Contains constant and diurnal cycle SST forcing model codes (SAM version 6.10.8) and scripts to analize the model simulations.
The model domain is square, doubly periodic in both horizontal directions x and y. We run simulations with a domain size of 768 km x 768 km. The horizontal resolution is 3 km, and the model's lowest level is defined at 37.5 m with vertical grid spacing increasing gradually with height from a few tens of meters to 500 m in the free troposphere. The model time step is 10 seconds. The Coriolis parameter is ignored in all simulations.
We force the model with a mean SST of 303 K either constant in time, or oscillating with a maximum amplitude of $\pm$3 K for the diurnal cycle simulations. We initialize the model with a sounding obtained from the domain-mean equilibrium profiles of a smaller-domain simulation with same SST (303 K). The forcing SST, temporally fixed or oscillating, is spatially homogeneous at a given time. Our experiment design eliminates any SST gradient effect since SST gradients force convection to cluster over warmer SSTs [Müller and Hohenegger, 2020; Shamekh et al., 2020a; Tompkins, 2001]. No mean wind is imposed and domain mean winds are nudged to zero. All the simulations are 100 days long.
Four different experiments are made. The first simulation is forced by fixed SST in time and no diurnal cycle in the incoming solar radiation (a constant insolation in space and time, with the same incident flux and zenith angle used by Tompkins and Craig (1998), that is a solar constant of 685 .0 W m\textsuperscript{-2} and a zenith angle of 51.7 \degree). We call this experiment CTRL. In the second experiment, we allow a diurnal cycle in the incoming solar radiation (with the same diurnal mean as CTRL) keeping everything else the same as CTRL. We call this experiment CTRL\textsubscript{sw}. The third experiment, EXPT, is the same as CTRL except SST is diurnally oscillating with a maximum amplitude of $\pm$3 K. The fourth experiment, EXPT\textsubscript{sw}, is the same as EXPT except with a diurnal cycle in the incoming solar radiation. In EXPT\textsubscript{sw}, the diurnal peak of SST oscillation lags that of the incoming solar radiation by 6 hours. This lag is qualitatively consistent (although perhaps slightly longer than average) with observed time lag between maximum incoming solar radiation and SST maximum. Moreover, the addition of the second and third type experiments, CTRL\textsubscript{sw} and EXPT, helps us to distill the effects of variations in forcing on resulting CSA. The designs CTRL\textsubscript{sw} and EXPT are unrealistic in a way since varying incoming solar radiation or SST are inseparable in reality. However, we add those diurnal variations one by one from CTRL to EXPT\textsubscript{sw} to evaluate the influence of each forcing categorically.
Additionally, we performed 7 ensemble simulations each for CTRL and EXPT\textsubscript{sw} to check the robustness of early onset of CSA in case of the diurnally varying SST forcing. For these simulations, we used a mean SST value of 303 K as before, as well as additional experiments with mean SST 300 K, constant for the CTRL simulations and diurnally oscillating (mean SST $\pm$3 K) for the EXPT\textsubscript{sw} simulations, in order to assess the impact of mean SST on the variability of aggregation timing.