Nonstationary Frequency Analysis of Extreme Precipitation based on Weather Types

Abstract:

Theoretical arguments and climate projections suggest that extreme precipitation (EP) is expected to increase in a warmer climate. Observational studies have started to confirm these predictions, indicating the need to update infrastructure design standards by accounting for EP nonstationarity. Here, I present a statistical framework to quantify EP frequency that incorporates climate nonstationarity through changes in weather type (WT) occurrence. The framework is based on mixed populations of peak-over-threshold (POT) series associated with the dominant WTs in a given region. The occurrence of the WTs is modeled through a Poisson distribution with time-varying parameters, while the POT series via the Generalized Pareto distribution with constant parameters. To demonstrate the value of the proposed method, I use long-term daily precipitation records from the Global Historical Climatology Network in the U.S. Midwest and the WTs recently presented by Zhang and Villarini (2019), which have shown that occurrence of the WT related to heavy precipitation in the region has been increasing since 1949. I first derive WTs from the NCEP-NCAR reanalysis and show that the statistical uncertainty of the nonstationary framework is comparable to a stationary approach based on the Generalized Extreme distribution fitted to annual precipitation maxima, often used in current design. I then use historical and future climate simulations of a set of general circulation models from CMIP6 to quantify projected changes in EP frequency in the region, along with the associated uncertainty.

Zhang, W., and G. Villarini, 2019: On the weather types that shape the precipitation patterns across the U.S. Midwest. Clim. Dyn., 53, 4217–4232.