Aubière, France
Aubière, France

Time filter

Source Type

Armand P.,CEA DAM Ile-de-France | Brocheton F.,NUMTECH | Poulet D.,NUMTECH | Vendel F.,Sillages Environnement | And 2 more authors.
Atmospheric Environment | Year: 2014

This paper is an original contribution to uncertainty quantification in atmospheric transport & dispersion (AT&D) at the local scale (1-10km). It is proposed to account for the imprecise knowledge of the meteorological and release conditions in the case of an accidental hazardous atmospheric emission. The aim is to produce probabilistic risk maps instead of a deterministic toxic load map in order to help the stakeholders making their decisions. Due to the urge attached to such situations, the proposed methodology is able to produce such maps in a limited amount of time. It resorts to a Lagrangian particle dispersion model (LPDM) using wind fields interpolated from a pre-established database that collects the results from a computational fluid dynamics (CFD) model. This enables a decoupling of the CFD simulations from the dispersion analysis, thus a considerable saving of computational time. In order to make the Monte-Carlo-sampling-based estimation of the probability field even faster, it is also proposed to recourse to the use of a vector Gaussian process surrogate model together with high performance computing (HPC) resources. The Gaussian process (GP) surrogate modelling technique is coupled with a probabilistic principal component analysis (PCA) for reducing the number of GP predictors to fit, store and predict. The design of experiments (DOE) from which the surrogate model is built, is run over a cluster of PCs for making the total production time as short as possible. The use of GP predictors is validated by comparing the results produced by this technique with those obtained by crude Monte Carlo sampling. © 2014 Elsevier Ltd.


AlRFOBEP is the association in charge of the air quality monitoring in the Etang de Berre area. AIRFOBEP is managing a network of ten sensors to monitor the PM10 particulate pollution. This network is updated once a year according to the Air Quality Monitoring Plan (PSQA). Optimizing this network needs to know how the particulate pollution is distributed in the area. In other words, to determine the limits of homogeneous zones of PM10 pollution. The aim of the project presented in this article is to produce a map of homogeneous zones of PM10 pollution in the Etang de Berre area. The project was carried out in two steps: PM10 atmospheric dispersion modeling, using a ADMS-URBAN software, Statistic classification, based on the well known Hierarchical Ascending Classification (HAC) technique. Results of the atmospheric dispersion modeling was namely adjusted using an original technique for the "background PM10 pollution" computation. Good performances have been obtained when comparing modeling and measurements data. Finally, a set of five homogeneous zones was found to well describe the PM 10 pollution level distribution in the Etang de Berre area. Air quality modeling. PM10 pollution.


Brocheton F.,NUMTECH | Mesbah B.,AIRFOBEP | Jacquinot M.,AIRFOBEP | Buisson E.,NUMTECH
HARMO 2010 - Proceedings of the 13th International Conference on Harmonisation within Atmospheric Dispersion Modelling for Regulatory Purposes | Year: 2010

AIRFOBEP is the regional air quality agency in charge of the survey of the air pollution over the Etang de Berre region, which is one of the two main industrial areas in France. From several years, AIRFOBEP has decided to develop an operational automated platform which routinely monitors and forecasts air pollution over its territory. This paper discusses the operational tools associated with particle matter (PM10) and sulfur dioxide (SO2). The particularity of these tools is that the evaluation of the pollution associated with each pollutant is based on local air dispersion modelling (ADMS4 and ADMS-Urban for SO2 and PM10, respectively) to account for numerous local emission sources, considering a large simulation domain. A description of each tool which has been developed will be given. An overall view of the performance of the system in terms of ground-level concentration prediction will also be shown.


Mallet V.,French Institute for Research in Computer Science and Automation | Mallet V.,ParisTech National School of Bridges and Roads | Tilloy A.,French Institute for Research in Computer Science and Automation | Tilloy A.,ParisTech National School of Bridges and Roads | And 2 more authors.
Proceedings of the 15th International Conference on Harmonisation within Atmospheric Dispersion Modelling for Regulatory Purposes, HARMO 2013 | Year: 2013

ADMS Urban is a non-linear static model whose input data p varies one simulated hour after the other. The model computes a high-dimensional concentration vector y = M(p) which can contain 105 concentrations. A full-year simulation of NO2 concentrations can take dozens of days of computations, which greatly limits the range of methods that can be applied to the model, especially for uncertainty quantification. This work proposes a method to replace ADMS Urban with a so-called emulator, i.e., a close approximation of ADMS Urban whose computational cost is negligible. First, the output concentration field y is projected on a few modes of a proper orthogonal decomposition[Ψ1 ⋯ ΨN], so that y ≃ ΣNj=1 αj Ψj where αj is the projection coefficient on j-th mode and N smaller than 10. Then, the reduced model f(p) = ΨT M(p) is replaced by a statistical emulator fso that f(p) ≃ f(p) and the computational cost of f(p) is negligible.


Tilloy A.,French Institute for Research in Computer Science and Automation | Tilloy A.,ParisTech National School of Bridges and Roads | Mallet V.,French Institute for Research in Computer Science and Automation | Mallet V.,ParisTech National School of Bridges and Roads | And 3 more authors.
Journal of Geophysical Research: Atmospheres | Year: 2013

We aim at optimally combining air quality computations, from the Gaussian model ADMS Urban, and ground observations at urban scale. An ADMS simulation generated NO2 concentration fields across Clermont-Ferrand (France) down to street level, every 3 h for the full year 2008. A monitoring network composed of nine fixed stations provided hourly observations to be assimilated. Every 3 h, we compute the so-called BLUE (best linear unbiased estimator), which is a concentration field merging ADMS outputs and ground observations. Its error variance is supposed to be minimal under given assumptions regarding the errors on observations and model simulations. A key step lies in the modeling of error covariances between the computed NO2 concentrations across the city. We introduce a parameterized covariance which heavily relies on the road network. The covariance between two locations depends on the distance of each location to the road network and on the distance between the locations along the road network. Efficient parameters for the covariances are primarily chosen according to prior assumptions, χ2 diagnosis and leave-one-out cross-validations. According to the cross-validations, the improvements due to the assimilation seem moderately far from the observation network, but the root mean square error roughly decreases by 30-50% in the main city where the station density is high. The method is computationally tractable for the generation of improved concentration fields over a long period, or for day-to-day forecasts. Key PointsBLUE-based data assimilation is carried out at urban scale.Background error covariances are parameterized and depend on the road network.Cross validation shows 30% to 50% error decrease at urban stations. © 2013. American Geophysical Union. All Rights Reserved.


Sadek R.,École Centrale Lyon | Soulhac L.,École Centrale Lyon | Brocheton F.,NUMTECH | Buisson E.,NUMTECH
Proceedings of the 15th International Conference on Harmonisation within Atmospheric Dispersion Modelling for Regulatory Purposes, HARMO 2013 | Year: 2013

The RANS Reynolds Stress Model is one of the most comprehensive and powerful turbulence models available, allowing the simulation of the anisotropic nature of turbulence by solving transport equations for the Reynolds stresses. However, the generally used set of empirical constants, contained in the Reynolds stresses equations, are not appropriate for atmospheric simulations. We, therefore, present in this paper a new method for determining the empirical constants in order to achieve atmospheric levels of turbulence. A special focus is placed on the constants of the pressure-strain term, an important component contained in the Reynolds stress equation. We choose the Gibson and Launder (1979) linear modeling of the pressure-strain term, which is the one used in the commercial code Fluent. Results with the new set of constants are then validated against theoretical and empirical results of atmospheric conditions in flat terrain.


Sadek R.,École Centrale Lyon | Soulhac L.,École Centrale Lyon | Brocheton F.,NUMTECH | Buisson E.,NUMTECH
Proceedings of the 15th International Conference on Harmonisation within Atmospheric Dispersion Modelling for Regulatory Purposes, HARMO 2013 | Year: 2013

This paper investigates how to consider in numerical CFD modeling the influence of the sampling time of atmospheric turbulence characteristics. In wind-tunnel measurements, all of the turbulent time scales can be captured by a finite sampling duration because eddy sizes are limited by the spatial dimensions of the wind tunnel. In the atmosphere, the sampling duration is generally larger because the spectrum of velocity fluctuations is unlimited. This difference has led to the development of two sets of empirical constants for the RANS turbulence model: standard constants (generally applicable for small-scale simulations) and atmospheric constants. This paper aims a more general methodology for the simulation of turbulent levels. We therefore present in this paper a model capable of calculating the turbulent characteristics for any given sampling duration, using the data from a simulation carried out by taking into account all time scales of atmospheric turbulence. We also demonstrate that this kind of simulation can be performed by using a RANS model, provided that a proper set of constants is taken into account. Such atmospheric constants are proposed by Duynkerke (1988) for the RANS k-ε model, and by Sadek et al. (2013) for the RANS Reynolds Stress Model. Finally, the new developed model is validated against atmospheric measurements on the SIRTA site, in France.


Michelot N.,University of Nice Sophia Antipolis | Pesin C.,NUMTECH | Carrega P.,University of Nice Sophia Antipolis
Pollution Atmospherique | Year: 2013

In the souht-east of France, mediterranean climate, the study area is located in the central Siagne valley, characterized by a contrast and complex topography. It is under influence of breezes and thermic inversions during stable meteorology conditions. Modeling of PM10 using the ADMS-Urban software has been implemented in this area in order to test its ability to simulate, over complex terrain, topo-climate effects on levels of PM 10. This paper also presents the first use of the French Spatial National Inventory (INS). This one had integrated PM10 emissions. The results show there is a good model representation of topo-climate effects in time and space on the PM10 concentrations. However, a large difference of it appears compared to the measured data. Out due, the area is subject generally to an important unknown local sources and external contribution: 15.4 ug/m3 on average over the study period.


Sadek R.,École Centrale Lyon | Soulhac L.,École Centrale Lyon | Brocheton F.,NUMTECH | Buisson E.,NUMTECH
HARMO 2011 - Proceedings of the 14th International Conference on Harmonisation within Atmospheric Dispersion Modelling for Regulatory Purposes | Year: 2011

The simulation of high resolution complex terrain effects (grid resolution of about hundred meters) is of great importance to model atmospheric dispersion at local scale (simulation domain less than 10 km), for air quality assessments. These effects include acceleration at the lee side and over the hill top, deceleration downstream of the hill and an eventual separation depending on the steepness and roughness of the hill. This paper investigates the performances of two air flow models, a linearizeddiagnostic model (Flowstar) and a CFD code (Fluent), in capturing complex terrain effects. We compare these two codes with measurements from two wind tunnel experiments in the presence of hills and valleys of different shapes and roughness (Khurshudyan et al., 1981, and Almeida et al., 1992). Results are presented in terms of wind, turbulence and relative acceleration (speed-up). We also focus on the ability of these models to simulate recirculation regions downstream of the valleys and steepest hills.Overall, we find a better prediction with Fluent, especially for rough and steep hills and in recirculation regions. We also present a limit of applicability for each code. Furthermore, the air flow simulated by the two codes is used to drive the Safety Lagrangian Atmospheric Model (SLAM) developed at EcoleCentrale de Lyon. The impact of the precision of the simulated air flow is evaluated and an intercomparison with the Gaussian plume dispersion model ADMS, using the Flowstar wind field, is presented. Finally, we find considerable improvement when we use the coupling of a CFD wind field code with a Lagrangian dispersion model, especially in highly complex terrain simulations.


Sadek R.,INSA Lyon | Soulhac L.,INSA Lyon | Brocheton F.,NUMTECH | Buisson E.,NUMTECH
HARMO 2011 - Proceedings of the 14th International Conference on Harmonisation within Atmospheric Dispersion Modelling for Regulatory Purposes | Year: 2011

In modeling atmospheric dispersion over complex terrain (relief, roughness or heat flux change), Computational Fluid Dynamics (CFD) codes can be a powerful tool for simulating air flow with very high spatial resolution. However, the well-known drawback of the CFD approach is that it is time consuming. In this paper, a method which uses partially converged CFD solutions as a way of reducing CPU time, while keeping the precision of the solution at an acceptable level, is presented. We therefore demonstrate that it is possible to reach a wind field solution very close to the converged solution, in a small fraction of the CPU time needed to reach the fully converged solution. We present an optimum point of convergence,depending on the complexity of the terrain, for several cases of simulations with the commercial CFD code Fluent. Such complexities include steepness of hills and valleys, roughness of terrain and thermal stratification. We present an estimate of the error in comparison to the fully converged solution and evaluate the gain of CPU time following each case study.Finally, we strengthen our conclusions by a comparison with wind tunnel experiments in the presence of hills.

Loading NUMTECH collaborators
Loading NUMTECH collaborators