Stochastic behavior of tropical convection in observations and a 1 multicloud model

5 The aim for a more accurate representation of tropical convection in global circulation models 6 is a long-standing issue. Here, we investigate the relationships between largeand convective 7 scales in observations and a Stochastic Multicloud Model (SMCM) to ultimately support 8 the design of a novel convection parametrization with stochastic elements. Observations of 9 tropical convection obtained at Darwin and Kwajalein are used here. We find that the vari10 ability of observed tropical convection generally decreases with increasing large-scale forcing, 11 implying a transition from stochastic to more deterministic behaviour with increasing forc12 ing. Convection shows to yield a more systematic relationship with measures related to 13 large-scale convergence compared to measures related to energetics, e.g. CAPE. Using the 14 observations, we adjust the parameters in the SMCM, force it with the time series of the 15 observed large scale state and compare the simulated convective behaviour to that observed. 16 We find that the SMCM-modelled cloud fields compare better with observations when using 17 predictors related to convergence rather than energetics. Furthermore, the underlying frame18 work of the SMCM is able to reproduce the observed functional dependencies of convective 19 variability on the imposed large-scale state – an encouraging result on the road towards a 20 novel convection parametrization approach. However, establishing sound cause-and-effect re21 lationships between tropical convection and the large-scale environment remains problematic 22 and warrants further research. 23