In the last post we looked at counts of weather-related natural catstrophes according to Munich Re, the giant re-insurance corporation. It was mentioned that part of the increasing trend could be due to changes in the way catastrophes are counted, and in the number of people and value of property which is susceptible to such catastrophes. Although Munich Re adjusts their classification each decade to account for this, any such attempt is bound to be imperfect.
But it was also pointed out that Munich Re tracks catastrophes which are not weather-related (earthquake, tsunami, volcanic eruption), classifying them as “geophysical.” Do these geophysical events also show a trend? If so, are they increasing as fast as weather-related catastrophes?
Munich Re provides a nice graph of the number of all catastrophes, including geophysical:
Geophysical catastrophes are shown in red, the bottom part of each bar. There’s no visually obvious trend, but that’s a question which shouldn’t be decided on the basis of visual inspection so I digitized this graph to get some data. It turns out that there is a trend in geophysical catastrophes, and it’s statistically significant:
The overall rate of increase is only 0.8 per year per year, but it’s not zero. In fact, the overall trend is not linear, a quadratic trend giving a much better fit:
It’s not impossible that entirely natural changes have caused significant decadal variation in geophysical catastrophes. It’s not impossible that global warming has done so! The disappearance of landfast ice caused by man-made climate change affects tectonic stresses, which could impact the likelihood of earthquakes, tsunamis, even volcanic activity.
But it’s also possible that the observed changes are due to demographic/financial factors, the number of people and value of property vulnerable to catastrophes, or due to the way catastrophes are classified and counted. It would be nice to be able to “normalize” catastrophe counts to remove such influences, but properly accounting for changes in population, property value, building codes, advances in early-warning and in recovery-relief technology, even political changes which can impact disaster preparedness and response, is a very thorny issue indeed. When such normalizations are done by those with an agenda, they are too easily skewed one way or the other by the preferred outcome.
But there is a way to normalize weather-related catastrophe counts which does not depend on assumptions about the impact of all those complex factors. It is, to compare the changes in weather-related catastrophes to those which are not weather related. Just this approach was taken by Munich Re itself, as shown in this oustanding video; at the 5:35 mark you can hear the voice of Peter Hoppe from Munich Re, and at 6:00 he directly compares weather-related to geophysical catastrophes, noting that “… they have gone up by far more, and we only can explain this by a change also on the hazard side, which means a change in the frequency and intensity and as you have seen in the slide before we definitely have this.”
Essentially, we could use the counts of geophysical catastrophes to normalize the other types of catastrophe. Just as an experiment, let’s do so — and see what happens.
I decided to “normalize” the weather-related catastrophes by a smoothed estimate of the number of geophysical catstrophes. How to smooth? I’m very fond of my “modified lowess smooth” and also of the plain-vanilla lowess smooth. But there are other ways too, for instance something as simple as a Gaussian smooth (a moving average with Gaussian weights). Here are smoothed values using each of those three choices:
The Gaussian smooth shows less variation at the endpoints because instead of continuing the trend, it tends to do the opposite. Since it’s more “conservative,” showing less variation, I adopted the Gaussian smooth for normalization. I divided the smoothed values by their average so that the normalization function would have a mean value of 1:
Then I divided the value of each of the three types of weather-related catastrophes by this function to define normalized counts for weather-related catastrophes per year. And here they are:
All three classes of weather-related catastrophes show an increasing trend. All three trends are statistically significant. Of course the rates of increase are slightly less than when using non-normalized data. Catastrophic storms have increased by about 5 per year. Catastrophic floods/mass movements by about 6.6 per year. Heat/drought/fire catastrophes about 2.5 per year. For total weather-related catastrophes we’re seeing an extra 14 catastrophes per year:
Even when normalized by geophysical counts, the number of weather-related catastrophes has more than doubled in a mere 30 years.
I’m certain that some people will find an excuse to “explain away” this rather obvious trend. That’s what those in denial do.
I’ll agree with Peter Hoppe, that “… we only can explain this by a change also on the hazard side, which means a change in the frequency and intensity …”