The relationship between water column stratification, pelagic habitat heterogeneity and plankton diversity in a neritic, river-dominated environment

Christian Briseño-Avena, University of San Diego, Environmental and Ocean Sciences, San Diego, CA, United States, Adam T Greer, The University of Southern Mississippi, Division of Marine Science, Stennis Space Center, MS, United States, Luciano Chiaverano, University of Southern Mississippi, Marine Science, Stennis Space Center, MS, United States, William Monty Graham, University of Southern Mississippi, Marine Science, Stennis Space Center MS, United States and Robert Cowen, Oregon State University, Hatfield Marine Science Center, Newport, OR, United States
Abstract:
Fine-scale responses of plankton to highly complex and productive river-dominated systems are poorly understood, yet these systems are prevalent and vital to coastal economies around the world. In the northern Gulf of Mexico, the Mississippi Bight is one such ecosystem, where river input primarily from the Mobile Bay watershed and winds likely influence the distribution and abundance of plankton across multiple time scales (hourly to seasonal). To address this knowledge gap, a major goal of the CONsortium for COastal, River-Dominated Ecosystems (CONCORDE) was to establish a baseline of fine-scale plankton distributions in the Mississippi Bight and examine community responses to variability in freshwater input and local oceanographic circulation. To resolve the distributions of plankton in relation to oceanography, we towed an in situ imaging system equipped with a suite of environmental sensors (CTD, fluorometer, dissolved oxygen, and PAR), undulating the system from 2 m below the surface to 2-4 m above the seafloor along three parallel, ~60 km-long meridional transects over three seasons (fall 2015, spring and summer 2016). Sampling was designed to capture the riverine-influenced gradation of the shelf (within season) and three distinct river discharge regimes. The imagery collected resulted in >6 billion individual image segments (particles and plankton). Each segment was identified to the lowest taxonomic level possible by implementing an automated image processing and identification pipeline that uses a spatially-sparse Convolutional Neural Net (sCNN). Here we present the results of the sCNN performance trained on a library of 173 taxonomic and other classes (e.g., detritus, bubbles, and noise). The unprecedented spatial and taxonomic resolution attained allowed us to map full-water column “seascapes” which can be used to test the hypothesis that water column stratification produces pelagic habitat heterogeneity, which may lead to increased plankton diversity.