A Small Journey With Some Data Analysis
No conclusions, just an interesting learning experience, and, some technical stuff for spice.
Recently I read a series of interesting articles (1) (2) (3) by Jaime Jessop on her always interesting site “Climate & Covid: Challenging Unenlightenment 'Science“. This site has given me a source of alternative views on both climate/weather and the covid experience. These particular articles, linked above, spurred me to take a small data driven adventure. So, this little note is just my journey, based on Jaime’s, well reasoned, analysis of what might be the cause of the rise in temperature in the recent past and present.
Another reason to take this journey, was my desire to learn how to use “Jupyter lab” and the python environment, to do some data analysis. This has been a heroic procrastination exercise, by me, for quite some time. The articles gave me a push in the right direction and a chance to play with some real world data, on an interesting topic I just read about and see where I might get to.
I will attempt to summarize the articles, but, please read the original articles by Jaime, linked above, for a better understanding. This summary is the bare bones, and, probably represents poorly an intriguing story. In any case, the short story is, there is a temperature spike upwards (proceeded by a long linear trend of temperature rise). The spike starts around the beginning of 2023. This is a nice little spike as can be seen, on the plot, below (to the right of the red vertical line). I added a zoomed in view below the first plot to make this clearer. You can also go to the official site for more details, not to mention the excellent articles from Jamie, linked above.
The sudden rise in temperature, as seen from the average satellite global temperatures, just right of the red bar above is certainly an interesting data excursion, though, not without previous (differential) excursions near this magnitude (thus far). This, reasonably special heating event, gave rise to three hypothesis which attempts to explain this curiosity, they are as follows:
This is the direct result of the nasient el-ninio effect causing warming.
This is the effect of the massive Honga Tonga volcanic eruption spewing massive amounts of water into the stratosphere.
Perhaps it’s a little of both.
Up front, I liked the analysis given by Jaime, so I favor number two (I actually mean the numeral above, so there is no confusion). However, there are more than a few scientists who favor number one. I decided to down load the datasets relevant to this problem and have a look myself.
The two main datasets are:
The UAH satellite temperature dataset, which can be accessed here.
The El Ninio SOI (Southern Oscillator Index) dataset, here.
Explanation of the datasets can be read from the sites, and will not be repeated here.
There is one thing that I observed, specifically from the SOI data, and that is, The SOI is a differential measurement (namely pressure differences between Tahiti and Darwin, both places I would love to visit. This makes the fluctuations less sensitive to longer term and more global changes, as that would be the same at both locations, this would imply that the longer term linear trend, in the temperature dataset, will likely not be seen in the El ninio dataset.
This can be seen in the plot below, of the el-ninio SOI dataset. The sSLP curve is equivalent to the SOI (It took me awhile to realize this). You can see that the data bounces around zero without much, if any, longer term changes present.
The sSLP vs SOI issue was concluded, for me, after checking values from their site for both curves, The sSlp data values are the same as their SOI values. So without confusion, I shall use the sSlp time series as the El Ninio (SOI) time series in this analysis. Also, it is important to observe that the SOI moves negatively to the temperature data, so the spike (peak) we see in the previous temperature plot is an obvious through, or negative value, here. Even at this stage, we can see the trough is not exceptional compared to other troughs in this time series, which might be an early indication of its functional importance ( or more likely, lack of importance) in the causation of the current temperature spike.
With the preamble about the linear trend in the temperature data not being present in the El Ninio data. I proceed to try to remove the linear trend in the temperature data to make the two datasets more comprable. The plots below shows the linear fit to the temperature data and its subsequent removal.
Well, that looks ok. I will now put the two datasets together and see what we see.
As can be seen above, the El ninio SOI data values (yellow curve) are much more extreme than the temperature (green curve). It oscillates across a much larger interval of values than the temperature data. However, the good news is, they seem to fluctuate inversely from each other, suggesting a causal relationship.
To remedy difference in scale of the two time series datasets, I will equalize the RMS (root mean squared) amplitudes of the two data sets. I will use the temperature as the reference and scale the SOI dataset to fit. Equalizing the RMS values makes the energy of the two series alike. I will also multiply the El ninio data by negative one, so that they are more directly comparable. Having done this the plot above becomes the plot below.
Generally the basic shapes of the two are looking reasonable. We can see that the El Ninio SOI data (in yellow) tends towards more extreme values, in other words, to over shoot the satellite temperature (in green), this is with the large obvious exception of the data at the far right. It is under the green curve as can be seen after the Honga Tonga Eruption (in red). The trend there, has the El nino data very much less than the satellite temperature. This is unusual compared to the rest of the time series, again, pointing at the el ninio effect as not being the whole explanation.
Another observation can be seen on the plot above. The yellow curve (SOI)- appears to be lagging behind the green curve (temperature). To see the magnitude of this lag, I computed the cross-correlation of the two time series and plotted it below.
A rough examination of the cross-correlation curve show there is a dominant peak, which is almost twice the magnitude than the other peaks. This suggests the two data sets have a good correlation at the red line. The red line is not centered around zero lag, it is moved to the right. This means the el ninio (SOI) lags behind the conditioned satellite global temperature data. To see this more clearly we will zoom in around the peak value.
From the zoomed in view, we can clearly see the central peak indicated by the red vertical line is indeed shifted from zero lag. In fact, it is shifted by about four months. This has been confirmed by looking at the data values.
When this shift is applied to the el ninio (SOI) dataset, the result can be seen on the following plot.
As can be seen from the plot above the peaks and troughs, in general, have better alignment. To me, this chart shows the current miss-match even more clearly. There does not appear to be any other peak with this level of miss-match. A less obvious observation is the rise of the lagged SOI curve does not align with the rise of the temperature peak, which is quite consistent among the previous other peaks on the graph.
That my friend, for me, is a strong indication that the current el ninio episode does not fully explain the observed temperature increase. A likely alternative candidate for the additional temperature rise, as expressed so well by Jaime Jessop, is potentially due to the Honga Tonga eruption. Please read her articles linked above for her lucid explanation. If I try to summarize, it would just make it less clear. As I said the articles are linked at the beginning of this note.
There is one more display, that I find informative, and that is a cross-plot of the temperature directly with the SOI datasets.
On this plot, one can see that most of the values are clustered around the origin. As you move towards the NE you are heading into El ninio territory (higher temperatures and negative SOI), and, if you move towards the SW you are heading towards la ninia territory (lower temperature along with lower negative SOI). I highlighted the last three months where I have joint data for the two time series. These three points are colored in yellow (Aug. 2023), orange (Sep. 2023) and red (Oct. 2023). As you can see this trajectory is quite anomalous. The trajectory is towards the east (meaning no real el ninio contribution) and the red spot is in no-mans land. The Sep. point is anomalous from the trend (potentially due to noise in the datasets), however, it is not a part of the pack either. So viewing the datasets in this format, and you are inclined to explain the temperature spike as dominated by the current el ninio, you are fighting the preponderance of data contained in these time series. This observation, again suggest there is other forces in play. At this time, I am inclined towards hypothesis three with a strong dose of hypothesis 2. In other words, it seems to me the el ninio has some effect, but, there is a stronger driver that is not el ninio. I think the Hunga Tonga eruption is a very strong candidate for this driver.
I should add, all this is dependent of the manipulations of the time series I have undertaken, however, all that I have done is quite straight forward. Any other choice would require more hands on alteration than was done here. But if there is expert knowledge that would allow this, then, so be it. I am satisfied with my current view and will wait for additional information before I change my mind.
Well, that was fun. I have used up a lot of my play time on this. Guilt is setting in, so I best do some physical work and get our little home ready for winter.
Jaime Jessop finds climate scientist’s late admission of Hunga Tonga effect