GFDL - Geophysical Fluid Dynamics Laboratory

Global Modeling of Tropical Cyclone Activity

The model:

As part of the development of a new generation of atmospheric models with finer spatial resolution, we have been examining a variety of issues related to tropical cyclone statistics, including inter-annual variability, seasonal predictability, and the effects of global warming. The cornerstone of this research is an atmospheric model with approximately 50km horizontal resolution that provides an impressive simulation of the climatology of tropical cyclones, including their geographical distribution and seasonal cycle in different ocean basins.  It is intriguing that a model of this resolution can do so well in this regard, and it is this success that has allowed us to address a number of important issues related to tropical cyclones.  

The model and it simulation of various tropical cyclone statistics is described in

, Isaac
, Shian-Jiann
, and Gabriel
A Vecchi
, December 2009: Simulations of global hurricane
climatology, interannual variability, and response to global warming
using a 50km resolution GCM
. Journal of Climate, 22(24),
doi:10.1175/2009JCLI3049.1.   [ PDF

This model, which we refer to as C180HiRAM, shares many features with the AM2.1 atmospheric model (and the LM2 land model) that served as the atmospheric (and land) components of the GFDL’s CM2.1 model. But it also has some features shared with the newer AM3 model underdevelopment for the CMIP5/AR5 archive, particularly the underlying atmospheric “dynamical core”  residing on a “cubed-sphere” grid.  The convection scheme has been optimized with the simulation of tropical storms as one of the key targets.  Details can be found in the link above.   The type of model we are referring to is best appreciated by watching an animation of part of the output of the model.  The model evolves the state of the atmosphere and the land surface in time, constrained only by the observed surface ocean temperatures and the seasonally varying pattern of solar radiation incident at the top of the model.  The temperature, water vapor, cloud, and wind fields are all defined at 32 levels in the vertical, ranging from the surface to the lower stratosphere.  The field shown here is the east-west component of the horizontal wind field at the model level nearest the surface (white => winds from the west; black => winds from the east; colors saturate at 33 m/s, the formal definition of hurricane strength winds).  The frames are 6 hours apart, and the entire loop covers the months of June-November from a year for which the prescribed ocean temperatures create a particularly active Atlantic hurricane season.

The model does very well at simulating the frequency of cyclones of hurricane strength (33 m/s) or greater.  In fact, it seems to do a bit better at this than in simulating the number of weaker vortices, perhaps because stronger vortices are more accurately counted in observations.  But very “major hurricanes” of intensity greater than 50m/s are not simulated, a deficiency that is not surprising for a model of this resolution.

Internannual variability and trend in the Atlantic:

When provided with the observed sea surface temperatures, the
model produces an impressive simulation of the observed year-to-year variations in Atlantic hurricane frequency, as shown below. The red line is the observed number of North Atlantic hurricanes for each year shown (from the IBTracs data set).  The gray shading indicates the range of hurricane counts generated by the C180HIRAM model in 4 different realizations of the model climate (the simulations differ only in the initial condition prior to 1980; the model is free running from these initial conditions and knows what year it is only from the prescribed ocean temperatures.) The blue line is the mean over these realizations. On average, the model produces about 17% fewer North Atlantic hurricanes than observed over this period.  We have multiplied all model results plotted by 1.17 so that the mean of the model results agrees with the observations.


This result, combined with similar encouraging results with a few other models around the world, indicates that aspects of the seasonal statistics of the Atlantic hurricane season are quite predictable, given the ocean temperatures.  It also suggests that the upward trend in Atlantic hurricanes over this period can understood given the trends in the global ocean temperature field. 

Continuing work with this model, partly described in Zhao et al (2009), suggests further that this trend is primarily a consequence of the relative warming of the tropical Atlantic with respect to the typical oceans as a whole that has occurred in this period.  This differential warming is unlikely to have been generated by increasing greenhouse gases in the atmosphere. (More to come)


comments welcome — Isaac.Held at