Abstract
The influence of large-scale forcing on the high-resolution simulation of Tropical Storm Ivan (2004) in the southern Appalachians was investigated using the Weather Research and Forecasting model (WRF). Two forcing datasets were employed: the North American Regional Reanalysis (NARR; 32 km × 32 km) and the NCEP Final Operational Global Analysis (NCEP FNL; 1° × 1°). Simulated fields were evaluated against rain gauge, radar, and satellite data; sounding observations; and the best track from the National Hurricane Center (NHC). Overall, the NCEP FNL forced simulation (WRF_FNL) captures storm structure and evolution more accurately than the NARR forced simulation (WRF_NARR), benefiting from the hurricane initialization scheme in the NCEP FNL. Further, the performance of WRF_NARR is also negatively affected by a previously documented low-level warm bias in NARR. These factors lead to excessive precipitation in the Piedmont region, delayed rainfall in Alabama, as well as spatially displaced and unrealistically extreme rainbands during its passage over the southern Appalachians. Spatial filtering of the simulated precipitation fields confirms that the storm characteristics inherited from the forcing are critical to capture the storm’s impact at local places. Compared with the NHC observations, the storm is weaker in both NARR and NCEP FNL (up to Δp ~ 5 hPa), yet it is persistently deeper in all WRF simulations forced by either dataset. The surface wind fields are largely overestimated. This is attributed to the underestimation of surface roughness length over land, leading to underestimation of surface drag, reducing low-level convergence, and weakening the dissipation of the simulated cyclone.
Large Atmospheric Computation on the Earth Simulator: The LACES Project