Human-induced fires contribute significantly to CO2 emissions and forest loss in the brazilian Amazon. In this context, synthetic aperture radars emerge as a viable alternative for monitoring tropical regions under constant influence of adverse weather conditions. Thus, the present study aims to evaluate the performance of different filters and filtering windows for ALOS-2/PALSAR-2 data for subsequent detection of burned areas in a portion of the Tapajos Nati- ´ onal Forest - Para. Quantitative and qualitative analyses indicated that the Lee 9x9 filter was the most effective for pre-fire images, while the Gamma 7x7 filter performed better in post-fire images.