Estimating the distribution of extreme events



[1] Modeling the distribution of extreme events such as major floods, severe rainfall, and high wind speeds can be challenging. Researchers in many fields commonly use the generalized extreme value (GEV) distribution to characterize how these events are distributed over time. However, for many types of extreme events, there are limited data available, making it difficult to fit the GEV distribution. To help deal with this problem, Ailliot et al. generalized two common methods (maximum likelihood and method of moments) for estimating the parameters and high quantiles of the GEV distribution. Using simple parameter constraints, the scientists developed efficient and numerically robust estimation procedures that are suitable for automatic batch processing of many small-to medium-sized data sets. They illustrate their methods by applying them to annual maximum rainfall data in New Zealand. (Water Resources Research, doi:10.1029/2010WR009417, 2011)