Clusters in profiles of nutrients and their correlation with phytoplankton communities

[gmanuch]

Title
Clusters in profiles of nutrients and their correlation with phytoplankton communities

Summary
Hoping to identify clusters in the distribution of nutrients across the ocean and use them to identify which phytoplankton communities might be prevalent in those clusters. We will attempt to use machine learning to crack this problem: this would be a combination of an unsupervised clustering problem, with a follow-up supervised regression.

Personnel
Georgy Manucharyan, University of Washington, School of Oceanography (I'm happy to lead the machine learning component of the study). Need a biologist and/or chemical oceanographer as an additional lead!

[Efeogis]

I would love to be part of this project

[Tartomas]

Hi @gmanuch which is the data set background to perform this project? Did you know the Phyto functional Types from CMEMS?
Best

[gmanuch]

All ideas for datasets are welcome! I'm not a biologist but we've been looking into various datasets on CMAP https://simonscmap.com/. We could start with HOT or BATS time series datasets, which I believe come with some phytoplankton data. Have a look at the tutorial we had for our BIO day during the undergraduate program here: https://github.com/gmanuch/DataScienceOceanography/blob/main/Tutorials/BO/BO_Tutorial.ipynb

[7yl4r]

Where is the phyto occurrence data coming from? If you want to get occurrences from OBIS, we have been doing a lot of work on iobis/pyOBIS to help make accessing, analyzing, and visualizing species occurrence easier.

[gmanuch]

See my reply to a similar question above: all ideas for datasets are welcome! So far, I've been looking into CMAP, but it would be great to see other ideas. Looking for nutrient profiles co-located with phytoplankton characteristics.

[ANetTow]

C-MAP has digested a large number of data sets. There is SeaFlow flow cytometry data as well as IFCB data. You can co-localize for nutrients, temperature, etc.:

https://simonscmap.com/