Internal variability plays a dominant role in global climate projections of temperature and precipitation extremes

Tags

review, uncertainty

Blanusa, M.L., López-Zurita, C.J. & Rasp, S. Internal variability plays a dominant role in global climate projections of temperature and precipitation extremes. Clim Dyn (2023). https://doi.org/10.1007/s00382-023-06664-3

It’s a paper on the relative uncertainties in predictions (not projections as the title suggests as it treats scenarios as an uncertainty) of extreme events arising from: internal variability, model uncertainty and scenario uncertainty. The extreme events are once in a decade daily precip or Tmax values and they are asking about the predictability of the exact number of events over a particular time window. Unsurprisingly, internal variability is a large component of this mix in the near future. For precip it remains a large component in many regions out to the end of the century. The intro says “If internal variability makes up a large fraction of the total variability, even a significant model improvement would only lead to a minor reduction in total uncertainty”. As internal variability constitutes a large fraction of the variability, they conclude that there is less to be gained from reducing model uncertainty and that ensembles are important. I agree with the latter claim.

I have one major overarching criticism, but upfront I’ll say this is a well written paper, clearly expressed and illustrated and it looks like a good deal of thought went into assessing sensitivity to various choices made, at least within the boundaries set by the original question.

Advertisement

Quantifying Daytime Heating Biases in Marine Air Temperature Observations from Ships

Tags

review, uncertainty

Cropper, T. E., Berry, D. I., Cornes, R. C., & Kent, E. C. (2023). Quantifying Daytime Heating Biases in Marine Air Temperature Observations from Ships, Journal of Atmospheric and Oceanic Technology https://doi.org/10.1175/JTECH-D-22-0080.1

On seeing what’s not there and the practical exigencies of flying suddenly without an engine

There is a story which goes something like this:

During World War II, the military wanted to add armour to their planes to stop them being shot down. Armour is heavy, so you can’t put it everywhere. To decide where to put it, they looked at the planes that came back and analysed where the bullet holes were. They decided to add armour to the areas with most bullet holes. “Not so fast”, said a passing statistician*, “These are the planes that came back. If you want to know where a bullet hole will bring a plane down, you need to look at where the bullet holes aren’t“. And so, minds blown, the air force put armour where there were no bullet holes and no plane was ever shot down again.

I suppose, there are various conclusions or lessons one can draw from this parable. When I first heard it, I was delighted by the idea that no data can be as informative as some data. Some conclude that the navy were dumb (they were clearly wise enough to keep a statistician around). Others see it as a symptom of genius seeing what everyone else missed. It’s a neat example of survivorship bias as well. Too neat, as it turns out.

Homogenization is for milk

Homogenization is widely, and sometimes wilfully, misunderstood.

How not to predict El Niño Modoki

In September 2021, Nature* Scientific Reports** published an article titled “El Niño Modoki*** can be mostly predicted more than 10 years ahead of time” [my asterisks]. El Niño, in its non-Modoki form, is notoriously hard to forecast even a year ahead. Why one of its variants should be predictable several years in advance is a question that piques ones interest so hard it screams.

AI, ai ai ai!

More adventures with AI chatters. I’m still trying to find an interesting use for the AI chatbots* to the extent that I would happily pay for them. I am yet to find it.

Was 2022 the warmest year on record?

No*.

Another Year of…

Tags

review

Cheng et al. have published their annual “Another year of record heat for the oceans” paper. I love the weariness of the title. Again. It happened, again, and it will go on happening again and again until greenhouse gas concentrations stop rising. Anyway…

One in a series of (very very) occasional paper reviews.

Teaching the next almost-right wrong thing

I’m still recovering from reading WUWT. This is… therapy. In the discussion about error and uncertainty various links and youtube videos were suggested as a guide to what was meant. A number of these referred to what is taught at A-level in the UK. A-levels are the exams you typically study for between the age of 16 and 18 (just before going to university if that’s your thing).

One of the things you learn when you go to university to study Physics (I imagine it’s true for other subjects too) is that your teachers lied to you all through A-level. They did it in the most well meaning and instructive way possible, but what they told you was wrong. As an undergraduate, they teach you a whole bunch of new stuff that is definitely 100% correct and true*. You have to kind of sneak up on the truth. The euphemism is that these are “successive approximations to the truth”, which sounds nicer and reassuringly sciencey.

Rounding errors – extreme edition

A couple of recent posts have dealt with rounding errors – the kind of measurement errors that occur when a thermometer is read and recorded to the nearest whole degree – and their associated uncertainties and how those should be propagated when performing simple operations such as addition and division that are used to calculate means. My discussion assumed, for the sake of simplicity, that temperatures were uniformly distributed within the rounding window – from half a degree below to half a degree above the whole number being read off the thermometer. This won’t always be the case of course.

If temperature variability has a small variance relative to the size of the rounding window then there can be appreciable non-uniformity within the rounding window. In that case, the putative distribution of errors associated with a measurement has a non-zero mean. Even if the temperature variability is large, the non-uniformity will still be there even if it is typically small enough to ignore.

One way to address this is to combine the measurement (in whole degrees) with some information about the distribution of temperatures. This can be done using Bayes theorem, but it involves horrible integrals of things over discontinuous intervals and… ugh.

It’s much easier to generate random numbers from a distribution, round them, and add them, many many times. This gives an idea of the shapes of the distributions and likely biases. The following plots are generated by selecting a set of measurements (whole numbers), generating values from a gaussian distribution – our prior for the “true” temperature distribution with the stated mean and standard deviation – which result in those measurements and then adding each group together separately (measurements and “true” values) and taking the difference. This gives a plot of the measurement errors. Repeated a million times to get an idea of the shape of the distributions.