2004 LOM Workshop Monday 9:15 - 9:35 a.m.
HIM algorithmic developments required to make GFDL's global isopycnal ocean model work
Robert Hallberg
NOAA/GFDL
Robert.Hallberg@noaa.gov
ABSTRACT
GFDL is in the process of building an IPCC-class coupled model using a global version of HIM at 1-degree, 48 layer resolution. In the course of this exercise, a number of algorithmic improvements have proven necessary. This talk will briefly cover several of these; some are brief updates on topics discussed in past LOM workshops. The pressure gradient formulation has been changed to avoid the thermobaric instability that arises in global isopycnal models with high resolution in density space and the full nonlinear equation state. This is the resolution to the problem discussed at the 2003 LOM workshop. An additional source of bottom mixing from the energy extracted by the bottom drag has proven necessary to avoid having thin, viscous plumes of overflow water sink to the bottom of the ocean. The use of a 2-layer refined bulk mixed layer enables shear-driven restratification and Ekman restratifation and destabilization, and may reduce the differences in the mixed layer dynamics between traditional bulk mixed layers and pressure-space hybrid coordinate mixed layers. Changes in the reconsilation of the internal and external estimates of the free surface height may also be described. The simulations generated by this model will be described in a separate talk by B. Arbic.
2004 LOM Workshop Monday 9:35 - 9:55 a.m.
Isopycnal, orthobaric and neutral coordinates
Trevor J McDougall
CSIRO Marine Research
Trevor.McDougall@csiro.au
ABSTRACT
The extent to which isopycnals, patched isopycnals, and orthobaric density surfaces attain the neutral property will be discussed and quantified. This leads into an evaluation of the quasi-non-material aspect of both orthobaric surfaces and neutral density surfaces (the quasi-non-material property causes flow across surfaces even in the absence of diabatic processes).
2004 LOM Workshop Monday 9:55 - 10:15 a.m.
The T/S/rho conservation conundrum in isopycnic models
Rainer Bleck
Los Alamos National Laboratory
bleck@lanl.gov
ABSTRACT
Temperature and salinity fronts in the ocean tend to adjust so as to minimize their joint impact on the buoyancy field. This phenomenon, known as T/S compensation, renders fronts dynamically invisible and long-lived. Numerically imperfect T/S advection in a circulation model can interfere with this T/S compensation. In an isopycnic model, lateral T/S transport errors are particularly troubling because they lead to nonphysical perturbations in layer thickness as the model struggles to restore the target density in affected layers. In MICOM and HYCOM, T/S compensation has traditionally been "built in" by diagnosing T from S & rho, rather than solving a prognostic equation for T. Due to the nonlinearity of the equation of state, this approach implies a small heat gain whenever S diffuses laterally, be it for numerical or physical reasons. Thermodynamic variables in a circulation model are subject to both transport and mixing. While enforcing T/S compensation during transport has clear physical advantages, our past strategy to treat T as a diagnostic variable during lateral mixing was strictly a matter of computational convenience. Experiments to reduce the internal heat source by retaining S and rho as the variables being transported, but evaluating lateral mixing in terms of T and S, have been successful. The resulting blend of S/rho advection and T/S mixing greatly improves the performance of MICOM and HYCOM in long-term climate simulations.
2004 LOM Workshop Monday 10:15 - 10:35 a.m.
Numerical Cabbeling of Dissipative and Monotone Tracer Advection Schemes
Mohamed Iskandarani, Eric Chassignet and Trevor J. McDougall
UM
miskandarani@rsmas.miami.edu
ABSTRACT
We investigate the spurious, and numerically induced, cabbeling caused by the inherent numerical dissipation of tracer advection schemes. This spurious cabbeling is the numerical analog of natural cabbeling, the process by which the nonlinear equation of states leads to denser water when water particles of nearly equal densities, but with different temperature and salt properties, mix. The spurious mixing can be exacerbated by the flux- and slope-limiting mechanisms used to prevent Gibbs oscillations from occuring. The impact of four classical advection schemes (Centered Difference, MPDATA, UTOPIA and WENO) on numerical cabbeling is studied. The experiment consists of advecting two passive tracers, that initially combine to yield a constant density, in a square basin driven by a double-gyre circulation. Most of the spurious cabbeling is caused by the deformational flow field, and occurs primarily in the western boundary current region and the eastward jet. Centered schemes with no numerical dissipation can lead to the production of negative anomalies, whereas only positive density anomalies are seen with dissipative schemes like WENO and UTOPIA. At coarse resolution, numerical cabbeling is most significant in deep layers with weak stratification, irrespective of the numerical scheme used.
2004 LOM Workshop Monday 10:50 - 11:10 a.m.
Some Idealized Thermobaric Solutions
Roland de Szoeke and Scott Springer
Oregon State University
szoeke@coas.oregonstate.edu
ABSTRACT
We discretized the equations of motion and thermodynamics, written in terms of orthobaric specific volume (reciprocal of density) as vertical coordinate, following the principles of Hsu and Arakawa (1990), so that no spurious forces or energy sources are created. Because orthobaric sp. vol. is not materially conserved, even in the absence of diffusion and friction, there are unavoidable, but thermodynamically reversible, mass fluxes across orthobaric isopycnals (or discrete layer interfaces), because of the thermobaric character of seawater, coupled with geographic variation of the ocean's T-S relation. We will show some idealized test solutions designed to highlight thermobaric effects and T-S variation. First of all, we show rest states (stable, level in situ and orthobaric isopycnals) in which potential density isopycnals exhibit quite bizarre behavior. We show time-dependent, nonlinear solutions, such as solitary waves and hydraulic jumps, in which the essential nonlinearity is supplied by the thermobaricity of the equation of state. Variants of such behavior can be demonstrated in both the low-frequency planetary-wave realm, and in the ultra-Coriolis frequency band of internal gravity waves. The thermobaric behavior is evident even in simple two-layer idealizations. These tests should be useful benchmarks for numerical model codes that include thermobaric effects in their equations of state.
2004 LOM Workshop Monday 11:10 - 11:30 a.m.
A Global Ocean Model Based on Icosahedral-Hexagonal Grids With a Hybrid Vertical Coordinate
Todd Ringler, David Randall, John Baumgardner, Don Stark, Bert Semtner
Colorado State University
todd@atmos.colostate.edu
ABSTRACT
This talk will present results from a new global ocean general circulation model (OGCM). The ocean model uses an icosahedral-hexagonal grid to tile the surface of the sphere and hybrid (floating) coordinates to discretize the vertical depth of the ocean. The use of icosahedral-hexagonal grids leads to a highly uniform and isotropic discretization of the sphere and eliminates problematic grid singularities found in other grid systems. The hybrid coordinate used to discretize the vertical direction spans the limits between an Lagrangian-coordinate, such as isopycnal coordinates, and an Eulerian-coordinate, such as z-level coordinates. As a result, the model can be integrated forward in time for several days using Lagrangian vertical coordinates to minimize vertical diffusion and dispersion. At fixed time intervals, the vertical layers are mapped back to z-level coordinates and the integration continues forward in time. This type of coordinate is referred to as an Arbitrary Lagrangian Eulerian coordinate and is being developed at LANL. The model contains much of the functionality of a full production ocean model including convective adjustment, implicit vertical mixing, UNESCO equation of state, realistic bathymetry, and sub-grid scale mixing parameterizations. The results show realistic ocean circulations using 40962 grids cells in the horizontal (nominal 1 degree grid) and 33 layers in the vertical. The ocean is forced using monthly mean NCEP wind stresses and restoring to Levitus temperature and salinity in the top 5 meter layer. The results demonstrate that not only is this a viable approach to ocean modeling, but also provides an alternative approach to overcome many of the classical problems in ocean modeling.
2004 LOM Workshop Monday 11:30 - 11:50 a.m.
Toward the Implementation of Ripa's Inhomogeneous Layer Model
F. J. Beron-Vera, M. J. Olascoaga and J. Zavala-Garay
RSMAS/AMP, UMiami, 4600 Rickenbacker Cwy., Miami, FL 33149 USA
fberon@rsmas.miami.edu
ABSTRACT
Inhomogeneous layer models in which the velocity and buoyancy fields are allowed to vary only in the horizontal position and time have been very extensively exploited in ocean modeling. One example is the widely used Miami Isopycnic-Coordinate Model, whose upper layer is chosen as an inhomogeneous layer of this kind. These so-called "slab" models have the ability to partially incorporate thermodynamic processes, which are of fundamental importance in the ocean. For instance, in addition to momentum fluxes, these models can accommodate nonuniform heat and freshwater fluxes through the ocean surface. However, the slab models are known to have several limitations and deficiencies. In particular: (i) they cannot represent explicitly the thermal-wind balance which dominates at low frequencies; (ii) they have a zero-frequency mode not present in the exact fully three-dimensional model; and, in close relation to this, (iii) they cannot prevent spurious instabilities from developing. To cure the slab model limitations and deficiencies, Ripa proposed an improved closure to incorporate thermodynamic processes in a one-layer model. In addition to allowing arbitrary velocity and buoyancy variations in horizontal position and time, Ripa's model allows the velocity and buoyancy fields to vary \emph{linearly} with depth. Ripa's model enjoys a number of properties which make it very promising: (i) it represents explicitly the thermal-wind balance at low frequencies; (ii) the free waves supported by the model (Poincaré, Rossby, midlatitude coastal Kelvin, equatorial, etc.) are a very good approximation to the first and second vertical modes in the fully three-dimensional model; and, very importantly, (iii) in the absence of dissipation and external forcing, Ripa's model has a general invariant, quadratic in the departure from a state of rest (or at most with a uniform current), which is positive definite. This property, which is present in the fully three-dimensional model, prevents the system to explode by itself, unlike the slab models for which this integral of motion is nonnegative definite. In this work we generalize Ripa's model to an arbitrary number of layers, including the possibility of a free surface and irregular bottom topography or the (mathematically equivalent) case in which the stack of layers floats on top of a quiescent infinitely deep layer. As a test we consider the problem of ageostrophic upper-ocean baroclinic instability, by allowing for the bottom boundary in the classical Stone's model to move freely.
2004 LOM Workshop Monday 11:50 a.m. - 12:10 p.m.
The Slippery Sacks Ocean Model
Patrick Haertel
University of North Dakota
haertel@aero.und.edu
ABSTRACT
Over the past several years I have been developing a new Lagrangian ocean model. The model is based on the slippery sacks (SS) numerical method, under which a fluid is represented as a pile of conforming sacks. During this talk I will introduce the SS method and present tests in which SS simulations are compared to analytic solutions and simulations carried out with other models. The simulated circulations will range in complexity from idealized waves to the three dimensional flow in a large lake. I will discuss the numerical accuracy, the computational efficiency (for both single- and multi-processor simulations), and the advantages and disadvantages of the SS model.
2004 LOM Workshop Monday 1:20 - 1:40 p.m.
An improved vertical coordinate of the potential density type
John Dukowicz
Los Alamos National Laboratory
duk@lanl.gov
ABSTRACT
There are two types of quasi-neutral density functions suitable for use as vertical coordinates: Functions of density and pressure, such as orthobaric density, and functions of potential temperature and salinity, such as potential density. In this talk I will discuss two functions of the potential density type that improve on potential density as a materially-conserved, quasi-neutral vertical coordinate.
2004 LOM Workshop Monday 1:40 - 2:00 p.m.
HYCOM code development
Alan Wallcraft
NRL
Alan.Wallcraft@nrlssc.navy.mil
ABSTRACT
HYCOM is one of the generalized vertical coordinate ocean models that is a precursor to HOME (Hybrid Ocean Modeling Environment). The performance and features of HYCOM 2.2 will be described, as will plans for future development and preparation for HOME.
2004 LOM Workshop Monday 2:00 - 2:45 p.m.
HOME overview, plans, and best-practice studies
Robert Hallberg, R. Bleck, E. Chassignet, R. deSzoeke, S. Griffies, P. Schopf, A. Wallcraft
NOAA/GFDL
Robert.Hallberg@noaa.gov
ABSTRACT
This talk will be an overview of the Hybrid Ocean Modeling Environment (HOME). HOME will provide a versatile community open-source, ocean modeling environment using a predominantly Lagrangian vertical coordinate. The HOME development effort will also identify and refine best practices or describe trade-offs between alternatives for simulating a range of important ocean processes. Many of the principle developers of HYCOM, HIM, Poseidon, POSUM are actively engaged in the HOME effort. While we will maintain much of the existing algorithmic diversity within HOME, it is anticipated that new development and new model applications from each of these groups will be based primarily upon the HOME code within the next 3 years. HOME is intended to be a community model in the fullest sense, and we will strive to facilitate community participation in and contributions to the HOME development effort. This talk will provide an overview of HOME, the rationale behind it and the synergies that will emerge from it. Both the near term plans for HOME and the long-term prospects for HOME to act as a catalyst of broader collaborations with the users and developers of other classes of ocean models will be discussed. A particular emphasis will be placed on plans to conduct a broad series of "best-practice" studies for Lagrangian vertical coordinate ocean models.
2004 LOM Workshop Monday 3:15 - 3:45 p.m.
HOME Architecture and Structure
Paul Schopf, Rainer Bleck, Eric Chassignet, Robert Hallberg, Roland deSzoeke, Alan Wallcraft
George Mason University
schopf@scs.gmu.edu
ABSTRACT
This talk will discuss the architectural problems confronting HOME. As an environment designed to replicate the precursor models while being built upon technologies such as the Earth System Modeling Framework (ESMF), a careful analysis of the extant codes, the ESMF standards, and desires for future functionality will need to be undertaken. This talk will discuss the currently- known problems confronting the team and the emerging strategies to solve them.