Learning ocean circulation models with reservoir computing

PLOS ONE, 2016

Lake Victoria provides important ecosystem services including transport, water for domestic and industrial uses and fisheries to about 33 million inhabitants in three East African countries. The lake plays an important role in modulating regional climate. Its thermodynamics and hydrodynamics are also influenced by prevailing climatic and weather conditions on diel, seasonal and annual scales. However, information on water temperature and circulation in the lake is limited in space and time. We use a Regional Oceanographic Model System (ROMS) to simulate these processes from 1 st January 2000 to 31 st December 2014. The model is based on real bathymetry, river runoff and atmospheric forcing data using the bulk flux algorithm. Simulations show that the water column exhibits annual cycles of thermo-stratification (September-May) and mixing (June-August). Surface water currents take different patterns ranging from a lake-wide northward flow to gyres that vary in size and number. An under flow exists that leads to the formation of upwelling and downwelling regions. Current velocities are highest at the center of the lake and on the western inshore waters indicating enhanced water circulation in those areas. However, there is little exchange of water between the major gulfs (especially Nyanza) and the open lake, a factor that could be responsible for the different water quality reported in those regions. Findings of the present study enhance understanding of the physical processes (temperature and currents) that have an effect on diel, seasonal, and annual variations in stratification, vertical mixing, inshore-offshore exchanges and fluxes of nutrients that ultimately influence the biotic distribution and trophic structure. For instance information on areas/timing of upwelling and vertical mixing obtained from this study will help predict locations/seasons of high primary production and ultimately fisheries productivity in Lake Victoria.

We present neural network surrogates that provide extremely fast and accurate emulation of a large-scale circulation model for the coupled Columbia River, its estuary and near ocean regions. The circulation model has O(10 7 ) degrees of freedom, is highly nonlinear and is driven by ocean, atmospheric and river influences at its boundaries. The surrogates provide accurate emulation of the full circulation code and run over 1000 times faster. Such fast dynamic surrogates will enable significant advances in ensemble forecasts in oceanography and weather.

HAL (Le Centre pour la Communication Scientifique Directe), 2002

The ocean numerical model is one of the three essential components of an ocean forecasting system. Observational data, via data assimilation, set the stage for the model forecast. The quality of the forecast will primarily depend on the ability of the ocean numerical model to faithfully represent the ocean physics and dynamics. Even the use of an infinite amount of data to constrain the initial conditions will not necessarily improve the forecast against persistence of a poorly performing ocean numerical model. In this paper, we briefly review the present state of the art of numerical models within the context of operational global ocean prediction systems, discuss their limitations, and present some of the challenges associated with global ocean modeling. We also briefly address how ocean model development can benefit from such operational systems.

Climate Dynamics, 1992

The global ocean circulation with a seasonal cycle has been simulated with a two-and-a-half layer upper-ocean model. This model was developed for the purpose of coupling to an atmospheric general circulation model for climate studies on decadal time scales. The horizontal resolution is 4 ° latitude by 5 ° longitude and is thus not eddy-resolving. Effects of bottom topography are neglected. In the vertical, the model resolves the oceanic mixed layer and the thermocline. A thermodynamic sea-ice model is coupled to the mixed layer. The model is forced at the surface with seasonally varying (a) observed wind stress, (b) heat fluxes, as defined by an atmospheric equilibrium temperature, and (c) Newtonian-type surface salt fluxes. The second layer is coupled to the underlying deep ocean through Newtonian-type diffusive heat and salt fluxes, convective overturning, and mass entrainment in the upwelling regions of the subpolar gyres. The overall global distributions of mixed layer temperature, salinity and thickness are favorably reproduced. Inherent limitations due to coarse horizontal resolution result in large mixed-layer temperature errors near continental boundaries and in weak current systems. Sea ice distributions agree well with observations except in the interiors of the Ross and Weddell Seas. A realistic time rate of change of heat storage is simulated. There is also realistic heat transport from low to high latitudes.

Journal of Computational Physics, 2008

Systematic improvements in algorithmic design of regional ocean circulation models have led to significant enhancement in simulation ability across a wide range of space/time scales and marine system types. As an example, we briefly review the Regional Ocean Modeling System, a member of a general class of three-dimensional, free-surface, terrain-following numerical models. Noteworthy characteristics of the ROMS computational kernel include: consistent temporal averaging of the barotropic mode to guarantee both exact conservation and constancy preservation properties for tracers; redefined barotropic pressure-gradient terms to account for local variations in the density field; vertical interpolation performed using conservative parabolic splines; and higher-order, quasi-monotone advection algorithms. Examples of quantitative skill assessment are shown for a tidally driven estuary, an ice-covered high-latitude sea, a wind-and buoyancy-forced continental shelf, and a mid-latitude ocean basin. The combination of moderate-order spatial approximations, enhanced conservation properties, and quasi-monotone advection produces both more robust and accurate, and less diffusive, solutions than those produced in earlier terrain-following ocean models. Together with advanced methods of data

Journal of Marine Science and Technology

The assessment of marine environmental risk necessitates the simulation of a series of phenomena related to the risk as well as a measurement of creatures exposed to the risk. As a practical tool, the simulation is based on the establishment of a numerical ocean model. Although several decades have passed since the numerical model for ocean dynamics has been presented, there remains room for fundamental approaches to refine the method for computing solutions. This paper is a report of the development of a novel algorithm of the model. In this algorithm, discrete variables are positioned in a grid to maximally elicit the advantages of a numerical scheme adopted to each term in the governing equations and simplify the program structure. The implemented program is applied to a tidal flow and riverine buoyant plume in the Hinchinbrook Channel in the eastern coast of the Australian Continent. The computation reproduces the observed strong oscillatory flows and low-salinity water dynamics...

Ocean Dynamics, 2014

This paper investigates the use of large-scale circulation patterns (El Niñ o-Southern Oscillation and the equatorial Indian Ocean Oscillation), local outgoing longwave radiation (OLR), and previous streamflow information for short-term (weekly) basin-scale streamflow forecasting. To model the complex relationship between these inputs and basin-scale streamflow, an artificial intelligence approach-genetic programming (GP)-has been employed. Research findings of this study indicate that the use of large-scale atmospheric circulation information and streamflow at previous time steps, along with OLR as a local meteorological input, potentially improves the performance of weekly basin-scale streamflow prediction. The genetic programming approach is found to capture the complex relationship between the weekly streamflow and various inputs. Different input variable combinations were explored to come up with the best one. The observed and predicted streamflows were found to correspond well with each other with a coefficient of determination of 0.653 (correlation coefficient r 5 0.808), which may appear attractive for such a complex system.

Journal of Marine Science and Engineering / MDPI, 2025

Modelling hydrodynamic circulation in the marine environment is one of the most challenging topics in the marine sciences. Several aspects of the marine environment (e.g., biochemical, ecological, and geological processes) are profoundly influenced by the prevailing ocean circulation patterns and, more generally, by the distribution of seawater’s physical properties and processes in both the open ocean and coastal zones. However, several gaps and challenges exist in this field. Hence, the present SI covers a variety of topics related to modern techniques of numerical modelling in marine hydrodynamics, such as the following: -Operational forecast models. -The combination of local, regional, and global numerical datasets for detailed simulations of coastal circulation processes. -Model nesting and data assimilation techniques. -Long-term climatic-type numerical studies of marine hydrodynamics. -Advances in wave-induced circulation modelling and hydrodynamic applications. -Coastal hydrodynamic circulation modelling for sediment transport. -Process-oriented studies of physical processes (e.g., water renewal in atolls). -Coupling between atmospheric, wave, and hydrodynamic models. -Advanced Lagrangian numerical techniques. -New approaches in the modelling of spectral wave hydrodynamics.

1994

We describe new parallelization techniques applied to a highly parallel ocean circulation application code – the Miami Isopycnic Ocean Coordinate Model or MICOM. We compare three parallel architectures executing MICOM: vector, massively parallel, and a multiprocessor workstation. Results from a high resolution 0.08 MICOM North Atlantic basin calculation on the Cray T3D are described briefly. This high-resolution calculation gives qualitatively different results than lower resolution runs implying that a resolution threshold must be crossed to retain fidelity to observed ocean currents.

This paper presents the development of a general-purpose parallel ocean circulation model, for use on a wide range of computer platforms, from traditional scalar machines to workstation clusters and massively parallel processors. Parallelism is provided, as a modular option, via high-level message-passing routines, thus hiding the technical intricacies from the user. An initial implementation highlights that the parallel eciency of the model is adversely aected by a number of factors, for which optimisations are discussed and implemented. The resulting ocean code is portable and, in particular, allows science to be achieved on local workstations that could otherwise only be undertaken on state-of-the-art supercomputers.

2009

Abstract: This research is concerned with the fundamental understanding and modeling of complex physical, acoustical and biogeochemical oceanic dynamics and processes. New mathematical models and computational methods are created, developed and utilized for i) ocean predictions and dynamical diagnostics, ii) data assimilation and data-model comparisons, and, iii) optimization and control of autonomous ocean observation systems.

2018

This paper describes the implementation of three different simplified ocean models on a GPU (graphics processing unit) using Python and PyOpenCL. The three models are all based on the solving the shallow water equations on Cartesian grids, and our work is motivated by the aim of running very large ensembles of forecast models for fully nonlinear data assimilation. The models are the linearized shallow water equations, the non-linear shallow water equations, and the two-layer non-linear shallow water equations, respectively, and they contain progressively more physical properties of the ocean dynamics. We show how these models are discretized to run efficiently on a GPU, discuss how to implement them, and show some simulation results. The implementation is available online under an open source license, and may serve as a starting point for others to implement similar oceanographic models.

We overview problems and prospects in ocean circulation models, with emphasis on certain developments aiming to enhance the physical integrity and flexibility of large-scale models used to study global climate. We also consider elements of observational measures rendering information to help evaluate simulations and to guide development priorities.