Abstract
River plumes discharging into continental shelf waters have the potential to influence the distributions, predator-prey relationships, and thus survival of nearshore marine fish larvae, but few studies have been able to characterize the plume environment at sufficiently fine scales to resolve the underlying mechanisms. We used a high-resolution plankton imaging system and a sparse convolutional neural network to automate image classification of larval fishes, their planktonic prey (calanoid copepods), and gelatinous planktonic predators (ctenophores, hydromedusae, and siphonophores) over broad spatial scales (km) and multiple pulses of estuarine water exiting Mobile Bay (Alabama, USA) into the northern Gulf of Mexico from 9-11 April 2016. Fine-scale (1 m) plankton distributions were examined to analyze predator-prey relationships across 3 distinct plume regimes that varied by degree of wind-forcing and mixing rates. In calm wind conditions, the water column was highly stratified, and fish larvae and zooplankton were observed aggregating in a region of river plume-derived hydrodynamic convergence. As winds strengthened, the water column was subjected to downwelling and highly turbulent conditions, and there was decreasing spatial overlap between larval fishes and their zooplankton prey and predators. Our results indicate that high-discharge plume regimes characterized by strong wind-forcing and turbulence can rapidly shift the physical and trophic environments from favorable to unfavorable for fish larvae. Multiple pathways for both nearshore retention and advective dispersal of fish larvae were also identified. Documenting this variability is a first step toward understanding how high discharge events and physical forcing can affect fisheries production in river-dominated coastal ecosystems worldwide.