Satellite-derived ocean thermal structure for the North Atlantic hurricane season

Abstract

This paper describes a new model (method) called Satellite-derived North Atlantic Profiles (SNAP) that seeks to provide a high-resolution, near-real-time ocean thermal field to aid tropical cyclone (TC) forecasting. Using about 139 000 observed temperature profiles, a spatially dependent regression model is developed for the North Atlantic Ocean during hurricane season. A new step introduced in this work is that the daily mixed layer depth is derived from the output of a one-dimensional Price–Weller–Pinkel ocean mixed layer model with time-dependent surface forcing.

The accuracy of SNAP is assessed by comparison to 19 076 independent Argo profiles from the hurricane seasons of 2011 and 2013. The rms differences of the SNAP-estimated isotherm depths are found to be 10–25 m for upper thermocline isotherms (29°–19°C), 35–55 m for middle isotherms (18°–7°C), and 60–100 m for lower isotherms (6°–4°C). The primary error sources include uncertainty of sea surface height anomaly (SSHA), high-frequency fluctuations of isotherm depths, salinity effects, and the barotropic component of SSHA. These account for roughly 29%, 25%, 19%, and 10% of the estimation error, respectively. The rms differences of TC-related ocean parameters, upper-ocean heat content, and averaged temperature of the upper 100 m, are ~10 kJ cm−2 and ~0.8°C, respectively, over the North Atlantic basin. These errors are typical also of the open ocean underlying the majority of TC tracks. Errors are somewhat larger over regions of greatest mesoscale variability (i.e., the Gulf Stream and the Loop Current within the Gulf of Mexico).