A Stochastic Extreme Sea Level Model for the German Baltic Sea Coast

This paper describes a framework in which artificial extreme sea levels (ESLs) can be generated for use in flood risk analyses. Such analyses require large numbers of events to accurately assess the risk associated with certain return water levels and quantify uncertainties surrounding the temporal variability of ESL events. Stochastic models satisfy this requirement as they are computationally inexpensive, and thus, many thousands of events may be generated over a very short period of time. As a case study, we have developed a stochastic model for the German Baltic Sea coast capable of simulating the temporal behavior of ESLs. To do this, high-resolution water level data from 45 tide-gauges have been used as model input. At each location, observed ESLs are identified and parameterized. Artificial ESLs are generated using Monte Carlo simulations based on the parametric distribution functions fitted to the parameterized observed ESLs. We show that the method outlined here provides an accurate representation of ESLs at all tide-gauges tested. However, the model is limited by the availability, length, and quality of high-resolution water level data. Due to the rarity of ESLs in the German Baltic Sea, including historical measurements into the stochastic procedure allows for the generation of artificial ESLs more in-line with past extremes. Plain Language Summary Extreme sea levels leading to coastal flooding pose a major threat to coastal populations. With global exposure to such events expected to increase due to a number of factors, including sea level rise and population growth, it is important that an accurate assessment of flood risk is conducted to inform future decisions on coastal defenses and adaptation options. However, extreme sea levels are rare and in regions where water level records are short, it is difficult to accurately determine the level of risk. In this study, we develop an extreme sea level model capable of producing thousands of artificial extreme events to be used in flood risk analyses. The modeled events are based on a number of observed extreme sea level statistics and are therefore representative of recorded events. Our results highlight the need for further digitization of past extreme events so that they may contribute to a better understanding of extreme sea levels.