source demand data were collected by Smart Card system along one bus service route over one year. After correlation analysis, three temporal relevant pattern time series are generated, namely, the weekly, daily and hourly pattern time series. Then statistical pattern models are developed to capture different time series patterns. Finally, an amended IMM algorithm is applied to dynamically combine the pattern models estimations to output the final demand prediction. The proposed IMMPH model is validated by comparing with statistical methods and an artificial neural network based hybrid model. The results suggest that the IMMPH model provides a better forecast performance than its alternatives, including prediction accuracy, robustness, explanatory power and model complexity. The proposed approach can be potentially extended to other short-term time series forecast applications as well, such as traffic flow forecast.     

Assessment and prediction of the impact of road transport on ambient concentrations of particulate matter PM10

The main challenge facing the air quality management authorities in most cities is meeting the air quality limits and objectives in areas where road traffic is high. The difficulty and uncertainties associated with the estimation and prediction of the road traffic contribution to the overall air quality levels is the major contributing factor. In this paper, particulate matter (PM₁₀) data from 10 monitoring sites in London was investigated with a view to estimating and developing Artificial Neural Network models (ANN) for predicting the impact of the road traffic on the levels of PM₁₀ concentration in London. Twin studies in conjunction with bivariate polar plots were used to identify and estimate the contribution of road traffic and other sources of PM₁₀ at the monitoring sites. The road traffic was found to have contributed between 24% and 62% of the hourly average roadside PM₁₀ concentrations. The ANN models performed well in predicting the road contributions with their R-values ranging between 0.6 and 0.9, FAC2 between 0.6 and 0.95, and the normalised mean bias between 0.01 and 0.11. The hourly emission rates of the vehicles were found to be the most contributing input variables to the outputs of the ANN models followed by background PM₁₀, gaseous pollutants and meteorological variables respectively.     

Prediction of hourly air pollutant concentrations near urban arterials using artificial neural network approach

This paper applies artificial neural network to predict hourly air pollutant concentrations near an arterial in Guangzhou, China. Factors that influence pollutant concentrations are classified into four categories: traffic-related, background concentration, meteorological and geographical. The hourly averages of these influential factors and concentrations of carbon monoxide, nitrogen dioxide, particular matter and ozone were measured at three selected sites near the arterial using vehicular automatic monitoring equipments. Models based on back-propagation neural network were trained, validated and tested using the collected data. It is demonstrated that the models are able to produce accurate prediction of hourly concentrations of the pollutants respectively more than 10 h in advance. A comparison study shows that the neural network models outperform multiple linear regression models and the California line source dispersion model.     

Short-term speed predictions exploiting big data on large urban road networks

Big data from floating cars supply a frequent, ubiquitous sampling of traffic conditions on the road network and provide great opportunities for enhanced short-term traffic predictions based on real-time information on the whole network. Two network-based machine learning models, a Bayesian network and a neural network, are formulated with a double star framework that reflects time and space correlation among traffic variables and because of its modular structure is suitable for an automatic implementation on large road networks. Among different mono-dimensional time-series models, a seasonal autoregressive moving average model (SARMA) is selected for comparison. The time-series model is also used in a hybrid modeling framework to provide the Bayesian network with an a priori estimation of the predicted speed, which is then corrected exploiting the information collected on other links. A large floating car data set on a sub-area of the road network of Rome is used for validation. To account for the variable accuracy of the speed estimated from floating car data, a new error indicator is introduced that relates accuracy of prediction to accuracy of measure. Validation results highlighted that the spatial architecture of the Bayesian network is advantageous in standard conditions, where a priori knowledge is more significant, while mono-dimensional time series revealed to be more valuable in the few cases of non-recurrent congestion conditions observed in the data set. The results obtained suggested introducing a supervisor framework that selects the most suitable prediction depending on the detected traffic regimes.     

A novel passenger flow prediction model using deep learning methods

Currently, deep learning has been successfully applied in many fields and achieved amazing results. Meanwhile, big data has revolutionized the transportation industry over the past several years. These two hot topics have inspired us to reconsider the traditional issue of passenger flow prediction. As a special structure of deep neural network (DNN), an autoencoder can deeply and abstractly extract the nonlinear features embedded in the input without any labels. By exploiting its remarkable capabilities, a novel hourly passenger flow prediction model using deep learning methods is proposed in this paper. Temporal features including the day of a week, the hour of a day, and holidays, the scenario features including inbound and outbound, and tickets and cards, and the passenger flow features including the previous average passenger flow and real-time passenger flow, are defined as the input features. These features are combined and trained as different stacked autoencoders (SAE) in the first stage. Then, the pre-trained SAE are further used to initialize the supervised DNN with the real-time passenger flow as the label data in the second stage. The hybrid model (SAE-DNN) is applied and evaluated with a case study of passenger flow prediction for four bus rapid transit (BRT) stations of Xiamen in the third stage. The experimental results show that the proposed method has the capability to provide a more accurate and universal passenger flow prediction model for different BRT stations with different passenger flow profiles.     

Linking an activity-based travel demand model with traffic emission and dispersion models: Transport’s contribution to air pollution in Toronto

This paper describes the development of an integrated approach for assessing ambient air quality and population exposure as a result of road passenger transportation in large urban areas. A microsimulation activity-based travel demand model for the Greater Toronto Area – the Travel Activity Scheduler for Household Agents – is extended with capabilities for modelling and mapping of traffic emissions and atmospheric dispersion. Hourly link-based emissions and zone-based soak emissions were estimated. In addition, hourly roadway emissions were dispersed at a high spatial resolution and the resulting ambient air concentrations were linked with individual time-activity patterns derived from the model to assess person-level daily exposure. The method results in an explicit representation of the temporal and spatial variation in emissions, ambient air quality, and population exposure.     

Probabilistic multi-aircraft conflict detection approach for trajectory-based operation

Conflict detection (CD) is one of the key functions used to ensure air transport safety and efficiency. In trajectory-based operation (TBO), aircraft are provided with more flexibility in en route trajectory planning and more responsibility for self-separation. The high flexibility in trajectory planning enables random changes in pilot intent, thus increasing the uncertainty in trajectory prediction and CD. This study proposes a novel probabilistic CD approach for TBO in which the uncertainty of pilot intent is taken into account by quantifying the aircraft reachable domain constrained by the flight plan. First, a probabilistic model for aircraft trajectory prediction is developed using the truncated Brownian bridge method. Based on this model, a novel conflict probability estimation method is developed. Finally, the performance of the proposed probabilistic CD approach is demonstrated through an illustrative air traffic scenario.     

Short-term traffic flow prediction using time-varying Vasicek model

This paper provides a two-step approach based on the stochastic differential equations (SDEs) to improve short-term prediction. In the first step of this framework, a Hull-White (HW) model is applied to obtain a baseline prediction model from previous days. Then, the extended Vasicek model (EV) is employed for modeling the difference between observations and baseline predictions (residuals) during an individual day. The parameters of this time-varying model are estimated at each sample using the residuals in a short duration of time before the time point of prediction; so it provides a real time prediction. The extracted model recovers the valuable local variation information during each day. The performance of our method in comparison with other methods improves significantly in terms of root mean squared error (RMSE), mean absolute error (MAE) and mean relative error (MRE) for real data from Tehran’s highways and the open-access PeMS database. We also demonstrate that the proposed model is appropriate for imputing the missing data in traffic dataset and it is more efficient than the probabilistic principal component analysis (PPCA) and k-Nearest neighbors (k-NN) methods.     

Enhanced, multi criteria based site selection to measure mobile source toxic air pollutants

A multi-criteria-based site selection process that was developed for the Las Vegas near-road study is used to select the most appropriate near-road measurement sites in the Detroit area. The study measures particulate matter with aerodynamic diameters ?2.5 μm and mobile source air toxic compounds, and to document the rationale for selecting or rejecting specific sites. An application of this multi-criteria decision analysis may be the recent US Environmental Protection Agency rule requiring the siting of NO₂ monitors within 50 m of major roads as well as siting for the measurement of community-wide NO₂ concentrations.     

Adding temporal information to direct-demand models: Hourly estimation of bicycle and pedestrian traffic in Blacksburg,VA

Cycling and walking are environmentally-friendly transport modes, providing alternatives to automobility. However, exposure to hazards (e.g., crashes) may influence the choice to walk or cycle for risk-averse populations, minimizing non-motorized travel as an alternative to driving. Most models to estimate non-motorized traffic volumes (and subsequently hazard exposure) are based on specific time periods (e.g., peak-hour) or long-term averages (e.g., Annual Average Daily Traffic), which do not allow for estimating hazard exposure by time of day. We calculated Annual Average Hourly Traffic estimates of bicycles and pedestrians from a comprehensive traffic monitoring campaign in a small university town (Blacksburg, VA) to develop hourly direct-demand models that account for both spatial (e.g., land use, transportation) and temporal (i.e., time of day) factors. We developed two types of models: (1) hour-specific models (i.e., one model for each hour of the day) and (2) a single spatiotemporal model that directly incorporates temporal variables. Our model results were reasonable (adj-R² for the hour-specific [spatiotemporal] bicycle model: ∼0.47 [0.49]; pedestrian model: ∼0.69 [0.72]). We found correlation among non-motorized traffic, land use (e.g., population density), and transportation (e.g., on-street facility) variables. Temporal variables had a similar magnitude of correlation as the spatial variables. We produced spatial estimates that vary by time of day to illustrate spatiotemporal traffic patterns for the entire network. Our temporally-resolved models could be used to assess exposure to hazards (e.g. air pollution, crashes) or locate safety-related infrastructure (e.g., striping, lights) based on targeted time periods (e.g., peak-hour, nighttime) that temporally averaged estimates cannot.     

Estimating near-road pollutant dispersion: A model inter-comparison

A model inter-comparison study to assess the abilities of steady-state Gaussian dispersion models to capture near-road pollutant dispersion has been carried out with four models (AERMOD, run with both the area-source and volume-source options to represent roadways, CALINE, versions 3 and 4, ADMS and RLINE). Two field tracer studies are used: the Idaho Falls tracer study and the Caltrans Highway 99 tracer study. Model performance measures are calculated using concentrations (observed and estimated) that are paired in time and space, since many of the health related questions involve outcomes associated with spatially and temporally distributed human activities. All four models showed an ability to estimate the majority of downwind concentrations within a factor of two of the observations. RLINE, AERMOD-V, and ADMS, also have the capability to predict concentrations upwind of the roadway that result from low-speed meandering of the plume. Generally, RLINE, ADMS, and AERMOD (both source types) had overall performance statistics that were broadly similar, while CALINE 3 and 4 both produced a larger degree of scatter in their concentration estimates. The models performed best for near-neutral conditions in both tracer studies, but had mixed results under convective and stable conditions.     

Temporal and spatial variation in allocating annual traffic activity across an urban region and implications for air quality assessments

Patterns of traffic activity, including changes in the volume and speed of vehicles, vary over time and across urban areas and can substantially affect vehicle emissions of air pollutants. Time-resolved activity at the street scale typically is derived using temporal allocation factors (TAFs) that allow the development of emissions inventories needed to predict concentrations of traffic-related air pollutants. This study examines the spatial and temporal variation of TAFs, and characterizes prediction errors resulting from their use. Methods are presented to estimate TAFs and their spatial and temporal variability and used to analyze total, commercial and non-commercial traffic in the Detroit, Michigan, U.S. metropolitan area. The variability of total volume estimates, quantified by the coefficient of variation (COV) representing the percentage departure from expected hourly volume, was 21%, 33%, 24% and 33% for weekdays, Saturdays, Sundays and holidays, respectively. Prediction errors mostly resulted from hour-to-hour variability on weekdays and Saturdays, and from day-to-day variability on Sundays and holidays. Spatial variability was limited across the study roads, most of which were large freeways. Commercial traffic had different temporal patterns and greater variability than non-commercial vehicle traffic, e.g., the weekday variability of hourly commercial volume was 28%. The results indicate that TAFs for a metropolitan region can provide reasonably accurate estimates of hourly vehicle volume on major roads. While vehicle volume is only one of many factors that govern on-road emission rates, air quality analyses would be strengthened by incorporating information regarding the uncertainty and variability of traffic activity.     

Investigating effects of asphalt pavement conditions on traffic accidents in Tennessee based on the pavement management system (PMS)

Pavement maintenance is essential for ensuring good riding quality and avoiding traffic congestion, air pollution, and accidents. Improving road safety is one of the most important objectives for pavement management systems. This study utilized the Tennessee Pavement Management System (PMS) and Accident History Database (AHD) to investigate the relationship between accident frequency and pavement distress variables. Focusing on four urban interstates with asphalt pavements, divided median types, and 55 mph speed limits, 21 Negative Binomial Regression models were developed for predicting various types of traffic accident frequencies based on different pavement condition variables, including rut depth (RD), International Roughness Index (IRI), and Present Serviceability Index (PSI). The modeling results indicated that the RD models did not perform well, except for predicting accidents at night and accidents under rain weather conditions; whereas, IRI and PSI were always significant prediction variables in all types of accident models. Comparing the models goodness‐of‐fit results, it was found that the PSI models had a better performance in crash frequency prediction than the RD models and IRI models. This study suggests that the PSI accident prediction models should be considered as a comprehensive approach to integrate the highway safety factors into the pavement management system. Copyright © 2010 John Wiley & Sons, Ltd.     

Estimating traffic volumes on intercity road locations using roadway attributes,socioeconomic features and other work-related activity characteristics

Traffic volume data are key inputs to many applications in highway design and planning. But these data are collected in only a limited number of road locations due to the cost involved. This paper presents an approach for estimating daily and hourly traffic volumes on intercity road locations combining clustering and regression modelling techniques. With the aim of applying the procedure to any road location, it proposes the use of roadway attributes and socioeconomic characteristics of nearby cities as explanatory variables, together with a set of previously discovered patterns with the hourly traffic percent distribution. Test results show that the proposed approach significantly produces accurate estimates of daily volumes for most locations. The accuracy at hourly level is a bit more reduced but, for periods when traffic is significant, more than half of the estimates are within 20% of absolute percentage error. Moreover, the main peak period is approximately identified for most cases. These findings together with its great applicability make this approach attractive for planners when no traffic data are available and an estimate is helpful.     

A hybrid machine learning model for short-term estimated time of arrival prediction in terminal manoeuvring area

4D trajectory prediction is the core element of future air transportation system, which is intended to improve the operational ability and the predictability of air traffic. In this paper, we introduce a novel hybrid model to address the short-term trajectory prediction problem in Terminal Manoeuvring Area (TMA) by application of machine learning methods. The proposed model consists of two parts: clustering-based preprocessing and Multi-Cells Neural Network (MCNN)-based prediction. Firstly, in the preprocessing part, after data cleaning, filtering and data re-sampling, we applied principal Component Analysis (PCA) to reduce the dimension of trajectory vector variable. Then, the trajectories are clustered into several patterns by clustering algorithm. Using nested cross validation, MCNN model is trained to find out the appropriate prediction model of Estimated Time of Arrival (ETA) for each individual cluster cell. Finally, the predicted ETA for each new flight is generated in different cluster cells classified by decision trees. To assess the performance of MCNN model, the Multiple Linear Regression (MLR) model is proposed as the comparison learning model, and K-means++ and DBSCAN are proposed as two comparison clustering models in preprocessing part. With real 4D trajectory data in Beijing TMA, experimental results demonstrate that our proposed model MCNN with DBSCAN in preprocessing is the most effective and robust hybrid machine learning model, both in trajectory clustering and short-term 4D trajectory prediction. In addition, it can make an accurate trajectory prediction in terms of Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE) with regards to comparison models.     

A gradient boosting logit model to investigate driver’s stop-or-run behavior at signalized intersections using high-resolution traffic data

Driver’s stop-or-run behavior at signalized intersection has become a major concern for the intersection safety. While many studies were undertaken to model and predict drivers’ stop-or-run (SoR) behaviors including Yellow-Light-Running (YLR) and Red-Light-Running (RLR) using traditional statistical regression models, a critical problem for these models is that the relative influences of predictor variables on driver’s SoR behavior could not be evaluated. To address this challenge, this research proposes a new approach which applies a recently developed data mining approach called gradient boosting logit model to handle different types of predictor variables, fit complex nonlinear relationships among variables, and automatically disentangle interaction effects between influential factors using high-resolution traffic and signal event data collected from loop detectors. Particularly, this research will first identify a series of related influential factors including signal timing information, surrounding traffic information, and surrounding drivers’ behaviors using thousands drivers’ decision events including YLR, RLR, and first-to-stop (FSTP) extracted from high-resolution loop detector data from three intersections. Then the research applies the proposed data mining approach to search for the optimal prediction model for each intersection. Furthermore, a comparison was conducted to compare the proposed new method with the traditional statistical regression model. The results show that the gradient boosting logit model has superior performance in terms of prediction accuracy. In contrast to other machine learning methods which usually apply ‘black-box’ procedures, the gradient boosting logit model can identify and rank the relative importance of influential factors on driver’s stop-or-run behavior prediction. This study brings great potential for future practical applications since loops have been widely implemented in many intersections and can collect data in real time. This research is expected to contribute to the improvement of intersection safety significantly.     

Long-memory characteristics of urban roadside air quality

Carbon monoxide is a major contributor to air pollution in urban cities, particularly at the roadside. Hourly, monthly and seasonal mean carbon monoxide concentration data are collected from a roadside air monitoring station in Hong Kong over 7-years. The station is a few metres from a major intersection surrounded by tall buildings. In particular, hourly patterns of concentrations on different days of the week are investigated. The data show that hourly carbon monoxide concentrations resemble the traffic pattern of the area and tend to be lower in the summer. Using a seasonal autoregressive integrated moving average models shows that the daily traffic cycle strongly influences concentrations. Further, it is found that urban roadside carbon monoxide monitoring data exhibits a long-memory process, suggesting that a model incorporating long memory and seasonality effects is needed simulate urban roadside air quality.