## Task 2 - Wildfire behavior simulation for management

Duration: 14 months

Start date: January 19th 2020

Person * months: 31

Researchers involved: Adrian Pascual Arranz; Brigite Botequim; José Borges; Paulo Fernandes; Susete Marques and 2 Masters to be hired

Consulting: Marc McDill and José Pereira

This task aims at estimating ignition probability, stand flammability and wildfire spread probability in order to characterize wildfire behavior in a forested landscape under climate change and provide a forest management planning wildfire simulator. For that purpose, Task 2 requires computing resources, takes the data and information from Task 1 and is structured into three subtasks:

2.1 Wildfire ignition probability

In this sub-task, the ignition probability will be estimated using multivariate logistic regression models that will take as dependent variables the distance to roads, the fuel type, the elevation and the population density (Catry et al. 2009). It will read the decision space generated in Task 1 and it will produce the ignition probabilities for each stand in ZIF_VS in each planning period under each climate change scenario.

2.2 Stand flammability

In this sub-task, we will parametrize stand flammability by using logistic regression models that predict wildfire risk for the most relevant forest species in Portugal. These models will have as dependent variables environmental variables as well as biometrical variables available from Task 1. The modeling work will read the decision space generated in Task 1 and it will generate the flammability (probability of a stand to burn and wildfire damage if it occurs) of each stand in each planning period in each climate scenario.

2.3 Wildfire spread probability

In this sub-task, available fire spread model (e.g. Wei et al. 2008) will help to estimate the wildfire spread probability from each stand considering data collected in Task 1 - topographical data (e.g. slope, aspect), meteorological data (e.g. wind direction) and topological data (e.g. the relative position of the adjacent stands). The wildfire spread model results will be validated by the comparison between the results of the application of the wildfire spread model and simulations performed by spatial wildfire behavior simulators. For that purpose, we will consider FlamMap as it is the only available option to estimate maximum ﬁre potential. FlamMap simulations will be performed in each year of the planning horizon based on temporal projections in surface and canopy fuels obtained in Task 1. FlamMap system will explore fire behavior characteristics namely, the wildfire spread in ZIF_VS using data collected in Task 1 and considering a “no climate change” scenario (BAU) under extreme “wildfire conditions” that represent fuel moisture and wind speed of 4% and 40 km h-1, respectively. It will explore further weather conditions derived from climate scenarios (Task 1). In addition, to validate 2.1 and 2.2 the spatial variation among fires ignited in different locations of the forested landscape including burnt probability metrics will be explored using FlamMap. Fire-ignition locations in FlamMap will be based on the application of risk models for Portuguese stands conditions using biometric variables from each inventories plot (Task 1).

The expected outcome of the three subtasks will be a forest management planning wildfire simulator that estimates ignition probability, stand flammability and wildfire spread probability for all stands in ZIF-VS in any time period under each climate scenario. This simulator may be used in a variety of landscape configurations with contrasting vegetation, terrain, and climate conditions.