Abstract: [Objective] Exploring the burned biomass estimation method by using the long time series of satellite remote sensing data products according to forest types at large scale, and getting the burned biomass estimation result by annual for different forest types, were the major objective of the current work.[Method] The MODIS satellite active fire detection products(MOD14A2)covering the land of P. R. China in 2001-2014 have been selected as the datasets, the feature of power law distribution of FRP(fire radiative power)has been analyzed on three forest types; meanwhile, the burned forest biomass estimation models have been developed on the forest types. To get the scaling parameter m, the linear regression fit to probability distribution function in log scales has been applied, and about 10 times forest fire every year have been selected to modify the annual fire duration d. The national burned forest biomass has been estimated on an annual basis. At the same time, the study result calculated from MODIS burned dataset(MCD45A1)were compared with those derived from forest fire emissions calculation model to validate the efficacy of the current estimation method.[Result] The value of FRP of broadleaf forest, coniferous forest and shrub forest follow the power law distribution. In the 14 years, the annual burned biomass of broadleaf forest was in 0.94-1.37 Mt, with a burned biomass in 0.80-1.92 Mt for coniferous forest, shrub forest's burns in 0.37-0.53 Mt. The result of the two methods were significantly different in some years, even certain values derived from forest fire emissions model were far beyond the total burned biomass of the observed 14 years calculated from the method developed in the current work. The result and inter annual variation were more consistent with the statistics of fire numbers and burned area issued by the National Bureau of Statistics.[Conclusion] The value of FRP of the broadleaf forest, coniferous forest and shrub forest follow the power-law distribution characteristics, the national burned forest biomass estimation models have been developed based on the feature by forest type in the 14 years. The method based on power law distribution is more accurate than the result of fire emissions model estimation method.

Key words: satellite remote sensing data, forest fires, biomass burning estimation, FRP