Exploring the Behavior of the Community Atmosphere Model in the Tropics: The Single-Column Root of a Hierarchy

I-Kuan Hu

Modeling of the tropical atmosphere has been a major challenge because the physical processes are non-adiabatic and inherently multi-scale. The tropical circulation is driven largely by flow-dependent latent heat release in localized turbulent cloudy convective systems, and the associated radiative impacts of water vapor and clouds are too important to ignore. With increasing computing power, software models that time-integrate full suites of approximate representations of these diverse processes have advanced substantially in terms of its numerical simulation. However, those models still suffer from various long-lasting errors, and seeing ways forward to im- proving them through a more holistic conceptual understanding of the tropical moist dynamics is still lacking. The goal of this work is to understand the interactions among subgrid and diabatic physical parameterizations, and between them and re- solved dynamics, by employing a climate model hierarchy that connects simple and complex configurations. In the Community Atmosphere Model (CAM), the focus of this work, we built two new inter-related configurations of the single-column version (called SCAM) at the lowest ”root” of the model hierarchy, for evaluation of CAM’s subgrid and diabatic physical process parameterizations. The ultimate goal is to find some experimental protocol with SCAM that is systematically predictive of some aspects of CAM’s behavior, drastically shortening the loop of model development and testing relative to full 3D simulation tests. Conventionally, SCAM is forced with observed, time-evolving advective tendencies, and the output states are compared to observed state variables. In this approach, however, genuine interactions among physical parameterizations are often overwhelmingly interfered by the external advective ”forcing”, limiting the transferability of lessons learned from SCAM to CAM. This work’s two new SCAM configurations are i) the SCAM-RCE configuration that allows SCAM to be run to its own radiative-convective equilibrium (RCE) as a one-dimensional climate model; and ii) the SCAM-PLSD configuration in which the column interacts with parameterized large-scale dynamics (PLSD) of the tropics. The SCAM-RCE configuration is advection-forced SCAM with zero horizontal transport and vertical velocity (as is true of the average over the whole Earth, or approximately over the broad tropical belt). The SCAM-PLSD configuration is formulated as deviations around the time-mean climate of SCAM-RCE, and may hopefully bridge the gap between SCAM and the most idealized 3D CAM configurations (aqua planets with uniform or zonally symmetric boundary conditions and forcings) in the hierarchy. For our PLSD treatment, this work uses the damped gravity wave (DGW) method, a mathematical approximation to the observed phenomenon of convectively-coupled Kelvin waves (CCKWs). In both the SCAM-RCE and SCAM-DGW configurations, we investigated how the standard CAM5 and CAM6 physics packages behave, and how their behaviors depend on vertical resolution and distribution, time step, advection scheme, and some parameter variations as well as sea surface temperature (SST), a representative of external climate parameters. That wide sweep of internal and external parameters yielded a wide range of results about the model’s behaviors, which include the following: 1. Both the climate state and temporal variability of SCAM-RCE are especially sensitive to vertical resolution and moist convection parameterization. In terms of vertical resolution, a series of grid-level switching experiments reveal that the vertical resolution of the middle troposphere (300-760 hPa) plays a crucial role. In terms of moist convection parameterization, the Zhang-McFarlane (ZM) deep convection scheme plays a key role in determining model behaviors. The updraft detrainment of the ZM scheme is important, driving layer structures of moisture and humidity in the middle troposphere. In addition, the different numbers of negative buoyancy level allowed in the algorithm for determining the deep convection top (ZMnumcin) primarily explains the different results between the SCAM5- and SCAM6-RCE. Future work should test whether this subtle internal parameter sensitivity corresponds to differences in the performance between CAM5 and CAM6, or is a numerical artifact. 2. One interesting phenomenon in the SCAM-RCE simulations is the presence of spontaneous free oscillations with multi-day periods, characterized by deepenings and shoalings of deep convection’s top along with rainfall rate oscillations. These oscillations seem to appear only when a finer-resolved vertical grid and the ZM scheme is used at the same time. When using a coarser-resolved vertical grid or turning the ZM scheme off, the SCAM-RCE tends to show random fluctuations. Column-integrated moist static energy (MSE) analysis indicates that the oscillations include both internal storage of MSE and conversion of latent heat to enthalpy. Moreover, time-pressure cross sections suggest that stratus clouds and overshooting of deep convection together modulate the column-integrated radiative cooling (and thus the column MSE). However, mechanism-denial experiments suggest that the cloud-radiative effects (CREs) only strengthen the oscillations rather than create ones; when the CREs are removed, the oscillating amplitudes of precipitation diminish, but the vertical structures of thermodynamics remain similar to those of the control simulations. It remains unclear whether the vertical-resolution and parameter dependence of the SCAM-RCE behaviors might map usefully onto any corresponding behaviors in 3D CAM, but selecting our SCAM configurations with the most robust behavior distinctions might offer a good way to select prospects for future 3D tests to establish such a mapping. 3. The SCAM-DGW coupled systems (anomalous SCAM convection + DGW- associated vertical velocity) exhibit a wide range of behaviors across the choices of radiation, parameterization of large-scale horizontal advection of moisture, upper boundary condition, wavelengths (5000, 7500, 10000 km), and damping time scales (2, 4, 6, and 8 days). By prescribing an idealized radiative cooling profile similar to that of the tropical mean, the SCAM-DGW coupled system is unstable for almost all the cases. Some cases exhibit a runaway bias with persistent vertical velocity, reaching a steady state with damping or sometimes even crashing the model’s numerics when damping is weak. Most generate oscillatory waves that are superposed with a bias component (of ascent, in most cases). By additionally relaxing moisture toward its reference profile in convergent layers (as a way of parameterizing large-scale horizontal advection of moisture), all the SCAM-DGW coupled systems with the selected wavelengths and damping time scales complete their simulations of 200-day coupling successfully (i.e. no model failure), allowing us to explore the simulated CCKWs across model physics, vertical resolution, wavelength, and damping strength. Some cases produce the desired result, with the coupled waves oscillating centered on the RCE climate reference state, while most other cases still feature a bias component. Further inspections of the 10000-km wavelength set of simulations reveal that, among those wavy results, the phase speed of ∼ 11, 14, and 17 ms−1 are seen, roughly consistent with the observed phase speed range of CCKWs. In addition, the vertical structures of the DGW-associated vertical velocity and thermodynamics resemble those of the CCKWs from observations and the simulation of a cloud-resolving model coupled with the DGW. Two specific set of simulations produces lower frequency waves: the CAM5 physics package with a finer-resolved 60-level grid, and the CAM6 physics package with CAM5’s ZMnumcin value and with the default 32-level grid. These results pose testable predictions about CAM5 vs. CAM6 tropical waves which, if verified, might begin to fulfill the ultimate goal outlined above.

Exploring the Behavior of the Community Atmosphere Model in the Tropics: The Single-Column Root of a Hierarchy

Abstract

Files and links (1)

Metrics

Details