Antarctic Sea Ice Index (part 4)

I explore the relationship between sea ice extent, my newly-minted grand continental anomaly, sea surface temperature, NSST v6.0 ocean/land anomalies & atmospheric CO2

This is where I pull together the BIG SIX anomaly series. I fancy starting with something simple to understand like a correlation matrix and follow through with exotic cross-correlation function (CCF) plots to see how pairs of series bounce off each other over time. With six time series there are a total of 15 paired CCF plots to be crayoned but since this is all about sea ice I’m going to whittle that down to just five. The time period will be 1979 – 2018 (n=40), which should be plenty to detect significant correlations.

To understand how the ‘Dee’ variables have been derived you’ll need to look back through my publication archive landing here for sea surface temperature and here for land surface temperature. Right then, here’s that first spanner, being an all-in-the-pot correlation matrix – alarmists had better look away now:

This might look busy but ignore everything apart from the first column headed ‘Mean Sea Ice Extent (NSIDC SII)’. If you run your eyeballs down that column you’ll see a statistically significant negative correlation with Sea Surface Temperature (Dee) of r = -0.531 (p<0.001) and a statistically significant negative correlation with Lower Troposphere Ocean (NSST v6.0) of r = -0.440 (p<0.001). Nothing else pops up and sings a song. In plain English this means the temperature of the surrounding wet stuff is determining what is happening to sea ice, which makes perfect sense. There is no evidence of an overall association with land surface temperature and there is no evidence of an overall association with atmospheric CO2 – there’s a tasty result right there!

If we stick with Sea Surface Temperature (Dee) and move along to the second column we discover a negative correlation between Sea Surface Temperature (Dee) and Atmospheric CO2 (Mauna Loa) of r = -0.645 (p<0.001). That’s right… a negative correlation! Thus, as atmospheric CO2 has been rising the Antarctic seas have been cooling. I appreciate this is one of those WERGH moments so we better take a look at the raw data to convince ourselves that everything is hunky dory:

So there it is in blue and grey! What we have here are combined mean annual sea surface temperature anomalies based on the Hadley, ICOADS and NOAA ERSST datasets plotted against the Scripps Mauna Loa primary record for atmospheric CO2. There is no doubting the overall negative relationship but there are a couple of interesting dynamics that are picked out by the snaky green LOESS line, along with a wacko outlier for 2017.

When we see a point sticking out sorely like 2017 does we ought to be asking questions as to the reliability of not one, not two but three of the biggest and best sea surface temperature datasets in existence. Right now I’d rather believe the Scripps crew at Mauna Loa than the Hadley, ICOADS and NOAA crews. Rubbing the salt in this wound is lack of a wacko relationship with atmospheric CO2 for the NSST v6.0 ocean anomaly (r = -0.226, p=161). So what are the Hadley, NOAA and ICOADS crews doing with their SST gridded data products that NSST’s lower troposphere series is not?  Mighty interesting, to say the least!

Given that the CO2 record has been consistently rising with each passing year then we can also use this axis as a rough proxy for the passage of time. We thus arrive at a somewhat surprising conclusion that Antarctic seas have been cooling whilst atmospheric CO2 has been rising. How strange is that? Well, it’s not that strange when you stop following the science™ and start thinking about how complex the climate really is compared to our current level of understanding.

So how does the UAH NSST v6.0 Lower Troposphere South Pole ocean region satellite data fare in comparison, especially given an insignificant correlation of r = -0.226 (p=0.161) with atmospheric CO2? Let us take a look…

A beard stroking moment! We’re back to pancake batter platters. Looking at this we would conclude there is no relationship between atmospheric CO2 and the temperature of the Antarctic region oceans. In fact, if we dare look down the column for atmospheric CO2 we discover absence of a statistically significant correlation with Land Surface Temperature (Dee) (p=0.387) and Lower Troposphere Land (NSST) (p=0.198). It would seem that the rise in atmospheric CO2 counts for nothing when it comes to the Antarctic climate, and I suggest alarmists sit quietly with some hot, sweet tea for a while.

The Second Spanner

I guess what we better do next is rollout the cross-correlation spanner to take account of lags in thermal transfer from atmosphere to ocean to sea-ice just in case some relationships are hiding away.

SIDE NOTE: Positive-going bars peeking beyond the 95% upper confidence limit indicate a statistically significant positive correlation. Negative-going bars peeking beyond the 95% lower confidence limit indicate a statistically significant negative correlation. Bars at positive lags mean the variable of interest (dependent) is changing in time after the independent variable that is listed first, and thus the relationship may be causal. Bars at negative lags indicate the dependent variable is changing in time before the independent and therefore the correlation cannot be causal.

Here we go…

We plough straight in with a cross-correlation plot of atmospheric CO2 with mean sea ice extent. This looks very exciting, with a good-size positive correlation hitting r = +0.591 (p<0.001) but a positive correlation means sea ice increases in extent with increasing atmospheric CO2. As if that wasn’t whoopsy enough please note this peak correlation arrives at a negative lag of 3 years. What is happening, then, is that Antarctic sea ice increases in extent a full 3 years before atmospheric CO2 rises; and it will also decrease in extent before atmospheric CO2 decreases. In a nutshell sea ice is doing its own thing first and foremost and atmospheric CO2 is following, though the relationship between the two may well be coincidental, with the passage of time serving to create an illusion of causality - what I call ‘The Time Trap’.

---

We follow through with a cross-correlation plot of sea surface temperature derived by myself from a combination of Hadley, ERSST and ICOADS datasets with mean sea ice extent. This also looks very exciting with a good-size negative correlation r = -0.531 (p<0.001). A negative correlation is what we expect to see for it means sea ice decreases in extent with increasing sea surface temperature. We may note the peak correlation is centred at zero lag, which means the effect is instantaneous; that is, it occurs within the same year. This result at zero lag is the exact equivalent to the value of r = -0.531 we saw in the bivariate correlation matrix at the outset – a munch on a biscuit or two at this point might help clarify the connection between the two spanners. The eagle-eyed and curious will spot some spill over into a second year (lag +1) and a most bizarre situation in which there is spill over going back in time (lag -1). This is totally normal behaviour when using cross-correlation on two oscillating series.

---

We arrive at the cross-correlation plot for my grand Antarctic land surface temperature anomaly with sea ice extent only to discover nothing is popping up as statistically significant. This drills home the point that sea ice doesn’t give two figs for what the land surface temperature is doing and only listens to the sea in which it floats. This sure makes sense to me but try telling that to the BBC!

---

I confess to being rather taken with the UAH NSST v6.0 satellite dataset and here it is in a cross-correlation with sea ice extent, this version using the lower troposphere South Pole ocean anomaly. We may note the smashing negative correlation of r = -0.440 (p=0.005) at lag zero matching the initial matrix which tells us that sea ice melts away the same year the oceans warm. The interplay between ocean surface temperature and lower troposphere temperature sitting directly above it is going to be complex and worthy of a future article series.

---

The final slide in the series of five and we observe what we already know, and that is Antarctic sea ice doesn’t care what the land temperature is doing.

On The Stove

Well, that’s a darn useful start for the Antarctic sea ice model wot I am developing! Clearly the sea is king so I need to figure what might be affecting sea surface temperatures. Alarmists will point directly to CO2 but we’ve unearthed a mighty strange and somewhat inverted relationship between atmospheric CO2 and sea surface temperature (as estimated by Hadley/NOAA ERSST/ICOADS), so I better run another cross-correlation plot to get a handle on this:

Well, well, well! There’s a whole palisade of negative correlation coefficients at lags from +3 years to -12 years, with the majority of big correlations sitting around lags of -6 years. This means sea surface temperature, as estimated by the Hadley/NOAA/ICOADS crews for their gridded products, starts to change before changes in the atmospheric CO2 lock into the same pattern. Since CO2 is rising persistently then we’re looking at a drop in Antarctic sea surface temperature that correlates with a rise in CO2 several years later. This is well bonkers!

I can understand atmospheric CO2 reacting to a change in sea surface temperature through gas exchange but warming oceans result in outgassing and therefore increased atmospheric CO2; but this well-established relationship is flipped on its head! My guess is that there is something radically wrong with the data and I am inclined to believe the Scripps crew at Mauna Loa got it right. This points a scrawny finger of suspicion at the three main organisations that bring us sea surface temperature data - just how much are they making up or fudging ‘correcting’ these days? This would be a conspiratorially-flavoured frippery were it not for climategate emails like this:

For the record we don’t have this same dodgy issue with the UAH NSST v6.0 lower troposphere south pole ocean anomaly:

It strikes me that if I am to develop a half-decent model for Antarctic sea ice I better trust the NSST satellite data! Until then…

Kettle On!