Synoptic patterns associated with high-frequency sea level extremes in the Adriatic Sea

Abstract. The study focuses on classification of synoptic conditions leading to episodes of extreme high-frequency sea level oscillations in the Adriatic Sea (Mediterranean). Two types of extreme episodes were obtained from sea level time series measured at six tide gauge stations: (i) HF extremes, extracted from high-frequency components (periods shorter than 2 hours) of sea level time series and defined as periods in which high-frequency component was above a threshold value; and (ii) Compound extremes, extracted from residual (de-tided) time series, and defined as periods in which both high-frequency and residual components were above their respective thresholds. Characteristic synoptic situations preceding both types of extremes were determined using the k-medoid clustering method applied on the ERA5 reanalysis data (mean sea level pressure, temperature at 850 hPa, and geopotential height of 500 hPa level). The structural similarity index measure (SSIM) was used as a distance metric. The data were divided into a training set (from the start of measurements to the beginning of 2018) and a testing set (from the beginning of 2018 to the end of 2020). For each station, k-medoid was used to obtain first 2 and then 3 clusters with characteristic synoptic patterns called medoids. Two distinct patterns, related to HF and Compound extremes were identified at all stations: (i) “summer-type” pattern – characterized by non-gradient mean sea level pressure, warm air advection from the south-southwest at 850 hPa, and a presence of a jet stream at the 500 hPa height, with all three conditions previously found to favour development of meteorological tsunamis, i.e., the strongest of atmospherically triggered high-frequency sea level oscillations; (ii) “winter-type” pattern characterized by pronounced mean sea level pressure gradients favouring winds which induce storm surges, colder low troposphere, and a presence of a jet stream at the 500 hPa level. Including the third cluster into the analysis led to extraction of either a novel “Bora-type” pattern involving strong northeast winds at stations Bakar and Rovinj, or an additional cluster with a medoid which represents refinement of summer- or winter-type patterns. The extracted medoids of clusters were used to label all days of the testing period. It was shown that HF or Compound episodes recorded in the testing period mostly appeared during synoptic situations which highly resembled extracted medoids. The potential of using k-medoid method for forecasting high-frequency sea level oscillations is discussed.