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.

scholarly journals HIDRA 1.0: deep-learning-based ensemble sea level forecasting in the northern Adriatic

2021 ◽  
Vol 14 (4) ◽  
pp. 2057-2074
Author(s):  
Lojze Žust ◽  
Anja Fettich ◽  
Matej Kristan ◽  
Matjaž Ličer
Keyword(s):  
Time Series ◽  
Sea Level ◽  
Ocean Circulation ◽  
Tide Gauge ◽  
Circulation Model ◽  
Building Blocks ◽  
Lead Times ◽  
Atmospheric Forcing ◽  
Northern Adriatic ◽  
Set Up

Abstract. Interactions between atmospheric forcing, topographic constraints to air and water flow, and resonant character of the basin make sea level modelling in the Adriatic a challenging problem. In this study we present an ensemble deep-neural-network-based sea level forecasting method HIDRA, which outperforms our set-up 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 2×10-6). HIDRA exhibits larger bias but lower RMSE than our set-up 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 analysed 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 10-year (2006–2016) time series 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 1-year (2019) hourly time series from a tide gauge in Koper (Slovenia). Spectral and continuous wavelet analysis of the forecasts at the semi-diurnal frequency (12 h)−1 and at the ground-state basin seiche frequency (21.5 h)−1 is performed. The energy at the basin seiche in the HIDRA forecast is close to that observed, while our set-up 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.

scholarly journals 2-D reconstruction of past sea level (1950–2003) using tide gauge records and spatial patterns from a general ocean circulation model

2009 ◽  
Vol 5 (2) ◽  
pp. 1109-1132 ◽  
Author(s):  
W. Llovel ◽  
A. Cazenave ◽  
P. Rogel ◽  
A. Lombard ◽  
M. Bergé-Nguyen
Keyword(s):  
Time Series ◽  
Sea Level ◽  
Ocean Circulation ◽  
Tide Gauge ◽  
Circulation Model ◽  
Global Ocean ◽  
Tide Gauges ◽  
Ocean Circulation Model ◽  
Tide Gauge Records ◽  
Function Decomposition

Abstract. A two-dimensional reconstruction of past sea level is proposed at yearly interval over the period 1950–2003 using tide gauge records at 99 selected sites and 44-year long (1960–2003) 2°×2° gridded dynamic heights from the OPA/NEMO global ocean circulation model with data assimilation. An Empirical Orthogonal Function decomposition of the reconstructed sea level over 1950–2003 displays leading modes that reflect two main components: a long-term (multi-decadal) but regionally variable signal and interannual fluctuations dominated by the signature of El Nino-Southern Oscillation. Tests show that spatial trend patterns of the 54-year long reconstructed sea level (1950–2003) significantly depend on the length of the gridded OPA/NEMO time series used to compute spatial covariance signal used for the reconstruction (i.e., the length of the gridded OPA/NEMO time series). On the other hand, the interannual variability is well reconstructed, even with ~10-year long of the OPA/NEMO model or satellite altimetry-based sea level grids. The robustness of the results is assessed, leaving out successively each of the 99 tide gauges when reconstructing the sea level signal and then comparing observed and reconstructed time series at the non contributing tide gauge site. The reconstruction performs well at most tide gauges, especially at interannual frequency.

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>

scholarly journals Evaluation of ocean circulation models in the computation of the mean dynamic topography for geodetic applications. Case study in the Greek seas

2019 ◽  
Vol 9 (1) ◽  
pp. 154-173
Author(s):  
I. Mintourakis ◽  
G. Panou ◽  
D. Paradissis
Keyword(s):  
Sea Level ◽  
Ocean Circulation ◽  
Tide Gauge ◽  
Dynamic Topography ◽  
Tide Gauge Records ◽  
Mean Dynamic Topography ◽  
Case Study Area ◽  
Extensive Evaluation ◽  
Precise Knowledge

Abstract Precise knowledge of the oceanic Mean Dynamic Topography (MDT) is crucial for a number of geodetic applications, such as vertical datum unification and marine geoid modelling. The lack of gravity surveys over many regions of the Greek seas and the incapacity of the space borne gradiometry/gravity missions to resolve the small and medium wavelengths of the geoid led to the investigation of the oceanographic approach for computing the MDT. We compute two new regional MDT surfaces after averaging, for given epochs, the periodic gridded solutions of the Dynamic Ocean Topography (DOT) provided by two ocean circulation models. These newly developed regional MDT surfaces are compared to three state-of-theart models, which represent the oceanographic, the geodetic and the mixed oceanographic/geodetic approaches in the implementation of the MDT, respectively. Based on these comparisons, we discuss the differences between the three approaches for the case study area and we present some valuable findings regarding the computation of the regional MDT. Furthermore, in order to have an estimate of the precision of the oceanographic approach, we apply extensive evaluation tests on the ability of the two regional ocean circulation models to track the sea level variations by comparing their solutions to tide gauge records and satellite altimetry Sea Level Anomalies (SLA) data. The overall findings support the claim that, for the computation of the MDT surface due to the lack of geodetic data and to limitations of the Global Geopotential Models (GGMs) in the case study area, the oceanographic approach is preferable over the geodetic or the mixed oceano-graphic/geodetic approaches.

scholarly journals “Grey swan” storm surges pose a greater coastal flood hazard than climate change

2021 ◽  
Author(s):  
Kevin Horsburgh ◽  
Ivan D. Haigh ◽  
Jane Williams ◽  
Michela De Dominicis ◽  
Judith Wolf ◽  
...  
Keyword(s):  
Sea Level ◽  
Tide Gauge ◽  
Storm Surges ◽  
Natural Variability ◽  
Major Incident ◽  
Dynamical Response ◽  
Mean Sea Level ◽  
Coastal Flood ◽  
Extreme Storm ◽  
Observational Record

AbstractIn this paper, we show that over the next few decades, the natural variability of mid-latitude storm systems is likely to be a more important driver of coastal extreme sea levels than either mean sea level rise or climatically induced changes to storminess. Due to their episodic nature, the variability of local sea level response, and our short observational record, understanding the natural variability of storm surges is at least as important as understanding projected long-term mean sea level changes due to global warming. Using the December 2013 North Atlantic Storm Xaver as a baseline, we used a meteorological forecast modification tool to create “grey swan” events, whilst maintaining key physical properties of the storm system. Here we define “grey swan” to mean an event which is expected on the grounds of natural variability but is not within the observational record. For each of these synthesised storm events, we simulated storm tides and waves in the North Sea using hydrodynamic models that are routinely used in operational forecasting systems. The grey swan storms produced storm surges that were consistently higher than those experienced during the December 2013 event at all analysed tide gauge locations along the UK east coast. The additional storm surge elevations obtained in our simulations are comparable to high-end projected mean sea level rises for the year 2100 for the European coastline. Our results indicate strongly that mid-latitude storms, capable of generating more extreme storm surges and waves than ever observed, are likely due to natural variability. We confirmed previous observations that more extreme storm surges in semi-enclosed basins can be caused by slowing down the speed of movement of the storm, and we provide a novel explanation in terms of slower storm propagation allowing the dynamical response to approach equilibrium. We did not find any significant changes to maximum wave heights at the coast, with changes largely confined to deeper water. Many other regions of the world experience storm surges driven by mid-latitude weather systems. Our approach could therefore be adopted more widely to identify physically plausible, low probability, potentially catastrophic coastal flood events and to assist with major incident planning.

A Near-Uniform Basin-Wide Sea Level Fluctuation of the Mediterranean Sea

2007 ◽  
Vol 37 (2) ◽  
pp. 338-358 ◽  
Author(s):  
Ichiro Fukumori ◽  
Dimitris Menemenlis ◽  
Tong Lee
Keyword(s):  
Atlantic Ocean ◽  
Mediterranean Sea ◽  
Sea Level ◽  
Ocean Circulation ◽  
Large Scale ◽  
Circulation Model ◽  
Level Fluctuation ◽  
Strait Of Gibraltar ◽  
Sea Level Fluctuation ◽  
The Mediterranean

Abstract A new basin-wide oscillation of the Mediterranean Sea is identified and analyzed using sea level observations from the Ocean Topography Experiment (TOPEX)/Poseidon satellite altimeter and a numerical ocean circulation model. More than 50% of the large-scale, nontidal, and non-pressure-driven variance of sea level can be attributed to this oscillation, which is nearly uniform in phase and amplitude across the entire basin. The oscillation has periods ranging from 10 days to several years and has a magnitude as large as 10 cm. The model suggests that the fluctuations are driven by winds at the Strait of Gibraltar and its neighboring region, including the Alboran Sea and a part of the Atlantic Ocean immediately to the west of the strait. Winds in this region force a net mass flux through the Strait of Gibraltar to which the Mediterranean Sea adjusts almost uniformly across its entire basin with depth-independent pressure perturbations. The wind-driven response can be explained in part by wind setup; a near-stationary balance is established between the along-strait wind in this forcing region and the sea level difference between the Mediterranean Sea and the Atlantic Ocean. The amplitude of this basin-wide wind-driven sea level fluctuation is inversely proportional to the setup region’s depth but is insensitive to its width including that of Gibraltar Strait. The wind-driven fluctuation is coherent with atmospheric pressure over the basin and contributes to the apparent deviation of the Mediterranean Sea from an inverse barometer response.

The northern and the eastern Adriatic Sea floods: connecting the extreme sea-levels to cyclone pathways 

2021 ◽  
Author(s):  
Tihana Dević ◽  
Jadranka Šepić ◽  
Darko Koračin
Keyword(s):  
Sea Level ◽  
Tide Gauge ◽  
Middle Eastern ◽  
Origin Time ◽  
Northern Adriatic ◽  
Sea Levels ◽  
North West ◽  
Mediterranean Cyclones ◽  
Extreme Sea Levels ◽  
The Moment

<p>An objective method for tracking pathways of cyclone centres over Europe was developed and applied to the ERA-Interim reanalysis atmospheric data (1979-2014). The method was used to determine trajectories of those Mediterranean cyclones which generated extreme sea levels along the northern and the eastern Adriatic coast during the period from 1979 to 2014. Extreme events were defined as periods during which sea level was above 99.95 percentile value of time series of hourly sea-level data measured at the Venice (northern Adriatic), Split (middle eastern Adriatic) and Dubrovnik (south-eastern Adriatic) tide-gauge stations. The cyclone pathways were tracked backwards from the moment closest to the moment of maximum sea level up to the cyclone origin time, or at most, up to 72 hours prior the occurrence of the sea-level maximum.</p><p>Our results point out that extreme sea levels in Venice normally appear during synoptic situations in which a cyclone centre is located to the south-west and north-west of Venice, i.e., when it can be found over the Gulf of Genoa, or the Alps. On the contrary, extreme sea levels in Dubrovnik are usually associates with cyclone centres above the middle Adriatic, whereas floods in Split seem to appear during both above-described types of situations.</p><p>Occurrence times and intensity of cyclones and extreme sea-levels was further associated with the NAO index. It has been shown that the deepest cyclones and corresponding extreme floods tend to occur during the negative NAO phase.   </p>

scholarly journals Annual Sea Level Changes on the North American Northeast Coast: Influence of Local Winds and Barotropic Motions

2016 ◽  
Vol 29 (13) ◽  
pp. 4801-4816 ◽  
Author(s):  
Christopher G. Piecuch ◽  
Sönke Dangendorf ◽  
Rui M. Ponte ◽  
Marta Marcos
Keyword(s):  
Sea Level ◽  
Wind Stress ◽  
Ocean Circulation ◽  
Tide Gauge ◽  
Sea Level Changes ◽  
Tide Gauge Records ◽  
Barotropic Model ◽  
The North ◽  
Ocean Reanalyses ◽  
Coastal Sea

Abstract Understanding the relationship between coastal sea level and the variable ocean circulation is crucial for interpreting tide gauge records and projecting sea level rise. In this study, annual sea level records (adjusted for the inverted barometer effect) from tide gauges along the North American northeast coast over 1980–2010 are compared to a set of data-assimilating ocean reanalysis products as well as a global barotropic model solution forced with wind stress and barometric pressure. Correspondence between models and data depends strongly on model and location. At sites north of Cape Hatteras, the barotropic model shows as much (if not more) skill than ocean reanalyses, explaining about 50% of the variance in the adjusted annual tide gauge sea level records. Additional numerical experiments show that annual sea level changes along this coast from the barotropic model are driven by local wind stress over the continental shelf and slope. This result is interpreted in the light of a simple dynamic framework, wherein bottom friction balances surface wind stress in the alongshore direction and geostrophy holds in the across-shore direction. Results highlight the importance of barotropic dynamics on coastal sea level changes on interannual and decadal time scales; they also have implications for diagnosing the uncertainties in current ocean reanalyses, using tide gauge records to infer past changes in ocean circulation, and identifying the physical mechanisms responsible for projected future regional sea level rise.

scholarly journals A century of sea level data and the UK's 2013/14 storm surges: an assessment of extremes and clustering using the Newlyn tide gauge record

2014 ◽  
Vol 11 (4) ◽  
pp. 1995-2028 ◽  
Author(s):  
M. P. Wadey ◽  
I. D. Haigh ◽  
J. M. Brown
Keyword(s):  
Sea Level ◽  
Return Period ◽  
Tide Gauge ◽  
Storm Surges ◽  
High Water ◽  
Lower Probability ◽  
Tide Gauge Record ◽  
Storm Activity ◽  
Sea Levels ◽  
Level Data

Abstract. For the UK's longest and most complete sea level record (Newlyn), we assess extreme high water events and their temporal clustering; prompted by the 2013/2014 winter of flooding and storms. These are set into context against this almost 100 yr record. We define annual periods for which storm activity, tides and sea levels can be compared on a year-by-year basis. Amongst the storms and high tides which affected Newlyn the recent winter produced the largest recorded high water (3 February 2014) and five others above a 1 in 1 yr return period. The large magnitude of tide and mean sea level, and the close inter-event spacings (of large return period high waters), suggests that the 2013/2014 high water "season" may be considered the most extreme on record. However, storm and sea level events may be classified in different ways. For example in the context of sea level rise (which we calculate linearly as 1.81 ± 0.1 mm yr−1 from 1915 to 2014), a lower probability combination of surge and tide occurred on 29 January 1948, whilst 1995/1996 storm surge season saw the most high waters of ≥ 1 in 1 yr return period. We provide a basic categorisation of five types of high water cluster, ranging from consecutive tidal cycles to multiple years. The assessment is extended to other UK sites (with shorter sea level records and different tide-surge characteristics), which suggests 2013/2014 was extreme, although further work should assess clustering mechanisms and flood system "memory".

Sign in / Sign up

Export Citation Format

Share Document