Mike Flannigan Projects

Landscape Fire Modelling

This section, in part, is taken from G. Cary, R., Keane and M. Flannigan. 2007. Classifying and comparing spatial models of fire dynamics. iLEAPS Newsletter. 4:26-27

Wildland fire is a significant disturbance in many ecosystems worldwide (Crutzen and Goldammer 1993) and the interaction of fire with climate and vegetation over long time spans has major effects on vegetation dynamics, ecosystem carbon budgets, and patterns of biodiversity. Landscape-fire-succession models, that simulate the linked processes of fire and vegetation development in a spatial domain, are one of the few tools that can be used to explore the interaction of fire, weather and vegetation over long time scales (Keane and Finney 2003). There is a diverse set of approaches to predicting fire regimes and vegetation dynamics over long time scales, due in large part to the variety of landscapes, fuels and climatic patterns that foster frequent forest fires, and variation in modeller’s approaches to representing them in models.

Over recent years, an international group of scientists working under the auspices of Global Change and Terrestrial Ecosystems and funded by the US National Centre for Ecological Analysis and Synthesis classified an extensive set of spatial models of fire and vegetation dynamics and compared the behaviour of a subset of them to determine the relative sensitivity of simulated area burned to variation in terrain, fuel pattern, climate and weather. A set of recommendations for the incorporation of fire dynamics into global dynamic vegetation models was also developed.

Model classification

Keane et al.(2004) identified and classified 44 published spatial models of fire and vegetation dynamics. The models were evaluated according to four components (succession, fire ignition, fire spread, and fire effects) by the three evaluation gradients (stochasticity, complexity, and mechanism) using a scale from zero to 10 (zero meant that it was not modelled or applicable and 10 represented the highest level of stochasticity, mechanism, or complexity). Rankings were assigned by model developers and Keane et al.'s (2004) own review of publications of the model.

A general classification of spatial fire models was developed by combining results from a principal components analysis and a TWINSPAN clustering analysis of the evaluation element data. Given the high variance in evaluation elements across models, explanatory categories such as approach, strategy, scale and other descriptive elements were assigned to models. The frequency of keywords for each category across all LFSMs was then analysed to qualitatively identify similar characteristics and natural clusters.

The classification had 12 hierarchically nested classes that are distinguished by their scale of application (coarse vs fine), representation of vegetation (individual plant cohorts vs framed-based community), simulation of succession (empirical, gap diameter, age, or successional pathway) and the explicit or implicit simulation of fire spread. A full description of the classification results are presented in Keane et al. (2004).

Model comparison

Direct comparison among models can be difficult (Bugmann et al. 1996). Cary et al. (2006) developed an approach to compare the sensitivity of modelled area burned to a range of factors in a standardised design across a subset of the spatial models of fire dynamics (EMBYR, FIRESCAPE, LAMOS(DS), LANDSUM and SEM-LAND). Ideally, the comparison would have selected models from all categories of the classification, however, model selection was also constrained by the availability of modellers with sufficient resources to implement the design. The five models represented a spectrum of complexity in model formulation and represented three out of the twelve classification categories presented by Keane et al. (2004).

Variation in terrain was introduced by varying the minimum and maximum elevation of the simulation landscape so that flat, undulating and mountainous landscapes had relief of 0, 1250 and 2500m respectively. Fuel pattern was varied to represent finely clumped (25 ha patches) and coarsely clumped (625 ha patches) patterns of varying fuel age. Weather and climate are essentially different phenomena at fine temporal scales and were treated as orthogonal. Variation in weather was introduced by selecting ten representative years of daily weather records for the landscape where the model has undergone most rigorous validation. Three types of climate were included in the design, including observed, warmer/wetter (+3.6oC, +20% precipitation), and warmer/drier (+3.6oC, -20% precipitation) climate. In this experiment, simulations were limited to one year and vegetation dynamics were not invoked.

Modelled area burned was most sensitive to climate and variation in weather, with four models sensitive to each of these factors and three models sensitive to their interaction (Table 1), giving similar results to the findings of Hely et al. (2001). Models generally exhibited a trend of increasing area burned from observed, through warmer and wetter, to warmer and drier climates. Area burned was sensitive to fuel pattern for EMBYR and terrain for FIRESCAPE which was the only model that incorporated the effect of elevation on site weather by invoking lapse rates in temperature, humidity and precipitation.

Importance for higher order models

These findings have particular significance for the inclusion of fire in Dynamic Global Vegetation Models (DGVMs). The lack of sensitivity of area burned to fine scale fuel pattern indicates that coarse scale DGVMs may not need to incorporate pattern of vegetation within simulation cells, although this depends on the importance of vegetation succession on area burned, which was not tested in this experiment. Also, the general finding of the importance of inter-annual variability in weather (compared with climate) has important implications for the inclusion of fire into DGVMs, because an increase in the year-to-year variation in weather may translate into large effects on area burned as long-term changes in mean temperature and precipitation brought about by climate change. On the other hand, landscape scale pattern in terrain was demonstrated to be important by the one landscape-fire-succession model that incorporates the effect of terrain on weather.

Table 1

Important sources of variation () in area burned in five spatial models of fire and vegetation dyanmics. Variation in terrain (Terrain), fuel pattern (Fuel), climate (Climate) and weather (Weather) factors, and their interactions, was considered important if they explained more than 0.05 and 0.025 of total variation within a model respectively.

Terrain x Fuel     
Terrain x Climate     
Fuel x Climate     
Terrain x Fuel x Climate     
Terrain x Weather    
Fuel x Weather    
Terrain x Fuel x Weather     
Climate x Weather   
Terrain x Climate x Weather    
Fuel x Climate x Weather     
Terrain x Fuel x Climate x Weather     
Relative importance of fuel management, ignition management and weather for area burned: Evidence from five landscape-fire-succession models

The behaviour of five landscape fire models (CAFÉ, FIRESCAPE, LAMOS(HS), LANDSUM and SEM-LAND) was compared in a standardised modelling experiment. The importance of fuel management approach, fuel management effort, ignition management effort and weather in determining variation in area burned and number of edge pixels burned (a measure of potential impact on assets adjacent to fire-prone landscapes) was quantified for a standardised modelling landscape. Importance was measured as the proportion of variation in area or edge pixels burned explained by each factor and all interactions among them. Weather and ignition management were consistently more important for explaining variation in area burned than fuel management approach and effort, which were found to be statistically unimportant. For the number of edge pixels burned, weather and ignition management were generally more important than fuel management approach and effort. Increased ignition management effort resulted in decreased area burned in all models and decreased number of edge pixels burned in three models. When variation in total pixels burned was analysed separately for each of the different fuel management approaches, variation in fuel management effort was found to be important for the random fuel management approach in only one model. By comparison, when the number of edge pixels burned was analysed separately for each fuel management approach, fuel management effort was found to be important for the edge-based fuel management approach in three of the five models. The findings demonstrate that year-to-year variation in weather and the success of ignition management consistently prevail over the effects of fuel management on area burned in a range of modelled ecosystems.


Bugmann H. K. M., Yan X.D., Sykes M.T., Martin P., Lindner M., Desanker P.V and Cumming S.G. 1996. A comparison of forest gap models: model structure and behaviour. Climatic Change, 34, 289-313.

Cary, G.J., Keane, R.K., Gardner, R.H., Lavorel, S., Flannigan, M., Davies, I.D., Li, C., Lenihan, J.M., Rupp, T.S. & Mouillot, F. 2006 Comparison of the sensitivity of landscape-fire-succession models to variation in terrain, fuel pattern, climate and weather. Landscape Ecology, 21, 121-137.

Crutzen, P. J., and J.G. Goldammer. 1993. Fire in the Environment: The ecological, atmospheric and climatic importance of vegetation fires. John Wiley and Sons, New York, NY, USA.

Hely, C., Flannigan, M.D., Bergeron, Y. and McRae, D. 2001. Role of vegetation and weather on fire behavior in the Canadian Mixedwood boreal forest using two fire behavior prediction systems. Canadian Journal of Forest Research. 31:430-441.

Keane R. E., Cary G.J., Davies I. D., Flannigan M. D., Gardner R. H., Lavorel S., Lenihan J. M., Li C., and Rupp S. T. 2004. A classification of landscape fire succession models: spatial simulations of fire and vegetation dynamics. Ecological Modelling, 179, 3-27.

Keane R. E. and Finney M. A. 2003. The simulation of landscape fire, climate, and ecosystem dynamics. In Veblen T. T., Baker W. L., Montenegro G., and Swetnam T. W. (eds.) Fire and Global Change in Temperate Ecosystems of the Western Americas, pp. 32-66. Springer-Verlag, New York, New York, USA.

  • Unordered list

********************************************************** ***************** Facebook Twitter *********************** **********************************************************

Abonnez-vous à
l'Infolettre du CEF!

********************************************************** ************* Écoles d'été et formation **************************** **********************************************************

Écoles d'été et formations

École d'été : Aménagement des écosystèmes forestiers 
30 juillet au 10 août 2018, FERLD (Abitibi)
École d'été : Modélisation de la biodiversité 
19 au 24 août 2018, Orford

********************************************************** ***************** Pub - Carapace ****************** **********************************************************

********************************************************** ***************** Pub - Budworm ****************** **********************************************************

********************************************************** ***************** Pub - Colibri **************************** **********************************************************

********************************************************** ********** Pub 6 - Au coeur de l'arbre *********** **********************************************************

...Une exposition
virtuelle sur l'arbre!

********************************************************** ***************** Boîte à trucs *************** **********************************************************

La référence vedette !

Jérémie Alluard (2016) Les statistiques au moments de la rédaction 

  • Ce document a pour but de guider les étudiants à intégrer de manière appropriée une analyse statistique dans leur rapport de recherche.

Voir les autres...