Benutzerhandbuch Datenschutzhinweis Impressum Kontakt





Power law distributions of wildfires across Europe: benchmarking a land surface model with observed data


Migliavacca,  Mirco
Biosphere-Atmosphere Interactions and Experimentation, Dr. M. Migliavacca, Department Biogeochemical Integration, Dr. M. Reichstein, Max Planck Institute for Biogeochemistry, Max Planck Society;

Externe Ressourcen
Volltexte (frei zugänglich)

(Verlagsversion), 2MB

Ergänzendes Material (frei zugänglich)
Es sind keine frei zugänglichen Ergänzenden Materialien verfügbar

Mauro, B. D., Fava, F., Frattini, P., Camia, A., Colombo, R., & Migliavacca, M. (2015). Power law distributions of wildfires across Europe: benchmarking a land surface model with observed data. Nonlinear processes in geophysics discussions, 2, 1553-1586. doi:10.5194/npgd-2-1553-2015.

Monthly wildfire burned area frequency is here modeled with a power law distribution and scaling exponent across dierent European biomes are estimated. Data sets, spanning from 2000 to 2009, comprehend the inventory of monthly burned areas from 5 the European Forest Fire Information System (EFFIS) and simulated monthly burned areas from a recent parameterization of a Land Surface Model (LSM), that is the Community Land Model (CLM). Power law exponents are estimated with a Maximum Likelihood Estimation (MLE) for dierent European biomes. The characteristic fire size (CFS), i.e. the area that most contributes to the total burned area, was also calcu10 lated both from EFFIS and CLM data set. We used the power law fitting and the CFS analysis to benchmark CLM model against the EFFIS observational wildfires data set available for Europe. Results for the EFFIS data showed that power law fittings holds for 2–3 orders of magnitude in the Boreal and Continental ecoregions, whereas the distribution of the 15 Alpine, Atlantic are fitted only in the upper tail. Power law instead is not a suitable model for fitting CLM simulations. CLM benchmarking analysis showed that the model strongly overestimates burned areas and fails in reproducing size-frequency distribution of observed EFFIS wildfires. This benchmarking analysis showed that some refinements in CLM structure (in par20 ticular regarding the anthropogenic influence) are needed for predicting future wildfires scenarios, since the low spatial resolution of the model and dierences in relative frequency of small and large fires can aect the reliability of the predictions.