01 November 2012

Normalized US Hurricane Damage 1900-2012, Including Sandy

The graph above shows normalized US hurricane damage, based on data from ICAT, which applies an extension to the methodology of Pielke et al. 2008. The 2012 estimate for Sandy comes from Moody's, and is an estimate.  The red line represents a linear best fit to the data -- it is flat.


  1. I've been closely following your line of thinking on this for some time. At first, I was very sceptical - everyone seems to be reporting the increase in storminess. The Moody's link says 'If the $50bn total turns out to be accurate, it will make Hurricane Sandy the fourth costliest US disaster behind Hurricane Katrina, the attacks of September 11, and Hurricane Andrew.' But of course they aren't normalising to PL05. Which brings me to my question.

    How are you arriving at the PL05 normalised Sandy estimate of $30B? The EQECAT insured loss estimate of $10-20B? It is interesting to think about relative wealth per capita on relatively wealthy Long Island vs Hoboken. To close my opening thought, i was originally very sceptical but it is almost impossible for a layman to see any underlying numbers for the repeated headlines of '100 year storm strikes two years in a row!' Which makes me yearn for hard numbers such as yours.

  2. I have two complaints about this graph.

    1) I'd expect that (proportionally) more infrastructure was built along low lying areas in 1900 then it was now. This mainly because I expect that we relied more on shipping for trade and transportation than we do now. So, I would expect more (normalized) damage to be done in 1900 than in 2012 by an equal sized storm. Of course, this may not be the case as I am not an expert. Nonetheless, you need to argue that normalized hurricane damage is a good way to measure storm strength (since this is what this is really about), and I'd be surprised if it is.

    2) I'm not sure why the fact that it is flat is surprising. If a line were drawn with much positive slope at all, the line would cross the x-axis and be negative when evaluated at 1900. Maybe a linear fit is not the way to go?

    Can you address these two complains? Please justify why you expect hurricane damage to be proportional to strength or acknowledge that this is not likely to be the case, and please explain why a linear fit is suitable to analyze this data.

  3. -2-Unknown

    Thanks for the questions, some replies:

    1. The method is not designed to say anything about storm strength, so no worries there. There is good data since 1900 on storm frequency and intensity, so look there for trends in storms.

    2. The linear fit covers 1900-2012 and is a simple least squares regression. It is not to be extended forward or backward.


  4. You probably don't care, but in case you hadn't noticed Phil Plait smeared you as a "Climate Change Denier" when this post was linked to in his recent blog posting linking climate change to Hurricane Sandy: rpielkejr

  5. -4-Morgan Holt

    Thanks, but you are correct, I don't care;-)

  6. Roger, the GDP adjustment to convert 2005 dollars into 2012 dollars is approximately 1.15. I notice in the graph above, 1926 normalized damages have increased by greater than that 1.15 ratio relative to the values reported in Pielke et al, 2008. In contrast, 2005 normalized damages have increased by less than the GDP adjustment. Can you confirm that this difference is because population (and I presume property prices) have continued to rise in Miami, whereas they have fallen relative to pre-Katrina levels in New Orleans?

  7. -6-Tom Curtis

    Thanks ... The post-2005 adjustments comes from ICAT extending our methods, so I'd point you to them for exact details.

    I too noticed that Katrina was essentially unchanged from 2005, which makes sense. In the past we used a simple annual adjustment everywhere to extend the results, but a more tailored approach makes a lot of sense. We will probably do another comprehensive update around 2015.


  8. It seems to me that the damage in dollars caused by a hurricane will change over time due to a variety of reasons. One of the biggest variables (in my opinion) would be the trajectory of the storm. As each of these storms have different paths, the damage in dollars is not necessarily a function of the size or wind speeds of the storm.

    Also, I'd expect as a function of time the damages from each storm to decrease. Between 1900 and now we have much better building codes and materials, and know much more about how to design buildings to withstand large storms. There is also an economic incentive to not rebuild where a storm just hit, so two equivalent storms a few years apart probably wouldn't do the same damage in dollars.

    Is there another way of comparing hurricanes that might more accurately represent their size/strength? Basically, did they have any way of knowing how big the storms were in the 1920's?

  9. It also looks like one outlier can skew the entire graph. Is there a trend if you exclude the top 5% and bottom 5% damaging hurricanes?