Climate v Weather
A big collection of articles trying to ascertain the role of global warming came out recently in the Bulletin of the American Meteorological Society, inspiring a host of articles in the popular media ranging from Fox News utterly ignoring it (instead highlighting a paper published in PLOS One arguing that changes in sea surface temperatures in the eastern Pacific were natural) to the New York Times slugging it “Scientists Trace Extreme Heat in Australia to Climate Change” to Ars Technica’s middle of the road piece highlighting possible connections and disconnects down to the local paper here in Boulder highlighting the absence of a connection of last year’s Colorado floods to global warming. Most stories appear to have the same emphasis as the New York Times.
Ars leads off its story with the following claim: “Often lost in the public discussion is that determining the role of climate change in a specific weather event is a challenging but interesting scientific problem. It’s also one with immense practical implications. As regions rebuild after a damaging event, it’s important that these efforts be informed by what we should expect in the future.” OK, so what are the immense practical implications? Better still, what is the point?
This is where GG gets lost. Are we trying to convince people that global warming is real by saying “look, this wouldn’t have happened without global warming”? Are we trying to learn what the specific consequences are of global warming?
Really, if a strong record of global warming as shown from temperature records, from the strong asymmetry in the occurrence of cold and warm record temperatures, as shown from changes in times of ice breakups in harbors and rivers, as shown from changes in the volume of the polar ice cap, as illustrated by the changing ranges of various plants and animals as well as their changing blooming or migrating times–if none of that has convinced people of global warming, are some numerical tests of the rate of occurrence of specific weather phenomena run in models with and without modern CO2 levels really going to convince them? Of course not. (well, we hope not; if so, there are some seriously deceived climate modelers out there).
Are we trying to convince people how costly global warming could be? “Look, here’s a heat wave/drought/flood that caused $XX in damage and it became more probable because of global warming”? Well, the problem with that is that you have to be able to say just how much more common will it be, which hasn’t really been the way these studies are posed.
Yes, it is very helpful to know what kinds of things are going to occur more often as climate changes, but just how helpful is it to determine if a particular event in the past was influenced by global warming? That the Australian heat waves were both emphasized in the news stories and most robustly connected to global warming is no surprise: when global temperatures are rising above what they otherwise would be, it is pretty clear that heat waves, which are of long duration and build directly off of mean global temperatures, are the things that all most easily clear the bar as being influenced by warming. But many of the other events are short-lived, maybe a few days in the case of the Colorado floods or the Dakotas’ blizzard. And, of course, the scientists involved recognize (well, we hope) that all they can say is that the probability of an event has increased (or decreased, as in the case of the blizzard) in the face of global warming; “yes” or “no” is actually not a real outcome despite the media slant.
Here is where you have to question the emphasis on this kind of retroforecasting. Small probability events are still not well understood: we don’t have a long enough weather history to know what the tail of the distribution of rain events out in the once in a thousand year to once in ten thousand year range really look like. How much do we gain by focusing effort on trying to see if they are more or less likely when, realistically, we don’t understand right now how likely they really are? In a sense, this was the nub of the argument between Roger Pielke Jr. and several climate scientists over Pielke’s piece in Fivethirtyeight. Pielke was arguing that you do not see a signal from climate change in the cost of disasters, and while some critics pointed out some real limits to the data he presented, many took his criticism to mean that he was saying that climate change was not occurring or that it would not produce (or is not producing) serious harm. The problem in this case is that financial damage from disasters is an extremely noisy time series; finding a reliable trend would be very hard. Extreme weather events are by their nature rarer and more erratic than deviations from the means; as such, their record is quite noisy. For instance, look at the record of numbers of violent tornadoes in the U.S. over time (I reproduce the NOAA graph below; these are presumably more reliably reported due to the damage from them)
You have bad years and good years and trying to find a trend in here is going to be tough. At a simple level you could argue that with more moisture in the air, as appears to be a simple and direct consequence of global warming, we will have more energy in the atmosphere capable of driving storms like this. On the flip side, these are frequently driven by the meeting of warm, moist air masses with cold, dry ones, but as higher latitude sources of the cold, dry air will be warming and moistening more that the low latitude ones, we might expect that the frequency and intensity of warm/cold interactions to decrease and reduce the occurrence of tornadoes. Worse yet, there is no prehistoric record we can use to inform us of natural variability of tornadoes; our baseline is awfully short. While we can understand the sexiness of arguing about these extreme events, by the time we get a signal on this that is out of the noise, we will have seen far more profound average changes in things like sea level, total air conditioning days, and early snow melts that will be far more robust indicators that climate has changed.
Consider another case that came up in the recent set of papers: the ongoing drought in California. It is roughly comparable to the 1977-1978 drought at present, so right from the get-go you get the sense that this could just be regular noise in the climate system. When you consider the paleo record of drought in California (also this blog discussion), with profound decadal droughts, it is clear that even without climate change there would be a high risk of a drought like this or worse. Natural lakes in the Sierra such as Fallen Leaf Lake that historically have never dropped below their outlet fell 150 to 200 feet and long enough for large trees to grow at the lower shoreline; such century-long droughts occurred without the benefit of human-created CO2 pollution. Do we truly have a grip on what the climate system does to create such droughts? If that aspect of the climate system is missing from the models being used to test whether the current drought is “natural” or “artificial”, does the new research have any validity? Um, no, it doesn’t.
What might be missing from these models? Well, for instance, El Nino events greatly effect the rainfall in California; would these be less frequent or more frequent? A 2014 paper suggests no change but strong events would occur about twice as often; however, as the linked news article notes, a number of other researchers suggest that the models used have done a very poor job of forecasting El Nino events and appear to be missing other climate elements affecting El Nino. The emerging disappointingly small El Nino right now (after earlier forecasts of a strong El Nino) tends to underscore that problem. Many long-period oscillations in the atmosphere (ENSO being the big one, but the North Atlantic Decadal Oscillation being another) are still proving hard to forecast and their interactions are challenging climate researchers. If there is another level to this, of oscillations on the century or millennial scale that we are not yet clued in on, we might be very far from accurately forecasting any patterns in extreme weather like droughts.
So how robust are these studies anyways? That the researchers disagree in most cases should provide a quick answer: not very. Do they inform us of the probable consequences of global warming? Not particularly well; arguably a better approach (and one that generally informs the IPCC reports) is to run lots of models with lots of variability with and without anthropogenic CO2 and the differences that occur consistently across the models are the ones to be most worried about. CSI Extreme Weather may make for media attention but it is lousy science.