Multifractals and Physically Based Estimates of Extreme Floods (Phase 1)

The aim of this project is to better predict floods by using a physically-based approach established on systems that respect a scale symmetry over a wide range of space-time scales. This approach will determine the relationship between flood magnitude and return period for a wide range of aggregation periods.

The major activities of the project include data selection, data analysis and methodology development. It aims to resolve several problems encountered in existing flood studies related to the data, i.e. non-stationarity, long range dependencies and the clustering of extremes, which often results in fat-tailed (i.e., algebraic type) probability distributions. The techniques for handling such non-classical variability over wide ranges of time and space scales exist and the authors believe that they should be applied to dam management throughout the world.

The ambition of this project is to investigate very large data sets of reasonable quality (e.g., daily stream flow data recorded for at least 20 years for several thousands of gauges distributed all over Canada and the USA). At the same time, one of the main objectives of the project remains the ability of using very scarce historical data, very short or incomplete data records for reliable statistical flood predictions. The validation of the Multifractal Flood Frequency Analysis has been developed and performed on time series over both time ranges. The relationship between the estimates of multifractal parameters of the daily data and their yearly maxima was quantified.

This phase evaluates the uncertainty in the results of Phases 1A and 1B.