Changes in the distribution of hydro-climatic extremes in a non-stationary framework

Taha B M J Ouarda,Christian Charron

doi:10.1038/s41598-019-44603-7

Taha B M J Ouarda, Christian Charron

Open Access

https://doi.org/10.1038/s41598-019-44603-7

Copy DOI

Abstract

Hydro-climatic extremes are influenced by climate change and climate variability associated to large-scale oscillations. Non-stationary frequency models integrate trends and climate variability by introducing covariates in the distribution parameters. These models often assume that the distribution function and shape of the distribution do not change. However, these assumptions are rarely verified in practice. We propose here an approach based on L-moment ratio diagrams to analyze changes in the distribution function and shape parameter of hydro-climate extremes. We found that important changes occur in the distribution of annual maximum streamflow and extreme temperatures. Eventual relations between the shapes of the distributions of extremes and climate indices are also identified. We provide an example of a non-stationary frequency model applied to flood flows. Results show that a model with a shape parameter dependent on climate indices in combination with a scale parameter dependent on time improves significantly the goodness-of-fit.

Highlights

An adequate knowledge of the characteristics of hydro-climatic extremes is essential for structure design and management
These diagrams have commonly been used for the selection of the appropriate probability distribution function to fit a given sample data[16]
The L-moment ratio diagrams for station 08MG005 are presented in Fig. 2 where 40-year samples are related to the climate indices Southern Oscillation Index (SOI), Pacific Decadal Oscillation (PDO) and Pacific North American (PNA) (See details in Methods)

Summary

Introduction

An adequate knowledge of the characteristics of hydro-climatic extremes is essential for structure design and management. In the vast majority of non-stationary hydro-climate models, the location and/or scale parameters of the probability function are made dependent on covariates. We propose here to verify these hypotheses using L-moment ratio diagrams These diagrams have commonly been used for the selection of the appropriate probability distribution function to fit a given sample data[16]. They are used in the present study to evaluate the temporal evolution of the scale and shape of hydro-climatic extremes. The usefulness of introducing climate indices in the shape parameter of the probability distribution function in a non-stationary framework is tested. Stations Quatsino (50.53°N, 127.65°W) and Fort St-James (54.46°N, 124.29°W) are selected because of their long record periods: from 1896 to 2010 and 1895 to 2010 respectively

Methods

Results

Conclusion