Publications

2024

• Étudier et promouvoir les langues de la Bretagne
  
  • Thomas Mannaïg
  • Ôbrée Bèrtran
  • Lefranc Yannick
  Cahiers du plurilinguisme européen, Presses universitaires de Strasbourg, 2024 (16). Cette contribution restitue les échanges d’une rencontre entre deux spécialistes des langues et cultures de Bretagne, sollicités pour partager les résultats de leurs analyses sur la place du breton et du gallo, reconnus aujourd’hui comme des langues de Bretagne, aux côtés du français. Attentifs aux problèmes théoriques et pratiques que posent la description et la diffusion de ces idiomes, les échanges ont porté sur les recherches dans les universités bretonnes, sur les collectes qui les ont précédées ou accompagnées, mais aussi sur les politiques publiques qui promeuvent ces langues dans la région. Il a également été question de la transmission ainsi que de l’écriture de ces langues et de leurs variétés. (10.57086/cpe.1724)
  DOI : 10.57086/cpe.1724
• Bilan d’activités sur la modélisation et la simulation numérique des écoulements multiphasiques
  
  • Hérard Jean-Marc
  , 2024. Cette note fournit une synthèse d'actions, menées sur la période 2001-2024 dans le lot de RD numérique du projet quadripartite -CEA-EDF-Framatome-IRSN- NEPTUNE, ainsi que dans d’autres projets d’EDF R&D, concernant la modélisation et la simulation numérique des écoulements transitoires diphasiques et multiphasiques. Une première partie décrit le développement de modèles hors-équilibre (EDP) et de lois de fermeture pour la représentation des écoulements multi-phasiques, à composants miscibles ou non miscibles, en milieu libre ou poreux, et pour des lois d’état thermodynamique quelconques par phase, en s’appuyant sur un cahier des charges strict. Les modèles totalement hors-équilibre sont associés au cadre : - diphasique eau-vapeur en milieu libre ; - diphasique eau-vapeur en milieu poreux ; - diphasique gaz-solide ; - diphasique hybride à trois champs (eau liquide, vapeur d’eau et gaz incondensable) ; - triphasique immiscible à trois champs (métal liquide, eau liquide et vapeur d’eau) ; - triphasique hybride à quatre champs (métal liquide, eau liquide, vapeur d’eau et gaz incondensable). Une seconde partie des actions a concerné la construction, le développement et la vérification de schémas d’approximation des solutions de ces modèles d’EDP en situation hors-équilibre, et d’éléments de validation des modèles.
• Multifractal analysis of wind turbine power and rainfall from an operational wind farm -Part 1: Wind turbine power and the associated biases
  
  • Jose Jerry
  • Gires Auguste
  • Roustan Yelva
  • Schnorenberger Ernani
  • Tchiguirinskaia Ioulia
  • Schertzer Daniel
  Nonlinear Processes in Geophysics, European Geosciences Union (EGU), 2024, 31, pp.587-602. The inherent variability in atmospheric fields, which extends over a wide range of temporal and spatial scales, is also transferred to energy fields extracted from them. In the specific case of wind power generation, this can be seen in the theoretical power available for extraction and the empirical power produced by turbines. To model and analyse them, it is important to quantify their variability, intermittency, and correlations with other interacting fields across scales. To understand the uncertainties involved in power production, power outputs from four 2 MW turbines are analysed (from an operational wind farm at Pay d'Othe, 110 km south-east of Paris, France) using the scale-invariant framework of universal multifractals (UM). Their scaling properties were compared with power available at the same location from simultaneously measured wind velocity.<p>While statistically analysing the turbine output, the rated power acts like an upper threshold that results in biased estimators. This is identified and quantified here using the theoretical framework of UM and validated using numerical simulations. Understanding the effect of instrumental thresholds in statistical analysis is important in retrieving actual fields and modelling them, more so in wind power production, where the uncertainties due to turbulence are already a leading challenge. This is expanded in Part 2, where the influence of rainfall on power production is studied across scales using UM and joint multifractals.</p> (10.5194/npg-31-587-2024)
  DOI : 10.5194/npg-31-587-2024
• Multifractal analysis of wind turbine power and rainfall from an operational wind farm -Part 2: Joint analysis of available wind power and rain intensity
  
  • Jose Jerry
  • Gires Auguste
  • Schnorenberger Ernani
  • Roustan Yelva
  • Schertzer Daniel
  • Tchiguirinskaia Ioulia
  Nonlinear Processes in Geophysics, European Geosciences Union (EGU), 2024, 31, pp.603-624. In the increasing global transition towards renewable and carbon-neutral energy, understanding the uncertainties associated with wind power production is extremely important. In addition to the widely acknowledged uncertainties from turbulence and wind intermittency, further complexity arises from the influence of rainfall, which only a limited number of studies have addressed so far. To understand this, multiple 3D sonic anemometers, mini meteorological stations, and optical disdrometers were employed on a meteorological mast on the Pays d'Othe wind farm (110 km south-east of Paris, France) in the framework of the Rainfall Wind Turbine or Turbulence (RW-Turb) project (https://hmco.enpc.fr/portfolio-archive/rw-turb/, last access: 26 November 2024). With these simultaneously measured data, wind power and its associated atmospheric fields were studied under various rainy conditions.<p>Variations of the wind velocity, power available on the wind farm, power produced by wind turbines, and air density are examined here, under rainy and dry conditions, using the scale-invariant framework of universal multifractals (UM). Since rated power acts like an upper threshold in statistical analysis of turbine power (discussed in Part 1), theoretically available power was used as a proxy. From an event-based analysis, differences in UM parameters were observed between rainy and dry conditions for the fields. This is explored further using joint multifractal analysis, which revealed an increase in the correlation exponent between various fields with the rain rate. Here we also examine the possibility of variation in power production by rainy conditions (convective or stratiform) as well as by regimes of wind velocity. While examining time steps according to wind velocity, turbine power curves showed different regions of departure from the state curve according to the rain rate.</p> (10.5194/npg-31-603-2024)
  DOI : 10.5194/npg-31-603-2024
• Deep learning, data assimilation and sea-ice dynamics
  
  • Durand Charlotte
  , 2024. The polar regions, Earth's natural thermostat, are undergoing rapid transformations due to climate change, with sea ice being a key indicator. Sea ice influences global temperatures, ocean circulation, and supports ecosystems and human communities. Predicting the sea-ice evolution is crucial but challenging due to its complex interactions with the atmosphere and ocean. The evolution of sea-ice depends on thermodynamic processes, mechanics, and fluid dynamics, which are challenging to model.On the one hand, deep learning has emerged as a powerful tool for modeling complex relationships in large datasets, showing promise in capturing patterns at a fraction of the computational cost of physics-based modeling. In sea-ice modeling, deep learning can enhance predictions, complement geophysical models and even potentially replace the models. Although still developing, these approaches offer potential for sea-ice forecasts at a moderate computational cost. On the other hand, data assimilation, which combines observational data with prediction models, is widely used in meteorology and oceanography to improve predictions.Merging deep learning with data assimilation offers a promising approach to sea-ice modeling. By combining data-driven and observation-based methods, this thesis aims to propose new methods for sea-ice prediction, which go beyond the state-of-the-art and enable a path forward to improve sea-ice prediction systems.In this thesis, we use deep learning to emulate neXtSIM, the sea-ice model developed by the SASIP project, and evaluate its use in variational data assimilation. Specifically, we develop a model capable of predicting the sea-ice thickness across the entire Arctic, showing improvements of up to 50% in forecast error over a persistence forecast, with stability maintained across several months. Similar results hold when the sea-ice concentration is emulated with improvement up to 20% over persistence in forecast skills.Next, by assimilating simulated data in a four-dimensional variational data assimilation scheme (4D-Var), we demonstrate the capabilities of a novel 4D-Var system built on the developed emulator. These methods, rarely applied in sea-ice modeling, require the model’s adjoint, which can be automatically computed with deep learning models. When ingesting real observations, our data assimilation system performs on par with the operational neXtSIM-F sea-ice forecasting system. These results pave the way for innovative 4D-Var systems for sea-ice models. (10.70675/1f32cd74z97b2z4760zb7d7zbfc518295376)
  DOI : 10.70675/1f32cd74z97b2z4760zb7d7zbfc518295376
• Generative Diffusion for Regional Surrogate Models From Sea‐Ice Simulations
  
  • Finn Tobias Sebastian
  • Durand Charlotte
  • Farchi Alban
  • Bocquet Marc
  • Rampal Pierre
  • Carrassi Alberto
  Journal of Advances in Modeling Earth Systems, American Geophysical Union, 2024, 16 (10), pp.e2024MS004395. We introduce deep generative diffusion for multivariate and regional surrogate modeling learned from sea-ice simulations. Given initial conditions and atmospheric forcings, the model is trained to generate forecasts for a 12-hr lead time from simulations by the state-of-the-art sea-ice model neXtSIM. For our regional model setup, the diffusion model outperforms as ensemble forecast all other tested models, including a free-drift model and a stochastic extension of a deterministic data-driven surrogate model. The diffusion model additionally retains information at all scales, resolving smoothing issues of deterministic models. Furthermore, by generating physically consistent forecasts, previously unseen for such kind of completely data-driven surrogates, the model can almost match the scaling properties of neXtSIM, as similarly deduced from sea-ice observations. With these results, we provide a strong indication that diffusion models can achieve similar results as traditional geophysical models with the significant advantage of being orders of magnitude faster and solely learned from data. (10.1029/2024ms004395)
  DOI : 10.1029/2024ms004395
• Assessing emissions from ship plumes in three French harbours: a glimpse at the PIRATE (Port Inventories ReAl TimE) project
  
  • Riffault Véronique
  • D’anna Barbara
  • Roustan Yelva
  • Landry Clara
  • Cesano Mara
  • Armengaud Alexandre
  • Yoo Jiye
  • Lee Yongchan
  • Lee Heekwan
  , 2024. 1.Context The impact of maritime transport on air quality and climate change is a critical area of research. Ships emit significant amounts of greenhouse gases (GHGs), volatile organic compounds (VOCs), NOx, and particulate matter (PM), contributing to both global and local pollution. While the overall contribution of maritime transport to pollution is relatively low compared to the volume of goods and passengers moved, its impact is particularly pronounced in port areas and coastal cities. 2.Methodology Three field campaigns of at least one month each were conducted in three French harbours: Toulon (Mediterranean Sea, Sept. 2021), Dunkirk (North Sea, Sept. 2022 and Sept.-Oct. 2023), and Le Havre (English Channel, Apr.-May 2023). Regional air quality monitoring networks measured regulated pollutant concentrations during each campaign. Additional state-of-the-art instruments were deployed in Dunkirk’s 1st campaign (focusing on particulate chemical/physical properties and VOCs), and Le Havre (primarily VOCs). The Port Air Quality management system (PAQman©) tracked ship movements in harbor areas throughout the campaigns. 3.Results and discussion PAQman© tracked vessels in real-time, generating emission maps using dynamic and static data. Field measurements identified periods of ship emission influence to validate modelling, aiding in understanding ship contribution to urban air pollution in port cities, which is crucial for developing effective pollution mitigation strategies.
• ACTRIS et les environnements urbains : Présentation du projet PEPR VDBI RESILIENCE
  
  • D’anna Barbara
  • Riffault Véronique
  • Vicente Jerome
  • Fasquelle Thomas
  • Albert Cécile H.
  • Xueref-Remy Irène
  • Hernandez Frédérique
  • Sartelet Karine
  • Occelli Florent
  • Frère Séverine
  , 2024.
• Population exposure to outdoor NO2 , black carbon, particle mass, and number concentrations over Paris with multi-scale modelling down to the street scale
  
  • Park Soo-Jin
  • Lugon Lya
  • Jacquot Oscar
  • Kim Youngseob
  • Baudic Alexia
  • D’anna Barbara
  • Di Antonio Ludovico
  • Di Biagio Claudia
  • Dugay Fabrice
  • Favez Olivier
  • Ghersi Véronique
  • Gratien Aline
  • Kammer Julien
  • Petit Jean-Eudes
  • Sanchez Olivier
  • Valari Myrto
  • Vigneron Jérémy
  • Sartelet Karine
  Atmospheric Chemistry and Physics Discussions, European Geosciences Union, 2024. This study focuses on mapping the concentrations of pollutants of health interest (NO2, black carbon (BC), PM2.5, number of particles (PN)) down to the street scale to represent as accurately as possible the population exposure. Simulations are performed over the Greater Paris area with the WRF-CHIMERE/MUNICH/SSH-aerosol chain, using either the top-down inventory EMEP or the bottom-up inventory Airparif with correction of the traffic flow. The concentrations of the pollutants are higher in streets than in the regional-scale urban background, due to the strong influence of road-traffic emissions locally. Model-to-data comparisons were performed at urban background and traffic stations, and evaluated using two performance criteria from the literature. For BC, harmonized equivalent BC (eBC) concentrations were estimated from concomitant mea-surements of eBC and elemental carbon. Using the bottom-up inventory with corrected road-traffic flow, the strictest criteria are met for NO2, eBC, PM2.5, and PN. Using the EMEP top-down inventory, the strictest criteria are also met for NO2, eBC and PM2.5, but errors tend to be larger than with the bottom-up inventory for NO2, eBC and PN. Using the top-down inventory, the concentrations tend to be lower along the streets than those simulated using the bottom-up inventory, especially for NO2 con-centrations, resulting in less urban heterogeneities. The impact of the size-distribution of non-exhaust emissions was analyzed at both regional and local scales, and it is higher in heavy-traffic streets. To assess exposure, a french database detailing the number of inhabitants in each building was used. The population-weighted concentration (PWC) was calculated by weighting populations by the outdoor concentrations to which they are exposed at the precise location of their home. An exposure scaling factor (ESF) was determined for each pollutant to estimate the ratio needed to correct urban background concentrations in order to assess exposure. The average ESF in Paris and Paris Ring Road is higher than 1 for NO2, eBC, PM2.5, PN, because the concentrations simulated at the local scale in streets are higher than those modelled at the regional scale. It indicates that the Parisian population exposure is under-estimated using regional-scale concentrations. Although this underestimation is low for PM2.5, with an ESF of 1.04, it is very high for NO2 (1.26), eBC (between 1.22 and 1.24), and PN (1.12). This shows that urban heterogeneities are important to be considered in order to represent the population exposure to NO2, eBC, and PN, but less so for PM2.5. (10.5194/egusphere-2024-2120)
  DOI : 10.5194/egusphere-2024-2120
• Multi-domain encoder–decoder neural networks for latent data assimilation in dynamical systems
  
  • Cheng Sibo
  • Zhuang Yilin
  • Kahouadji Lyes
  • Liu Che
  • Chen Jianhua
  • Matar Omar K
  • Arcucci Rossella
  Computer Methods in Applied Mechanics and Engineering, Elsevier, 2024, 430, pp.117201. High-dimensional dynamical systems often require computationally intensive physics-based simulations, making full physical space data assimilation impractical. Latent data assimilation methods perform assimilation in reduced-order latent space for efficiency but struggle with complex, nonlinear state-observation mappings. Recent solutions like Generalized Latent Data Assimilation (GLA) and Latent Space Data Assimilation (LSDA) address heterogeneous latent spaces by incorporating surrogate mapping functions but introduce computational costs and uncertainties. Furthermore, current algorithms that integrate data assimilation and deep learning still face limitations when it comes to handling non-explicit mapping functions. To address these challenges, this paper introduces a novel deep-learning-based data assimilation scheme, named Multi-domain Encoder–Decoder Latent Data Assimilation (MEDLA), capable of handling diverse data sources by sharing a common latent space. The proposed approach significantly reduces the computational burden since the complex mapping functions are mimicked by the multi-domain encoder–decoder neural network. It also enhances assimilation accuracy by minimizing interpolation and approximation errors. Extensive numerical experiments from three different test cases assess MEDLA’s performance in high dimensional dynamical systems, benchmarking it against state-of-the-art latent data assimilation methods. The numerical results consistently underscore MEDLA’s superiority in managing multi-scale observational data and tackling intricate, non-explicit mapping functions. (10.1016/j.cma.2024.117201)
  DOI : 10.1016/j.cma.2024.117201
• Accurate deep learning-based filtering for chaotic dynamics by identifying instabilities without an ensemble
  
  • Bocquet Marc
  • Farchi Alban
  • Finn Tobias
  • Durand Charlotte
  • Cheng Sibo
  • Chen Yumeng
  • Pasmans Ivo
  • Carrassi Alberto
  Chaos: An Interdisciplinary Journal of Nonlinear Science, American Institute of Physics, 2024, 34 (9). We investigate the ability to discover data assimilation (DA) schemes meant for chaotic dynamics with deep learning. The focus is on learning the analysis step of sequential DA, from state trajectories and their observations, using a simple residual convolutional neural network, while assuming the dynamics to be known. Experiments are performed with the Lorenz 96 dynamics, which display spatiotemporal chaos and for which solid benchmarks for DA performance exist. The accuracy of the states obtained from the learned analysis approaches that of the best possibly tuned ensemble Kalman filter and is far better than that of variational DA alternatives. Critically, this can be achieved while propagating even just a single state in the forecast step. We investigate the reason for achieving ensemble filtering accuracy without an ensemble. We diagnose that the analysis scheme actually identifies key dynamical perturbations, mildly aligned with the unstable subspace, from the forecast state alone, without any ensemble-based covariances representation. This reveals that the analysis scheme has learned some multiplicative ergodic theorem associated to the DA process seen as a non-autonomous random dynamical system. (10.1063/5.0230837)
  DOI : 10.1063/5.0230837
• Representation learning with unconditional denoising diffusion models for dynamical systems
  
  • Finn Tobias Sebastian
  • Disson Lucas
  • Farchi Alban
  • Bocquet Marc
  • Durand Charlotte
  Nonlinear Processes in Geophysics, European Geosciences Union (EGU), 2024, 31 (3), pp.409-431. We propose denoising diffusion models for data-driven representation learning of dynamical systems. In this type of generative deep learning, a neural network is trained to denoise and reverse a diffusion process, where Gaussian noise is added to states from the attractor of a dynamical system. Iteratively applied, the neural network can then map samples from isotropic Gaussian noise to the state distribution. We showcase the potential of such neural networks in proof-of-concept experiments with the Lorenz 1963 system. Trained for state generation, the neural network can produce samples that are almost indistinguishable from those on the attractor. The model has thereby learned an internal representation of the system, applicable for different tasks other than state generation. As a first task, we fine-tune the pre-trained neural network for surrogate modelling by retraining its last layer and keeping the remaining network as a fixed feature extractor. In these low-dimensional settings, such fine-tuned models perform similarly to deep neural networks trained from scratch. As a second task, we apply the pre-trained model to generate an ensemble out of a deterministic run. Diffusing the run, and then iteratively applying the neural network, conditions the state generation, which allows us to sample from the attractor in the run's neighbouring region. To control the resulting ensemble spread and Gaussianity, we tune the diffusion time and, thus, the sampled portion of the attractor. While easier to tune, this proposed ensemble sampler can outperform tuned static covariances in ensemble optimal interpolation. Therefore, these two applications show that denoising diffusion models are a promising way towards representation learning for dynamical systems. (10.5194/npg-31-409-2024)
  DOI : 10.5194/npg-31-409-2024
• Editorial: Workshop “New Trends in Complex Flows”
  
  • Faccanoni Gloria
  • Grec Bérénice
  • Hérard Jean-Marc
  • Hurisse Olivier
  • Jung Jonathan
  • Kokh Samuel
  • Mathis Hélène
  • Ndjinga Michael
  • Seguin Nicolas
  ESAIM: Proceedings and Surveys, EDP Sciences, 2024, 76, pp.1-1. (10.1051/proc/202476001)
  DOI : 10.1051/proc/202476001
• Impact of gas dry deposition parameterization on secondary particle formation in an urban canyon
  
  • Lin Chao
  • Ooka Ryozo
  • Kikumoto Hideki
  • Kim Youngseob
  • Zhang Yang
  • Flageul Cédric
  • Sartelet Karine
  Atmospheric Environment, Elsevier, 2024, 333, pp.120633. This study investigates the impact of the parameterization of dry deposition on the local gas-particle partitioning between ammonium nitrate NH₄NO₃ and its precursor gases (ammonia, NH₃ and nitric acid, HNO₃) through chemistry-coupled large-eddy simulations on the dispersion of reactive gaseous and particulate pollutant in an idealized street canyon. A key factor in the parameterization of dry deposition is the effective Henry's Law constant <i>H*</i>. Three distinct characterizations of <i>H*</i> are compared: two drawn from existing literature (referred to as Model 1 and Model 2) and one typical of low pH conditions (referred to as Model 3), which could be more representative of urban areas. Model 3 shows contrasting gas-particle partitioning results between NH₃, HNO₃, and NH₄NO3 with Model 1 and Model 2. In detail, the NH₄NO₃ concentrations in the street of Model 1 and Model 2 are smaller than the background NH₄NO₃ concentration. However, Model 3 shows higher NH4NO₃ concentration in the street than the background NH₄NO₃ concentration. This is because the dry deposition fluxes of NH₃ and HNO₃ are higher in Model 1 and Model 2 than in Model 3, resulting in less available NH₃ and HNO₃ for NH₄NO₃ formation. These findings highlight the importance of selecting an appropriate <i>H*</i> characterization that is tailored to the specific environmental conditions under investigation. (10.1016/j.atmosenv.2024.120633)
  DOI : 10.1016/j.atmosenv.2024.120633
• A CFD model for heat and mass transfer leading to plume formation within Wet Cooling Towers
  
  • Favre Luc
  • Ferrand Martin
  , 2024. The crucial role played by Wet Cooling Towers (WCT) in many electricity production plants (e.g. nuclear power plants) make them a key parameter in the industrial design of such facilities. Their impact over the cooling water consumption and surrounding atmosphere through the formation and dispersion of a humid air plume has pushed the need to obtain proper models and simulations in order to anticipate those effects. In this work, we tackle this issue through a dedicated modelling in the CFD solver code_saturne. Specific modeling includes heat and mass transfer (convection and evaporation) between the injected water and the air flow that are validated against experimental results obtained in a reduced scale WCT experimental loop. Satisfying agreement is obtained for several parameters such as air and water exit temperatures, evaporation mass flow rate and total exchanged thermal power. This constitutes an important first step for detailed CFD predictions of WCT water consumption and humid air plume atmospheric dispersion. (10.11159/htff24.209)
  DOI : 10.11159/htff24.209
• Development of a detailed gaseous oxidation scheme of naphthalene for secondary organic aerosol (SOA) formation and speciation
  
  • Lannuque Victor
  • Sartelet Karine
  Atmospheric Chemistry and Physics, European Geosciences Union, 2024, 24 (15), pp.8589-8606. Naphthalene is the most abundant polycyclic aromatic hydrocarbon (PAH) in vehicle emissions and polluted urban areas. Its atmospheric oxidation products are oxygenated compounds that are potentially harmful for health and/or contribute to secondary organic aerosol (SOA) formation. Despite its impact on air quality, its complex structure and a lack of data mean that no detailed scheme of naphthalene gaseous oxidation for SOA formation and speciation has been established yet. This study presents the construction of the first near-explicit chemical scheme for naphthalene oxidation by OH, including kinetic and mechanistic data. The scheme redundantly represents all the classical steps of atmospheric organic chemistry (i.e., oxidation of stable species, peroxy radical formation and reaction, and alkoxy radical evolution), thus integrating fragmentation or functionalization pathways and the influence of NOx on secondary compound formation. Missing kinetic and mechanistic data were estimated using structure–activity relationships (SARs) or by analogy with existing experimental or theoretical data. The proposed chemical scheme involves 383 species (231 stable species, including 93 % of the major molar masses observed in previous experimental studies) and 484 reactions with products. A first simulation reproducing experimental oxidation in an oxidation flow reactor under high-NOx conditions shows a simulated SOA mass on the same order of magnitude as has been observed experimentally, with an error of −9 %. (10.5194/acp-24-8589-2024)
  DOI : 10.5194/acp-24-8589-2024
• The CHIMERE chemistry-transport model v2023r1
  
  • Menut Laurent
  • Cholakian Arineh
  • Pennel Romain
  • Siour Guillaume
  • Mailler Sylvain
  • Valari Myrto
  • Lugon Lya
  • Meurdesoif Yann
  Geoscientific Model Development, European Geosciences Union, 2024, 17 (14), pp.5431-5457. A new version of the CHIMERE model is presented. This version contains both computational and physico-chemical changes. The computational changes make it easy to choose the variables to be extracted as a result, including values of maximum sub-hourly concentrations. Performance tests show that the model is 1.5 to 2 times faster than the previous version for the same setup. Processes such as turbulence, transport schemes and dry deposition have been modified and updated. Optimization was also performed for the management of emissions such as anthropogenic and mineral dust. The impact of fires on wind speed, soil properties and leaf area index (LAI) was added. Pollen emissions, transport and deposition were added for birch, ragweed, olive and grass. The model is validated with a simulation covering Europe with a 60 km × 60 km resolution and the entire year of 2019. Results are compared to various measurements, and statistical scores show that the model provides better results than the previous versions. (10.5194/gmd-17-5431-2024)
  DOI : 10.5194/gmd-17-5431-2024
• Bridging classical data assimilation and optimal transport: the 3D-Var case
  
  • Bocquet Marc
  • Vanderbecken Pierre J
  • Farchi Alban
  • Dumont Le Brazidec Joffrey
  • Roustan Yelva
  Nonlinear Processes in Geophysics, European Geosciences Union (EGU), 2024, 31, pp.335 - 357. <div><p>Because optimal transport (OT) acts as displacement interpolation in physical space rather than as interpolation in value space, it can avoid double-penalty errors generated by mislocations of geophysical fields. As such, it provides a very attractive metric for non-negative, sharp field comparison -the Wasserstein distance -which could further be used in data assimilation (DA) for the geosciences. However, the algorithmic and numerical implementations of such a distance are not straightforward. Moreover, its theoretical formulation within typical DA problems faces conceptual challenges, resulting in scarce contributions on the topic in the literature.</p><p>We formulate the problem in a way that offers a unified view with respect to both classical DA and OT. The resulting OTDA framework accounts for both the classical source of prior errors, background and observation, and a Wasserstein barycentre in between states which are pre-images of the background state and observation vector. We show that the hybrid OTDA analysis can be decomposed as a simpler OTDA problem involving a single Wasserstein distance, followed by a Wasserstein barycentre problem that ignores the prior errors and can be seen as a McCann interpolant. We also propose a less enlightening but straightforward solution to the full OTDA problem, which includes the derivation of its analysis error covariance matrix. Thanks to these theoretical developments, we are able to extend the classical 3D-Var/BLUE (best linear unbiased estimator) paradigm at the core of most classical DA schemes. The resulting formalism is very flexible and can account for sparse, noisy observations and non-Gaussian error statistics. It is illustrated by simple one-and two-dimensional examples that show the richness of the new types of analysis offered by this unification.</p></div> (10.5194/npg-31-335-2024)
  DOI : 10.5194/npg-31-335-2024
• Machine Learning and Physics-Driven Modelling and Simulation of Multiphase Systems
  
  • Basha Nausheen
  • Arcucci Rossella
  • Angeli Panagiota
  • Anastasiou Charitos
  • Abadie Thomas
  • Casas César Quilodrán
  • Chen Jianhua
  • Cheng Sibo
  • Chagot Loïc
  • Galvanin Federico
  • Heaney Claire
  • Hossein Fria
  • Hu Jinwei
  • Kovalchuk Nina
  • Kalli Maria
  • Kahouadji Lyes
  • Kerhouant Morgan
  • Lavino Alessio
  • Liang Fuyue
  • Nathanael Konstantia
  • Magri Luca
  • Lettieri Paola
  • Materazzi Massimiliano
  • Erigo Matteo
  • Pico Paula
  • Pain Christopher
  • Shams Mosayeb
  • Simmons Mark
  • Traverso Tullio
  • Valdes Juan Pablo
  • Wolffs Zef
  • Zhu Kewei
  • Zhuang Yilin
  • Matar Omar K
  International Journal of Multiphase Flow, Elsevier, 2024, 179, pp.104936. (10.1016/j.ijmultiphaseflow.2024.104936)
  DOI : 10.1016/j.ijmultiphaseflow.2024.104936
• CFD modeling of heat and mass transfer in cooling towers for humid air plume formation prediction
  
  • Favre Luc
  • Ferrand Martin
  , 2024. The crucial role played by Wet Cooling Towers (WCT) in many electricity production plants (e.g. nuclear power plants) make them a key parameter in the industrial design of such facilities. Their impact over the cooling water consumption and surrounding atmosphere through the formation and dispersion of a humid air plume has pushed the need to obtain proper models and simulations in order to anticipate those effects. In this work, we tackle this issue through a dedicated modelling in the CFD solver code_saturne. Specific modeling includes heat and mass transfer (convection and evaporation) between the injected water and the air flow that are validated against experimental results obtained in a reduced scale WCT experimental loop. Satisfying agreement is obtained for several parameters such as air and water exit temperatures, evaporation mass flow rate and total exchanged thermal power. This constitutes an important first step for detailed CFD predictions of WCT water consumption and humid air plume atmospheric dispersion.
• CFD study of PM10 dispersion in a sports stadium using a mesh based on geometry obtained from a 3-D cloud of laser points
  
  • Amino Hector
  • Flageul Cédric
  • Carissimo Bertrand
  • Ferrand Martin
  , 2024. This work presents a dispersion study of a multi-sport stadium using local scale simulation (CFD) using the open-source software code_saturne. A recently developed time scheme for indoor airflow is used. A high-fidelity numerical mesh is built from a cloud of points and used. Besides providing the local dynamics in the stadium, simulations results are compared to experimental PM10 concentration data from a handball game, where firework were lighted, and a 0-D model. CFD results were shown to correctly reproduce the PM10 variation.
• Contrasting effects of urban trees on air quality: From the aerodynamic effects in streets to impacts of biogenic emissions in cities
  
  • Maison Alice
  • Lugon Lya
  • Park Soo-Jin
  • Boissard Christophe
  • Faucheux Aurélien
  • Gros Valérie
  • Kalalian Carmen
  • Kim Youngseob
  • Leymarie Juliette
  • Petit Jean-Eudes
  • Roustan Yelva
  • Sanchez Olivier
  • Squarcioni Alexis
  • Valari Myrto
  • Viatte Camille
  • Vigneron Jérémy
  • Tuzet Andrée
  • Sartelet Karine
  Science of the Total Environment, Elsevier, 2024, 946, pp.174116. Urban trees are often not considered in air-quality models although they can significantly impact the concentrations of pollutants. Gas and particles can deposit on leaf surfaces, lowering their concentrations, but the tree crown aerodynamic effect is antagonist, limiting the dispersion of pollutants in streets. Furthermore, trees emit Biogenic Volatile Organic Compounds (BVOCs) that react with other compounds to form ozone and secondary organic aerosols. This study aims to quantify the impacts of these three tree effects (dry deposition, aerodynamic effect and BVOC emissions) on air quality from the regional to the street scale over Paris city. Each tree effect is added in the model chain CHIMERE/MUNICH/SSH-aerosol. The tree location and characteristics are determined using the Paris tree inventory, combined with allometric equations. The air-quality simulations are performed over June and July 2022. The results show that the aerodynamic tree effect increases the concentrations of gas and particles emitted in streets, such as NOx (+4.6 % on average in streets with trees and up to +37 % for NO2). This effect increases with the tree Leaf Area Index and it is more important in streets with high traffic, suggesting to limit the planting of trees with large crowns on high-traffic streets. The effect of dry deposition of gas and particles on leaves is very limited, reducing the concentrations of O3 concentrations by −0.6 % on average and at most −2.5 %. Tree biogenic emissions largely increase the isoprene and monoterpene concentrations, bringing the simulated concentrations closer to observations. Over the two-week sensitivity analysis, biogenic emissions induce an increase of O3, organic particles and PM2.5 street concentrations by respectively +1.1, +2.4 and + 0.5 % on average over all streets. This concentration increase may reach locally +3.5, +12.3 and + 2.9 % respectively for O3, organic particles and PM2.5, suggesting to prefer the plantation of low-emitting VOC species in cities. (10.1016/j.scitotenv.2024.174116)
  DOI : 10.1016/j.scitotenv.2024.174116
• Evolution des particules fines en champ proche du trafic maritime
  
  • Le Berre Lise
  • Dufresne Marvin
  • Oppo Sonia
  • Tinel Liselotte
  • Ferreira de Brito Joel
  • Salameh Thérèse
  • Temime-Roussel Brice
  • Marchand Nicolas
  • D’anna Barbara
  • Lanzafame Grazia Maria
  • Wortham Henri
  • Armengaud Alexandre
  • Roustan Yelva
  • Sauvage Stephane
  , 2024, pp.144. Les émissions liées au trafic maritime ont un impact avéré sur la qualité de l’air en particulier sur les zones côtières. La réglementation sur la composition des carburants utilisés a évolué notamment pour le soufre mais reste peu contraignante pour les particules fines et certains de leurs précurseurs. Les inventaires d’émissions prennent en compte le trafic maritime mais affichent une variabilité importante et une connaissance encore parcellaire quant à la composition des particules émises et des principaux précurseurs d’aérosols secondaires. Ce Projet PAREA a pour but d’améliorer les connaissances sur les émissions du transport maritime. Sur la base d’une approche combinant observation et modélisation, l’objectif est de caractériser les particules fines émises par le trafic maritime en termes de tailles et de composition chimique, de caractériser les précurseurs de particules secondaires émis par le trafic maritime et de documenter l’évolution de ces caractéristiques en champs proches de la zone portuaire. Une campagne de mesure intensive a été menée à Marseille pour l’obtention d’une base de données uniques de 145 composés gazeux et particulaires. La comparaison des concentrations mesurées sur le site de fond urbain et sur les sites implantés au cœur de la zone portuaire montrent clairement une influence des activités portuaires sur les niveaux de concentrations et notamment les maximas avec des évènements de pollution courts mais intenses. Plus de 350 panaches de navires ont été identifiés pour déterminer des Facteurs d’Emissions (EF). Ces EF constituent des données précieuses, issus de mesures en champs proches et donc représentatives d’un ensemble de navires sur la zone portuaire de Marseille. Les résultats mesurés peuvent être jusqu'à cinq fois inférieurs à ceux utilisés dans les cadastres d'émissions, en particulier pour les NOX et le SO2. Concernant la phase particulaire, les quantités de particules émises par les navires peuvent varier d’un facteur 3 entre les différentes phases opérationnelles avec davantage d’émissions lors des phases de navigation et/ou de manœuvre comparativement aux phases de stationnement à quai. L’étude montre que la combinaison de la granulométrie et de la composition chimique des particules fournit des informations clés pour l’amélioration des inventaires d’émission et l’identification des différents types de carburants utilisés ainsi que l’emploi de systèmes d'épuration des fumées. Une analyse factorielle a été conduite montrant que les émissions liées aux navires représentent environ 9 % de la masse totale des métaux mesurés en champs proche (zone portuaire) et 4 % en zone urbaine plus éloignée. En considérant la nature chimique, ces émissions contribuent spécifiquement à plus de 80 % des concentrations en nickel (Ni) et en vanadium (V), métaux reconnus pour leurs effets néfastes sur la santé des populations. Cette même analyse menée sur les COV montre une contribution totale de 18 % des concentrations en COV mesurés en champs proches et 11% en zone urbaine plus éloignée. En deuxième approche, la modélisation déterministe a été mise en œuvre pour cartographier l’impact des activités maritimes à Marseille. Le constat est en cohérence avec celui des observations avec un impact chronique des navires peu important sur la ville mais l’occurrence de panaches de courte durée qui peuvent avoir localement un impact important. Le détail des contributions par famille d’espèce chimique montre que ces émissions peuvent avoir un impact différencié en zone éloignée pour la formation d’aérosols secondaires. L’étude montre toute la difficulté pour l’obtention d’une bonne représentation des émissions de la sources trafic maritime et de ses impacts. L’amélioration des inventaires d’émissions en termes de spéciation chimique et d’évolution temporelle de cette source est essentielle pour évaluer des actions de remédiation.
• Machine learning, data assimilation and dynamical systems
  
  • Malartic Quentin
  , 2024. The chaotic dynamics and the sparse and noisy observations of geophysical systems, particularly in the domains of meteorology, climate science, and oceanography, demand sophisticated methodologies for accurate state or parameter estimation. This thesis explores, both theoretically and experimentally, the synergy between traditional Data Assimilation (DA) techniques and the recent surge in Machine Learning (ML). My focus is specifically on the joint estimation of both state variables and parameters, and on the training of ML models aimed to be later used in DA setups.Such ML models, that are non entirely physical, and in some cases even fully statistical, are referred to as surrogate models. The relevancy of such a model can lie either in its improved accuracy, or in its computational efficiency, for example if it can achieve a similar accuracy at a reduced computational cost.In the pursuit of more accurate state estimations and forecasts, requiring better surrogate models, as well as computationally cheaper surrogate models, the development of methodologies combining ML and DA becomes crucial. The flexibility of ML techniques, ranging from conventional statistical methods to advanced deep learning architectures, in synergy with the well established methods of DA, form a powerful toolkit to significantly enhance prediction quality in the context of sparse and noisy observational data of chaotic dynamics.Typical geophysical systems like the ocean and the atmosphere are governed by local equations, where the temporal evolution of the system at a given point in space only depends on its neighboring state. The presented methods, both for DA and ML, not only take into account this particularity, but are crafted around it, with the aim of producing better models, making more accurate predictions and estimations, as well as getting an algorithmic complexity advantage from it.On the one hand, the combination of DA and ML will be studied in the context of joint estimation, where both the ML model and the dynamical system state are estimated, and updated, in an online fashion, as new observational data is acquired. The developed algorithms will be tested extensively in toy, fictive experimental setups. Nevertheless, the methods and experimental setups developed will be focused on numerical weather prediction (NWP) applications, and will be designed in a way that makes them scalable in this context.On the other hand, the training of ML models on past observational data will be studied, and their performance will be evaluated both in term of future forecasts, and in the context of classical DA experiments. In this case, the experimentation will be done using a Quasi Geostrophic model, following the implementation of Marshall and Molteni (1993), model representing the large scale atmospheric dynamics, being relevant especially in the context of boreal winter in the mid-latitudes D'Andrea and Vautard (2001). (10.70675/d7feb169z7520z4c44za5efzed912837dfc6)
  DOI : 10.70675/d7feb169z7520z4c44za5efzed912837dfc6
• Impact of solid road barriers on reactive pollutant dispersion in an idealized urban canyon: A large-eddy simulation coupled with chemistry
  
  • Lin Chao
  • Ooka Ryozo
  • Kikumoto Hideki
  • Flageul Cédric
  • Kim Youngseob
  • Zhang Yang
  • Sartelet Karine
  Urban Climate, Elsevier, 2024, 55, pp.101989. This study conducts chemistry-coupled large-eddy simulations on reactive gaseous and particulate pollutant dispersions in an idealized street canyon. Four road-barrier configurations are considered: barrier-free, side barriers, center barrier, and the combination of side and center barriers. Regarding the formation of secondary aerosols, the center barrier reduces the canyon-averaged mass concentration of particulate matter (PM) but increases the inorganic particle concentration. This is because nitric acid (HNO3) is limited in the formation of ammonium nitrate due to the long residence time in the street, and the center barrier increases the HNO3 inflow from the background air. The side barriers increase the canyon-averaged PM10 mass concentration but reduce PM number concentration due to sufficient time for coagulation. Using combined barriers reduces the most PM10 mass concentration and number concentration from the barrier-free case. Regarding the formation of secondary gases, the side barriers enhance the NO2 formation due to the worse ventilation. In contrast, the center barrier largely reduces the canyon-averaged NO concentration but slightly reduces the NO2 concentration from the barrier-free case, because the center barrier increases the O3 inflow from the background air favoring NO2 formation. Additionally, the combined barriers show smaller canyon-averaged NO2 and O3 concentrations than the center barrier. From the view of controlling reactive pollutants, urban planners are recommended to apply the combined barriers to enhance better air quality in street canyons, rather than using either side barriers or a center barrier alone. (10.1016/j.uclim.2024.101989)
  DOI : 10.1016/j.uclim.2024.101989