The NOAA CPO Modeling, Analysis, Prediction, and Projections (MAPP) program hosted a webinar on the topic of Addressing a Major Model Challenge: Improving the Representation of Clouds in Climate and Earth System Models on Wednesday, February 25, 2015. The announcement is provided below; you are invited to remotely join the session.
Steve Krueger - Global models parameterize the effects of subgrid-scale processes. Current global forecast models use grid spacings of a few tens of kilometers. In the next few years, the grid size is expected to be less than 10 kilometers. Such models will resolve mesocale circulations and the largest deep convective clouds. However, smaller clouds such as shallow cumuli will not be even partially resolved in the foreseeable future. We have developed an economical PDF-based SGS turbulence and cloudiness scheme that unifies the representation of turbulence and SGS cloud processes, and that is scale-aware: it does not require any explicit grid-size information. SHOC (Simplified Higher-Order Closure, Bogenschutz and Krueger 2013) combines the assumed joint PDF method with a prognostic SGS TKE equation, a diagnostic second- and third-moment closure, and a turbulence length scale related to the subgrid-scale turbulence kinetic energy and eddy length scales. The turbulence closure requires only one prognostic equation. This makes it economical, portable, and well-behaved. Our closure also uses a novel turbulence length scale that produces excellent scalablity with horizontal resolution. We implemented SHOC in a cloud-resolving model (CRM) and tested it against large-eddy simulation (LES) results. We also implemented it in a global model based on the Multiscale Modeling Framework (MMF) and evaluated the results using global observations. Our evaluations indicate that SHOC can realistically represent many boundary layer cloud regimes in coarse-grid CRMs, while in the global model, shallow cumulus and subtropical stratocumulus were both improved. We are currently installing it in the NCEP GFS.
Roger Marchand - In recent years the combination of increasing computational capability and uncertainties in climate simulations due to clouds (or more broadly un- or under-resolved processes that must be approximated or parameterized) has led to interest in higher resolution global models. In simple terms, the expectation is that significantly better simulations can be obtained by resolving cloud scale motions and relying less on cloud parameterizations. However, simulations capable of resolving cloud scale motions on global or large regional domains are computationally challenging. In particular for stratocumulus clouds, simulations run on small domains (where very fine grids can be applied) have demonstrated that horizontal grid spacing of less than 1 km and vertical grid spacing as small as 10 m may be required to simulate stratocumulus clouds. From a climate modeling perspective, accurately capturing stratocumulus and the transition between stratocumulus and cumulus is critical because of the large role that these clouds play in the Earth radiation balance. Decreasing the grid spacing in order to resolve boundary layer cloud processes and stratocumulus is (at best) marginally practical for global cloud resolving models, or the so called superparameterized or Multiscale Modling Framework (MMF) model. In the MMF, a two- dimensional or small three-dimensional cloud resolving model is embedded into each grid cell of a traditional climate model, replacing many of the cloud parameterizations normally used by climate models. One potential approach to increasing vertical resolution with only modest increases in computational costs is to use an adaptive grid. In this approach, additional grid points are added to the (relatively coarse) model base grid only where needed as determined by the model simulation itself. In this presentation we report on efforts to implement an Adaptive Vertical Grid in a MMF model.
Chris Golaz - Development of the next generation GFDL climate model (CM4) is an on-going multi-year effort. Prototype configurations for the atmospheric component (AM4) are currently under active development. These configurations are rooted in previous generation GFDL models, but with increased resolution and updated physics. The target horizontal resolution for AM4 is 0.5 deg, with an option for a lower resolution version of 1 deg for certain applications. The reduction of long-standing systematic errors is also a focus of the development efforts through updates to the cloud physics. For example, an updated convective parameterization improves the representation of the MJO. Modifications to the boundary layer and cloud parameterizations show potential in significantly reducing marine stratocumulus clouds biases.