[gvidas]

Estimating the potential impact of climate change on Hurricane Florence

Methodology.

The global atmospheric model CAM5 is set up in a variable-resolution configuration with a base grid spacing of ~100 km, similar to conventional atmospheric general circulation models, and a refined region over the North Atlantic basin with a grid spacing of ~28 km. The model is initialized with atmospheric analyses from NOAA’s Global Forecast System (GFS) following the technique outlined in Zarzycki and Jablonowski (2015) and run for 7 days and the first 5 days are analyzed. For Hurricane Florence, the model is initialized on 9/11 at 00z. To account for model uncertainty in storm characteristics, a 10*member ensemble is created by randomly varying three parameters (c0_ocn, tau, and dmpdz) in the deep convective parameterization (Zhang and McFarlane 1995). TC tracks from the forecast runs are generated using the TempestExtremes algorithm (Ullrich and Zarzycki 2017). For modified forecasts with the climate change signal removed, the methodology follows the the framework of Wehner et al. (2018). In particular, the air temperature, specific humidity, and sea surface temperature from the observed initial conditions are modified to remove climate change effect. Data from the C20C Detection and Attribution project ( portal.nersc.gov/c20c ) define the initial conditions for the counterfactual “storm that might have been”. Differences between global simulations driven by observed boundary conditions and simulations driven by conditions with the human induced climate change removed are calculated for September over the 1996-2016 period and approximate the change in the large scale environment attributable to climate change. Additionally, the greenhouse gas concentrations, solar radiation conditions, ozone concentration, and aerosol concentrations are all set to pre-industrial levels for the modified forecasts.