Spin Snow, not Sea Ice: the AMO is Real!

How unfortunate. Back in 2000, yes, that’s not a typo, in 2000, the Independent wrote that:

“According to Dr David Viner, a senior research scientist at the climatic research unit (CRU) of the University of East Anglia, within a few years winter snowfall will become ‘a very rare and exciting event’.

‘Children just aren’t going to know what snow is,’ he said.”

In Viner’s defence, he did go on to say that rare snow events would cause chaos.

It’s No Joke

For a long time it’s seemed to me that one problem global warming (GW) is likely to throw up is that snow events, like other forms of precipitation, will become more extreme. That is, when it does snow, it’ll be heavier.

A commenter on one of my recent posts suggested I go and do some statistical analysis on temperature measurement data to see if trends are significant. In actual fact, about 5 years ago, I did exactly this with data on snowfall. If I recollect correctly, I found that there was a statistically significant trend in the number of heavy snow days (above a particular depth) in the middle of winter (i.e. not in months when, due to GW, some pf what would otherwise have been snow might fall as rain) in the data I found on the net for a particular Rocky Mountain ski resort. If I come across my notes I’ll bring the analysis up to date.

Here’s the real concern. A few decades down the line, the planet will be a lot warmer and we’ll be seeing much heavier precipitation in some regions. Some of this will be snow. Furthermore, there’s always the chance of a cold snap, for example, when a volcano goes off (and we really should be worrying more about this climate risk, IMHOP – more another time, maybe). Or after a geo-engineering accident (sorry, couldn’t resist). At the start of the cold event at least, the oceans will still be warm, because of stored cumulative GW heat, and they will therefore continue to pump moisture into the atmosphere. But the dust shroud will rapidly cool land areas, so that some places used to dealing with just heavier rain suddenly find themselves trying to cope with a foot or two of the white stuff.

It’s a shame climate scientists haven’t been warning people about the vulnerability of flat roofs to heavy snow.

Skating on Thin Ice

On the other hand, there’s been a worrying tendency over the last few years to treat the continually diminishing amount of Arctic sea ice each year (at the minimum extent in September) as a GW canary in the coal-mine, like glaciers.

It would have been better to stick to glaciers. Because changes in Arctic sea ice may well be part of a natural cycle. Of course, there’s an underlying warming trend tending to reduce the amount of Arctic sea ice. But if and when the natural cycle starts to dominate, sceptics will have another field day.

It’s worse than this. The cycle – which is called the Atlantic Multi-Decadal Oscillation, or AMO for short – could affect the temperature of the entire Northern Hemisphere (NH). [See previous post Musings of the Hemispheres – there may be similar processes in the SH, but I’m not going to discuss those just now].

Before I go on, there was a fuss a while back – serious stuff: letters to the Guardian editor, that kind of thing – when a Professor Latif was accused of explaining GW with AMO. His position, like mine, is that both GW and AMO affect the climate. I just want to make it clear that I’m with the Professor on this, even if simplistic sceptic brains find this position a logical contortion.

Evidence for the AMO (1): IPCC Data

Consider the following graph from the IPCC (AR4, the most recent report):

Global mean surface temperature relative to 1901-50, compared to climate models (IPCC Fig TS.23)

What gets me about the IPCC data is the anomaly around 1940. The average temperature was simply too high, and this is not adequately explained (if it was, I guess the models would be corrected).

We can drill down a little further:

Continental-scale breakdown of actual and modelled temperatures compared to 1901-50 (IPCC Fig TS.22)

Here we see that by and large the models represent land temperature fairly well, but that ocean temperatures were outside what are presumably intended to be some kind of confidence limits – for what looks like around an entire decade (just before mid-century).

This is not a very satisfactory state of affairs.

Note from Fig TS.22 above that the land temperature range over the past century has been around 1C, and that of the oceans perhaps 0.7C.

Consider what’s happened in the North Atlantic:

AMO from 1850-2005 (temperature relative to 1961-90 (IPCC Fig. 3.33)

The North Atlantic sea surface temperature (SST) (top graph) has increased by nearly 1C since its lowest point soon after the turn of the 20th century.

A 1C increase in ocean temperature is unsustainable. Land has a lower heat capacity (i.e. you have to put in less heat for a 1C temperature rise) than ocean, so must warm faster. The North Atlantic heat will have to dissipate.

Evidence for the AMO (2): The Historical Record

If I were a climate specialist about to make a song and dance over a particular piece of evidence for GW, I think I’d make pretty sure the phenomenon in question hadn’t happened before.

It just so happens that the area of Arctic sea ice has shrunk dramatically before, and not so long ago.

Yeap, you’ve guessed it, the Arctic warmed from around 1920 to 1940. Here’s the Abstract of a paper The early twentieth century warm period in the European Arctic that looks kosher – it must be, it costs $42! A site, www.arctic-warming.com seems to be devoted to the issue (particularly of warming around Spitsbergen in 1918-22) and cites some other papers discussing the 1920-40 episode, “one of the most spectacular climate events of the 20th century”. There’s even a book about the event.

None of these sites offer a clear explanation for the Arctic warming, so I’m going to have a bash.

Explaining the AMO

The point is that loss of Arctic sea ice – absence in summer and thinning year round – is not just a symptom of warming. It is part of a cyclic causal mechanism.

As I pointed out in a previous post, The Earth is a Fridge, the less sea ice there is at the start of winter (the Arctic ice extent is at a minimum around mid-September!), the more heat the Arctic waters can lose to the atmosphere and hence into space during the winter. Water covered by ice can’t lose heat because ice is an insulator, and the process of freezing is itself an important mechanism for losing heat.

Clearly the Arctic waters will lose most heat in winter when there is no summer ice. In a steadily warming world, you might expect first the summer ice to disappear, at which point the Arctic would have reached it’s maximum effectiveness in getting rid of heat (imported in currents from lower latitudes) and gradually the maximum extent of ice each year would reduce.

But there is an oscillation in the system.

Modelling the AMO

At first I was going to simply draw a curve on a piece of paper and scan it in, but my better half is a bit of an Excel whizz and persuaded me to do something a bit more sophisticated.

It was astonishingly easy.  Here’s the result, first without taking account of global warming (GW):

I can’t emphasise enough how easy it was to produce this graph. It’s hugely simplified, including as it does just two ocean masses and nothing else and making no attempt to distinguish between heat and temperature, and between temperatures at different times of year.  But I don’t see why it isn’t qualitatively valid – it produces the asynchronous sinusoidal temperature curves I’d deduced anyway, but with the added theoretical basis of generating them by heat exchange between the Arctic and the NA.  And since I’ve tied the temperature curves very roughly to historic data, the timescale of future temperature changes could conceivably be roughly correct.  The fact that what I wanted to show drops so easily out of the spreadsheet suggests some underlying veracity – I claim no more than that – at least to me.  End of disclaimer.

All I’ve done is calculate the temperature of the Arctic (purple line) in a given year as its temperature the previous year (times a cooling factor) plus the North Atlantic (NA) temperature the previous year times a factor (15% in this instance).  All I’m assuming is that the warmer the NA is, the warmer the Arctic will be.  After all, we know surface water flows from the NA to the Arctic.

So far, so simple.  The next bit is the critical point.

I’ve calculated the temperature of the NA (green line) similarly, but included a negative feedback.  In the model, the NA temperature is equal to its temperature the previous year (times a cooling factor) minus the Arctic temperature the previous year times a factor (6% in this instance, less than the 15% for the reverse case because the NA is bigger than the Arctic).

The minus in this calculation says that the warmer the Arctic is, the more NA heat it can absorb and disperse ultimately into space.    Remember, my argument is that the thinner and less extensive the Arctic ice, i.e. the warmer it is on average over the year, and in particular at the start of winter, the more NA heat it can disperse over the year, but in particular in winter.  [A more complex model could try to model the Arctic temperature at different times of year].

Obviously I’ve adjusted the numbers and starting conditions to fit the graph roughly to the historical record.  (The anomaly on the vertical axis is arbitrary, 0 is intended to be the long-term equilibrium – if you start with 0 for both anomalies, the graph is flat).

As well as the Arctic and NA temperatures I’ve included in my schematic an indication of the Northern Hemisphere (NH) temperature, produced by simply adding the NA and Arctic values (yellow line).  This shows a peak in 1940, which is what we’re trying to explain, as well as a peak around 2005 and, as predicted by Professor Latif, subsequent cooling for quite some time.

The good news is that we won’t have to wait too long to find out whether the AMO is real.  The bad news is, that, if it is, it’ll be like putting rocket fuel in the sceptic bandwagon.

I thought I’d go a little further and see if my model predicts anything else.  I’ve therefore included an “Arctic Oscillation” (AO) (blue line) which I’ve calculated by subtracting the NA temperature from the Arctic temperature.  The AO – represented by real-world indicators such as the North Atlantic Oscillation (NAO) and the Northern Annular Mode (NAM) – is an atmospheric phenomenon which correlates with the nature of NH winters.  My logic is that the higher the temperature of the Arctic compared to the NA, the lower the air pressure will be over the Arctic in comparison with the NA, which is in principle what the NAO and NAM measure.

Anyway, here, again from the IPCC, is the actual historical record of the NAO/NAM:

NAO/NAM indices (IPCC Fig 3.31)

Compare these real-world measurements with my model which (blue line) predicted a positive AO from 1900 to the 1930s and again from the 1960s to around 2000.  Could they possibly fit together?

Future temperatures, Global Dimming and Global Warming

I have to say I’m rather alarmed that, based on the timescales of the historic 20th century AMO cycle, my model shows temperatures falling for another 15 years.  I thought I’d better factor in a bit of global warming, so I played around in Excel a bit more:

This time I’ve allowed for GW by adding an arithmetically progressively larger term into the NA and Arctic temperatures each year.  As in the previous figure, the vertical anomaly scale is entirely arbitrary and not intended to map to real temperature deviations.

I’ve also extended the model to 2050 and calculated the NH temperature (yellow line) by adding the NA temperature (green line) to a halved, rather than the whole, Arctic temperature (purple line), since the NA is bigger than the Arctic.  Clearly the temperature cycles still exist, it’s just that the AMO is imposed on an underlying trend, so both peaks and troughs in the temperature curves are higher.

In this very rough calculation, we still see NH temperatures declining for a couple of decades.  Worrying.

I should add that the usual explanation for the cooling period from around 1940 to 1970 is “global dimming”, i.e. the blocking of sunlight by industrial pollution.   The AMO hypothesis suggests that at least some of this cooling was caused by a natural cycle.

Next Steps

A perfect computer model would accurately represent sea ice melting and freezing and the resultant exchanges of heat between the sea and the atmosphere and effect on oceanic circulation.  It would therefore predict long-term natural climate variability such as may – and I stress “may” – be caused by the AMO.

Current climate models do not correctly retrodict (i.e. predict known data) the warming up to 1940 and they have under-estimated the Arctic warming that has occurred over the last decade or so.

It seems to me that – prior to the IPCC’s next report on the science, AR5 – serious effort needs to be made to evaluate the evidence and theoretical basis for an AMO, and take account of it in projections of the future climate.

I used to be highly sceptical of long-term natural climate variability, but now I’ve realised there could be feedbacks between Arctic ice-melt and NA temperatures, I’m suddenly convinced.  I’d like to see some serious modelling of the AMO and similar decadal variability that logically should also occur in the SH.

Maybe the effect of GW will be to completely swamp the natural AMO.  But I’d like to see proof of that.

A failure to explain the AMO would lead to increased climate scepticism and a loss of political will to deal with GW.  We could be left totally unprepared for a steep rise in temperatures starting in a decade or two’s time.