Toward Identifying and Understanding Errors in Numerically Simulated Tropical Cyclones: A Case Study of Hurricane Edouard (2014)

Over the past several years, a widespread coordinated effort has resulted in improved numerical weather prediction (NWP) model forecasts of tropical cyclone (TC) intensity – as measured by maximum 1-minute sustained 10-m wind speed. However, there is often substantial disagreement between observed TCs and their simulated counterparts in other important areas such as storm structure and air-sea thermodynamic fields. Hurricane Edouard (2014), for which an unprecedented volume and diversity of observational data exists, is used as a case study to identify and better understand the abovementioned model deficiencies. Specifically, observations are compared with analysis fields from three cycling regional-scale NWP models: (i) the operational HWRF, with vortex adjustment/relocation; (ii) HWRF but without vortex adjustment/relocation; and (iii) a version of the Advanced Research WRF (ARW) without vortex adjustment/relocation. Because we are interested for now in comparing observations with model analyses (not forecasts), we tested the effect of initializing each of the three regional models from the GFS global model at different lead times (of 0, 6, 12, and 18 hours before the final analysis time of 18Z 15 September). Not surprisingly, using a short-term forecast from the regional model as the first guess (i.e., prior to data assimilation) yields a much better analysis than when the GFS is used as the prior (i.e., a cold-start). Detailed analyses of these simulations will be presented, with a particular focus on air-sea interaction.