Tero Partanen (FMI)
As a continuation of the fire prediction method developed within the ERA4CS SERV_FORFIRE project and presented at the 12th EARSeL Forest Fires SIG Workshop in Rome, Italy, October 2019, an extension of its results are briefly reported here.
The predicted quantity of interest is the time-dependence of the fire radiative power (FRP) emitted by the fires of an area, which depends on the time of day, day of the year, area, and weather. Based on the prediction method developed and high temporal resolution SEVIRI FRP data, it was found that in areas with large numbers of wildfires on a daily basis during the dry seasons from year to year, the temporal FRP is quite predictable.
On the other hand, in areas in which these conditions are not met, the temporal FRP is not predictable. To demonstrate, founded on parameterization to 2010 data, the map of Fig. 1 illustrates the correlations between the observed and predicted fire radiative energies (FREs, i.e., the temporal integrals of FRPs) on cloudless days over the dry season in 2018 for each 1.5° x 1.5° grid area in the savanna region of south-central Africa, whose observed fire seasonal FRE exceeds 1 PJ amount of energy to ensure adequate predictability.
In most cases the value exceeds 0.6, which indicates a moderate to very strong correlation between observations and predictions and that the method possesses some true predictive power. The ratio of the sums of observed and predicted fire seasonal FREs over all the 106 areas depicted in the right panel of Fig. 1 is 1.39. The method was also tested in Europe using nearly the entire Iberian Peninsula as a fire area of size 6.5° x 6.5°. Similarly to the unpredicted areas, with low FRE, in Fig. 1, it turned out that wildfires, which occur only in scarce numbers even within the vast Iberian area, on sporadic days are either randomly ignited or not.
To predict such fires at a daily level is a game of pure luck, and thus cannot be done sensibly. The results make evident that with the exception of some specific areas of the world, predicting wildfires in general with a reasonable spatiotemporal resolution is practically impossible.