High-throughput quantification of community evolution in planktonic foraminifera across the K-Pg mass extinction using automated morphometrics and convolutional neural networks
The K/Pg bolide impact had a dramatic effect on oceans, including disrupting the biological pump, a vital component of the global carbon cycle. The short timescale of the perturbations makes the K/Pg a unique analogue for understanding how oceans might respond to rapid anthropogenic climate change. Planktonic foraminifera are single-celled organisms that build a CaCO3 shell (or “test”), which form a uniquely rich fossil record. Test morphology has been shown to correlate with environmental parameters. As such, planktonic foraminifera are an ideal group for studying the response of the biotic realm to environmental perturbation.
We will build a high-resolution record of community morphology of planktonic foraminifera across the K/Pg boundary. Using software that we developed for high-throughput image processing and morphological data extraction, in combination with convolutional neural networks, we will generate data for hundreds of thousands of planktonic foraminifera from 0.5 Myr before the boundary to 2 Myrs after. This time period covers the late Maastrichtian warming, the bolide impact event, and the full duration of the recovery of the biological pump in the aftermath of global extinction. We will compare this new morphological record with existing geochemical records to see how patterns of morphological evolution correlate with environmental perturbation and recovery. This will allow us to develop planktonic foraminiferal morphology as a proxy for ocean ecosystem health.