Assessing the Uncertainty in Tropical Cyclone Simulations in NCAR's Community Atmosphere Model

Abstract

The paper explores the impact of the initial-data, parameter and structuralmodel uncertainty on the simulation of a tropical cyclone-like vortexin the National Center for Atmospheric Research's (NCAR) CommunityAtmosphere Model (CAM). An analytic technique is used to initializethe model with an idealized weak vortex that develops into a tropicalcyclone over ten simulation days. A total of 78 ensemble simulationsare performed at horizontal grid spacings of 1.0��, 0.5�� and 0.25��using two recently released versions of the model, CAM 4 and CAM5. The ensemble members represent simulations with random small-amplitudeperturbations of the initial conditions, small shifts in the longitudinalposition of the initial vortex and runs with slightly altered modelparameters. The main distinction between CAM 4 and CAM 5 lies withinthe physical parameterization suite, and the simulations with bothCAM versions at the varying resolutions assess the structural modeluncertainty. At all resolutions storms are produced with many tropicalcyclone-like characteristics. The CAM 5 simulations exhibit moreintense storms than CAM 4 by day 10 at the 0.5�� and 0.25�� grid spacings,while the CAM 4 storm at 1.0�� is stronger. There are also distinctdifferences in the shapes and vertical profiles of the storms inthe two variants of CAM. The ensemble members show no distinctionbetween the initial-data and parameter uncertainty simulations. Atday 10 they produce ensemble root-mean-square deviations from anunperturbed control simulation on the order of 1-5 m s�|1 for themaximum low-level wind speed and 2-10 hPa for the minimum surfacepressure. However, there are large differences between the two CAMversions at identical horizontal resolutions. It suggests that thestructural uncertainty is more dominant than the initial-data andparameter uncertainties in this study. The uncertainty among theensemble members is assessed and quantified.