Dispersion Of Airborne Pollutants Research Articles

Rapid prediction for the transient dispersion of leaked airborne pollutant in urban environment based on graph neural networks

Rapidly predicting airborne pollutant dispersion in urban is vital for ventilation design and evacuation planning. Computational fluid dynamics (CFD) simulations are commonly used to provide accurate predictions, but the computational cost is too high. Although graph neural networks (GNNs) provide fast predictions of flow fields by manipulating unstructured mesh on GPU, they suffer from high memory usage and accuracy decreases when applied to large-scale urban scenes. Moreover, it is difficult for GNNs to learn the coupled relationship between wind field and pollutant concentration field. We propose a multi-objective GNN model as CFD surrogate to rapidly predict the transient dispersion of airborne pollutant under the influence of complex wind field patterns in urban environment. Based on random urban layouts generated by a 2D bin packing algorithm, we employ a validated CFD model to construct a sample dataset of wind fields and concentration fields. We leverage graph pooling and multi-scale feature fusion to improve prediction accuracy, and subgraph partitioning of both wind field and concentration field to reduce GPU memory usage. The results show that our GNN model at its best runs 1–2 orders of magnitude faster than CFD simulation with accuracy evaluation metrics R2=0.92, and achieves 70 % GPU memory reduction.

Deep learning methods evaluation to predict air quality based on Computational Fluid Dynamics

Air quality is a major health issue for densified cities nowadays. To evaluate and act upon it, modeling alongside sensors has proved to be a powerful tool. Among the different available models, Computational Fluid Dynamics (CFD) has proved to be formidable to evaluate airborne pollutant dispersion locally in urban areas since it is able to consider buildings and others complexes phenomenon at the scale of the meter. Nevertheless, this method has a major drawback, it is computationally expensive and cannot be applied in real time or over large areas. To overcome this issue, several state-of-the-art deep learning methods to treat spatial information have been trained based on CFD results to predict airborne pollutant dispersion. Among these models, multiResUnet architecture was proved to be the best on overall over seven metrics. It managed to have two out of three air quality metrics within satisfactory range for a good air quality model. These results are obtained in a mere matter of tens of seconds against several hours for CFD.

Generalizability evaluation of k-ε models calibrated by using ensemble Kalman filtering for urban airflow and airborne contaminant dispersion

The low accuracy of Reynolds-averaged Navier-Stokes (RANS) modeling of urban airflow and airborne pollutant dispersion is attributed to model flaws and uncertainty contributed by closure coefficients. Previous studies have attempted to improve the performance of RANS modeling by ad hoc calibration of the turbulence model coefficients specifically for urban problems. However, these models failed to accurately reproduce the key features like the reattachment lengths. In addition, there was a lack of generalizability evaluations of the calibrated models. To optimize the accuracy and generalizability of the turbulence model, this study considered the effects of optimization objectives and model formulations. Six k-ε based turbulence models calibrated using the ensemble Kalman filtering (EnKF) approach were compared. A wind tunnel experiment consisting of key features of the airflow around a single-building model was conducted to provide the training data. The proposed optimization objective prioritizing the reproduction of the reattachment lengths in the calibrations enabled the calibrated models to capture the key features of the airflow with better accuracy. The six turbulence models before and after the calibrations were compared for a single-building test case and a building-array test case with respect to the reattachment lengths, velocity, turbulence kinetic energy, and airborne contaminant concentration. The calibrated models with different formulations exhibit distinctive generalizability. The calibrated Murakami-Mochida-Kondo k-ε model (MMK) exhibited strong potential for generalizability to single-building problems. However, the generalization from the single-building isolated roughness flow regime to the street canyon skimming flow regime is limited.

Centrality metric for the vulnerability of urban networks to toxic releases.

The dispersion of airborne pollutants in the urban atmosphere is a complex, canopy-driven process. The intricate structure of the city, the high number of potential sources, and the large spatial domain make it difficult to predict dispersion patterns, to simulate a great number of scenarios, and to identify the high-impact emission areas. Here we show that these complex transport dynamics can be efficiently characterized by adopting a complex network approach. The urban canopy layer is represented as a complex network. Street canyons and their intersections shape the spatial structure of the network. The direction and the transport capacity of the flow in the streets define the direction and the weight of the links. Within this perspective, pollutant contamination from a source is modeled as a spreading process on a network, and the most dangerous areas in a city are identified as the best spreading nodes. To this aim, we derive a centrality metric tailored to mass transport in flow networks. By means of the proposed approach, vulnerability maps of cities are rapidly depicted, revealing the nontrivial relation between urban topology, transport capacity of the street canyons, and forcing of the external wind. The network formalism provides promising insight in the comprehensive analysis of the fragility of cities to air pollution.

Effect of pollutant source location on air pollutant dispersion around a high-rise building

This article investigates the dispersion of airborne pollutants emitted from different locations near a high-rise building. A Computational Fluid Dynamics (CFD) model for simulating the wind flow field and the pollutant dispersion was developed and validated by wind tunnel data. Then the spreading of the pollutant emitted from different locations to a rectangular-shaped high-rise residential (HRR) building was numerically studied. The pollutant source location was set in a wide range of the position angle and distance between the source and the building. It was found that the pollutant concentration on the building decreases with an increase in the emission distance whereas the effect of the position angle is more complicated. Interestingly, there is a critical range of the position angle from which the emitted pollutants will not spread to the building in a significant way. The effect of the source location was linked to the wind flow field around the building, particularly with several major flows. The vertical distributions of the pollutant concentration on different faces were also investigated, and it was found that these are more affected by the vertical flow near each face. Finally, a mathematical model was developed to evaluate the pollutant concentration as a function of the emission distance and position angle. These findings are helpful to the understanding of the dispersion of airborne pollutants around high-rise buildings and the related hazard management in urban design.

Street canyon ventilation and airborne pollutant dispersion: 2-D versus 3-D CFD simulations

Urban ventilation is important for building a healthy urban living environment. 2-D CFD simulation has been used widely for street canyon ventilation due to its high computational efficiency, but its applicability for a 3-D simulation has never been studied. This paper tried to answer the question: if and under what conditions, the widely-adopted 2-D CFD simulations on street canyon ventilation can represent a 3-D scenarios? 3-D simulations on street canyons with various street lengths and corresponding 2-D simulations are carried out with RNG k-ε model. Our study identified two important ventilation mechanism for controlling ventilation and dispersion in a 3-D street canyon, i.e., canyon vortex on the canyon top and the corner vortices at the street ends. The relative importance of these two driving forces will change with the street length/street width ratio (B/W). For isolated street canyon, when B/W is higher than 20 (for H/W = 1) and 70 (H/W = 2), the street canyon ventilation will be dominated by canyon vortex, and 3-D street canyon ventilation could be simplified as a 2-D case. For multiple street canyon, the threshold of B/W will become 20 when H/W = 1, and 50 when H/W = 2. The findings in this study could improve our approaches for simulating urban ventilation.

Estimating distributions of airborne contaminants released by sources with dynamic strength and dynamic location by a probabilistic model

Characteristics of contaminant release source usually contribute significant impacts to dispersion of airborne pollutants. It is necessary and also desirable to grasp detailed information of contaminant distributions under various release patterns of contaminant sources. This study developed a Markov chain based probabilistic model to simulate transport and dispersion of airborne contaminants released by sources with dynamic strengths and locations. The proposed model takes the dynamic source strengths and locations into account by replacing elements of the state vector with values in accord with specific functions, which indicate strength and location variations. The particle phase simulations need to be initialized by comprehensive data of the flow field, which can be solved by CFD tools in advance. For the first step, the developed model is validated by experiment data from a 2D turbulence duct flow with a particulate source instantaneously releasing airborne particles. Then the new model is adopted to simulate contaminant dispersion of gaseous (or particles with small diameters) contaminant released by sources with dynamic strength and location. Results indicate that the proposed model is capable of predicting the behaviors of contaminant sources under various release patterns. The effects of dynamic contaminant sources on downstream regions in unidirectional flow are limited by the distances between the target region and source location. Accuracy of the present model is significantly affected by temporal intervals (the specified time step Δt) for sources with dynamic strengths and spatial intervals (computational grid scales) for sources with dynamic locations.

Room airborne pollutant separation by the use of air curtains in the large building enclosure: Infiltration efficiency and partial enclosure ventilation rate

Suppression of airborne pollutant infiltration into the occupation region by the air curtain was numerically studied in the present work. The computational analysis was depending on the two-dimensional conservation equations of mass, momentum and species concentration for the incompressible isothermal turbulent flow with a standard k-ε two-equation turbulent model adopted for turbulence closure. Time averaged turbulent stream-functions and species-functions were respectively defined to vividly visualize the fluid flow and pollutant dispersion. The effects of ejection velocity, air curtain width and enclosure height on the turbulent fluid flow and the sealing performance of the air curtain were investigated. Detailed correlations of air curtain sealing efficiency with those governing parameters were presented through multiple linear regression analysis. Numerical results indicate that air curtain sealing efficiency is heavily influenced by the enclosure height, while little by the air curtain width. Present investigation could benefit the future development of air curtain implemented in the control of airborne pollutant dispersion in the large building spaces.

The Impact of Planting Trees on NOx Concentrations: The Case of the Plaza de la Cruz Neighborhood in Pamplona (Spain)

In this paper, the role of trees on airborne pollutant dispersion in a real neighborhood in Pamplona (Spain) is discussed. A Computational Fluid Dynamics (CFD) model is employed and evaluated against concentrations measured during the last part of winter season at a monitoring station located in the study area. Aerodynamic and deposition effects of trees are jointly considered, which has only been done in few recent studies. Specifically, the impact on NOx concentration of: (a) tree-foliage; and (b) introducing new vegetation in a tree-free street is analyzed considering several deposition velocities and Leaf Area Densities (LAD) to model deciduous and evergreen vegetation. Results show that the higher the LAD, the higher the deposition (concentration reduction) and the blocking aerodynamic effect (concentration increase). Regardless of foliage or deposition rates, results suggest the predominance of aerodynamic effects which induce concentration increases up to a maximum of 7.2%, while deposition induces concentration decreases up to a maximum of 6.9%. The inclusion of new trees in one street modifies the distribution of pollutant, not only in that street, but also in nearby locations with concentration increase or decrease. This finding suggests that planting trees in street with traffic as an air pollution reduction strategy seems to be not appropriate in general, highlighting the necessity of ad hoc studies for each particular case to select the suitable location of new vegetation.

Fluid mechanical dispersion of airborne pollutants inside urban street canyons subjecting to multi-component ventilation and unstable thermal stratifications

The pedestrian level pollutant transport in street canyons with multiple aspect ratios (H/W) is numerically investigated in the present work, regarding of various unstable thermal stratification scenarios and plain surrounding. Non-isothermal turbulent wind flow, temperature field and pollutant spread within and above the street canyons are solved by the realizable k-ε turbulence model along with the enhanced wall treatment. One-vortex flow regime is observed for shallow canyons with H/W=0.5, whereas multi-vortex flow regime is observed for deep canyons with H/W=2.0. Both one-vortex and multi-vortex regimes could be observed for the street canyons with H/W=1.0, where the secondary vortex could be initiated by the flow separation and intensified by unstable thermal stratification. Air exchange rate (AER) and pollutant retention time are adopted to respectively evaluate the street canyon ventilation and pollutant removal performance. A second-order polynomial functional relationship is established between AER and Richardson number (Ri). Similar functional relationship could be established between retention time and Ri, and it is only valid for canyons with one-vortex flow regime. In addition, retention time could be prolonged abruptly for canyons with multi-vortex flow regime. Very weak secondary vortex is presented at the ground level of deep canyons with mild stratification, where pollutants are highly accumulated. However, with the decrease of Ri, pollutant concentration adjacent to the ground reduces accordingly. Present research could be applied to guide the urban design and city planning for enhancing pedestrian environment.

Prediction of the upper tail of concentration distributions of a continuous point source release in urban environments

The peak values observed in a measured concentration time series of a dispersing gaseous pollutant released continuously from a point source in urban environments, and the hazard level associated with them, demonstrate the necessity of predicting the upper tail of concentration distributions. For the prediction of concentration distributions statistical models are preferably employed which provide information about the probability of occurrence. In this paper a concentration database pertaining to a field experiment is used for the selection of the statistical distribution. The inverses of the gamma cumulative distribution function (cdf) for 75th–99th percentiles of concentration are found to be more consistent with the experimental data than those of the log-normal distribution. The experimental values have been derived from measured high frequency time series by sorting first the concentrations and then finding the concentration which corresponds to each probability. Then the concentration mean and variance that are predicted with Computational Fluid Dynamics-Reynolds Averaged Navier–Stokes (RANS) methodology are used to construct the gamma distribution. The proposed model (“RANS-gamma”) is included in the framework of a computational code (ADREA-HF) suitable for simulating the dispersion of airborne pollutants over complex geometries. The methodology is validated by comparing the inverses of the model cdfs with the observed ones from two wind tunnel experiments. The evaluation is performed in the form of validation metrics such as the fractional bias, the normalized mean square error and the factor-of-two percentage. From the above comparisons it is concluded that the overall model performance for the present cases is satisfactory.

Assessment of Pollutant Dispersion in the Re-entrance Space of a High-rise Residential Building, Using Wind Tunnel Simulations

This paper investigates the likely route of dispersion of airborne pollutants and pathogenic agents in the re-entrance space of a high-rise residential (HRR) building in Hong Kong. A 1 : 150 scaled HRR building, a 30 storeys tall model was tested in a wind tunnel with simulated atmospheric boundary layer flow. Tracer gas, propane released from the lower, middle and upper parts of the building, was measured by flame ionisation detection and this was used to simulate the air pollutant dispersion within the re-entrance space. There were a total of 495 measurements, tested under five different wind directions at 99 measurement points on windows locations of every floor of the modelled building. The tracer gas concentration profiles obtained would illustrate the pollutant dispersion and transmission routes under the different wind effect. The results showed that the airborne pollutants or pathogenic agents could be dispersed mainly along the vertical routes, posing a potential risk to the adjacent neighbours immediately above the source location. Wind direction could also affect the dispersion. The findings of this study should inform the development of ventilation strategies for HRR buildings and would guide building design to minimise transmission of airborne diseases or harmful pollutants within buildings.

Predicting Gaseous Pollutant Dispersion Around a Workplace

Predicting the space-time evolution of a gaseous or particulate pollutant concentration in a ventilated room where a process operation is performed is imperative in hazardous activities, such as chemical or nuclear ones. This study presents a prediction of the space-time evolution of airborne pollutant dispersion following the accidental rupture of a containment enclosure (fume cupboard, glove box, pressurized gas duct, etc.). The final model is written as correlations inspired by the free turbulent jet theory, giving the space-time evolution of a pollutant concentration c (x,y,z,t) that has been formulated as a correlated function of various parameters: leak geometry (slot or round opening), emission type (continuous or transient), emission duration and initial emission velocity. These correlations are based on gas tracing experiments and on multidimensional simulations using computational fluid dynamics (CFD) tools. An instrumented experimental facility was used to simulate pressurized gas industrial failure, and the measurements performed gave the real-time evolution of a tracer gas concentration. Transient leak simulations were run in parallel with a CFD code. Comparisons between experimental and numerical results largely agree. A semiempirical model was built using a methodical parametric study of all the simulation results. This model is easy to use in safety evaluations of radioactive material containment and radiological protection inside nuclear facilities and for evaluating toxic gaseous compounds in the chemical industry.

Parallel Implementation of a Lagrangian Stochastic Model for Pollutant Dispersion

Lagrangian dispersion models have shown to be effective and reliable tools for simulating the airborne pollutant dispersion. However, the main drawback for their use as regulatory models is the associated high computational costs. Consequently, in this paper a parallel version of a Lagrangian particle model LAMBDA--is developed using the MPI message passing communication library. Performance tests were executed in a distributed memory parallel machine, a multicomputer based on IA-32 architecture. Portions of the pollutant in the air emitted from its source are simulated as fictitious particles whose trajectories evolve under stochastic forcing. This yields independent evolution equations for each particle of the model that can be computed by a different processor in a parallel implementation. Speed-up results show that the parallel implementation is suitable for the used architecture.

An overview about the study of downwash effects on dispersion of airborne pollutants

The aim of the present work is to give an overview about the state of the art of the study of the downwash phenomena of airborne pollutants. Accounting for the downwash of pollutant dispersion is of interest because it can contribute to the prevention of dangerous situations by determining in advance what configuration of buildings, stacks, and effluents could cause a high concentration of harmful effluents in a particular area. Recent and less recent studies concerning both building and stack downwash are presented. Some of the presented models are well established and implemented in regulatory air pollutions codes, while other ones are more sophisticated and still under development.

Methodology of model testing and application to dispersion simulation above complex terrain

This paper concerns an integrated model testing exercise focussed on the use of model performance statistical indices. SAFE_AIR is a multisource model mainly designed to describe dispersion of airborne pollutants from point sources at local-to-regional scales over complex terrain in non-stationary meteorological and emission conditions. It was applied in the region of Ilo, Peru, a complex coastal area where a tracer (SF6 ) was emitted from a copper smelter plant to collect field data about pollutant dispersion in the area. To compare measured concentrations with predicted concentrations, the statistical indices suggested in the Model Evaluation Kit (MEAN, SIGMA, BIAS, NMSE, COR, FA2, FB and FS) and two less common ones (WNNR and NNR) were used.

Influence of valley winds on transport and dispersion of airborne pollutants in the Freiburg‐Schauinsland area

The change of the concentration of trace gases in a natural environment depends on both chemical and meteorological conditions. To quantify the contribution of meteorological processes, such as transport and dispersion, chemical and meteorological processes must be assessed separately. In June 1996 the Schauinsland Ozone Precursor Experiment (SLOPE) was carried out to investigate the behavior of ozone and its precursors as a function of time in a valley between Freiburg in the Rhine Valley and Schauinsland in the Black Forest [Volz‐Thomas et al., this issue]. The main purpose of this contribution is to calculate the influence of transport and dispersion on the change of the concentration of airborne pollutants in a propagating air mass. Therefore (1) an inert tracer, SF6, was released, and (2) meteorological measurements were performed to investigate the valley wind regimes and the evolution of the mixed layer. Here are the main results of the experiment: Clear‐sky conditions caused upvalley winds to be generated in the various valleys in the morning. These upvalley winds led to the transport of polluted air and SF6 from Freiburg through the Zarten Basin and the side valley(Großes Tal) up to the Schauinsland hill. The propagation through the Großes Tal took 90± 5 min, which gives a propagation rate of approximately 1.3 m s−1. The dilution factor for SF6 between the entrance and the outlet of the valley was approximately 9. This dilution factor could not only be explained by the continuous growth of the boundary layer during the propagation time of the tracer from the entrance to the outlet of the valley, but additionally by flow splitting and mountain venting [Fiedler et al., this issue].

Numerical modeling of three-dimensional flow and pollutant dispersion around structures

This paper describes a numerical modeling approach that can be used to provide estimates of emissions at industrial sites. In particular the models presented are capable of simulating the wind flow and dispersion of airborne pollutants around surface-mounted structures such as buildings or building complexes. The calculational procedure in this approach consists of two sequential steps, namely: (i) prediction of the mean flow via a turbulent flow model; and (ii) employment of the calculated flow field to drive a particle-in-cell transport and diffusion model. A benchmark simulation is performed in which numerical results from the flow model are compared with other numerical models and with experimental data for flow over a backward-facing step. Results from three-dimensional simulations of flow and dispersion over a two-building complex are also presented.

Simulation of atmospheric diffusion in low windspeed meandering conditions by a Monte Carlo dispersion model

Conditions of low windspeed associated to strong stability and air stagnation correspond to highly nonstationary and inhomogeneous diffusion situations. In such «poor» diffusion cases, integrated models (like Gaussian models) are no longer valid. New numerical methodologies, like Monte Carlo methods, are needed. Recently a Monte Carlo particle model for airborne pollutant dispersion has been developed by our team. Its ability to predict ground-level concentrations in convective and neutral conditions with a high degree of accuracy was verified against tracer experiments. In the present paper our model is used to simulate the dispersion under low windspeed conditions. Particular emphasis is given to the problem of wind meandering. We have demonstrated the need of recording the wind statistics over averaging periods of the order of a few minutes and anad hoc algorythm to treat low windspeed situation when only hourly average data are available is suggested. The model results are compared to tracer data obtained by the Idaho National Engineering Laboratory (U.S.A.) in 1974.

Comparison between the results of a Monte Carlo atmospheric diffusion model and tracer experiments

A Lagrangian statistical (Monte Carlo) model for airborne pollutant dispersion is presented. Its ability to simulate the atmospheric dispersion both in homogeneous and inhomogeneous turbulence by comparison with an analytical solution and with the Willis and Deardorff water tank experiments, respectively, has been stated in previous papers. In the present paper the model is used to simulate dispersion in the real atmospheric PBL. The numerical results obtained are verified against experimental data from the Karlsruhe Nuclear Research Center tracer experiments. The model is applied to the problem of predicting the ground level concentration of two different tracers simultaneously released from two heights (160 and 195 m) at the Karlsruhe meteorological tower. Convectively unstable and neutral conditions were prevailing during the two tracer experiments which have been simulated. Model performance was evaluated through two statistical indexes: relative mean bias and normalized mean square error. The cumulative frequency distribution of the point-by-point ratio between observed and predicted ground level concentrations (glcs) was also computed. The simulated concentrations agree very well with observations. The tracer data were also compared to the simulations of 10 Gaussian models. They differed one another for the choice of dispersion sigma curves and for the way to insert the wind speed and direction. None of them proved to perform better than our particle model in all the exercises.