On Wednesday at the annual fall meeting of the American Geophysical Union, along with Kerry Emanuel I’ll be presiding over a special session on hurricane Katrina. As part of that session I will be presenting preliminary results of a comprehensive update of a 1998 paper that Chris Landsea and I collaborated on titled, “Normalized Hurricane Damages in the United States: 1925-95″ (available here as PDF). On the update I am collaborating again with Chris, and also Joel Gratz from here at the University of Colorado.

First, some of the features of the updated analysis.

1. We have extended the dataset forward to 2005 from 1995 and back to 1900 from 1925, representing 35 years of additional data. (Our analysis of the period 1900-1924 is still in process.) NOTE! This data has not been peer reviewed and is preliminary. But it does give a sense of where we are at in the analysis and some early numbers.

2. We used updated inflation for all factors used in the normalization: inflation data (source: 2005 Economic Report of the President), population data (source: 2000 U.S. Census), and wealth data (Source: Bureau of Economic Analysis).


3. We have included 2005 storms damages as twice the current insured loss total, following the conventional practice of the National Hurricane Center (NHC). These numbers will be closely looked at, particularly Katrina. Generally, flood damage is not included in the hurricane damage tabulation, so how exactly to deal with New Orleans will be a subject of considerable attention. In what you see below, consider 2005 numbers preliminary, but are an effort to make an apples-to-apples comparison with the data 1900-2004.

4. We will be paying close attention to uncertainties in the data. In particular, we are investigating the possibility that pre- ~1960 data may significantly underestimate the actual losses for two reasons. One is that in this period there was very little or no public assistance provided after disasters (e.g., federal disaster assistance) and thus such costs would not have appeared in the totals during that period, yet they do appear in every large event today. Second, at some point before 1987 the NHC began using insured losses as the basis for calculating total losses, multiplying insured losses by a factor of two to arrive at a total estimate. This raises the possibility of a step function in the loss data when this practice began, meaning that earlier losses could in fact be significantly higher than we report, perhaps by as much as a factor of two. We are confident that the data since 1960 is fairly robust. Note that all of these errors work in the direction of magnifying losses earlier in the century. The possibility exists that we may continue to underestimate the hurricane loss potential

Some Data

Top 10 Damaging Storms 1900-2005 (note 1900-1924 incomplete)

1. 1926 Great Miami $129,700,000,000
2. 2005 Katrina $80,000,000,000
3. 1900 Galveston $53,100,000,000
4. 1992 Andrew $50,800,000,000
5. 1915 Storm 2 $50,200,000,000
6. 1938 New England $35,000,000,000
7. 1944 Storm 9 $34,300,000,000
8. 1928 Lake Okeechobee $29,600,000,000
9. 1965 Donna $23,900,000,000
10. 1969 Camille $19,200,000,000

Note that Katrina I ranked number 2 here. Even with a larger estimate for Katrina this would hold. The 1926 Miami storm estimate may be conservative for reasons discussed above. Note that none of the 4 storms from 2004 or Wilma appears in the top 10. Andrew is ranked 4th. Andrew and Katrina are the only storms in the top 10 that occurred in the past 35 years.

Top 10 Years 1900-2005 (note 1900-1924 incomplete)

1. 1926 $141,400,000,000
2. 2005 $100,000,000,000
3. 1900 $53,100,000,000
4. 1992 $52,500,000,000
5. 1915 $52,200,000,000
6. 1944 $45,900,000,000
7. 2004 $45,100,000,000
8. 1938 $35,000,000,000
9. 1954 $32,700,000,000
10. 1928 $29,600,000,000

Looking at yearly damages, 1926 is number 1 with 2005 not far behind. 2004 is in 7th place and 1992 is in 4th. The next most recent year in the rankings in 1954.

Top 10 10-Year Periods 1935-2005

1. 1935 $193,300,000,000
2. 2005 $175,700,000,000
3. 1947 $125,500,000,000
4. 1950 $114,400,000,000
5. 1951 $112,600,000,000
6. 1952 $110,600,000,000
7. 1949 $107,700,000,000
8. 1953 $107,400,000,000
9. 1946 $106,000,000,000
10. 1955 $103,000,000,000

When we look at the cumulative damages over 10-year intervals 1935-2005 we see the effects of many large events in the 1940s and 1950s. When we eventually extend the database back to 1900 comprehensively, I expect to see additional years in this listing around 1926 and before 1910. The take-home point here is that single large events are not the only way to look at hurricane losses, but decadal periods with consistently high losses. The 1980s and 1990s appear at the bottom of such a ranking. Our thinking about hurricane damage is shaped by our experiences in an anomalous several decades of very low losses when considered in historical perspective.

Differences between 1998 and 2005 analyses, and a look to the future

Across the board the average per-storm increase in losses between our 1998 study and the present effort is 104%. Inflation accounts for about 13% of this increase, meaning that losses have increased in real terms by 91% or in other words, have just about doubled.

Consider this, if the pace of population growth and increasing wealth continues at these rates until 2020, then in 2004 dollars (i.e., today’s dollars) we should expect a repeat of the 1926 storm to cause about $500 billion in damages in 2020, all else being equal. Katrina would be $320 billion, Andrew $200 billion. These numbers boggle the mind, but provide some indication where we are headed.