A comparison of operational Lagrangian particle and adaptive puff models for plume dispersion forecasting

Transport and dispersion of pollutants in the lower atmosphere are predicted by using both a Lagrangian particle model (LPM) and an adaptive puff model (APM2) coupled to the same mesoscale meteorological prediction model PMETEO. LPM and APM2 apply the same numerical solutions for plume rise; but, for advection and plume growth, LPM uses a stochastic surrogate to the pollutant conservation equation, and APM2 applies interpolated winds and standard deviations from the meteorological model, using a step-wise Gaussian approach. The results of both models in forecasting the SO₂ ground level concentration (glc) around the 1400 MWe coal-fired As Pontes Power Plant are compared under unstable conditions. In addition, meteorological and SO₂ glc numerical results are compared to field measurements provided by 17 fully automated SO₂ glc remote stations, nine meteorological towers and one Remtech PA-3 SODAR, from a meteorological and air quality monitoring network located 30 km around the power plant.     

Evaluation of the effect of meteorological data resolution on Lagrangian particle dispersion simulations using the ETEX experiment

This paper presents results from a series of numerical experiments designed to evaluate operational long-range dispersion model simulations, and to investigate the effect of different temporal and spatial resolution of meteorological data from numerical weather prediction models on these simulations. Results of Lagrangian particle dispersion simulations of the first tracer release of the European Tracer Experiment (ETEX) are presented and compared with measured tracer concentrations. The use of analyzed data of higher resolution from the European Center for Medium-Range Weather Forecasts (ECMWF) model produced significantly better agreement between the concentrations predicted with the dispersion model and the ETEX measurements than the use of lower resolution Navy Operational Global Atmospheric Prediction System (NOGAPS) forecast data. Numerical experiments were performed in which the ECMWF model data with lower vertical resolution (4 instead of 7 levels below 500 mb), lower temporal resolution (12 h instead of 6 h intervals), and lower horizontal resolution (2.5° instead of 0.5°) were used. Degrading the horizontal or temporal resolution of the ECMWF data resulted in decreased accuracy of the dispersion simulations. These results indicate that flow features resolved by the numerical weather prediction model data at approximately 45 km horizontal grid spacing and 6 h time intervals, but not resolved at 225 km spacing and 12 h intervals, made an important contribution to the long-range dispersion.     

A new Lagrangian particle model for the simulation of dense gas dispersion

A Lagrangian stochastic model (MicroSpray), able to simulate the airborne dispersion in complex terrain and in presence of obstacles, was modified to simulate the dispersion of dense gas clouds. This is accomplished by taking into account the following processes: negative buoyancy, gravity spreading and the particle's reflection at the bottom computational boundary. Elevated and ground level sources, continuous and instantaneous emissions, time varying sources, plumes with initial momentum (horizontal, vertical or oblique in any direction), plumes without initial momentum are considered. MicroSpray is part of the model system MSS, which also includes the diagnostic MicroSwift model for the reconstruction of the 3-D wind field in presence of obstacles and orography. To evaluate the MSS ability to simulate the dispersion of heavy gases, its simulation performances are compared in detail to two field experiments (Thorney Island and Kit Fox) and to a chlorine railway accident (Macdona). Then, a comprehensive analysis considering several experiments of the Modelers Data Archive is presented. The statistical analysis on the overall available data reveals that the performance of the new MicroSpray version for dense gas releases is generally reliable. For instance, the agreement between concentration predictions and observations is within a factor of two in the 72% up to 99% of the occurrences for the case studies considered. The values of other performance measures, such as correlation coefficient, geometric mean bias and geometric variance, mostly set in the ranges indicated as good-model performances in the specialized literature.     

An efficient Lagrangian stochastic model of vertical dispersion in the convective boundary layer

We consider the one-dimensional case of vertical dispersion in the convective boundary layer (CBL) assuming that the turbulence field is stationary and horizontally homogeneous. The dispersion process is simulated by following Lagrangian trajectories of many independent tracer particles in the turbulent flow field, leading to a prediction of the mean concentration. The particle acceleration is determined using a stochastic differential equation, assuming that the joint evolution of the particle velocity and position is a Markov process. The equation consists of a deterministic term and a random term. While the formulation is standard, attention has been focused in recent years on various ways of calculating the deterministic term using the well-mixed condition incorporating the Fokker–Planck equation. Here we propose a simple parameterisation for the deterministic acceleration term by approximating it as a quadratic function of velocity. Such a function is shown to represent well the acceleration under moderate velocity skewness conditions observed in the CBL. The coefficients in the quadratic form are determined in terms of given turbulence statistics by directly integrating the Fokker–Planck equation. An advantage of this approach is that, unlike in existing Lagrangian stochastic models for the CBL, the use of the turbulence statistics up to the fourth order can be made without assuming any predefined form for the probability distribution function (PDF) of the velocity. The main strength of the model, however, lies in its simplicity and computational efficiency. The dispersion results obtained from the new model are compared with existing laboratory data as well as with those obtained from a more complex Lagrangian model in which the deterministic acceleration term is based on a bi-Gaussian velocity PDF. The comparison shows that the new model performs well.     

Comparison of PDFs, closure schemes and turbulence parameterisations in Lagrangian stochastic models

Six one-dimensional models, based on the Ito-type stochastic equation, are presented and compared. Four of these take into account up to the fourth order moment of vertical velocity fluctuations, and two up to the third order moment. Four models make use of a bi-Gaussian probability density function (PDF) and the other two are based on a Gram-Charlier series expansion truncated to the third or fourth order. All the models were run with a parameterisation of input turbulence (i.e. w², w³, and τ profiles). Concerning the fourth order moment w⁴, two different parameterisations were considered. Comparisons are made between ground-level concentrations, plume height and plume width observed in the Willis and Deardorff water tank experiments and those predicted by the different models here considered. The goal of this study was to find the models that give greater confidence in their applicability in dispersion studies and to verify the importance of considering the fourth order moment. The main conclusions are: simulation results largely depend on the turbulence parameterisation chosen; the Gram-Charlier PDF gives the best agreement with observations; some combinations of models and turbulence parameterisations perform well in simulating the shape of the ground-level concentration (g.1.c.) trend but fail in correctly simulating the form of the plume (plume height and vertical width); in the case of the Gram-Charlier PDF, the fourth order model reproduced the vertical plume width better than the third order one, whereas the two schemes yielded similar g.1.c. distributions.     

A comparison of three highway line source dispersion models

This paper compares three idealized line source dispersion models that predict carbon monoxide concentrations near highways. The models are EPA HIWAY (Zimmerman, 1974), the original California Line Source (Beaton, 1972), and CALINE2 (Ward, 1975). The comparison includes a sensitivity analysis and model validation. A sensitivity analysis refers to an analysis of the dependence of normalized concentration to variations of several independent input parameters. Model validation is accomplished by comparing measured carbon monoxide concentrations with concentrations predicted by the models. The sensitivity analysis indicates that the EPA HIWAY model predicts higher pollutant concentrations than the two California models for oblique and crosswind cases. For parallel wind conditions, the California Line Source model predicts higher pollutant concentrations. A comparison of predicted and measured concentrations shows that all of the models overestimate concentrations for parallel wind conditions and underestimate concentrations for oblique and crosswind conditions.     

A concentration correction scheme for Lagrangian particle model and its application in street canyon air dispersion modelling

Pollutant dispersion in street canyons with various configurations was simulated by discharging a large number of particles into the computation domain after developing a time-dependent wind field. Trajectory of the released particles was predicted using a Lagrangian particle model developed in an earlier study. A concentration correction scheme, based on the concept of “visibility”, was adopted for the Lagrangian particle model to correct the calculated pollutant concentration field in street canyons. The corrected concentrations compared favourably with those from wind tunnel experiments and a linear relationship between the computed concentrations and wind tunnel data were found. The developed model was then applied to four simulations to test for the suitability of the correction scheme and to study pollutant distribution in street canyons with different configurations. For those cases with obstacles presence in the computation domain, the correction scheme gives more reasonable results compared with the one without using it. Different flow regimes are observed in the street canyons, which depend on building configurations. A counter-clockwise rotating vortex may appear in a two-building case with wind flow from left to right, causing lower pollutant concentration at the leeward side of upstream building and higher concentration at the windward side of downstream building. On the other hand, a stable clockwise rotating vortex is formed in the street canyon with multiple identical buildings, resulting in poor natural ventilation in the street canyon. Moreover, particles emitted in the downstream canyon formed by buildings with large height-to-width ratios will be transported to upstream canyons.     

Shared- and distributed-memory parallelization of a Lagrangian atmospheric dispersion model

This paper describes parallelization of a 3-D Lagrangian stochastic atmospheric dispersion model using both distributed- and shared-memory methods. Shared-memory parallelism is implemented through the use of OpenMP compiler directives. Distributed-memory parallelism relies on the MPI message-passing library. One or both (using MPI for inter-node and OpenMP for intra-node communication) of the parallel modes can be used depending upon the requirements of the problem and the computational platform available. The distributed-memory version achieves a nearly linear decrease in execution time as the number of processors is increased. As the number of particles per processor is lowered, performance is limited by the decrease in work per processor and by the need to produce one set of output files. The shared-memory version achieves a speedup factor of ∼1.4 running on machines with four processors.     

The effects of parametric uncertainties in simulations of a reactive plume using a Lagrangian stochastic model

A combined Lagrangian stochastic model with micro-mixing and chemical sub-models is used to investigate a reactive plume of nitrogen oxides (NO_x) released into a turbulent grid flow doped with ozone (O₃). Sensitivities to the model input parameters are explored for different source NO_x scenarios. The wind tunnel experiments of Brown and Bilger (1996) provide the simulation conditions for the first case study where photolysis reactions are not included and the main uncertainties occur in parameters defining the turbulence scales, source size and reaction rate of NO with O₃. Using nominal values of the parameters from previous studies, the model gives a good representation of the radial profile of the conserved mean scalar ${\overline{\Gamma }}_{{\text{NO}}_{\text{x}}}$ although slightly over predicts peak mean NO₂ concentrations ${\overline{\Gamma }}_{{\text{NO}}_{\text{2}}}$ compared to the experiments. The high dimensional model representation (HDMR) method is used to investigate the effects of uncertainties in model inputs on the simulation of chemical species concentrations. For this scenario, the Lagrangian velocity structure function coefficient has the largest impact on simulated ${\overline{\Gamma }}_{{\text{NO}}_{\text{x}}}$ profiles. Photolysis reactions are then included in a chemical scheme consisting of eight reactions between species NO, O, O₃ and NO₂. Independent and interactive effects of 22 input parameters are studied for two source NO_x scenarios using HDMR, including turbulence parameters, temperature dependant rate parameters, photolysis rates, temperature, fraction of NO in total NO_x at the source and background ozone concentration [O₃]. For this reactive case, the variance in the predicted mean plume centre ${\overline{\Gamma }}_{{\text{O}}_{\text{3}}}$ is caused by parameters describing both physical (mixing time-scale coefficient) and chemical processes (activation energy for the reaction O₃+NO). The variance in predicted plume centre ${\overline{\Gamma }}_{{\text{NO}}_{\text{2}}}$ and root mean square NO₂ concentration ${\gamma }_{{\text{NO}}_{\text{2}}}^{\prime }$, is strongly influenced by the fraction of NO in the source NO_x, and to a lesser extent the mixing time-scale coefficient. Adjusting the latter gives improved agreement with the Brown and Bilger experiment. Some weak parameter interactions are observed.     

Lagrangian dispersion model and its application to monitor nuclear power plants

The Lagrangian dispersion model and its advantages while applying it in monitoring nuclear power plants in complex terrain at varying meteorological conditions is explained. The software developed has been installed at the Bavarian State Authority to monitor its six nuclear power plants. Input data are routinely measured meteorological data as well as emission data for iodine, aerosols and noble gas.     

Determination of PM2.5 sources using time-resolved integrated source and receptor models

Multivariate statistical techniques are applied to particulate matter (PM) and meteorological data to identify the sources responsible for evening PM spikes at Sunland Park, NM (USA). The statistical techniques applied are principal components analysis (PCA), redundancy analysis (RDA), and absolute principal components scores analysis (APCSA), and the data evaluated are 3-h average (6–9 p.m.) PM_2.5 mass and chemical composition and 1-h average PM_2.5 and PM₁₀ mass and environmental data collected in the winter of 2002. Although the interpretation of the data was complicated by the presence of sources which are likely changing in time (e.g. brick kilns), the multivariate analyses indicate that the evening high PM_2.5 is associated with burning-activities occurring to the south of Sunland Park, and these emissions are characterized by elevated Sb, Cl⁻, and elemental carbon; 68% of the PM_2.5 mass can be attributed to this source. The PM₁₀ evening peaks, on the other hand, are mainly caused by resuspended dust generated by vehicular movements south of the site and transported by the local terrain-induced drainage flow.     

Determination of PM2.5 sources using time-resolved integrated source and receptor models

Multivariate statistical techniques are applied to particulate matter (PM) and meteorological data to identify the sources responsible for evening PM spikes at Sunland Park, NM (USA). The statistical techniques applied are principal components analysis (PCA), redundancy analysis (RDA), and absolute principal components scores analysis (APCSA), and the data evaluated are 3-h average (6–9 p.m.) PM_2.5 mass and chemical composition and 1-h average PM_2.5 and PM₁₀ mass and environmental data collected in the winter of 2002. Although the interpretation of the data was complicated by the presence of sources which are likely changing in time (e.g. brick kilns), the multivariate analyses indicate that the evening high PM_2.5 is associated with burning-activities occurring to the south of Sunland Park, and these emissions are characterized by elevated Sb, Cl⁻, and elemental carbon; ∼68% of the PM_2.5 mass can be attributed to this source. The PM₁₀ evening peaks, on the other hand, are mainly caused by resuspended dust generated by vehicular movements south of the site and transported by the local terrain-induced drainage flow.     

Simulating industrial emissions using atmospheric dispersion modeling system: model performance and source emission factors

In this paper, the Gaussian Atmospheric Dispersion Modeling System (ADMS4) was coupled with field observations of surface meteorology and concentrations of several air quality indicators (nitrogen oxides (NOx), carbon monoxide (CO), fine particulate matter (PM10) and sulfur dioxide (SO2)) to test the applicability of source emission factors set by the European Environment Agency (EEA) and the United States Environmental Protection Agency (USEPA) at an industrial complex. Best emission factors and data groupings based on receptor location, type of terrain and wind speed, were relied upon to examine model performance using statistical analyses of simulated and observed data. The model performance was deemed satisfactory for several scenarios when receptors were located at downwind sites with index of agreement 'd' values reaching 0.58, fractional bias 'FB' and geometric mean bias 'MG' values approaching 0 and 1, respectively, and normalized mean square error 'NMSE' values as low as 2.17. However, median ratios of predicted to observed concentrations 'Cp/Co' at variable downstream distances were 0.01, 0.36, 0.76 and 0.19 for NOx, CO, PM10 and SO2, respectively, and the fraction of predictions within a factor of two of observations 'FAC2' values were lower than 0.5, indicating that the model could not adequately replicate all observed variations in emittant concentrations. Also, the model was found to be significantly sensitive to the input emission factor bringing into light the deficiency in regulatory compliance modeling which often uses internationally reported emission factors without testing their applicability.     

Comparison of the industrial source complex and AERMOD dispersion models: case study for human health risk assessment

Air quality models are typically used to predict the fate and transport of air emissions from industrial sources to comply with federal and state regulatory requirements and environmental standards, as well as to determine pollution control requirements. For many years, the U.S. Environmental Protection Agency (EPA) widely used the Industrial Source Complex (ISC) model because of its broad applicability to multiple source types. Recently, EPA adopted a new rule that replaces ISC with AERMOD, a state-of-the-practice air dispersion model, in many air quality impact assessments. This study compared the two models as well as their enhanced versions that incorporate the Plume Rise Model Enhancements (PRIME) algorithm. PRIME takes into account the effects of building downwash on plume dispersion. The comparison used actual point, area, and volume sources located on two separate facilities in conjunction with site-specific terrain and meteorological data. The modeled maximum total period average ground-level air concentrations were used to calculate potential health effects for human receptors. The results show that the switch from ISC to AERMOD and the incorporation of the PRIME algorithm tend to generate lower concentration estimates at the point of maximum ground-level concentration. However, the magnitude of difference varies from insignificant to significant depending on the types of the sources and the site-specific conditions. The differences in human health effects, predicted using results from the two models, mirror the concentrations predicted by the models.     

Validation and application of obstacle-resolving urban dispersion models

The development of micro-scale meteorological models has progressed in recent years. Some of them are already commercially available. With little hesitation, consulting engineers apply them to complex real-world problems. How accurate are the results? Using the example of urban dispersion models, the paper tries to give a critical assessment of the present ‘state of application’.     

Numerical simulation of scalar dispersion downstream of a square obstacle using gradient-transport type models

We present a numerical study of scalar transport released from a line source downstream of a square obstacle to investigate the capabilities and limitations of gradient-transport modeling in predicting atmospheric dispersion. The standard k–? and k–ω models and a Reynolds Stress Transport closure are employed and compared to predict the time-averaged turbulent flow field, while a standard gradient–diffusion model is initially adopted to relate the scalar flux to mean gradients of the concentration field. The analysis of two algebraic closures for turbulent scalar fluxes based on the generalized-gradient–diffusion hypothesis and its quadratic extension is also presented. In spite of the rather simple flow setup, where both the flow and the scalar fields can be assumed homogeneous in the spanwise direction, the analysis clarifies several critical issues concerning gradient-transport type models. We established the dominant role of predicted turbulent kinetic energy on scalar dispersion when a scalar diffusivity is employed, irrespectively of the Reynolds stress closure adopted for the averaged momentum equation. Moreover, the standard gradient–diffusion hypothesis failed to predict the streamwise component of the scalar flux, which is characterized by a counter-gradient-transport mechanism. Although the resulting contribution in the averaged scalar transport equation is small in the present flow configuration, this limitation can become severe for strongly inhomogeneous flows in the presence of point sources, where the spread of the scalar plume is essentially three-dimensional. The predictive capabilities of gradient-transport type modeling are found clearly improved using algebraic closures, which appear to represent a promising tool for predicting atmospheric dispersion in complex flows when unsteady transport mechanisms are not dominant.     

Sensitivity of two dispersion models (AERMOD and ISCST3) to input parameters for a rural ground-level area source

As of December 2006, the American Meteorological Society/U.S. Environmental Protection Agency (EPA) Regulatory Model with Plume Rise Model Enhancements (AERMOD-PRIME; hereafter AERMOD) replaced the Industrial Source Complex Short Term Version 3 (ISCST3) as the EPA-preferred regulatory model. The change from ISCST3 to AERMOD will affect Prevention of Significant Deterioration (PSD) increment consumption as well as permit compliance in states where regulatory agencies limit property line concentrations using modeling analysis. Because of differences in model formulation and the treatment of terrain features, one cannot predict a priori whether ISCST3 or AERMOD will predict higher or lower pollutant concentrations downwind of a source. The objectives of this paper were to determine the sensitivity of AERMOD to various inputs and compare the highest downwind concentrations from a ground-level area source (GLAS) predicted by AERMOD to those predicted by ISCST3. Concentrations predicted using ISCST3 were sensitive to changes in wind speed, temperature, solar radiation (as it affects stability class), and mixing heights below 160 m. Surface roughness also affected downwind concentrations predicted by ISCST3. AERMOD was sensitive to changes in albedo, surface roughness, wind speed, temperature, and cloud cover. Bowen ratio did not affect the results from AERMOD. These results demonstrate AERMOD's sensitivity to small changes in wind speed and surface roughness. When AERMOD is used to determine property line concentrations, small changes in these variables may affect the distance within which concentration limits are exceeded by several hundred meters.