Abstract ID: 191

A probabilistic model of fire ignition and spreading for the Xingu Headwaters

Fire model is a key tool to assess forest resilience in face of a changing environment due to global warming and increasingly anthropogenic disturbance. We have developed a spatially explicit model that simulates both fire ignition and spreading processes. Fire ignition is a function of land use, depicted by spatial determinants - such as distances to deforested land, roads and towns, altitude and land use restriction - and climatic seasonality represented by monthly VPD data (vapor pressure deficit). Annual land use and monthly climate probabilities maps were combined to produce a space-time model for fire occurrence, which was calibrated using NOAA-12 night satellite hot pixels for 2003 and then validated for the years 2002, 2004 and 2005. The fire spreading component employs a cellular automata model to simulate fire spreading as a function of distance to ignition sources, as set by the fire occurrence model, terrain features - such as river channels and topography -, land cover, plus climatic conditions inside the forest and biomass fuel availability output from CARLUC. The fire spreading model was calibrated and validated using fire scars obtained from Landsat imagery for 2005. Both models showed good validation results with respect to fire frequency and location. Fire events are passed on to CARLUC, which keeps track of the carbon fluxes among the forest components and from the forest to the atmosphere. In a next phase, these coupled models will be applied to simulate future fire regimes, under various scenarios of Amazon dynamics, to determine forest resilience thresholds and thereby predict the Amazon ecosystem tipping point.

Session:  Fire - Fire, drought, and changes in vegetation structure and composition

Presentation Type:  Oral