Actually, the monthly time series of air temperature in Stockholm in the winter season is quite highly correlated (negatively) to the monthly precipitation in the Mediterranean region: in the winter months in which the Azores High is strong, the stronger westerly winds brings warmer and more humid air to Northern Europe and deflect the rain-bringing Atlantic depressions that normally would hit the Mediterranean at this time of the year. Last winter, the NAO was in a very negative state, with a very weak Azores High. Temperatures in Northern Europe were low and precipitation was very high in the Mediterranean. We can appreciate this correlation in the figure below. The map shows the correlation coefficient between the time series of monthly winter precipitation in one particular location and the simultaneous temperature in Stockholm. This plot shows that although the word 'teleconnection' has been quite derided in some blogs, it indeed has some physical basis.
Map showing the correlation coefficients between monthly winter temperature in Stockholm and the simultaneous monthly mean precipitation in Europe in the period 1950-2010. Links, from observations; right from a climate model simulation started in year 1000. The right figure displays the model grid.
Incidentally, a similar calculation performed with data from a climate model simulations shows pretty much the same pattern: in the climate model 'Stockholm' temperatures are also strongly (negatively) correlated to Mediterranean precipitation in wintertime. I think this similarity is quite remarkable - the climate model has not been tuned to produce this pattern. It emerges from the model's own atmospheric dynamics, and shows that climate models, though not perfect, can produce climate patterns that closely reassemble the observed.
How is climate change affecting the NAO, or in other words the strength of the Azores High ? Most climate models predict an intensification of the NAO with increasing concentration of greenhouse gases. This is why the Mediterranean precipitation is one the few areas not located at high-latitudes for which climate models agree in their predictions for winter precipitation. However, the NAO index displayed a mostly negative trend through the 20th century until about 1975, then an upward trend until the early 1990s and a negative trend again thereafter. It is not very often mentioned that climate models are not able to replicate either this long-term behaviour nor the observed amplitude of short-term variations of the NAO index.
As it is becoming too common in climate research, uncertainty opens the window for an inflation of hypothesis to identify the mechanisms behind the NAO variability and its possible evolution in the future. Last winter a paper put forward the hypothesis that the low temperatures in Northern Europe caused by the extreme low state of the NAO were caused by anthropogenic climate change. The reasoning went something like this: a diminished sea-ice cover in some regions of the Arctic would tend to disturb the atmospheric circulation towards a more negative NAO state. As it usually happens, Nature has promptly responded this winter so far with a high NAO index - the opposite to the predictions in this paper.
Monthly Index of the North Atlantic Oscillation in winter (November-March)
Several institutions issue extended weather forecast or seasonal predictions focused on the NAO, for instance NOAA . Seasonal forecasts pose different questions from climate projections. In the former, a good knowledge of the initial state of the atmosphere and the ocean is critical to predict their evolution in the following weeks or months, climate projections are to a very large extent determined by the external forcing, like greenhouse gasses, volcanism and solar activity. However, I think there is a value in seasonal forecasts, like the NAO forecast issued by NOAA, that goes beyond the the mere short-term prediction. This requires a more lengthy exposition, but in my opinion a seasonal forecast that happens to be usually correct would clearly strengthen the credibility of climate models in general. The skill of seasonal forecast is still quite poor, unfortunately. We can see for instance that the NOAA forecast for this winter NAO, even the forecast with a lead time of 14 days, underestimates the high values of the NAO index observed.
In mid-November, I made a bet with my sister that the UK would have a wet and windy winter. So far I'm on track to collect my winnings: 10p.
ReplyDeleteI need snow in Scandinavia in March, though. Can anyone arrange that? I'll pay you 10p.
"(...) right from a climate model simulation started in year 1000."
ReplyDeleteI simply don't understand the need for this.
@ 2
ReplyDeleteNo, there is no need to use a simulation that long, and the model was not run with the purpose of creating this figure. That would be very costly indeed. I wanted, however, to avoid that anyone would think that the model has been somehow launched, or tuned, to produce this result.
Thank you eduardo, I see.
ReplyDeleteOne thing I fail to grasp regularly however ist the combined logic that is so often presented.
In this case it's combining what _most_ models say about a shift in NOA with the documented skill of _one_ single model.
If we have some confidence that this model indeed has skill, then shouldn't we better combine its pattern with what _the same_ model says about future NOA shifts?
... it's NAO of course, stupid me.
ReplyDelete@4
ReplyDeleteI would generally agree with you that one way to reduce the uncertainty in the projections and thus provide more accurate and maybe more falsifiable projections is to reduce the number of models that are deemed good enough. There are clearly suboptimal models in the IPCC suite that are just kept in the pack for political rather than scientific reasons. It is obvious that if we include all possible models, good and bad, the model ensemble will be always compatible with any observation. However, I have the impression that I am in the minority here.
In practice the question is more tough that just deciding whether or not a model has skill. Usually, one model maybe better than another model in some aspects, say precipitation in the extratropics, but worse in another aspect, for instance ENSO. Also, what this post shows is simply that the correlation between the Stockholm temperature and European precipitation is, in this model, quite realistic, but this not a guarantee that this model will properly simulate the effect of greenhouse gases on the NAO.
the interpretation of models and ensemble of models is a subtle question. Models are quite useful and sometimes surprisingly realistic, contrary to what one can read in many skeptic blogs. But they are a long way from perfect
eduartdo,
ReplyDeletethank you for your reply. I knew about some basic points you mentioned generally but didn't know about the views you hold.
There is strong criticism in what you say in the first section about not dropping bad models. I wish you would elaborate more on that.
"It is obvious that if we include all possible models, good and bad, the model ensemble will be always compatible with any observation."
This is true but isn't there another side of the story? Doesn't including these models OTOH result in diminishing confidence regarding all results?
But leaving politics and qui bono aside for now I would like to learn more about how you validate models to conclude that some are clearly inferior and should be dropped.
Because what I've learned so far was that regarding models there is no champions league (as you mention in section two) so I suspected there wouldn't be a bunglers league either.
eduardo,
ReplyDeletesorry for the misspelling of your first name ... the more fluently I appear to type, the more typing errors simply escape me.
@ 7
ReplyDeleteI am preparing a longer posting, since a comment would be too short for this subtle question. I hope I can post it soon
I'm also curious about the answer to #7 ;-)
ReplyDeleteThere was also an article in the newspapers about "climate prediction" based measurements to predict the AMOC:
Hamburger Abendblatt
Original report in Science:
Multiyear Prediction of Monthly Mean Atlantic Meridional Overturning Circulation at 26.5°N
This kind of measurements and predictions might be also interesting for currents further north possibly affecting also the NAO as slowly changing boundary conditions. But as stated by Marotzke in the article, the climate further north gets more chaotic compared to the (sub-)tropics.