[see update at bottom of post]
A new article has just been published in the January 22, 2010 issue of Science magazine which finds that there will be a large increase in the frequency of the strongest hurricanes in the Atlantic basin as the climate changes from increasing anthropogenic greenhouse gas emissions. But a closer look at the results shows that this model-based result is produced by a hurricane model which under-simulates the frequency of strong storms in today’s climate. And that, despite the projected increase in intense hurricanes, the frequency of those storms projected by the model to occur by the end of the 21st century is considerably less than the frequency of intense hurricanes actually observed in the current climate. If the model doesn’t work for the present, why should we trust it for the future?
The Science article summarizes the research findings of hurricane modelers Morris Bender and colleagues from Princeton’s Geophysical Fluid Dynamics Laboratory and Old Dominion University’s Center for Coastal Physical Oceanography.
Since standard general circulation models (GCMs) are too spatially coarse to produce actual hurricanes, they can’t be used directly to assess how the characteristics of hurricanes (frequency, intensity, preferred tracks, etc.) may change under a changing climate warming. Instead, the storm changes generally must be inferred from our understanding of the processes of hurricane formation and development and how these processes may be impacted by the environmental changes projected to occur in the future (for example, sea surface temperature changes, vertical wind shear, etc.). But direct p;rojections of future hurricane behavior in have been limited.
Bender and colleagues attempt to overcome this limitation by using a series of models with progressively more spatial resolution and each better designed to model hurricane development. They took the output of future climate conditions produced by the GCMs, and fed that into a more detailed model which could create hurricanes, and then fed those hurricanes into an even more detailed model which could forecast the development, track, and intensity of the storms. Comparing the hurricanes created from environmental conditions at the end of the 21st century (assuming a mid-range emission scenario, SRES A1B), with those created from the actual conditions during the period 1980-2006, allowed the researchers to assess how the behavior of Atlantic hurricanes may potentially change as a result of anthropogenic influences on the earth’s climate.
In general they reported the following findings:
1) the total number of hurricanes forming in the Atlantic basin will decline in the future by about one-third;
2) the total number of very intense hurricanes (category 4 and 5) will increase in the future by 81% while the frequency of the most intense storms will increase by 250%;
3) the total damage in the U.S. from hurricanes will increase by about 30% as the increase in the frequency of intense storms more than makes up for the decline in the total number of hurricanes;
4) that changes in hurricane characteristics observed over the past 30 years don’t match well with their model expectations, sugegsting that “global warming: is not the cause; and
5) that most of the expected changes in hurricane characteristics will be undetectable until late until at least the second half of this century.
Of the above, the ones that are surely to get most of the attention are numbers 2 and 3—that the frequency of intense hurricanes is supposed to dramatically increase which will lead to more hurricane-related damages in the U.S.
However, these two findings seem to have the weakest legs.
First off, the Bender et al. models don’t do such a good job in simulating the known frequency of intense hurricanes. When they applied their models to the conditions characterizing the known period from 1980-2006, they simulated more than 50% fewer category 4&5 hurricanes than were actually observed (and more than 4 times fewer of the most intense hurricanes). Instead of 40 total category 4&5 storms that were observed during the 27 years from 1980-2006, their models simulated somewhere around 16 (depending on their particular model). When the Bender et al. models were run under end-of-the-21st-century conditions, they produced about 29 category 4&5 hurricanes. This is a 81% increase in the model world, but in fact, 25% below what the real world has produced under current conditions.
This major discrepancy is illustrated in Figure 1. The top panel in Figure 1 shows the tracks of all hurricanes that reached at least category 4 strength in the Bender et al. simulation of the 1980-2006 period. The middle panel of Figure 1 shows the category 4 and greater storms generated for the same period assuming end-of-the-century environmental conditions. The overall increase in the number of category 4 or stronger storms is quite obvious. We’ve added the bottom panel to Figure 1, which shows the tracks of the actual category 4 or stronger storms that occurred during the 1980-2006 period (mapped using the tool from the NOAA Coastal Services Center ). In fact there were more category 4&5 storms observed from 1980-2006 than were modeled to have occurred if conditions were as warm as projected by the end of the century. Clearly the Bender et al. models are not working so well in simulating intense hurricanes.
Figure 1. Tracks of all category 4 and 5 hurricanes simulated by Bender et al. for environmental conditions as observed from 1980-2006 (top) and for conditions projected at the end of the century (middle). The bottom chart shows the tracks from all actual category 4 and 5 hurricanes during the same 1980-2006 period (source: top and middle panel from Bender et al., 2010; bottom panel from data available from the NOAA Coastal Services Center).
So, perhaps an appropriate headline should read, somewhat paradoxically, “New climate models project 81% more intense hurricanes in the future, but 25% fewer than today.”
Is this a good thing or a bad thing?
And what does it mean for the hurricane-related damage estimates? Already the authors use a rather questionable assumption that future damages to the U.S. (from a particular category of hurricane) will change in the same proportion as the frequency of storms (of that category) in the entire Atlantic basin. But just because a storm exists in the Atlantic basin somewhere doesn’t mean that it will hit the U.S. And, the future preferred storm tracks may not be the same as the past storm tracks. So the assumption that future U.S. landfalls increase in the same percentage as the basin frequency seems like it is a bit flimsy and needs a bit more evaluation.
[update: Roger Pielke Jr. writes in to tell us that the ~30% increase in estimated damages from storm changes should be compared against the roughly 640% increase in hurricane damages estimated to take place by the end of the century from societal changes (assuming no changes in climate). In Roger’s view, when estimating future hurricane damages, societal changes dominate climate changes by a factor of about 20 to 1 (for more information about how Roger came to his conclusions, see here).]
The bottom line is that if a model doesn’t perform very well in a particular area of interest—in this case in simulating the observed frequency of intense hurricanes—then that model shouldn’t be considered a reliable estimator of how the characteristics of that area of interest will change in the future. And yet that is exactly what the Bender et al. article attempts to do. We at World Climate report, need a bit more convincing.
Bender, M.A., et al., 2010. Modeled Impact of Anthropogenic Warming on the Frequency of Intense Atlantic Hurricanes. Science, 327, 454-458.