Search

Background Identifying if and how climatic and non-climatic factors drive local changes in fire regimes is, as in many other human-dominated landscapes, challenging in south-eastern France where both heterogeneous spatial patterns and complex fire trends are observed. Aim We sought to identify the factors driving the spatial-temporal patterns of fire activity in southeastern France. Methods We incorporated several non-climatic variables into the probabilistic Firelihood model of fire activity and implemented an enhanced spatio-temporal component to quantitatively assess remaining unexplained variations in fire activity. Key results Several non-climatic drivers (i.e. orography, land cover and human activities) contributed as much as fire-weather to the distribution of fire occurrence (>1 ha) but less to larger fires (>10, 100 and 1000 ha). Over the past decades, increased fire-weather induced a strong increase in wildfire probabilities, which was actually observed on the western part of the region but not so in the east and Corsican Island, most likely due to reinforced suppression policies. Conclusions While spatial patterns in fire activity are driven by land-use and land-cover factors, temporal patterns were mostly driven by changes in fire-weather and unexplained effects potentially related to suppression policies but with large differences between regions.

Background An increase in fire weather is expected in a warming climate, but its translation to fire activity (fire numbers and sizes) remains largely unknown. Additionally, disentangling the extent to which geographic and seasonal extensions as well as intensification contribute to future fire activity remain largely unknown. Aims We aimed to assess the impact of future climate change on fire activity in southeastern France and estimate changes in spatial and seasonal distributions. Methods We projected future fire activities using a Bayesian modelling framework combined with ensemble climate simulations. Changes in numbers of escaped fires (>1 ha), large fires (>100 ha) and burned area were studied for different emission scenarios or degrees of global warming. Key results Fire activity could increase by up to +180% for +4°C of global warming, with large expansions of fire-prone regions and long seasonal lengthenings. Overall, changes will be dominated by intensification within the historical fire niche, representing two-thirds of additional future fire activity, half of this occurring during the high fire season. Conclusions This study confirms that major changes in fire niches would be expected in Euro-Mediterranean regions. Implications Long-term strategic policies for adapting prevention and suppression resources and ecosystems are needed to account for such changes.

Vapor Pressure Deficit (VPD, atmospheric drought) and soil water potential (Ψsoil, soil drought) have both been reported to affect terrestrial plant water stress, plant functions (growth, stomatal conductance, transpiration) and vulnerability to ecosystem disturbances (mortality or vulnerability to wildfires). Which of atmospheric drought or soil drought has the greatest influence on these responses is yet an unresolved question. Using a state-of-the-art soil-plant-atmosphere hydraulic model, we conducted an in-silico experiment where VPD and Ψsoil were manipulated one at a time to quantify the relative importance of atmospheric vs soil drought on most critical plant functions. The model simulates the combined effects of soil drought and atmospheric drought on plant water potential (ΨPlant), a physiologically meaningful metric of plant water status driving plant turgor, stomatal conductance, hydraulic conductance or water content, and thus mortality and fire risks. Contrary to expectations, we showed that VPD had a weaker effect than Ψsoil on tree water stress and forest disturbances risk (i.e leaf moisture content). While physiological responses associated with low water stress such as stomatal closure or turgor loss could be driven by both VPD or soil drought, consequences of extreme water stress such as hydraulic failure, leaf desiccation and vulnerability to wildfires were almost exclusively driven by low Ψsoil. Our results therefore suggest that most plant functions are affected by VPD through its cumulative effect on Ψsoil via increased plant transpiration, rather than through a direct instantaneous effect on plant water potential. We argue that plant hydraulics provide a strong foundation for predicting tree and terrestrial ecosystem responses to climate changes and propose a list of explanations and testable hypotheses to reconcile plant hydraulic theory and observations of soil and atmospheric drought effects on plant functions.

Fuel moisture content (FMC) is a crucial driver of forest fires in many regions world-wide. Yet, the dynamics of FMC in forest canopies as well as their physiological and environmental determinants remain poorly understood, especially under extreme drought.We embedded a FMC module in the trait-based, plant-hydraulic SurEau-Ecos model to provide innovative process-based predictions of leaf live fuel moisture content (LFMC) and canopy fuel moisture content (CFMC) based on leaf water potential (psi Leaf$$ {\psi}_{\mathrm{Leaf}} $$). SurEau-Ecos-FMC relies on pressure-volume (p-v) curves to simulate LFMC and vulnerability curves to cavitation to simulate foliage mortality.SurEau-Ecos-FMC accurately reproduced psi Leaf$$ {\psi}_{\mathrm{Leaf}} $$ and LFMC dynamics as well as the occurrence of foliage mortality in a Mediterranean Quercus ilex forest. Several traits related to water use (leaf area index, available soil water, and transpiration regulation), vulnerability to cavitation, and p-v curves (full turgor osmotic potential) had the greatest influence on LFMC and CFMC dynamics. As the climate gets drier, our results showed that drought-induced foliage mortality is expected to increase, thereby significantly decreasing CFMC.Our results represent an important advance in our capacity to understand and predict the sensitivity of forests to wildfires.

Accurate spatiotemporal modeling of conditions leading to moderate and large wildfires provides better understanding of mechanisms driving fire-prone ecosystems and improves risk management. We here develop a joint model for the occurrence intensity and the wildfire size distribution by combining extreme-value theory and point processes within a novel Bayesian hierarchical model, and use it to study daily summer wildfire data for the French Mediterranean basin during 1995--2018. The occurrence component models wildfire ignitions as a spatiotemporal log-Gaussian Cox process. Burnt areas are numerical marks attached to points and are considered as extreme if they exceed a high threshold. The size component is a two-component mixture varying in space and time that jointly models moderate and extreme fires. We capture non-linear influence of covariates (Fire Weather Index, forest cover) through component-specific smooth functions, which may vary with season. We propose estimating shared random effects between model components to reveal and interpret common drivers of different aspects of wildfire activity. This leads to increased parsimony and reduced estimation uncertainty with better predictions. Specific stratified subsampling of zero counts is implemented to cope with large observation vectors. We compare and validate models through predictive scores and visual diagnostics. Our methodology provides a holistic approach to explaining and predicting the drivers of wildfire activity and associated uncertainties.

Many decisions in the field of agriculture, forestry and/or hydrology can get profit from seasonal forecast. However, the skill of such forecast is a critical issue to promote their use in operational context and get profitable decisions. If many methods to assess meteorological forecast performances are available, they are mostly implemented on raw climate variables, while their implementation in sectorial application remains limited to some case studies. In this study a wide range of indicators covering most of the decision-making needs in agriculture, forestry and in some extent to hydrology were considered. These indicators are either direct climate variables, a combination of climate variables, or variables calculated by dynamic models (e.g. a crop model). The study was implemented in southern France using the Méteo-France system 6 1993-2016 hindcast, downscaled using the UERRA reanalysis and the ADAMONT methods available in CS-Tools R package developed in the frame of the MEDSCOPE project. These computed indicators need various climate variables as wind speed, radiation and air humidity while most of the downscaling methods were designed for air temperature and precipitation. The main results are the following.We showed that all variables led to comparable level of accuracy. Seasonal forecasts provide added value compared to climatological forecasts with Brier Skill Scores between 0.05 and 0.20. The predictability of the number of rainy days or the number of days with temperature above a threshold is comparable to those of the corresponding scalar quantities such as cumulative precipitation or mean air temperature. However seasonal forecast of extreme events such as heat waves or drought episodes was not possible. Indicators combining several climatic information such as potential evapotranspiration or fire weather index have comparable predictability than the individual climate variables used in the calculation. With indicators based on dynamic models, the memory effect, i.e. the effect of the system state at the beginning of the forecast period, has a strong impact on the skill scores. We propose a methodology based on an ANOVA to qualify this memory effect by using the F-value. It is shown that when the memory effect is strong (F-value >10) the seasonal forecast does not bring any added value compared to the climatological forecast. An evaluation of the interest of a seasonal forecast in a decision-making framework was carried out by an economic approach. We have based our analysis on the decision making based on the forecast of an event. We show that there is a generic relationship between the AUC score and the gain from the forecast. We show that this relationship depends on the frequency of the decision event, the rarer the event the higher the AUC value must be to have a profitable decision. In our case, a decision based on the detection of a tercile leads to a profitable decision in more than half of the indicators while no indicator leads to a profitable decision when it is based on the detection of a quintile.

A widespread increase in tree mortality has been observed around the globe, and this trend is likely to continue because of ongoing climate-induced increases in drought frequency and intensity. This raises the need to identify regions and ecosystems that are likely to experience the most frequent and significant damages. We present SurEau-Ecos, a trait-based, plant-hydraulic model designed to predict tree desiccation and mortality at scales from stand to region. SurEau-Ecos draws on the general principles of the process-based, soil-plant-atmosphere SurEau model. It also introduces a simplified representation of plant architecture and alternative numerical schemes; both additions made to facilitate model parameterization and large-scale applications. In SurEau-Ecos, the water fluxes from the soil to the atmosphere are represented through two plant organs (a leaf and a stem, which includes the volume of the trunk, roots and branches) as the product of an interface conductance and the difference of water potentials. Each organ is described by its symplasmic and apoplasmic compartments. The dynamics of plant’s water status beyond the point of stomatal closure are explicitly represented via residual transpiration flow, plant cavitation and solicitation of plants’ water reservoirs. In addition to the "explicit” numerical scheme of SurEau, we implemented a “semi-implicit” and “implicit” scheme. Both schemes led to a substantial gain in computing time compared to the “explicit” scheme (>10,000 times), and the implicit scheme was the most accurate. We also observed similar plant water dynamics between SurEau-Ecos and SurEau but slight disparities in infra-daily variations of plant water potentials that we attributed to the differences in the representation of plant architecture between models. A global model’s sensitivity analysis revealed that factors controlling plant desiccation rates differ depending on whether leaf water potential is below or above the point of stomatal closure. Total available water for the plant, leaf area index, and the leaf water potential at 50 % stomatal closure mostly drove the time needed to reach stomatal closure. Once stomata are closed, resistance to cavitation, residual cuticular transpiration and plant water stocks mostly determined the time to hydraulic failure. Finally, we illustrated the potential of SurEau-Ecos to simulate regional drought-induced mortality over France. SurEau-Ecos is a promising tool to perform regional-scale predictions of drought-induced hydraulic failure, determine the most vulnerable areas and ecosystems to drying conditions, and asses the dynamics of forest flammability.

Une recrudescence des incendies extrêmes est observée sur tous les continents. Trois principaux facteurs aggravants sont discutés. La croissance démographique et l’étalement urbain augmentent l’exposition des biens et des personnes, et multiplient les mises à feu accidentelles. Les changements d’usage des sols favorisent le développement de la biomasse combustible. Enfin, le changement climatique et les sécheresses associées accroissent la combustibilité des végétaux et, de fait, la propagation des feux. Les interactions entre le risque d’incendie et d’autres risques d’origines biotique et abiotique sont ensuite illustrées, notamment l’effet retour des incendies extrêmes sur le climat par la libération massive de gaz à effet de serre. Nous concluons sur des recommandations pour la France en matière de gestion massive du combustible, de planification territoriale et de responsabilisation des citoyens. L’efficacité de ces mesures sera jugée à l’aune des dommages socio-écologiques évités plutôt que des gains en termes de superficie brûlée.

Identifying the drivers of fire activity’s spatio-temporal variability is challenging in densely populated and fire prone landscapes. Human usage and climate affect the local fire regime in contrasting ways. The identification of these drivers is further complicated due to the stochastic nature of fire activity. Fire regimes in Mediterranean France show contrasted spatial patterns and temporal changes at decadal scales. While overall, the number of fires decreased over the last thirty years, certain zones suffered local increases in fire activity. To describe and understand the drivers of those changes and the spatial variability, we introduced several improvements in the Firelihood model - a probabilistic framework capable of prediction fire occurrence of >1ha fires, and exceedance probabilities of 10 and 100 ha thresholds - by incorporating Land-Use Land-Cover (LULC) explanatory variables, as well as by enhancing its spatio-temporal components to account for unexplained variability in models. The novel model - fitted on a 2km-pixel grid, but relying on variables aggregated at various spatial aggregations (2, 4, 8 and 16km) - is used to explain the observed spatial patterns of fire activity during the last 27 years, as well as the regional and local changes observed between two decades with contrasted fire activities by running counterfactual scenarios. LULC variables, including road density, wildland-urban interface, or expert-based fuel type rating explain a significant part (as much as fire-weather) of the variability in fire occurrence (>1ha), thereby reducing the effect of unexplained spatial variability. The selected occurrence model uses only 2km-resolution variables, as local factors have a high influence on fire ignition and initial spread. The occurrence of larger fire (>10 ha or >100 ha) is largely driven by fire-weather, followed by unexplained spatial variability; selected models for larger fires uses a few LULC variables aggregated at 4, 8 and/or 16 km. This indicates the influence of surrounding factors on fire size extension. The spatial effect for fire occurrence presents contrasted hot and cold-spots throughout the area, while it has a clear east to west decreasing trend for fire size. Regarding temporal changes in fire activity between the two decades, changes in fire weather induced a strong increase in fire probability in many hot spots throughout the region, but this effect was overcompensated by a negative trend associated with unexplained temporal factors (and of larger magnitude than fire weather). LULC variables had negligible effect on the fire regime’s temporal trends. Moreover, an east-to-west gradient appears for the spatial trends of the larger fires, and for the temporal trends in all sizes, highlighting the increase in fire activity in the western side of the region. Those results suggest that observed temporal changes in fire activity are the result of a changing socio-economic or policy frame, probably related to reinforced suppression policies following the year 2003, and the increasing agricultural abandonment.

Understanding and predicting fuel moisture content (FMC) is a crucial prerequisite to increase our knowledge of forest’s vulnerability to fires in a changing climate. While live fuel moisture content (LFMC) is a main driver of fire behavior and activity in crown fires in forests and shrublands, it remains poorly understood and predicted, especially under extreme drought. A major reason for that is that LFMC sensitivity to climate is mediated by a range of location-specific factors, including soil characteristics and plant response to drought. Another reason is that LFMC is often simulated at the leaf scale while, from a fire danger perspective, canopy scale fuel moisture content (CFMC) is more relevant. Here we introduced a FMC module in the plant-hydraulic SurEau-Ecos model to simulate the dynamics of both LFMC and CFMC as a function of leaf water potential (psi_{Leaf}). CFMC integrates the impacts on moisture content of foliage mortality that can occur under extreme drought because of leaf embolism. SurEau-Ecos-FMC relies on two main mechanisms. The relationship between psi_{Leaf} and leaf relative water content of the symplasm is modeled through pressure volume curves. Percent loss of leaf conductance (PLC) is derived from psi_{Leaf} through vulnerability curves to cavitation and affects the dynamics of fuel moisture content in two different ways. At the leaf level, PLC dictates the dynamics of the leaf apoplasmic reservoir. At the canopy level, PLC drives the proportion of dead fuel within the canopy. SurEau-Ecos-FMC captured relatively well the temporal dynamics of LFMC measured in a Quercus ilex stand at the Puéchabon site over a three-years period. The model explained 69 % (RMSE = 4.27) and 74 % (RMSE = 4.92) of the variance in the minimum and maximum daily observed LFMC, respectively. The model was also able to capture the dynamics in CFMC resulting from leaf mortality during the summer drought. Indeed, we showed that the years when SurEau-Ecos-FMC predicted leaf mortality were the ones with the highest anomaly in NDVI. Multi-model projections of fire danger indices based on CFMC showed a general increase of fire danger over the next century. Under RCP8.5, the averaged minimum CFMC reached during the year is expected to decrease from 65 % to 32 % and the fire season length (number of day when CFMC

Anticipating future fire activity at global and regional scales is critical in a changing climate. Indeed, fire seasons are expected to lengthen and fire prone areas are expected to extend, but the magnitude, location and timing of such increases remain uncertain. Moreover, an intensification is expected during the core of the fire season of already fire-prone regions. However, quantifying seasonal and spatial impacts of climate change on fire activity is challenging. Here, we projected future fire activities in Southern France using the Firelihood model. This Bayesian probabilistic model operates on a daily basis in 8-km pixels, allowing to analyze both seasonal and spatial distributions of fire activities in a framework integrating stochasticity. Projections were computed for 13 GCM-RCM couples under two RCP scenarios (4.5 and 8.5), assuming that the only factor of change in future fire activity was the daily fire weather. The fire season was defined as the period with fire-activity level higher than the level of the 15th of July of the present period. The fire prone region corresponded to locations with fire-activity levels higher than the 2nd level of a 5-level fire-activity scale derived from numbers of fires larger than 1ha, 100ha (N1ha and N100ha) and burnt areas (BA). Simulations under RCP8.5 show that large increases in fire activity should be expected from the mid-century and that the rate of increase should then accelerate, leading to up to three-fold increases for number of fires larger than 100ha by the end of the century. In particular, all metrics except N1ha increased faster than the mean FWI and even the mean DSR. Such increases were partly caused by a massive seasonal lengthening from 45-50 days to up to 125 days, equally distributed between spring and autumn. However, the intensification during the present fire season was found to contribute slightly more to the overall increase than the lengthening itself. For example, for N100ha, the intensification would represent a 280 % increase in fire activity with respect to the present seasonal reference, whereas the lengthening outside of the present season would represent +230%. The fire prone area would increase by 168%, shifting from 22 to 56% of region total area. However, the intensification inside the already fire-prone region was found to contribute more to the increase than the spatial extension. For example, for N100ha, the intensification would represent a 190% increase with respect to the present fire-prone regional reference, whereas the extension outside of this area would represent +110%. These drastic increases provide a good indication of the potential lengthening of the fire season, spatial extension and intensification of future fire activities under RCP 8.5, all three being importantly concerned, but dominated by intensification. Extending and lengthening suppression policies may allow to mitigate projected increases, but the intensification of fire activity during the core of the fire season overwhelm current fire suppression capacities.

Data are available in PROMÉTHÉE2: Forest fires database for Mediterranean area in France, and can be extracted at https://www.promethee.com/en/incendies using the following selection criteria: fire type, Forêt; date range, 1 January 1995 to 31 December 2018.; International audience; Modeling wildfire activity is crucial for informing science-based risk management and understanding the spatiotemporal dynamics of fire-prone ecosystems worldwide. Models help disentangle the relative influences of different factors, understand wildfire predictability, and provide insights into specific events. Here, we develop Firelihood, a two-component, Bayesian, hierarchically structured, probabilistic model of daily fire activity, which is modeled as the outcome of a marked point process: individual fires are the points (occurrence component), and fire sizes are the marks (size component). The space-time Poisson model for occurrence is adjusted to gridded fire counts using the integrated nested Laplace approximation (INLA) combined with the stochastic partial differential equation (SPDE) approach. The size model is based on piecewise-estimated Pareto and generalized Pareto distributions, adjusted with INLA. The Fire Weather Index (FWI) and forest area are the main explanatory variables. Temporal and spatial residuals are included to improve the consistency of the relationship between weather and fire occurrence. The posterior distribution of the Bayesian model provided 1,000 replications of fire activity that were compared with observations at various temporal and spatial scales in Mediterranean France. The number of fires larger than 1 ha across the region was coarsely reproduced at the daily scale, and was more accurately predicted on a weekly basis or longer. The regional weekly total number of larger fires (10-100 ha) was predicted as well, but the accuracy degraded with size, as the model uncertainty increased with event rareness. Local predictions of fire numbers or burned areas also required a longer aggregation period to maintain model accuracy. The estimation of fires larger than 1 ha was also consistent with observations during the extreme fire season of the 2003 unprecedented heat wave, but the model systematically underrepresented large fires and burned areas, which suggests that the FWI does not consistently rate the actual danger of large fire occurrence during heat waves. Firelihood enabled a novel analysis of the stochasticity underlying fire hazard, and offers a variety of applications, including fire hazard predictions for management and projections in the context of climate change.

International audience; Highlights:• A new method for LAD estimation from T-LiDAR in occluded areas is developed.• This method use kriging on LAD estimator (LAD-Kriging) and relies on TLS data only.• LAD-Kriging can be applied to all voxels whatever their reliability.• LAD-Kriging retrieves reliable estimates in not-explored and poorly-sampled voxels.• This method can be fitted to other unbiased estimators extracting metrics in point clouds.Abstract:Terrestrial Laser Scanning (TLS) has been used during the past decade to capture the complexity of 3D forest canopy structures, especially Leaf or Plant Area Density (LAD/PAD). TLS data, i.e. point cloud, can be divided into voxels to estimate the three-dimensional distribution of LAD/PAD. However, the combination effects of vegetation occlusion and shooting pattern of TLS scanners lead to a highly heterogeneous sampling, which limits the reliability of some local estimates, since several voxels are either not explored or poorly-explored by laser beams. In practice, recommendations vary regarding the minimum number of beams crossing voxels or the minimum path lengths required to provide reliable predictions. In addition, assigning a value to non-explored and poorly-explored voxels is still an open question. The present work proposes a new method, called LAD-kriging, to mitigate the impact of non-uniform sampling and to increase the accuracy of LAD estimates in non-explored and poorly-explored voxels. The method takes advantage of i) an unbiased LAD estimator of known variance, which was recently developed; ii) the spatial correlation of the LAD field, which derives from vegetation clumping. LAD-kriging computes kriging weights from mathematical derivations, which takes into account both spatial dependencies in the LAD field and the reliability of the estimate available in each voxel. It was evaluated through numerical experiments, which enabled to validate the algorithm and to evaluate its performance through comparison with true references. An example application to field data shows that such spatial correlations truly exist in the field and that LAD-kriging entails to reduce sampling errors (with respect to full resolution scanning). Although a real validation is impossible due to the lack of precise references at the voxel scale, this example allows to gain confidence for its application to field data. In our realistic numerical experiment, up to 25-30% of voxels can be explored by

Water content in living vegetation (or live fuel moisture content, LFMC), is increasingly recognized as a key factor linked to vegetation mortality and wildfire ignition and spread. Most often, empirical indices are used as surrogates for direct LFMC measurements.In this paper, we explore the functional and ecophysiological drivers of LFMC during drought at the leaf and canopy scale using the SurEau-Ecos model, and a three years dataset of leaf and canopy scale measurements on a mature Quercus ilex forest, including an extreme drought. The model is based on forest hydrology and plant hydraulics and allows to simulate temporal variations of water potential and content at a daily time step. At leaf level, it simulates the relationship between water potential and water content by separating the apoplasm and the symplasm. Symplasm water content is modeled using the pressure volume curve theory, and apoplasm water content is modelled using the xylem vulnerability to cavitation. Fuel moisture content was upscaled to the canopy level by accounting for foliage mortality estimated from drought induced cavitation.The model was parameterized either with site-measured traits or using a calibration procedure, and compared with water potential and LFMC measured at leaf level, and NDVI variation measured at canopy level and taken as a surrogate for foliage mortality.At leaf level, LFMC prediction using measured hydraulic traits could be improved by considering year-to-year osmotic adjustments. At canopy level, foliage mortality due to drought induced cavitation was a key driver of LFMC decline during the most extreme drought.A sensitivity analysis showed that parameters driving soil water balance (leaf area index, soil water capacity, and regulation of transpiration) and parameters determining pressure volume curves are key traits driving LFMC dynamics at leaf level. At the canopy level, parameters that drives hydraulic failure were the most sensitive and included, both soil water balance parameters and hydraulic traits (the leaf vulnerability to cavitation) were the main drivers of LFMC decline during extreme drought.We also showed that under normal historic weather conditions, most variation of LFMC are linked to reversible symplasm dehydration, however under future, hotter and dryer conditions, most variations are due to the decline canopy of LFMC driven by foliage mortality.

Modelling wildfire activity is crucial for informing science-based risk management and understanding fire-prone ecosystem functioning worldwide. Models also help to disentangle the relative roles of different factors, to understand wildfire predictability or to provide insights into specific events.Here, we develop a two-component Bayesian hierarchically-structured probabilistic model of daily fire activity, which are modelled as the outcome of a marked point process in which individual fires are the points (occurrence component) and the fire sizes are the marks (size component). The space-time Poisson model for occurrence is adjusted to gridded fire counts using the integrated nested Laplace approximation (INLA) combined with the Stochastic Partial Differential Equation (SPDE) approach. The size model is based on piecewise-estimated Pareto and Generalized-Pareto distributions, also adjusted with INLA. The Fire Weather Index (FWI) and Forest Area are the main explanatory variables. Seasonal and spatial residuals as well as a post-2003 effect are included to improve the consistency of the relationship between climate and fire occurrence, in accordance with parsimonious criteria.A set of 1000 simulations of the posterior model of fire activity is evaluated at various temporal and spatial scales in Mediterranean France. The number of escaped fires (≥1ha) across the region can be coarsely reproduced at the daily scale, and is more accurately predicted on a weekly basis or longer. The regional weekly total number of larger fires (10 to 100 ha) can be predicted as well, but the accuracy decays with size, as the model uncertainty increases with event rareness. Local predictions of fire numbers or burnt areas likewise require a longer aggregation period to maintain model accuracy.Regarding the year 2003 -which was characterized by an extreme burnt area in France associated with a heat wave-, the estimation of the number of escaped fires was consistent with observations, but the model systematically underrepresents larger fires and burnt areas, which suggests that the FWI does not consistently rate the danger of large fire occurrence during heat waves.Our study sheds new light on the stochastic processes underlying fire hazard, and is promising for predicting and projecting future fire hazard in the context of climate change.

Wildfire is the main disturbance in forested ecosystems of southern Europe and is due to complex interactions between climate-weather, fuels and people. Warmer and drier conditions projected in this region are expected to profoundly affect wildfires, which will impact ecosystems and humans. We review the scientific literature addressing the assessment of climate change impacts on wildfires in southern Europe, with a twofold objective: (i) report the trends in wildfire danger and activity projected under warming climate in southern Europe and (ii) discuss the limitations of wildfire projections under the specific biogeographical context of southern Europe.We identified 22 projection studies that examined future wildfire danger or wildfire activity at local, regional or continental scale. Under the scenario with the highest greenhouse gas emissions, we found that projections studies estimate an increase in future fire danger and burnt areas varying, on average, from 2 to 4 % and from 15 to 25 % per decade, respectively. Fire-prone area expansion to the north and to Mediterranean mountains is a concern, while climate-induced burnt area increase might be limited by fuel availability in the most arid areas. While all studies agreed on the direction of changes, further comparisons on the magnitude of increase remained challenging because of heterogeneous methodological choices between projections studies (climate models, projection period, spatial scale and fire metrics). We then described three main sources of uncertainty that may affect the reliability of wildfire projections: climate projections, climate-fire models, and the influences of fuel load/structure and human related factors on the climate-fire relationships. We finally suggest research directions to address some of these issues for the purpose of refining fire danger and fire activity projections in southern Europe.

International audience; Highlights:• Stand-scale estimates of leaf density from voxelized LiDAR data can be biased.• We identified a negative bias caused by the oversampling of light voxels.• Its magnitude depends on location, vegetation structure, scan design and voxel size• Voxel-scale measurement accuracy is too low in small voxels.• 0.5 m voxel size is the best option because corrections are more straightforward.Abstract:The need for fine scale description of vegetation structure is increasing as Leaf Area Density (LAD, m(2)/m(3)) becomes a critical parameter to understand ecosystem functioning and energy and mass fluxes in heterogeneous ecosystems. Terrestrial Laser Scanning (TLS) has shown great potential for retrieving the foliage area at stand, plant or voxel scales. Several sources of measurement errors have been identified and corrected over the past years. However, measurements remain sensitive to several factors, including, 1) voxel size and vegetation structure within voxels, 2) heterogeneity in sampling from TLS instrument (occlusion and shooting pattern), the consequences of which have been seldom analyzed at the scale of forest plots. In the present paper, we aimed at disentangling biases and errors in plot-scale measurements of LAD with TLS in a simulated vegetation scene. Two negative biases were formerly attributed to (i) the unsampled voxels and to (ii) the subgrid vegetation heterogeneity (i.e. clumping effect), and then quantified, thanks to a the simulation experiment providing known LAD references at voxel scale, vegetation manipulations and unbiased point estimators. We used confidence intervals to evaluate voxel-scale measurement accuracy. We found that the unsampled voxel effect (i) led to underestimations with the "mean layer" method -commonly used to fill unsampled voxels- for small voxels (0.1-0.2 m) and/or low number of scans (

Wildfire is the main disturbance in forested ecosystems of southern Europe and is due to complex interactions between climate-weather, fuels and people. Warmer and drier conditions projected in this region are expected to profoundly affect wildfires, which will impact ecosystems and humans. We review the scientific literature addressing the assessment of climate change impacts on wildfires in southern Europe, with a twofold objective: (i) report the trends in wildfire danger and activity projected under warming climate in southern Europe and (ii) discuss the limitations of wildfire projections under the specific biogeographical context of southern Europe.We identified 22 projection studies that examined future wildfire danger or wildfire activity at local, regional or continental scale. Under the scenario with the highest greenhouse gas emissions, we found that projections studies estimate an increase in future fire danger and burnt areas varying, on average, from 2 to 4 % and from 15 to 25 % per decade, respectively. Fire-prone area expansion to the north and to Mediterranean mountains is a concern, while climate-induced burnt area increase might be limited by fuel availability in the most arid areas. While all studies agreed on the direction of changes, further comparisons on the magnitude of increase remained challenging because of heterogeneous methodological choices between projections studies (climate models, projection period, spatial scale and fire metrics). We then described three main sources of uncertainty that may affect the reliability of wildfire projections: climate projections, climate-fire models, and the influences of fuel load/structure and human related factors on the climate-fire relationships. We finally suggest research directions to address some of these issues for the purpose of refining fire danger and fire activity projections in southern Europe.

According to climate projections, global warming is associated with increasing temperatures and dry spells in some parts of the world, especially the Mediterranean area. However, it is quite challenging for the scientific community to assess the intensity of these changes, because (i) the trend relies on the greenhouse gases (GHG) emission scenario and (ii) fire occurrence depends on multiple factors (including climate, but not only). A proper assessment of the trend in terms of fire activity still lacks, especially for the French territory. Our study refines traditional approaches of fire risk projection under climate change by translating a danger index into potential fire activity (fire occurrence above 1 ha and burnt area).

International audience; According to climate projections, global warming is associated with increasing temperatures and dry spells in some parts of the world, especially the Mediterranean area. This climate change has already triggered increases in wildfire danger and fire season length in Southern Europe and is expected to amplify in the forthcoming decades. However, it is quite challenging for the scientific community to assess the intensity of these changes, because (i) the trend relies on the greenhouse gases (GHG) emission scenario and (ii) fire occurrence depends on multiple factors (including climate, but not only). A proper assessment of the trend in terms of fire occurrence and of uncertainties associated with this increasing trend, still lacks, especially for the French territory. Our study refines traditional approaches of fire risk projection under climate change on two aspects: (i) the impact of climate prediction uncertainties on the prediction of fire danger, and (ii) the translation of a danger index into a fire occurrence (per size classes).

International audience; Live Fuel Moisture Content (LFMC) is a critical variable affecting fire ignition, behavior and severity in many ecosystems. Although the use of meteorological drought indices as proxies for LFMC is a straightforward and widespread approach, it is largely unknown whether it can provide reliable estimates of LFMC, either for local or spatial applications. We address this issue by evaluating the capacity of drought indices to predict LFMC quantitative variations and critical values. LFMC observations used for reference were measured on six different Mediterranean shrub species for 15 years in 20 different sites in Southern France. Six drought indices were evaluated: the Duff Moisture Code (DMC) and Drought Code (DC) of the Canadian Forest Fire Weather Index System, the Keetch-Byram Drought Index (KBDI), the Nesterov Index (NI) and the Relative Water Content (RWC) of the soil derived from a forest water balance model for low (80 mm) and high (160 mm) field capacities. The species were classified in two groups according to their seasonal variability: high and low responding species. We found large differences in the capacity of drought indices to predict LFMC, with indices that simulate long-term drought dynamics (DC, RWC and KBDI) generally performing better than others (NI and DMC). Once calibrated at stand scale, drought indices showed a good potential for predicting LFMC of high responding species, although large variations between sites were observed. In contrast, spatial predictability was limited with a RMSE and R2 on the order of 20% and 0.3, respectively (for high responding species). Our results suggest that drought indices should therefore be used with caution for spatial applications in wildfire research and operational fire management. Because they can explicitly consider environmental (soil, climate) and biological (species traits related to dehydration) factors, mechanistic indices have a great potential to improve LFMC predictions.

Summarization: Droughts can have strong environmental and socio-economic impacts in the Mediterranean region, in particular for countries relying on rain-fed agricultural production, but also in areas in which irrigation plays an important role and in which natural vegetation has been modified or is subject to water stress. The purpose of this review is to provide an assessment of the complexity of the drought phenomenon in the Mediterranean region and present various perspectives on drought in the present and under future climate change scenarios. The projections of various model experiments on future climate change scenarios strongly agree on an increased frequency and severity of droughts in the Mediterranean basin. Nevertheless, given the complexity of the phenomenon, with different types of droughts and complex interrelated impacts, significant future uncertainties remain. For example, uncertainties are stronger for hydrological droughts than meteorological droughts due to human influences and water withdrawal. Significant drought impacts are expected in the future, in particular for developing countries in the southern and eastern parts of the Mediterranean basin. To improve the resilience and adaptive capacities of societies and environments faced with drought, we aim to provide an overview of the key issues in research on climate change impacts on droughts, with a specific focus on the Mediterranean region, in order to: i) redefine more meaningful drought metrics tailored to the Mediterranean context, ii) better take into account vegetation and its feedback on droughts, iii) improve the modelling and forecasting of drought events through remote sensing and land surface models, and iv) promote a more integrated vision of droughts taking into account both water availability and water use. This overview reflects the complexity of the problem and the need to combine scientific research with adaptation solutions to deal with drought in the future. Presented on

Scientists and managers critically need ways to assess how fuel treatments alter fire behavior, yet few tools currently exist for this purpose. We present a spatially-explicit-fuel-modeling system, FuelManager, which models fuels, vegetation growth, fire behavior (using a physics-based model, FIRETEC), and fire effects. FuelManager's flexible approach facilitates modeling fuels across a wide range of detail. Large trees or shrubs with specific coordinates are modeled as individual "Plants", while understory plants are modeled as collections of plants called "LayerSets". Both Plants and LayerSets contain various fuel particles (leaves, needles, twigs) with various properties including shape, size and surface area to volume ratio. A wide range of vegetation and treatments can be modeled, analyzed quantitatively and visualized in a 3D viewer. We describe the modeling approach and demonstrate fuel modeling at different levels of detail, fuel treatment and fire effects capabilities. Detailed model equations are provided in the Appendices. Display Omitted Comprehensive description of fuel modeling for a physics-based fire model.Different fuel items can be defined to model trees and plant collections.Fuel treatment, growth and fire effects can be simulated.Several applications of the model are demonstrated.

International audience; The complex interactions between the inclined terrain and the flow generated by the fire make the slope one of the most influencing factors on fire spread. In order to gain a deeper understanding of the mechanisms involved in wildfires spreading upslope, the investigation of flow dynamics and heat transfers is fundamental. This paper reports a series of fire spread experiments conducted across a porous bed of excelsior in a large-scale facility, under both no-slope and 30° up-slope conditions. The coupling of particle image velocimetry and video imaging allowed characterizing the flow pattern with respect to the fire front. Simultaneous heat flux measurements with high scan rate were also performed at the edge of the fuel bed. From the collected data, the increase of the rate of spread with increasing slope is attributed to a major change in fluid dynamics surrounding the flame. For horizontal fire spread, flame fronts exhibit quasi-vertical plume resulting from the buoyancy forces generated by the fire. These buoyancy effects induce an inward flow of ambient air that is entrained laterally into the fire from both sides. Flame radiation is the dominant fuel preheating mechanism. Under upslope conditions, the fire plume is tilted toward the unburnt vegetation, increasing radiation levels. The air entrainment at the burnt side of the fire strongly influences the downstream flow, which becomes attached to the surface over a characteristic length scale. Ahead of the flame front, the induced wind blows away from the fire rather than toward it, enhancing convective heating. Periodical forward bursts of flame combined with distant fuel ignitions were also observed. The heat flux measurements confirmed the existence of such convective mechanisms.

International audience; Key message We model the dynamics of fuel characteristics in shrub strata dominated by Quercus coccifera L. with data gathered in available literature. The model expresses the variability of this important fire-prone fuel type thanks to yield classes, and it can be used to investigate management scenarios. The approach could easily be applied to other shrub communities. Context Characterizing fuel is a basic requirement for fire hazard assessment. Quercus coccifera L. is present in several Mediterranean fire-prone communities, and its fuel characteristics have been studied over various Mediterranean countries, but no general model describes its dynamics. Aims Herein, we present such a general model, initially developed for operational purposes at the French Forest Service. Methods We review available literature and fit statistical relationships to predict the dynamics of fuel height and biomass, by size categories of fine fuel elements. Results The model estimates fuel characteristics from shrub-strata age, overstorey cover, and yield class with a reasonable degree of accuracy considering the heterogeneity of the datasets. It shows that bulk density is highly sensitive to overstorey, and in a lesser extent to strata age, which could lead to significant bias when assessing fuel properties from general allometries. The model is integrated in the FuelManager software, which is devoted to fuel modeling for physics-based-fire-behavior models. Conclusion This simple approach enables to provide a fuel model for the Quercus coccifera L. shrub strata in the Mediterranean basin. It is more general than the existing relationships available for local data. This approach could be generalized to other fire-prone communities.

International audience; Correction to: Annals of Forest Science https://doi.org/10.1007/s13595-017-0686-2 In Table 2 in the original article, torching index (TI) and crowning index (CI) values were incorrectly expressed in km/s. The correct units are km/h. The original article has been corrected. The online version of the original article can be found at https://doi.org/10.1007/s13595-017-0686-2

Key message We describe a modeling system that enables detailed, 3D fire simulations in forest fuels. Using data from three sites, we analyze thinning fuel treatments on fire behavior and fire effects and compare outputs with a more commonly used model. Context Thinning is considered useful in altering fire behavior, reducing fire severity, and restoring resilient ecosystems. Yet, few tools currently exist that enable detailed analysis of such efforts. Aims The study aims to describe and demonstrate a new modeling system. A second goal is to put its capabilities in context of previous work through comparisons with established models. Methods The modeling system, built in Python and Java, uses data from a widely used forest model to develop spatially explicit fuel inputs to two 3D physics-based fire models. Using forest data from three sites in Montana, USA, we explore effects of thinning on fire behavior and fire effects and compare model outputs. Results The study demonstrates new capabilities in assessing fire behavior and fire effects changes from thinning. While both models showed some increases in fire behavior relating to higher winds within the stand following thinning, results were quite different in terms of tree mortality. These different outcomes illustrate the need for continuing refinement of decision support tools for forest management. Conclusion This system enables researchers and managers to use measured forest fuel data in dynamic, 3D fire simulations, improving capabilities for quantitative assessment of fuel treatments, and facilitating further refinement in physics-based fire modeling.

International audience; This study examines the spread of a natural fire over an inclinable bench for several vegetative fuel-load and slope conditions and on an intermediate scale (18 m2). A pioneering combination of heat flux measurements and a large-scale particle image velocimetry system is presented. Three different fuel loads and two slopes (0° and 30°) are investigated. The results provide new insights into the fluid mechanics of fire spread, elucidating the air entrained flow motion generated by the fire as a function of the fuel load and slope in particular. For horizontal fires, increases in the vertical velocities and the size of the mixing layer, which is formed by the elevation of hot gases in the atmosphere, are observed with increasing fuel load. The air ahead of the fire is aspired far from the fire, suggesting convective cooling of the unburned vegetation. For 30°-upslope fires, the measurements show a convective heat transfer generated by hot gases advancing towards the unburned vegetation, which is supported by the elongation of the fire head. The experimental technique presented in this study is, therefore, effective for extracting valuable data for fire modeling.

Climate change is often expected to lead to an increase in wildfire danger. The Forest Fire Weather Index (FWI) and its subcomponents, like the Drought Code (DC), are widely used to forecast and project climatic fire danger. However, applications of such wildfire danger indices are subjected to different uncertainties. A first one is related to the ability of these indices to accurately predict empirical data of observed wildfires (“empirical uncertainty”); whereas a second lies in the projection of these indices, which is inherently affected by the spread among climatic input data available from different general and regional circulation models (“projection uncertainty”). It is crucial to take this aspect into account as uncertainties pile up in the modelling chains, what has seldom been done in the context of wildfires. Our study aims at exploring both types of uncertainties, and comparing their respective contributions to the global prediction. First, we analyse uncertainties related to the predictions of an empirical record of past fire observations (number of fire) over French Mediterranean area (Prométhée) using the FWI computed at an 8km resolution (using SAFRAN climatic input) for the historical period 1995-2015. We then investigate the uncertainty linked to climatic projections by projecting FWI and fire occurrence (based on the empirical relation calibrated under historical conditions) under two IPCC scenarios and using two GCM/RCM model combinations for the period 2005-2100. Predictions of fire occurrence under historical conditions reveal important bias both spatially and temporally, highlighting potential uncertainties in the global use of FWI to predict wildfire. FWI and fire occurrence projected under different RCP scenario also reveal an important spread. Ongoing work includes computation with additional models and the assessment of the relative weight of different uncertainty sources in the projection of wildfire risk.

International audience; Estimating leaf and plant area density with Terrestrial Laser Scanners (TLS) continues to be more and more popular, as tridimensional point clouds appear as an appealing measurement technique for heterogeneous environments. Some approaches implement a discretization of the point cloud in a grid (referred to as “voxelbased”) to account for this vegetation heterogeneity and significant work has been done in this recent research field, but no general theoretical analysis is available. Although estimators have been proposed and several causes of biases have been identified, their unbiasedness (zero bias) and efficiency (smallest error) have not been evaluated. Also, confidence intervals are almost never provided. In the present paper, we assumed that the vegetation elements were randomly distributed within voxels and that TLS beams were infinitely thin, in order to focus on the remaining sources of biases and errors. In this simplified context, we both solve the transmittance equation and use the Maximum Likelihood Estimation (MLE), to derive some new estimators for the attenuation coefficient, which is proportional to leaf area density at voxel scale in this idealized context. These estimators include bias corrections and confidence intervals, and account for the number of beams crossing the voxel (beam number), the inequality of path lengths in voxel, the size of vegetation elements, as well as for the variability of element positions between vegetation samples. These theoretical derivations are complemented by numerous numerical simulations for the evaluation of estimator bias and efficiency, as well as the assessment of the coverage probabilities of confidence intervals. Our simulations reveal that the usual estimators are biased and exhibit 95% confidence intervals on the order of±100% of the estimate, when the beam number is smaller than 30. Second, our bias-corrected estimators -especially the bias-corrected MLE- are truly unbiased and efficient in a wider range of validity than the usual ones, even for beam number as low as 5. Third, we found that the confidence intervals can be as high as ≈ ± 50% when the projected area of a single element was on the order of 10% of voxel cross-sectional area and vegetation was dense (optical depth of the voxel equal to 2), even for a beam number larger than 1000. This is explained by the variability of element positions between vegetation samples, which implies that a significant part of residual error is caused by random effects. When LAD estimates are averaged over several small voxels -typically to determine a vertical profile at plot scale or to compute the total leaf area of a single plant-, confidence intervals are typically on the order of±5 to 10% with bias-corrected estimators, which is twice as small as with usual estimators.Our study provides some new ready-to-use estimators and confidence intervals for attenuation coefficients,which are unbiased and efficient within a fairly large range of parameter values. The unbiasedness is achieved for a fairly low beam number, which is promising for application to airborne LiDAR data. They permit to raise the level of understanding and confidence on LAD estimation. Among other applications, their usage should help determine the most suitable voxel size, for given vegetation types and scanning density, whereas existing guidelines are highly variable among studies, probably because of differences in vegetation, scanning design and estimators. The impact of other sources of biases and errors, such as vegetation heterogeneity inside voxels or TLS specifications are not addressed in the present manuscript and would require further investigations.

Experimental fires often aim to relate fire behavior to fuel and weather conditions, such as wind speeds. These experiments are typically limited to short durations (∼300 s) and small lateral extents (∼100 m). Although most studies include measurements of wind velocities, such measurements are often taken at some distance from the fire experiments, and may not represent conditions at the fire location. This disparity may potentially introduce errors of unknown magnitude in empirical models based upon the data collected. At present, little guidance is available regarding how well remote anemometry measurements are actually representative of wind velocities at the fire front. A number of factors may affect this representativeness, including the fire itself (size, spread rate and duration), the reference height for fire wind measurement, the sensors (number and location), the vegetation, and weather conditions (wind speed and atmospheric stability). In the present study, we use large-eddy simulations of wind flows to compute fire-front wind (at a virtual moving fire line) and measured wind (at anemometer locations) corresponding to hypothetical fire experiments. Replicates of these hypothetical experiments were used to quantify wind measurement representativeness, by computing the errors resulting from the estimation of the fire-front wind by remote anemometers. We then examine the sensitivity of these errors to the factors mentioned above. We found that the main factors were the size of the experiment, the reference height for wind measurement, the ratio of ambient wind speed to expected spread rate, and the number of sensors. Convective instability and distance between anemometers and fire plots played a minor role in most cases. We propose a simple model to characterize this error as it is influenced by the main factors. The simple model reproduces and generalizes outcomes reported by an earlier field study and shows a clear picture of the respective role of the factors cited above. It can be used to estimate errors in wind measurement in completed experiments. Practical guidelines are provided to apply this model to the design of future experiments.

Key message Fuel moisture and chemical content affecting live plant flammability can be measured through laboratory and field techniques, or remotely assessed. Standardization of methodologies and a better understanding of plant attributes and phenological status can improve models for fire management. Context Wildland fire management is subject to manifold sources of uncertainty. Beyond the difficulties of predicting accurately the fire behavior, uncertainty stems from incomplete understanding of ecological susceptibility to fire. Aims We aimed at reviewing current knowledge of (i) plant attributes and flammability fuel moisture and chemical content in leaves; (ii) experimental evaluation of flammability in the laboratory and in the field; and (iii) proxy evaluation of flammability vegetation cover assessment at large scale, fuel seasonality, and biomass distribution using remote sensing and Light Detection and Ranging (LiDAR) techniques. Methods We conducted a review of scientific literature from the last two decades on the three selected issues, with a specific focus on the Mediterranean region. Results We have evidenced important knowledge gaps: (i) developing standardized methodologies for laboratory- and field-scale assessment of vegetation flammability; (ii) introducing reliable approaches to test the impact of biogenic volatile organic compounds on fire spread; (iii) improving the analysis of spatiotemporal changes in vegetation dynamics, acknowledging distinctive vegetation phenological status as a relevant driver affecting leaf biomass and moisture contents; and (iv) further exploring the processes that shape fuel dynamics to understand how fuel characteristics change over time and space. Conclusion: We propose some improvements in the current knowledge of vegetation science and wildland fire ecology, aiming to generate more realistic models and effective planning in support of fire management in the Mediterranean basin. � 2017, INRA and Springer-Verlag France.

International audience

National audience

[Departement_IRSTEA]Territoires [TR1_IRSTEA]SEDYVIN; International audience; Past fire recurrence impacts the vegetation structure, and it is consequently hypothesized to alter its future fire behaviour. We examined the fire behaviour in shrubland-forest mosaics of southeastern France, which were organized along a range of fire frequency (0 to 3–4 fires along the past 50 years) and had different time intervals between fires. The mosaic was dominated by Quercus suber L. and Erica–Cistus shrubland communities. We described the vegetation structure through measurements of tree height, base of tree crown or shrub layer, mean diameter, cover, plant water content and bulk density. We used the physical model Firetec to simulate the fire behaviour. Fire intensity, fire spread, plant water content and biomass loss varied significantly according to fire recurrence and vegetation structure, mainly linked to the time since the last fire, then the number of fires. These results confirm that past fire recurrence affects future fire behaviour, with multi-layered vegetation (particularly high shrublands) producing more intense fires, contrary to submature Quercus woodlands that have not burnt since 1959 and that are unlikely to reburn. Further simulations, with more vegetation scenes according to shrub and canopy covers, will complete this study in order to discuss the fire propagation risk in heterogeneous vegetation, particularly in the Mediterranean area, with a view to a local management of these ecosystems.; La combustibilité de 51 placettes correspondant à différentes récurrences d’incendies depuis 1959 (0 à 4 feux) a été analysée en Provence siliceuse. Les écosystèmes étudiés sont des suberaies dominées par Quercus suber L., et des maquis à Erica arborea L. et à cistes. Les données descriptives de la végétation combustible (composition, biomasse, recouvrement, hauteur) ont été entrées dans un modèle physique du feu (Firetec) afin d’obtenir des variables descriptives du comportement de feu. Les pertes en masse et en eau de la végétation, la vitesse de propagation et l’intensité du feu augmentent de façon significative avec la continuité verticale du combustible, qui dépend principalement du temps depuis le dernier feu, puis du nombre de feux. Les résultats confirment l’hypothèse selon laquelle le feu est particulièrement intense et rapide dans des peuplements multi-stratifiés à forte charge en éléments fins et forte connexion spatiale entre les individus, tels que les maquis hauts développés après un seul feu intense en cinquante ans. Au contraire, le feu ne se propage pas dans les forêts de chênes submatures n’ayant pas brûlé depuis au moins cinquante ans et caractérisées par une faible connectivité verticale entre la canopée et le sous-étage, favorisant ainsi un effet « auto-protecteur » contre les incendies futurs. Ces informations permettent de discuter de la gestion de la végétation, en particulier la surveillance ou le débroussaillement local des maquis hauts grandement combustibles, pour diminuer le risque de propagation des feux à l’échelle de la parcelle.