One Step at a Time: How Model Time Step Significantly Affects Convection‐Permitting Simulations

Abstract

We show that there is a strong sensitivity of cloud microphysics to model time step in idealized convection‐permitting simulations using the COnsortium for Small‐scale MOdeling model. Specifically, we found a 53% reduction in precipitation when the time step is increased from 1 to 15 s, changes to the location of precipitation and hail reaching the surface, and changes to the vertical distribution of hydrometeors. The effect of cloud condensation nuclei perturbations on precipitation also changes both magnitude and sign with the changing model time step. The sensitivity arises because of the numerical implementation of processes in the model, specifically the so‐called “splitting” of the dynamics (e.g., advection and diffusion) and the parameterized physics (e.g., microphysics scheme). Calculating one step at a time (sequential‐update splitting) gives a significant time step dependence because large supersaturation with respect to liquid is generated in updraft regions, which strongly affect parameterized microphysical process rates—in particular, ice nucleation. In comparison, calculating both dynamics and microphysics using the same inputs of temperature and water vapor (hybrid parallel splitting) or adding an additional saturation adjustment within the dynamics reduces the time step sensitivity of surface precipitation by limiting the supersaturation seen by the microphysics, although sensitivity to time step remains for some processes.

Improving Time Step Convergence in an Atmosphere Model With Simplified Physics: The Impacts of Closure Assumption and Process Coupling

“…Given the uncertainties associated with model formulation and the constraints on computational resources, numerical accuracy is often not considered the highest priority (Beljaars, 1991;Beljaars et al, 2018). A number of recent studies have shown that time integration errors in the parameterizations or process coupling can be sufficiently large to dominate the total error of the numerical solutions (e.g., Barrett et al, 2019;Zhang et al, 2012Zhang et al, , 2018. These results suggest that more attention is needed to identify and reduce numerical artifacts related to the parameterizations.…”

Marine Low Clouds and their Parameterization in Climate Models

“…For instance, the time step used in MRI-ESM2 (Yukimoto et al 2019) for TL159 simulations submitted to CMIP6 is minutes. However, a time step should be less than several tens of seconds for an appropriate calculation of cloud microphysics (e.g., Barrett et al 2019, Posselt and Lohmann 2008, Michibata et al 2019. One solution to the problem of long time steps is for short time-scale processes including cloud microphysics and turbulence to be calculated several times using sub-time-steps within one model integration time step (e.g., Posselt and Lohmann 2008, Gettelman et al 2015, Michibata et al 2019.…”

Development of “Physical Parametrizations with PYthon” (PPPY, version 1.1) and its usage to reduce the time-step dependency in a microphysical scheme

“…A similar time-step sensitivity test is done regarding the sedimentation scheme used in the model (all other parametrizations, including the microphysical processes, were turned off). Two schemes are available: the AROME operational one (Bouteloup et al, 2010;BSB2010 hereafter), which is a statistical scheme and the Eulerian scheme included in the original version of the ICE scheme. In order to remain stable, the Eulerian scheme uses a time-splitting technique with an upstream differencing scheme.…”

“…In the COnsortium for Small-scale MOdeling (COSMO) model, Barrett et al (2019) also observed a time-step dependency on rain and hail accumulations traced back mainly to the interaction between the dynamics and the physics of the model and, to a lesser extent, to some microphysical processes. The example shown in their paper demonstrates that a significant part of the time-step dependency can also be explained by the microphysical scheme itself.…”