Wildland fires are globally widespread, constituting the primary forest disturbance in many ecosystems. Burn severity (fire-induced change to vegetation and soils) has short-term impacts on erosion and post-fire environments, and persistent effects on forest regeneration, making burn severity data important for managers and scientists. Analysts can create atlases of historical and recent burn severity, represented by changes in surface reflectance following fire, using satellite imagery and fire perimeters. Burn severity atlas production has been limited by diverse constraints outside the US. We demonstrate the development and validation of a burn severity atlas using the Google Earth Engine platform and image catalogue. We automated mapping of three burn severity metrics using mean compositing (averaging reflectance values) of pixels for all large (≥200 ha) fires in Alberta, Canada. We share the resulting atlas and code. We compared burn severity datasets produced using mean compositing with data from paired images (one pre- and post-fire image). There was no meaningful difference in model correspondence to field data between the two approaches, but mean compositing saved time and increased the area mapped. This approach could be applied and tested worldwide, and is ideal for regions with small staffs and budgets, and areas with frequent cloud.