Scaling Laws for Extreme Precipitations: Assessing the Impacts of Datasets Characteristics on Extreme Distribution Estimation.

English Abstract:

Characterizing extreme precipitations at small spatial and temporal scales is crucial in order to evaluate and predict the impacts of natural hazards on regional ecosystems. Available datasets, including climate model simulations and reanalysis products, still have many deficiencies related to their temporal and spatial resolution. Biases and uncertainties are also important, especially when considering extreme precipitations.

The present study aims at validating the use of scaling models for the description of the spatio-temporal structure of extreme precipitations in North-America. By means of scaling models, the statistical distribution of the extremes estimated at specific spatial and temporal scales is related to the distribution at other scales. It is therefore possible to assess extreme precipitation distribution at temporal and spatial scales which are only partially or not sampled. Hence, a consistent and parsimonious construction of IDF curves is possible.

The influence of datasets characteristics (e.g., their temporal and spatial resolution, or spatial coverage) on the scaling properties of sub-daily and daily precipitations is investigated through the comparison of different available datasets (station network series, NCEP Stage IV dataset, and reanalysis series). The range of validity, the magnitude, and the variability of the estimated scaling laws are compared among observed datasets and reanalysis having different spatial resolutions. The spatial distribution of scaling estimates is presented. The influence of climatic and geographic characteristics of Canadian region on scaling estimates is also evaluated. The objective is to validate the use of scaling models to estimate IDF curves for precipitation extremes over Canada. Preliminary results will be presented.