Last week the Museum internally launched its new strategy to 2031. It called us to make our data, insight, knowledge and expertise openly available. Strategic priorities included transforming the study of natural history to benefit people and planet as well as training future generations of scientists.
I immediately felt compelled to write about a paper we jointly published on-line this week in the journal Paleoceanography and Paleoclimatology that features our collections. We provided data for a project run by Yale University to create a portal of calcareous plankton called Foraminifera. It can be used to train the scientists of the future and test a machine learning classifier that could generate large datasets vital for research on our oceans.
Planktonic Foraminifera
Planktonic Foraminifera are single celled microscopic organisms that form a calcareous shell. They are major components of the oceanic microplankton community and are generally less than a millimetre in length.
As they grow they use calcium carbonate in the ocean to form their shells so analysing their chemistry can tell us a lot about the composition of the ocean in which they lived and hence the climate. Each species has a surface water temperature tolerance so studying variations in assemblage composition also allows us to make inferences about past climatic changes.
Planktonic foraminifera are incredibly abundant in the oceans of today and the past. When they die, their shells sink to the bottom of the ocean and get incorporated into sediments, sometimes constituting almost 100% of the sediment. Studying foraminifera in sediments cored from the bottom of the ocean can tell us an almost complete history of changes in the surface of the oceans over millions of years.
The contribution of our collections
We have extensive collections of modern planktonic foraminifera from oceans and seas across the globe. Many of these were derived from our Ocean Bottom Deposit Collection and compiled by the former curator of this collection, Henry Buckley (1939-2002).
For her PhD, former NHM postgraduate student Marina Rillo was studying Henry Buckley’s collection and provided images of all of the specimens in the collection as a contribution to a project called Endless Forams being run at Yale University.
The Museum collections were able to provide details of some species not present and from geographical areas not covered in the Yale dataset.
Quality data
Delivering up to date identifications of our collections is an issue as many of our legacy records reflect identifications that are taxonomically out of date and some are in need of refining by experts. Disagreements between identifiers can also be problematic even when there are relatively few species of modern planktonic Foraminifera and taxonomic schemes are relatively well established.
A recent repeatability study paper based on a set of 100 modern planktonic foraminiferal specimens that are now housed in our collections, aimed to assess the importance of various reasons for these disagreements. A range of foraminiferal researchers from students to expert were asked to identify each of the specimens. It showed that in general, identifications are reproducible but that it can be difficult to discern closely related species.
To help to compile a dataset of well categorised specimens, the Endless Forams project set up an on-line portal using the Zooniverse platform to enable 34,640 images of modern planktonic foraminifera to be vetted by a panel of experts that included Marina Rillo and former Museum postdoctoral worker Lyndsey Fox.
Marina Rillo has also been assessing Henry Buckley’s identifications and workflows by returning to the original samples, reprocessing them and re-identifying the slides of tiny fossils he made from the collection. In general he did a good job identifying planktonic Foraminifera although there were some biases towards choosing larger specimens and taxonomic concepts have changed slightly since Buckley’s time.
Training scientists for the future
Prior to the Endless Forams project there has been a lack of on-line resources to help identify foraminifera with the Mikrotax site a notable exception. The Mikrotax site includes images of some of our type specimens and covers fossil as well as modern species. In addition, the Museum has recently hosted NERC funded training courses for students at the beginning of their studies on planktonic foraminifera.
The images of specimens that were agreed by the Endless Forams Project experts have been delivered to an on-line portal. There is also a set of these images on the citizen science platform Zooniverse that can be used as a training tool for learning the identification of modern planktonic Foraminifera.
Machine learning
As part of the Endless Forams project a supervised machine learning classifier was trained with ~27,000 images of these identified planktonic foraminifera. This was able to provide the correct species name 87.4% of the time which compares well to the results from the repeatability study published last year.
This new technique potentially allows high quality data covering a large area of our oceans to be generated very quickly. Details of planktonic foraminiferal assemblages from the ocean bottom floor and from ocean bottom sediment cores can provide evidence for both anthropogenic and deep time changes in surface temperatures across the globe.
Finally it is worth noting that Henry Buckley intended to create an atlas of living planktonic foraminifera using his collection. He was discouraged by the Museum from carrying out this research so he would perhaps be surprised to hear that his collection and the data it contains very much fits with current Museum strategy. The NHM contribution to the Endless Forams project is in part dedicated to Henry Buckley.