For example, if historical records show that there are, on average, two potentially damaging earthquakes per year in your area, you could model the time to the next earthquake using the exponential distribution. If the time of day of the earthquake makes a difference to the risk, you could model the time of the earthquake using the uniform distribution. If you know the cost of a large earthquake is between 5 m and 100 m with a most likely figure of 20 m. then maybe the best you can do is to use a triangular distribution until more information is available.