There has been a growing number of climatologies and observational case studies of extratropical transitions (ETs) of tropical storms; however, there have been very few high resolution simulations of particular events. This paper examines Floyd's ET along the U.S. East Coast on 16-17 September 1999 using the Penn State/NCAR MM5 down to 1.33-km grid-spacing. The MM5 was able to reproduce the track of Floyd, the development of a deep and intense baroclinic zone along the coast, and the tendency for heavy (> 30 cm) precipitation to fall within a relatively narrow (30-40 km wide) band just inland of the coast. A combination of strong deformation frontogenesis up to 400 mb, moist symmetric instability below 800 mb, and slantwise neutrality aloft resulted in the narrow and intense band of precipitation just inland of the coast. A separate simulation without the Appalachians and coastal terrain had little effect on Floyd's wind, temperature, and precipitation evolution.

A frontogenesis budget revealed that the upper-level baroclinic zone was enhanced by a horizontal gradient in mid-level latent heating between the heavy precipitation near the coast and the lighter precipitation further inland. This result is consistent with a simulation without latent heating, which resulted in much less baroclinicity and downstream ridging aloft. Without latent heating, the central pressure of Floyd was 25 mb weaker than the control run, and the storm only slowly propagated up the coast. Without evaporative effects from precipitation, the low-level front was 10-20% weaker than the control, and Floyd was about 4 mb weaker. Another simulation without surface heat fluxes resulted in a 4-5 mb weaker cyclone and 10-20% less precipitation than the control. The importance of diabatic effects is also revealed by the large spread in model precipitation amounts near the coast when using different PBL and convective parameterizations.