Improving Ocean State Initialization in Coupled Tropical Cyclone Forecast Models

Coupled ocean-atmosphere models do not accurately forecast tropical cyclone (TC) intensity at present, partly due to inadequate representation of the ocean. Intensity depends in part on sea surface temperature (SST) since latent heat provided by sea surface evaporation is the primary TC energy source. Assuming constant SST, the additional power required by an intensifying storm is supplied by the increase in evaporation rate as wind speed increases. This tendency for the ocean to support intensification will be reversed if SST cools rapidly enough beneath the storm. As a result, SST along a projected TC path cannot be used alone to predict whether the ocean will promote or inhibit intensification. The thickness and temperature (heat content) of the upper-ocean warm layer along with the entrainment rate of colder water into the ocean surface mixed layer are important factors determining SST cooling rate and the net oceanic contribution to intensity change. A key factor for improving TC intensity forecasts by coupled models is accurate initialization of the state of the ocean. We explore this issue in the northwest Caribbean and southeast Gulf of Mexico for September 2002 just prior to hurricane Isidore. Initialization from climatology is inadequate because: (1) climatologies do not accurately represent the location of variable oceanographic features associated with different heat content such as the warm Loop Current; and (2) very anomalous conditions were present in the upper ocean prior to Isidore. In particular, the warm layer was much thicker than normal in the northwest Caribbean and in the Loop Current as documented by pre-storm Airborne Expendable Conductivity Temperature and Depth (AXCTD) profiles. We evaluate one Global Ocean Data Assimilation Experiment (GODAE) nowcast product for providing the initial ocean state, specifically the Atlantic Ocean nowcast produced by Planning Systems, Inc. and the Naval Research Laboratory that assimilates satellite altimetry (with vertical information projection) and SST into the HYbrid Coordinate Ocean Model (HYCOM). Comparison of nowcast temperature and salinity fields to aircraft observations and to Navy Modular Ocean Data Assimilation System (MODAS) analysis fields demonstrates that the present HYCOM-GODAE product accurately represents the location of important features such as the Loop Current, but does not reproduce the anomalous ocean state as accurately as the MODAS analysis. Salinity is inaccurately represented in both HYCOM and MODAS due to insufficient observations. Significant improvement is expected in the next generation HYCOM-GODAE product because it will use a more sophisticated assimilation procedure that also assimilates subsurface ocean observations. This study demonstrates the critical importance of sufficient observational coverage to initialize, evaluate, and improve the ocean component of TC prediction models for improving intensity forecasts.