MLO and MEI

Guest Post by Willis Eschenbach

In my last post, which was about the Mauna Loa Observatory (MLO) in Hawaii, Dr. Richard Keen and others noted that for a good comparison, there was a need to remove the variations due to El Nino. Dr. Keen said that he uses the Multivariate ENSO Index (MEI) for such removal.

And what is the MEI when it is at home? Here’s the description from NOAA:

El Niño/Southern Oscillation (ENSO) is the most important coupled ocean-atmosphere phenomenon to cause global climate variability on interannual time scales. Here we attempt to monitor ENSO by basing the Multivariate ENSO Index (MEI) on the six main observed variables over the tropical Pacific. These six variables are:

sea-level pressure (P),

zonal (U) and

meridional (V) components of the surface wind,

sea surface temperature (S),

surface air temperature (A),

and total cloudiness fraction of the sky (C).

Now me, I’m a bit wary of the MEI, because of the possibility of it sharing a variable with something that I’m investigating. For example, in the post I did on the Mauna Loa Observatory in Hawaii, cloudiness was a variable because I was looking at solar radiation. However, I used it because Dr. Keen used it, and because for the purposes of my post it turned out the considerations didn’t matter.

The effects of the El Nino don’t happen immediately, of course. In general, the effect of the El Nino on the global temperature lags the El Nino by a couple of months. You can determine the lag by using a “cross-correlation analysis”, which shows the correlation between the El Nino and the variable of interest over a wide range of lags.

So imagine my surprise when I did the cross-correlation between the MEI and the Mauna Loa Observatory temperature and got the following result:

Figure 1. Cross-correlation between Mauna Loa Observatory (MLO) temperatures and the Multivariate ENSO Index (MEI).

Zowie … I was expecting a two or three-month lag, but the peak correlation is not lagged a couple months, a couple of quarters, or even a full year. Instead, peak correlation is at no less than a fifteen-month lag.

Now the joy of science is in the surprises. When I get surprised, I don’t sleep right until I learn more about what it was that surprised me. I couldn’t figure out how it was that Hawaii got basically no correlation for six months, and then after that, the correlation kept increasing until it peaked at fifteen months.

So I made an investigation of the correlation of the MEI with the individual 1° latitude x 1° longitude gridcells of the planetary surface. As you might imagine, at a lag of one month you have the strongest correlation between the MEI and the tropical Pacific. Here’s that map:

Figure 2. Correlation, MEI and 1°x1° gridcells. The dark blue lines outline the areas where the correlation is less than minus 0.2. Red outlines the areas where the correlation is greater than plus 0.2.

You can see that the area of the central Equatorial Pacific has the highest correlation with the MEI. The light blue rectangle shows the NINO 3.4 area, which is used in the same way as the MEI is used, to diagnose the state of the El Nino. So the high correlation there makes sense.

Figure 2 also reveals why the correlation with Hawaii is so low at the one-month lag. It is because Hawaii (black dot above the left side of the light blue rectangle) is very near the edge between the red and the blue areas, where the correlation is small.

To investigate the longer lags, I decided to make a movie so I could understand the evolution of the El Nino variations as they spread out and affected other parts of the world. Here’s that movie. It shows the correlation of the MEI and the individual gridcells at periods from one month to 24 months and then back down again to one month.

Again, more surprises. The correlation dies away quickly in some areas, but in Hawaii it builds until about fifteen months, and then decreases after that.

.

How amazing is that? If you want to know what the temperatures at the Mauna Loa Observatory will be doing fifteen months from now, you can look at the MEI today.

To demonstrate this odd fact, here are the MLO temperatures compared to the Multivariate ENSO Index lagged by 15 months.

And that’s why I love climate science … because there is so much to learn about it.

[UPDATE] I got to thinking that I should do the same thing using the NINO34 Index … here is that result. As you can see, it is extremely similar to the MEI graph above.

Here in the forest, after doing most of the work in this post, I was moving my computer files into new computer folders yesterday evening and I managed to destroy about half of them … and my last backup was three weeks ago.

So I was up until 2 AM saying bad words and reconstructing lost functions that went into the bit bucket. Grrrr … then I spent all day today beating my files back into submission so that I could recreate the work that I’d already done on this post. However, it allowed me to clean up some poorly written functions, and I suppose anything worth doing is worth doing twice.

It was also another demonstration of my rule of thumb gained from living about 20 years on tropical Pacific islands, which is:

The Universe doesn’t really give a damn what I think should happen next.

Factcheck: True … the good news is in the corollary to that rule of thumb, which is:

I do have a choice in the matter: I can dig it or whine about it.

Endeavoring to follow my own maxims re my monumental computer blunder, I remain,

Yr. Obdt. Svt.,

w.

I Know This Gets Old But: When you comment, please quote the exact words you are discussing. Misunderstandings abound on the intarwebs. Clarity about your subject can minimize them.