That’s not the raw data. The original recordings were made by merchants on parchment. They measured the volume of water in a wooden box, to set the buoyancy for their loads
For the interested, here[1] is an article on an attempt to recreate and verify measurements made during the HMS Challenger expedition in the 1870s.
It was recently done so the full results aren't out, but one aspect they noted was that the traditionally-created hemp rope stretched about 10% so temperatures were taken at slightly deeper depths than expected. This can be used to calibrate the data from HMS Challenger.
I appreciated your comment because more discussion will better help everyone understand the various tranches of surface temperature observations.
I did a quick review, and appreciated the article because they were clear about how their methods different from the recordings. For one using different pressure sensors, and they mentioned the depth differential they measured would lead to variability in the ocean temp readings.
Hah. Shall I present it to you on a silver platter then?
If you read the NASA page, they explicitly cite GHCNd, a raw surface temperature and precipitation dataset that goes back quite far. There's many other similar datasets you can find if you're willing to look.
Check out the readme for the csv format description, and /by-year for the raw rows:
picked four stations at random[0] and it's just precip numbers, no temps, no humidity, no insolation, etc.
are you sure you linked what you think you linked?
[0] /by-station and then unclutched my scroll wheel and spun it for arbitrary amount of time, re-engaged clutch and clicked what was under the cursor. repeated 3 more times. i did a fifth, where the one i was looking at was identical to the fourth one, but had a 1 in the least significant portion of the station ID instead of a 4, in case it was like, "4" is precip, "1" is temps, and i happened to click "4" 4 times in a row.
HAHA you're completely right! or, and this is just some advice: don't tell strangers to look up data, link the data, and it not be what you said it was.
If i promise you punch and pie, you'd be pretty upset if it wasn't.
There are tons of raw data available freely and publicly. In my estimation, there is no comparable scientific discipline with a better curated data environment.
What exact raw data would you want? I am sure ChatGPT can throw together some python that will download the relevant data.
If you want the raw data, you'll have to go dig in the archives to find the log books and card decks.
This[1] paper goes into some detail on how the digital records were constructed from the log books, card decks and such. This[2] paper deals with an update of those digital records, including new digitization efforts. You can download the raw digital data from ICOADS here[3].
Regardless, ascii encoding isn’t raw data. You’re making software engineer assumptions. Statistical noise is introduced 4-5 steps before the data is recorded digitally.
Even after it’s digitized, more noise is introduced through recording errors and normalization.
To understand the original distribution, the entire workflow needs to have been recorded
Climate deniers are perfectly trained for finding some weak spots in any data anytime they want. It would be better for them to be trained enough to show at least any links to any studies though. It is so hard to convince a climate denier to give at least one climate-denying source for the sake of experiencing some laugher together.
you're right this is much to complicated and important for anyone to understand. just take our word for it that we have to make things more expensive, raise taxes, and restrict freedoms to fix it.
If you are serious about this, you know full well the data is out there. So stop asking for it and just go get it, go write some code and process it, then come back here and report your results.
i went and got the raw data, wrote some code to parse it and then process it, and my results:
MMTS locations are so close to heat sources and heat sinks, at least in the US, that any sort of debiasing appears to be a "guess."
statistically. 96% of them. thankfully NWS/NOAA/NASA/etc have started deploying wireless sensors, but unless they admonish the volunteers for placing their (NWS/NOAA/etc) dumb MMTS designs so close to heat sinks and sources, as if it was their fault, demanding that volunteers move the sensors to a location 20 meters from said sinks and sources...
you're just gunna have and continue to have decades of literally unusable data. But hey, hottest year on record!
I am not mad at volunteers. it isn't their fault the MMTS devices only came with 10 meters or whatever of cabling for the indoor-outdoor data. I would, however, like to see the rationale and meeting notes and design documents (and the reasoning and arguments thereof) for the MMTS; explicitly for use tracking climate trends.
anecdote: i have multi channel humidity and temp sensors that log to SD card. they have been logging for a long time. My outdoor sensors, as well as our cars, etc, show that our location is always 7F cooler than the nearby metro (20 miles) during the warm months. If we used my temperature data, i'd believe the trends. if we use the temperature data from the sensors they use at the airport, it's going to show warming - and i submit you can't de-bias that using the methods used for the IPCC and other reports. and when i say 7F cooler here than there, i mean on the thermometers on our cars (and multimeters, or even a liquid-in-glass carried around!). I also mean my location is consistently cooler than the forecast temperatures for the city.
There’s a ton of raw data here. Not sure what you’re looking for specifically but if you look for it the vast majority of the data used for studies like this is public.
related: I like to scare myself with the nice graphics of daily global sea surface temperature from the Climate Change Institute at the University of Maine.
If the Lightning OUTSOLD Tesla, is that really losing to them? Feels to me like an indictment of the scale that Tesla actually operates: an order of magnitude less than the big car makers. If Ford declares a truck that sold better than CT as a failure, it's because for their size it didn't sell enough. If that lesser number IS enough for Tesla, they're simply not a player in the same league as Ford.
This reflects a very common pronunciation of syllable-final Ls in English, called a vocalised L, but I've never seen it reflected in spelling in such a way. Very cool!
I'm extremely curious - did you go for that spelling as an intentional stylistic variation, or was it a typo reflecting your usual pronunciation?
The Toyota Rav4 outsold the Model Y in 2025. Toyota led all brands worldwide with 10.3 million cars sold by November. Tesla which sold 1.64 million, didn't make the cut for top ten brands.
Here’s the Tesla source.
Everything I’m seeing for the rav4 is dated before the end of 2025, so the data is projected for the rav4 to make a headline, not actual data.
