Klimaspirale Part 1: From a Video to My Own Data – Simply Explained
One morning a video climbed the ranks in a forum where people share things that grab their attention. A ring turns further outwards year by year, from blue to red. At the end the flat disc tips to the side and becomes a funnel. The video already had over a thousand upvotes. It was a reupload, and where it originally came from was no longer traceable.
One revolution is one year, January at the top. You don’t read numbers, you see a movement: for a hundred years the line trembles around the same spot, then it drifts, and at the end it drifts fast.
The data ends in 2021. I looked for a newer version and kept finding the same clip. I’d come across a tool called HyperFrames shortly before — it films a web page frame by frame and turns it into a video file. The spiral looked to me like exactly the case for it. I sent the link to Claude with one sentence:
“The video only goes to 2021 or something… can we rebuild that, maybe with HyperFrames?”
A mistake about the tool
Claude’s answer began with a rebuff:
“Straight talk up front: Hyperframes or any AI video generator is the wrong way to go here. The climate spiral isn’t a ‘generatable’ animation, it’s an exact data-driven visualization — every point is a real monthly value. An AI model would hallucinate you a pretty but factually wrong spiral.”
The warning was fair in substance. An image generator would paint something that looks like climate data without a single real measured value in it. But the premise was wrong: HyperFrames isn’t an AI video generator at all. Claude had taken the name for one and passed judgment without looking it up.
I objected mid-sentence:
“Of course I want to render it, I’m not stupid. Go research HyperFrames…”
Then came the research, and then the retraction:
“I was wrong. HyperFrames isn’t a generative AI video generator at all.”
What it actually is: an open-source tool that takes a web page and captures it frame by frame, like a very patient screenshot machine. Same input, same output, every time. No creativity, no surprises. Exactly what you want when every dot represents a real measurement.
Claude’s warning was right on the substance and aimed at the wrong tool.
The numbers
The numbers came first. A spiral without solid values is just a graphic.
The source is GISTEMP, NASA’s global temperature series — a public file with one number per month going back to 1880. The automated download came up empty, so we took a continuously maintained copy and pulled 1,752 values out of it: one for every month from January 1880 to December 2025. At that point NASA’s data already ran to February 2026, but we stopped at a full year.
Second-hand numbers are a problem when you go on to call them “NASA data.” So we ran spot checks against the original table, and they matched. The copy was a transport route, not a second truth.
What “warmer” means here
One year per line, one month, one number in degrees. What matters is what that number isn’t.
If it says +1.48 degrees, it wasn’t 1.48 degrees warm. It means: that month was 1.48 degrees warmer than usual, and “usual” here is the average of the years 1951 to 1980. Climate scientists call this a temperature anomaly — a difference, not a temperature.
Why so roundabout? Think of a house. The cellar is always cooler, up under the roof always warmer. An average for the whole house depends on which rooms have a thermometer in them, and you can’t copy the cellar’s value from the room next door.
With the Earth it’s the same, only bigger. Measuring stations sit in mountains, deserts and out at sea, delivering very different temperatures. But “two degrees warmer than usual at this time of year” works the same everywhere. That’s why the scale carries a difference, not a temperature.
Why the series starts in 1880
People certainly read thermometers before 1880. Individual records reach far back into the 19th century. But they sit almost entirely in the north, and for large parts of the Earth there are no values at all.
It’s like a survey about a whole country in which only one region answered: you have answers, just none that speak for everybody. For earlier periods you would need indirect evidence — tree rings, ice cores.
So 1880 is no special year in climate history. It’s the point from which enough measuring stations are spread across the Earth to make a global statement. Other series draw the line differently: Ed Hawkins, who originated the spiral as a form of presentation, starts his version in 1850.
The build day
Claude first built a quick draft to prove the numbers work: a simple animated plot. It looked right, but two things were missing. The tipping from the flat disc into the funnel wasn’t in it. And I wanted scale rings at 1, 1.5 and 2 degrees.
Claude itself noted that its own draft was a dead end for the tipping and recommended a rebuild in Three.js (a tool for 3D graphics in the browser). The reason: there, the flat spiral and the funnel are the same object seen from different angles. The transition is just a camera movement, not a second graphic.
At its core the graphic is a conversion of table rows into places in space. The month sets the direction. The deviation sets the distance from the middle — cold inside, warm outside. And time sets the height: 1880 at the bottom, today at the top. From above it’s a flat spiral. Tip the camera and the same construction becomes a funnel.
Then came the look of it. A white ring for the long-term average, bluish below it, orange-red above it. A slowdown at the end so the most recent years stand still instead of flitting past. A switch for English and German. Every change cost a new run, and one minute of video consists of about eighteen hundred frames. A good part of the day was waiting.
In between, a value readout kept wobbling back and forth on every frame, because in a normal typeface a 1 is narrower than an 8. The fix: a typeface where every digit takes up the same width.
The side question
While waiting for a render I asked whether the spiral could also exist as a web page — something you turn and tilt with the mouse. The camera move at the end of the video is really just a fixed path through a space you could also move around in freely. If the same scene runs in the browser, you decide where to look from.
That question was on the table before anyone on the project knew whether more current spirals existed. In the end it became the actual release.
From the same waiting time came one more decision:
“Let’s do the translucent cylinder only at 1.5 degrees (because that used to be the climate target).”
The cylinder is still there today. I didn’t think any further about the half-sentence in the brackets at the time.
The finding
A day later the decision to publish was made. Before that I wanted two things checked:
“Do two more things first: 1. Plausibility check: don’t want our display of the data to have an error in it. That would be embarrassing. 2. Is the spiral really not out there in a more current version than 2021/2022? That would also be embarrassing.”
There is a current climate spiral, from NASA itself. Their Scientific Visualization Studio keeps a version that gets updated monthly. Ed Hawkins, who originated the form of presentation, maintains his version too.
The occasion for the project rested on a mistake. I had searched myself and found nothing apart from the old file. That there’s a maintained page where a new version appears every single month was in none of my search results.
What’s left is a smaller and more interesting statement: the version in circulation just lags behind. The clip that gets shared everywhere ends at 2021. The lead image in the Wikipedia article is a still from 2017, calculated on an older data series. What goes through the feeds as “the climate spiral” is five to nine years old, while the maintained version keeps growing every month.
It got built anyway. The question about the web page never depended on being up to date. What NASA publishes is a video: sixty seconds, a fixed camera move, a viewing angle somebody else picked out.
The version I built runs in the browser. You can turn the spiral yourself, zoom in and out, pause and rewind, click individual months and switch the scale. That makes the data something you experience instead of just watching.
Part 1 of 4 in the Klimaspirale series.