Search

The ‘Balbi model’ is a simplified rate of fire spread model aimed at providing computationally fast and accurate simulations of fire propagation that can be used by fire managers under operational conditions. This model describes the steady-state spread rate of surface fires by accounting for both radiation and convection heat transfer processes. In the present work the original Balbi model developed for laboratory conditions is improved by addressing specificities of outdoor fires, such as fuel complexes with a mix of live and dead materials, a larger scale and an open environment. The model is calibrated against a small training dataset (n = 25) of shrubland fires conducted in Turkey. A sensitivity analysis of model output is presented and its predictive capacity against a larger independent dataset of experimental fires in shrubland fuels from different regions of the world (Europe, Australia, New Zealand and South Africa) is tested. A comparison with older versions of the model and a generic empirical model is also conducted with encouraging results. The improved model remains physics-based, faster than real time and fully predictive.

International audience

International audience; Nowadays, the needs for decision making tools useful for people involved in firefighting and/or in landscape management becomes more and more crucial, especially with the dramatic increase of the fire dangerousness and fire severity. These tools have to be accurate enough and faster than real time. Up to now, simulators and other tools are mainly based on empirical or semi-empirical models but the lack of physics in their formulation is a major flaw. The Balbi model is a simplified physical propagation model for surface fires which explicitly depends on the topography, the wind velocity and several fuel characteristics. It is a set of algebraic equations built from usual physical conservation laws (mass, momentum etc.) with some strong assumptions. This work aims at providing a new version of the Balbi model in which the resolution of the rate of spread (ROS) does not need any iterative method any more. This simplification is helpful in implementing the equations set into a fire propagation simulator or a coupled fire-atmosphere simulator. It needs a complete change in the structure of the model and the predicted ROS was tested at the field scale against 179 shrubland fires (burnt in Australia, South Africa, Turkey, Portugal, Spain, New Zealand) and 178 Australian grassland fires with a very good agreement with the observed ROS. Two statistical tools are used to check this agreement (Normalized Mean Square Error, NMSE and Mean Absolute Percentage Error, MAPE) and the Fractional Bias (FB) aims at understanding when the model over-predicts or under-predicts the ROS. The proposed model is accurate and its model parameters are calibrated against a small training dataset which makes it fully predictive whatever the environmental and topographic conditions and the fuel bed characteristics. Its more simple structure allows it to be a good candidate for the heart of a simulation or land management decision making tool.

International audience; The GOLIAT project is a consortium of academics and firefighting operators and land-use planning professionals of Corsica. One goal of GOLIAT project is to provide four operational decision support tools. To reach this goal, a survey of past fires occurred in Corsica since the twentieth century beginning is made. This inventory contributes to build up a database with a web display interface easy to use as fire patterns history. A fire behavior and impact simulator prototype for vegetation fires, a geolocation tool for hot spots using UAV images, and a guide of good practices of prescribed fires in the undergrowth are building. At the same time, experimental fires are carried out to improve knowledge about high intensity fire and the experimental results were compared to the predictions provided by a complete physical 3D model, namely FireStar3D.

International audience; Eruptive fires are one category of extreme fire behaviour. They are characterized by a sudden and unpredictable change in the fire behaviour which represents an extreme danger for people involved in firefighting. The major point is about the mechanism that turns a usual fire behaviour into an eruptive fire behaviour. Among the different explanations found in the literature, the pioneering interpretation consisting in a feedback effect caused by the convective flow induced by the fire under wind and/or slope conditions, has never been disproved with an example of fire accident. The main goal of this work lies in proposing a physical modelling of this fire induced wind. This modelling attempt is derived from the brand-new version of the Balbi model, which is a simplified physical model for surface fires at the field scale that explicitly depends on the triangle of fire (fuel bed, wind and slope). This work is a first step to the modelling of fire eruption. The model tries to represent accurately the acceleration of the fire rate of spread propagating on different sloped terrain under no-wind or weak wind conditions. It is tested against three sets of experiments carried out at the laboratory scale without external wind and against a high intensity experimental fire spreading on a steep sloped terrain and conducted under weak wind conditions in the north-western of Corsica. Some statistical tools are used to compare predicted and observed rate of spread (NMSE, Normalized Mean Square Error and MAPE, Mean Absolute Percentage Error) and to understand the model’s under-predictions or over-predictions trends (FB, Fractional Bias).

Dimensioning a fuelbreak remains always a challenging problem. For a long time, this problem was tackled using an empirical approach from the experience of operational users such as the fire fighters and the foresters. During the last decades, new approaches coming from fire safety engineering have completed the set of tools adapted to study this problem. These tools are all based on physical considerations, more or- less sophisticated. The simplest ones, consist in assimilating the flame as a radiant panel, calculating the distribution of radiant heat flux as a function of the distance separating the flame to a potential target and defining at what distance this heat flux reached a critical threshold level susceptible to produce damages on this target (pain for people or ignition for materials). The most complex ones, consist in solving the conservation equations (mass, momentum, energy ...) governing the behaviour of complex coupled problem formed by the vegetation, the flame front and the surrounding atmosphere. This new generation of engineering tool, based on CFD approach allows to directly predict the behaviour of a fire front propagating toward a fuelbreak, in order to evaluate its efficiency as a function of the amount of surface fuel (grass, shrubs) removed to reduce locally the fuel load and therefore the intensity of an incoming fire. These two approaches are fully complementary, only the first one has the potentiality to be spread operationally on the field, whereas the second one can contribute to improve the first one and to study with more detail some very sensitive situations such as those encountered in the wildland urban interface (WUI). The main part of this study concerns numerical simulations of the propagation of a fire front through a homogeneous vegetation layer (a grassland) in the vicinity of a fuelbreak represented by a band more or less wide inside which all the fuel was removed. The simulations were performed using a fully physical wildfire model (FIRESTAR3D), three variable parameters were considered in this study: the 1m open wind speed (U1 ranged between 3 and 10 m/s), the fuel height (HFuel ranged between 0.25 and 1m) and the fuelbreak width (LFB). With these conditions, the simulations covered a large range of values of the Byram’s convective number NC (0.3 < NC < 60) in order to explore wind as well driven fires (NC < 2) and plume dominated fires (NC > 10). The 72 simulations carried out in this study have been classified in three categories: 1/ Propagation (if the fire has crossed the fuelbreak with a propagation after); 2/ Overshooting or Marginal (if the fire has crossed the fuelbreak without a propagation after); 3/ No-propagation (if the fuelbreak has stopped the fire). The main objective of this study was to determine the optimal fuelbreak width LFBx separating between the Propagation and the No-propagation regimes, in order to generalize the conclusion, the results have been presented in dimensionless form (similitude theory) in representing as an example the ratio LFBx/HFuel versus the Byram’s convective number NC.

This article presents a first attempt of implementation of a lab scale methodology proposed to describe the thermal decomposition behavior of vegetative fuels, by using thermogravimetric (TG) methods, with respect to forest fire. Experiments within this method include testing the usual grinded form of samples compared to a continues fuel layer (intact form), which can render reliable and useful information to improve the understanding of the ignitability and combustibility of forest fuels along with predicting fire propagation. Moreover, slow (20 and 40 °C/min) and fast (60, 80 and 100 °C/min) heating rate were chosen to get close from the actual conditions of wildland fires (preheating/smoldering and flame region, respectively). In this work, we report the thermal decomposition behavior under inert (N2) and oxidative (air) atmospheres of Genista Salzmannii Needles (GSN) involved in fires of Mediterranean forests. TG analyses were performed to assess thermal reactivity and combustion indices (ignition, devolatilization, combustion and burnout indices) of grinded and cut GSN samples at fives heating rates of 20, 40, 60, 80 and 100 °C/min. The use of the set of thermal indices practical investigation, ignitability, combustibility and sustainability of forest fuels are important properties to be determined when talking to efficient wildland fire management. Moreover, Activation energy was calculated by means of two iso-conversional methods. The results showed that the cut GSN was characterized by a higher mass loss rate compared to grinded samples. As the heating rate increases, reactivity and combustion indices increase linearly. At low heating rates (20 and 40 °C/min), these parameters are quite similar for both samples, while the gap becomes more and more significant at elevated heating rates (60, 80 and 100 °C/min). High correlation coefficients (R2 > 0.96) were obtained, which indicate a good degree of fitting reliability between combustion characteristics and the tested heating rates. Furthermore, the variation of activation energy (Ea) with the conversion rate has exhibited a quite similar behavior during the whole pyrolysis and combustion process. Char formation, at the end of an intense devolatilization, was the most complex process. The oxidation of the remaining char was characterized by a significant decrease of Ea, which is not considered as a major event in the combustion process. Finally, the obtained data provide systematic knowledge for understanding the thermal decomposition of forest fuels (considered as the first stage in fire process), while the developed methodology can be used for comparison and classification of vegetative fuels related to fire risk potential.

Purpose The purpose of this paper is to discuss how, despite increasing data availability from a wide range of sources unlocks unprecedented opportunities for disaster risk reduction, data interoperability remains a challenge due to a number of barriers. As a first step to enhancing data interoperability for disaster risk reduction is to identify major barriers, this paper presents a case study on data interoperability in disaster risk reduction in Europe, linking current barriers to the regional initiative of the European Science and Technology Advisory Group. Design/methodology/approach In support of Priority 2 (“Strengthening disaster risk governance to manage disaster risk”) of the Sendai Framework and SDG17 (“Partnerships for the goals”), this paper presents a case study on barriers to data interoperability in Europe based on a series of reviews, surveys and interviews with National Sendai Focal Points and stakeholders in science and research, governmental agencies, non-governmental organizations and industry. Findings For a number of European countries, there remains a clear imbalance between long-term disaster risk reduction and short-term preparation and the dominant role of emergency relief, response and recovery, pointing to the potential of investments in ex ante measures with better inclusion and exploitation of data. Originality/value Modern society is facing a digital revolution. As highlighted by the International Council of Science and the Committee on Data for Science and Technology, digital technology offers profound opportunities for science to discover unsuspected patterns and relationships in nature and society, on scales from the molecular to the cosmic, from local health systems to global sustainability. It has created the potential for disciplines of science to synergize into a holistic understanding of the complex challenges currently confronting humanity; the Sustainable Development Goals are a direct reflectance of this. Interdisciplinary is obtained with integration of data across relevant disciplines. However, a barrier to realization and exploitation of this potential arises from the incompatible data standards and nomenclatures used in different disciplines. Although the problem has been addressed by several initiatives, the following challenge still remains: to make online data integration a routine.

International audience; The effectiveness of a fuelbreak, created in a homogeneous grassland on a flat terrain, was studied numerically. The analysis relies on 3D numerical simulations that were performed using a detailed physical-fire-model (FIRESTAR3D) based on a multiphase formulation. To avoid border effects, calculations were carried out by imposing periodic boundary conditions along the two lateral sides of the computational domain, reproducing that way a quasi-infinitely long fire front. A total of 72 simulations were carried out for various wind speeds, fuel heights, and fuelbreak widths, which allowed to cover a large spectrum of fire behaviour, ranging from plume-dominated fires to wind-driven fires. The results were classified in three main categories: 1- “Propagation” if fire crossed the fuelbreak with a continuous fire front, 2- “Overshooting” and “Marginal” if fire marginally crosses the fuelbreak with the formation of burning pockets, and 3- “No propagation” if fire does not cross at all the fuelbreak. The ratio of fuelbreak width to fuel height, marking the “Propagation”/“No propagation” transition, was found to be scaled with Byram's convection number Nc as 75.07 × Nc−0.46. The numerical results were also compared to an operational wildfire engineering tool (DIMZAL) dedicated to fuelbreaks dimensioning.

International audience; This paper reported a high intensity experimental fire conducted during a field-scale experiment on a steep sloped terrain (28°) as part of a winter prescribed burns campaign managed by the local firefighter service in the northwestern region of Corsica. The rate of spread (ROS) of fire, measured using UAV cameras (thermal and visible), was evaluated at 0.45 m/s. The experiment was numerically reproduced using a completely physical 2D model, namely FireStar2D, and the comparison with the experimental measurements mainly concerned the fire ROS and the heat fluxes received by three distant targets placed at the end of the plot. The results analysis shows that the considered fire has a wind-driven regime of propagation with a fire intensity higher than 7 MW/m. The numerical results are in fairly good agreement with the experimental measurements, within 11% difference for the ROS and 5% for the heat fluxes, validating consequently the relevance of the numerical approach to tackle such high-intensity wildfires. Despite the unfavorable wind and humidity conditions for fire propagation (U = 1.67 m/s and RH = 82%), this experiment confirms that such fire can exhibit a dangerous behavior due to the steep slope of the terrain.

International audience; The safety during prescribed burnings could be achieved by conducting these operations under marginal conditions of fire propagation. This type of fire can or cannot propagate on account of small deviations of the burning conditions, mainly the wind speed, the fuel load, and the fuel moisture-content. In this context, numerical simulations of grassland fires were conducted under marginal conditions in order to relate the moisturecontent threshold of propagation success to the wind speed and the fuel load. The simulations were conducted using FireStar2D, a complete physical 2D fire simulator based on a multiphase modelling approach. The 10 mopen wind speed ranged from 0 to 10 m/s and the fuel load varied from 0.1 kg/m 2 to 0.7 kg/m 2. The effects of wind speed and fuel moisture-content on the fire behaviour and on the flame parameters are discussed. The results show that the moisture threshold increases with the fuel load until it reaches a value beyond which there is no dependence. A similar dependence of the moisture threshold on the wind speed is also observed. Finally, empirical formulae were constructed to relate the fuel moisture content threshold to the wind speed and the fuel loading implicitly through Byram's convective number.

International audience; The 'Balbi model' is a simplified steady-state physical propagation model for surface fires that considers radiative heat transfer from the surface area of burning fuel particles as well as from the flame body. In this work, a completely new version of this propagation model for wildand fires is proposed. Even if, in the present work, this model is confined to laboratory experiments, its purpose is to be used at a larger scale in the field under operational conditions. This model was constructed from a radiative propagation model with the addition of a convective heat transfer term resulting from the impingement of packets of hot reacting gases on unburnt fuel elements located at the base of the flame. The flame inside the fuel bed is seen as the 'fingers of fire' described in the literature. The proposed model is physics-based, faster than real time and fully predictive, which means that model parameters do not change from one experiment to another. The predicted rate of spread is applied to a large set of laboratory experiments (through homogeneous pine needles and excelsior fuel beds) and is compared with the predictions of both a very simple empirical model (Catchpole) and a detailed physical model (FireStar2D).

Disaster risk reduction (DRR) involves complex processes with different stakeholders at all administrative levels. Two elements in particular play a major role: the need to ensure stakeholders interoperability through an efficient exchange of data, and the inclusion of socioeconomic factors which may influence disaster risk reduction processes. This report aims at contributing to increase the global knowledge on these two elements, focusing on the related issues affecting disaster risk reduction throughout the European continent. The information contained in this report would support national authorities and DRR stakeholders in the continued implementation of the Sendai Framework, and to address the specificities identified by the EFDRR Road Map.

International audience; The scope of this work is to present a multidisciplinary study in order to propose a tool called DIMZAL. DIMZAL forecasts fuelbreak safety zone sizes. To evaluate a safety zone and to prevent injury, the Acceptable Safety Distance (ASD) between the fire and firefighters is required. This distance is usually set thanks to a general rule-of-thumb: it should be at least 4 times the maximum flame length. A common assumption considers an empirical relationship between fireline intensity and flame length. In the current work which follows on from an oral presentation held at the VII International Conference on Forest Fire Research in Coimbra in 2014, an alternative way is proposed: a closed physical model is applied in order to quantize the ASD. This model is integrated in a software tool, which uses a simulation framework based on Discrete EVent system Specification formalism (DEVS), a 3D physical real-time model of surface fires developed at the University of Corsica and a mobile application based on a Google SDK to display the results.

International audience; Several studies in the literature explore the connection between rate of spread (ROS) and wind in wildland fires. These studies show very different positions about the role of radiation and convection as heat transfer mechanisms. In the case when the fuel bed is well-ordered and vertically-oriented, there seems to be a consensus leading to suggest that convective heating is the dominant heat transfer mode in that case. The purpose of this work is to propose a convective semi-physical model for the behaviour of the rate of spread in wind, when the fuel bed is vertically-oriented. Due to a specific fuel bed arrangement, flame radiation -i.e. radiation from the part of the flame above the vegetal stratum is neglected. Only horizontal radiation from the fuel burning particles area and convective heating are taken into account. Convective heat transfer is assumed to be the primary heat transfer mechanism. The proposed model is confronted to 172 laboratory fires with a wide range of fuel characteristics. The predicted results are also compared with two simplified models from the literature. Statistical tools are used to check the agreement between the predicted ROS and the observed one where a strong agreement is generally observed, irrespective of fuel bed characteristics.

In a wildfire, radiative heat transfer is often the main thermal impact on people fighting the fire or on structures. Thus, the estimation of the radiation from the fire front and the heating of a target is of primary importance for forest and urban managers. An analytical formulation of this radiative heat transfer, based on a solid-flame assumption, is used. The realistic description of finite fire-front widths allows the proposal of a new criterion for the estimation of the radiative impact of the fire, which is based on the ratio of the fire-front width to the flame length, which is opposite to the classical approach of considering only the flame length. A numerical solution is necessary to calculate the safety distance for a fixed radiative threshold value, so an analytical approximation is proposed to obtain a simple and useful formulation of this Acceptable Safety Distance. A sensitivity analysis is conducted on the different physical and geometrical parameters used to define the flame front. This analysis shows that the flame temperature is the most sensitive parameter. The results of the analytical model are compared with the numerical solution of the flame model and previous approaches based only on flame length. The results show that the analytical model is a good approximation of the numerical approach and displays realistic estimations of the Acceptable Safety Distance for different fire-front characteristics.

Acoustic scattering by a pair of identical parallel cylinders is studied by emphasizing the role of the symmetries of the scatterer. Incident and scattered fields are expanded over the different irreducible representations ofC2v, the symmetry group of the scatterer. Then, from the boundary conditions, one obtains an infinite set of four linear complex algebraic equations (each one associated with a representation) where the unknown coefficients of the scattered fields are uncoupled. This method significantly simplifies the numerical treatment of the problem. As a consequence, positions of the scatterer resonances are determined in the complex plane of the reduced frequency and a partial algebraic classification of the resonances is obtained for various boundary conditions (soft cylinders, hard cylinders and elastic cylinders immersed in water). A physical interpretation of certain resonances in terms of trapped geometrical paths is provided.

International audience; This paper presents a work conducted in the field of image processing and computer vision in order to measure geometrical characteristics of spreading fires; this allows understanding the phenomenon occurring during the propagation and improving and/or validating mathematical models by comparing experimental data with numerical ones. In order to develop a metrological system based on vision for wildland fires, it is necessary to solve scientific locks that are also presented in this paper.

International audience; This paper proposes a comparison between two simplified flame models. The first flame model uses the radiant surface approach with a new analytical expression for the heat flux. The second one is derived from the Radiative Transfer Equation. The fire front has been considered as a line characterized by some geometric and physical parameters. Two assumptions are used to model the flame either a radiant plane or a volumetric flame. The flame parameters have been identified from experiments using video records and applying an inverse method. These two models were tested against fires carried out in a fire tunnel and found to perform very well considering the complicate nature of the flame geometry and flame characteristics. The need to determine the heat flux from a large scale fire has lead to make a number of assumptions. By means of the proposed modeling, this study tries to determine the extent to which the range of assumptions made disqualifies some simplified flame models from use.

International audience; The physical model we developed in a previous work (Balbi et al., 2007) predicts fire behaviour with a computational time that is faster than real time. However it has the inconvenient of introducing an empirical law (McCaffrey, 1979) which provides flame height according to heat release rate. The present study introduces two improvements of this model: the triangular flame hypothesis and a modification taking into account the air cooling on the rear fire front. To test this variant of our simplified model, it was compared with important experimental results (Viegas, 2004). The experiments were performed in homogeneous and plane fuel beds made with dead Pinus pinaster needles under high values of wind and slope. In spite of these high values this physical model provides a good approximation of the fire front perimeter.

International audience; The scope of this work was the determination of kinetic parameters of the thermal oxidative degradation of a Mediterranean scrub using a hybrid method developed at the laboratory. DSC and TGA were used in this study under air sweeping to record oxidative reactions. Two dominating and overlapped exothermic peaks were recorded in DSC and individualized using an experimental and numerical separation. This first stage allowed obtaining the enthalpy variation of each exothermic phenomenon. In a second time, a model free method was applied on each isolated curve to determine the apparent activation energies. A reactional kinetic scheme was proposed for the global exotherm composed of two independent and consecutive reactions. In fine mean values of enthalpy variation and apparent activation energy previously determined were injected in a model fitting method to obtain the reaction order and the preexponential factor of each oxidative reaction. We plan to use these data in a sub-model to be integrated in a wildland fire spread model.

International audience; This work presents a new modelling approach to the elaboration of a simple model of surface fire spread. This model runs more fastly than real-time and will be integrated in management tools. Up to now, models used in such tools have been based on empirical relationship. These tools may be efficient for conditions that are comparable to those of test-fires but the absence of a physical description makes them inapplicable to other situations. This paper proposes a physical 3D model of surface fire able to predict fire behaviour faster than real-time. This model is tested on experiments carried out across fuel beds under slope and wind conditions.

International audience; A previous termfirenext term spread experiment was conducted in the field under wind-blown conditions. The fuel consists of tall and dense Mediterranean shrub vegetation. The plot area was about 30 m wide and 80 m long. This experiment was conducted not only in order to increase the knowledge and understanding of the previous termfirenext term behaviour in the field but to provide data for the validation of physics based models of previous termfirenext term spread. In particular, the effects of wind on the geometric and thermal properties of the flame front in the field were investigated. The flame temperature along the vertical direction and the radiation emitted ahead of the flame front, were measured. The methodology employed in this experiment and some quantitative measurements of wind velocity and direction, flame geometric properties, are also presented and discussed. The measurements and observations exhibit that the behaviour of the previous termfirenext term and the flame structure character are very different from the one encountered at laboratory scale. These preliminary results show that large scale turbulence influence previous termfirenext term spread and affects the flame shape, temperature and radiation emission.

International audience; The AQUAmark 200® device is used to reduce the incidental by-catch of small cetaceans in fishing gear. Little information is known about the signal features of this system despite the fact that such knowledge is important to evaluate its efficiency. The AQUAmark 200 device emits eight different signals composed of one fundamental and several others harmonics with energy in the 5-160 kHz band. The mathematical equation of the temporal evolution of the frequency of each signal has been determined by the authors, and the resulting function has been used to build the six nonconstant evolutions

Wildfire suppression begins before the start of a fire incidence with the application of many proactive prevention policies. One of the most common method is the reduction of the fuel load accumulation by using controlled fires techniques also called prescribed fires. These fires are often conducted in forests and their intensity must be low to prevent any major damage on trees and on the soil. Thus, they must propagate under marginal conditions. So, a slight modification of the propagation conditions can cause their extinction. Indeed, the probability of sustainability of this type of fire depends on several natural factors (ambient temperature, relative humidity, wind speed, etc.) and also on the fuel properties (water content, shape, size, density, specific heat of combustion…). The aim of this study is to identify the fuel properties allowing fire propagation under these particular conditions. Also, the influence of some fuel properties, and of meteorological conditions (wind speed, ambient temperature, etc.) on the behavior of marginal fires and especially on the fuel moisture content of extinction are investigated.During a first work about the determination of the threshold value of the fuel moisture content involving fire extinction under no wind and no slope conditions, three models were used: a complete physical model, based on a multiphase formulation (“FireStar2D”), an analytical model published in 2014 by Balbi et al., and two experimental models. The first one is based on experimental tests carried out in the UMR SPE 6134 laboratory at the University of Corsica and the second one is an empirical model proposed by Wilson in 1985. Thus, field scale simulations were carried out using FireStar2D, with grassland and shrubs as fuel under conditions of light wind and no slope. Moreover, laboratory scale experimental burnings were conducted with excelsior fuel beds. The different results of the simulations and the burning tests are in agreement with those obtained with the semi-physical model. In addition, the effect of fuel load and the LAI, « Leaf Area Index », on the fuel moisture extinction threshold are studied.In a second study, several field scale numerical simulations are conducted with grassland under non-zero wind conditions in order to relate the fuel moisture content threshold to wind speed and to fuel load. This study also allows us to evaluate the influence of wind speed and fuel moisture content on marginal fire behavior and especially on flame characteristics.The results are encouraging and interesting to operational people working in fire management and firefighting fields, given the importance of marginal fire applications in these two domains., La lutte contre les incendies de forêt commence très en amont des départs de feux par une politique volontariste de prévention du risque. Une des méthodes la plus utilisée est la réduction de la charge du combustible en utilisant la technique des feux contrôlés appelés aussi feux dirigés. Ces feux sont souvent menés en forêt et doivent être à faible intensité pour ne pas causer des dégâts importants sur les arbres et sur le sol. Ainsi, ils sont toujours conduits avec des conditions de propagation marginales. Ainsi une légère modification de ces conditions peut provoquer leur extinction. En effet, la probabilité de succès de ce type de feux dépend de plusieurs facteurs naturels (température ambiante, humidité relative, vitesse de vent…) et également des propriétés du couvert végétal (teneur en eau, forme, taille, densité, …). Le but de cette étude est d’identifier les principales propriétés du combustible permettant une propagation du feu dans ces conditions particulières, ainsi que l’influence de certaines propriétés du combustible et des conditions météorologiques (vitesse de vent, température ambiante, …) sur le comportement des feux marginaux. La valeur seuil de la teneur en eau provoquant une extinction sera plus particulièrement investiguée.Dans un premier travail portant sur la détermination de la teneur en eau d’extinction sous conditions de vent et pente nuls, trois modèles ont été utilisés : un modèle physique complet, basé sur une approche multiphasique (« FireStar2D »), un modèle analytique publié en 2014 par Balbi et al., et deux modèles expérimentaux, dont le premier est basé sur des séries de tests de brûlages réalisés au sein du laboratoire UMR CNRS SPE 6134 de l’université de Corse et dont le deuxième, purement empirique, a été proposé par Wilson en 1985. Ainsi, des simulations ont été réalisées avec FireStar2D à l’échelle du terrain, pour l’herbe et le maquis sous des conditions de vent faible et de pente nulle. En outre, des tests de brûlages sont menés à partir de la frisure de bois à l’échelle du laboratoire. Les différents résultats des simulations et des tests expérimentaux sont en adéquation avec ceux obtenus avec le modèle semi-physique. De plus, cette étude a permis de mettre en évidence l’effet de la charge surfacique et du LAI, « Leaf Area Index », sur le seuil d’extinction relié à la teneur en eau de la végétation.La deuxième partie de recherche a porté sur la détermination de la teneur en eau d’extinction sous condition d’un vent modéré ou fort. Des simulations numériques sont menées à l’échelle du terrain avec de l’herbe sous condition de vent afin de relier la teneur en eau d’extinction à la vitesse du vent et à la charge surfacique du combustible. Cette étude a permis également d’évaluer l’influence du vent et de la teneur en eau sur le comportement d’un feu marginal et plus particulièrement sur les caractéristiques de la flamme.Les résultats de ce travail de recherche sont encourageants et très intéressants pour les opérationnels qui travaillent dans les domaines de la gestion de terrain et de la lutte contre les incendies, vu l’importance des applications des feux marginaux dans ces deux domaines.

Modeling the spread of forest fires is a complex affair which can be examined in different ways (physical, semi-physical, empirical). Numerous simulators exist, each one bearing its own particularities, benefits and weak points. The work which has led to the writing of this thesis aims at improving the simplified physical model of fire propagation (Balbi et al, 2007), as well as at laying down the first foundation stones of what will favor an evolution towards a fire behavioral pattern. This study is based upon three main themes :• Modeling the spread of a surface fire with the main engine the radiative heat transfer. This modeling leads to achieve improvements in the formulation of its speed on flat ground, without wind. It finally led to the characterization of two criteria of extinction with characterization of critical values of two parameters (leaf area index and the water content of the fuel).• Quantification of the energy radiating from the flame front. This quantification was performed assuming the solid flame model. This approach has led to obtaining an analytical relationship for the evaluation of Admissible Safety Distance (ASD) in a very short time.• Sensitivity analysis of the different results attained in order to confirm them as a whole and identify the relative influence of various parameters., La modélisation de la propagation des feux de forêt est une matière complexe pouvant être abordée de différentes manières (approche physique, semi-physique, empirique). Nombre de simulateurs existent avec chacun leurs particularités, leurs avantages et leurs points faibles. Les travaux qui ont conduits à la rédaction de cette thèse visent à l’amélioration du modèle physique simplifié de propagation de Balbi et al, 2007 ainsi qu’à la pose des premières pierres visant à le faire évoluer vers un modèle de comportement du feu. Ils se sont articulés autour de 3 grands axes :• La modélisation de la propagation d’un feu de surface avec pour principal moteur le transfert de chaleur radiatif. Cette modélisation à conduit à réaliser des améliorations concernant la formulation de la vitesse à plat sans vent. Elle a enfin conduit à caractériser de deux critères d’extinction portant sur des valeurs critique de deux paramètres du combustible (l’indice de surface foliaire et de la teneur en eau du combustible).• La quantification de l’énergie rayonnée par le front de flamme. Cette quantification, a été réalisée en prenant pour hypothèse le modèle de flamme solide. Cette approche a conduit à l’obtention d’une relation analytique permettant l’évaluation des Distances de Sécurité Admissibles (DSA) dans un temps très restreint.• L’étude de sensibilité des différents résultats obtenus afin de valider l’ensemble de ceux-ci et identifier l’influence relative des différents paramètres