A Generalized Framework for Process-Informed Nonstationary Extreme Value Analysis

Document Type


Publication Date



Evolving climate conditions and anthropogenic factors, such as CO2 emissions, urbanization and population growth, can cause changes in weather and climate extremes. Most current risk assessment models rely on the assumption of stationarity (i.e., no temporal change in statistics of extremes). Most nonstationary modeling studies focus primarily on changes in extremes over time. Here, we present Process-informed Nonstationary Extreme Value Analysis (ProNEVA) as a generalized tool for incorporating different types of physical drivers (i.e., underlying processes), stationary and nonstationary concepts, and extreme value analysis methods (i.e., annual maxima, peak-over-threshold). ProNEVA builds upon a newly-developed hybrid evolution Markov Chain Monte Carlo (MCMC) approach for numerical parameters estimation and uncertainty assessment. This offers more robust uncertainty estimates of return periods of climatic extremes under both stationary and nonstationary assumptions. ProNEVA is designed as a generalized tool allowing using different types of data and nonstationarity concepts physically-based or purely statistical) into account. In this paper, we show a wide range of applications describing changes in: annual maxima river discharge in response to urbanization, annual maxima sea levels over time, annual maxima temperatures in response to CO2 emissions in the atmosphere, and precipitation with a peak-over-threshold approach. ProNEVA is freely available to the public and includes a user-friendly Graphical User Interface (GUI) to enhance its implementation.