Optimization of the Eddy‐Diffusivity/Mass‐Flux Shallow Cumulus and Boundary‐Layer Parameterization Using Surrogate Models

Abstract

Physical parameterizations in global atmospheric and ocean models typically include free parameters that are not theoretically or empirically constrained. New methods are required to determine the optimal parameter combinations for such models in an objective, exhaustive, yet computationally feasible manner. Here we propose to apply computationally inexpensive radial basis function (RBF) surrogate models to minimize a "cost," or error, function of an atmospheric model or a physical parameterization. The RBF is iteratively updated as more input-output pairs are obtained during the optimization. The approach is used to optimize the eddy-diffusivity/mass-flux boundary-layer parameterization of the convective boundary-layer in a single-column model framework. The optimization based on surrogate models is able to identify parameter combinations that reduce the error of the untuned default setting by 41%. The probability to detect a low-error solution increases rapidly especially over the first tens of single-column model evaluations. In comparison, a quadratic polynomial model only yields an error reduction of 17% since (a) high-order parameter interactions are not accounted for and (b) the construction of the polynomial is not based on an iterative sampling approach. The RBF surrogate models achieve this 17% error reduction for 40% of the polynomial model's cost. Interestingly, one of the emerging optimal settings describes a pure mass flux parameterization without eddy-diffusivity component. A second optimal solution is characterized by a plume fractional area of 20-30% and an eddy-mixing time scale of ∼700 s. Key Points:• Radial basis function (RBF) surrogate models provide an efficient framework to optimize physics parameterizations such as EDMF • The RBF surrogate models detect solutions much better than those provided by a commonly used quadratic polynomial model and are less costly • A multiplume model without eddy-diffusivity flux emerges as one of the optimal settings for the cloud-free and cloud-topped convective PBL

Time Evolution and Diurnal Variability of the Parametric Sensitivity of Turbine‐Height Winds in the MYNN‐EDMF Parameterization

“…Suselj et al (2020) examined the sensitivity of the Jet Propulsion Laboratory version of the EDMF parameterization and found strong sensitivity to the entrainment in the mass-flux part of the parameterization for case studies with boundary layer clouds. Langhans et al (2019) also looked at the parametric sensitivity of the EDMF parameterization in a single column model and found sensitivity to the plume entrainment and the initial plume properties.…”

Unified Entrainment and Detrainment Closures for Extended Eddy‐Diffusivity Mass‐Flux Schemes

“…(2012), and Langhans et al. (2019) in shallow convection parameterizations. Gregory (2001) analyzed LES of shallow convection and suggested , which was used by Tan et al. (2018) for shallow convection.…”

“…Using a perturbation response experiment in LES of shallow convection,Tian and Kuang (2016) found ϵ i0 ∝ 1=ðw i τÞ with a mixing time scale τ. Such an entrainment rate was used byNeggers et al (2002),Sušelj et al (2012), andLanghans et al (2019) in shallow convection parameterizations.•Gregory (2001) analyzed LES of shallow convection and suggested ϵ i0 ∝ b i =w 2…”

Compressible EULAG Dynamical Core in COSMO: Convective-Scale Alpine Weather Forecasts

“…In operational practice with the COSMO model at grid intervals of 2.2 and 1.1 km (e.g., at MeteoSwiss; see Schmidli et al 2018), the TKE scheme is used with its ability to provide a downgradient transport present in deep convective flows (Verrelle et al 2017). Following these considerations, the TKE scheme is used for simulations at 2.2-and 1.1-km grid while at 0.22-and 0.1-km grid, the three-dimensional Smagorinsky scheme (Langhans et al 2012;Baldauf and Brdar 2016) is applied. At a 0.55-km grid, both schemes are tested.…”