An ensemble modeling approach to examining the impact of the Saharan air layer on the evolution of an idealized tropical cyclone

ABSTRACT The Saharan air layer (SAL) is a warm, dry, dusty layer of air that resides over the tropical Atlantic Ocean. While some studies have found that the SAL may indirectly promote tropical cyclone (TC) development, others have identified potential inhibiting factors for TC development. One uncertainty is whether TC evolution depends on increases in cloud condensation nuclei, thought to occur in the SAL, in a predictable manner. In this study, the Regional Atmospheric Modeling System (RAMS) was used to test if a systematic dependence of the evolution of TC intensity on cloud condensation nuclei (CCN) concentration exists in the context of uncertainties induced by perturbations in other parameters describing the meteorological conditions and properties of the initial vortex. First, RAMSv6.0 was used to show that minor differences in computational architecture across platforms resulted in maximum surface wind speed, Vmax, variations of 8.1 ms-1 as early as 12 hours into simulations with identical initial conditions. Results were identical when simulations with the same initial conditions were conducted on the same compute node. Second, RAMSv6.0 simulations on the same compute node examined how TC evolution responded to changes in the initial warm bubble temperature used to initialize convection and in the initial radius of maximum winds (RMW) in the vortex. Two and three grid simulations assessed the importance of horizontal resolution on the magnitude of the sensitivity to changes in these parameters. The sensitivity was greater in the 2 grid simulations than in the 3 grid simulations with a spread in Vmax at 96 hours of 14.4 ms-1 and 8.8 ms-1 respectively for the RMW tests. Finally, RAMSv4.3 was used to assess the impacts of dust acting as CCN on TC evolution in the context of changes in other initial conditions by conducting an ensemble of simulations. For CCN of 100, 101, 1000 and 2000 cm-3, a series of simulations with varying environmental temperature, relative humidity, warm bubble temperature, vortex height, and RMW were conducted. A monotonic relationship was seen such that the mean MSLP increased and mean Vmax decreased with increasing CCN. Although the difference in mean Vmax of TCs simulated with CCN concentrations of 100 vs. 101 cm-3 was larger than one standard deviation in the first 24 hours of simulated time, beyond 36 hours the mean values were within a standard deviation of each other. With a few exceptions, the mean Vmax and MSLP for simulations with CCN of 100, 1000 and 2000 cm-3 differed by more than one standard deviation from each other. Thus, the analysis suggests that the effect of CCN on TC intensity is greater than uncertainties in intensity induced by non-linear amplification of noise in the initial conditions