A nonstationary peaks-over-threshold approach for modelling daily precipitation with covariate-dependent thresholds

Abstract

The estimation of extreme precipitation events is a topic of growing interest and concern, particularly in highly urbanized areas. The Fifth Assessment Report by the Intergovernmental Panel on Climate Change (IPCC) reported that land-surface air temperatures have increased and that more extreme precipitation events are expected in a warmer climate. Hydrometeorological data often display nonstationary characteristics in a changing climate, prompting numerous studies worldwide. This research proposes a peaks-over-threshold (POT) approach using both stationary and covariate-dependent nonstationary thresholds (bivariate and multivariate models). The generalized Pareto (GP) distribution is fit to threshold exceedance data, which accounts for potential nonstationarity in the variability of the extreme events. The analysis herein focuses on coastal British Columbia (BC), which includes the Greater Vancouver Regional District (Metro Vancouver), due to the potential impacts of climate change in the region. Results from quantile estimates and an uncertainty analysis indicate that in the winter and summer months, there is stronger evidence of stationarity in 50-year quantiles. A trend analysis is also performed on the magnitude and frequency of POT events for winter and summer for two time periods (1976–2014 and 1986–2014). Results of this analysis indicate that globally significant trends are found in the threshold exceedance and frequency of peak events data.