A new paper led by Sebastian Sippel just appeared in Nature arguing that ocean temperature measurements in the early 20th century have a cold bias. It's a fun story illustrating the process of scientific discovery, so let me talk about it a bit. 🧵 https://t.co/hAzvDYoDyZ
Before I begin, I should note that I am one of the 11 coauthors on this new paper. My contributions are actually fairly modest, and Sebastian and others deserve the lion's share of the credit for developing this work over more than 2 years. 2/
This story of scientific discovery begins, as so many do, by noticing a small discrepancy in the data. Through most of the last 170 years, the land measurements and ocean measurements show a similar pattern of global warming, but not in the early 20th century. 3/
The land warms about twice as fast as the ocean, but once you rescale for that basic difference, the patterns have usually been quite similar. But not during the early 20th century, when the ocean measurements were much cooler than land observations would expect. 4/
Now, this isn't actually a new observation. This inconsistency had been noticed before, but had generally been waved off by people mumbling, "something, internal variability, something something...". 5/
But continuing the journey of exploration, let's think a bit more about what this difference might mean. To begin, let's ask: Do climate models show similar behavior? Short answer, no they do not. 6/
In climate models (reddish dots) the land warming and the ocean warming are always tightly coupled on multidecadal scales. The models never show the kind of divergence between land warming and ocean warming that the measurements have been suggesting. 7/
One can also look at the model's general behavior in the early 20th century. While the global average of observations (gray curves) dip in the early 20th century before rebounding, the average climate model has a smooth rise (red curves). https://t.co/P27UWNPuaS 8/
Even seeing these differences, faith in the measurements tended to remain strong. So much so that failing to reproduce this "unforced natural variability" in the land/ocean system was sometimes used as a major criticism of the climate models. 9/
Since models will never fully be trusted, scientists needed to be another way to attack this problem. Sippel and colleagues turned to their attention to proxy measurements of temperature. 10/
Proxies, for the unfamiliar, are physical systems that respond to temperature changes in their environment and leave a record behind of those environmental changes that can be discerned later. 11/
The most common proxies referenced by this study are tree ring growth patterns on land and coral isotopic and chemical composition in the ocean. These are less precise than a thermometer but both record changes based on the temperatures present during their growth period. 12/
Long story made short, the proxy data from the early 20th century broadly align with the land-based thermometer data, and don't show the cold excursions implied by in the oceanic measurements. 13/
So, if you are following this journey, you might now suspect that the ocean measurements in the early 20th could be faulty, with some sort of cold bias. But bare suspicion isn't very satisfying, so lets think about this more. 14/
Amateurs always talk about urban heat islands and other weather station biases, but the reality is that the biases that affect land measurements are tame compared to what happens in the oceans. 15/
Outside specialists, most people do not realize that ocean uncertainties are the main limitation on global average temperature measurement accuracy for most of the last 170 years. 16/
Weather stations obligingly stay in one place and churn out long records of measurement. So, when a weather station goes wrong, it is often easy to discover the problem, for example, by comparing against neighboring stations. 17/
Also, even if a weather station error is overlooked, it will generally only affect estimates of temperature in its immediate vicinity. Ocean-going ships and buoys are not nearly so cooperative. 18/
Ships move around almost constantly, and many have depressingly short measurement histories. This limits the ability to do after-the-fact tests for biases. It also means any errors that do occur can get spread through the whole ocean. 19/
To make matters worse, the way that ocean measurements have been made has changed... a lot. And every change introduces a chance for biases that must be corrected to ensure that different types of measurement are comparable. 20/
And we know that changes in ocean measuring technology and methods made a difference. Different nations often deployed different techniques. When comparing ocean measurements collected at the same time by different nations (colors), significant differences exist. 21/
Prior to ~1930, essentially all ocean temperature measurement were made by throwing a bucket in the water, hauling it up on deck, and sticking a thermometer in it. 22/
Distressingly though, the "bucket" technology changed over time. And almost no one recorded what type of bucket they were using. 23/
Canvas sacks, a common choice in the early 20th century, had much less insulation than wooden or rubber buckets. As a result, the simple act of hauling the bucket onto the deck might have allowed a bit of cooling. 24/
How much cooling? Maybe a couple tenths of a degree? It's a good guess, but still limited by the fact no one recorded the type of bucket they were using. 25/
Helpfully, a different paper (appearing a few months ago), may help resolve some of these issues. Duo Chan and colleagues, published a new global temperature reconstruction leveraging additional techniques to constrain ocean biases. https://t.co/KlSG9nHJQm 26/
Their work did many things, but one of the things they did was assume that ships near shore and in port should observe similar temperature variations to nearby coastal / island weather stations. Expecting such coastal consistency helped to constrain ship-based biases. 27/
When they did that, Chan and colleagues found that early 20th century cold dip essentially vanished (red curve). 28/
Will this by the final word on early 20th century temperatures and potential ocean biases? Certainly not. But I'd say the preponderance of evidence is now pointing towards a more gradual warming trend in the early 20th century, without that cooling excursion. 29/
Does this matter? For the story of global warming, not so much. The estimates of total warming since preindustrial is little changed by these revisions. But it does highlight the need to still better understand past changes, especially in the oceans. 30/
One area where is does matter is in the discussion of internal variability. The "unforced natural variability" of the early 20th century was a major line of argument for the possibility of large, natural multi-decadal variations internal to Earth's climate. 31/