An ensemble Kalman filter for atmospheric data assimilation: Application to wind tunnel data

In the previous work (Zheng et al., 2007, Zheng et al., 2009), a data assimilation method, based on ensemble Kalman filter, has been applied to a Monte Carlo Dispersion Model (MCDM). The results were encouraging when the method was tested by the twin experiment and a short-range field experiment. In this technical note, the measured data collected in a wind tunnel experiment have been assimilated into the Monte Carlo dispersion model. The uncertain parameters in the dispersion model, including source term, release height, turbulence intensity and wind direction have been considered. The 3D parameters, i.e. the turbulence intensity and wind direction, have been perturbed by 3D random fields. In order to find the factors which may influence the assimilation results, eight tests with different specifications were carried out. Two strategies of constructing the 3D perturbation field of wind direction were proposed, and the result shows that the two level strategy performs better than the one level strategy. It is also found that proper standard deviation and the correlation radius of the perturbation field play an important role for the data assimilation results.     

Assessing sensitivity of source term estimation

Source term estimation algorithms compute unknown atmospheric transport and dispersion modeling variables from concentration observations made by sensors in the field. Insufficient spatial and temporal resolution in the meteorological data as well as inherent uncertainty in the wind field data make source term estimation and the prediction of subsequent transport and dispersion extremely difficult. This work addresses the question: how many sensors are necessary in order to successfully estimate the source term and meteorological variables required for atmospheric transport and dispersion modeling?The source term estimation system presented here uses a robust optimization technique – a genetic algorithm (GA) – to find the combination of source location, source height, source strength, surface wind direction, surface wind speed, and time of release that produces a concentration field that best matches the sensor observations. The approach is validated using the Gaussian puff as the dispersion model in identical twin numerical experiments. The limits of the system are tested by incorporating additive and multiplicative noise into the synthetic data. The minimum requirements for data quantity and quality are determined by an extensive grid sensitivity analysis. Finally, a metric is developed for quantifying the minimum number of sensors necessary to accurately estimate the source term and to obtain the relevant wind information.     

Online update of model state and parameters of a Monte Carlo atmospheric dispersion model by using ensemble Kalman filter

For an atmospheric dispersion model designed for the assessment of nuclear accident consequences, some uncertain model parameters, such as source term and weather conditions, may influence the reliability of model predictions. In this respect, good estimations of both model state and uncertain parameters are required. In this paper, an ensemble Kalman filter (EnKF) based method for simultaneous state and parameter estimation, using off-site radiation monitoring data, is presented. This method is based on a stochastic state space model, which resembles the parameter errors with stochastic quantities. Three imperfect parameters, including the source release rate, wind direction and turbulence intensity were perturbed simultaneously, and multiple parameter estimation were performed. Having been tested against both simulated and real radiation monitoring data, the method was found to be able to realistically reconstruct the real scene of dispersion, as well as the uncertain parameters. The estimated parameters given by EnKF nicely converge to the true values, and the method also tracks the temporal variation of those parameters.     

Assimilation of conventional and satellite wind observations in a mesoscale atmospheric model for studying atmospheric dispersion

A mesoscale atmospheric model PSU/NCAR MM5 is used to provide operational weather forecasts for a nuclear emergency response decision support system on the southeast coast of India. In this study the performance of the MM5 model with assimilation of conventional surface and upper-air observations along with satellite derived 2-d surface wind data from QuickSCAT sources is examined. Two numerical experiments with MM5 are conducted: one with static initialization using NCEP FNL data and second with dynamic initialization by assimilation of observations using four dimensional data assimilation (FDDA) analysis nudging for a pre-forecast period of 12 h. Dispersion simulations are conducted for a hypothetical source at Kalpakkam location with the HYSPLIT Lagrangian particle model using simulated wind field from the above experiments. The present paper brings out the differences in the atmospheric model predictions and the differences in dispersion model results from control and assimilation runs. An improvement is noted in the atmospheric fields from the assimilation experiment which has led to significant alteration in the trajectory positions, plume orientation and its distribution pattern. Sensitivity tests using different PBL and surface parameterizations indicated the simple first order closure schemes (Blackadar, MRF) coupled with the simple soil model have given better results for various atmospheric fields. The study illustrates the impact of the assimilation of the scatterometer wind and automated weather stations (AWS) observations on the meteorological model predictions and the dispersion results.     

A statistical approach for estimating uncertainty in dispersion modeling: An example of application in southwestern USA

A method based on a statistical approach of estimating uncertainty in simulating the transport and dispersion of atmospheric pollutants is developed using observations and modeling results from a tracer experiment in the complex terrain of the southwestern USA. The method takes into account the compensating nature of the error components by representing all terms, except dispersion error and variance of stochastic processes. Dispersion error and the variance of the stochastic error are estimated using the maximum likelihood estimation technique applied to the equation for the fractional error. Mesoscale Model 5 (MM5) and a Lagrangian random particle dispersion model with three optional turbulence parameterizations were used as a test bed for method application. Modeled concentrations compared well with the measurements (correlation coefficients on the order of 0.8). The effects of changing two structural components (the turbulence parameterization and the model grid vertical resolution) on the magnitude of the dispersion error also were examined. The expected normalized dispersion error appears to be quite large (up to a factor of three) among model runs with various turbulence schemes. Tests with increased vertical resolution of the atmospheric model (MM5) improved most of the dispersion model statistical performance measures, but to a lesser extent compared to selection of a turbulence parameterization. Method results confirm that structural components of the dispersion model, namely turbulence parameterizations, have the most influence on the expected dispersion error.     

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.     

Source estimation methods for atmospheric dispersion

Both forward and backward transport modeling methods are being developed for characterization of sources in atmospheric releases of toxic agents. Forward modeling methods, which describe the atmospheric transport from sources to receptors, use forward-running transport and dispersion models or computational fluid dynamics models which are run many times, and the resulting dispersion field is compared to observations from multiple sensors. Forward modeling methods include Bayesian updating and inference schemes using stochastic Monte Carlo or Markov Chain Monte Carlo sampling techniques. Backward or inverse modeling methods use only one model run in the reverse direction from the receptors to estimate the upwind sources. Inverse modeling methods include adjoint and tangent linear models, Kalman filters, and variational data assimilation, among others.This survey paper discusses these source estimation methods and lists the key references. The need for assessing uncertainties in the characterization of sources using atmospheric transport and dispersion models is emphasized.     

Multi-model vs. EPS-based ensemble atmospheric dispersion simulations: A quantitative assessment on the ETEX-1 tracer experiment case

Several techniques have been developed over the last decade for the ensemble treatment of atmospheric dispersion model predictions. Among them two have received most of the attention, the multi-model and the ensemble prediction system (EPS) modeling. The multi-model approach relies on model simulations produced by different atmospheric dispersion models using meteorological data from potentially different weather prediction systems. The EPS-based ensemble is generated by running a single atmospheric dispersion model with the ensemble weather prediction members. In the paper we compare both approaches with the help of statistical indicators, using the simulations performed for the ETEX-1 tracer experiment. Both ensembles are also evaluated against measurement data. Among the most relevant results is that the multi-model median and the mean of EPS-based ensemble produced the best results, hence we consider a combination of multi-model and EPS-based approaches as an interesting suggestion for further research.     

Data assimilation in meteorological pre-processors: Effects on atmospheric dispersion simulations

In previous work [Kovalets, I., Andronopoulos, S., Bartzis, J.G., Gounaris, N., Kushchan, A., 2004. Introduction of data assimilation procedures in the meteorological pre-processor of atmospheric dispersion models used in emergency response systems. Atmospheric Environment 38, 457–467.] the authors have developed data assimilation (DA) procedures and implemented them in the frames of a diagnostic meteorological pre-processor (MPP) to enable simultaneous use of meteorological measurements with numerical weather prediction (NWP) data. The DA techniques were directly validated showing a clear improvement of the MPP output quality in comparison with meteorological measurement data. In the current paper it is demonstrated that the application of DA procedures in the MPP, to combine meteorological measurements with NWP data, has a noticeable positive effect on the performance of an atmospheric dispersion model (ADM) driven by the MPP output. This result is particularly important for emergency response systems used for accidental releases of pollutants, because it provides the possibility to combine meteorological measurements with NWP data in order to achieve more reliable dispersion predictions. This is also an indirect way to validate the DA procedures applied in the MPP. The above goal is achieved by applying the Lagrangian ADM DIPCOT driven by meteorological data calculated by the MPP code both with and without the use of DA procedures to simulate the first European tracer experiment (ETEX I). The performance of the ADM in each case was evaluated by comparing the predicted and the experimental concentrations with the use of statistical indices and concentration plots. The comparison of resulting concentrations using the different sets of meteorological data showed that the activation of DA in the MPP code clearly improves the performance of dispersion calculations in terms of plume shape and dimensions, location of maximum concentrations, statistical indices and time variation of concentration at the detectors locations.     

An atmospheric dispersion model for the environmental impact assessment of thermal power plants in Japan--a method for evaluating topographical effects

An atmospheric dispersion model was developed for the environmental impact assessment of thermal power plants in Japan, and a method for evaluating topographical effects using this model was proposed. The atmospheric dispersion model consists of an airflow model with a turbulence closure model based on the algebraic Reynolds stress model and a Lagrangian particle dispersion model (LPDM). The evaluation of the maximum concentration of air pollutants such as SO2, NOx, and suspended particulate matter is usually considered of primary importance for environmental impact assessment. Three indices were therefore estimated by the atmospheric dispersion model: the ratios (alpha and beta, respectively) of the maximum concentration and the distance of the point of the maximum concentration from the source over topography to the respective values over a flat plane, and the relative concentration distribution [gamma(x)] along the ground surface projection of the plume axis normalized by the maximum concentration over a flat plane. The atmospheric dispersion model was applied to the topography around a power plant with a maximum elevation of more than 1,000 m. The values of alpha and beta evaluated by the atmospheric dispersion model varied between 1 and 3 and between 1 and 0.4, respectively, depending on the topographical features. These results and the calculated distributions of y(x) were highly similar to the results of the wind tunnel experiment. Therefore, when the slope of a hill or mountain is similar to the topography considered in this study, it is possible to evaluate topographical effects on exhaust gas dispersion with reasonable accuracy using the atmospheric dispersion model as well as wind tunnel experiments.     

The effect of improved nowcasting of precipitation on air quality modeling

The predictive potential of air quality models and thus their value in emergency management and public health support are critically dependent on the quality of their meteorological inputs. The atmospheric flow is the primary cause of the dispersion of airborne substances. The scavenging of pollutants by cloud particles and precipitation is an important sink of atmospheric pollution and subsequently determines the spatial distribution of the deposition of pollutants. The long-standing problem of the spin-up of clouds and precipitation in numerical weather prediction models limits the accuracy of the prediction of short-range dispersion and deposition from local sources. The resulting errors in the atmospheric concentration of pollutants also affect the initial conditions for the calculation of the long-range transport of these pollutants. Customary the spin-up problem is avoided by only using NWP (Numerical Weather Prediction) forecasts with a lead time greater than the spin-up time of the model. Due to the increase of uncertainty with forecast range this reduces the quality of the associated forecasts of the atmospheric flow.In this article recent improvements through diabatic initialization in the spin-up of large-scale precipitation in the Hirlam NWP model are discussed. In a synthetic example using a puff dispersion model the effect is demonstrated of these improvements on the deposition and dispersion of pollutants with a high scavenging coefficient, such as sulphur, and a low scavenging coefficient, such as cesium-137. The analysis presented in this article leads to the conclusion that, at least for situations where large-scale precipitation dominates, the improved model has a limited spin-up so that its full forecast range can be used. The implication for dispersion modeling is that the improved model is particularly useful for short-range forecasts and the calculation of local deposition. The sensitivity of the hydrological processes to proper initialization implies that the spin-up problem may reoccur with changes in the model and increased model resolution. Spin-up should be an ongoing concern for atmospheric modelers.     

Source function estimate by means of variational data assimilation applied to the ETEX-I tracer experiment

The ETEX data set opens new possibilities to develop data assimilation procedures in the area of long-range transport. This paper illustrates the possibilities using a variational approach, where the source term for ETEX-I was reconstructed. The MATCH model (Robertson et al., 1996) has been the basis for this attempt. The timing of the derived emission rates are in accordance with the time period for the ETEX-I release, and a cross validation, with observations beyond the selected assimilation period, shows that the source term gained holds for the entire ETEX-I experiment. A poor-man variational approach was shown to perform nearly as good as a fully variational data assimilation. The issue of quality control has not been considered in this attempt but will be an important part that has to be addressed in future work.     

An Atmospheric Dispersion Model for the Environmental Impact Assessment of Thermal Power Plants in Japan—A Method for Evaluating Topographical Effects

Abstract

An atmospheric dispersion model was developed for the environmental impact assessment of thermal power plants in Japan, and a method for evaluating topographical effects using this model was proposed. The atmospheric dispersion model consists of an airflow model with a turbulence closure model based on the algebraic Reynolds stress model and a Lagrangian particle dispersion model (LPDM). The evaluation of the maximum concentration of air pollutants such as SO₂, NO_x, and suspended particulate matter is usually considered of primary importance for environmental impact assessment. Three indices were therefore estimated by the atmospheric dispersion model: the ratios (α and β, respectively) of the maximum concentration and the distance of the point of the maximum concentration from the source over topography to the respective values over a flat plane, and the relative concentration distribution [γ(x)] along the ground surface projection of the plume axis normalized by the maximum concentration over a flat plane. The atmospheric dispersion model was applied to the topography around a power plant with a maximum elevation of more than 1000 m. The values of α and β evaluated by the atmospheric dispersion model varied between 1 and 3 and between 1 and 0.4, respectively, depending on the topographical features. These results and the calculated distributions of γ(x) were highly similar to the results of the wind tunnel experiment. Therefore, when the slope of a hill or mountain is similar to the topography considered in this study, it is possible to evaluate topographical effects on exhaust gas dispersion with reasonable accuracy using the atmospheric dispersion model as well as wind tunnel experiments.     

On the influence of the urban roughness sublayer on turbulence and dispersion

The concept of the urban roughness sublayer is discussed and this lowest atmospheric layer over a rough surface is shown to have a non-negligible vertical extension over typical urban surfaces. The existing knowledge on the turbulence and flow structure within an urban roughness sublayer is reviewed, focusing on the height dependence of turbulent fluxes and a scaling approach for turbulence statistics, such as velocity variances, in the above-roof part of the roughness sublayer. Finally, the implication of this turbulence and flow structure upon dispersion characteristics is investigated. The most prominent difference of explicitly taking into account the roughness sublayer in a dispersion simulation (as compared to assuming a `constant flux layer') is a clearly enhanced ground level concentration far downwind from the source. For the example of a tracer release experiment over a (sub) urban surface (Copenhagen) it is shown that introducing the roughness sublayer clearly improves the model performance.     

Targeting of observations for accidental atmospheric release monitoring

In the event of an accidental atmospheric release of radionuclides from a nuclear power plant, accurate real-time forecasting of the activity concentrations of radionuclides is acutely required by the decision makers for the preparation of adequate countermeasures. Yet, the accuracy of the forecasted plume is highly dependent on the source term estimation. Inverse modelling and data assimilation techniques should help in that respect. However the plume can locally be thin and could avoid a significant part of the radiological monitoring network surrounding the plant. Deploying mobile measuring stations following the accident could help to improve the source term estimation. In this paper, a method is proposed for the sequential reconstruction of the plume, by coupling a sequential data assimilation algorithm based on inverse modelling with an observation targeting strategy. The targeting design strategy consists in seeking the optimal locations of the mobile monitors at time t + 1 based on all available observations up to time t.The performance of the sequential assimilation with and without targeting of observations has been assessed in a realistic framework. It focuses on the Bugey nuclear power plant (France) and its surroundings within 50 km from the plant. The existing surveillance network is used and realistic observational errors are assumed. The targeting scheme leads to a better estimation of the source term as well as the activity concentrations in the domain. The mobile stations tend to be deployed along plume contours, where activity concentration gradients are important. It is shown that the information carried by the targeted observations is very significant, as compared to the information content of fixed observations. A simple test on the impact of model error from meteorology shows that the targeting strategy is still very useful in a more uncertain context.     

Prediction of atmospheric dispersion of pollutants in an airport environment

In this article we discuss the development of a methodology to predict atmospheric turbulent dispersion of pollutants generated from air traffic in an airport. It is based on the Lagrangian stochastic model (LSM), developed by Das and Durbin [2005. A Lagrangian stochastic model for dispersion in stratified turbulence, Physics of Fluids 17, 025109]. The approach is via the backward trajectory formulation of the model. The sources and receptors in an airport type problem are modeled as spheres and procedures have been derived for concentration calculation by both forward and backward trajectory methods. Some tests are performed to highlight certain features of the method. The turbulence statistics that are required as input are provided in terms of similarity profiles. The airport domain is partitioned to make the required search algorithms efficient. Pollutant concentration profiles are calculated over a range of meteorological data.     

Adherence of Sulfur Dioxide Concentrations in the Vicinity of a Steam Plant to Plume Dispersion Models

Air monitoring data for a calendar year at one of the TVA power plants has been used to evaluate the appropriateness of the Sutton, the Bosanquet and Pearson, and the USPHS-TVA atmospheric dispersion models to predict ground level concentrations of sulfur dioxide from emission and meterological data. Aerometric data included one half hourly average sulfur dioxide concentrations, recorded by four Thomas autometers, and the necessary meterological parameters for the solving of atmospheric dispersion models. Based on these meterological parameters and observed plume rise data, over 4000 one half hourly average maximum and minimum expected ground line sulfur dioxide concentrations were predicted for each of the above dispersion models by the use of computer techniques. The plant is a line source; however, an empirical correction was applied to emission data to reduce them to emissions for an equivalent point source. The predicted sulfur dioxide levels for each of the dispersion models were compared to the measured levels throughout the year. Three different sets of diffusion coefficients were applied to the Sutton model and successful predictions, according to a criterion utilizing an acceptable range of concentration, varied from 66 to 93%. The Bosanquet and Pearson model produced successful predictions 90% of the time, while the USPHS-TVA model was successful 94% of the time.Unsuccessful predictions were primarily overestimates.     

Developing intake fraction estimates with limited data: Comparison of methods in Mexico City

In order to estimate the health benefits of reducing mobile source emissions, analysts typically use detailed atmospheric models to estimate the change in population exposure that results from a given change in emissions. However, this may not be feasible in settings where data are limited or policy decisions are needed in the short term. Intake fraction (iF), defined as the fraction of emissions of a pollutant or its precursor that is inhaled by the population, is a metric that can be used to compare exposure assessment methods in a health benefits analysis context. To clarify the utility of rapid-assessment methods, we calculate particulate matter iFs for the Mexico City Metropolitan Area using five methods, some more resource intensive than others. First, we create two simple box models to describe dispersion of primary fine particulate matter (PM_2.5) in the Mexico City basin. Second, we extrapolate iFs for primary PM_2.5, ammonium sulfate, and ammonium nitrate from US values using a regression model. Third, we calculate iFs by assuming a linear relationship between emissions and population-weighted concentrations of primary PM_2.5, ammonium nitrate, and ammonium sulfate (a particle composition method). Finally, we estimate PM iFs from detailed atmospheric dispersion and chemistry models run for only a short period of time. Intake fractions vary by up to a factor of five, from 23 to 120 per million for primary PM_2.5. Estimates of 60, 7, and 0.7 per million for primary PM, secondary ammonium sulfate, and secondary ammonium nitrate, respectively, represent credible central estimates, with an approximate factor of two uncertainty surrounding each estimate. Our results emphasize that multiple rapid-assessment methods can provide meaningful estimates of iFs in resource-limited environments, and that formal uncertainty analysis, with special attention to model biases and uncertainty, would be important for health benefits analyses.     

A hybrid plume model for local-scale atmospheric dispersion

This paper presents a local-scale dispersion model, based on atmospheric boundary layer scaling theory. In the vicinity of the source, Gaussian equations are used in both the horizontal and vertical directions. After a specified transition distance, gradient transfer theory is applied in the vertical direction, while the horizontal dispersion is still assumed to be Gaussian. The dispersion parameters and eddy diffusivity are modelled in a form, which facilitates the use of a meteorological pre-processor. We present a novel model of the vertical eddy diffusivity (K_z), which is a continuous function of height in various atmospheric scaling regions. The model also includes a treatment of the dry deposition of gases and particulate matter. The accuracy of the numerical model was analysed by comparing the model predictions with two analytical solutions; the numerical deviations from these solutions were less than 2% for the computational regime. The model has been tested against the Kincaid experimental field data. The agreement of the predictions and the data is good on the average, although the internal variation of the predictions versus data scatter plot is substantial.     

Simulation of the evolution of particle size distributions containing coarse particulate in the atmospheric surface layer with a simple convection-diffusion-sedimentation model

The Fugitive Dust Model (FDM) and Industrial Source Complex (ISC), widely used coarse particulate dispersion models, have been shown inaccurate due to the neglect of vertical variations in atmospheric wind speed and turbulent diffusivity (Vesovic et al., 2001), omission of the gravitational advection velocity, and an underestimation of the ground deposition velocity (Kim and Larson, 2001). A simple, transient two-dimensional convection-diffusion-sedimentation model is proposed to simulate the evolution in particle size distribution of an aerosol ‘puff’ containing coarse particulate in the atmospheric surface layer. Monin-Okhubov similarity theory, accompanied by empirical observations made by Businger et al. (1971), is adopted to characterize the surface layer wind speed and turbulent diffusivity profiles over a wide range of atmospheric conditions. A first order analysis of the crossing trajectories effect suggests simulation data presented here are not significantly affected by particle inertia. The model is validated against Suffield experimental data in which coarse particulate deposition was measured out to a distance of 800 m from the source (Walker, 1965). Good agreement is found for the decay in ground deposits with distance from the source for stable atmospheres. Deposition data was also simulated for unstable atmospheric stratification and the current model was determined to modestly underestimate the peak concentration with increasing accuracy further downwind of the release. The current model's effective deposition velocity was compared to that suggested by Kim et al. (2000) and shows improvement with respect to FDM. Lastly, the model was used to simulate the dispersion of nine lognormal aerosol puffs in the lowest 50 m of the atmospheric surface layer for four classes of atmospheric stability. The simulated mass median aerodynamic diameters (MMAD) at multiple downwind sampling locations were calculated and plotted with distance from the source. The first 50 m from the source was found to have a substantial impact on the evolution of MMAD for stable atmospheric conditions. Away from the source, it was observed that particle size distributions were truncated by removal of all particles larger than about 60 μm. A particle Peclet number was also defined to quantify the relative importance of turbulent dispersion and sedimentation on particle motion in the vertical direction.