The risk due to natural phenomena is a contingent liability and, therefore, a sovereign risk for governments, making it necessary to evaluate in a form suitable potential losses. From this evaluation can be designed policies and strategies to reduce, retain and transfer disaster risk effectively.
Therefore, in this thesis is proposed a methodology for risk assessment disaster based in two different aspects. On the one hand, there is the empirical evaluation of losses, also called retrospective evaluation by using the historical information available in databases of disasters to assess losses from past events and to estimate the risk of a similar used way in the industry vehicle insurance, life, etc. In this case, the fact occured in the past is used as a predictor for the future, at least in the short term, considering stable conditions and trends. On another hand, probabilistic assessments are used to estimate losses in larger future events, even catastrophic, for which historical information is limited or inadequate.