Monitoring wildfires has become increasingly criticaldue to the sharp rise in wildfire incidents in recent years.Optical satellites like Sentinel-2 and Landsat are extensivelyutilized for mapping burned areas. However, the effectivenessof optical sensors is compromised by clouds and smoke, whichobstruct the detection of burned areas. Thus, satellites equippedwith Synthetic Aperture Radar (SAR), such as dual-polarizationSentinel-1 and quad-polarization RADARSAT-1/-2 C-band SAR,which can penetrate clouds and smoke, are investigated formapping burned areas. However, there is limited research onusing compact polarisation (compact-pol) C-Band RADARSATConstellation Mission (RCM) SAR data for this purpose. Thisstudy aims to investigate the use of compact polarisation RCMdata for burned area mapping through deep learning. Compactpolm-χ decomposition and Compact-pol Radar VegetationIndex (CpRVI) are derived from the RCM Multi-look Complexproduct. A deep-learning-based processing pipeline incorporatingConvNet-based and Transformer-based models is applied forburned area mapping, with three different input settings: usingonly log-ratio images, using only compact-pol decompositionplus CpRVI, and using all three data sources. The test datasetincludes seven wildfire events from the 2023 and 2024 Canadianwildfire seasons in Quebec, British Columbia, and the NorthwestTerritories. The results demonstrate that compact-pol m-χ decomposition and CpRVI images significantly complementlog-ratio images for burned area mapping. The best-performingTransformer-based model, UNETR, trained with log-ratio, m-χ decomposition, and CpRVI data, achieved an F1 Score of0.718 and an IoU Score of 0.565, showing a notable improvementcompared to the same model trained using only log-ratio images(F1 Score: 0.684, IoU Score: 0.557).