The capability of meteorological models in identifying the conditions associated with large scale fire events

Numerical meteorological models are widely used for predicting severe weather and rainfall extremes and hence providing early warning of potentially high impact events. The linkage between meteorological conditions and wildfire potential as estimated through wildfire danger rating systems is also well-established. In this paper an investigation of the capability of numerical meteorological model outputs alone to provide indicators of large scale fire occurrence has been carried out. In many vegetation types large scale fires occur in extremely fuel dry conditions determined by persistent pattern of very low relative humidity. A number of large scale fires occurring over the last 5 years in locations around the globe have been analyzed in terms of the accompanying dynamics of the main meteorological variables provided by a global circulation model. At the local level (in Italy), the capability of predicting large scale fire events is further investigated through a comparison of several meteorological models with different spatial and temporal resolution, spanning form the ECMWF to MOLOCH and WRF. In this local analysis, all potentially extreme impact fires (burned area > 500 ha) have been considered over a period from 1st of January, 2007 to 31st of December 2011, both during summer and winter. A sensitivity analysis with respect to the spatial resolution of meteorological models has been performed. The accuracy of model forecasts has been assessed by comparison with hourly observations gathered by an extensive network of remotely automated weather stations (RAWS) network operated by the Italian National Civil Protection Department in collaboration with regional meteorological services.