UC Irvine

Numerous studies show that climatic extremes have increased substantially in the second half of the 20th century. For this reason, analysis of extremes under a non-stationary assumption has received a great deal of attention. In this dissertation, a methodology is developed for deriving non-stationary return levels, return periods, and climate risk assessment using a Bayesian approach. The methodology is presented in the Non-stationary Extreme Value Analysis (hereafter, NEVA) software package. The methodology offers the confidence intervals and uncertainty bounds of estimated non-stationary return levels using both constant and time varying exceedance probability methods. Both stationary and non-stationary components of NEVA are validated for a number of case studies, and have been validated using empirical return levels. The results show that NEVA reliably describes extremes and their return levels. The methodology has been applied for assessing non-stationary extreme return levels in CMIP5 multi-model simulations. Furthermore, the model has been applied for non-stationary precipitation Intensity-Frequency-Duration (IDF). Beyond univariate non-stationary analysis, a novel framework named empirical Bayes conditional extreme value analysis model has been developed for modeling concurrent and conditional extremes. The methodology has been used for detecting potential changes in the hydrological cycle, and assessing joint occurrences of extreme events.