Visualising 16 years of phytoplankton community dynamics

My esteemed colleague Jean-Yves Dias recently published an article on phytoplankton communities in French coastal waters. We worked on it together with 2 other co-authors, and it’s the result of more than 2 years of hard work (especially on Jean-Yves’ part) so I’m quite proud to advertise it on the blog!

The paper is (objectively :)) really nice, and you can check it out here: https://doi.org/10.1093/ismeco/ycag174

In it, we analysed data from long-term phytoplankton time series to study how blooms impact the structure of the phytoplankton community. One problem that we faced was: “How do we visualise 16 years of phytoplankton community dynamics?”

The answer we settled for is my favourite figure in the article, and I’m now going to propose a detailed analysis of it, as an introduction for our paper.

(It was also the occasion to learn how to make interactive charts in R with the plotly package, which was a lot of fun)

The data

The data we used comes from the REPHY time series, a monitoring survey of phytoplankton in coastal waters operated by IFREMER¹ since 1987. Every 2 weeks, at designated places along the French coastline, scientists go and sample some phytoplankton, then identify and count the organisms in the sample, and record their observations. As a result, we have a dataset that documents how the phytoplankton community evolves over months, years, and decades!

Because the REPHY operators in IFREMER labs all around France have been following a standard protocol for nearly 40 years now, this dataset is of exceptional value for ecology studies. Of course, the survey has evolved over the years², so it’s not perfectly consistent. For our analysis, we decided to focus on 19 sampling sites in 3 different regions, over a period extending from March 2007 to August 2022.

Phytoplankton cell numbers

Phytoplankton is very diverse, and the REPHY dataset includes a lot of species. To make our visualisation clearer, we will focus on the 12 most abundant and frequent genera³ in the 3 regions. All other taxa will be grouped together in a “Others” category. Of course, the price for this clarification is that it sets aside a lot of the fine-scale variations in the community, because many rare taxa will not be properly represented.

First, we will simply display the number of cells counted (in cells per liter) in each region, for each month in the dataset:

This is an interactive chart so you can explore with the little menu in the top right of the image. What can we see here?

The most obvious pattern is in the Eastern Channel – North Sea region, where Phaeocystis (in yellow) appears in very large numbers every year in spring.

Phaeocystis is a microalga of the haptophyte clade, that can form colonies of many cells embedded in a sphere of mucus. It is well known in northern France because its colonies aggregate in large amounts of foam that clog the nets of fishermen. People in the region call it “Vert de mai”⁴, and indeed we see that the Phaeocystis bloom generally occurs around April-May.

The Phaeocystis bloom is a good example of a very strong seasonal pattern structuring the phytoplankton community. We also notice that there is generally much less cells in winter, especially in regions 2 and 3. But this visualisation masks the more subtle changes, because occasional surges in cell numbers dwarf the rest.

Log10 transformation

Let’s look at the same data, except this time we transform it with a base 10 logarithm:

This new visualisation reveals new features of the community, for example:

• in all 3 regions, the community seems generally structured around a few dominant genera of diatoms (in blues)
• the constant presence of cryptophytes (in pink) in the Atlantic – Western Channel region
• the recurrence of Akashiwo and Azadinium (dinoflagellates) and Chrysochromulina (haptophyte) in recent years in the Mediterranean region

Cryptophytes are not endemic to the Atlantic. Their quasi-absence in the other regions is probably due to the fact that these very small cells are not systematically counted in the survey. People in the IFREMER labs of the Atlantic region have, for some reason⁵, been historically interested in cryptophytes. So patterns in the data can also emerge because of human factors! Another very good example of this is the gap you can observe in March 2020 in regions 2 and 3: the sampling activity was temporarily suspended because of the COVID-19 pandemic.

The problem with the log10 transformation is that it squeezes big numbers so much that it equalises all taxa, when in reality some are much more numerous than others (to see this, look at the Phaeocystis bloom in region 2, compared to the previous chart).

Proportions

We can instead look at the proportion of each taxon in the community (which is the option we chose for the paper). This completely erases the overall variations in cell numbers, but it allows us to see the composition of the community and how it evolves:

We can see that in the Mediterranean, the dominance regularly shifts between the different diatom genera: the community is in fact very dynamic!

Some species regularly bloom on a seasonal basis, this is the case of Phaeocystis in the North Sea as we saw before. But sometimes, very transient blooms dominate the community during exceptional events: see for instance Chrysochromulina in June 2012 (53% of cells counted) or Azadinium in February 2020 (57%), both in the Atlantic region.

Conclusion

Of course, our approach has its flaws. The very broad “Others” category obfuscates a lot of the interesting diversity within the phytoplankton. Because we work with cell numbers and not biovolumes, small organisms will be vastly overrepresented (and this matters when phytoplankton cells can vary in size from 5 to 100+ micrometers). Also, pooling together data from several sampling sites as we did helps to apprehend general patterns but dampens the local variability.

Nonetheless, these rather simple data visualisations highlight interesting features of the phytoplanktonic ecosystem: seasonal variations in cell numbers, recurrent blooms that take over the community (Phaeocystis), and chaotic⁶ dynamics with sudden and transient dominance by rare taxa (Azadinium and Chrysochromulina).

Finally, there is a nice feeling in bringing to life a time series that was born from the dedication of so many people over so many years. Like a mosaic, each sample may not represent much taken separately, but together they compose a decades-long mural of microbial life in the ocean!

As I mentioned, this post is an introduction to and an elaboration on the article Dias et al. (2026) “Diatoms vs dinoflagellates: a temporal network analysis of bloom impacts on phytoplankton diversity and community structure in French coastal waters”

The R scripts I used to produce the interactive charts are available on the github repository associated with the paper: https://github.com/J-YDi/Diatoms-vs-Dinoflagellates

The 3 figures in html format can be found on my github: https://github.com/vpochic/REPHYTO_interactive_charts

If I managed to spark your curiosity about phytoplankton community dynamics, I encourage you to go read the full paper! Good reads on the subject by other authors: Picoche and Barraquand 2020 (also with the REPHY dataset) Deutschmann et al. 2023 (with another time series from the Mediterranean sea).

Thanks to the co-authors of the paper: Jean-Yves Dias (his github page), Samuel Chaffron and Pierre Gernez. Many thanks to the operators and coordinators of the REPHY survey for producing such an invaluable ecological dataset.

This blog post and images it contains are under a Creative Commons Attribution 4.0 license and you can quote, reuse or redistribute it freely as long as the author(s) are properly acknowledged.

1. The French Institute for Sea Research. ↩︎
2. Some sampling sites have been discontinued, others created, some phytoplankton species are better recognised now than they were 20 years ago, etc. ↩︎
3. The genus (plural: genera) is the taxonomic level just above the species. In our visualisation, some genera will be pulled together into even higher taxonomic ranks: families (e.g. Chaetocerotaceae, Cryptophyceae). ↩︎
4. “May green” ↩︎
5. I strongly suspect this interest stems from cryptophytes belonging to the Dinophysis trophic chain… ↩︎
6. Here, “chaotic” has a precise mathematical meaning related to chaos theory, that apparently applies to the dynamics of some ecological communities, including plankton (see Rogers et al. 2022, Mallmin et al. 2024) ↩︎