Hacker News Reader

Comment by vehementi | Hacker News Reader

> Five hundred data points collected from 13 vessels sounds impressive, but David Johns, head of the Continuous Plankton Recorder Survey, describes it as "a literal drop in the ocean." Johns would know—the Continuous Plankton Recorder Survey has been running since 1958 and has accumulated more than 265,000 samples.

That uh isn't how stats works. 500 data points is a significant amount if they were being taken in a proper (randomized) manner. Maybe Johns is being taken out of context (i.e. his point might be that the sampling methodology was poor) but this too sounds like BS.

View on HN

The problem is that the ocean is not amenable to random sampling in the same way that, say, Twitter accounts are; phytoplankton are going to be differentially-distributed depending on where in the euphotic zone you're sampling, which latitudes, distance to land, time of year, and so on, which is why longitudinal biodiversity studies are carried out in specific areas. It's vanishingly unlikely that 500 water samples of unknown provenance (and I'm curious why they weren't sampling using the standard method of towed plankton nets) can demonstrate a global extinction event of all plankton, though it could indicate a loss of biodiversity in certain areas. God knows we're working hard to empty out our oceans, but I'm not sure that this paper demonstrates what they claim.

> and I'm curious why they weren't sampling using the standard method of towed plankton nets

Because the GOES Foundation use volunteer yachts to collect the data:

https://www.goesfoundation.com/citizen-science-project/

>Twice a day, (if possible) we want you to take a 0.5 litre of sea water, put it through a GOES filter (developed by Dr.Jesus Ramon Barriuso Diez), count plankton, microplastics (fibres and beads) and any other particles which are over 20 microns.

I believe they went with this method because it's easier for non-scientists to contribute to their study.

They mention nets, but you're right in being suspicious in the method of collection. It does not seem like it uses best practices.

Not a marine biologist, but this seems like it would explain their entire findings. Yachts tend to hang out more around the tropics and less around murky waters in, say, the North Atlantic. Clear water, such as you'd find in the Bahamas, is clear precisely because of lower plankton populations[1]. By collecting data off of pleasure boats, they're corrupting the data right off the bat!

1: https://www.businessinsider.com/why-some-beaches-have-clear-...

Great points, and additionally pleasure yachts stay relatively near coasts most of the time and don't navigate the deep water.

> I believe they went with this method because it's easier for non-scientists to contribute to their study.

Which should probably cast their study in an even more negative light. Even ignoring the fact that the samples were collected contrary to a standard method, we can't be sure the samples collected by random people were actually all done in a similar way as each other.

Climate change could not only affect the viability of plankton, but also shift ocean and air currents. Sampling at the same GPS coordinates could give very different results from one year to the next.

If the plankton counts suddenly shot up 2 orders of magnitude in a location, I expect that would cause a lot of questions to be asked that might result in more accurate results. But a decline or increase of 25% might just be assumed to be population fluctuations instead of a shift.

Here is another guy with a domain PhD, who had this to say

>Also "13 vessels and more than 500 data points" for a finding this sweeping in its assertions is enough to make any microbial oceanographer fall off their lab bench laughing.

https://twitter.com/wang_seaver/status/1548751068640686080

The ocean is 41 million square miles. That's 82,000 square miles per sample, from only 13 ships, or over 3 million square miles per ship.

The accuracy of their model would have to be incredible, assuming those boats weren't jet powered racing boats.

It would be interesting to see the coordinates of each sample. I assume we'll see something unimpressive since, as the article points out, the data containing many times the samples does not agree, suggesting this did not use the appropriate statistical methods/sampling.

Yeah, I found it really fishy that they didn’t counter with their data. They weren’t like yeah, it’s only 40%. They just said we’ve collected more data, so they are wrong.

Reply on news.ycombinator.com