Seasonal Shift in Storm Surges at Brest Revealed by Extreme Value Analysis

Abstract

Climate is changing due to global warming and many of the observed changes since the 1950s are unprecedented over many centuries to many thousands of years (IPCC, 2021). Since the end of the 20th century, the frequency and intensity of the strongest storms have been increasing in the North Atlantic (IPCC, 2013). Damage resulting from storm surges, sea level rise, and coastal flooding presents a major risk for Europe (IPCC, 2014). It is thus essential to investigate how extreme sea levels and storm surges change in a warming climate, in the perspective of predicting them and adapting coastal areas accordingly to future changes.Extreme high sea levels are the joint effect of mean sea level (MSL), tide, and storm surges. Storm surges are generated during extreme weather events such as extra-tropical storms or cyclones, and result from strong, large-scale atmospheric forcing (e.g., Dangendorf et al., 2016). The European coasts are regularly impacted by mid-latitude extra-tropical storms, which cause large surges, i.e., greater than 1 m. This may lead to huge economic losses and sometimes loss of human life. For example, the storm Xynthia hit the French coast severely on February 27 and 28, 2010, causing a large surge of 1.53 m in the harbor of La Rochelle (see location on Figure 1). This was the highest surge ever observed since the installation of the tide gauge in 1997; its return period was estimated to be greater than 100 yr (Pineau-Guillou et al., 2012). This exceptional storm event caused a major coastal flooding (Bertin et al., 2014). Forty-seven people were killed, around 10,000 people had to be evacuated, and the losses were estimated to more than 2.5 billion Euros (Genovese & Przyluski, 2013).

Shift of the storm surge season in Europe due to climate variability

“…The first method is exactly the same as the one already applied to Brest, and fully described in Reinert et al 21 , 31 (and also referred as method 1 in their paper). Following previous studies 1 , 10 and extreme value theory 32 , a non-stationnary Generalised Extreme Value (GEV) distribution is fitted by Maximum Likelihood on monthly maxima.…”

“…The cumulative density function of the GEV is: with the location parameter, the scale parameter and the shape parameter. Following Reinert et al 21 , is constant, because this parameter is generally hard to estimate 32 and does not vary a lot with time 1 , 11 . However, neither nor can be considered constant, because their variability is significant.…”

“…Similarly to Reinert et al 21 , we used the following model for the GEV parameters : The parameters test a possible linear trend, especially the one due to MSL rise and mentioned earlier (MSL being part of the GESLA-2 surge data, see the Data section). The parameters then help capture the amplitude and phase of the annual cycle, with the annual pulsation.…”

“…The second reason is that it allows to minimize the degree of freedom in our model, and to focus on the annual cycle. As in Reinert et al 21 , the GEV parameters are estimated with a Maximum Likelihood Estimator (MLE), and we checked that the parameters contribute significantly to the model, with a deviance statistic test at the 5%-level of significance 32 . The Q–Q plots are displayed in the Supplementary Fig.…”

“…Storm surges more likely occur between December and January 10 . More precisely, at Brest (France), Reinert et al 21 reported a clear shift in the storm surge timing. From the analysis of more than 150 years of tide gauge data, extreme events were found to occur three weeks earlier (mid-December instead of beginning of January) in the winter 2000 than in the 1950s.…”