A Stochastic Parameterization of Organized Tropical Convection Using Cellular Automata for Global Forecasts in NOAA's Unified Forecast System

In the atmosphere, convection can organize from smaller scale updrafts into more coherent structures on various scales. In bulk-plume cumulus convection parameterizations, this type of organization has to be represented in terms of how the resolved flow would "feel" convection if more coherent structures were present on the subgrid. This type of subgrid organization acts as building blocks for larger scale tropical convective organization known to modulate local and remote weather. In this work a parameterization for subgrid (and cross-grid) organization in a bulk-plume convection scheme is proposed using the stochastic, self-organizing, properties of cellular automata (CA). We investigate the effects of using a CA which can interact with three different components of the bulk-plume scheme that modulate convective activity: entrainment, triggering, and closure. The impacts of the revised schemes are studied in terms of the model's ability to organize convectively coupled equatorial waves (CCEWs). The differing impacts of adopting the stochastic CA scheme, as compared to the widely used Stochastically Perturbed Physics Tendency (SPPT) scheme, are also assessed. Results show that with the CA scheme, precipitation is more spatially and temporally organized, and there is a systematic shift in equatorial wave phase speed not seen with SPPT. Previous studies have noted a linear relationship between Gross Moist Stability (GMS) and Kelvin wave phase speed. Analysis of GMS in this study shows an increase in Kelvin wave phase speed and an increase in GMS with the CA scheme, which is tied to a shift from large-scale precipitation to convective precipitation. Plain Language Summary Vertical heat transfer between the ocean and atmosphere, called atmospheric convection, can organize over a variety of different scales, ranging from small-scale fair-weather cumulus clouds, rain showers and thunderstorms, to large scale "convectively coupled" equatorial waves. In traditional weather and climate models, such organization is not well represented, as only the mean effect of all possible convective types occurring in a model grid-box is represented by a onedimensional plume model. In this study we explore the impact of representing convective organization in weather and climate models using cellular automata-a discrete model often used to describe selforganizing behavior in physical systems. We are particularly interested in the impact of the proposed method in the tropics since the weather in the tropics is dominated by organized atmospheric convection, and serves as the engine of the Earth's atmospheric circulation pattern. We find that when we let the cellular automata initiate atmospheric convection in the nearby environment of existing precipitating convection, precipitation becomes more organized in space and time, and there is an improvement in the model's ability to reproduce the observed large-scale organization of convectively coupled tropical variability, important for improving predictions of weather and climate across the world.