Convergence of the simulated tropical convergence zone and its response to climate change with resolution

Abstract

More than 50 years have passed since the first attempts to understand the atmospheric general circulation in a comprehensive climate model. General circulation models (GCMs) have played an essential role in modeling and understanding the atmospheric general circulation and its response to climate change. However, their construction (horizontal grid resolutions in the order of 100km and subgrid processes representation through parameterization) has raised criticism as uncertainties in global warming have remained almost the same along the different Coupled Model Intercomparison Projects (CMIPs, e.g., Meehl et al. 2020; Zelinka et al. 2020). One of the main contributors to inter-model spread in global warming in response to increased CO2 concentration is the change of clouds, attributable to differences in cumulus parameterizations. However, clouds are complex features of the atmosphere that span a multi-scale range of processes, which makes it difficult to adequately represent physically at horizontal grid resolutions in the order of 100km. To address the underlying uncertainty, more complex parameterization or explicitly resolving convection at adequate horizontal grid spacing in the order of one kilometer can be used; the so-called convection-permitting models or global storm resolving models (GSRM). However, the computational cost to perform such global simulations increases. In this dissertation, I first explore, in a state-of-the-art GSRM, if, by increasing the horizontal resolution, the atmospheric general circulation displays physical convergence as uncertainties and discretization errors reduce. Secondly, I explore if its response to forcing, mimicking climate warming, and the responses to climate change of the atmospheric general circulation display convergence.

In the first part of this dissertation, I focus on evaluating physical convergence of the atmospheric general circulation and increasing horizontal resolution using the ICOsahedral Nonhydrostatic (ICON) model from 160km to 1.25km horizontal grid spacing. I develop a methodology based on the Richardson extrapolation method to assess physical convergence in an idealized setup, which retain basic atmospheric features of Earth's general circulation, response to warming, and required reduced computational resources to achieve robust statistics. As I increase the horizontal grid spacing, a better representation of clouds and zonal distribution of water vapor drives convergence in the energy and water budget towards kilometer-scale [O(1km)] horizontal grid spacing. The atmospheric general circulation displays convergence in its structure at kilometer-scale horizontal grid spacing, but its intensity requires finer horizontal grid spacing. However, shallow marine boundary layer clouds and their effect on the shortwave radiation components of the energy budget would require hectometer-scale horizontal grid spacing to achieve convergence, according to Large-Eddy Simulation (LES) findings (e.g., Stevens et al., 2020). ICON displays physical convergence with increasing horizontal resolution, and at 2.5-5km horizontal grid spacing, uncertainties from numerical errors are significantly smaller in the large-scale structure of the general circulation.

In the second part of this dissertation, I investigate the model's response to a uniform increase in sea surface temperature, mimicking global warming. Across resolution and according to previous studies in GCMs, the hydrological cycle intensifies slower than the increase of precipitable water, the Hadley cell expands, and the anvil clouds shift to higher altitudes. The hydrological cycle intensification and longwave climate feedback converge faster than other metrics at 10km horizontal grid resolution, in agreement with observations and other GSRMs. The response of the width of the deep tropics varies according to the Intertropical Convergence Zone (ITCZ) structure, a contraction for a single ITCZ, and an expansion otherwise, converging at 5km horizontal grid resolution. Across resolution, the boundary layer becomes drier while the middle and upper troposphere become moister and significantly warmer. The amount of anvil clouds in the deep tropics and shallow clouds in the subtropics do not show significant changes. My results show that ICON displays robust responses to warming with increasing horizontal resolution and converging for most large-scale features at 5km, even if shallow marine boundary layer clouds have not converged at those resolutions.

In summary, this dissertation demonstrates that a GSRM (ICON) displays convergence with increasing horizontal grid spacing. Thus, the results raise confidence in using GSMR to investigate the climate response to warming. However, if one GSRM converges, it does not imply that another will or that it arrives at the same climate. For this reason, the methodology developed serves as a tool to assess GSRM implementations and further development.