I say this because I'm pretty sure that the Francis/Vavrus paper
The key paper is "Revisiting the evidence linking Arctic Amplification to extreme weather in midlatitudes", Dr Elizabeth Barnes, 2013 GRL. A PDF copy is here, but I've quoted the abstract in it's entirety.
Previous studies have suggested that Arctic Amplification has caused planetary-scale
waves to elongate meridionally and slow-down, resulting in more frequent blocking patterns and extreme weather. Here, trends in the meridional extent of atmospheric waves over North America and the North Atlantic are investigated in three reanalyses, and it is demonstrated that previously reported positive trends are an artifact of the methodology. No significant decrease in planetary-scale wave phase speeds are found except in OND, but this trend is sensitive to the analysis parameters. Moreover, the frequency of blocking occurrence exhibits no significant increase in any season in any of the three reanalyses, further supporting the lack of trends in wave speed and meridional extent. This work highlights that observed trends in midlatitude weather patterns are complex and likely not simply understood in terms of Arctic Amplification alone.
Additionally there is Screen & Simmonds, 2013, "Exploring links between Arctic amplification and mid-latitude weather.", Abstract. A paper that finds "that possible connections between Arctic amplification and planetary waves, and implications of these, are sensitive to how waves are conceptualized."
The Francis/Vavrus paper linking Arctic Amplification to increased amplitude of waves in the Jetstream and 'stuck' weather patterns is now looking like it is wrong. In the Barnes paper it is noted that "metrics that focus on a narrow range of isopleths to track the ridges and troughs of a passing wave will incorrectly interpret a shift of the geopotential height field as a change in wave extent. When this shift is accounted for, no significant trend is found."
In other words, with the warming of the northern hemisphere the atmosphere has been expanding, shifting geopotential heights upwards (geopotential height being the height at which a certain pressure level is found). And causing a northward shift of geopotential height. Barnes attacks the problem in a very intelligent way, she defines two indices, SeaMaxMin is the seasonal excursion of peaks and troughs in the 500mb GPH field, where GPH is 5700m up in the atmosphere, DayMaxMin is the same index but on a daily basis. In line with Francis and Vavrus SeaMaxMin has an upward trend, but crucially DayMaxMin doesn't. This is fatal problem for the Francis/Vavrus paper, because if their hypothesis is correct both indices should show an upwards trend.
One problem I have with the Barnes paper is that while I can't fault the reasoning behind the following finding:
The lack of increasing trends in the atmospheric blocking patterns further supports the lack of trends in the wave meridional extents and wave phase speeds, and suggests that Arctic Amplification over the past 30 years has not had a quantifiable impact on slow-moving weather patterns over North America or the North Atlantic.In my examination of the blocking data kindly made public by Dr Anthony Lupo, the increase in blocking in that data was particularly over Europe, image, but that's another issue.
None of this seems to me to affect Dr Judah Cohen's work on winter links to reduced sea ice mediated by snow line advance in October. But it may help to throw some light on the Greenland blocking and the summer pattern. I'm still considering that. The pattern is real, I don't see how it affects Overland et al's finding of an atmospheric shift in the Arctic post 2007, and I don't see that it invalidates my related obsession with the post 2007 summer pattern. If anyone thinks I'm being blinkered in this, that's what the comments are for.
22 comments:
Chris, I've read the Barnes paper and have to say I wasn't impressed. Maybe I'm just suspicious of anything that comes out of CSU (Pielke Sr.).
Jennifer Frances actually seemed a bit personally annoyed by the paper -- read this.
I should explain the "I wasn't impressed." Even before I'd read Dr. Frances' comments I got the impression that Barnes was going out of her way to spin the data. I thought it actually supporteed Ftrances' conclusions and not Barnes' conclusions.
I was going to post a comment asking you to verify some of my own misgivings, but first checked to see if Frances had commented on Barnes' paper. Frances not only verified my own misgivings, but added a couple I missed.
The data is good - the conclusions are the opposite of what they should be.
Thanks Kevin,
I hadn't seen Dr Francis's response, her comments seem reasonable and I will look at this after work. But given the region Barnes selected, it tallies with my previous findings looking at NH Blocking - there's been an increase in blocking, but far less than in Eurasia. Had Barnes have chosen Eurasia she probably wouldn't have been able to conclude no impact from AA, but in the context of the region studied I think the conclusion is supported.
I have made clear my doubts, S&S finding that approach strongly affects results, and the difference between DayMaxMin and SeaMaxMin, I'd expect greater slope for DayMaxMin. I still think these are critical issues that mean I will need to see more research before being happy with the findings of Francis/Vavrus.
Kevin,
I've had more time to consider this today, the above comment was made in a rush this morning before leaving for work.
I have changed a line in the opening of the post as I think the original line was an unduly harsh judgment.
I think it is reasonable for Barnes to look at the region she did as that is a subset of the region used by FV12. This does however raise the question: Why use only a subset, why not use the whole region?
Here I should note that I checked Barnes's affiliation (University) and immediately made the Pielke connection. However I have read duff papers from that stable, and their flaws have been all too readily apparent. After reading the Barnes paper I had only the above concern (region), and did not see an obviously flawed paper.
The issues I have concerns over remain. I would expect to see more of a positive trend in the daymaxmin index. That seamaxmin shows an increase does not support FV12 in my opinion, this is because it is selected to be a similar index to that used in FV12. That daymaxmin shows a much smaller increase is of concern with regards to FV12 and brings into play Screen & Simmonds (ref'd in the above post). Thus it is hard to attribute strength to the findings of FV12.
This however brings into play Francis's comment in the WSJ. She rightly notes that the process is in its early stages. So it may be that FV12 have identified a process in its early stages with a less than optimal approach - on reflection today I suspect this is the case.
I expect that both the Screen/Simmonds and Barnes paper will spur Francis and Vavrus to revisit and strengthen FV12. None of this prejudices me against those authors, I just don't think now that FV12 is anything more than suggestive of the increase of Rossby wave amplitude, slowing of Rossby waves and consequent linkage to persistent (stuck) weather patterns.
The Screen/Simmonds paper mentioned in the lead post is available for free here: It's via a 'handler' page, so I can't give a direct link to PDF, not sure it will work properly.
Chris - In case it's of interest, a friendly local neighbourhood climate scientist has pointed me in the direction of this overview of the Francis v Barnes debate:
http://www.eenews.net/stories/1059986669
Thanks Jim, and please pass my thanks on to your contact. That's a very useful explanation of the situation.
E&E though...
That made me wonder if it was the spoof journal Energy & Environment, producers of some of the most atrocious pseudo science I've ever read. ;) Turns out it's Environment & Energy, which from the intelligent, balanced and reasonable piece you linked to seems to be un-associated.
Chris
In the Environment and Energy overview 'statistically significant' seems to be a key concept. In gambling terms, am I correct in thinking this refers to a bet of 1/19 on? i.e. to be statistically significant a gambler would find it sensible to lay out £19 to win £1?
Hi Geoff,
I don't know much about gambling, despite having recently offered someone out on a bet (which he declined).
Statistically significant typically means 95% confidence, in other words a 1/20 likelihood of being wrong. Tighter significance can be specified, but 95% seems most common to me.
Thanks Chris
So the "evidence so far" says that the point where there is only a 5% chance of Francis/Vavrus being wrong.
But if it were the case that there is now (at this stage of the evidence) a 33% chance of them being wrong, would that mean there is a 67% chance that they are right?
Correction ...
Thanks Chris
So the "evidence so far" says that the point where there is only a 5% chance of Francis/Vavrus being wrong has not yet been reached.
But if it were the case that there is now (at this stage of the evidence) a 33% chance of them being wrong, would that mean there is a 67% chance that they are right?
Wow Geoff,
Now I am really not following you! I don't get your math (33/67) at all and we're obviously talking at cross purposes.
I'd say more like 5% of Francis/Vavrus being right (at this stage), 95% that they're wrong. Basically from my reading of the Barnes paper F/V 2012 is now in serious question, for what it's worth; I will not be relying on that paper unless further research addresses the issues raised in both Barnes and Screen & Simmonds. Right now it's hard to see how F/V can be recovered. Were I in Dr Francis' shoes I'd go right back to the drawing board and see if there is a more robust way of determining whether the effect is robustly detectable at present, with a reasonable degree of certainty.
Note - I say 'at this stage' because I suspect that, as Dr Francis states, they may be too early in looking for the effect. However that means one cannot claim the effect is real, in that instance one must wait for more data, and for the process to continue.
Chris
The (33/67) was meant as an example not an estimation. My question was really about the use of the law of excluded middle. You seem to apply it in saying "5% right,95% wrong" and there is no other possibilities than right or wrong.
I'm nervous of this particularly when it's used to imply (not by you) statements like "It's not statistically significant so it must be wrong."
In your "note" you do seem to intoduce a third possibility "not proven" - similar to the third verdict available in Scottish courts.
You are one of the people whose judgment I take very seriously so I would like to be clear.
It seems to me that Dr Francis has a theoretical reason to expect the effects she predicts: starting... If the Arctic warms then there will be changes in height differences in the atmosphere &etc. Should arguments like this modify the expectations we have now? i.e before extra evidence is found.
Geoff,
If a result is 'not statistically significant' it is not wrong, only a complete idiot would say that. Results that are not stat sig can still be useful, although they cannot be relied upon with confidence one might ascribe to a significant result.
I think Dr Francis may end up being right, but that it is too early to get a stat sig result. Despite that possibility the papers I refer to in the main post pose problems for the method, so IMO a new method needs to be devised, one not subject to the issues raised in Barnes and Screen & Simmonds.
Not proven would suggest that the accused walk free. In this instance 'not proven' implies that the research cannot be relied upon - the 'waves in the jet' walk free, not found guilty of involvement in the pole/equator temperature gradient and extreme weather events.
The following paper may be relevant:
Screen JA 2013 Influence of Arctic sea ice on European summer precipitation
2013 Environ. Res. Lett. 8 044015
http://iopscience.iop.org/1748-9326/8/4/044015/
It is open access.
Thanks Timothy,
I posted a new blog post on that paper last night.
Chris, as an expert Pielkeologist, allow Eli to say that UC Boulder is a huge place with lots of people, not only at UC, but adjunct people from NCAR and the NOAA lab and those in CIRES.
Even the political scientists want nothing to do with Pielke, Jr. Eli remembers a fit he threw when they would not give him a dual appointment.
This is off topic, but hopefully people will not diss UC Boulder because he is there. Lots of good scientists are...one polemicist amidst the weeds is what you often get.
Thanks Eli,
Given comments up-thread your contribution is quite on topic. And as you've seen - Pielke prejudice didn't stop me from concluding Barnes has made a good case.
I would find the LUPO data highly suspicious. It seems to be missing some blocking data from the north pacific. Anyone who would actively contribute to an anti-science document as the following has lost any credibility in my book.
•Lupo, A.R. (Contributing Author Ch 6 only), 2009:
Heartland Institute, 2009: Climate Change Reconsidered: The Report of the Nongovernmental International Panel on Climate Change (NIPCC), edited by Craig Idso, Ph.D., and S. Fred Singer, Ph.D. (released 2 June 2009)- 2%.
Anon,
Everything that comes out of Heartland is garbage, which is why you won't find me blogging on it. People's motives may be suspicious, they may be flat out wrong on some issues, but that doesn't mean all their science is garbage. That said as a simple act of time-efficiency I have stopped reading Pielke and some of the other high profile denialists, if they say something worthwhile it will be cited and I will revisit it.
There are a large variety of approaches to blocking, some better than others, but others for which it is hard to decide which is better, and how one should judge that. I was asked to look at the Lupo data, you've seen the result. If you think there's a better, publicly available, dataset please feel free to direct me to it.
It's also worth noting that CMIP5 climate models predict a substantial decrease in high-latitude blocking with strong agreement. The wording of the AR5 largely agrees with the CMIP5 models.
Thanks Andrew.
Post a Comment