Evaluating changes in flood regime in Canadian watersheds using peaks over threshold approach

Flood behaviour is likely to alter as a result of the impacts of climate change. This paper investigates the non-stationarity in the flood regime of several Canadian rivers through an analysis of peaks over threshold (POT) data. Identification of thresholds and ensuring the independence of POT events are the two major challenges in the implementation of a POT approach. In the present research, a semi-automatic approach based on the POT package in R–an open-source software environment for statistical computing has been used. A total of 127 hydrometric reference streamflow gauging stations that are reasonably free from human intervention have been considered in the present analysis. This ensured the reliability of the trends identified in the flow and timing measures considered herein. The POT data have been extracted from daily streamflow data for each hydrometric station. Four flow and timing measures, such as (1) duration of POT events, (2) volume of POT events, (3) annual sum of durations in POT events, and (4) annual sum of volumes of POT events, were extracted using the daily flow data. The trends in the flow and timing measures were investigated using the Mann–Kendall nonparametric test, and the significance of trends was evaluated using the bootstrap resampling approach. The results of the analysis clearly indicated a predominance of decreasing trends at a greater number of stations in all the four measures of flow and timings considered. It can be concluded that the present analysis indicates a remarkable change in the flood regime of the Canadian watersheds considered herein.