Assessing the Potential of Embedding Vegetation Dynamics into a Fire Behaviour Model : LPJ-GUESS-FARSITE

Abstract: Disturbances such as wildfires are key players involved in the shape, structure and function of the ecosystems. Fire is rarely included in Dynamic global vegetation models due to their difficulty in implementing its processes and impacts associated. Therefore, it is essential to understand the variables and processes involved in fire, and to evaluate the strengths and weaknesses before going forward in global fire modelling. LPJ-GUESS-SPITFIRE allows the calculation of vegetation in a daily-time-step manner. However, the fire module has revealed some flaws in performance. For this reason, an alternative fire area simulator (FARSITE), a robust and semi-empirical model widely used worldwide, has been taken into account. The aim of this study is to assess a potential embedment of vegetation dynamic (LPJ-GUESS-SPITFIRE) into spatial-explicit fire behaviour modelling (FARSITE): LPJ-GUESS-FARSITE. The study includes: (1) a comparison between simulated vegetation and observed vegetation in Mediterranean regions and, to what extent to fire recurrence affects vegetation; (2) the evaluation and comparison of fuel- and tree-related variables from the observed data, and (3) the comparison of fire behaviour performed by each model. Simulations have shown that Quercus coccifera and C3 grasses are dominant at 25 years fire return interval. Besides, the fire return interval influences largely the successional stage of the vegetation. Biomass tends to increase whereas leaf area index and net primary production decrease from short to long fire recurrence periods. Dead fuel loading, fuel depth, fuel moisture 1hr and live grass, simulated in LPJ-GUESS-SPITFIRE, tend to underestimate field measurements. On contrary fuel moisture 10hr and 100hr are overestimated. Fire behaviour results from both models have underestimated field experimental results. FARSITE results, followed by LPJ-GUESS-FARSITE, have been closer related to field data than LPJ-GUESS-SPITFIRE. The results also showed evidence of more intense fires in LPJ-GUESS-FARSITE than in LPJ-GUESS-SPITFIRE, with identical input data. This thesis concludes that both FARSITE and LPJ-GUESS-FARSITE fire behaviour’s outputs are expected to be more realistic than LPJ-GUESS-SPITFIRE. Even though results do still underestimate real observations, there is enough evidence to say that the LPJ-GUESS framework could be improved. The substitution of the SPITFIRE module by FARSITE model, together with an increase of litter and fuel loading and a decrease of fuel moisture, reflects the promising advantages in creating the meta-model LPJ-GUESS-FARSITE.