Are cities on track to net-zero emissions? The next generation of greenhouse gas information systems

06/09/2021 00:00

Impact of SST Fronts on the Atmospheric Water Cycle and Storm-track Activity

12/06/2025 14:00

Séminaire du LMD à l’ENS.

Essential roles of synoptic-scale transient eddies in frontal air-sea interactions

12/06/2025 11:30

Séminaire du LMD à l’ENS.

Reexamination of the Southern Reexamination of the Southern Hemisphere baroclinic annular mode by separating sub-weekly and lower-frequency disturbances

12/06/2025 11:00

Séminaire du LMD à l’ENS.

Impact of the spatial and temporal variability of the Mistral on dense water formation in the Mediterranean Sea

21/10/2022 14:00

Deep convection is the process where surface water is cooled to the point it becomes dense enough that it sinks to the seafloor. In the Gulf of Lion, the dense water it produces helps the general circulation of the Mediterranean Sea. In this region, the cooling is mainly caused by the Mistral winds and seasonal change of the atmosphere. This thesis explores the individual and relative cooling provided by these two sources, finding that both sources contribute to the cooling, with the seasonal change providing more than the Mistral. However, in the future neither are able to overcome the predicted changes to the gulf due to climate change, and deep convection is believed to stop altogether. Due to this, part of the thesis looks at methanol producing islands; devices that could help mitigate climate change and energy availability, especially in the Mediterranean.

Local and remote sources of Arctic air pollution

09/09/2022 10:00

The Arctic region is warming faster than any other region on Earth due to the effect of greenhouse gases, notably CO₂, and short-lived climate forcers of anthropogenic origin, such as black carbon (BC). Over the last 20-30 years, remote anthropogenic emissions over mid-latitude regions have been decreasing. Anthropogenic emissions within the Arctic are also contributing and might increase in the future and further affect Arctic air pollution and climate. Natural emissions, such as sea-spray aerosols, also might increase due to on-going climate change. However, the processes and sources influencing Arctic aerosols and trace gases are poorly quantified, especially in wintertime.

In this thesis, quasi-hemispheric and regional simulations are performed using the Weather Research Forecast model, coupled with chemistry (WRF-Chem). The model is used to investigate atmospheric composition over the wider Arctic and during two field campaigns, one in northern Alaska at Barrow, Utqiaġvik in January and February 2014 and the second in Fairbanks, central Alaska in November and December 2019 during the French pre-ALPACA (Alaskan Layered Pollution And Chemical Analysis) campaign.

First, modelled inorganic and sea-spray (SSA) aerosols are evaluated at remote Arctic sites during wintertime. Then, the model is improved with respect to SSA treatments, following evaluation against Barrow field campaign data, and their contribution to the total aerosol burden within the Arctic region is quantified. A series of sensitivity runs are performed over northern Alaska, revealing model uncertainties in processes influencing SSA in the Arctic such as the presence of sea-ice and open leads. Second, a sensitivity analysis is performed to investigate processes and sources influencing wintertime BC over the wider Arctic and over northern Alaska, with a focus on removal treatments and regional emissions. Variations in model sensitivity to wet and dry deposition are found across the Arctic and could explain model biases. Over northern Alaska, regional emissions from petroleum extraction are found to make an important contribution to observed BC.

Model results are also sensitive to planetary boundary layer parameterisation schemes. Third, the improved version of the model is used to investigate the contribution of regional and local sources on air pollution in the Fairbanks area in winter 2019. Using up-to-date emissions, the model performs better in winter 2019 than in winter 2014, when compared to observations at background sites across Alaska. Underestimations in modelled BC and sulphate aerosols can be partly explained by lacking local and regional anthropogenic emissions. In the case of sulphate, additional secondary aerosol formation mechanisms under dark/cold conditions also need to be considered.

Approches statistiques multivariées pour l'ajustement des biais des simulations climatiques et l'analyse des événements composés

23/09/2022 10:00

Le climat est un système complexe qui est le résultat de multiples interactions entre ses différentes composantes et ses multiples variables. Cette thèse a pour but d’évaluer si et comment l’utilisation d’approches statistiques multivariées pour l’étude des simulations climatiques peut contribuer à une compréhension plus approfondie du changement climatique et des événements climatiques à forts impacts sur la société. Pour répondre à ces questions, je propose et applique de nouveaux outils statistiques multivariés pour, d’une part, la correction de biais des simulations climatiques, et d’autre part, l’étude des changements de probabilités d’événements conjoints à forts impacts. Le travail s’articule autour de trois objectifs : (i) comparer des méthodes de correction de biais multivariés (MBC) déjà existantes, (ii) développer une nouvelle méthode MBC pour l’ajustement des dépendances spatiales des simulations climatiques, (iii) évaluer la période d’émergence des probabilités d’événements conjoints et quantifier la contribution des propriétés univariées et multivariées aux changements de ces probabilités.

La comparaison de méthodes de correction de biais a permis d’une part d’informer les utilisateurs de leurs avantages et leurs inconvénients mais aussi d’identifier des pistes de développements pour de nouvelles méthodes. Une nouvelle méthode, basée sur une technique de Machine Learning appelée réseaux adverses génératifs (CycleGAN), a été développée. Elle donne des résultats satisfaisants, montrant ainsi le potentiel des techniques de Machine Learning pour la correction de biais multivariés. L’évaluation de la période d’émergence des probabilités d’événements conjoints, ainsi que la quantification de la contribution des propriétés univariées et multivariées aux changements de ces probabilités se révèlent être une procédure pertinente pour améliorer la compréhension de tels phénomènes climatiques. Il est trouvé que la non-stationnarité de la structure de dépendance inter-variable dans un contexte de changement climatique peut jouer un rôle important dans les probabilités futures d’événements conjoints.

Les travaux réalisés dans cette thèse ouvrent des perspectives pertinentes en termes méthodologiques mais participent aussi à une amélioration de la compréhension du climat et de ses évolutions en fournissant des outils statistiques adaptés à la nature intrinsèquement multivariée du système climatique.