3rd workshop on Physics Dynamics Coupling (PDC18) - Abstracts (Oral)

Representation of air-sea interactions on an idealized coupled atmosphere-ocean model with focus on the Western Baltic Sea

Tobias Bauer (Leibniz Institute for Tropospheric Research - TROPOS)


Air-sea interactions in Earth system models are generally covered by the boundary conditions of the atmosphere and ocean components. The main idea of this strategy is to utilize each component of an Earth system model independently. A coupler usually controls the data exchange between the components.

In order to understand the physics at the air-sea interface, an idealized coupled atmosphere-ocean model with focus on the Western Baltic Sea is developed. At first, simplified independent atmosphere and ocean models are created. Furthermore, those are combined and the air-sea interactions, such as heat exchange, evaporation, precipitation as well as transfer of momentum, are considered within the set of equations.

The presentation will start with an overview of the prognostic and diagnostic variables of the model. Moreover, the physical representation of the air-sea interactions within the coupled idealized model will be shown as well as how they are numerical realized.

Understanding couplings between the boundary layer and the large-scale dynamics

Robert Beare (University of Exeter)


The couplings between the boundary layer, moist convection and the large-scale dynamics form a central part of the physics-dynamics coupling problem. Ekman pumping is a common explanation of how the boundary layer interacts with the dynamics. However, it is less common to place Ekman pumping fully in context of the time-evolving dynamics. Here we describe a theory (a modification of semi-geostrophic theory) that allows us to understand more fully the interaction of the Ekman layer with the interior dynamics. We give examples of applying the theory to frontal jets, baroclinic waves, and coupling to convectively-driven circulations. We also show how the theory can expose deficiencies in the time-stepping of the boundary-layer parametrization.

Reduced numerical precision guided by physics-dynamics coupling

Matthew Chantry (Oxford University)


The impact of stochastic elements in physics-dynamics coupling has now been leverage by many weather centres running ensemble forecasts. The improvements they bring highlight the imperfections in both the physics and dynamics, introduced by unresolved lengthscales and processes. Resolving these elements requires computational power unavailable at this time. Despite the uncertainty of these schemes, high numerical precision is used to calculate and store the necessary computations. Driven by machine learning, where high numerical precision is generally wasted, emerging hardware now supports half-precision floats, four times smaller than the double-precision standard. In light of the stochastic element to the physics-dynamics coupling we ask whether double precision computations are necessary to produce a high quality forecast. We examine a hierarchy of models, finishing with ECMWF’s IFS, to investigate this question. Of particular interest is whether the regions or lengthscales most impacted by the physics, and thus the most stochastically perturbed, can be calculated at reduced precision. The computational costs saved by these reductions can be reinvested in higher resolutions or further model complexity.

Physics-dynamics coupling experiments in the UM

Terry Davies (Met Office)


To increase possible task parallelism in the UM experiments are being conducted to finds the limits (resolution and time step) to running all the individual parametrizations in parallel. In the UKV (1.5km resolution, 1  minute time step and no deep convection) it appears that all the physics can be run in parallel with no loss of skill. However lower resolution global models fail after a few hours. Is this because deep convection cannot run in parallel or is there a limit on the time step?

Parallel Physics-Dynamics Coupling in an Atmosphere Model

Aaron Donahue (Lawrence Livermore National Laboratory)


One of the major goals of Earth System Model development is to design computational paradigms that are best suited to take advantage of the next generation of high performance super computing platforms.  As the computing community drives forward towards exascale machines it is our responsibility to make sure our models are positioned to exploit these new and more powerful machines.  This talk will discuss one strategy that shows great promise for extending the scalability of earth system models by separating the physics and dynamics calculations onto separate computational cores and running in parallel.

This talk will discuss preliminary results from application of this physics/dynamics coupling approach implemented in the DOE Energy Exascale Earth System Model (E3SM).  Through the proper balance of computational cores devoted to dynamics solved and cores devoted to physics cores it is possible to accomplish improved performance at both low and high computational core count simulations over the current default coupling mechanism in the model.  There are, however, stability concerns that arise from adopting a parallel physics/dynamics coupling approach which require careful attention.

Variational-metriplectic formulations of compressible, multiphase, multicomponent geophysical fluids

Christopher Eldred (AIRSEA, INRIA/Laboratoire Jean Kuntzmann)


The equations of reversible (inviscid, adiabatic) fluid dynamics have a well-known geometric description in terms of a Lagrangian formulation, and its associated Hamiltonian formulation (see, for example, [1,2]). This structure underlies many of the most basic principles that we know about geophysical fluid flows, such as conservation laws. Structure-preserving numerical models can be developed that emulate these characteristics utilizing either the Hamiltonian [3,4,5,6] or Lagrangian [7] formulations. However, real geophysical fluids (and numerical models for them) must also incorporate irreversible processes. Recent work [8,9] has demonstrated that the Lagrangian variational approach can be extended to include irreversible processes (such as dissipation, diffusion and phase changes). This variational approach is based on the new concept of thermodynamic displacement, and by design, it inherently satisfies both the 1st and 2nd law of thermodynamics. The corresponding Hamiltonian-type formulation is a metriplectic formulation [10], which combines a Hamiltonian structure for the reversible dynamics with a metric structure for the irreversible dynamics. This talk will present the metriplectic formulation of fully compressible, multicomponent geophysical fluids, associated to the variational formulation. The work of Almut Gassmann [11] in fact implicitly contained such a structure, and it is shown that the metriplectic formulation presented here is a generalization of her approach. The end goal of this work is a mathematically consistent set of equations to describe turbulence-averaged, multicomponent, multiphase geophysical fluid flows with irreversible processes. Some limitations of the current approach (especially with regards to precipitation and nonlocal processes such as convection) in achieving this goal and ideas to overcome them will be discussed.

[1] G. Badin and F. Crisciani. Variational Formulation of Fluid and Geophysical Fluid Dynamics: Mechanics, Symmetries and Conservation Laws (Advances in Geophysical and Environmental Mechanics and Mathematics). Springer, 2017.

[2] R. Salmon. Lectures on Geophysical Fluid Dynamics. Oxford University Press, 1998.

[3] T. Dubos, S. Dubey, M. Tort, R. Mittal, Y. Meurdesoif, and F. Hourdin. Dynamico-1.0, an icosahedral hydrostatic dynamical core designed for consistency and versatility. Geoscientific Model Development, 8(10):3131–3150, 2015.

[4] A. Gassmann. A global hexagonal c-grid non-hydrostatic dynamical core (icon-iap) designed for energetic consistency. Quarterly Journal of the Royal Meteorological Society, 139(670):152–175, 2013.

[5] R. Salmon. Poisson-bracket approach to the construction of energy- and potential-enstrophy-conserving algorithms for the shallow-water equations. Journal of the Atmospheric Sciences, 61(16):2016–2036, 2004.

[6] M. Tort, T. Dubos, and T. Melvin. Energy-conserving finite-difference schemes for quasi-hydrostatic equations. Quarterly Journal of the Royal Meteorological Society, 141(693):3056–3075, 2015

[7] M. Desbrun, E. S. Gawlik, F. Gay-Balmaz, and V. Zeitlin. Variational discretization for rotating stratified fluids. Discrete and Continuous Dynamical Systems - A, 34:477, 2014

[8] F. Gay-Balmaz. A variational derivation of the thermodynamics of a moist atmosphere with irreversible processes. ArXiv e-prints, Jan. 2017.

[9] F. Gay-Balmaz and H. Yoshimura. A lagrangian variational formulation for nonequilibrium thermodynamics. part ii: Continuum systems. Journal of Geometry and Physics, 111:194 – 212, 2017.

[10] P. J. Morrison. Thoughts on brackets and dissipation: Old and new. Journal of Physics: Conference Series, 169(1):012006, 2009.

[11] A. Gassmann and H.-J. Herzog. How is local material entropy production represented in a numerical model? Quarterly Journal of the Royal Meteorological Society, 141(688):854–869, 2015.

On the coupling of the compressible and incompressible Navier-Stokes equations

Fatemeh Ghasemi (Linköping University)


Understanding the ocean-atmosphere coupling is critical for predicting changes in global temperature patterns and global warming. This multiphase coupled problem can be handled with three different methodologies: one can treat both phases as compressible, both phases as incompressible or the gas (atmosphere) as compressible and the liquid (ocean) as incompressible.

The fully compressible model is limited to liquids for which there are acceptable models for their compressible evolution. The fully incompressible model is valid as long as the density variations in the gas phase can be neglected. These two models can easily be coupled since the same mathematical model is used. The compressible-incompressible coupling is more realistic and general, but more complicated than the previous ones, since the mathematical models are different.

In this paper, we consider the last approach and aim for interface conditions such that the coupled problem is well-posed and conservative. Next, we discretize using high order finite differences on summation-by-parts form and include the derived coupling conditions weakly. Finally, the requirements for discrete conservation and stability will be determined.

Across the ‘grey zone’ of ocean model resolutions

Helene Hewitt (Met Office Hadley Centre)


As the horizontal resolution of global ocean models increases, the issue arises of how to deal with the grey zone where mesoscale features are present but not fully resolved. The impacts (both positive and negative) of increasing ocean resolution are reviewed. This includes the regime of parameterised mesoscale eddies (~1 degree) to a regime of resolved mesoscale eddies (~1/12 degree) via the so-called grey zone (~1/4 degree). Even in the grey zone, as we move from a viscous to an inertial regime, the representation of western boundary currents is significantly improved relative to lower resolution. This has implications for both downstream circulation and coupling with the atmosphere. As resolution increases beyond the grey zone, smaller scale features in the bathymetry are also resolved which impacts on the steering of the circulation with potential feedbacks on sea surface temperature. In contrast, the impact of the grey zone is very apparent in interior heat and salt budgets highlighting that mesoscale features are neither sufficiently resolved nor parameterised. This deficiency has implications for the drifts of ocean water masses away from their initial conditions. Approaches to dealing with the grey zone are discussed.

Analyzing physics-dynamics coupling in an ensemble of simplified GCMs

Christiane Jablonowski (University of Michigan)


The paper will discuss the results of the 2016 Dynamical Core Model Intercomparison Project 2016 (DCMIP-2016) that paired the newest generation of nonhydrostatic dynamical cores with simplified physical processes, and in particular a simple warm-rain precipitation parameterization (a so-called ‘Kessler’ scheme). In addition, the models can also be paired with an even simpler ‘large-scale condensation’ precipitation parameterization.

Twelve modeling groups participated in DCMIP-2016. The idealized DCMIP test cases included a baroclinic wave (with dry and moist configurations), a tropical cyclone and the evolution of a supercell on a small Earth. The paper provides an in-depth discussion of the baroclinic wave intercomparison and pays special attention to the impact of moisture on the evolution of the baroclinic wave. An ensemble of dry and moist simulations will be assessed. The paper thereby sheds light on the physics-dynamics interplay and coupling strategies in the DCMIP models.

Improving climate model coupling through complete mesh representation

Robert Jacob (Argonne National Laboratory)


The U.S. DOE Climate Model Development and Validation program is increasing the capability of the current coupler in E3SM by replacing many of the communications and mapping functions from the Model Coupling Toolkit [1] with near-equivalent functions from the Mesh Oriented dAtaBase (MOAB) library [2].  Unlike MCT, MOAB allows a complete description of the numerical grid (mesh) used by each submodel in E3SM along with serialization of the discrete coupled solution fields, which in turn are utilized to compute the mesh intersection, interpolation and remapping weights through conservative schemes.  Computing the remapping weights through files loaded from disk [3, 4] presents several hurdles in terms of the scientific workflow and productivity.  Utilizing Fortran compatible interfaces in MOAB to directly link the submodels in E3SM with fully online coupling removes workflow bottlenecks and exposes opportunities for improving parallel scalability with minimal modifications. In this talk, we present some of the recent and ongoing developments in E3SM with the MOAB-TempestRemap based coupling interface and describe the workflow flexibility that can be leveraged by unifying the mesh and remapping descriptions.

[1] Larson, J. Walter, et al. "The model coupling toolkit." International Conference on Computational Science. Springer, Berlin, Heidelberg, 2001.

[2] Tautges, Timothy James, et al. MOAB: a mesh-oriented database. No. SAND2004-1592. Sandia National Laboratories, 2004.

[3] Hill, Chris, et al. "The architecture of the earth system modeling framework." Computing in Science & Engineering6.1 (2004): 18-28.

[4] Ullrich, Paul A., and Mark A. Taylor. "Arbitrary-order conservative and consistent remapping and a theory of linear maps: Part I." Monthly Weather Review 143.6 (2015): 2419-2440.

Coupling isobaric physics with isochoric dynamics

Kohei Kawano (Japan Meteorological Agency)


The Japan Meteorological Agency has been operating a regional nonhydrostatic model which employs constant height-based coordinate. The model couples physical processes that used to be adopted by the former operational models with hydrostatic assumption and pressure-based vertical coordinate. In converting tendencies evaluated by physics to those of prognostic variables of the dynamical core, differences of some assumptions between the dynamics and the physics have to be taken into account. Because the dynamics core adopts the finite volume method and employs total density as a prognostic variable, it is the simplest way to assume that total density in each cell is kept constant. With this coupling method, latent heat released by the microphysics process makes the pressure increased in constant volume cells, and then the local high pressure is mitigated through the following dynamics steps, which often predicts unphysical strong updrafts. Since this representation of the process seems to be unrealistic, we have just started to explore ways to achieve the consistency in the coupling, such as incorporating the change of cell volume in physical processes.

 In the presentation, we will talk about our recent attempt on the issues

Convergence and performance aspects of physics-dynamics coupling in US-DOE research

Dorothy Koch (US Department of Energy)


We present an overview of activities within the US-Department of Energy research portfolio to address physics-dynamics coupling, with particular attention to improving solution convergence and to optimizing model performance on multiple high-performance computer platforms.  Current activities connected to the Energy Exascale Earth System Model (E3SM) span the Earth system. For the atmosphere, we will highlight efforts to run physics and dynamics in parallel in order to improve model performance without degrading accuracy; strategies to improve physics solution convergence; partitioning processes to utilize heterogeneous computational architectures; and verification efforts to assure the correctness of our implementations.  For the ocean, we will highlight efforts to adjust solution for variable resolutions, including time-stepping approaches and treatments of eddies. For the cryosphere, we will discuss new sea-ice dynamics modeling methods appropriate for high resolution, variable and adaptively refined meshing strategies and scalable solution algorithms for land ice, and overall computational performance strategies. For land, we will present early work on process coupling and biogeochemistry solution convergence.  Finally, we will present new approaches for Earth system coupling that will improve solution accuracy as well as computational performance. We will conclude with some common themes, priorities and programmatic strategies.

Time integration methods and dynamic-physics coupling

Oswald Knoth (Leibniz Institute for Tropospheric Research)


From a mathematical point of view dynamical and physical processes in numerical weather models constitute a unique mathematical model. Therefore all individual processes should be part of a consistent approximation of the mathematical model in space and time. Following a method of lines approach a large system of ordinary differential-algebraic equations has to be solved in time in a second final step after spatial discretization.

We will review common practice in weather range forecast models with often sophisticated dynamical core time integration schemes and a brute-force coupling to other "physical" processes. These physical processes by itself are transport processes  like vertical and horizontal diffusion, convection parameterizations, surface fluxes and microphysical processes. Some of these forcings are time step dependent and the convergence of the whole scheme is doubtful in case of decreasing time steps.

We will propose smooth physical process modeling and time integration schemes which may better suited in this context. Multirate time integration methods with implicit stages or Rosenbrock-W-methods are possible classes of algorithms. An exemplary implementation of these methods in the atmospherical code ASAM  is presented and illustrated with numerical examples.

Physics coupling with the Finite-Volume Module of the IFS

Christian Kuehnlein (ECMWF)


The Finite-Volume Module of the IFS represents an alternative dynamical core module that complements the operational spectral-transform dynamical core of the IFS with new capabilities, such as a compact-stencil finite-volume discretisation, conservative non-oscillatory advective transport and nonhydrostatic governing equations.

The presentation highlights the current state of development of FVM at ECMWF.  The emphasis will be on aspects concerning the coupling of the FVM semi-implicit integration to the IFS physics.

Physics-dynamics coupling with element-based high-order Galerkin methods: quasi equal-area physics grid

Peter H. Lauritzen (National Center for Atmospheric Research (NCAR))


Traditionally the state of the atmosphere passed to the sub-grid-scale parameterizations (also referred to as physics) for models based on finite-volume and finite- difference methods has been the cell-averaged state in each control volume and the grid-point value, respectively. For the regular latitude-longitude, cubed-sphere and icosahedral grids the distance between the grid-points is gradually varying for finite-volume/finite-difference discretizations. If the same physics-dynamics coupling paradigm is applied to high-order element-based Galerkin methods, the state of the atmosphere passed to physics would be evaluated at the quadrature points. In the case of spectral-element method these are the Gauss-Lobatto-Legendre (GLL) quadrature points. Having the physics and dynamics grids coincide is obviously convenient since no interpolation is needed (which could disrupt conservation properties) and the number of degrees of freedom on both grids is exactly the same. That said, a unique aspect of the high-order quadrature rules is that the nodes within an element are not equally spaced; in fact, the higher the order of the quadrature rule the less equi-distant are the quadrature points. GLL quadrature points cluster near the edges and, in particular, the corners of the elements.

In this talk we present a version of the NCAR’s Community Atmosphere Model (CAM) using the spectral-element (SE) dynamical core in which we have separated the physics and dynamics (GLL) grids so that the state passed to physics is an integral of the state over quasi-equal area control volumes rather than non-equidistant GLL point values.

On the spatial and temporal discretization of vertical diffusion in the turbulent planetary boundary layer

Florian Lemarié (INRIA, University Grenoble-Alpes)


It is well known that model behavior can be very sensitive to changes in time-step and spatial resolution [1]. In particular such dependency can be observed for turbulent planetary boundary layer schemes when running standard single column experiments (e.g. for the Kato & Phillips, Willis & Deardorff, or GABLS1 experiments). In this talk we propose to investigate if part of this sensitivity could be attributed to inadequate spatial and temporal discretizations of the nonlinear vertical diffusion term in state-of-the art oceanic and atmospheric numerical models. In particular the usual second-order centered discretizations are generally not suitable for problems characterized by boundary layers where the gradients of the solution are large. Various finite- difference and finite-volume alternatives, with similar stencil as the second-order scheme, are investigated. As far as the temporal discretization is concerned, the very large vertical parabolic Courant numbers usually found in numerical simulations make it hard to formulate schemes that are robust to changes in model parameters since some degree of implicitness is required. For large parabolic Courant numbers and implicit Euler scheme, it can be shown that the amount of diffusion we would expect from physical principles (i.e. as diagnosed by the PBL scheme) is very different from the amount of diffusion actually ”seen” by the model because of numerical errors [2]. Alternatives proposed by [3] and [4] could help to mitigate this issue even if they have been derived mainly with nonlinear stability constraints in mind. We will conclude by discussing how a finite volume approach for the vertical diffusion term could allow to jointly ensure the proper regularity of the numerical solutions while satisfying the underlying physical principles (e.g. the Monin-Obukhov theory). Numerical results using standard single column experiments are shown to illustrate our findings for 0-equation and 1-equation turbulent schemes.


[1] Gross M. et al. (2018) Recent progress and review of Physics Dynamics Coupling in geo- physical models. Mon. Wea. Rev., in revision

[2] F. Lemarié, L. Debreu, G. Madec, J. Demange, J.-M. Molines, and M. Honnorat Stability constraints for oceanic numerical models: implications for the formulation of time and space discretizations. Ocean Model. (2015)

[3] Wood N., M. Diamantakis and A. Staniforth A monotonically-damping second-order- accurate unconditionally-stable numerical scheme for diffusion. Q. J. Roy. Met. Soc. (2007)

[4] Nazari F., A. Mohammadian, A. Zadra, and M. Charron. A stable and accurate scheme for nonlinear diffusion equations: application to atmospheric boundary layer. J. Comp. Phys. (2013)

An efficient integrated dynamics-physics coupling strategy for global cloud-resolving models

Shian-Jiann Lin (NOAA/Geophysical Fluid Dynamics Laboratory)


The FV3 group at the NOAA/Geophysical Fluid Dynamics laboratory is developing a new type of Global Cloud-Resolving Model (GCRM) based on an integrated dynamics-physics concept, in which several fast-acting physics (e.g., cloud microphysics) are incorporated into a new FV3 (Finite-Volume Dynamical Core on the Cubed-sphere) framework. This new modeling framework improves the dynamics-physics interaction and increases the overall computational efficiency due to the separation of the fast-acting physics from the slow-physics, allowing a near tenfold increase in overall time step. We have also built some of the SubGrid Orographically (SGO) forced processes into the new FV3 dynamics, which unavoidably breaks the traditional boundary between "dynamics" and "physics". We believe the boundary between the "dynamics" and "physics" set by the traditional modeling framework is one reason that limits modeling advancements in the past few decades. 

A preliminary version of this new type of GCRM is used for the DYAMOUD project.  We will carry out several 40-day "convective-parameterization-free" experiments across the gray-zone at three different horizontal resolutions: 13, 6.5, and 3.25 km. As a potential tool for sub-seasonal predictions, we shall analyze the forecast skill (first 10 days) as well as the systematic "climate basis" for the last 30 days.

Exploring the impacts of stochastic representations of model uncertainties

Sarah-Jane Lock (ECMWF)


ECMWF ensemble forecasts include stochastic representations of model uncertainties that are attributed to the parametrisation of atmospheric physics processes. Significant forecast skill is derived from the inclusion of the Stochastically Perturbed Parametrisation Tendencies (SPPT) scheme. As its name suggests, the scheme works by introducing random perturbations to the tendencies from the atmospheric physics schemes. Recent work has focussed on an alternative approach – the Stochastically Perturbed Parametrisations (SPP) scheme – which acts by randomly perturbing quantities within the physics schemes, thereby taking the representation of model uncertainty closer to its known sources.

We use model tendency outputs from the dynamics and physics components to explore the impact of the stochastic perturbations. In this talk, we will focus on the interactions between the physics and dynamics tendencies, and how they are affected by stochastic perturbations applied to different physical processes.

Introducing net mass transport in the mass flux parametrisation of convection of the IFS

Sylvie Malardel (ECMWF)


In the mass flux formulation used in the IFS convection parametrisation, the subgrid vertical transport of air in a buoyant convective updraft is systematically compensated in term of mass, inside each grid box, by a subsiding flux of air with the thermodynamics and momentum characteristics of the environment. Thus, the subgrid transport in the parametrisation of convection does not result in a net transport of mass in a grid box, or, in other words, the convection scheme does not produce any direct tendency of hydrostatic pressure in a grid box.

The original idea of Kuell et al (2007) has been adapted to the hydrostatic IFS in order to allow a net mass transport by the convection parametrisation and delegate the compensating subsidence, which is necessary to close the continuity equation, to the advection scheme in the dynamics. With this new formulation, the compensating subsidence is not constrained to take place in the same vertical column as the updraft. The proposed formulation is expected to improve the interaction between the convection and the large scale circulations on shorter time scales and in the grey zone of convection where convective updraughts are still insufficiently represented by the resolved flow.

Preliminary results for academic test cases and "operational'' IFS configurations will be shown and the pro and con of the new method will be discussed.

Scalar conservation mapping in physics dynamics coupling (PDC)

Timbwaoga Ouermi (University of Utah)


Physics dynamics coupling (PDC) is a procedure that lies at the heart of geophysical models. While traditional coupling approaches rely on an assumption of a common physics and dynamics grid, the advent of higher-order numerical methods such as spectral elements, may require the use of different computational grids.  Therefore, an interpolation algorithm must be used to transfer the modeled quantities between the two grids. It is vital to have interpolation algorithms that are consistent and preserve certain properties, such as positivity. The numerical approaches employed for mapping between physics and dynamics may introduce errors and unphysical values, such as negatives for a positive-definite quantity, which may degrade the dynamics and physics calculations. For this reason, a new arbitrary order provably positivity preserving mapping for the interpolation between physics and dynamics is introduced. Numerical results demonstrate both accuracy, and positivity. This method adaptively interpolates physical values using polynomial of degree up to n with n+1 given points. A comparative study between different methods for conservation of density, potential temperature, and energy is made in the context of the US Navy spectral element code NEPTUNE.

Evaluating time step and resolution sensitivities in the GEOS analysis and forecast system

Bill Putman (NASA)


The GEOS model is used to study the sensitivity of cloud forcing, microphysical processes and NWP forecast skill due to changes in the physics/dynamics coupling timestep and horizontal resolution. The GEOS model is run in ‘reforecast mode’ at three uniform global resolutions of 13-km (c720), 25-km (c360) and 50-km (c180) with 72 vertical levels up to 0.01mb starting from our production 13-km global analysis system.  Our physics/dynamics coupling timestep in production systems with GEOS has been fixed at 450 seconds in the MERRA-2 50-km reanalysis, the 25-km former production analysis/forecast system and the recently upgraded 13-km production version of GEOS. As we pursue higher resolution and non-hydrostatic regimes with GEOS, this coupling frequency has been shortened to as fine as 60s. We will explore the range of coupling frequency from 450s to 60s at 50-, 25- and 13-km resolutions.

The model physics uses the finite-volume cubed-sphere (FV3) dynamical core and the Grell-Freitas scale-aware convection scheme to dynamically reduce the role of parameterized deep convection as resolved scale processes in the model take over at higher resolutions.  The microphysics consists of options for single moment cloud microphysics (Bacmeister et al, 2006) and two moment Morrison-Gettleman-Barahona cloud microphysics.  Options to explore the GFDL microphysics are available in GEOS and developments from GFDL are pursuing a more tightly coupled version of these microphysics within the FV3 dynamical core. Should this more tightly coupled implementation become available in time our experiments will be repeated with this option.

Assessing and improving the numerical solution of atmospheric physics in an Earth System model

Philip Rasch (Pacific Northwest National Laboratory)


Physics parameterizations in atmosphere general circulation models are known to be “noisy” in time compared to the behavior of the dynamical core, even when the parameterizations are formulated with deterministic partial differential equations instead of as stochastic processes. Is this the expected behavior or an indication of numerical artifacts? If the latter is true, what are the impacts and how can the numerical solutions be improved? This presentation will introduce our efforts on developing systematic methods to evaluate and improve the accuracy and convergence of numerical solution of the atmospheric physics parameterizations, in particular those related to clouds, in the Energy Exascale Earth System Model (E3SM) of the US Department of Energy. The objectives of our efforts are to (1) understand the cause of recently revealed strong time-step sensitivity and poor time-step convergence in the E3SM atmosphere model, (2) develop new time-discretization and process-integration methods to improve convergence and accuracy in the physics parameterizations related to clouds, and (3) provide an assessment of the feasibility of using a stochastic perspective to develop time-integration methods for complex atmospheric processes. A hierarchy of models with differing levels of complexity is used in our investigation, which facilitates the analysis from a mathematical perspective using the theories of deterministic and stochastic differential equations.

Reduced complexity frameworks for exploring physics dynamics coupling sensitivities

Kevin Reed (Stony Brook University)


The use of idealized model configurations has had a long history in the understanding of the atmosphere. As global atmospheric models become more complex (i.e., higher resolution, shorter time steps, improved parameterizations, higher-order dynamics packages), the use of idealized modeling for process studies remains as vital as ever. This work presents a hierarchy of reduced complexity testbeds that have been used to explore model sensitivities of precipitation processes at reduced computational expense. The role of physics dynamics coupling, in particular the choice of the coupling frequency, at high horizontal resolution and its impacts on circulations and precipitation processes are explored as they represent large uncertainties in current-generation global models. The National Center for Atmospheric Research’s Community Atmosphere Model (CAM) is configured in various configurations, including moist bubble, radiative-convective equilibrium and aquaplanet setups, to investigate the simulation of organized convection and precipitation at next-generation horizontal resolutions (< 30 km grid spacing globally) for climate-scale modeling.

Outcomes from the PDC sessions at the Second Pan-GASS meeting

Ben Shipway (Met Office)


The Second Pan-GASS meeting 'Understanding and Modelling Atmospheric Processes', took place February 2018 in Lorne, Victoria, Australia.  This meeting brought together approximately 160 scientists whose primary mission is to develop numerical models of the atmosphere for climate research and NWP.  At this years' meeting, a plenary session and a break-out discussion session were dedicated to issues of physics-dynamics coupling. In particular, the break-out session sought ideas and interest for a potential international model intercomparison project that might benefit the community in both understanding and improving physics-dynamics coupling strategies.  Interest in this was very strong and we will report back on the some of the ideas with a view to further discussion at PDC2018.

A multi-fluid approach for the representation of convection

John Thuburn (University of Exeter)


A new approach for the representation of convection in numerical models is proposed, motivated by the potential to address several aspects of physics-dynamics coupling, including the dynamical memory of convection, the horizontal location of compensating subsidence, the horizontal propagation of convective systems, and the feedbacks that help control the intensity of convection. The governing equations are derived systematically by `conditional filtering'; they resemble the equations used in modelling of multi-phase flow, and have attractive conservation and normal mode properties. The approach may be considered a natural extension of the mass flux approach for representing convection, in which the same prognostic quantities (volume fraction, density, momentum, moisture, and entropy) are retained in the updrafts as in the environment. The principal novelty of the new approach is that it permits the large-scale-average dynamics of the convective updrafts to be handled by the dynamical core, rather than parameterized. In other words, the traditional but artificial boundary between `physics' and `dynamics' is shifted. Processes such as entrainment and detrainment, however, must still be parameterized. The new approach has been tested in a single column model of the dry convective boundary layer, with parameterizations for eddy diffusion and entrainment based on those used in eddy-diffusivity mass-flux (EDMF) schemes. Some care is needed with the numerical methods to ensure adequate conservation and stability. However, the results are similar to those obtained with the EDMF approach, which encourages us to extend the testing to the moist case and to three dimensions, where the benefits of the new approach should be seen.

The future of coupled modeling at the NWS

Hendrik Tolman (NOAA / NWS / Office of Science and Technology Integration)


Recently, the US National Weather Service (NWS) has completed a set of four documents laying out both a strategy and implementation plan for its environmental modeling enterprise. The first is a vision document on Unified Modeling for all of NOAA, ranging for operational weather modeling to fish stock assessments. The second is a Strategic Vision document for the Physical Environmental Modeling Enterprise at NOAA for the next decade. The third is a Roadmap for the evolution of the operational production suite of models at NOAA to move toward a Unified Forecast System (UFS), and accelerate its rate of improvement over the next 10 years. The UFS is intended to be coupled and ensemble based at all of its application scales.  The fourth and final document is a Strategic Implementation Plan (SIP) focusing on moving the present operational production suite toward the Strategic Vision and Roadmap goals over the next three years. The presentation will review the four documents, with a focus on coupling environmental components in the UFS.

Coupling convection with the continuity equation

Hillary Weller (University of Reading)


Many assumptions go into parameterisations of convection and the outcome is that these schemes perform badly, especially in the grey zone. A useful assumption that decouples convection schemes from the continuity equation is that convection schemes do not directly lead to net vertical mass transport. This assumption is clearly ridiculous but necessary to prevent convection schemes from generating acoustic and gravity waves outside the implicit solver. If convection schemes provided source terms to the continuity equation which were treated explicitly, the model would become unstable due to explicit treatment of acoustic and gravity waves.

This talk will describe how conditional filtering can be used to derive a full set of multi-fluid equations which will allow sub-grid scale convection in one fluid and stable air in another. This equation set includes interactions between convection and the continuity equation. A semi-implicit solution technique is presented which includes interactions of both fluids with the continuity equation and which is therefore stable.