Quantifying exposedness of the eastern Baltic Sea shores with respect to extremely high and low water levels

Abstract

Extremely high and low water levels are major sources of marine-induced danger to low-lying coastal areas and shallow waters. We explore options for quantification of this danger using parameters of the relevant Generalised Extreme Value (GEV) distributions and an ensemble of projections along the shores of Lithuania, Latvia and Estonia, including the Gulf of Riga, in the eastern Baltic Sea. The novel aspect is the exploration of the likelihood to have a greatly larger magnitude of water level extremes, both maxima and minima, in the future. The analysis is based on the outcome of circulation models RCO for 1961–2005 and RCA4-NEMO for 1961–2009 and water level recordings in 12 locations. The parameters of the GEV distributions for block maxima and minima over 12 months long time periods are evaluated using four different approaches for 367 coastal sectors along about 1400 km long coastal stretch of complicated shape and highly varying exposedness. An ensemble of 40 projections is developed to estimate water level maxima that occur once in 50 yr. The shape parameter of the GEV distribution for water level maxima varies considerably along the study area. Very high water levels in the eastern Gulf of Finland and eastern Gulf of Riga follow a Fréchet distribution and thus may rapidly increase even in a statistically stationary climate. A Gumbel or a 3-parameter Weibull distribution govern water level maxima in the rest of the study area and water level minima in the entire study area. The alongshore variation in the shape parameter can thus be a natural indicator of the potential increase in extremely large water levels. A Gumbel distribution serves as an acceptable substitute to the GEV distribution for water level maxima in most of study area, except for eastern bayheads. We show for the first time that extremely low water levels mostly follow a 3-parameter Weibull distribution. Substantial variability in estimates of parameters of the GEV distribution and extensive mismatch of these estimates using modelled and recorded water levels call for future research into physics and impact of local drivers on recorded water levels.