Tag Archives: Extreme Weather Events

Explaining Extreme Events of 2013: Limitations of the BAMS Report

The American Meteorological Society, in it’s Bulletin of the American Meteorological Society (BAMS), has released a report called “Explaining Extreme Events of 2013 from a Climate Perspective.” Three studies looked at excessive heat in Australia, three at drought or dry conditions in California, and 14 looked at various other extreme events (though some of those events may overlap or be related) for a total of about 15 different phenomena.

There was a pattern in the results. The studies looking at heat all suggested a link to anthropogenic global warming (AGW). This is not surprising because AGW has involved a global increase in average temperature which is manifest across a variable climate, so even a modest increase in global temperature, bunched up in to places that are a bit cooler or warmer than average (at a given moment in time) is going to be blatantly obvious when picking out heat events. Some of the studies that looked at the California drought and drought in New Zealand attributed these conditions to climate change, others were more ambiguous or suggested that there was no link. All of the studies that looked at extreme precipitation events concluded that there was no way to make a connection, except one (in Northern India) which as ambiguous.

Michael Mann has pointed out that there is a basic problem with the BAMS study. Many of the extreme climate events of recent years have been linked by various researchers to climate change, but a) none of these researchers seem to have been invited to contribute to this collection of papers, and b) some of those specific events have been linked to climate change by some of those researchers.

It might be tempting to view this volume as an authoritative statement by the scientific community on the role climate change may or may not have had in some high profile, devastating recent extreme weather events. But that would be misguided. The BAMS special issue is not a representative, community-wide scientific assessment like those published by the National Academy of Sciences or the Intergovernmental Panel on Climate Change. The editors, instead, have solicited contributions from a relatively small number of groups, so the findings do not necessarily reflect the range of views of the broader scientific community. Some leading climate scientists who were not included in the effort have presented evidence of a greater role for climate change in several of the events dismissed or downplayed by the BAMS articles (see e.g. Kevin Trenberth of NCAR on the September 2013 Colorado floods, Stefan Rahmstorf of the University of Potsdam on the June 2013 Central European floods and Jennifer Francis of Rutgers on the 2013/2014 California Drought).

I’ve provided references and links to the studies mentioned by Mann, as well as other studies, below.

Mann points out, and I had noticed this when I first read the BAMS volume, that the studies that looked at extreme precipitation pretty much leave out the mechanism implicated by the above mentioned researchers. In fact, I would say that the basic methodology used to examine these events is flawed in two ways. Before describing that, here is summary information about the studies.

The spatial and temporal context of the studies, and attribution

In order to evaluate the studies in the BAM report, I ranked attribution from 1 to 5. 1 means no effect of climate change at all, 5 means climate change is a major contributor or THE explanation. 3 is the nickpoint; a 3 means maybe maybe not, or serious ambiguity. So 4 and 5 are yes, climate change mattered, 1 and 2 are no, climate change did not matter (but 1 is more strongly stated) and 3 means you can’t say but maybe.

The four studies that look at dryness and drought (including more than one for California so these are not all independent data in that respect) had attribution vales of 2, 3, 4, and 4. The minimum surface area of the climatic events evaluated was about 268,021 km2, and the maximum about 423,970 km2. These events are generally long term. Drought or dry periods are large and slow moving things, and the period of time over which they happen ranges from several months to years. Notably, some meteorologists such as Paul Douglas have noticed a shorter term event, which Douglas calls a “Flash Drought,” a period of little or no rain where there usually is some rain lasting for several weeks, sufficient to disrupt crop growing but not sufficient to lead to long term effects such as depletion of ground water supplies.

The 10 studies that looked at heat (all but one excessive heat, one cold) had attribution values of 1 (the cold in UK), 4, 4, 5, 5, 5, 5, 5, 5 and 5, with a surface area ranging from about 220,000 km2
to 133,453,480 km2. Heat waves usually cover large areas, but can be very short term, lasting several days to several weeks.

The 8 studies that looked at extreme precip (most rain, one snow) had attribution values of 1,1,1,1,1,2,3, and 3 … no case was attributed unambiguously to climate change, most not at all. These areas covered a region of between 27,980 km2 and 8,080,464 km2, but most clustered near the lower end of that range. Most extreme precipitation events cover small areas (though the extreme rain experienced during the summer of 2014 in North America may have been a physically very large event running form the Upper Plains to the Ohio Vally, and beyond). Extreme rain and snow storms normally occur over a matter of several days.

Spacial-temporal bias in the BAMS study

The size and lifespan of the events under consideration is the best single predictor of level of attribution given by the individuals studies. For the most part, relatively small short term events were not attributed to climate change, while large slow moving events were more often attributed to climate change. Exactly parallel to this is, of course, the nature of the event. It isn’t just the size and lifespan of the event, but the kind of event that matters, which in turn relates to the size and time frame. But, the size and lifespan of the precipitation events may be part of the explanation for why attribution is low.

This is why. The studies that looked at precipitation used a number of different approaches but for the most part they had the same characteristics. Underlying the application of the various analytical techniques is the question of probability. If a certain kind of event (a large amount of rainfall in one place over a contiguous number of days) is rare, and climate change makes is somewhat less rare, it may be impossible to detect this probabilistically unless the sampling is done right. It is difficult to measure, with statistical confidence, the difference between something that is very very rare vs. merely very rare. Since the studies essentially tried to do this (for the most part), it is not surprising the study results did not attribute those events to change over time.

It may the case that a year by year study of changes in probability of rare events will not detect a change until the change is huge.

And now for a brief thought experiment.

Imagine for a moment that all climate events are the result of the distribution and behavior of small imaginary objects that float around in the air. We’ll call them climaticulus. Climaticules have two attributes: temperature and moisture. Unspecified processes alter these attributes. If enough climaticules in a region are dry for long enough, you might get a drought. If enough are wet, you might get lots of rain or snow. If a lot are warm, you might get a heat wave.

With the climaticule thought-model, variation in two attributes can generate a spatially large long term event (like a drought) or a spatially concentrated short term event (like flooding rains). At the surface, these events look like qualitatively distinct events, but underlying them is a simple system. In this system, the smaller and shorter term events are going to present statistical distributions that are different than the statistical distributions of the larger scale events. Even if they are the same kind of distribution, they will be scaled very differently. It is quite possible that a numerical change in one category of event will remain invisible while others are latent, given similar basic approaches such as “what happened over a year’s time” or “what happened in a particular space.”

That thought experiment may or may not have been helpful, but I can put it another way: Under climate change, the climaticules are sending out a signal that results from changes in average temperature and moisture. When the signal comes as a large slow moving event, it is hard to miss. When it comes as a small ephemeral event, it is easy to miss.

Mechanistic bias in the BAMS study.

As pointed out by Mann (see above), the BAMs study fails to consider an already proposed and reasonably well supported mechanism for increased occurrence of extreme precipitation events. This is the change in the patterns of trade winds and jet streams that seems to result from regionally increased sea surface temperatures and/or relatively more warming in the Arctic.

Under normal conditions, in the northern Temperate zone, air masses move from west to east, between two jet streams. The jet streams guide the air masses and the air masses push around the jet stream … they can be thought of as two aspects of the same system that arises from a a combination of the Earth’s rotation and the movement of heat from equatorial regions north towards the pole.

Under these conditions, the air masses vary in barometric pressure and moisture, and this variation causes rainy atmosphere to move at the large scale from west to east at a fairly rapid clip. So if you are sitting there in Iowa, it might be sunny mid day, than a front comes through bringing some storms, then the sky clear again, over several hours. Or, you might get a larger, wetter, air mass coming along that brings a day of variable amounts of rain.

It is thought that recently it is more common for the jet stream to for giant curves, which relates to the temperate air masses bulging northwards or being pushed southwards. The jet stream slows down. So, air masses that might bring precipitation are either blocked from their west to east movement or move very slowly (and in a somewhat different direction) than they normally would. So, a stormy, wet, rainy air mass may take two or three times longer to move across a given region, causing much more rain to fall there. At the same time, other regions may experience long term lack of precipitation. This is all further complicated by the changes in where the air mostly comes from and goes to, allowing some air to be dryer than usual and other air to be wetter than usual. And, because of AGW, the air is on average warmer so it can hold more moisture.

So, you get a bunch of extra moist air arriving in a place where contact with colder air masses and changes in pressure cause it to be rain-producing, and it sticks around for several days in one spot, or moves very slowly, and you end up with flooding like we saw in Calgary, or Boulder, or long periods of continuous storm formation and rainfall over a large area like we had in the upper Midwest in June of 2014.

What was the jet stream doing for each of the studied extreme events?

Focusing only on temperate regions in the US and Europe, I assembled a set of wind stream maps (from here) that more or less show the behavior of the jet stream at the time of each event. These are a little hard to read but you will notice that the location of the event is in every case up against a wavy part of the jet stream at the time. For reference, I took one of the events, September 13th, and picked out several earlier wind stream maps (every five years for several decades in the past). This is not a systematic sample, but it shows that typically the jet stream, not so long ago, was flatter (probably) than it was during these extreme events. These graphics are all pasted below, and I’m sorry if it takes a while for them to load.

The examination of the relationship between climate change and extreme weather events is tricky, in in its infancy, but we are beyond the point where we should be ignoring emerging research pertaning to the link. The BAMS report ignores that research. Also, the examination of extreme precipitation events should be done in larger blocks of time than one year. Looking year by year, and event by event, almost guarantees not making the link because of the bias in spatial and temporal features of these events.

Colorado storm, September:

Screen Shot 2014-10-03 at 11.28.48 AM

South Dakota Blizzard, October 4-5:

141003163023

Wet southern European winter, 2013:

141003163409

Heavy precipitation, May-June, Upper Danube and Elbe Basins:

141003163536

Extreme snow, Western Spanish Pyrenees:

141003163717

Violent Storm in Northern Germany and Denmark, 28 October:

141003163820

For comparison, September 13th from several prior years:

1970:

141003164506

1975:

141003164518
1980:
141003164537
1985:
141003164547

Prior work related to climate change and extreme weather events:

Linking Weather Extremes to Global Warming
More Research Linking Global Warming To Bad Weather Events
Is Global Warming Behind the Polar Vortex?
Extreme Jet Stream Pattern Triggers Historic European Floods
Are you ready for more floods and wildfires?
Drunken Arctic Goes Head Over Heels