Abstract
Mesoscale organized convection is generally misrepresented in the large-scale convective parameterizations used in contemporary climate models. This impacts extreme weather events (e.g., Madden-Jullian Oscillation (MJO)) and the general circulation driven by the significant amount of latent heat released from mesoscale organized convection. Studies show that the missing processes could be partially recovered by embedding a 2D cloud-resolving model in each GCM columns, i.e., super- parameterization. Despite successfully resolving the MJO, the study of mesoscale organization mechanism across the CRM and GCM cells remains sparse in this multiscale modeling framework. We applied rigorous detection and hierarchical clustering algorithms on the 3-hourly 2D resolved Mesoscale Convective Systems (MCSs) embedded in the cloud-permitting Super-Parameterized Community Atmosphere Model 5.2 (SPCAM). We then examined the fields of a long-lived and large MCS cluster at the central Pacific. The MCS cluster shows a squall line-like circulation throughout the lifecycle in SPCAM. The growth of deep shear surrounding the MCS cluster is presumably caused by the upgradient momentum transport of squall line organization. We simultaneously obtained the 3-hourly CAM cumulus parameterization outputs based on the timestep-wise perfect initial conditions given by SPCAM. This allows pure model physics comparison without introducing initial condition errors. The results show that CAM has a systematically biased stratiform cooling and moistening response below 3km to the given SPCAM deep convection favoring conditions. We showed that this bias is mainly due to the CAM’s stratiform microphysics scheme. The mesoscale organization in SPCAM thus provides a baseline for improvements of convective parameterization of CAM. The details of the systematic differences are revealed by a composite analysis. Results show SPCAM has a realistic growth to decay deep convection mode dominant in the large-scale heating and moisture sink in the composite. On the other hand, CAM strongly favors a steady stratiform dipole mode. CAM also shows little to no deep convection variation in the MCS organized environment given by SPCAM. The lack of variation commonly seen in deterministically parameterized large-scale models is often remedied by stochastic parameterization to represent the missing subgrid processes. By decomposing CAM’s cumulus parameterization schemes, we proposed a stochastic scheme to represent the mesoscale organization processes in CAM using the results from SPCAM. We performed a simple model proof-of-concept study to reach the goal of applying the stochastic scheme to a complex climate model. The time-independent perturbed parameter scheme shows comparable forecast skill when compared to the standard stochastic schemes. This time-independent perturbation scheme allows us to treat the forecast model as a black box with minimal intrusion to the model codes. The major standout of the new scheme is its ability to significantly reduce the simulation cost by building a stochastic spectral “surrogate model”, i.e., Polynomial Chaos Expansion (PCE). The surrogate model thus performs ensemble forecasts without the need to integrate the actual forecast model. The results show more reliable prediction and extended predictability compared to a deterministic scheme. The new scheme also shows comparable forecast skill when compared to the well known additive stochastic parametrization scheme. We incorporated the necessary components (i.e., Bayesian posterior sampling algorithm and global sensitivity analysis) on top of the stochastic scheme to form a more comprehensive forecast system for uncertainty quantification of the model parameters and tendencies. This study is the building block to the future application of the new forecast system to the climate model in the hope of representing convective organization.