Quantification of the climate change-induced variations in Intensity–Duration–Frequency curves in the Canadian Prairies

Abstract

Intensity–Duration–Frequency (IDF) curves are typically used as a standard design tool for various engineering applications, such as storm water management systems. Warming climate, however, changes the extreme precipitation quantiles represented by the IDF curves. This study attempts to construct the future IDF curves in Saskatoon, Canada, under possible climate change scenarios. For this purpose, a stochastic weather generator was used to spatially downscale the daily projections of Global Climate Models (GCMs) from coarse grid resolution to the local point scale. The stochastically downscaled daily precipitation realizations were further disaggregated to ensembles of hourly and sub-hourly (as fine as 5-min) series using a disaggregation scheme developed based on the K-nearest neighbor (K-NN) technique. This framework was applied to construct the future IDF curves in the city of Saskatoon. The sensitivity of the K-NN disaggregation model to the number of nearest neighbors (i.e. window size) was evaluated during the baseline period (1961–1990). The optimum window size was assigned based on the performance in reproducing the historical IDF curves. By using the simulated hourly and sub-hourly precipitation series and the Generalized Extreme Value (GEV) distribution, future changes in IDF curves and associated uncertainties were quantified using a large ensemble of projections obtained from eight GCMs and three representative concentration pathways – RCP2.6, RCP4.5, and RCP8.5. The constructed IDF curves were then compared with the corresponding historical ones and the IDF curves constructed using another genetic programming-based published method. The results show that the sign and the magnitude of future variations in extreme precipitation quantiles are sensitive to the selection of GCMs and/or RCPs, which seem to get intensified toward the end of the 21st century. The quantification of uncertainties suggests that GCMs are the main contributor to the uncertainty, followed by RCPs and the downscaling method.