Effects of a stochastic convective parameterization on tropical intraseasonal variability

Generally, convective parameterizations used in general circulation models (GCMs) aim only to simulate the mean or first-order moment of convection. Higher-order moment sub-grid convective variability is not explicitly considered in these parameterizations. In this study, a convective parameterization is developed that attempts to include higher-order moments by representing convection as a stochastic process. The parameterization uses a traditional GCM convective scheme (i.e. the Betts-Miller scheme) to calculate an expected value of convective precipitation (and thus the vertical integral of convective heating) at each time step. This expected value is then used to set the parameters of a probability distribution function that has been pre-calculated based upon observations. The convective heating is chosen as a random number from this distribution, which has been constrained by large-scale variables, but now also includes higher-moment variability of convection. This stochastic convective parameterization is implemented in an intermediate-complexity tropical atmospheric model.

Preliminary results indicate that the inclusion of the stochastic convective parameterization noticeably affects tropical variability. However, the quantitative effects, including those on intraseasonal variability, do appear to have a fairly strong dependence on the details of how the stochastic effects are parameterized. For instance, increased autocorrelation time in the stochastic process tends to increase the influence of stochastic effects.