Quantitative Assessment of Tropical Cylcone Simulation Sensitivity in the Community Atmosphere Model.

Tropical cyclone;Sensitivity analysis;Community Atmosphere Model;Atmospheric;Oceanic and Space Sciences;Science;Atmospheric, Oceanic and Space Sciences

He, FeiJablonowski, Christiane ;

University of Michigan

关键词: Tropical cyclone;    Sensitivity analysis;    Community Atmosphere Model;    Atmospheric;    Oceanic and Space Sciences;    Science;    Atmospheric, Oceanic and Space Sciences;   

Others  :  https://deepblue.lib.umich.edu/bitstream/handle/2027.42/133448/hefei_1.pdf?sequence=1&isAllowed=y

瑞士|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

This work conducted nearly two thousand idealized AGCM simulations to systematically assess the sensitivities of simulated Tropical cyclone (TC) characteristics to changes in model input and evaluate the performance of three surrogate models for approximating the behavior of numerical models. The TC characteristics are intensity, precipitation rate, longwave cloud radiative forcing (LWCF), shortwave cloud radiative forcing (SWCF), cloud liquid water path (LWP) and ice water path (IWP). The National Center for Atmospheric Research (NCAR)/Department of Energy (DOE) Community Atmosphere Model (CAM) version 5.1.1 is adopted. First, the Reed-Jablonowski TC test case was upgraded to a version with background vertical wind shear, in which the well-known shear-induced intensity change and structural asymmetry in tropical cyclones are well captured. Then, a statistical framework, consisting of a combination of Latin Hypercube Sampling (LHS) and surrogate models, is used to investigate the sensitivities of the six simulated TC characteristics to five model initial conditions: initial size and intensity of vortex seed, sea surface temperature, vertical lapse rate and mid-level relative humidity. The surrogate models are shown to successfully reproduce the response of CAM to changes in input conditions, and serve as powerful tools for quantifying numerous model input-output relationships with reduced computational burden. Finally, we examined the impact of parameterized physical processes on TC simulation and quantified the relative importance of 24 physical parameters on the six TC characteristics, respectively. The response function between TC characteristics and the associated most sensitive parameters are characterized. A group of ensemble simulations showed that the interactive effect among physical parameters greatly enlarges the uncertainty of simulated TC precipitation, LWCF, SWCF and IWP. Parameter uncertainty in simulated TC intensity is comparable to uncertainty resulting from changes in model initial conditions and model resolution. The Gaussian Spatial Process Model (GaSP) produced robust fits to CAM model responses in TC intensity, LWCF and SWCF, but experienced some difficulty reproducing TC precipitation rate, LWP and IWP.