Risk assessment in regions with low earthquake activity is important for reinsurance companies and governmental building authorities. They need a complete picture of the possible risks. Even contradictive opinions have to be taken into account. Data for this kind of analysis, especially in natural disasters, is of poor quality. Standard statistical analysis is not possible. Extreme values are rare and it is therefore not possible to postulate certain distributions. We present a methodology, where we integrate expert estimates and statistical models. The result will be used for risk assessment. We demonstrate further, how to take individual and automated decisions and how to implement them efficiently. Keywords. fuzzy probability, risk assessment, decision support, fuzzyorder