Dependence of aqua‐planet simulations on time step

Abstract

AbstractAqua‐planet simulations with Eulerian and semi‐Lagrangian dynamical cores coupled to the NCAR CCM3 parametrization suite produce very different zonal average precipitation patterns. The model with the Eulerian core forms a narrow single precipitation peak centred on the sea surface temperature (SST) maximum. The one with the semi‐Lagrangian core forms a broad structure often with a double peak straddling the SST maximum with a precipitation minimum centred on the SST maximum. The different structure is shown to be caused primarily by the different time step adopted by each core and its effect on the parametrizations rather than by different truncation errors introduced by the dynamical cores themselves. With a longer discrete time step, the surface exchange parametrization deposits more moisture in the atmosphere in a single time step, resulting in convection being initiated farther from the equator, closer to the maximum source. Different diffusive smoothing associated with different spectral resolutions is a secondary effect influencing the strength of the double structure. When the semi‐Lagrangian core is configured to match the Eulerian with the same time step, a three‐time‐level formulation and same spectral truncation it produces precipitation fields similar to those from the Eulerian.It is argued that the broad and double structure forms in this model with the longer time step because more water is put into the atmosphere over a longer discrete time step, the evaporation rate being the same. The additional water vapour in the region of equatorial moisture convergence results in more convective available potential energy farther from the equator which allows convection to initiate farther from the equator. The resulting heating drives upward vertical motion and low‐level convergence away from the equator, resulting in much weaker upward motion at the equator. The feedback between the convective heating and dynamics reduces the instability at the equator and decreases the convection there. The behaviour of the parametrizations depends on the amount of water inserted into the troposphere during a discrete time step. This is as important as the rate of insertion. Experiments are described that support this explanation. Copyright © 2003 Royal Meteorological Society