There are few different, related, ways in which climate change, including anthropogenic global warming, can cause extreme weather events. One is that climate zones move. This may result in “normal” weather for a different location occurring elsewhere. For example, if southern warm air system shift north, than the frequency of low and high temperatures, and their distribution throughout the year, can change. Another is the rise of entirely new conditions that were previously either rare or virtually unknown. One example of this might be the steering of Hurricane Sandy into the northeastern U.S. coast a couple of years ago. Hurricanes do plow into that region now and then, but they almost always come from the south and bump into land in the narrowing North Atlantic. Sandy did something different, moving north out at sea in the Atlantic, like many Atlantic hurricanes do, but then making an abrupt left turn, owing to an unusual configuration of the atmosphere, plowing into Connecticut, New York and New Jersey. That was a single unusual event, to be sure, but if such air patterns become “normal” (even occurring only every few years rather than almost never during hurricane season), that would be a qualitative shift in weather patterns. If that shift is caused by the phenomenon of Arctic Amplification (the relatively increased warming of the Arctic as the entire planet warms) that would be a shift in the kinds of weather patterns we have due to global warming. A third kind of change is what is often called “loading of the dice.” This is where events that have a low probability of happening simply happen more often. The dice analogy is tricky because it is often used differently by different people; one idea is that a rare event is rolling two sixes with two die. That would be rare. But climate change adds one or two more die, allowing for a greater chance of two of them coming up as a six. That’s a difficult analogy because there really isn’t an equivalent to extra die in the climate system. The point here is that probabilities of rare events changes.

My distinction between zone shifting, qualitative shifts, and probability changes is not something climate scientists or meteorologists generally say; this is just my way of talking about changes in variance of weather patterns. Also, these three different things are not necessarily that different, but rather, three faces to the same multi dimensional coin.

People used to say, and fortunately this is becoming rare, that you can’t attribute a given weather event to climate change. That has never been true. The reason people said that is not because weather events are somehow unlinked to climate change (they can’t be; weather and climate are the same thing at different scales of time and space). Rather, people said that because of the statistical difficulty of teasing out a given event from climate change. The fallacy behind this statement, which has been co-opted by “false balancers” and science denialists to reduce the importance of climate change, is easily exposed by asking a few simple questions.

Go to the desert in Arizona. Measure the temperature, daily, throughout the year for a few years. At the same time, have your friend go to the east slope of the Canadian Rockies, and also measure the temperature every day for a few years. Summarize your data by averaging across years per month. Now, go back to your study sites and measure the temperature on a given day and look up the time and place (month, Canadian Rockies vs. Arizona Desert). Compare the temperatures you’ve measured with the summary of data. Do this a few times. Notice a pattern? Yes, of course. The temperatures in Arizona will generally be higher than Canada, and this fits with the two or three years of data you’ve collected. Can you attribute the difference between your new measurements in Canada and Arizona to the differences between these locations based on your long term data? Yes, you can. The variation you see in your current measurements of the weather is patterned by the climate you estimated from your long term measurements. Climate predicts weather. Weather matches climate, plus or minus. Climate is weather with variation attenuated by greater sampling. You can attribute the weather you observe to the climate you are observing it in.

If you start in Arizona and measure the weather for a few days, then fly up to Canada and measure the weather for a few days, the differences in your measurements will reflect a difference that is explained by the longer term observations you made. The difference in weather you observe is explained by the different patterns of climate you characterized with your long term collection of information. So, if you change the climate, the weather will change, and you can attribute that to the climate change as well. It was never true that you “can’t attribute a single weather event to climate change” because it is always true that you can attribute all of the weather you observe to the climate you are observing it in. The weather is simply a low frequency sampling of the climate, so it will vary a lot more from observation to observation than will multi-year data. So while it was never true that you could pretend there was no link between climate and weather, it was always true that you could not ever separate observed weather from the region’s climate. They’ve always been linked.

But, there is a problem and it comes back to that word “attribute.” To meaningfully and quantitatively attribute daily weather observations to a change in any given variable is difficult because there are so many variables that affect weather. If we want to attribute a certain frequency of rare events such as major floods or killer heat waves to a given change in the climate in a way that allows us to convincingly and quantitatively link the that change in climate to the change in frequency of the events, we could observe for a very long time. Instead of just measuring temperature and rainfall for a few years, which would give us pretty good climatic dat, we’d have to observe and measure rare events for a very long time. For example, if we want to see if a theoretical “thousand year flood” has become more common so it is now a “hundred year flood” we’d have to observe floods for many decades in order to get enough data to re-calibrate flood frequency.

This presents two major problems if we want to understand the relationship between global warming and weather events. First, we will have to observe the weather for so long that policy makers waiting for our scientifically valid conclusion will not be able to act on the basis of the data in a timely manner. The second problem is that climate change may be happening so fast that zonal, qualitative, and quantitative shifts in climate may roll right past our humble data collection enterprise. If the climate fundamentally changes fast enough that every decade is different from the previous decade, than it will be impossible to get a nice twenty year long sample of any given phenomenon. Some aspects of climate change seem to be moving along at this rate, a great example being the annual rate of Arctic Sea ice melting. There is no twenty year period that reflects the current rate because the rate has gone up so fast. We can’t develop a “climatology” of Arctic Sea ice melt based on stable well behaved 20 year periods because there aren’t any.

One way to handle this problem is with simulation studies. If we have a good model that can simulate a year’s worth of climate and weather activity then we can run that model a large number of times and see how often particular weather events occur. Since this is a model run in a computer (or several computers) we can simulate a year of climate with and without the global warming related changes, and compare those two years. Thousands of times. This way we get a version of climatology, long term measurements, that is statistically better than any real life measurements would allow. Climate models, that run the Earth’s climate in a computer on demand, are good enough to do this.

So let’s do that. Let’s get a computer program that runs climate simulations, change the variables to reflect climate change vs. no climate change scenarios, run the model a gazillion times, and see if weather events like the historic flooding in the United Kingdom this year are likely to occur at a higher frequency with global warming, and if so, estimate what that frequency might be. Check it out:

A new citizen science project launched by climate researchers at the University of Oxford will determine in the next month or so whether global warming made this winter’s extreme deluge more likely to occur, or not. …

The weather@home project allows you to donate your spare computer time in return for helping turn speculation over the role of climate change in extreme weather into statistical fact….

The Weather@home 2014 project is located HERE and you can sign up to help.

Here’s a video explaining the project:

I’ll probably set up a computer to be used mainly for this purpose and give them a few days of processing time. I’ve read through the requirements and all the important information needed and it looks pretty straight forward. If I do this I’ll let you know how it goes.

Comments

  1. #1 JMI
    March 4, 2014

    Really interesting Greg, thanks​!​

    I think that you would be really interested in some of the most cutting-edge research that I have come across explaining crowds, open innovation, and citizen science.​

    http://papers.ssrn.com/sol3/cf_dev/AbsByAuth.cfm?per_id=1919614

    And you may also enjoy this blog about the same too:
    https://thecrowdsociety.jux.com/

    Powerful stuff, no?

  2. #2 jane
    March 5, 2014

    A possible alternative way of wording the dice analogy would be, e.g., to say that instead of rolling 12 on 2d6 to get a particular outcome, you now only had to roll 10 to “succeed”. Or if you want to be ultra-ultra-ultra-nerdy, it’s like having a critical hit range of 19-20 or better on a d20 vs. just 20.