BEST Just Proved I'm Right(?)

I've been writing more about the Berkeley Earth temperature record (BEST) lately than I planned to. I hope you'll bear with me though. There's just been a fascinating development.

Not long ago, the Berkeley Earth Twitter account posted two informative images. They seem to prove I've been right about BEST all along. Six months ago I said:

But here’s the thing. BEST is supposed to be the best temperature record. It has a website encouraging people to look at data on as fine a scale as individual cities. WHY?! If BEST can’t come close to getting things right for the state of Illinois, why should anyone care at what it says about the city of Springfield, Illinois?

At what scale does BEST stop giving imaginary results and start getting a right answer? It doesn’t at the city level. It doesn’t at the state level. What about at the regional level? Could it get temperatures right for something like, say, Southeast United States? Nope.

We can’t pick areas much larger than that. The Southeast United States is about half the size of Australia. It’s about a third the size of Europe. If BEST can’t get it right, how could it get Australia or Europe right? And that’s ignoring the fact the Southeast United States has far more temperature stations (per area) than either of those!

I get almost everybody seems to agree BEST gets things right at the global scale, but couldn’t we all agree there’s a problem if BEST can’t come close to the right answer when looking at entire continents?

Compare what I said to this image posted by the Berkeley Earth Twitter account:

2-9-BEST-Homog-Maps

The image contains three figures. The first figure shows the temperature trends of the planet for 1900-2014 if BEST doesn't perform any breakpoint calculations. It shows a great deal of variation. It even shows the cooling trend in the southeast United States I referred to

The second figure shows what happens if BEST only estimates breakpoints by using station metadata. That is, it shows what happens if BEST only messes with the data when it has some documented reason to think it may need to. The result is quite similar to the first figure

The third figure shows what happens when BEST performs its "empirical breakpoint" calculations. The results are radically different. Instead of showing a great deal of variation like the last two figures did, the third figure shows entire continents being almost identical

I've showed the incredibly limited spatial variation in BEST's results before, and I've suggested it is caused by BEST's breakpoint calculations, but now we have proof straight from the horse's mouth. As I said before:

I get almost everybody seems to agree BEST gets things right at the global scale, but couldn’t we all agree there’s a problem if BEST can’t come close to the right answer when looking at entire continents?

But it turns out the problem may be worse than I thought. This is another image the Berkeley Earth Twitter account posted:

Best-Homogenization

It shows BEST's breakpoint calculations have almost no effect on their global temperature results since 1960. They do, however, have a significant effect as one goes further back in time. The effect grows to several tenths of a degree as one gets back to 1850. It might get even larger as one goes further back, though I can't be sure since BEST didn't include the full record.

Now, it's obviously interesting to know ~20% of the warming in the BEST temperature record is a product of their adjustments. It's especially interesting since the main adjustments are "empirical." BEST compares stations to their neighbors to find problems in the data. Apparently what BEST finds when it compares stations to their neighbors is, "The stations don't show enough warming."

I think that's something BEST should have highlighted, or at least disclosed. I think BEST should have been up-front about this and said, "About 20% of the warming we find is due to us adjusting our data." I think it should have been made clear most of those adjustments are made without any documented reason for them.

Maybe that's just me. I don't know. What I do know is 24 hours ago, I never would have guessed BEST homogenizing its data would cause its global temperature results to change this much:

Best-Homogenization-mod

That's their graph with things removed to focus on the differences introduced by BEST's homogenization process. It shows BEST's homogenization warms temperatures a tiny bit after ~1940 and cools temperatures quite a bit before that point.

7 comments

  1. I was surprised how little of an effect homogenization of any form had post 1920.

    I also expected homogenization to flatten the curve prior to 1940. As you can see, it goes the other direction.

    I was glad to see that metadata based correction handles the issues well enough by itself.

    That also means that, if they want, they can just drop the empirical homogenization method from their list of tricks.

  2. Carrick, as I mentioned in the previous topic, I believe the reason homogenization has little effect in more recent times is due, at least in part, to baseline issues. The climate field BEST calculates is a form of taking a baseline. Because it takes that baseline only over a modern period, its variance is biased low in that modern period. The further you get from the modern period, the greater the variance will be. That's the exact pattern we see here.

    If I'm right, we could get more significant homogenization effects in the modern period by rerunning BEST's calculations but using an earlier period to calculate the climate field. If that's true, pointing to the limited effect of homogenization in modern times doesn't address anything. It's just pointing out an artifact caused by the baseline period.

    I was glad to see that metadata based correction handles the issues well enough by itself.

    Do you really know it does? I can't see a problem with their metadata breakpoints in these graphs, but that doesn't mean they're actually working properly.

    That also means that, if they want, they can just drop the empirical homogenization method from their list of tricks.

    I think they ought to. Failing that, I think they ought to publish their results with and without their "empirical breakpoints" and explain how calculating those breakpoints affects their results. Once they've done so, they can try to explain why we should prefer the results generated with the empirical breakpoints to those generated without them.

  3. ... main adjustments are “empirical.” ... Or possibly, "empriical" going by the graph titles.

    Tihs is conssitant with my buisness carere, mentroing juinor analsyts, who could spell prefectly well in mane body text, but seemed domed to inculde typos in the capitons of their powerpinot charts. We all now what their suposed to say, after all...

    Typically in version 1.0 of the documents. Which is why we'd have everybody on a team, regardless of technical ability, lay eyeballs on and provide comment.

    Humor and advocacy for crowd-sourced editing all aside, what BEST's graphics indicate to me is that global "warming" is not global at all, and the warming measured is happening in places that could easily benefit from some warming. Siberia needs warming like Mars needs women. AND, in regions where the temperatures average below freezing (near the interior Siberia arctic) an increase from, say -30 degrees to -20 degrees is NOT going to "melt" icecaps, drain glaciers into the ocean, and raise sea-levels. Instead we're talking about extending very brief growing seasons, providing a little more natural irrigation, and by and large helping eco-systems in marginally habitable zone. At the other end, the "hotspot" glowing in Antarctica seems to be adjusted OUT of the danger zone once meta-and-empriical considerations are considered. So much for the oft-stated fear that 1/6th (or whatever) of the entire planet's freshwater supply will suddenly be dumped into the seas.

    What is the picture intended to tell citizens of Europe, Australia, and the lower parts of North America that will inspire us to re-think our lifestyles, goals, and political preferences?

  4. Pouncer, I have no idea how good/bad global warming will be overall. I've spent some time looking at the economic predictions on it due to getting involved with Richard Tol's work on the subject, but I don't think that field provides much, if any, useful information on the subject. I currently view it like I view GCMs, but worse.

    That said, I don't think the overall effect is what will wind up mattering in discussions. I think regional changes are going to be people's primary focus. BEST is pretty much useless for that right now due to how much smearing it involves. There might be other things which allow us to draw useful conclusions about what effects will be seen at the regional level, but if so, I'm not very aware of them. I'm also not very interested in them. I'm not convinced we know anything particularly useful about what sort of benefits/harms will come from global warming at a regional level. With how shoddy our information is on the subject, it just winds up boring me.

  5. When the climate records were released in the wake of Climategate, there was I believe commentary to the effect that a third of the records showed warming, a third showed cooling, and the rest were flat. I should have saved this, because for me it is the killer. We live in specific environments rather than a global average. My country has a temperature that varies about 10k every day. How would I even be aware of a 2 degree change? How does this 2 degrees manifest? Will some places continue to cool while others heat up?

    Global averaging is a terrible way of communicating the impacts of extra co2 in the atmosphere.

  6. Diogenes, the details of the distribution of cooling/warming depends on the data set and how long you require the records be. Even so, I think it's always skewed toward there being more warming stations. I can't think of a time I've seen a data set with as many cooling stations as warming stations. Still, there are a substantial number of cooling ones.

    Global averaging is a terrible way of communicating the impacts of extra co2 in the atmosphere.

    Aye. I think the reason global averaging has been used so much in the global warming discussion is just how simple it is. People wanted to boil things down to simple talking points, and this is what happened. There's pretty much no scientific reason to focus on a single temperature value for the entire planet, yet that's what most people do now.

Leave a Reply

Your email address will not be published. Required fields are marked *