Fuel load mapping in the Brazilian Cerrado in support of integrated fire management

The Brazilian Cerrado is considered to be the most species-rich savannah region in the world, covering ~2 million km2. Uncontrolled late season fires promote deforestation, produce greenhouse gases (~25% of Brazil's land-use related CO2 emissions between 2003 and 2005) and are a major threat to the conservation of biodiversity in protected areas. Governmental institutions therefore implemented early dry season (EDS) prescribed burnings as part of integrated fire management (IFM) in protected areas of the Cerrado, with the aim to reduce the area and severity of late dry season (LDS) fires. The planning and implementation of EDS prescribed burning is supported by satellite-based geo-information on fuel conditions, derived from Landsat 8 and Sentinel-2 data. The Mixture Tuned Matched Filtering algorithm was used to analyse the data, and the relationship between the resulting matched fractions (dry vegetation, green vegetation and soil) and in situ surface fuel samples was assessed. The linear regression of in situ data versus matched filter scores (MF scores) of dry vegetation showed an r2 of 0.81 (RMSE = 0.15) and in situ data versus MF scores of soil showed an r2 of 0.65 (RMSE = 0.38). To predict quantitative fuel load, a multiple linear regression analysis was carried out with MF scores of NPV and soil as predictors (adjusted r2 = 0.86; p