Search for: All records

Abstract We assess whether a supervised machine learning algorithm, specifically a convolutional neural network (CNN), achieves higher accuracy on planktonic image classification when including non‐plankton and ancillary plankton during the training procedure. We focus on the case of optimizing the CNN for a single planktonic image source, while considering ancillary images to be plankton images from other instruments. We conducted two sets of experiments with three different types of plankton images (from aZooglider, Underwater Vision Profiler 5, and Zooscan), and our results held across all three image types. First, we considered whether single‐stage transfer learning using non‐plankton images was beneficial. For this assessment, we used ImageNet images and the 2015 ImageNet contest‐winning model, ResNet‐152. We found increased accuracy using a ResNet‐152 model pretrained on ImageNet, provided the entire network was retrained rather than retraining only the fully connected layers. Next, we combined all three plankton image types into a single dataset with 3.3 million images (despite their differences in contrast, resolution, and pixel pitch) and conducted a multistage transfer learning assessment. We executed a transfer learning stage from ImageNet to the merged ancillary plankton dataset, then a second transfer learning stage from that merged plankton model to a single instrument dataset. We found that multistage transfer learning resulted in additional accuracy gains. These results should have generality for other image classification tasks. 

Abstract The effects of environmental change on zooplankton communities, and more broadly, pelagic ecosystems are difficult to predict due to the high diversity of ecological strategies and complex interspecific interactions within the zooplankton. Trait‐based approaches can define zooplankton functional groups with distinct responses to environmental change. Analyses across multiple mesozooplankton groups can help identify key organizing traits. Here, we use the pronounced cross‐shore environmental gradient within the California Current Ecosystem in a space‐for‐time substitution to test potential effects of ocean warming and increased stratification on zooplankton communities. Along a horizontal gradient in sea‐surface temperature, water column stratification, and light attenuation, we test whether there are changes in zooplankton species composition, trait composition, and vertical habitat use. We employ DNA metabarcoding at two loci (18S‐V4 and COI) and digital ZooScan imaging of zooplankton sampled in a Lagrangian manner. We find that vertical distributions of many mesozooplankton taxa shift to deeper depths in the cross‐shore direction, and light attenuation is the strongest predictor of magnitude of change. Vertical habitat shifts vary among functional groups, with changes in vertical distribution most pronounced among carnivorous taxa. Herbivorous taxa remain associated with the chlorophyll maximum, especially in clear offshore waters. Our results suggest that increased stratification of this ocean region will lead to deeper depths occupied by some components of epipelagic mesozooplankton communities, and may result in zooplankton communities with more specialized feeding strategies, increased egg brooding, and more asexual reproduction. 

Abstract The Rhizaria is a super‐group of amoeboid protists with ubiquitous distributions, from the euphotic zone to the twilight zone and beyond. While rhizarians have been recently described as important contributors to both biogenic silica and carbon fluxes, we lack the most basic information about their ecological habitats and preferences. Here, using in situ imaging (Underwater Vision Profiler 5), we characterize the vertical ecological niches of different test‐bearing pelagic rhizarian taxa in the southernCalifornia Current Ecosystem. We define three vertical layers between 0 and 500 m occupied, respectively, by (1) surface dwelling and mostly symbiont‐bearing rhizarians (Acantharia and Collodaria), (2) flux‐feeding phaeodarians in the lower epipelagic (100–200 m), and (3) Foraminifera and Phaeodaria populations adjacent to the oxygen minimum zone. We then use Generalized Additive Models to analyze the response of each rhizarian category to a suite of environmental variables. The models explain between 9% and 93% of the total variance observed for the different groups. While temperature and the depth of the deep chlorophyll maximum appear as the main abiotic factors influencing populations in the upper 200 m, dissolved silicon concentration is related to the abundance of mesopelagic phaeodarians, though it explains only a portion of the variance. The importance of biotic interactions (e.g., prey availability, predation, parasitism, symbiosis) is still to be considered, in order to fully incorporate the dynamics of test‐bearing pelagic rhizarians in ecological and biogeochemical models. 

Abstract Plankton imaging systems supported by automated classification and analysis have improved ecologists' ability to observe aquatic ecosystems. Today, we are on the cusp of reliably tracking plankton populations with a suite of lab‐based and in situ tools, collecting imaging data at unprecedentedly fine spatial and temporal scales. But these data have potential well beyond examining the abundances of different taxa; the individual images themselves contain a wealth of information on functional traits. Here, we outline traits that could be measured from image data, suggest machine learning and computer vision approaches to extract functional trait information from the images, and discuss promising avenues for novel studies. The approaches we discuss are data agnostic and are broadly applicable to imagery of other aquatic or terrestrial organisms.