Posted by Jeff Id on April 4, 2010
Often here, I’ve made the point that the 06-07-08 sea ice reductions in the Arctic were created by winds and water flow rather than temperature. These reductions look pretty dramatic in video, but the key is that the wind patterns are very much visible during the summer melt season – especially in 07.
It’s like the southern portion of the Arctic sea ice shown above, is attacked by September winds. Now in my previous post, I made a point WRT a paper on Arctic winds which was critical of certain claims – without even reading the paper. Frank left a message criticizing my interpretation on the thread especially considering I hadn’t read the paper. A reasonable point for sure. After all, there is the possibility that I have surmised incorrectly.
The paper claimed that 1/3 of the minimal Arctic melting trend was due to winds and the recent 07 losses were 50 percent due to winds, leaving the rest for interpretation. Now from the video, I would estimate any differences from normal (whatever that is) in 07 to be 99% wind and water currents but that’s just a thumbnail.
Here’s what I wrote:
Of course the 1/3 or 1/2 numbers in the quoted paragraph are completely bogus in my opinion. I’m not saying fraudulent, there is a difference, but I’m extraordinarily skeptical of anyone who claims to know what percentage of sea ice was melted due to one effect or another. It is almost guaranteed to be nothing but handwaiving with some loosely defined regressions for evidence. Most of us are familiar with that trend in climatology – (some pun).
This is what reader Frank wrote in reply.
This isn’t a reason to suggest that information in a science paper is bogus. The paper, since you obviously haven’t read it, can be found at: http://www.jamstec.go.jp/frsgc/research/d2/masayo.ogi/2009GL042356-pip.pdf
In Figure 1, there are some very nice maps showing the anomaly in arctic summer and winter wind patterns multiplied by the anomaly in sea ice extent. The maps clearly show that reduced sea ice extent is associated with stronger winds blowing towards the North Atlantic. These maps would be worth posting on your site. The regressions show that about 50% of the variation in sea ice extent can be explained by how closely winds reflect the pattern in Figure 1. It isn’t clear to me how one can perform a “loosely defined regression” – the numbers must come from a defined location. In this case, they consider winds over the ocean everywhere north of 65 degrees. If they had considered winds from the most relevant areas in Figure 1, perhaps the correlation would have been stronger. However, with wind from an infinite combination of grid cells to choose from and only 30 years of ice data, it would be easy to overfit the data.
Now I’m not at all being critical of Franks response, his point was clear and he generously provided a link to the paper. However, the paper is exactly what I predicted from just a few short sentences. The authors, took raw wind data and turned it into an index value and regressed that directly against sea ice extent. This means we have the standard assumption of linearity, which makes the math simpler but in this case cannot be true. Here is a quote from the paper, my bold.
First, the 925-hPa wind fields are regressed on September SIE to determine the seasonally-varying wind pattern that is linearly related to the subsequent September SIE.
Now they have taken a complex wind pattern (see figures at the end of the above link) and decided it’s linearly related to extent. It’s right in the assumption. However, there will be a big difference between a wind coming straight in from North Russia, vs coming from the Baring Strait. All we have to do is imagine wind across land (Russia) vs pushing a bit of slightly warmer water – a powerful ice melter – right up the strait. The energy transfer from a bit of wind driven water flow is many times greater than a bit of warmer southern air.
Another quote from the paper:
We then use the standardized time series of these winter and summer wind indices as predictors in a linear model to predict difference in September SIE from one year to the next (ΔSIE).
Now the standardized time series are just a means of representing the wind in a linear quantity. There is no method to account for the expected melting of the Arctic ice when the wind comes from different directions, nor is there an attempt to correct for the flow of the ocean. Again, it’s easy to understand that these effects are in no way linear.
However, that does not make this paper a bad one. The authors clearly found a relationship between wind and ice minima. Where they go to far (as often happens in climatology) is to claim that a certain percentage of the change can be attributed to wind. From this work, we have absolutely no idea how much the sea ice is affected by wind, we only know that it IS. — No real surprise there.
So in using the wind field directly, we are regressing what are multiple effects against sea ice extents. Thus we have a loosely defined yet reasonable relationship and what I termed- a loosely defined linear regression. The result of which cannot tell us the true amount of affect wind patterns had on sea ice.
You might ask, how can we do it better? I would say, you could include ever more effects in more complex regressions but the main improvement I would suggest, without resorting to an expanded study, would be to change the conclusion. Perhaps to say: These results indicate a substantial relationship between wind field and ice melt. Through linear regressions against wind indices, we have been able to explain 50% of the long term variance of Arctic sea ice indicating a strong relationship. However, inclusion of non-linear effects in complex wind patterns and geography could lead to a substantial improvement in the determination of the wind’s total influence on sea ice extent.
Now how did I know the paper would be like this without reading it? Because, it’s common in climatology to regress one factor against another to define relationships. It’s common to assume linearity and it’s common to overconclude that some % of the natural effect is then explained. I already know sea ice is very complex as it basically acts a fluid with non-Newtonian properties (a bit of stiffness). Currents flow and change through the Arctic based on weather patterns and on top of that there has been a warming trend.
Anyway, in my opinion, claims of 1/3 or 1/2 of ice loss being due to winds are spurious at best. Claims that these authors have demonstrated that winds have a substantial effect on sea ice are appropriate tho.