Burned area mapping is an essential task for assessing the damages caused by the wildfires. Optical satellite data have often been used for burned area mapping. However, due to cloud cover and smoke, it is sometimes difficult to use optical data for accurate burned area mapping at global scale. The objective of this research is to evaluate the PALSAR- 2 instrument onboard Advanced Land Observation Satellite-2 (ALOS-2) for burned area mapping at several sites around the world. In this study, a global scale bi-temporal dataset using the L-band PALSAR-2 instrument was created for the research of burned area mapping. Then a U-Net segmentation model was adopted to segment the burned area from the bi-temporal dataset. The results showed that L-Band SAR images can effectively map burned area using the segmentation model. Moreover, ablation study on the impact of different backbones revealed that the average U-Net yielded the best performance while U-Net with Resnext101 as the backbone outperforms U-Net for Grassland and Broadleaf Forest.