Microphysics dependency in 3.5km NICAM DYAMOND phase 2 experiments

The elimination of convective parameterizations has made global cloud-resolving models (GCRM) powerful tools to realistically simulate and enable detailed investigations of convective systems. However, considerable uncertainties remain in the formulations of cloud microphysics that neither observation nor theory can adequately constrain. To address these issues and provide a platform for intercomparison of GCRMs, the DYAMOND initiative was launched in 2017 as the first GCRM model intercomparison project. Here we present results from experiments conducted on NICAM for participation in the DYAMOND project phase 2. Following the DYAMOND2 protocol, 40-day integrations were run on NICAM at a horizontal resolution of 3.5 km from Jan. 20, 2020. We conducted the integrations in two microphysics settings: one optimized for the reproducibility of the Madden-Julian Oscillation (MJO; MJO-tuned) in sub-seasonal integrations, and another optimized for longer integrations (highresMIP), which prioritizes the global energy balance. The integration period included an Madden-Julian Oscillation (MJO) event from late Jan. 2020 to mid-Feb. 2020. The comparison of the simulation results of the two settings revealed that both successfully reproduce the eastward movement of the MJO in the first 20 days of the integration. However, past this date, locations of the convective activities of the two experiments start to diverge, and were maintained near the dateline, as in the observation, only for the MJO-tuned setting. These differences were related to the systematic differences in the convective systems reproduced between the two settings, in which the highresMIP setting was accompanied by more tropical depressions and MJO-tuned setting was accompanied by more sporadic convections. Analyses of the simulated mean states of the atmosphere suggested that this was attributed to differences in the partitioning of frozen hydrometeors, which in turn resulted in more statically stable and unstable mean tropospheric states for the highresMIP and MJO-tuned settings, respectively. Our results revealed the sensitivity of the population of the convective systems to the microphysics settings in GCRM and indicated that careful assessment of microphysics settings is essential for reliable simulation of the atmosphere with GCRMs.