Arctic sea ice changes: Natural variation vs human influence

[Archimid]

The fingerprint is the particular pressure pattern in the Arctic summer in high loss years and the weather patterns that connect it to particular temperature variations in the Pacific. The same pattern is seen in the control run, and the historically forced runs and the data. 

To be clear, FTA:

The mechanisms of this teleconnection appear to be similar in observations and models, but the specific source areas and path of wave activity underlying the establishment of the high pressure in the Arctic are displaced in the model.

That important caveat, combined with the failure of most of the model runs to account for the Arctic loses tell me that

The internal variability isn't the difference between the historical data and the ensemble mean, its the spread of the ensemble. Look at the band of grey lines in 1e. The internal variability is the difference between the top of that band and the bottom of that band.

To me this says that if the Models are sufficiently accurate representations of the system, then the variability is given by the spread of the models. The problem is that the models have significant difference with observations. They are not sufficiently good representations of the system. I think the paper does a fantastic job illustrating the big differences between the models and the observations. Fig. 4 is particularly interesting.

There's not enough historic data to pull the variability from it alone. That's one of the key points from this paper. In 30 years time it might be possible to assess the variability from the historic data, but the record is still too short to properly characterize the climate.

I agree with that. There is statistical uncertainty due to the short record. Luckily, mathematical statistics are not the only tool we have to inform our decision making process. We have physics that dictate that the warming will continue and will increase. There is every reason to believe that "internal variability" will vary as the climate changes.

[Archimid]

We know full well that extent and volume have behaved differently during the historical period. This paper provides strong evidence that conclusions drawn from a data span as short as the historical period are biassed. A plot of volume from the historical period only is doubly biassed, first because the period considered is too short to capture the range of variability and second because its volume rather than extent.

I think this is a good argument only without context. If we knew nothing about the past temperatures and conditions of the Arctic and the world, then yes, a mere volume plot would be doubly biased. But we do know quite a bit about the state of the Arctic for much longer time frames than 37 years. There is every reason to expect that it was very cold, central volume and area was much less variable. There is every reason to expect multiyear ice to dominate the Arctic for thousands of years.

The fundamental mechanics haven't changed. Heat is still being transferred by radiation, conduction and convection. Ice still melts and water freezes according to imbalances in energy transfer at the phase interfaces. Air and water are still moving in response to pressure gradients. The Navier Stokes equation is as valid then as it is now.

True, but all values in the fundamental mechanics are changing. Some of them are bounded by fundamental physics and some of them are bounded by changing conditions and phenomena. As these values change, the models diverge from observations.

The way ice moves in response to the stresses exerted on it by air and water is not well understood and its quite possible that the equations used for it and the parameters fitted to them work better in an Arctic dominated by MYI, but its also possible they work better in an Arctic dominated by FYI and the models are actually representing that aspect better.

I doubt it. For example, the models were created for a desert climate but the arctic is changing to a maritime climate. The models will need some extreme luck for changes in atmospheric and oceanic pattern that this climate change is bringing to result in better accuracy. New models can bring better accuracy. But if they show near term destruction, who is going to believe them?

There's strong evidence in this paper that the high melt years in 2007-2012 are at the extremes of melt below the trend and 2013 is on trend, and the reason that isn't obvious from the historical record is that the historical record is too short.

You must forgive my insistence, but this is so unbelievable. When one considers the changes in    thickness/volume it becomes even more unbelievable. I just think about the changes in ice  behaviors around the Bering and Chukchi and I admit I start getting emotional because the danger is so real. Let's not even mention the changes North of Svalbard.

2012 was obviously below the trend, but 2013 was not "on-trend". More likely that 2013  was a rebound year, literally caused by the 2012 ice loss. The mechanism being that in 2012 the Arctic ocean was colder. When it lost the ice cover for the first time in millennia the ocean gained a lot of heat and the atmosphere lost a lot of heat, creating perfect conditions for mass freezing. That massive freezing ceased as heat/salinity changes started having a greater impact. That heat keeps accumulating resulting in increasingly hot winters.

The physical basis for the 2013 refreeze and the behavior during the following years should be used to inform weak statistics.

We have a historical temperature record with multiple El-Ninos in it so its obvious when a misleading trend line is drawn in it by picking El-Nino/La Nina endpoints. If we only had one half of an ENSO period in the historical data it would be extremely difficult to get a reliable trend from it. Thats the situation we are in with regard to ice extent in the Arctic. The historical period is too short and trends in it are liable to substantial bias.

Context. Arctic average Temperatures since 1900

http://berkeleyearth.org/wp-content/uploads/2018/01/Arctic2017.png

Why would you expect for the ice to comeback or 2013 be on trend when temperature is way above trend? Isn't it more likely that the models are missing huge elements of a changing climate?

[Archimid]

The fingerprint is the particular pressure pattern in the Arctic summer in high loss years and the weather patterns that connect it to particular temperature variations in the Pacific. The same pattern is seen in the control run, and the historically forced runs and the data. 

To be clear, FTA:

The mechanisms of this teleconnection appear to be similar in observations and models, but the specific source areas and path of wave activity underlying the establishment of the high pressure in the Arctic are displaced in the model.

That important caveat, combined with the failure of most of the model runs to account for the Arctic loses tell me that the models are likely wrong.

The internal variability isn't the difference between the historical data and the ensemble mean, its the spread of the ensemble. Look at the band of grey lines in 1e. The internal variability is the difference between the top of that band and the bottom of that band.

To me this says that if the Models are sufficiently accurate representations of the system, then the variability is given by the spread of the models. The problem is that the models have significant difference with observations. They are not sufficiently good representations of the system. I think the paper does a fantastic job illustrating the big differences between the models and the observations. Fig. 4 is particularly interesting.

There's not enough historic data to pull the variability from it alone. That's one of the key points from this paper. In 30 years time it might be possible to assess the variability from the historic data, but the record is still too short to properly characterize the climate.

I agree with that. There is statistical uncertainty due to the short record. Luckily, mathematical statistics are not the only tool we have to inform our decision making process. We have physics that dictate that the warming will continue and will increase. There is every reason to believe that "internal variability" will vary as the climate changes.

[jai mitchell]

This lecture by Dr Charles Kennel Prof. Emeritus from Scripps discusses their paper currently under review in PNAS that correlates Arctic sea ice loss, East Siberian September Convection and the translation connection to 300 mb surface flows including increasing 'super strong' trade winds during December following the sea ice loss years.

This shows that the sea ice loss is the tail wagging the dog (ENSO) as opposed to the idea from the paper started in this thread, that 'natural variability' drove the ice loss.

https://www.youtube.com/watch?v=_b6k7lWOA6Y&t=32m56s

Started midway into lecture as he is discussing the findings.  I do recommend watching from the beginning though.

[Richard Rathbone]

I'd really need to see the final version of the paper, but at the moment this looks like another part of the same picture, not a different picture to me.

Pacific influences arctic in spring-summer, arctic influences pacific in autumn-winter, and that feedback loop powers multi-decadal variability.

[jai mitchell]

I was reading through this paper that was recently published by Michael Mann et  al on the aerosol forcing that has caused the AMO, and that the AMO was not actually natural variability.

We were discussing this earlier in this thread several years ago.  I attribute these to industrialization and low vs. high altitude aerosol forcing. 

In any event, the AMO appears to be an arosol-forced effect, not some kind of mystical 'natural' variability.

Here is the thread:  https://forum.arctic-sea-ice.net/index.php/topic,1920.msg110298.html#msg110298

here is the paper:

https://science.sciencemag.org/content/371/6533/1014.full

Multidecadal climate oscillations during the past millennium driven by volcanic forcing

Abstract
Past research argues for an internal multidecadal (40- to 60-year) oscillation distinct from climate noise. Recent studies have claimed that this so-termed Atlantic Multidecadal Oscillation is instead a manifestation of competing time-varying effects of anthropogenic greenhouse gases and sulfate aerosols. That conclusion is bolstered by the absence of robust multidecadal climate oscillations in control simulations of current-generation models. Paleoclimate data, however, do demonstrate multidecadal oscillatory behavior during the preindustrial era. By comparing control and forced "Last Millennium" simulations, we show that these apparent multidecadal oscillations are an artifact of pulses of volcanic activity during the preindustrial era that project markedly onto the multidecadal (50- to 70-year) frequency band. We conclude that there is no compelling evidence for internal multidecadal oscillations in the climate system.

[jai mitchell]

Well, it only took about 6 years for my prediction to be proven correct.

jai

early mortality projections, mainly in sub-sahara Africa  and SE Asia under a >2.0C warming scenario are in excess of 1 billion human beings.

I'd be interested in a source for that.

[binntho]

A link would have been useful!

[kassy]

It was discussed in a more appropriate thread:

Their number is a very raw estimate:

scroll up some posts for a short version of the story.
None of it relates to sea ice changes.