GAR Atlas risk results are presented in terms of probabilistic risk metrics. One of them correspond to the average annual loss (AAL) which accounts for a long-term overview of disaster risk by being the expected loss, averaged on an annual basis, that considers the occurrence of small, medium and extreme events. AAL can serve also as an indicator of the amount of savings a nation needs to set aside, each year, to cover the cost of long-term losses associated to a single peril or to several of them. Since the GAR Atlas risk results have been obtained at global scale, AAL at country level is providing an order of magnitude of the potential extent of losses in a country associated to different hazards. The hazards considered herein, at global scale, are: earthquakes, tropical cyclones (strong wind and storm surge), tsunami and riverine floods. Additionally, for the Pacific Region, volcanic ash-fall has been also considered. The fully probabilistic risk assessment methodology used herein (see Cardona et al., 2015 for details) accounts for the uncertainties and propagate them throughout the analysis. Anyhow, it should be recognized that although the most appropriate datasets available at the time of analysis for hazard, exposure and vulnerability were used, the results still have a level of uncertainty that arises from assumptions and simplifications needed given the global coverage of the modelling process. However, for the purposes of a global-scale analysis and country-to-country comparisons, the uncertainty level is considered as acceptable. These results should be thus considered an initial step towards understanding the extent of disaster losses that a country might face and towards determining further actions in risk identification processes, such as detailed country, subnational and local risk assessments. For the GAR Atlas, the risk was calculated with the CAPRA Team RC+ program, which is the most recent risk modelling tool of the CAPRA suite. The CAPRA model follows a state-of-the-art methodology for assessing catastrophe risk in a fully probabilistic and multi-hazard approach, besides being openly available and widely used at different scales in several regions of the world.