Aerosols and in particular their black carbon (BC) content influence the atmospheric heating rate and fog dissipation. Substantial improvements have been introduced to the solar scheme of the computational fluid dynamic model code_saturne to estimate fluxes and heating rates in the atmosphere. This solar scheme is applied to a well-documented case of a fog that evolves into a low stratus cloud. Different sensitivity tests are conducted. They show that aerosols have a major effect with an overestimation of the direct solar fluxes by 150 W m${}^{-2}$ when aerosols are not considered and a reduction of the heating of the layers. Aerosols lead to an increase of the heating rate by as much as 55% in the solar infrared (SIR) band and 100% in the Ultra-Violet visible (UV-vis) band. Taking into account the fraction of BC in cloud droplets also accentuates the heating in the layers at the top of the fog layer where water liquid content is maximum. When the BC fraction in cloud droplets is equal to 8.6 × 10${}^{-6}$, there is an increase of approximately 7.3 °C/day in the layers. Increasing the BC fraction leads to an increase of this heating in the layer, especially in the UV-vis band.

Solar radiation modelling is important for the evaluation and deployment of solar renewable energy systems. The amount of solar radiation reaching the ground is influenced by geographical parameters (seasons, latitude and local characteristics of the site) and meteorological and atmospheric parameters (like humidity, clouds or particles). Those parameters have important spatio-temporal variations that make solar radiation hard to model. Various radiation models exist in literature. Among them, the 1D radiation model part of the computational fluid dynamics software “Code_Saturne” estimates the global and direct solar irradiances at the ground. It takes into account the impact of meteorology, atmospheric gas, particles and clouds whose influence is represented using the two-stream approximation. The model showed satisfactory results during clear-sky days  but not during cloudy-sky days. It is a common problem in solar radiation modelling, because of the complexity to accurately represent  clouds, which are extremely variable in space and time and have a strong influence on the depletion of solar irradiance.  In the current study, the estimation of radiation during cloudy-sky days is improved by coupling the 1D radiation model of Code_Saturne with on-site and satellite measurements of the cloud optical properties. Meteorological data are obtained from the Weather Research and Forecasting (WRF) model, aerosol’s concentrations from the air-quality modelling platform Polyphemus, and on-site measurements from the SIRTA observational site (close to Paris). Two periods are simulated: 'august 2009' and 'year 2014'. It is shown that the introduction of the measured cloud properties in the computation of the surface radiation fluxes leads to a strong reduction of the simulated errors, compared to the case where these properties are derived from the WRF model.                    A sensitivity analysis on the parameters representing clouds in the model is conducted. It enabled us to identify the most influencing parameters - cloud optical thickness (COD) and cloud fraction - and instruments that are sufficient and mandatory for a good description of solar radiation during cloudy-sky days. A fitted model is developed to deduce the COD from liquid water path measurements. Satellite and radiometric measurements could both be used, although satellite measurements are not always available.  For the estimation of cloud fraction, the best results are obtained from shortwave radiometric measurements or from a sky imager. Moreover, large error cases in hourly values of solar fluxes are examined to understand their origin. For a large part of these error cases, there is a high variation within the hour of satellite or in situ measurements, or the presence of low clouds (in more than 50% of these cases in august 2009).  

Cloud condensation nuclei (CCN) are the subset of aerosol particles able to form cloud droplets. CNN activation is influenced by the size distribution, chemical composition and number of particles. They consequently impact the cloud microstructure, which affects the radiative properties of clouds, atmospheric circulation and thermodynamics, as well as radiative budgets. By influencing the single scattering albedos of clouds, some particles lead to an increase of the solar irradiation absorption and solar heating in the cloud layers. A good example of these absorbing particles is those made of black carbon (BC), which is emitted during the combustion of various types of fuel and non-exhaust traffic-related processes. The present study deals with the role of BC in a solar radiative scheme and its interaction with clouds during a well-documented case of a fog that evolves into a low stratus cloud. To do so, the solar scheme of the computational fluid dynamic model Code Saturne is used for the estimation of fluxes and heating rates in the atmosphere. It is based on the two-stream parameterization with calculations done in the ultraviolet-visible (UV-Vis) and solar infrared (SIR) bands. A special attention is given on the impact of BC on the dissipation of the fog. As expected, the introduction of BC in cloud droplets accentuates the heating in the layers at the top of the cloud where water liquid content is maximum. In the SIR band, there is an increase of approximatively 80 %. In the UV-Vis band, where absorption of solar irradiance by ozone is minor, the heating rate is now 10 times higher. The contribution of the UV-Vis band becomes more important. The augmentation of solar heating leads to a reduction of the liquid water content and, consecutively, to a faster dissipation of the fog and the stratus. Therefore, direct surface fluxes are also increased.When increasing the volume fraction of black carbon in cloud droplets, the water liquid content is furthermore reduced leading to a faster dissipation of the fog. However, this impact is small, because the fog is formed in the morning. At this time, the cooling rate due to thermal radiation is higher than the solar heating at the top of the cloud. We expect the impact of black carbon in cloud droplets to be higher for more persistent clouds or for a fog in the boundary layer of the urban atmosphere, where the fraction of BC in particles is higher.

Modélisation du rayonnement solaire pour l’optimisation photovoltaïque (PV): Un fort développement de la filière énergie solaire est prévu pour les prochaines années en France et dans le monde. Une modélisation précise du rayonnement solaire est nécessaire pour minimiser les incertitudes sur l’estimation du productible de futures fermes et optimiser leur conception. Cependant, le rayonnement solaire au sol est fortement influencé par plusieurs paramètres géographiques, météorologiques et atmosphériques, dont notamment les caractéristiques des nuages, et celles des aérosols. L’objectif de cette thèse est d’améliorer la modélisation du rayonnement solaire, en mettant l’accent sur celles dues à la présence de nuages et d’aérosols. Des améliorations ont d’abord été apportées au modèle de rayonnement 1D du logiciel CFD code_saturne. Ce modèle estime dorénavant le rayonnement global, direct et diffus, en prenant en compte l’impact des gaz atmosphériques, des aérosols et des nuages. Les simulations sont effectuées au SIRTA, observatoire atmosphérique en Ile-de-France, et comparées aux mesures de rayonnement effectuées dans cet observatoire. Une bonne estimation du rayonnement en ciel clair est obtenue en prenant en compte la présence d’aérosols dont les propriétés optiques sont estimées grâce à un chaînage avec la plateforme Polyphemus. Les nuages conduisent notamment à une forte atténuation du rayonnement et ils sont très complexes à modéliser en raison notamment de leur forte variabilité spatio-temporelle. Afin d’améliorer la modélisation en ciel nuageux, le modèle est couplé aux mesures sur site pour déterminer les propriétés optiques des nuages (épaisseur optique, fraction nuageuse). Différents tests de sensibilité sont conduits afin de mieux comprendre l’impact de ces propriétés et les instruments utilisables pour les mesurer. L’analyse de la comparaison entre rayonnement mesuré et calculé (réalisée pour le rayonnement global et sa composante directe) est basée à la fois sur des scores statistiques globaux et sur une étude détaillée des causes possibles des erreurs les plus importantes. Un second axe de la thèse a consisté à appliquer et valider ce modèle de rayonnement sur à un cas bien documenté de brouillard radiatif au SIRTA (campagne Paris-Fog) évoluant en stratus bas après sa dissipation. L’accent est porté sur la prise en compte des aérosols et notamment du carbone suie dans la composition des gouttelettes lors de la phase de dissipation, ainsi que sur l’hypothèse considérée pour la fraction nuageuse. Dans un troisième axe, des améliorations sont apportées au modèle de rayonnement 3D de code_saturne pour la prise en compte des aérosols et nuages et en vue de son application à la modélisation d’une ferme PV. Une vérification du comportement du modèle est faite sur un cas d’interaction avec un obstacle, avec des comparaisons entre les résultats des modèles 1D et 3D.