EARLY WARNING SYSTEM OF METEOROLOGICAL FIRE DANGER FOR PANTANAL BASED ON REMOTE SENSING INFORMATION

We present a statistical model that, using longer and shorter effects of atmospheric conditions on vegetation stress, predicts the likelihood of Fire Radiative Power (FRP) of a wildfire in case an ignition occurs. The model combines a lognormal distribution central body with two Generalized-Pareto distributions for the upper and lower tails and is calibrated using FRP as derived by satellite information, and values of Fire Weather Index as a covariate of the parameters of the distribution. The model is applied to the nine hydrological subregions of Pantanal and data cover the period from 2001 to 2019. Probability of exceeding a prefixed FRP value is then used to define five different classes of meteorological fire danger. The exceptional year of 2020 is used for model validation. The developed procedure is on the basis of an operational early warning system of fire danger for Pantanal that is now being set up.