HIDRA2: deep-learning ensemble sea level and storm tide forecasting in the presence of seiches – the case of the northern Adriatic

Abstract

Abstract. We propose a new deep-learning architecture HIDRA2 for sea level and storm tide modeling, which is extremely fast to train and apply and outperforms both our previous network design HIDRA1 and two state-of-the-art numerical ocean models (a NEMO engine with sea level data assimilation and a SCHISM ocean modeling system), over all sea level bins and all forecast lead times. The architecture of HIDRA2 employs novel atmospheric, tidal and sea surface height (SSH) feature encoders as well as a novel feature fusion and SSH regression block. HIDRA2 was trained on surface wind and pressure fields from a single member of the European Centre for Medium-Range Weather Forecasts (ECMWF) atmospheric ensemble and on Koper tide gauge observations. An extensive ablation study was performed to estimate the individual importance of input encoders and data streams. Compared to HIDRA1, the overall mean absolute forecast error is reduced by 13 %, while in storm events it is lower by an even larger margin of 25 %. Consistent superior performance over HIDRA1 as well as over general circulation models is observed in both tails of the sea level distribution: low tail forecasting is relevant for marine traffic scheduling to ports of the northern Adriatic, while high tail accuracy helps coastal flood response. Power spectrum analysis indicates that HIDRA2 most accurately represents the energy density peak centered on the ground state sea surface eigenmode (seiche) and comes a close second to SCHISM in the energy band of the first excited eigenmode. To assign model errors to specific frequency bands covering diurnal and semi-diurnal tides and the two lowest basin seiches, spectral decomposition of sea levels during several historic storms is performed. HIDRA2 accurately predicts amplitudes and temporal phases of the Adriatic basin seiches, which is an important forecasting benefit due to the high sensitivity of the Adriatic storm tide level to the temporal lag between peak tide and peak seiche.

Coastal Flood Risk Assessment: An Approach to Accurately Map Flooding through National Registry-Reported Events

“…In this respect, the combination of past flooding events' interpretation (in this study), sea level rise predictions [35,36], and changing atmospheric conditions (with a corresponding increase in the frequency, duration, and intensity of extreme weather events [1, 15,16,34]) should provide better insight into coastal vulnerability and damaging infrastructural effects. However, for Piran, desalination and mixed waterways currently are not appropriately considered within this cumulative climate hazard altogether, and, thus, the future effects for Piran as a coastal city under threat of flooding might be more devastating than currently expected: recent estimates of damage on infrastructure and humans might represent an underestimation [11,14,23].…”

“…Most scholars use sea level rise projections [17,18,34], sea behavior models, or a combination of both, e.g., HYDRA [35,36], to predict and project flooding areas and levels, in contrast to our methodology, which uses past spatial and temporal flooding patterns to map flooding probability zones in Piran. Nevertheless, the current flood-prone mapping results align with spatial patterns of these models' predictions.…”

How to disentangle sea level rise and a number of other processes influencing coastal floods?

“…These investigations enabled different forecasting approaches to be developed. One of them is based on numerical modeling of the Mediterranean Sea with the Adriatic Sea and the Venice Lagoon embedded into it (e.g., Bajo et al 2019) whereas the other relies on machine learning (Rus et al 2023). The performance of both approaches heavily depends on measurements, which are assimilated into the numerical models or are used to train the machine learning system.…”

“…1). The event was forecasted by the operational modeling system developed for the Mediterranean and Adriatic Seas and the Venice Lagoon (Cavaleri et al 2020) and was hindcasted by a machine learning system trained on data originating from a nearby station (Rus et al 2023), but was in both cases underestimated by about 40 cm.…”

DELWAVE 1.0: deep learning surrogate model of surface wave climate in the Adriatic Basin

“…Deep learning has been shown to promise great potential for addressing these issues across multiple fields of science, including machine vision and natural language processing, and, more recently, in various subfields of meteorology (Janssens and Hulshoff, 2022;Beucler et al, 2021;Rasp et al, 2018) and oceanography (Rus et al, 2023;Sonnewald et al, 2021;Boehme and Rosso, 2021;Žust et al, 2021;Mallett et al, 2018). With particular reference to wave dynamics applications, James et al (2018) proposed a machine Abbreviations used on the map are as follows: AA stands for Acqua Alta tower, OB (2 and 3) for Ortona buoy (2 and 3), and MB for Monopoli buoy.…”