Stationary vs non-stationary modelling of flood frequency distribution across northwest England

Sina Hesarkazzazi,Rezgar Arabzadeh,Mohsen Hajibabaei,Wolfgang Rauch,Thomas R Kjeldsen,Ilaria Prosdocimi,Attilio Castellarin,Robert Sitzenfrei

doi:10.1080/02626667.2021.1884685

Abstract

ABSTRACT Extraordinary flood events occurred recently in northwest England, with several severe floods in Cumbria, Lancashire and the Manchester area in 2004, 2009 and 2015. These clustered extraordinary events have raised the question of whether any changes in the magnitude and frequency of river flows in the region can be detected. For this purpose, the annual maximum series of 39 river gauging stations in the study area are analysed. In particular, non-stationary models that include time, annual rainfall and annual temperature as predictors are investigated. Most records demonstrate a marked non-stationary behaviour and an increase of up to 75% in flood quantile estimates during the study period. Annual rainfall explains the largest proportion of variability in the peak flow series relative to other predictors considered in our study, providing practitioners with a useful framework for updating flood quantile estimates based on the dynamics of this highly accessible and informative climate indicator.

Highlights

There is a perception that the frequency and magnitude of extreme flood events and storms have changed significantly over the last few decades around the world, mainly because of climate change, seasonal rainfall intensities, temperature varia tions, change in the land cover and deforestation (Coles et al 2001, Milly et al 2002, 2008, Vogel et al 2011, López and Francés 2013, Salas and Obeysekera 2014, Prosdocimi et al 2015)
annual maxima (AM) series for all 39 stations were tested for trends and sudden changes, or change-points, via the Mann-Kendall test (MKT) and Pettitt test (PT), respec tively
Note that the same six stations for which PT detected statistically significant sudden changes are flagged by MKT as series characterized by statistically significant upward trends

Summary

Introduction

There is a perception that the frequency and magnitude of extreme flood events and storms have changed significantly over the last few decades around the world, mainly because of climate change, seasonal rainfall intensities, temperature varia tions, change in the land cover and deforestation (Coles et al 2001, Milly et al 2002, 2008, Vogel et al 2011, López and Francés 2013, Salas and Obeysekera 2014, Prosdocimi et al 2015) These changes have in some cases altered the seasonality of flooding processes and the magnitude of flood flows across Europe, increasing remarkably the fluvial flood hazard in large European regions (i.e. North-Central Europe, see Blöschl et al 2017, 2019) and leading to a change in the characteristics of under lying distribution of river flood flows (non-stationarity). Among several studies carried out within the realm of non-stationarity, different fitting and goodness-of-fit approaches as well as different covari ates and frequency distributions have been utilized

Objectives

Methods

Results

Conclusion