Ensemble Approach to Deep Learning and Numerical Ocean Modelling of Adriatic Storm Surges

2020 ◽  
Author(s):  
Matjaz Licer ◽  
Lojze Žust ◽  
Matej Kristan
Keyword(s):  
Deep Learning ◽  
Sea Level ◽  
Ocean Circulation ◽  
Adriatic Sea ◽  
State Of The Art ◽  
Storm Surges ◽  
Phase Lag ◽  
Deterministic Models ◽  
Northern Adriatic ◽  
Ensemble Approach

<p>Storm surges are among the most serious threats to Venice, Chioggia, Piran and other historic coastal towns in Northern Adriatic. Adriatic Sea has a well defined lowest seiche period of approximately 22 hours and its amplitude decays on the scale of several days, reinforcing (or diminishing) the tidal signal, depending on the relative phase lag between tides and surges. This makes prediction of Adriatic sea level extremely difficult using conventional deterministic models. The current state-of-the-art predictions of sea surface height (SSH) hence involve numerical ocean models using ensemble forcing. These simulations are computationally-demanding and time consuming, making the method unsuitable for operational or civil rescue services with limited access to dedicated high-performance computing facilities.</p><p>Ensemble approach to deep learning offers a possible solution to the challenges described above. Even though training a deep network may involve substantial computational resources, the subsequent forecasting -- even ensemble forecasting -- is fast and delivers near-realtime SSH predictions (and associated error variances) on a personal computer. In this work we present an ensemble SSH forecast using new deep convolutional neural network for sea-level prediction in the Adriatic basin and compare it to the standard approach using state-of-the-art publicly available modelling components (NEMO ocean circulation model and TensorFlow libraries for deep learning).</p>

HIDRA 1.0: Deep-Learning-Based Ensemble Sea Level Forecasting in the Northern Adriatic 

2021 ◽  
Author(s):  
Lojze Žust ◽  
Matjaž Ličer ◽  
Anja Fettich ◽  
Matej Kristan
Keyword(s):  
Sea Level ◽  
Ocean Circulation ◽  
Tide Gauge ◽  
Feature Learning ◽  
Circulation Model ◽  
Storm Surges ◽  
Building Blocks ◽  
Lead Times ◽  
Atmospheric Forcing ◽  
Northern Adriatic

<p>Interactions between atmospheric forcing, topographic constraints to air and water flow, and resonant character of the basin make sea level modeling in Adriatic a challenging problem. In this study we present an ensemble deep-neural-network-based sea level forecasting method HIDRA, which outperforms our setup of the general ocean circulation model ensemble (NEMO v3.6) for all forecast lead times and at a minuscule fraction of the numerical cost (order of 2 × 10<sup>-6</sup>). HIDRA exhibits larger bias but lower RMSE than our setup of NEMO over most of the residual sea level bins. It introduces a trainable atmospheric spatial encoder and employs fusion of atmospheric and sea level features into a self-contained network which enables discriminative feature learning. HIDRA architecture building blocks are experimentally analyzed in detail and compared to alternative approaches. Results show the importance of sea level input for forecast lead times below 24 h and the importance of atmospheric input for longer lead times. The best performance is achieved by considering the input as the total sea level, split into disjoint sets of tidal and residual signals. This enables HIDRA to optimize the prediction fidelity with respect to atmospheric forcing while compensating for the errors in the tidal model. HIDRA is trained and analysed on a ten-year (2006-2016) timeseries of atmospheric surface fields from a single member of ECMWF atmospheric ensemble. In the testing phase, both HIDRA and NEMO ensemble systems are forced by the ECMWF atmospheric ensemble. Their performance is evaluated on a one-year (2019) hourly time series from tide gauge in Koper (Slovenia). Spectral and continuous wavelet analysis of the forecasts at the semi-diurnal frequency (12 h)<sup>-1</sup> and at the ground-state basin seiche frequency (21.5 h)<sup>-1</sup> is performed. The energy at the basin seiche in the HIDRA forecast is close to the observed, while our setup of NEMO underestimates it. Analyses of the January 2015 and November 2019 storm surges indicate that HIDRA has learned to mimic the timing and amplitude of basin seiches.</p>

scholarly journals HIDRA 1.0: Deep-Learning-Based Ensemble Sea Level Forecasting in the Northern Adriatic

2020 ◽  
Author(s):  
Lojze Žust ◽  
Anja Fettich ◽  
Matej Kristan ◽  
Matjaž Ličer
Keyword(s):  
Sea Level ◽  
Ocean Circulation ◽  
Tide Gauge ◽  
Feature Learning ◽  
Circulation Model ◽  
Storm Surges ◽  
Building Blocks ◽  
Lead Times ◽  
Atmospheric Forcing ◽  
Northern Adriatic

Abstract. Complex interactions between atmospheric forcing, topographic constraints to air and water flow, and resonant character of the basin, make sea level modeling in Adriatic a particularly challenging problem. In this study we present an ensemble deep-neural-network-based sea level forecasting method HIDRA, which outperforms our general ocean circulation model ensemble (NEMO v3.6) for all forecast lead times and at a minuscule fraction of the numerical cost (order of 2 × 10−6). HIDRA exhibits larger bias but lower RMSE than NEMO over most of the residual sea level bins. It introduces a trainable atmospheric spatial encoder and employs fusion of atmospheric and sea level features into a self-contained network which enables discriminative feature learning. The HIDRA architecture building blocks are experimentally analyzed in detail and compared to alternative approaches. Results show individual importance of sea level input for accurate forecast lead times below 24 h and of the atmospheric input for longer time leads. The best performance is achieved by considering the input as the total sea level, split into disjoint sets of tidal and residual signals. This enables HIDRA to optimize the prediction fidelity with respect to atmospheric forcing, while compensating for the errors in the tidal model. HIDRA is trained and analysed on a ten-year (2006–2016) timeseries of atmospheric surface fields from a single member of ECMWF atmospheric ensemble. In the testing phase, both HIDRA and NEMO ensemble systems are forced by the ECMWF atmospheric ensemble. Their performance is evaluated on a one-year (2019) hourly time series from tide gauge in Koper (Slovenia). Spectral analysis of the forecasts at semi-diurnal frequency (12 h)−1 and at ground-state basin seiche frequency (21.5 h)−1 is performed by a continuous wavelet transform. The energy at the basin seiche in the HIDRA forecast is close to the observed, while NEMO underestimates it. Analyses of the January 2015 and November 2019 storm surges indicate that HIDRA has learned to mimic timing and amplitude of resonant sea level excitations in the basin.

Seismic evidence of the Messinian events in the Adriatic basin

2021 ◽  
Author(s):  
Alessandra Lanzoni ◽  
Anna Del Ben ◽  
Edy Forlin ◽  
Federica Donda ◽  
Massimo Zecchin
Keyword(s):  
Sea Level ◽  
Adriatic Sea ◽  
Carbonate Platform ◽  
Primary Role ◽  
Seismic Profiles ◽  
Northern Adriatic ◽  
Platform Margin ◽  
Gargano Promontory ◽  
Po Delta ◽  
Adriatic Basin

<p>The Adriatic basin represents one of several restricted basins located in the Mediterranean Area. It consists of the foreland of three different orogenic belts: the Dinarides to the East, active during the Eocene, the Southern Alps to the North, active since the Cretaceous time, and the Apennines to the West, active since the Paleogene. The Apennines had a primary role during the Messinian Salinity Crisis (MSC), conditioning the connection between the Adriatic basin, the Ionian basin, and the proto-Tyrrhenian basin. During the Messinian, the present Adriatic Sea was characterized by shallow water domains, where gypsum evaporites initially deposited and often successively incised or outcropped. </p><p>In the past 50 years, a massive dataset, composed of 2D multichannel seismic data and boreholes, was collected, covering almost the whole Adriatic basin in the Italian offshore. In this work, we interpreted the Plio-Quaternary base (PQb), based on available public datasets and on seismic profiles present in literature, which provided regional information from the northernmost Trieste Gulf (Northern Adriatic Sea) to the Otranto Channel (Southern Adriatic Sea). Here, we propose the PQb time-structural map, obtained by analyzing more than 600 seismic profiles. The PQb represents both the Messinian erosion and/or the top of the Messinian evaporites. It is characterized by a high-amplitude reflector, commonly called “horizon M” in the old literature. Principal findings concerning the Messinian event are summarized as below: </p><p>-The Northern Adriatic (Gulf of Trieste, Gulf of Venice, Po delta, Kvarner Area) reveals widespread channelized systems produced by the initial decrease of the sea level, followed by subaerial erosion, related to further sea level decrease. High-grade erosion involved the nearby Adriatic carbonate platform in the Croatian offshore, where deep valleys, filled with Last Messinian or post- Messinian sediments, cut through the limestones.</p><p>-The Central Adriatic (from the Po delta to the Gargano Promontory) displays a higher evaporites accumulation than the northern sector. Meanwhile, the Mid-Adriatic Ridge was already developing, along with the Apennine Chain, which was in a westernmost position. Erosional features in the deeper area are related to channelized systems, which followed the evaporites deposition. Meanwhile, also the Mid-Adriatic Ridge was affected by erosion.</p><p>-The Southern Adriatic (from the Gargano Promontory to the Otranto Channel) is characterized by the Mesozoic Apulia carbonate platform, covered by a thin Cenozoic sequence affected by subaerial erosion or non-deposition. The platform margin and the slope leading to the deepest South Adriatic basin, where a Messinian gypsum layer, also recorded in the Albanian and Croatian offshore, shows a lower level of upper erosion.</p><p>In general, we notice strongly variable thicknesses of the horizon M, which is related to submarine erosion (channels), subaerial erosion (discontinuous surfaces), non-deposition (possible unconformity), and tilting toward the surrounding chains (deepening horizons). In this work, we evaluate these different components from a regional point of view.  </p>

Analysis of the eastern Adriatic sea-level extremes

2021 ◽  
Author(s):  
Marija Pervan ◽  
Jadranka Šepić
Keyword(s):  
Sea Level ◽  
Adriatic Sea ◽  
Hot Spot ◽  
Storm Surges ◽  
Low Pressure ◽  
Long Period ◽  
Short Period ◽  
Sea Level Oscillations ◽  
Low Pass ◽  
Level Height

<p>The Adriatic Sea is known to be under a high flooding risk due to both storm surges and meteorological tsunamis, with the latter defined as short-period sea-level oscillations alike to tsunamis but generated by atmospheric processes. In June 2017, a tide-gauge station with a 1-min sampling resolution has been installed at Stari Grad (middle Adriatic Sea), the well-known meteotsunami hot-spot, which is, also, often hit by storm surges. </p><p>Three years of corresponding sea-level measurements were analyzed, and 10 strongest episodes of each of the following extreme types were extracted from the residual series: (1) positive long-period (T > 210 min) extremes; (2) negative long-period (T > 210 min) extremes; (3) short-period (T < 210) extremes. Long-period extremes were defined as situations during which sea level surpasses (is lower than) 99.7 (i.e. 2) percentile of sea level height, and short-period extremes as situations during which variance of short-period sea-level oscillations is higher than 99.4 percentile of total variance[J1]  of short-period series. A strong seasonal signal was detected for all extremes, with most of the positive long-period extremes appearing during November to February, and most of the negative long-period extremes during January to February. As for the short-period extremes, these appear evenly throughout the year, but strongest events seem to appear during May to July.</p><p>All events were associated to characteristic atmospheric situations, using both local measurements of the atmospheric variables, and ERA5 Reanalysis dataset. It was shown that positive low-pass extremes commonly appear during presence of low pressure over the Adriatic associated with strong SE winds (“sirocco”), and negative low-pass extremes are associated to the high atmospheric pressure over the area associated with either strong NE winds (“bora”), or no winds at all. On the other hand, high-pass sea level extremes are noticed during two distinct types of atmospheric situations corresponding to both “bad” (low pressure, strong SE wind) and “nice” (high pressure, no wind) weather.</p><p>It is particularly interesting that short-period extremes, of which strongest are meteotsunamis, are occasionally coincident with positive long-period extremes contributing with up to 50 percent to total sea level height – thus implying existence of a double danger phenomena (meteotsunami + storm surge). </p>

Climatology and process-oriented analysis of the Adriatic sea-level extremes

2021 ◽  
Author(s):  
Jadranka Sepic ◽  
Mira Pasaric ◽  
Iva Medugorac ◽  
Ivica Vilibic ◽  
Maja Karlovic ◽  
...  
Keyword(s):  
Time Series ◽  
Sea Level ◽  
Adriatic Sea ◽  
The South ◽  
Northern Adriatic ◽  
Atmospheric Waves ◽  
Tidal Forcing ◽  
Atmospheric Processes ◽  
South Eastern ◽  
Adriatic Coast

<p>The northern and the eastern coast of the Adriatic Sea are occasionally affected by extreme sea-levels known to cause substantial material damage. These extremes appear due to the superposition of several ocean processes that occur at different periods, have different spatial extents, and are caused by distinct forcing mechanisms.</p><p>To better understand the extremes, hourly sea-level time series from six tide-gauge stations located along the northern and the eastern Adriatic coast (Venice, Trieste, Rovinj, Bakar, Split, Dubrovnik) were collected for the period of 1956 to 2015 (1984 to 2015 for Venice) and analysed. The time series have been checked for spurious data, and then decomposed using tidal analysis and filtering procedures. The following time series were thus obtained for each station: (1) trend; (2) seasonal signal; (3) tides; (4-7) sea-level oscillations at periods: (4) longer than 100 days, (5) from 10 to 100 days, (6) from 6 hours to 10 days, and (7) shorter than 6 hours. These bands correspond, respectively, to sea-level fluctuations dominantly forced by (but not restricted to): (1) climate change and land uplift and sinking; (2) seasonal changes; (3) tidal forcing; (4); quasi-stationary atmospheric and ocean circulation and climate variability patterns; (5) planetary atmospheric waves; (6) synoptic atmospheric processes; and (7) mesoscale atmospheric processes.</p><p>Positive sea-level extremes surpassing 99.95 and 99.99 percentile values, and negative sea-level extremes lower than 0.05 and 0.01 percentile values were extracted from the original time series for each station. It was shown that positive (negative) extremes are up to 50-100% higher (lower) in the northern than in the south-eastern Adriatic. Then, station-based distributions, return periods, seasonal distributions, event durations, and trends were estimated and assessed. It was shown that the northern Adriatic positive sea-level extremes are dominantly caused by synoptic atmospheric processes superimposed to positive tide (contributing jointly to ~70% of total extreme height), whereas more to the south-east, positive extremes are caused by planetary atmospheric waves, synoptic atmospheric processes, and tides (each contributing with an average of ~25%). As for the negative sea-level extremes, these are due to a combination of planetary atmospheric waves and tides: in the northern Adriatic tide provides the largest contribution (~60%) while in the south-eastern Adriatic the two processes are of similar impact (each contributing with an average of ~30%). The simultaneity of the events along the entire northern and eastern Adriatic coast was studied as well, revealing that positive extremes are strongly regional dependant, i.e. that they usually appear simultaneously only along one part of the coast, whereas negative extremes are more likely to appear along the entire coast at the same time.</p><p>Finally, it is suggested that the distribution of sea-level extremes along the south-eastern Adriatic coast can be explained as a superposition of tidal forcing and prevailing atmospheric processes, whereas for the northern Adriatic, strong topographic enhancement of sea-level extremes is also important.</p>

Multiple drivers of extreme sea levels in the northern Adriatic Sea

2021 ◽  
Author(s):  
Christian Ferrarin ◽  
Piero Lionello ◽  
Mirko Orlic ◽  
Fabio Raicich ◽  
Gianfausto Salvadori
Keyword(s):  
Sea Level Rise ◽  
Sea Level ◽  
Adriatic Sea ◽  
Driving Factors ◽  
Northern Adriatic Sea ◽  
Relative Sea Level ◽  
Northern Adriatic ◽  
Sea Levels ◽  
Extreme Sea Levels ◽  
Relative Sea Level Rise

<p><span><span>Extreme sea levels at the coast result from the combination of astronomical tides with atmospherically forced fluctuations at multiple time scales. Seiches, river floods, waves, inter-annual and inter-decad</span></span><span><span>al dynamics and relative sea-level rise can also contribute to the total sea level. While tides are usually well described and predicted, the effect of the different atmospheric contributions to the sea level and their trends are still not well understood. Meso-scale atmospheric disturbances, synoptic-scale phenomena and planetary atmospheric waves (PAW) act at different temporal and spatial scales and thus generate sea-level disturbances at different frequencies. In this study, we analyze the 1872-2019 sea-level time series in Venice (northern Adriatic Sea, Italy) to investigate the relative role of the different driving factors in the extreme sea levels distribution. The adopted approach consists in 1) isolating the different contributions to the sea level by applying least-squares fitting and Fourier decomposition; 2) performing a multivariate statistical analysis which enables the dependencies among driving factors and their joint probability of occurrence to be described; 3) analyzing temporal changes in extreme sea levels and extrapolating possible future tendencies. The results highlight the fact that the most extreme sea levels are mainly dominated by the non-tidal residual, while the tide plays a secondary role. The non-tidal residual of the extreme sea levels is attributed mostly to PAW surge and storm surge, with the latter component becoming dominant for the most extreme events. The results of temporal evolution analysis confirm previous studies according to which the relative sea-level rise is the major driver of the increase in the frequency of floods in Venice over the last century. However, also long term variability in the storm activity impacted the frequency and intensity of extreme sea levels and have contributed to an increase of floods in Venice during the fall and winter months of the last three decades.</span></span></p>

Sign in / Sign up

Export Citation Format

Share Document