A year of blogging

Sometime around November last year, I started blogging regularly. Partly this was due to a long backlog of unfinished blog posts, partly it was in response to the threatened tidal wave of LLM-generated waffle, and partly due to boredom and frustration caused by the decline of Twitter and the pain of having to say the same thing on three different platforms to reach half as many people. Like many others, I was rethinking the whole social media thing and concluded that I liked blogging. If I was going to spend time doing this, I thought, then I should spend the time writing things rather than copy-pasting little snippets from one tiny window to another. There was also the issue that the vast majority of people who came to my blog came to read a story about an incident involving an orange¹. This bothered me. While it’s nice that so many people had read the story and enjoyed it and even nicer that some of them felt a tiny bit bad about laughing so hard, I wondered if I could get at least an equal number of people to read something else I’d written.

Also, I like writing. It’s how I work things out. So there’s that.

I started off with a post on Facebook’s short-lived foray into the world of LLMs, Galactica. The subject of LLMs came back a couple of times. Like everyone else, I was messing about with Chat GPT and wondering what it could do for me. I didn’t find it very helpful, but I did use it to quickly prototype a new webpage. What it can do remains, to my mind at least, rather amazing, but its proliferation into our everyday tools – search engines, word processors etc – seems awfully hasty given everything it can’t do, or does unintentionally.

Early posts touched on subjects that I’ve since revisited. The question of what 1.5C means came up early with a paper by Blair Trewin. As 2023 proceeded, this question become unexpectedly pressing. The question of whether we exceeded 2C on a particular day isn’t one I’d anticipated, but came rushing to the fore. This was part of an unexpectedly warm year, which also – accurate impression or not – seemed to be peppered by an unusual number of extreme events. Early posts in June were more of a comment on the difficulties of daily monitoring in general. There were more specific posts on Antarctic sea ice, which was exceptionally low for most of the year, starting with how the low extents were being communicated using standard deviations. This unnecessary and, I felt, unhelpful abstraction seemed like a distraction but it took me a few more posts to work out what really irked me with lots of graphs and a bit of particle physics². By October, the spike in global temperatures was too large to ignore (not least because I would soon have to say something intelligent about it). I started gathering together explanations for why 2023 was so damned hot. This eventually came full circle back to the question of how we’d know if we exceeded 1.5C or whether we’d know and how, formally, to measure such things (this last one being a critique of a paper I was an author on).

Global temperatures came up time and again. First there was the perennial question: was 2022 the warmest year on record? To answer such questions and those raised by trying to monitor progress relative to the Paris Agreement, we need to understand uncertainty in global temperature. IPCC used a limited number of datasets, but there are more to draw on, such as this list of infilled versions of HadCRUT (which is missing a more recent one). There were some interesting movements towards synthesising multiple estimates of global temperature into one but frustratingly the authors continue to eschew engagement with anyone who knows something about the data. I think that more global temperature estimates are needed and these can’t just be based on existing datasets. We need brand new ground-up estimates to help answer difficult questions like what is climate sensitivity (qualitatively and quantitatively) from a more solid foundation. More estimates plus digitization of hard copy records, and more efficient pull through of new ideas into existing global datasets.

Extreme event attribution came up time and again. I have negligible experience or understanding in this area, but it fascinates me, particularly how we construct counter factual worlds and how this relates to how we³ frame questions and answers. Some concepts seem to be hotly contested without good reason. In other cases, the reasons are fairly transparent even if presented with terrible terrible analogies. Event attribution continues to be controversial for some people. I’ve sporadically asked the world, “if climate change doesn’t cause heatwaves, what does?” From the lack of answers I concluded that nothing causes heatwaves. Such questions provoked a comically sad scandal about California wildfires. Even if we’re not talking about attribution, simply communicating extreme events can be difficult, particularly when people generalise or over simplify. For my part, I continue to be slightly uncertain what use extreme event attribution actually has. People say things about risk and adaptation, but I can’t shake the feeling it’s mostly PR.

I read some papers this year and wrote up some notes on a selection of them. These notes aren’t always⁴ well informed or accurate, but cover a range of things that I found interesting, useful or provoked some other sort of reaction. Topics included:

• how uncertainty depends on what specific question you are asking, or how you frame the answers. Part of my continuing puzzlement with the distinction between predictions and projections and how these are communicated.
• Overfitting
• Ocean heat content in 2022
• Quantifying daytime heating biases in marine air temperature obs from ships which ought to free up a huge amount of currently unusable data.
• Calculating a pre-industrial baseline for Australian data.
• The magnificently named excusable ignorance objection and how it pertains to loss and damage etc.
• Sometimes papers raise more questions than they answer, like this one about warming in the mid-to-high latitudes of the southern hemisphere
• The climatologists obsession with oscillations.
• Cautionary tales about using mapped data which are often uncertain and easy to misuse.
• And sometimes, the papers come in threes.

In addition to these interesting and/or thought-provoking papers, there are the kinds of papers and blog posts that provoke a whole grab bag of negative responses: sarcasm, withering scorn, resignation. It’s always fun to write this kind of thing and these posts have proved quite popular, but at the same time my feelings about them are ambivalent. This started with a series of posts on a bad uncertainty analysis, nonsense about rounding errors, even more on rounding errors, bad logic and komodo dragons, why I don’t generally engage directly with sceptics (and again), embarrassingly basic science errors, the unique argumentation style of Pat Frank, a paper the Royal Society should be ashamed of, and finally the whole saga of a rebuttal of a paper in GRL.

Other posts aren’t so easily categorised and were more or less serious. There was a post looking into the background of the canonical example of sampling bias. A rambling essay about understanding and misunderstanding. My reflections on and experience of being interviewed, something that I hate very much. A longish post that exists only because “Weather is how climate happens to you” seemed too short on its own to constitute a whole blogpost. And some thoughts about the difficulty of understanding and predicting something (climate) that happens over decades.

And finally, there’s the complete nonsense. A sort of restaurant review about the horrors of German sushi (rescued from a twitter thread and expanded). A list of books I read in 2022 and only sketchily remembered. My unapologetically negative feelings about large conferences. A spurious list of reasons that climbing made me a better scientist – an allergic reaction to the notion that everything you do must be in service to your economic worth. A collection of misguided data visualizations. A review of the book of the film of the book “We can remember it for you wholesale”. And, last but not least, a footnote turned into a blogpost about new shoes.

Then there are the blog posts that never got written: “A bag of hair⁵“, “the canard in the coal mine”, “sigh comms”, “how the Swiss army destroyed the universe”, “more bad news about the good news about the bad news about the…”. The list of unfinished posts has grown faster than the list of completed posts, so in at least one sense, I have been completely unsuccessful. Though, as for that, I’m not sure exactly what success looks like.

Of all these posts, only one did better than the orange incident, largely because it got picked up on reddit. The second most popular was “Describing the water” which ended up on the front page of some aggregator website (and mailed to its subscribers) with a spectacularly misleading precis. After that, there is the continually growing post on why 2023 was so warm, it’s popularity due in large part to my repeated tweeting of it rather than any particular quality or usefulness. Altogether, the posts got far more views than the orange one, around 20,000 more. Most of that comes from Twitter, but quite a lot come via other blogs (mostly Moyhu, And then there’s physics and Sterna Paradisaea), which is heartening.

Occasionally, someone I meet will say they’ve read one of my blogposts or even found one useful. This always causes mild panic and an urgent need to apologise. Mostly though, I enjoyed writing them. If you did enjoy any of them or found them useful (if anyone is still reading at this point) that makes me happy too.

1. If I’d known it would have had that reach I would have given it a bit more of a polish. ↩︎
2. I was a particle physicist in a previous life. ↩︎
3. Not sure where the “we” came from here. I’d like to use this kind of information in monitoring reports, but there’s always some objection to doing so whether internal or external. For a start the whole framing of EEA seems misleading. Comparing a NAT world with an ALL world maximises the signal, but doesn’t strike me as the most pertinent comparison. ↩︎
4. Ever. ↩︎
5. This one appears twice for some reason. ↩︎