Wildfire risk in the wildland–urban interface: A simulation study in northwestern Wisconsin

The rapid growth of housing in and near the wildland–urban interface (WUI) increases wildfire risk to lives and structures. To reduce fire risk, it is necessary to identify WUI housing areas that are more susceptible to wildfire. This is challenging, because wildfire patterns depend on fire behavior and spread, which in turn depend on ignition locations, weather conditions, the spatial arrangement of fuels, and topography. The goal of our study was to assess wildfire risk to a 60,000 ha WUI area in northwestern Wisconsin while accounting for all of these factors. We conducted 6000 simulations with two dynamic fire models: Fire Area Simulator (FARSITE) and Minimum Travel Time (MTT) in order to map the spatial pattern of burn probabilities. Simulations were run under normal and extreme weather conditions to assess the effect of weather on fire spread, burn probability, and risk to structures. The resulting burn probability maps were intersected with maps of structure locations and land cover types. The simulations revealed clear hotspots of wildfire activity and a large range of wildfire risk to structures in the study area. As expected, the extreme weather conditions yielded higher burn probabilities over the entire landscape, as well as to different land cover classes and individual structures. Moreover, the spatial pattern of risk was significantly different between extreme and normal weather conditions. The results highlight the fact that extreme weather conditions not only produce higher fire risk than normal weather conditions, but also change the fine-scale locations of high risk areas in the landscape, which is of great importance for fire management in WUI areas. In addition, the choice of weather data may limit the potential for comparisons of risk maps for different areas and for extrapolating risk maps to future scenarios where weather conditions are unknown. Our approach to modeling wildfire risk to structures can aid fire risk reduction management activities by identifying areas with elevated wildfire risk and those most vulnerable under extreme weather conditions.     

Forest fuel mapping and evaluation of LANDFIRE fuel maps in Boulder County,Colorado, USA

A key challenge in modern wildfire mitigation and forest management is accurate mapping of forest fuels in order to determine spatial fire hazard, plan mitigation efforts, and manage active fires. This study quantified forest fuels of the montane zone of Boulder County, CO, USA in an effort to aid wildfire mitigation planning and provide a metric by which LANDFIRE national fuel maps may be compared. Using data from 196 randomly stratified field plots, pre-existing vegetation maps, and derived variables, predictive classification and regression tree models were created for four fuel parameters necessary for spatial fire simulation with FARSITE (surface fuel model, canopy bulk density, canopy base height, and stand height). These predictive models accounted for 56–62% of the variability in forest fuels and produced fuel maps that predicted 91.4% and 88.2% of the burned area of two historic fires simulated in the FARSITE model. Simulations of areas burned based on LANDFIRE national fuel maps were less accurate, burning 77.7% and 40.3% of the historic fire areas. Our results indicate that fuel mapping efforts that utilize local area information and biotic as well as abiotic predictors will more accurately simulate fire spread rates and reflect the inherent variability of forested environments than do current LANDFIRE data products.     

Post-fire analysis of pre-fire mapping of fire-risk: A recent case study from Mt. Carmel (Israel)

The recent devastating wildfire on Mt. Carmel provided a unique opportunity to evaluate a fire-risk map constructed for the region, published two years ago in this journal. This largest forest fire in the history of Israel, occurred during December 2010, covering 2180 ha, burning more than half-million trees and causing the loss of life of 45 people.A study of fire risk in this area was conducted between 2007 and 2009 utilizing a combination of Monte Carlo simulation of spatial spread of fire ignition with fire behavior model (FARSITE). The fire risk map produced in 2009 is assessed here with reference to the area burnt during December 2010. The results showed that most of burnt areas corresponded to high risk levels in the risk map. According to a null model, the five lower risk levels taken together would have corresponded to 50% of the burnt area, while in fact they were presented in only 5.6% of the area. In contrast, the three highest risk levels, for which the null model expectation would be a representation of 30%, were represented in 87% of the area. Comparing the fire risk map against the map of the real recent fire provided support to the general approach, and strengthened the confidence of our fire risk model.     

Land cover analysis in wildland-urban interfaces according to wildfire risk: A case study in the South of France

Each year, forest fires destroy about 500,000 ha of vegetation in Europe, predominantly in the Mediterranean region. Many large fires are linked to the land transformations that have taken place in the Mediterranean region in recent decades that have increased the risk of forest fires. On the one hand, agricultural fallows and orchards are slowly being colonized by vegetation, and on the other hand, the forest is not sufficiently used, both of which result in increased accumulation of fuel. In addition, urbanization combined with forest extension results in new spatial configurations called “wildland-urban interfaces” (WUI). WUI are commonly defined as “areas where urban areas meet and interact with rural lands, wildland vegetation and forests”. Spatial analyses were performed using a WUI typology based on two intertwined elements, the spatial organization of homes and the structure of fuel vegetation. The organization of the land cover in terms of representativeness, complexity or road density was evaluated for each type of WUI. Results showed that there were significant differences between the types of WUI in the study area. Three indicators (i) “fire ignition density”, derived from the distribution of fire ignition points, (ii) “wildfire density”, derived from the distribution of wildfire area and (iii) “burned area ratio”, derived from the proportion of the burned area to the total study area were then compared with each type of WUI. Assuming that the three indicators correspond to important aspects of fire risk, we showed that, at least in the south of France, WUI are at high risk of wildfire, and that of the different types of wildland-urban interfaces, isolated and scattered WUI were the most at risk. Their main land cover characteristics, i.e. low housing and road densities but a high density of country roads, and the availability of burnable vegetation such as forested stands and shrubland (garrigue) explain the high fire risk. Improving our knowledge of relationships between WUI environments and fire risk should increase the efficiency of wildfire prevention: to this end, suitable prevention actions and communication campaigns targeting the types of WUI at the highest risk are recommended.     

Modeling fire susceptibility in west central Alberta, Canada

Strategic modification of forest vegetation has become increasingly popular as one of the few preemptive activities that land managers can undertake to reduce the likelihood that an area will be burned by a wildfire. Directed use of prescribed fire or harvest planning can lead to changes in the type and arrangement of forest vegetation across the landscape that, in turn, may reduce fire susceptibility across large areas. While among the few variables that fire managers can influence, fuel conditions are only one of many factors that determine fire susceptibility. Variations in weather and topography, in combination with fuels, determine which areas are more likely to burn under a given fire regime. An understanding of these combined factors is necessary to identify high fire susceptibility areas for prioritizing and evaluating strategic fuel management activities, as well as informing other fire management activities, such as community protection planning and strategic level allocation of fire suppression resources across a management area. We used repeated fire growth simulations, automated in the Burn-P3 landscape-fire simulation model, to assess spatial variations in fire susceptibility across a 2.4 million ha study area in the province of Alberta, Canada. The results were used to develop a Fire Susceptibility Index (FSI). Multivariate statistical analyses were used to identify the key factors that determine variation in FSI across the study area and to describe the spatial scale at which these variables influence fire susceptibility at a given location. A fuel management scenario was used to assess the impact of prescribed fire treatments on FSI. Results indicated that modeled fire susceptibility was strongly influenced by fuel composition, fuel arrangement, and topography. The likelihood of high or extreme FSI values at a given location was strongly associated with the percent of conifer forest within a 2-km radius, and with elevation and ignition patterns within a 5-km radius. Results indicated that prescribed fire treatments can be effective at reducing forest fire susceptibility in community protection zones and that simulation modeling is an effective means of evaluating spatial variation in landscape fire susceptibility.     

Determination of fire risk to assist fire management for insular areas: the case of a small Greek island

Forest fire risk estimation constitutes an essential process to prevent high-intensity fires which are associated with severe implications to the natural and cultural environment. The primary aim of this research was to determine fire risk levels based on the local features of an island,namely, the impact of fuel structures, slope, aspects, as well as the impact of the road network and inhabited regions. The contribution of all the involved factors to forest fires ignition and behavior highlight certain regions which are highly vulnerable. In addition, the influence of both natural and anthropogenic factors to forest fire phenomena is explored. In this study, natural factors play a dominant role compared to anthropogenic factors. Hence essential preventative measures must focus on specific areas and established immediately. Indicative measures may include: the optimal allocation of watchtowers as well as the spatial optimization of mobile firefighting vehicles;and, forest fuel treatments in areas characterized by extremely high fire risk. The added value of this fire prediction tool is that it is highly flexible and could be adopted elsewhere with the necessary adjustments to local characteristics.     

Feedbacks between fuel reduction and landscape homogenisation determine fire regimes in three Mediterranean areas

In densely populated areas like the Mediterranean, wildfire extent is mostly limited by fire suppression and fuel fragmentation. Fire is known to spread more easily through high fuel loads and homogenous terrain and it is supposed to reduce fuel amount and continuity, creating a negative feedback. Here we combine information from administration fire records, satellite imagery fire scars and land use/cover maps to asses the effects of fire on landscape structure and vice versa for three areas in Catalonia (NE Spain). We worked with three spatial focuses: the actual fire scar, 1 km² squares and 10 km² squares. In these regions agriculture land abandonment has lead to increased fuel continuity, paralleled by an increment of fire size. We confirm that fire spread is facilitated by land use/cover types with high fuel load and by homogeneous terrain and that fire reduces fuel load by transforming forests into shrublands. But we also found that fire increased landscape homogeneity, creating a positive feedback on fire propagation. We argue that this is possible in landscapes with finer grain than fire alone would create. The lack of discontinuities in the fuel bed diminishes the extinction capacity of fire brigades and increases the risk of large fires. We recommend that fire management should focus more on conservation of the traditional rural mosaic in order to prevent further increases in fuel continuity and fire risk.     

Historic and current fire regimes in the Great Xing’an Mountains,northeastern China: Implications for long-term forest management

Understanding both historic and current fire regimes is indispensable to sustainable forest landscape management. In this paper, we use a spatially explicit landscape simulation model, LANDIS, to simulate historic and current fire regimes in the Great Xing’an Mountains, in northeastern China. We analyzed fire frequency, fire size, fire intensity, and spatial pattern of burnt patches. Our simulated results show that fire frequency under the current fire scenario is lower than under the historic fire scenario; total area burnt is larger with lower fire intensity under the historic fire scenario, and smaller with higher fire intensity under the current fire scenario. We also found most areas were burned by high intensity fires under the current fire scenario, but by low to moderate fires under the historic fire scenario. Burnt patches exhibit a different pattern between the two simulation scenarios. Large patches burnt by high intensity class fires dominate the landscape under the current fire scenario, and under historic fire scenario, patches burnt by low to moderate fire intensity fires have relatively larger size than those burnt by high intensity fires. Based on these simulated results, we suggest that prescribed burning or coarse woody debris reduction should be incorporated into forest management plans in this region, especially on north-facing slopes. Tree planting may be a better management option on these severely burned areas whereas prescribed burning after small area selective cutting, retaining dispersed seed trees, may be a sound forest management alternative in areas except for the severely burned patches.     

Using Monte Carlo simulations to estimate relative fire ignition danger in a low-to-medium fire-prone region

A comprehensive assessment of fire ignition danger is nowadays a basic step towards the prioritization of fire management measures. In this study we propose performing a fire selectivity analysis using Monte Carlo simulations to statistically estimate the relative fire ignition danger in a low-to-intermediate fire-prone region such as Canton Ticino, Switzerland. We define fire ignition danger as the likelihood that at a given place a fire will be ignited. For each 25 m × 25 m pixel of the study area, landscape characteristics that may be related to the probability of fire ignition such as vegetation type, elevation, aspect, slope, urban-forest interface were first split into 9-12 categories. The selectivity of each category with respect to fire ignition was then statistically tested by means of Monte Carlo simulations. Finally, we proposed two different approaches for calculating the ignition danger index: cumulating the scores of the Monte Carlo simulations to a final index or producing synthetic scores by performing a principal component analysis of the Monte Carlo results. The validation of the resulting fire danger indices highlights the suitability of both proposed approaches. The PCA-option allows a slightly better discrimination between ignition and non-ignition points and may be of more general application.     

Nomographs for predicting crown fire initiation in Aleppo pine (<Emphasis Type="Italic">Pinus halepensis</Emphasis> Mill.) forests

Nomographs that calculate the threshold values of surface fire parameters which lead to crown fire initiation were created by linking two separate fire behavior models: Van Wagner’s crown fire ignition criteria and Byram’s surface fire model. The nomographs were also based on the existing surface (fuel load, fuel heat content) and canopy (foliar moisture content, live crown base height) fuel models of Aleppo (Pinus halepensis Mill.) pine forests of Mediterranean Greece. The most important fire parameters for crown fire initiation that are calculated by the nomographs are the critical flame length and the forward spread rate of the surface fire. These parameters are readily observable in the field during fires. The nomographs provide a judicious way to assess whether a crown fire is likely to occur in a conifer forest stand. Although the fire behavior models used had limited testing and are based on certain assumptions, yet they are widely applied in forestry practice worldwide, as a basis for justified fire prevention and suppression planning.     

Sensitivity analysis of fire behavior modeling with LIDAR-derived surface fuel maps

Each year, wildland fires burn millions of hectares of forest worldwide. Fire managers need to provide effective methods for mapping fire fuels accurately. Fuel distribution is very important for predicting fire behavior. The overall aim of this project is to model fire behavior using FARSITE (Fire Area Simulator) and investigate differences in modeling outputs using fuel model maps, which differ in accuracy, in east Texas. This simulator model requires as input spatial data themes such as elevation, slope, aspect, surface fuel model, and canopy cover along with separate weather and wind data. Seven fuel models, including grass, brush, and timber models, are identified in the study area. To perform modeling sensitivity analysis, two different fuel model maps were used, one obtained by classifying a QuickBird image and the other obtained by classifying a LIDAR (LIght Detection and Ranging) and QuickBird fused data set. Our previous investigations showed that LIDAR improves the accuracy of fuel mapping by at least 13%. According to our new results, LIDAR-derived variables also provides more detailed information about characteristics of fire. This study will show the importance of using accurate maps of fuel models derived using new LIDAR remote sensing techniques.     

卫星遥感技术在林火管理与研究中的应用(英文)

卫星遥感已经成为森林火险等级预测、可燃物和火烧区制图、林火监测和火生态研究的一个主要数据来源。本文综述了这些研究领域的研究成果,分析了未来林火管理中采用的卫星遥感技术的发展趋势。根据卫星遥感数据制取的可燃物分布图可以满足林火管理在空间和时间尺度上的需要。单独采用遥感数据或结合地面气象数据可以生成一些火险指数,用于森林火险的预报。目前NOAA 和MODIS 卫星由于有高的时间分辨率已被广泛用于林火探测和监测,这些监测结果可以在许多林火网站上见到,这为世界各地的林火管理和研究提供了重要的参考资料。作为低成本的有效工具, 卫星遥感技术在确定火烧面积和过火区制图上发挥了重要作用。遥感技术的发展也可以用来推断火烧时间和估计火烧程度。卫星遥感也非常适合用来估计生物燃烧面积,这是估计全球或区域生物燃烧排放量和理解火对全球变化的影响的基础。本文还讨论了林火研究中采用的卫星类型。文章最后建议中国需要在卫星遥感技术的应用上进一步发展,提高我国的林火管理水平。参71。     

Weather and human impacts on forest fires: 100 years of fire history in two climatic regions of Switzerland

Understanding the factors driving past fire regimes is crucial in the context of global change as a basis for predicting future changes. In this study, we aimed to identify the impact of climate and human activities on fire occurrence in the most fire-prone regions of Switzerland. We considered forest fires, land use and meteorological data over the period 1904-2008 in the neighboring mountain cantons (states) of Valais and Ticino, which are characterized by distinct climatic regimes.The presence/absence of fire ignitions was analyzed using the Nesterov ignition index (as a proxy for fire weather), road density (for ignition sources), livestock density (for biomass removal), and change in forest area (for fire-prone abandoned agricultural areas).We found that fire weather played a key role in fire occurrence in both regions. Road and livestock densities had similar influences in the two cantons. However, while the increase in forest area was well correlated with fire occurrence in Ticino, no such correlation was evident in Valais, probably because land abandonment and forest cover change have been less extensive there. Our findings emphasize the non-linear nature of the relationships between fire occurrence and anthropogenic drivers, as we found thresholds above which road density was no longer correlated with fire occurrence. This implies that the projected future increase and spatial concentration of the human population may not result in a further increase in fire risk in intermediately to densely populated areas in both cantons.The driving factors behind fire activity differ slightly in the two cantons, in particular with increasing forest area enhancing fire occurrence in Ticino but not in Valais. These differences should be taken into account when assessing future fire risk, especially in Valais where the potential for an increase in the fire-prone area is still high. Fires are likely to become more frequent in a warmer climate, but future fire activity may develop differently in the two cantons. This should be taken into account when planning optimized fire prevention measures. This case study should help to better understand fire activity in highly populated regions where fire activity is still moderate but might markedly increase under a projected more fire-prone climate.     

Litter flammability in oak woodlands and shrublands of southeastern France

Characterizing the flammability of litter fuels is of major importance for assessing wildland fire ignition hazard. Here we compared the flammability of litter within a mosaic of Quercus suber (cork oak) woodlands and shrublands in a Mediterranean fire-prone area (Maures massif, southeastern France) to test whether the characteristics and the flammability of litter vary with the vegetation types. We tested experimentally the ignitability, the sustainability, the combustibility and the consumability of undisturbed (=non-reconstructed) litter samples with a point-source mode of ignition. Although the frequency of ignition was similar between all the vegetation types, we distinguished four groups having litter of specific composition and flammability: low and sparse shrublands dominated by Cistus species, medium shrublands with cork oak, high Erica shrublands with sparse cork oak woodlands, and mixed mature oak woodlands with Q. suber, Q. ilex and Q. pubescens. As these vegetation types corresponded to a specific range of past fire recurrence, we also tested the effect of the number of fires and the time since the last fire on litter flammability. Litters of plots recurrently burned had low ability to propagate flames and low flame sustainability. We discuss how the recent fire history can modify vegetation and litter flammability, and thus the fire ignition hazard.     

Fuel types and crown fire potential in <Emphasis Type="Italic">Pinus halepensis</Emphasis> forests

There is a lack of knowledge to identify and classify forest structures according to the risk of crown fires, especially in Mediterranean regions. In this study, for the first time, we use real information, obtained after a wildfire that burnt under extreme meteorological conditions, to classify forest structures of Pinus halepensis into fuel types as a function of crown fire potential. We identified fourteen forest structures which characterize many forest types in Western Mediterranean areas depending on canopy closure, number of tree layers, percent of each tree layer and overall tree density. By using the pattern of fire types that burnt the most numerous forest structures, we have identified four fire hazard groups of forest structures which are considered different fuel types. The first two had the lowest risk of active crown fires and they differed in the proportion of surface fires and passive crown fires. The third fuel type was the threshold between structures with low and high extreme fire behavior; while the fourth had a high risk of active crown fires. Firefighters and forest managers who are demanding this kind of schema, will test and upgrade this classification of fuel types in function of crown fire potential during future wildfires.     

Does forest biomass harvesting for energy reduce fire hazard in the Mediterranean basin? a case study in the Caroig Massif (Eastern Spain)

The Mediterranean basin is a fire-prone area and is expected to continue being so according to projected climate and socioeconomic changes. Sustainable exploitation of forest biomass could have a positive effect on wildfire hazard mitigation. A modelling approach was used to compare how four different Scenarios for biomass collection for energy use affect fire behaviour and potential burnt area at landscape level under extreme meteorological conditions in a typical Mediterranean Massif. A case study of Pinus halepensis stands in Valencia (Eastern Spain) was conducted. The FARSITE simulator was used to evaluate the burnt area and fire behaviour parameters. Simulations predicted a significant increase in the burnt area and the values of most fire behaviour parameters in a Scenario of rural abandonment, relative to the current situation. Biomass management through thinning reduced canopy bulk density; however, no differences in the values of the main fire behaviour parameters were detected. Thinning and understory clearing, including biomass collection in large shrub fuel model areas, significantly reduces fire hazard. Forest biomass sustainable harvesting for energy is expected to reduce fire hazard if management includes intense modification of fuel models, comprising management of shrub biomass at the landscape level. Strong modification of forest fuel models requires intensive silvicultural treatments. Therefore, forest biomass collection for energy in the Mediterranean basin reduces fire hazard only if both tree and shrub strata are managed at landscape level.     

Effects of fuel treatments on fire severity in an area of wildland–urban interface,Angora Fire,Lake Tahoe Basin,California

The Angora Fire burned 1243 ha of Jeffrey pine and mixed conifer forest in the Lake Tahoe Basin between June 24 and July 2, 2007. The Angora Fire burned at unusually high severity due to heavy fuels; strong winds; warm, dry weather; and unseasonably low fuel moistures. The fire destroyed 254 homes, and final loss and suppression cost estimates of $160,000,000 make the Angora Fire one of the ten costliest wildfires in US history. The Angora Fire burned into 194 ha of fuel treatments intended to modify fire behavior and protect private and public assets in the Angora Creek watershed. The fire thus provides a unique opportunity to quantitatively assess the effects of fuel treatments on wildfire severity in an area of wildland–urban interface. We measured fire effects on vegetation in treated and adjacent untreated areas within the Angora Fire perimeter, immediately after and one year after the fire. Our measures of fire severity included tree mortality; height of bole char, crown scorch, and crown torch; and percent crown scorch and torch. Unlike most studies of fuel treatment effectiveness, our study design included replication and implicitly controlled for variation in topography and weather. Our results show that fuel treatments generally performed as designed and substantially changed fire behavior and subsequent fire effects to forest vegetation. Exceptions include two treatment units where slope steepness led to lower levels of fuels removal due to local standards for erosion prevention. Hand-piled fuels in one of these two units had also not yet been burned. Excepting these units, bole char height and fire effects to the forest canopy (measured by crown scorching and torching) were significantly lower, and tree survival significantly higher, within sampled treatments than outside them. In most cases, crown fire behavior changed to surface fire within 50 m of encountering a fuel treatment. The Angora Fire underlines the important role that properly implemented fuel treatments can play in protecting assets, reducing fire severity and increasing forest resilience.     

The role of fire in forest dynamics in Daxinganling region

Recent fire statistics and preliminary fire history data suggest that fire has been historically responsible for maintaining the vegetative communities up to present in Daxinganling region. Forest types, and even tree species, arc dependent on the degree of fire intensity, fire size, depth of burn and fire frequency. Selected samples of larch, pine, birch and spruce forest were studied in terms of species composition as determined by fire frequency which mainly depends on topography and site conditions. Intervals between fires range between 6 and 170 years.     

Modeling prescribed burns to serve as regional firebreaks to allow wildfire activity in protected areas

Management around wilderness parks ideally requires thorough fire suppression in proximate settled and commercially exploited lands and natural fire within protected areas. To satisfy these requirements, we explored a potential regional firebreak (firewall) based on a series of prescribed burns in Quetico Provincial Park in northwestern Ontario, Canada. Fire managers were recruited each to independently devise a regional firebreak using simulated prescribed burns. The experts’ five designs consisted of between 9 and 25 prescribed burns, set over periods ranging from 3 to 8 years, and covering from 7900 to 26,100 ha. Each wildlife ignition was run after the entire firebreak was created and the vegetation was reclassified to account for post-fire vegetation re-growth. The potential efficacy of each design was tested using worst-case historical weather and 100 random ignitions in the Prometheus fire growth simulation model. Without a firewall, 100 ignitions resulted in 69 fires escaping the park and consuming 483,900 ha of forest beyond the park boundary. The firewall designs were all effective, reducing the area burned outside the park to between 15,400 and 35,400 ha. There was a 77–90% reduction in the number of fires escaping the firewall areas and an average reduction of fire area beyond the park of 92%. Moreover, one can map the geographic weak points in each design, which encourages iterative firebreak design improvements. For instance, firewalls set nearer the park boundary allowed fewer fires to start between the firewall and the boundary, so increasing firebreak effectiveness. The cost of the above systems can be regarded as taking preventative measures against the risk of future economic loss, and the modeling approach reduces the uncertainties in associated decision making.     

Factors influencing the formation of unburned forest islands within the perimeter of a large forest fire

Large forest fires have recently increased in frequency and severity in many ecosystems. Due to the heterogeneity in fuels, weather and topography, these large fires tend to form unburned islands of vegetation. This study focuses on a large forest fire that occurred in north-eastern Spain in 1998, which left large areas of unburned vegetation within its perimeter. Based on a satellite post-fire severity map we searched for the relative influence of biotic and abiotic factors leading to unburned island formation. We divided the area of the fire into individual units we called “slopes” which were meant to separate the differential microclimatic effects of contrasted aspects. The number of unburned islands and their areas were related to 12 variables that influence their formation (i.e. land cover composition, aspect, steepness, forest structure, two landscape indices and weather variables). We hypothesized that unburned vegetation islands would concentrate on northern aspects, in less flammable forests (i.e. broadleaf species) and higher fragmentation to interrupt the advance of fire. While north and western aspects did have a higher presence of unburned vegetation islands, our study suggests greater presence of islands in slopes that are larger (i.e. more continuous areas with relatively homogeneous aspect), with greater proportions of forest cover, with higher wood volumes and with lower proportions of broadleaf species. Climate also played a role, with relative humidity and wind speed positively and negatively correlated to island formation, respectively. Unburned vegetation was more frequent on slopes with lower diversity of land covers and higher dominance of one land cover in the slope. Since slopes with only one land cover (i.e. forests) had more islands than slopes with multiple cover types, we infer that under severe meteorological conditions, fragmented forests can be more affected by wind and by water stress, thus burning more readily than forests that are protected from this edge phenomenon. These results would reinforce forest management strategies that avoid linear features (fire-lines and fire-breaks), to enhance fuel treatments that focus on areas and minimize fragmentation.