Assimilation Of Sea Level Data Research Articles

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

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.

Generation of diurnal K1 internal tide in the Luzon Strait and its influence on surface tide in the South China Sea

The predominance of diurnal surface tides over semidiurnal surface tides in the South China Sea (SCS) has been attributed to the near‐resonance response of the former in the SCS. Recent observations further revealed vigorous internal tides in the northern SCS. Conceivably, internal tides generated in the Luzon Strait could modify the surface tide in the SCS. We use a three‐dimensional tide model of the East Asian seas to address this issue. With a typical summertime stratification of the SCS as the initial condition, energy budget indicates that one third of the incident K1 surface tide energy is converted to the baroclinic energy over topographic ridges in the Luzon Strait. In comparison with a global tidal model [Matsumoto et al., 2000], our numerical experiments that annihilated or reduced the K1 internal tide in the Luzon Strait led to up to 50% amplification of the simulated K1 surface tide in the SCS. This suggests that the baroclinic energy conversion substantially reduces the amplitude of K1 surface tide in the SCS. The simulated phases in the SCS differ little from those calculated from Matsumoto’s tide model, suggesting that the modification is primarily on the amplitude. Two‐dimensional surface tidal models lack baroclinic energy conversion in the Luzon Strait; the consequent overestimation of surface K1 tide can be reduced only through precise prescription of sea levels in the Luzon Strait or assimilation of sea level data.

Assimilation of sea level data over continental shelves: an ensemble method for the exploration of model errors due to uncertainties in bathymetry

Sea level model error statistics due to uncertainties in bathymetry in shallow seas are investigated through an original approach based on an ensemble method. The model is free-surface, barotropic and implemented over the entire European continental shelf. The main focus is the specific sea level response of the ocean to meteorological forcing in the presence of bathymetric errors. The introduction of such errors is generally unavoidable considering our insufficient knowledge of bottom topography in most shallow regions of the world. These errors are likely to impact on ocean modeling. In particular, coastal gravity waves, which are responsible for an important part of sea level variability due to atmospheric forcing over shelves, are sensitive to bathymetric errors. An ensemble of perturbed bathymetric solutions is generated by first examining differences between selected existing bathymetric databases and then randomly combining perturbations generated from these differences. The objective is that the ensemble of simulations obtained by running the model over these bathymetries best represents the probability density of model states due to this particular source of errors. Interesting space–time characteristics of sea level error covariances in the perspective of sea-level data assimilation are pointed out. These statistics are shown to be neither homogeneous over shelves, nor isotropic when approaching the coast. They are not even stationary, since they are very dependent on the meteorological regime. These features are of crucial importance since they impose heavy constraints on the choice of the scheme necessary to get dynamically consistent results from data assimilation over continental shelves. The question of the ability of this method to provide sensible bathymetry corrections is addressed in the last part of this paper. The correction is obtained by assimilating sea-level data generated by a simulated Wide Swath Ocean Altimeter system via a global inverse formulation using ensemble statistics. Twin experiments are carried out in that perspective. Results are encouraging, even if a significant part of bathymetric errors remains uncorrected after the analysis.

Sea Level Assimilation Experiments in the Tropical Pacific

Idealized twin experiments with the HOPE ocean model have been used to study the ability of sea level data assimilation to correct for errors in a model simulation of the tropical Pacific, using the Cooper and Haines method to project the surface height increments below the surface. This work should be seen in the context of the development of the comprehensive real-time ocean analysis system used at ECMWF for seasonal forecasting, which currently assimilates only thermal data. Errors in the model simulation from two sources are studied: those present in the initial state and those generated by errors in the surface forcing during the simulation. In the former, the assimilation of sea level data improves the convergence of the model toward its twin. Without assimilation convergence occurs more slowly on the equator, compared to an experiment using only correct surface forcing. With forcing errors present the sea level assimilation still significantly reduces the errors almost everywhere. An exception was in the central equatorial Pacific where assimilation of sea level did not correct the errors. This is mainly due to this region responding rapidly to errors in wind stress forcing and also to relatively large freshwater flux errors imposed here. These lead to errors in the mixed layer salinity, which the Cooper and Haines scheme is not designed to correct. It is argued that surface salinity analyses would strongly complement sea level assimilation here.

Fitting a finite element tidal model to sparse data: application to the Straits of Dover

The major difficulty faced by numerical models of tidal dynamics in the limited-area shelf basins concerns the treatment of open boundaries. We show that effective control of the open boundary conditions in a depth-averaged numerical model can be achieved by assimilation of data from coastal tide gauges in the interior of the basin, distant from its open boundaries. We use variational assimilation of sea level data at a number of coastal locations into a quasi-linearized finite element tidal model, and a “weak” formulation of the dynamical constraints. The effectiveness of the procedure is verified on a test problem involving tidal wave propagation in a zonally oriented, narrow, rotating channel with two open boundaries. The assimilation technique was then successfully applied to the reconstruction of the M 2 tide in the Dover Straits using coastal tidal measurements in the ports. A rigorous error analysis is performed using the explicit inversion of the Hessian matrix, associated with the assimilation scheme. As a result, an error chart for the amplitude of the optimized sea surface elevation is obtained.

Assimilation of TOPEX/Poseidon data into a seasonal forecast system

This paper describes first steps towards the quasi-operational assimilation of sea level anomalies from the TOPEX/POSEIDON and ERS1/2 missions into a global ocean model that is used to initialize a coupled oceanatmosphere seasonal forecast system. We estimate the impact that assimilation of sea level might have on the ocean analysis compared to the assimilation of subsurface temperature observations. As a first step satellite-observed sea level anomalies are compared with simulated sea level anomalies from model experiments. To assess the quality of the ocean analysis and the altimeter data further, model results and satellite data are compared to independent sea level observations from tide gauges. Model results are from three experiments: no assimilation, assimilation of subsurface temperature data, and assimilation of sea level data, respectively. The comparison is focused on the equatorial Pacific where subsurface temperatures are frequently sampled and impact of subsurface oceanic conditions on forecast skill is assumed to be large . It is shown that altimeter data assimilation can match subsurface temperature assimilation in this area.