Missions

The postdoctoral researcher will be part of the CECI team responsible for developing and evaluating a data assimilation workflow relevant for wildland fire behavior prediction. He/she will work under the supervision of Mélanie Rochoux within the framework of the ANR FIREFLY project. He/she will be a member of the collaboration with CNRM/Météo-France, University of Corsica, University Polytechnic of Catalonya and Worcester Polytechnic Institute, relating to uncertainty quantification and reduction of coupled atmosphere-fire modeling.

Activités

Context
Anticipating wildfire behavior has recently become a key operational and scientific issue due to the emergence of extreme wildfire events in many ecosystems across the world and in particular in mediterranean forest ecosystems that are particularly exposed to climate change. Extreme fire behavior features very high rate of spread (ROS) and fireline intensity, crowning, spotting and even pyroconvection, resulting from the combined effects of topography, meteorology and biomass fuel complexity on a wide range of scales. To assist in fire crisis management, there is a need to develop an on-demand simulation capability able to represent the complexity of wildland fire behavior, involved in both the fire spread and the fire plume, and to estimate the plausible fire scenarios at the scale of an event. Such a capability requires the fusion (or assimilation) of all available sources of information on a given wildland fire event, i.e. the best physically-based models and the available infrared measurements, and their associated uncertainties. As part of the ANR FIREFLY project, the postdoctoral researcher will contribute to the development a data assimilation framework relevant for wildland fire modeling combining coupled atmosphere-fire modeling and infrared imaging data.

Coupled atmosphere-fire modeling tool
At Cerfacs, in a close collaboration with CNRM, we work since 2020 on the development of the coupled Meso-NH/BLAZE atmosphere-fire model (Costes et al. 2021) to explore atmosphere-fire interactions at landscape to meteorological scales. Such a coupled model simulates the propagation of the fire front over the landscape and the micrometeorology in the vicinity of the wildland fire (in particular, the fire-induced wind near the surface).

Objectives
The novelty of the postdoctoral work lies in the development of an efficient and robust ensemble-based data assimilation process capable of correcting uncertain biomass fuel input parameters and the fire front position, and of handling fire front position errors to produce more accurate coupled atmosphere-fire model predictions for a given fire event.

The postdoctoral work is organized in four steps:
1) Sensitivity analysis to identify the most influential biomass fuel parameters on fire spread and fire-induced wind and that are the most important to infer through data assimilation (Rochoux et al. 2014; Allaire et al. 2020). Collaboration with Patrick Le Moigne and Margaux Peyrot (PhD student) from CNRM will be useful for this part, based on the ongoing work on fuel modeling.

2) Emulating the coupled atmosphere-fire model using machine learning adapted from the work of Lumet et al. (2025). The emulator will be trained offline to learn the response of the coupled atmosphere-fire model to changes in the selected uncertain biomass parameters from step 1.

3) Building the data assimilation workflow around the emulator from step 2 to accelerate the parameter estimation step and evaluating it using idealized data assimilation experiments. The way to best represent the observations within the data assimilation will be studied (front positions versus front arrival times).

4) Applying data assimilation to an experimental fire in a mixed forest environment in the New Jersey Pinelands National Reserve conducted by Worcester Polytechnic Institute, for which airborne infrared images are available at high resolution (Mueller et al. 2017).

References
- Allaire et al. (2020), Generation and evaluation of an ensemble of wildland fire simulations, International Journal of Wildland Fire http://dx.doi.org/10.1071/WF19073
- Costes et al. (2021), Subgrid-scale fire front reconstruction for ensemble coupled atmosphere-fire simulations of the FireFlux I experiment, Fire Safety Journal, http://dx.doi.org/10.1016/j.firesaf.2021.103475
- Lumet et al. (2025), Uncertainty-aware surrogate modeling for urban air pollutant dispersion prediction, Building and Environment, http://dx.doi.org/10.1016/j.buildenv.2024.112287
- Mueller et al. (2017), Utilization of remote sensing techniques for the quantification of fire behavior in two pine stands, Fire Safety Journal, https://doi.org/10.1016/j.firesaf.2017.03.076
- Rochoux et al. (2014), Towards predictive data-driven simulations of wildfire spread – Part I: Reduced-cost Ensemble Kalman Filter based on a Polynomial Chaos surrogate model for parameter estimation, Natural Hazards and Earth System Sciences, http://dx.doi.org/10.5194/nhess-14-2951-2014

Compétences

We are seeking a motivated and enthusiastic early-career researcher to join our team at CECI/Cerfacs and making progress on wildland fire modeling through data assimilation.

Candidates must hold a recent PhD in atmospheric science or related discipline with experience in applied mathematics, or a recent PhD in machine learning and/or data assimilation with strong interest in atmospheric and fire science questions.
Candidates must have proven research skills evidenced by a least one publication as the first author relating to the subject or activities of the project. Other essential criteria for this job are a sound knowledge of computer skills and numerical modeling.

Candidates must have a very positive attitude to working in a team. Fluency in spoken and written English is a requirement.

Applicants are asked to send a CV, a cover letter, and the names and e-mail addresses of two professional references through the CNRS job portal (emploi.cnrs.fr). An initial selection phase will be based on the application. Selected applicants will be contacted to an interview early September 2025.

Contexte de travail

The CECI research unit is a joint laboratory between the European Center for Research and Advanced Training in Scientific Computing (Cerfacs), the French National Center for Scientific Research (CNRS) and the French National Research Institute for Development (IRD).

CECI includes about 30 researchers and early-career researchers with strong expertise on climate and environmental models, high-performance computing, simulation workflows and data management. We conduct cutting edge research spanning from climate variability and prediction, oceanography and polar science, air-sea interaction, climate change detection and attribution and its impacts, to extreme events such as heat waves, intense precipitation events and droughts as well as environmental risks such as atmospheric pollutant dispersion, wildland fires and floods. We use a wide range of numerical models from large-eddy simulation to global Earth system models and associated algorithms (data assimilation, uncertainty quantification, machine learning, code coupling) to tackle our science challenges.

On the wildfire topic, CECI closely works with CNRM/Météo-France and more recently with CNES. CECI also benefits from collaborations with University of Corsica, Polytechnic University of Catalonia (UPC), INRAE and CESBIO (Observatoire Midi-Pyrénées).