Post-fire analysis of pre-fire mapping of fire-risk: A recent case study from Mt. Carmel (Israel)
Highlights
► Recent wildfire on Mt. Carmel provided an opportunity to evaluate a fire-risk map. ► Most of burnt areas corresponded significantly to high risk levels in the risk map. ► The three highest risk levels occupied 87% of the area, while the five lower risk levels were present in only 5.6% of the area.
Introduction
Fire risk assessments are crucial in forest ecosystems, where high ecological value coincides with dense population (Shoshany and Goldshleger, 2002). High resolution fire risk maps enable managers to plan long-term strategic fire prevention activities (Galtie et al., 2003), yet until recently, fire risk maps were produced at resolutions too coarse to allow strategic planning at local levels (Pastor et al., 2003, Scott and Burgan, 2005). In a recent paper, Carmel et al. (2009) presented an approach in which fire risk was a function of the multitude of factors affecting fire behavior at high resolution, using Monte Carlo simulation of a fire behavior model (FARSITE, Finney, 1998). The model used a variety of inputs such as topography data, fuel information, weather conditions and human activity areas, among others parameters.
In the present study, Mt. Carmel National Park forest (located in northwestern Israel, Fig. 1) served as a pilot area. Although FARSITE does not have its own fuel model, it allows using custom fuel models (e.g. Anderson, 1982, Scott and Burgan, 2005). The basis for the fuel layer was a canopy cover layer coupled to a detailed map of the Mediterranean vegetation formations on Mt. Carmel (Fig. 1; see also Table 2 and more details in Carmel et al., 2009). A single fuel model was assigned to each major vegetation formation and heuristic adjustments were made to two of the fuel models of Scott and Burgan (2005). Fuel models #4 (chaparral) and #1 (herbaceous vegetation), were suppressed by a factor of two while fuel model #10 (conifer forests) was applied to the Eastern Mediterranean pine forests with an adjustment factor of 4 (see Carmel et al., 2009 for details and justification). Fig. 1 illustrates the distribution of fuel models in the study area.
Using the Monte Carlo simulations of fire spread, for each simulation run, a calendar date, fire length, ignition location, weather data and other parameters were selected randomly from known distributions of these parameters. Distance from road served as a proxy for the probability of ignition. The resulting 1000 maps of fire distribution (the entire area burnt in a specific fire) were overlaid to produce a map of ‘hotspots’ and ‘coldspots’ of fire frequency. The findings revealed a clear pattern of fires that seems to be affected by several factors including the location of urban areas, microclimate, topography and the distribution of ignition locations. Despite the fact that the results demonstrated the complexities of fire behavior, they showed a very clear pattern of risk levels even at fine scales (Carmel et al., 2009) where the distance between areas with different risk levels is only hundreds of meters or few kilometers. Our approach was then adopted in several other case studies (e.g. Bar Massada et al., 2009, Ager et al., 2010, Lorz et al., 2010).
The validation of spatially explicit models is not trivial, and in particular fire risk models are difficult to evaluate. Most fire risk studies do not provide any estimate of the associations between model and reality, and such measures are indeed very difficult to obtain. In our study, historic fires in the region served as a partial indication of model accuracy. Yet, immediate validation of a numerical model of this type has not been reported.
A unique opportunity to evaluate the reliability of the fire risk map occurred recently when a severe forest wildfire, the largest in the history of the state of Israel (since 1948), occurred in Mt. Carmel, Israel, between the 2nd and the 5th of December 2010, burnt more than half-million trees (unofficial estimate of the Israel Forest Authority). Unfortunately, the fire caused loss of life and property as well as severe damage to parts of the forest. The size of the burnt area was 2180 ha (Malkinson and Wittenberg, 2011). For comparison, according to Wittenberg et al. (2007), besides dozens of small fires, eight large wildfires were recorded on Mt. Carmel during the last three decades, each consuming areas of 80–530 ha.
During the past several decades, a sharp increase in fire events in the Mediterranean forests has been observed, especially where the anthropogenic pressure is high (FAO, 2001). This tendency exists in Mt. Carmel in which experienced increasing numbers of forest fires, as a result of increasing human activities (Wittenberg et al., 2007). Another influencing factor is the increase in drought processes as a consequence of climate change (Moriondo et al., 2006, Carvalho et al., 2010). Indeed, the weather conditions in Mt. Carmel prior to the wildfire and during its occurrence were exceptional. Summer 2010 was the warmest on record and the following fall was the warmest and driest in the last 40 years with a precipitation amount of about 10% of the perennial average rate of the season. As a result, the vegetation was unusually dry for this time of the year. December is a rainy month in Israel and there are no records of forest fires on this month. However, during the days of the recent wildfire, the air temperature was very high and the relative humidity was extremely low, below 10% (IMS, 2010). These conditions, together with strong dry eastern winds, fanned the fire and resulted in rapid spread of the fire (burnt area of 2180 ha within three days), flame elevation of 60 m and high intensity fire.
Section snippets
Material and methods
The aim of the current study is to compare between the fire risk map and the map of the actual recent fire, in order to validate the reliability of the model above. In principle, an actual fire can be viewed as a single realization of the fire risk model. If the model is a good predictor of fire, then we expect most of the burnt areas to correspond to high risk classes in the model map. On the other extreme, if the model is not informative at all, then the different risk levels would be
Results and discussion
The number of simulated fires that burnt in a specific location was considered as a surrogate of fire risk at that location. The region was divided into ten risk levels using an ‘equal area’ algorithm. We ran 1000 fire simulations, ‘fire’ frequency varied between 0 and 52 fires in a given location. For example, half of the area (risk levels 1–5) had fewer than 18 ‘fires’ in any given location, while the two highest risk levels (covering 20% of the area) suffered 33 fires or more (Fig. 2).
Conclusions
Fire risk models describe and predict a distribution of events. A fire event is a single realization of the predicted distribution. Fire is a complex phenomenon, and it is therefore reasonable to model fire risk using a complex structure that accounts for the many factors that affect fire ignition and propagation. Numerical and computational models are often the only scientific means for understanding and predicting complex, non-linear environmental phenomena. For predictive purposes, producing
References (23)
- et al.
Comparison of landscape fuel treatment strategies to mitigate wildland fire risk in the urban interface and preserve old forest structure
Forest Ecology and Management
(2010) - et al.
Wildfire risk in the wildland–urban interface. A simulation study in northwestern Wisconsin
Forest Ecology and Management
(2009) - et al.
Assessing fire risk using Monte Carlo simulations of fire spread
Forest Ecology and Management
(2009) - et al.
Combining NDVI and surface temperature for the estimation of live fuel moisture content in forest fire danger rating
Remote Sensing of Environment
(2004) - et al.
Mathematical models and calculation systems for the study of wildland fire behaviour
Progress in Energy and Combustion Science
(2003) The predictive validation of ecological and environmental models
Ecological Modelling
(1993)- et al.
Land-use and population density changes in Israel – 1950 to 1990: Analysis of regional and local trends
Land Use Policy
(2002) - et al.
Spatial and temporal patterns of vegetation recovery following sequences of forest fires in a Mediterranean landscape, Mt. Carmel Israel
Catena
(2007) Aids to Determining Fuel Models for Estimating Fire Behavior
(1982)- et al.
The impact of spatial resolution on area burnt and fire occurrence projections in Portugal under climate change
Climatic Change
(2010)