A Multiple Model Examination of the Predictability of Hurricane Sandy (2012) and Hurricane Edouard (2014)

A multiple model approach using pseudo-operational implementations of the tropical cyclone Coupled Ocean/Atmospheric Mesoscale Prediction System (COAMPS-TC), Hurricane Weather Research and Forecasting model (HWRF), and the Weather Research and Forecasting (WRF-ARW) is used to examine the impact of different dynamic model cores on the predictability of Hurricane Sandy (2012) and Hurricane Edouard (2014). An inter-model comparison of track and intensity forecast errors are made to an intra-model multi-physics WRF-ARW ensemble to examine the impact of multi-physics and multi-core on track and intensity errors.

The Pennsylvania State University real-time cycling ensemble Kalman filter WRF-ARW data assimilation system (PSU WRF-EnKF) 60 member ensemble is used to initialize the three-model hindcasts of Hurricane Sandy at 00 UTC 26 October 2012 and Hurricane Edouard at 12 UTC 11 September 2014 with surface and lateral boundary conditions provided from the operational Global Forecast System (GFS) provided by NCEP. For simplicity, only the atmospheric component of each model is examined and the sea surface temperatures are fixed.

This study is the first systematic comparison performed for three state-of-the-art convection-permitting regional-scale hurricane prediction systems initialized with the same advanced data assimilation system initial conditions. The comparison in this study helps to elucidate the current limits of both deterministic and probabilistic practical predictability of two tropical cyclones, examines the impact of model error versus initial condition error, and explores the relative importance of model core versus model physics on hurricane forecast errors in terms of both track and intensity.