| Résumé | Subdaily precipitation gauging stations are limited and unevenly distributed across Canada. To support the design of sustainable stormwater infrastructure, especially in the data-sparse regions of Canada, this study presents a novel, gridded intensity-duration-frequency (IDF) dataset of precipitation storms of 5, 10, 15, 30, and 60 min and 1, 2, 6, 12, and 24 h durations and 2, 5, 10, 25, 50, and 100 year return periods. The dataset has been prepared using atmospheric variable (AVs) estimates from two reanalysis products: the North American Regional Reanalysis (NARR) and ERA-Interim. A state-of-the-art machine-learning algorithm, named a support vector machine (SVM), is used to establish the link between AVs and extreme precipitation magnitudes. First, the most relevant AVs shaping precipitation extremes in different parts of Canada are identified, and preliminary estimates of gridded IDFs are produced. The preliminary IDF estimates are corrected for systematic distribution of spatial errors to obtain corrected gridded IDF estimates. Modeled gridded IDF estimates are compared with observations and are found to exhibit a root mean squared error varying between 5% and 25% across different regions of Canada. The gridded IDFs are also found to capture the observed spatial pattern of extreme precipitation reasonably well. |
|---|
