Polarimetric Sar Forest Fire Scars Extraction And Burn Severity Estimation

Forest fire occurs frequently in global forest fire is an important disturbance factor for global forest ecosystem. It has important and profound impacts on forest succession, formation of biodiversity, carbon cycle in land and atmospheric, global warming. So accurately known the forest fire’s information, including occurrence time, location, extent and intensity, is the basic for understanding the complex relationship between forest fires and the atmosphere, biosphere, and human activities.Remote sensing is the only feasible means to obtain wide range forest fire’s information. Early forest fire analysis most focuses on optical or thermal infrared remote sensing, however, it is disturbed by cloud and weather. Synthetic Aperture Radar data is few affected by weather, it can observe all climates and all day. Since the 1990 s, studies had carried out at abroad, but not yet appeared in domestic. Most studies are based on single polarized channel data, few based on quad-polarization data. Forest fire not only changed the environmental parameters but also by changing the structure of forest canopy and ground vegetation to change the backscatter mechanism. Only the quad-polarization data can provide complete information about radar backscatter.Therefore, this paper use the quad-polarization data acquired before and after fire to carry out the following tasks:(1) Use the C and L band quad-polarization data acquired before and after the fire, explains the changes of fire scars scattering mechanism. Mainly through quantitative analysis the backscatter coefficient, depolarization and decomposition parameters changes before and after the fire to interpret the changes of fire scars scattering mechanism. Provide the basis for following fire scars extraction and burn severity estimation.(2) Comparison the C and L band quad-polarization data monitoring forest fire’s ability. By calculating parameters variation and transformed divergence to realize. Select the data more conductive to forest fire monitoring for following research.(3) In case of only has data acquired after fire, using the optimal polarization technology to extract fire scars. In case of has data acquired both before and after fire, using change detection method to extract fire scars. And extract one forest fire scar occurred in boreal forest as experiment, get high precision.(4) Using the quad-polarization parameters acquired before and after fire through regression analysis to establish a model to estimating burn severity. By estimating one forest fire’s severity in boreal forest confirmed the analyzed estimating model’s feasibility.