By Javier

A year ago I wrote an article at WUWT analyzing the recent upward trend in summer Arctic sea ice extent. Despite challenges of statistical irrelevancy, the trend has continued another year. Arctic ice experts, that have repeatedly predicted the demise of summer ice, don’t have an explanation for a 10-year trend that contradicts their predictions, beyond statistical variability or unexplained natural variability. They believe the upward trend will end any year, and there were high expectations that 2017 was going to be that year, due to the low maximum in March. As we will see a low maximum has no predictive value.

However, the upward trend was predicted by Divine & Dick in 2006, based on the analysis of Nordic sea melt-season ice changes for the period from 1750-2002, where they identified two periodicities of ~60-80 years and ~20-30 years.

“… our results suggest that the Arctic ice pack is now at the periodical apogee of the low-frequency variability. This could explain the strong negative trend in ice extent during the last decades as a possible superposition of natural low frequency variability and greenhouse gas induced warming of the last decades. However, a similar shrinkage of ice cover was observed in the 1920s– 1930s, during the previous warm phase of the LFO [Low Frequency Oscillation], when any anthropogenic influence is believed to have still been negligible. We suppose therefore that during the decades to come, as the negative phase of the thermohaline circulation evolves, the retreat of ice cover may change to an expansion.”

So, when nearly every expert was predicting the collapse of Arctic summer ice, these two Norway-based researchers correctly predicted the trend observed for the past 10 years.

In science your hypothesis can only be correct if it not only explains, but also predicts the behavior of the studied phenomena. Therefore, the hypothesis of Divine & Dick is superior to the more popular hypothesis that assigns sea-ice behavior to the anthropogenic effect. For this year’s article I have decided to examine the hypothesis of Divine & Dick to analyze the importance of natural variability on summer Arctic sea ice evolution.

I am using NSIDC monthly Arctic sea ice data for March and September available here. The data are plotted in figure 1.

Figure 1. Arctic sea ice extent

Then I define the melt value for the year X as the September X value minus the previous March X value, resulting in a negative number. The refreeze value for the same year X is the March (X+1) value minus the previous September X value, resulting in a positive number.

Plotting the Melt and Refreeze curves on the same graph produces the amazing result shown in figure 2.

Figure 2. Arctic sea ice melt-refreeze cycle.

Both curves are very close. So close that the winter growth in Arctic sea ice is >80% predictable based only on the ice extent lost in the previous melt season. In fact, I can predict that the Arctic will gain between 9.3 and 9.7 million square kilometers from this past September to March 2018.

I’m not sure how surprised you are by this result. I don’t doubt this must be known by plenty of ice researchers, but I haven’t seen it reported anywhere despite reading a great deal about Arctic Sea Ice. This result leads up to some very important conclusions:

• Arctic sea ice dynamics are driven by unpredictable melting. Freezing is reactive and largely predictable.

• This indicates a very strong negative feedback in action. A small melting is followed by a small refreezing, and a huge melting by a huge refreezing. Surprisingly this is not known by many ice experts that expressed surprise after the huge refreezing that followed the huge 2012 melting.

• The negative feedback stabilizes sea ice. Alarmism and spirals of death are unjustified.

• The much-touted albedo effect can only have a small effect in the Arctic, as the lost ice is recovered during the following “dark” season, during which albedo has no role. An example that evidence always trumps logic.

• Inter-annual changes in sea ice are due to the small residuals indicated in the figure by the colored areas. Red for decrease and blue for increase.

• Around 1998 Arctic sea ice changed its dynamics and entered a period of higher volatility. One possibility is that below a certain size the Arctic sea ice sheet becomes more unstable and sensitive to weather phenomena.

To continue, we must concentrate on the annual difference between melt and refreeze. I define the anomaly for a year as the summation of the melt that occurs on that year and the refreeze that starts on that year and ends in the next year. This produces another amazing chart.

Figure 3. Arctic sea ice extent anomaly

The anomaly graph is very homogeneous for the 38-year period analyzed, despite huge changes in Arctic sea ice. So, there are more interesting conclusions to be extracted from the data:

• The yearly anomaly appears to be range bound. No positive or negative changes bigger than 600,000 square kilometers are observed.

• Despite periods when the anomalies are skewed towards one side, overall the observed linear trend is flat at –53,000 square kilometers/year. This means no acceleration of the Arctic sea ice loss is observed for the 38-year period, during which atmospheric CO2 levels have increased enormously to values not observed in over a million years.

• This result supports the hypothesis that cyclical changes in ice cover, over time, average out. As opposed to the hypothesis that ice cover loss is accelerating due to an increasing anthropogenic effect.

Since the loss of ice during the melt season is the driving factor in the Arctic sea ice dynamics, I have constructed a very simple model to explore the relationship between natural and anthropogenic factors in Arctic changes. The model rests on unproven assumptions and is not intended to represent or predict Arctic sea ice changes. It is simply a learning tool that uses several of the proposed mechanisms acting on ice. The main assumption is that to be observable above the high noise of September ice data, the four main factors, thought to participate in the process, must be between 15 and 33% responsible for the observed changes.

Figure 4. Components of the Arctic sea ice melt model

The first component (A) is a 21.33-year sinusoidal oscillator that is set to explain 25% of the observed variability.

y = (-0.25) sin 0.2944 (x)

The lows of the cycle are identified at 1990 and 2012 based on local minimum ice values.

The second component (B) is a 65-year sinusoidal oscillator that is set to explain 33% of the observed variability.

y = (-0.35) sin 0.096664 (x+24) – 0.306

The low of the cycle is placed at 2007, when the current upward trend started, and when North Atlantic sea-surface temperatures started to decrease.

The third component (C) is the anthropogenic factor based on atmospheric CO₂ changes. It is set to explain 24% of the observed variability.

y = 0.5 – 3.2 Ln ([CO2]/290)

The fourth component (D) represents the long term natural variability, since the end of the LIA. It is essentially the ~ 1000-year cycle. Since it is very long term, it can be adequately represented for a short period with a line that is set to represent 17% of the observed variability.

y = (-0.34/32) x + 21.04

The model is initiated at 1980 at a melt of –8.1 million square kilometers

Such a simple model is not expected to adequately represent a complex phenomenon that likely responds to many more factors, but it reproduces the general shape and behavior of Arctic sea ice melt, and compares well with a polynomial fit to the data.

Figure 5. Arctic sea ice extent melt.

By comparing figure 5 and figure 1 we can see that the melt graph is extraordinarily similar to the September extent graph. As we have seen, Arctic sea ice dynamics are driven by the melting. The model therefore can be set to reproduce and project September Arctic sea ice data into the future. For that I have used RCP 4.5 scenario that contemplates a stabilization of CO₂ levels at around 540 ppm soon after 2100.

Figure 6. September Arctic sea ice extent.

While I don’t expect future Arctic sea ice data to follow the model, I do expect it to perform better than the models that are based mainly on anthropogenic factors. As I said the goal of the model is to examine the possible effect of the different natural and anthropogenic factors on sea ice dynamics. It can be seen as a graphical representation of Divine & Dick hypothesis with fictitious but reasonable values.

I do believe we are entering a period of Arctic sea ice stabilization, and even expansion, that should last until around 2042, and this is a prediction in stark contrast with IPCC’s ice models that see an end to summer Arctic sea ice by 2040-2080 for most scenarios and near constant decline until then.

High sea ice variability could produce some ice-free summers around 2075, but the conditions for the existence of summer Arctic sea ice are likely to remain for the foreseeable future. By 2100, atmospheric CO₂ levels are expected to stabilize in the more credible RCP 4.5 scenario, and the millennial cycle is expected to change phase, so there won’t be a net negative ice driver. From then on, Arctic sea ice should start growing for many centuries to come.