A Study of Low Cloud Climate Feedbacks Using a Generalized Higher-Order Closure Subgrid Model
May 15, 2013
Grant Firl
Committee: Dave Randall (advisor), Scott Denning, Dick Johnson, Paul Evangelista (Ecosystem Science and Sustainability)
Abstract
One of the biggest uncertainties in projections of future climate is whether and how low cloudiness will change and whether that change will feed back on the climate system. Much of the uncertainty revolves around the difference in scales between the processes that govern low cloudiness and the processes that can be resolved in climate models, a fact that relegates shallow convection to the parameterization realm with varying levels of success. A new subgrid-scale parameterization, named THOR, has been developed in an effort to improve the representation of low cloudiness via parameterization in climate models. THOR uses the higher-order closure approach to determine the statistics describing subgrid-scale processes. These statistics are used to determine a trivariate double-Gaussian PDF among vertical velocity, ice-liquid water potential temperature, and total water specific humidity. With this information, one can diagnose what portion of the grid cell is cloudy, subgrid-scale cloud water content, and subgrid-scale vertical cloud water flux. In addition, samples are drawn from the trivariate PDF in order to drive the microphysics and radiation schemes. Although schemes similar to THOR have been developed over the past decade, THOR includes several novel concepts, like the generalization of the saturation curve to include condensation over both ice and liquid substrates, the determination of the PDF parameters from the given turbulence statistics, the introduction of a stochastic parcel entrainment process for the turbulence length scale, and a sub-column approach for calculating radiative transfer using the PDF.The new model is validated by simulating five test cases spanning a wide range of boundary layer cloud types, from stratocumulus to cumulus and the transition between the two. The results are compared to an ensemble of LES models running the same cases, with particular attention paid to turbulence statistics and cloud structure. For all cloud types tested, THOR produces results that are generally within the range of LES results, indicating that the single-column THOR is able to reproduce the gross characteristics of boundary layer clouds nearly as well as three-dimensional LES. Sensitivity to vertical grid spacing, diagnostic/prognostic third- order moments, choice of turbulence length scale entrainment process, and whether or not PDF sampling is used to drive the microphysics and radiation schemes is assessed for all test cases. Simulation of the cumulus regime was degraded when vertical grid spacing exceeded 200 m, when more third-order moments were predicted, when higher parcel entrainment rates were assumed, and when PDF sampling for the microphysics scheme was omitted. Simulation of stratocumulus was degraded with grid spacing larger than 100 m, when PDF sampling for microphysics was omitted, and when PDF sampling for radiation was included.
Lastly, THOR is used to study low cloud climate feedbacks in the northeastern Pacific Ocean in the context of the CGILS project. Initial conditions and forcings are supplied at 13 points along the GPCI cross-section that spans from the ITCZ northeast to the coast of California transecting regions of shallow cumuli and stratocumuli, for both the current climate and a climate with a +2K SST perturbation. A change in net cloud radiative forcing of 0-8 W/m^2 was simulated along the cross-section for the perturbed climate, representing neutral to weak positive feedback. The responsible mechanism appeared to be increased boundary layer entrainment and stratocumulus decoupling leading to reduced maximum cloud cover in the cumulus regime and reduced liquid water path in the stratocumulus regime.