ESRL/PSD Seminar Series

Abstract

---

The superparameterized* (SP) Community Atmosphere Model version 3.0 (SPCAM) produces a remarkably realistic Madden Julian Oscillation in free-running simulations but why it succeeds where most global models fail is in debate. One hypothesis is that the physics of mesoscale organization are critical to MJO dynamics. Here, hindcast sensitivity experiments explore the internal sensitivities of the SPMJO. Results show that SPCAM's intrinsic MJO predictability is robust to CRM configurations with no mesoscale, and also robust to initial conditions with no mesoscale memory. Physically this argues against the hypothesis. Practically, it means there is untapped potential for faster superparameterized climate simulations using ultra-small-domain CRMs. Real-world hindcasts are further attempted to test the validity of the SPMJO as an analog for nature. However, meaningful MJO skill analysis is prohibited by imperfect initialization when using ERA-Inerim reanalysis as ICs. Strong systematic climate drift toward's SPCAM's eccentric attractor underscores a need for SP-data-assimilation, which now seems computationally feasible given the model speedup and the fact that the difficult problem of initializing the interior scale is avoidable. *The superparameterization technique for global atmospheric simulation uses thousands of large domain cloud resolving models in place of conventional cloud/boundary layer parameterizations.

---