by Roger Pielke
In the next chapter I’ll discuss some of these regional studies for specific phenomena in greater detail.
5. A Normalization of the EM-DAT Dataset
In 2014 researchers in the Netherlands and the UK published a paper which looked at a dataset kept by a Belgian research group. Their results, using a different dataset than that which was the focus of the other four papers reviewed above, were nonetheless very similar to those other studies.
Specifically, once the data was normalized, the researchers concluded:
The absence of trends in normalized disaster burden indicators appears to be largely consistent with the absence of trends in extreme weather events.[63]
The paper concluded that the lack of trends in extreme weather events and normalized losses indicates that, overall, vulnerability to losses has been constant over time. The lack of a detectable change in vulnerability at the global level does not preclude changes in vulnerability at more localized contexts, only that any such signal is not detectable at the global level once data has been normalized to account for changes in exposure to loss.
Conclusion
At the global level, it should be clear that the available evidence provides no support for claims that disaster losses have been increasing due to climate change, whether those changes are human-caused or not. Once societal factors are taken into consideration, there is no residual trend. In the language of the IPCC, detection has not been achieved. There is consequently no remaining increase in losses to be attributed to any factors beyond the various societal factors which lead to increasing disaster losses.
Given these studies, it should not come as a surprise that the IPCC’s 2012 special report on extreme events came to exactly this conclusion, which was based on regional studies as well as the global studies reviewed here:
Long-term trends in economic disaster losses adjusted for wealth and population increases have not been attributed to climate change, but a role for climate change has not been excluded (medium evidence, high agreement).[64]
Its 2014 report on impacts and vulnerability reinforced these conclusions:
· “Economic growth, including greater concentrations of people and wealth in periled areas and rising insurance penetration, is the most important driver of increasing losses.”
· “Apart from detection, loss trends have not been conclusively attributed to anthropogenic climate change; most such claims are not based on scientific attribution methods.”[65]
These conclusions are very strong (even though they also continue to fall back on the meaningless assertion that the negative has not been proven). Let’s now take a closer look at some of the various regional studies for the U.S. and the world, which will reinforce these conclusions at a more localized level.
5
Heat, Rain, Hurricanes, Floods, Tornadoes, Drought, Oh My!
This chapter surveys, in a rather dry and academic manner, some of the literature on climate change and extreme events with a focus on detection and attribution. I rely heavily on the recent assessments of the IPCC, which has done considerable work to assess the relevant literature. However, those passages relevant to the IPCC’s discussion of extreme events are spread over dozens of chapters in multiple reports. This section aims to provide a summary which is perfectly consistent with the findings of the IPCC. Consequently, I reproduce a number of extended excerpts from its reports.
Specifically, this chapter focuses on the following phenomena:
· Extreme heat
· Extreme precipitation
· Tropical cyclones (hurricanes)
· Floods
· Tornadoes
· Drought (including Australian bushfire)
This chapter covers a lot of territory, but only scratches the surface of what is available in the primary literature.
In 2011 Dutch researcher Laurens Bouwer wrote a review paper summarizing much of the literature on disaster losses and climate change available at that time. That review paper concluded:
The analysis of twenty-two disaster loss studies shows that economic losses from various weather related natural hazards, such as storms, tropical cyclones, floods, and small-scale weather events such as wildfires and hailstorms, have increased around the globe. The studies show no trends in losses, corrected for changes (increases) in population and capital at risk, that could be attributed to anthropogenic climate change. Therefore it can be concluded that anthropogenic climate change so far has not had a significant impact on losses from natural disasters.[66]
This conclusion should by now be familiar. But let’s now take a closer look at some of the studies of specific phenomena at the regional level. Those wanting to explore further are directed to Bouwer’s paper, and to the more recent literature summarized by the IPCC.
In the language of the IPCC, at the global scale detection of increases in extreme heat and extreme precipitation has been achieved. These increases have also been attributed to human causes, specifically to growing concentrations of greenhouse gases in the atmosphere. Of course the IPCC expresses all of its conclusions not with 100% certainty, but accompanied by a judgment call about the confidence they have in their conclusions.
With respect to tropical cyclones, floods, tornadoes, and drought, neither detection nor attribution has been achieved at the global scale. At selected regional and local scales there is evidence of trends over various time periods, which of course would be expected as the Earth’s climate system is highly variable. Attribution of small-scale, short-time period trends to anything other than variability of the climate system remains a challenge. Let’s now look at some of the IPCC’s findings and some of the underlying data for these various phenomena.
Extreme Temperature and Precipitation
Extreme temperatures and precipitation are not big drivers of disaster losses. There are nonetheless important phenomena with significant impacts on people and ecosystems, and they have been changing.
With respect to extreme temperatures, the IPCC Fifth Assessment Report (AR5) concludes:
[T]here is medium confidence that globally the length and frequency of warm spells, including heat waves, has increased since the middle of the 20th century although it is likely that heatwave frequency has increased during this period in large parts of Europe, Asia and Australia.[67]
In the dry prose of the IPCC “medium confidence” is certainly not as sensational as some characterizations of the IPCC conclusions. For the U.S., the IPCC concluded with respect to heat waves: “Medium confidence: increases in more regions than decreases but 1930s [Dust Bowl] dominates longer term trends in the USA.”
With respect to attribution, the IPCC surveys a large number of modeling studies which try to disentangle human forcing of the climate system from ongoing climate variability. The AR5 concludes from this research:
[N]ew results suggest more clearly the role of anthropogenic forcing on temperature extremes compared to results at the time of the SREX assessment. We assess that it is very likely that human influence has contributed to the observed changes in the frequency and intensity of daily temperature extremes on the global scale since the mid-20th century.[68]
In the jargon of the IPCC, “very likely” means that they are expressing a likelihood value of at least 90% probability that this claim is correct.[69]
With respect to extreme precipitation the IPCC’s conclusions are not nearly as strong as those for extreme temperatures. The IPCC concludes:
[I]t is likely that since 1951 there have been statistically significant increases in the number of heavy precipitation events (e.g., above the 95th percentile) in more regions than there have been statistically significant decreases, but there are strong regional and subregional variations in the trends.[70]
By “likely” the IPCC means that there is at least a 66% likelihood that this particular claim is correct. The IPCC thus judges that there is a two in three chance that there have been increases in heavy precipitation in more locations than have seen d
ecreases.
Given the large uncertainties in detection of trends, it is therefore not surprising that the IPCC expressed limited confidence in attribution:
[T]here is medium confidence that anthropogenic forcing has contributed to a global scale intensification of heavy precipitation over the second half of the 20th century in land regions where observational coverage is sufficient for assessment.[71]
The IPCC expressed “high confidence” in its conclusions that precipitation extremes had “very likely” increased in central North America, its strongest conclusion for any region.
Such nuanced, carefully expressed conclusions in shades of grey do not lend themselves to effective translation when politics deals in black and white.
Are Extreme Precipitation and Flooding the Same Thing?
A common confusion is that an increase in “extreme precipitation” necessarily implies or is directly associated with an increase in flooding. This is incorrect.
Setting aside uncertainties in detection and attribution and postulating that extreme precipitation has increased and can be attributed in some part to greenhouse gas emissions does not lead automatically to findings of corresponding increases in streamflow (floods) or damage.
How can this be?
Think of it like this: Precipitation is to flood damage as wind is to windstorm damage. It is not enough to say that it has become windier to make a connection to increased windstorm damage—you would need to show a specific increase in those specific wind events that actually cause damage. There are a lot of days that could be windier with no increase in damage; the same goes for precipitation.
Even though there have been increases in what scientists call “extreme precipitation” there is very little evidence to suggest that these increases have been accompanied by increasing floods. This is a robust finding across the literature, and further details are provided below in the discussion of floods.
Absent an increase in peak streamflows caused by increasing extreme precipitation, it is impossible to find a causal linkage between increasing precipitation and increasing floods, much less between precipitation and flood damage. There are of course good reasons why a linkage between increasing precipitation and peak streamflow would be difficult to make, such as the seasonality of the increase in rain or snow, the large variability of flooding, and the human influence on river systems. Those difficulties of course translate directly to a difficulty in connecting the effects of increasing greenhouse gases in the atmosphere to flood disasters.
Let’s now turn to the most damaging events, tropical cyclones, which are also among the most studied type of extreme event.
Tropical Cyclones
Hurricanes are “tropical cyclones” that occur in the North Atlantic and in the Eastern Pacific, often off the coast of Mexico. Tropical cyclones have different names in other parts of the world, like typhoon or cyclone, but they are all the same phenomena. Tropical cyclones are responsible for some of the greatest economic and human impacts from any type of extreme event. In 2005 Hurricane Katrina led to damage of more than $80 billion and in 1991 a tropical cyclone led to more than 138,000 deaths in Bangladesh.[72]
In a recent paper we found that U.S. hurricanes are responsible for almost 70% of the overall increase in disaster losses since 1980 in the Munich Re global loss dataset.[73] Consequently, if we can understand what is behind that increase we will have explained a majority of recent increases in the costs of disasters.
In 2008, a team of six researchers (including me) published a paper that asked an apparently simple question: If each hurricane season of the past took place with the level of development on the nation’s coasts of 2005, how much damage would have occurred in each year?
Figure 5.1: Normalized U.S. Hurricane Damage (1900-2013)
Source: Updated from R. Pielke, Jr., J. Gratz, C. Landsea, D. Collins, M. Saunders, and R. Musulin, “Normalized Hurricane Damage in the United States: 1900-2005,” Natural Hazards Review 9 (2008): pp. 29-42.
To answer this question we took a time series of loss data kept by the U.S. National Hurricane Center from 1900 and adjusted it for inflation, a measure of household wealth, property in coastal counties and population.[74] We used two different methods which arrived at substantially similar results. The graph above shows the results of this study, updated through 2013.[75] We call the adjusted data “normalized hurricane damage” to reflect the fact that it has been adjusted to a common base year.
The graph includes Superstorm Sandy, which by some measures was not technically a hurricane but a “post-tropical cyclone of hurricane strength.” There were also three other storms since 1900 which made landfall as “post-tropical cyclones of hurricane strength” which occurred in 1904, 1924 and 1925. We don’t have loss data for these three storms, so they appear in the dataset as placeholders at $5 billion each.[76]
Sandy made 2012 a bad year, but since 1900 we estimate that 8 other years would have had greater damage. The most costly year, in terms of normalized damage, was 1926 with more than $200 billion in estimated damage. This was mainly from the Great Miami Hurricane, which would devastate Miami today.
But how do we know if our estimates are any good? Maybe our methods are flawed or we have neglected to include important variables, like changing building practices, as described in Chapter 2.
We perform several independent checks on the analysis. One is simple. We know which years had exceptionally large losses: 1900 and 1915 in Galveston, 1926 and 1928 in Florida, 1938 in New England, Hugo in 1989, Andrew in 1992, and so on. These years show up clearly in our dataset with big losses.
A far more sophisticated check is to compare trends in the incidence of hurricanes with trends in damage. Because counts of hurricanes and measures of their strength are independent of damage estimates, they can serve as a basis for evaluating the appropriateness of our adjustments. Logically, we would expect that trends in normalized damage and trends in hurricane incidence would go in the same direction. It turns out that they do match up, almost perfectly.
Figure 5.2: U.S. Hurricane Landfalls (1900-2013)
Source: NOAA/NHC.
The graph above shows a count of U.S. hurricane landfalls from 1900.[77] It shows no evidence indicating that hurricane landfalls have increased since 1900, a finding that holds if one starts the analysis in 1851 (when NOAA’s dataset begins) or 1950. There is an upwards trend if the count is arbitrarily started in 1970, the lowest period of activity since 1900.
But that is storm frequency. What about storm intensity, in particular, the strength of storms when they make landfall?
The graph below shows the U.S. landfall intensity data for 1900 through 2013. There is no upwards trend since 1900, consistent with the trends in normalized losses. There is similarly no upwards trend in the data since 1950, but there is if the analysis is started in 1970, as is the case with landfall frequency.
Figure 5.3: Intensity of U.S. Hurricane Landfalls by Year (1900-2013)
Source: NOAA, provided courtesy of C. Landsea, NHC. The vertical axis is expressed as an index.
The data show that hurricanes have not increased in the U.S. in frequency, intensity, or normalized damage since at least 1900. The trends across these three datasets match up well. Based on this match here is what we concluded in our 2008 paper:
The lack of trend in twentieth century normalized hurricane losses is consistent with what one would expect to find given the lack of trends in hurricane frequency or intensity at landfall. This finding should add some confidence that, at least to a first degree, the normalization approach has successfully adjusted for changing societal conditions. Given the lack of trends in hurricanes themselves, any trend observed in the normalized losses would necessarily reflect some bias in the adjustment process, such as failing to recognize changes in adaptive capacity or misspecifying wealth. That we do not have a resulting bias suggests that any factors not included in the normalization methods do not have a resulting net large significance.[78]
<
br /> But what about the rest of the world?
The IPCC SREX report concluded “There is low confidence in any observed long-term (i.e., 40 years or more) increases in tropical cyclone activity (i.e., intensity, frequency, duration), after accounting for past changes in observing capabilities.”[79] The IPCC AR5 reaffirmed this conclusion:
Current datasets indicate no significant observed trends in global tropical cyclone frequency over the past century…. In summary, this assessment does not revise the SREX conclusion of low confidence that any reported long-term (centennial) increases in tropical cyclone activity are robust, after accounting for past changes in observing capabilities.[80]
The IPCC did find that tropical cyclone activity has increased in the North Atlantic since 1970. As shown above, U.S. normalized losses, landfall frequency and intensity have also increased in the U.S. over this period. However, the IPCC concluded that these increasing losses are a matter of the choice of the start date for the analysis, as the trends since 1970 do not exceed documented variability:
No robust trends in annual numbers of tropical storms, hurricanes and major hurricanes counts have been identified over the past 100 years in the North Atlantic basin.[81]
The IPCC SREX agrees with respect to observed damage:
Most studies related increases found in normalized hurricane losses in the United States since the 1970s (Miller et al., 2008; Schmidt et al., 2009; Nordhaus, 2010) to the natural variability observed since that time (Miller et al., 2008; Pielke Jr. et al., 2008). Bouwer and Botzen (2011) demonstrated that other normalized records of total economic and insured losses for the same series of hurricanes exhibit no significant trends in losses since 1900.[82]
