by Judith Curry
Climate scientists have made public statements attributing extreme events to global warming. The first such attribution that I recall was made by Kevin Trenberth, to the effect that 7% of Hurricane Katrina’s intensity and rainfalls could be attributed to global warming. Trenberth has subsequently made public statements about the attribution to global warming of the Russian heatwave, Pakistan floods, and Queensland floods. Others have made similar public statements, most recently Richard Somerville.
There was strong consensus among attendees that unusual or extreme climate events attract public attention through their impacts, and therefore, demand explanation. A conclusion of the workshop was that an effective attribution service must be timely if it is to be relevant to decision making, and to fill the void that could be easily filled by speculations. There was also concurrence that an attribution service activity must be scientifically sound and authoritative if the public and decision makers are to correctly understand the causes of extreme climate events, draw the proper inference from such events, and appreciate the context of their occurrence in a changing climate.
The requirements of such an attribution service were described to be:
The diversity of expertise yielded a rich perspective on the requirements for the development of a scientifically robust and relevant attribution service. It was agreed that the foundations of an authoritative explanation of extreme events must begin with a real-time monitoring and climate analysis capability, and availability of historical data sets, such that current events can be placed into a reliable and physically consistent historical context. Model simulations and experimentation were likewise seen as core elements which provide an essential tool in “connecting the dots” so as to establish plausible cause-effect relationships. The workshop attendees also emphasized that society, and decision makers also need to be provided with a clear statement of the meaning and implications of the scientific findings.
Some warnings are given with regards to premature attempts at attribution:
Nature itself provided no shortage of illustrations regarding the question of who cares about the causes for climate conditions and extreme events. Newsprint and media were suddenly deluged with discussions regarding the Russian heat wave, Pakistan floods, and China floods and of the concerns about the implications such events held for the immediate future, for example, on food supplies and commodity prices. The events impressed upon the attendees of this workshop the need for rapid, yet accurate, attribution information. The “teaching opportunity” that such extremes offer via a receptive window to educate the public about future climate change was appreciated by all, though the danger of premature attribution and misattribution were also recognized.
When I served as a member of the NOAA Climate Working Group, this idea of an attribution service was pushed hard by NOAA and discussed extensively by the Working Group. I was strongly and vocally critical. The assumption that extreme weather events “demand explanation” is strange in and of itself; most people assume that weather “happens” and are mainly concerned about having a good forecast with advanced warning. Those wondering about the climate connections might be interested in whether or not this was an El Nino year, or of course we expect more Atlantic hurricanes since it the active (warm) phase of the AMO, it was the “pineapple express,” there was a big blocking pattern that brought in warm air from Africa, etc.
Since Hurricane Katrina, the issue of whether global warming is contributing to specific extreme events has been discussed in the media by scientists. Exactly what would someone do with the information (if they could assume it accurate) that 5% less rainfall would have fell in Pakistan without global warming (95% of the rain would still cause massive flooding)? Would this help people adapt better to extreme events (there is already a large adaptation deficit in most places)? Would it provide fodder for litigation or the “blame game” to motivate more international humanitarian assistance? Would it help build political will to support CO2 mitigation policies? Can you think of other things people might do with such information? Do any of these reasons seem valid/useful and potentially worth U.S. taxpayer dollars to provide such attribution service?
Apart from the value of such attribution assessments, how valid are they? In his recent congressional testimony, Ben Santer provides an overview of the strategy for using climate models for attribution of extreme events:
Assessing Risks of Changes in Extreme Events. We are now capable of making informed scientific statements regarding the influence of human activities on the likelihood of extreme events (75, 76, 77).
As noted previously, computer models can be used to perform the control experiment (no human effects on climate) that we cannot perform in the real world. Using the “unforced” climate variability from a multi-century control run, it is possible to determine how many times an extreme event of a given magnitude should have been observed in the absence of human interference. The probability of obtaining the same extreme event is then calculated in a perturbed climate – for example, in a model experiment with historical or future increases in greenhouse gases, or under some specified change in mean climate (78). Comparison of the frequencies of extremes in the control and perturbed experiments allows climate scientists to make probabilistic statements about how human-induced climate change may have altered the likelihood of the extreme event (53, 78, 79). This is sometimes referred to as an assessment of “fractional attributable risk” (78).
Recently, a “fractional attributable risk” study of the 2003 European summer heat wave concluded that “there is a greater than 90% chance that over half the risk of European summer temperatures exceeding a threshold of 1.6 K is attributable to human influence on climate” (78).
This study (and related work) illustrates that the “D&A” community has moved beyond analysis of changes in the mean state of the climate. We now apply rigorous statistical methods to the problem of estimating how human activities may alter the probability of occurrence extreme events. The demonstration of human culpability in changing these risks is likely to have significant implications for the debate on policy responses to climate change.
So assuming that there is some valid point to the attribution of extreme event, what might be useful here? Well, the IMO the main thing that would be useful is a better understanding of the statistical envelope of extreme events as influenced by the main climate regimes and teleconnection indices (e.g. AMO, PDO, ENSO) on a regional basis. The engineering tables with the 50 year flood, etc. are based on a relatively recent 50 year period in the historical record. Assuming that a climate shift has occurred, the climate is currently dominated by the warm AMO and cool PDO, which was last seen in the 1950’s. The 1950’s is probably a better analogue for recent and forthcoming weather than the period 1960-2010.
But what about global warming? It is ~0.5C warmer now than in the 1950’s. Well I suspect that the large-scale circulation changes associated with the AMO and PDO have a much greater influence on the overall planetary dynamics and extreme events, but almost certainly there is at least some influence of global warming on extreme events. The idea behind the model-based attribution analysis described in Santer’s testimony is to look at the statistics of extreme events for the 20th century climate simulations, with and without the additional anthropogenic forcing. We have already discussed on the previous detection and attribution threads the difficulties and uncertainties in doing the attribution for global surface temperature, which is something that is much more straightforward than attempting to do a statistical attribution for extreme events.
By definition, extreme events are on the tail of the distribution. In forecasting heat waves, we define heat waves as a period of at least 5 days with temperatures exceeding 1.5 standard deviations from the monthly mean. Even with weather forecasts (which have higher resolution and are more accurate for such things), for a given location we have to adjust the individual ensemble simulations not only for the model bias, but also for distributional errors, otherwise we have no chance of catching what’s out there on the tail. Have the attribution scientists been attempting such adjustments? Not as far as I know. And how to make the adjustments for two different climate states isn’t straightforward; there is no reason to think that the distributional errors will be the same for the two different climate states. Presumably, they assume implicitly that the changes in distribution of extreme events is meaningful, even if there are absolute distributional errors; but this presupposes there is no change in the distributional errors.
A further issue is that future extreme events that are even more extreme than anything we’ve seen in the 20th century have to be classified as emergent phenomena from the model: well outside the range for which the model has been validated. While I made the statement of “overconfidence in IPCC’s 20th century attribution,” I have to say that I find no basis for confidence in the model-based attribution of extreme events. Yes, there is the issue of more water vapor in the air with warmer sea surface temperatures, but exactly how this gets translated into individual weather events is not at all straightforward (this will be the topic of a future post).
Summary: Not sure what the motive is for the attribution of extreme events, other than to build political will for climate change policies. More comprehensive analysis of regional extreme events (including those in the paleo records, of which we need more of) in the context of known modes of natural climate variability is probably the single most useful thing that could be done in this regard. In terms of attribution services and the broader issues of a National Climate Service, well don’t get me started (more on this in a future thread.)