If you want to know what climate change will look like, you need to know what Earth's climate looked like in the past — what air temperatures were like, for example, and what ocean currents and sea levels were doing. You need to know what polar ice caps and glaciers were up to and, crucially, how hot the oceans were.
"Most of the Earth is water," explains Peter Huybers, a climate scientist at Harvard University. "If you want to understand what global temperatures have been doing, you better understand, in detail, the rates that different parts of the ocean are warming."
Easier said than done.
To know how ocean temperature is changing today, scientists rely on more than a century's worth of temperature data gathered by sailors who used buckets to gather samples of water.
It's the best information available about how hot the oceans were before the middle of the 20th century, but it's full of errors and biases. Making the historical data more reliable led researchers on a wild investigation that involved advanced statistics and big data, along with early 20th century shipbuilding norms and Asian maritime history.
"I sometimes joke with my friends, 'I'm not only a climate scientist, I'm a detective!' " says Duo Chan, a graduate student who led much of the analysis for an influential study published early this year.
The underlying problem Chan and Huybers were dealing with is that different countries used buckets made of different materials, in different sizes, on different lengths of rope — all things that could change a temperature reading.
For example, the water in a midsize canvas bucket can lose up to 0.5 degree Celsius over the course of just a couple of minutes, says Chan.
"Half a degree doesn't sound like a big deal, right? However, if you look at the whole global warming, it's only, like, 1 degree," Chan explains. "Every 0.1 degree matters a lot."
He and Huybers set out to find and correct those tiny errors and biases within a massive database of historical sea surface temperature measurements maintained by the National Oceanic and Atmospheric Administration, with the help of researchers at the National Oceanography Centre in the United Kingdom.
The database has millions of entries from more than 100 sources, including Japanese whaling ships, Dutch naval vessels and a Norwegian Antarctic fleet. It's difficult to figure out how reliable any given measurement is.
"This is like if someone left you all their receipts that they had ever spent during their lives, and you were trying to piece together what they had been doing," Huybers says.
"It's a big data problem," says Chan, a "statistical nightmare."
They approached the nightmare from a novel angle: What if they could compare the measurements made by sailors from different countries, to see if some countries were systematically warmer or cooler in their temperature readings?
To do that, they paired up measurements that happened when ships were close to each other, so sailors were measuring the same part of the ocean at around the same time, and then looked for patterns. The starkest pattern had to do with temperatures taken by Japanese ships in the 1930s; the measurements were too cold.
But why? Chan hypothesized that bigger ships might be to blame — perhaps buckets full of water were swaying in the wind for longer as they made their way up to higher decks, losing heat in the process. He analyzed Japanese ship data and even learned to read Japanese to do it, and he found that, indeed, Japanese ships had gotten taller.
But that all ended up being a red herring. Bigger ships weren't to blame for the erroneously low temperature measurements. The answer was even more mundane and couldn't be solved using complex statistics or fancy computer models. Instead, the answer was written on an old U.S. Air Force document.
While Chan and Huybers were analyzing their millions of data pairs, researchers Elizabeth Kent and David Berry at the National Oceanography Centre in the United Kingdom had been going through thousands of pages of old documents about the data as part of their ongoing work to maintain the full sea surface temperature database.
The Japanese data in question had been digitized by the U.S. military after World War II, and one day Kent emailed Chan and Huybers with a clue.
"[She] sent us an email with just one PDF attached and it was a scan of a U.S. Air Force data sheet and she had circled part of the data sheet and she said, 'Hey, look at what we found here!' "
She had circled the word "truncation."
When the data was digitized, the U.S. military had dropped everything after the decimal point. A measurement of 15.1 degrees and a measurement of 15.9 degrees were both recorded as simply 15 degrees. Repeated over and over, those missing tenths of a degree added up to artificially cold measurements.
Kent says she had mixed feeling when she realized the data had been rounded down to the nearest degree.
"Of course, it was great that we knew how the data had been handled, so we can treat it appropriately in our analyses," she wrote in an email. "But it's also very frustrating that we don't have the full precision of each observation as it was recorded."
Still, knowing what happened allows scientists to understand what they're working with. The newly corrected data has far-reaching implications for climate science. Sea surface temperature is a big part of every major climate model, and for decades scientists have struggled to understand how the Pacific Ocean's relatively cool temperatures in the early 20th century fit into the overall trend of a warming planet.
"Often when you find errors in data it makes your life more complicated," says Huybers. "In this case it's actually the opposite."
Of course, if you zoom out, a warmer Pacific isn't particularly great news for humanity. "If you correct for the Japanese measurements, then basically you would warm up the global trend," says Chan. "That actually implies or suggests that maybe the human contribution is greater than what we used to think."
But, he says, having a more accurate understanding of the past climate is important if scientists want to understand what the future holds, which will be crucial if humans hope to avoid the most catastrophic effects of climate change.
DAVID GREENE, HOST:
Now we have the story of a strange and elaborate quest that all hinged on a couple decimal points. Scientists today need to figure out how much hotter the oceans are now than in the past, but the old measurements are unreliable. And figuring them out is more difficult than it might seem.
Here's NPR's Rebecca Hersher.
REBECCA HERSHER, BYLINE: If you want to know how much hotter the oceans are today than they were, say, in the 1800s, you need two things - information about how hot the oceans are now and information about how hot the oceans were then. And it's that second part, the historical measurements, that graduate student Duo Chan was interested in.
DUO CHAN: I sometimes joke with my friends - like, oh, I'm not only a climate scientist, I'm a detective.
HERSHER: His detective work started with a tome that sent him down a crazy rabbit hole. The book listed all the ocean temperature measurements taken by sailors going back to the 1800s.
CHAN: Yes. You'll see that there are, like, more than 100 different data sources.
CHAN: So what we...
HERSHER: Wait. Can I read a couple of them?
HERSHER: So a Japanese whaling fleet...
HERSHER: ...Norwegian Antarctic whaling factory ships, Netherlands marine.
CHAN: Yes. And here you have, like, the Japanese Kobe collection.
HERSHER: In all, there are millions of measurements from all over the world. And most of them, up until World War II, were collected using buckets.
(SOUNDBITE OF ARCHIVED RECORDING)
UNIDENTIFIED NARRATOR: First, the sea temperature is taken. The movement of warm and cold currents affect the weather a good deal.
HERSHER: This newsreel is from 1947.
(SOUNDBITE OF ARCHIVED RECORDING)
UNIDENTIFIED NARRATOR: The met assistant heaves a canvas bucket over the side and collects a sample of seawater and then takes its temperature.
HERSHER: The temperature got written in a logbook. But sailors generally didn't write down other information, like how big the bucket was or what it was made of or how long they waited before putting the thermometer in - all things that could change a temperature reading.
Peter Huybers is the lead scientist on the project.
PETER HUYBERS: This is like if someone left you all their receipts that they had ever spent during their lives and you were trying to piece together what they had been doing.
HERSHER: Duo Chan, the grad student, was undeterred. He had an idea. He took all the measurements and looked for ships that had passed near each other. So they had both measured the temperature of the same piece of ocean at the same time about. He found nearly 18 million pairs of measurements like that and looked for patterns and found one big one. The measurements from Japan in the 1930s seemed to be a little bit too cold - but why?
Duo thought maybe Japanese ships were getting taller, so the water cooled off as it was hoisted all the way up to higher decks. So he found records about Japanese ship sizes. And in order to read them...
CHAN: I learned Japanese.
HERSHER: You did?
CHAN: I did.
HERSHER: Turned out, ships had gotten slightly bigger. But all that work ended up being for nothing because ship size actually had nothing to do with the cold measurements. It was something much more mundane - rounding. A colleague in the U.K. happened to send them some vital information, an old Air Force document. After World War II, the U.S. military had taken all the ocean temperature measurements from Japanese ships and digitized them. And when they did that, they dropped the decimal place.
So if the water was 15.1 degrees on one day and 15.9 degrees on another, both temperatures were recorded as just 15 degrees. Overall, that made the measurements artificially cold by about half a degree.
CHAN: It's amazing. Like, people keep a record of what they did.
HERSHER: The team published their research this spring and it's made waves - because of their work, we now know that the Pacific Ocean used to be slightly warmer than we thought, which will help make climate models more accurate and potentially help avoid catastrophic climate change.
Rebecca Hersher, NPR News.
(SOUNDBITE OF ...OF SINKING SHIPS' "IT'S EASIER WITH NO DESTINATION") Transcript provided by NPR, Copyright NPR.