Forecasters at the National Hurricane Center use several different computer models to help forecast hurricane motion. Over the past 20 years errors in the official forecasts have decreased a little less than 1% per year on average, largely because of improved forecast models. The models fall into three classes: statistical models that try to relate the hurricanes' motion to experience with past storms, dynamical models that solve the fundamental physical equations that describe atmospheric motions under various simplifying assumptions, and statistical-dynamical models that use results from dynamical-model calculations as input to statistical descriptions of storm motion. The models being run in 1995 are:

**CLIPER**is the only purely statistical model still in use. The acronym stands for CLImatology and PERsistence. The only inputs to CLIPER are the storm's current and previous positions, its motion and intensity, and the time of the year. Because CLIPER knows nothing about the meteorological situation, forecasters use it as a "no skill" prediction against which other models are judged.-
**The Aviation Model**is a version of NCEP's (formerly NMC's) operational forecast model. It is a multilevel, global spectral model with T126 truncation, equivalent to 106 km horizontal resolution. Hurricane track forecasting capability is attained by inserting a synthetic vortex into the initial analysis at the observed hurricane position and tracking the vortex motion as the model calculation evolves. -
**BAM**(Beta and Advection Model) assumes that the hurricane moves with the aviation model's winds (vertically averaged and filtered to remove the hurricane) plus a drift toward the pole and westward due to the northward increase of the Coriolis parameter (the "beta effect"). BAM comes in three versions, shallow, medium, and deep, depending upon the depth over which the vertical average is computed. -
**The GFDL Model**is a full physics model, developed as a research tool at the Geophysical Fluid Dynamics Laboratory, that has become fully operational. With its fine resolution (about 19km) and other special adaptations to the hurricane problem, it has an excellent forecasting record. -
**VICBAR**is a specialized forecast model developed at HRD/AOML. It replaces the complex dynamics of the atmosphere with a layer of fluid with constant density, representing the average motions between 850 and 200 hPa. Despite its simplified dynamics and 50 km resolution, it is also a powerful forecast model. -
**NHC93**was developed at NHC. It is the only statistical-dynamical model now in use. It uses the output from the aviation model in a series of statistical regression relations to predict hurricane motion.

Bender, M. A., R. J. Ross, R. E. Tuleya, and Y. Kurihara,
1193: Improvements in tropical cyclone track and intensity
forecasts using the GFDL initialization scheme. *Mon. Wea.
Rev.*, **121**, 2046-2061.

Lord, S. J., 1993: Recent developments in tropical cyclone
track forecasting with the NMC global analysis and forecast
system. * Preprints, 20 Conference on Hurricanes and Tropical
Meteorology*, American Meteorological Society, 290-291.

McAdie, C. J., and M. B. Lawrence, 1993: Long-term trends
in National Hurricane Center track forecast errors in the
Atlantic basin.* Preprints, 20 Conference on Hurricanes and
Tropical Meteorology*, American Meteorological Society,
281-284.