The drag coefficient parameterization of wind stress is investigated for tropical storm conditions using model sensitivity studies. The Massachusetts Institute of Technology (MIT) Ocean General Circulation Model was run in a regional setting with realistic stratification and forcing fields representing Hurricane Frances, which in early September 2004 passed east of the Caribbean Leeward Island chain. The model was forced with a NOAA-HWIND wind speed product after converting it to wind stress using four different drag coefficient parameterizations. Respective model results were tested against in situ measurements of temperature profiles and velocity, available from an array of 22 surface drifters and 12 subsurface floats. Changing the drag coefficient parameterization from one that saturated at a value of 2.3 × 10 -3 to a constant drag coefficient of 1.2 × 10-3 reduced the standard deviation difference between the simulated minus the measured sea surface temperature change from 0.8°C to 0.3°C. Additionally, the standard deviation in the difference between simulated minus measured high pass filtered 15-m current speed reduced from 15 cm/s to 5 cm/s. The maximum difference in sea surface temperature response when two different turbulent mixing parameterizations were implemented was 0.3°C, i.e., only 11% of the maximum change of sea surface temperature caused by the storm. Copyright 2009 by the American Geophysical Union.