Recent Developments and Future Challenges in Cumulus Parameterization
The biggest development has been the realization from cloud-resolving model studies that convection sensitivity to moisture, and vice-versa, was a key missing element of the CMIP3 generation of GCMs. This led to rejection of the concept of undilute plumes, the implementation of stronger triggers for convection, and a focus on the effects of increased entrainment. Evidence has continued to build that prominent tropical phenomena such as the Madden-Julian Oscillation (MJO) are the direct result of moisture-cloud-convection interactions with the large scale and fundamentally a reflection of variations in moist static energy. Active remote sensing instruments including NASA CloudSat-CALIPSO and TRMM and DOE ground-based cloud radars have allowed the population of convective cloud depths to be precisely documented and related to water vapor and the dynamical state. Successful MJO simulation now seems feasible in operational GCMs, the result being overall shallower convective depths, more variety in whether convection dries or moistens the environment, and a delayed transition from shallow to deep convection. This potentially has implications for the vertical distribution of chemical tracers with surface sources in the tropics. The transition from shallow to deep convection is also one culprit in the tendency of many GCMs to initiate convective rain too early in the day, but this problem appears to require also that basic ideas about the closure assumptions that determine cumulus mass flux be revisited.
Some cumulus parameterizations represent only the cumulus mass flux, but for the transport of condensate particles with significant fall speeds, knowledge of the updraft speed is required and is increasingly being represented in GCMs. Updraft speed and entrainment are intertwined, and so recent developments in retrieving cloud-scale vertical velocities from vertically profiling radars and scanning Doppler radars are beginning to provide an independent constraint. Early results show that parameterized updraft speeds in weak-entrainment models are stronger than observed. Predicting updraft speed and its interaction with the convective microphysics is more challenging than adopting simple thresholds for converting condensate to precipitation and less likely to be rewarded in a model evaluation based on mean state metrics, but the convection-microphysics interaction must be confronted to address the question of whether changes in high cloud properties other than cloud top height are a significant contributor to cloud feedback in a climate change. Diagnosis of convective updraft speed in GCMs also offers the hope of more physically based parameterizations of lightning flash rate than the current cloud top height or mass flux-based approaches used by chemistry-climate models.
The recent emphasis on the need to more strongly suppress deep convection does not adequately account for those situations in which vigorous deep convection does exist and sustain itself, sometimes organizing on the mesoscale. A recent development has been to try to capture the beginning stages of that process by parameterizing the cold pools that form in the presence of precipitating downdrafts and their role in producing secondary convection at their boundaries, especially where multiple cold pools collide. The attractiveness of this idea in the GCM context is its introduction of memory into the system, its potential to allow for an interactive approach to entrainment that is sensitive to current conditions, and its role as the starting point for thinking about organization. To date both abstract and explicit cold pool schemes have been developed, without a clear picture emerging yet as to the best approach. The next step, representing the organization that follows, is still on the horizon for virtually all GCMs, though eventually mesoscale aspects of convection may be explicitly simulated as some GCMs approach convection-permitting resolution.