The Sensitivity of Simulated Tropical Cyclones to Tunable Physical Parameters in Community Atmosphere Model

Abstract:

The inability to explicitly resolve the sub-grid scale physical processes (e.g. cloud, precipitation and convection) of atmospheric general circulation models (AGCMs) greatly limits their performance in simulating tropical cyclones (TCs) and predicting their future changes. To address it, this study carried out a total of 92 simulations and investigated the sensitivity of TC simulation to 24 physical parameters that control the deep convection, shallow convection, turbulence, cloud microphysics and cloud macrophysics processes in Community Atmosphere Model version 5 (CAM5). The Reed-Jablonowski TC test case is utilized and run at horizontal resolution of 0.5^°×0.5^° with 30 vertical levels. The sensitivity is assessed by the uncertainty each parameter exerts on simulated TC while perturbing it from its minimum to maximum with other 23 parameters set to their default value. The uncertainty is characterized by changes on simulated TC intensity (measured by absolute maximum wind speed at 100 m above surface), precipitation rate, shortwave cloud radiative forcing (SWCF), longwave cloud radiative forcing (LWCF), cloud liquid water path (LWP) and cloud ice water path (IWP), the latter five of which are quantified by their area-weighted value over the tropical cyclone region. Both the relative importance among these 24 physical parameters on TC simulation and the response function describing how they affect the six TC characteristics are quantified. It is found that the simulated TC intensity is most sensitive to the parcel fractional mass entrainment rate in Zhang-McFarlane (ZM) deep convection scheme. Decreasing this parameter enables a change from tropical depression to Category-4 storm. In contrast, other 23 physical parameters cause intensity uncertainty within 10 m/s. The precipitation rate, SWCF, LWP and IWP are also found to receive major impact from parameters in ZM deep convection scheme while the LWCF is dominated by parameters both in ZM deep convection and cloud microphysics scheme. The tunable parameters in University of Washington (UW) moist turbulence scheme, UW shallow convection scheme and cloud macrophysics are found to generally exert minor impact on TC simulation in CAM5. This study sheds light on improving the TC representation in AGCMs by tuning the physical parameters.