How many specimens make a sufficient training set for automated three-dimensional feature extraction?

Deep learning has emerged as a robust tool for automating feature extraction from three-dimensional images, offering an efficient alternative to labour-intensive and potentially biased manual image segmentation methods. However, there has been limited exploration into the optimal training set sizes, including assessing whether artficial expansion by data augmentation can achieve consistent results in less time and how consistent these benefits are across different types of traits. In this study, we manually segmented 50 planktonic foraminifera specimens from the genus Menardella to determine the minimum number of training images required to produce accurate volumetric and shape data from internal and external structures. The results reveal unsurprisingly that deep learning models improve with a larger number of training images with eight specimens being required to achieve 95% accuracy. Furthermore, data augmentation can enhance network accuracy by up to 8.0%. Notably, predicting both volumetric and shape measurements for the internal structure poses a greater challenge compared with the external structure, owing to low contrast differences between different materials and increased geometric complexity. These results provide novel insight into optimal training set sizes for precise image segmentation of diverse traits and highlight the potential of data augmentation for enhancing multivariate feature extraction from three-dimensional images.

Five million years of Antarctic Circumpolar Current strength variability.

The Antarctic Circumpolar Current (ACC) represents the world's largest ocean-current system and affects global ocean circulation, climate and Antarctic ice-sheet stability1-3. Today, ACC dynamics are controlled by atmospheric forcing, oceanic density gradients and eddy activity4. Whereas palaeoceanographic reconstructions exhibit regional heterogeneity in ACC position and strength over Pleistocene glacial-interglacial cycles5-8, the long-term evolution of the ACC is poorly known. Here we document changes in ACC strength from sediment cores in the Pacific Southern Ocean. We find no linear long-term trend in ACC flow since 5.3 million years ago (Ma), in contrast to global cooling9 and increasing global ice volume10. Instead, we observe a reversal on a million-year timescale, from increasing ACC strength during Pliocene global cooling to a subsequent decrease with further Early Pleistocene cooling. This shift in the ACC regime coincided with a Southern Ocean reconfiguration that altered the sensitivity of the ACC to atmospheric and oceanic forcings11-13. We find ACC strength changes to be closely linked to 400,000-year eccentricity cycles, probably originating from modulation of precessional changes in the South Pacific jet stream linked to tropical Pacific temperature variability14. A persistent link between weaker ACC flow, equatorward-shifted opal deposition and reduced atmospheric CO2 during glacial periods first emerged during the Mid-Pleistocene Transition (MPT). The strongest ACC flow occurred during warmer-than-present intervals of the Plio-Pleistocene, providing evidence of potentially increasing ACC flow with future climate warming.

Temperature is a poor proxy for synergistic climate forcing of plankton evolution.

Changes in biodiversity at all levels from molecules to ecosystems are often linked to climate change, which is widely represented univariately by temperature. A global environmental driving mechanism of biodiversity dynamics is thus implied by the strong correlation between temperature proxies and diversity patterns in a wide variety of fauna and flora. Yet climate consists of many interacting variables. Species probably respond to the entire climate system as opposed to its individual facets. Here, we examine ecological and morphological traits of 12 633 individuals of two species of planktonic foraminifera with similar ecologies but contrasting evolutionary outcomes. Our results show that morphological and ecological changes are correlated to the interactions between multiple environmental factors. Models including interactions between climate variables explain at least twice as much variation in size, shape and abundance changes as models assuming that climate parameters operate independently. No dominant climatic driver can be identified: temperature alone explains remarkably little variation through our highly resolved temporal sequences, implying that a multivariate approach is required to understand evolutionary response to abiotic forcing. Our results caution against the use of a 'silver bullet' environmental parameter to represent global climate while studying evolutionary responses to abiotic change, and show that more comprehensive reconstruction of palaeobiological dynamics requires multiple biotic and abiotic dimensions.

The Breakdown of Static and Evolutionary Allometries during Climatic Upheaval.

The influence of within-species variation and covariation on evolutionary patterns is well established for generational and macroevolutionary processes, most prominently through genetic lines of least resistance. However, it is not known whether intraspecific phenotypic variation also directs microevolutionary trajectories into the long term when a species is subject to varying environmental conditions. Here we present a continuous, high-resolution bivariate record of size and shape changes among 12,633 individual planktonic foraminifera of a surviving and an extinct-going species over 500,000 years. Our study interval spans the late Pliocene to earliest Pleistocene intensification of northern hemisphere glaciation, an interval of profound climate upheaval that can be divided into three phases of increasing glacial intensity. Within each of these three Plio-Pleistocene climate phases, the within-population allometries predict evolutionary change from one time step to the next and that the within-phase among-population (i.e., evolutionary) allometries match their corresponding static (within-population) allometries. However, the evolutionary allometry across the three climate phases deviates significantly from the static and phase-specific evolutionary allometries in the extinct-going species. Although intraspecific variation leaves a clear signature on mean evolutionary change from one time step to the next, our study suggests that the link between intraspecific variation and longer-term micro- and macroevolutionary phenomena is prone to environmental perturbation that can overcome constraints induced by within-species trait covariation.