Variability of Future Extreme Rainfall Statistics: Comparison of Multiple IDF Projections

Abstract

A variety of potential approaches and data sets can be used to develop future rainfall intensity-duration-frequency (IDF) statistics at the local scale. The aim of this study was to characterize the variability in an ensemble of future IDF curves generated using a combination of five different climate models, three climate change scenarios, and two downscaling methods. These data sets were prepared for six rainfall stations across two local study sites in southern Ontario: the Toronto and Windsor areas. Several distribution functions used in the derivation of IDF curves were also tested and the best-fit, generalized extreme value (GEV), was used during downscaling. For the 2050s, there was statistically significant variability in the direction of change and magnitude among IDF projections at the Toronto Area stations, with some member cases showing increases and decreases in intensity values within the ensemble. At the Windsor stations, there was a statistically significant trend of increasing storm intensity for the future, but variability in the magnitude of change within the ensemble was apparent. In general, variability among IDF projections for both study areas increased with storm intensity. The variability due to the selection of climate model data sets was greater than that arising from spatial variability in extreme rainfall, with downscaling methods and radiative forcing/emission scenarios contributing far less to the variability.