Clément Caron

Clément Caron

Auto-generated profile, based on publicly available information

Clément Caron


Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
8 November 2024 in Building and Environment
Verified
Journal Article
Open access
CCClément Caron
Clément Caron
PLPhilippe Lauret
Philippe Lauret
ABAlain Bastide
Alain Bastide

22 June 2026
Verified
Conference Paper
Open access
FKFrederik Kurzrock
Frederik Kurzrock
DSDavid Schuhbauer
David Schuhbauer
CCClément Caron
Clément Caron
NSNicolas Schmutz
Nicolas Schmutz
Accurate forecasting of atmospheric stability is critical for predicting convection and associated severe weather hazards, including thunderstorms, hail, and damaging wind gusts. Atmospheric stability indices (SIs) are routinely used to assess convective potential, yet their reliability across different data sources remains incompletely understood. This study evaluates the forecast skill of stability indices derived from two operational numerical weather prediction (NWP) models: the Integrated Forecast System (IFS, ECMWF) and the Global Forecast System (GFS, NCEP) alongside satellite-derived estimates from the NOAA-LAP product, using high-resolution Global Climate Observing System Reference Upper Air Network (GRUAN) radiosonde observations at Lamont (USA) as reference truth. The analysis focuses on five commonly employed indices: K-Index, Total Totals, Lifted Index, Convective Available Potential Energy (CAPE), and Precipitable Water. Results demonstrate that forecast accuracy varies significantly across indices. Precipitable Water exhibits the highest correlation and lowest normalized RMSE, whereas CAPE proves most challenging to predict, with substantially higher errors across all data sources. Among the three products, IFS consistently delivers superior performance, achieving the highest correlation coefficients and lowest errors for nearly all indices. Notably, the satellite-based NOAA-LAP nowcasting product, despite its higher temporal refresh rate and spatial resolution, does not surpass the hourly forecasts from the global NWP models. This underscores the intrinsic uncertainty in estimating atmospheric stability from either approach. The IFS advantage is likely attributable to its finer vertical resolution (137 hybrid levels versus 20 standard pressure levels in GFS), which better captures the thermodynamic structure essential for parcel-based indices. These findings have important implications for severe weather forecasting and early warning systems. While stability indices remain valuable operational tools, their limitations in predicting convection suggest that reliance on conventional scalar metrics alone is insufficient. Future work should explore richer representations of the atmospheric column or probabilistic frameworks that preserve more thermodynamic information for convective onset prediction.

22 June 2026
Verified
Conference Paper
Open access
CCClément Caron
Clément Caron
AVAnthony Voitus
Anthony Voitus
BABenjamin Adrien
Benjamin Adrien
MTMathieu Turpin
Mathieu Turpin
NSNicolas Sébastien
Nicolas Sébastien
NSNicolas Schmutz
Nicolas Schmutz
Tropical regions exhibit high solar potential, making them well suited for photovoltaic deployment. Moreover, in non-interconnected zones, such as islands, solar power generation can contribute substantially to electricity supply. In this context, accurate intra-hour forecasts of surface solar irradiance facilitate the operation of photovoltaic plants and their smooth integration into the power grid. However, short-term (0-1h) irradiance forecasting remains particularly challenging on tropical islands due to the high variability of cloud cover.Cloud cover forecasting techniques based on satellite observations are of great value for ground-level solar irradiance forecasting. Among them, optical flow algorithms for deriving Cloud Motion Vectors (CMVs) are now considered a standard in the field. Recently, deep learning approaches, especially U-Net architectures, have shown strong promise in improving the short-term anticipation of cloud positions. However, the extent to which these models can enhance subsequent irradiance forecasting in tropical environments remains to be investigated.This study introduces a lightweight U-Net-based model designed to simultaneously forecast four future clear-sky index (Kc) maps, with lead times of up to one hour. The model takes four past Kc maps as input, which are processed from Meteosat-9 visible channels (Indian Ocean Data Coverage, 45.5°E). First, distinct models are trained using 2.5 years of satellite data. While sharing the same architecture and training procedure, each model is tailored to a specific territory in the South-West Indian Ocean through dedicated training data. Then, independent evaluations are performed on a separate year. Beyond the assessment of forecast Kc map quality, surface global horizontal irradiance (GHI) is derived from model outputs at specific station locations. These values are then compared with ground-based GHI measurements from the Indian Ocean Solar Network (IOS-net).The results demonstrate that our U-Net-based models improve spatial Kc map forecasts by approximately 10-15% over standard CMV methods, in terms of mean absolute error (MAE) and root mean square error (RMSE). Consequently, local GHI forecasts are typically enhanced by 5-10% across all territories tested. In addition, the models exhibit promising cross-island generalisation, suggesting that they capture robust location-invariant patterns. These findings point to the potential for developing a unified model for tropical islands, as well as for leveraging transfer learning approaches, which could simplify the practical deployment of such models from an operational perspective. Overall, this study establishes U-Net-based models as valuable candidates for improving intra-hour GHI forecasts over tropical islands in the South-West Indian Ocean, with the potential to further increase photovoltaic penetration in the local energy mixes.

9 April 2026
Verified
Thesis
Closed access
CCClément Caron
Clément Caron
La mécanique des fluides numérique apporte une contribution précieuse à l’amélioration de la conception des bâtiments, offrant la possibilité de simuler avec précision des écoulements complexes, tels que les mouvements d'air ou de fumée. À La Réunion, les particularités géographiques, météorologiques et architecturales renforcent l'intérêt pour ce type de modélisation, comme l'illustrent les conflits entre ventilation naturelle et désenfumage. Cependant, le temps de calcul demeure un obstacle majeur. Les récents progrès en intelligence artificielle incitent à se demander si des modèles statistiques sont capables de décrire avec suffisamment de précision ces phénomènes physiques, et ce à moindre coût. Dans ce contexte, cette thèse vise à examiner si les méthodes d’apprentissage automatique peuvent contribuer à accélérer les simulations de mécanique des fluides numérique pour des études d’ingénierie de désenfumage des bâtiments. L'analyse de la littérature met en évidence que, malgré de nombreuses initiatives, le manque de garanties physiques et de solutions adaptées à des cas réels freine encore le déploiement de ces approches. Pour répondre aux limitations identifiées, un solveur hybride intégrant un modèle statistique à la méthode des volumes finis a été développé. En particulier, nous avons montré que des prédictions locales de l'évolution du champ de pression peuvent fournir une approximation initiale pertinente pour accélérer la résolution itérative de l'équation de pression. Ce solveur hybride a été évalué sur des écoulements transitoires incompressibles de propagation d’un panache thermique, prouvant sa capacité à réduire les temps de calcul pour des simulations liées à un domaine spécifique tout en préservant la précision des résultats. Bien que cette démonstration s’inscrive dans un contexte académique, elle constitue une avancée encourageante vers une hybridation efficace entre modèles statistiques et numériques, capitalisant sur leurs forces respectives pour des applications plus ambitieuses.