Predicting dreissenid mussel abundance in nearshore waters using underwater imagery and deep learning

Abstract

Accurate and cost‐effective dreissenid mussel abundance maps are vital to assess their ecological roles in aquatic systems. A deep neural network (DNN) modeling framework using semantic segmentation was developed to automatically assess the abundance distribution of two invasive mussel species: zebra and quagga. DNN models were trained on images captured in Lake Erie and Lake Ontario using an underwater color imaging technique. The accuracy of the method was assessed relative to manual laboratory counts of harvested mussels, their dry biomass, and percentage live coverage estimated from fixed‐size quadrats. Assessments performed on a test set collected from 2016 to 2018 show that DNN‐based mussel coverage predictions explain 79% of the variance in log biomass, and 71% for log abundance (N = 125). For reference, live coverage estimated by scuba divers was transformed and found to be a better predictor of biomass (93%) and abundance (91%) (N = 725), leaving room for improvement of our automated method. When identical images were presented to eight human analysts and the DNN, the agreement in live mussel coverage prediction was 85% (N = 189). Models generalize well to diverse underwater illuminations, camera orientations, and resolutions, but are adversely impacted by occluding vegetation and suspended sediment. DNN models are an efficient and accurate solution for mapping mussel abundances at a scale that was previously impossible. The method may be integrated with other studies to assess the mussels' impacts in a variety of aquatic ecosystems. Source code: https://github.com/AngusG/deep-learning-dreissenid and data http://10.5683/SP3/MZEBOJ for reproducing our method are publicly available.

Open water dreissenid mussel control projects: lessons learned from a retrospective analysis

“…This observation, as compared to rapid response eradication projects, suggests that more effort needs to be spent de ning the spatial extent of an infestation area so project managers can better determine the necessary treatment area. Furthermore, the project managers should account for the uncertainty of dreissenid mussel detection efforts due to the inherent limitations 49 of locating and isolating all dreissenid mussel populations in lakes 49 . Limiting the treatment to an area immediately around known mussel locations may fail to capture an entire population, especially planktonic veligers, resulting in a need for additional treatments or a failed eradication attempt.…”

Open water dreissenid mussel control projects: lessons learned from a retrospective analysis

“…This observation, as compared to rapid response eradication projects, highlights the benefits of adequately defining the spatial extent of an infestation area so project managers can better determine the necessary treatment area. Furthermore, accounting for the uncertainty of dreissenid mussel detection efforts is warranted due to the inherent limitations 47 of locating and isolating all dreissenid mussel populations in lakes 55 . Limiting the treatment to an area immediately around known mussel locations may fail to capture an entire population, especially planktonic veligers, resulting in a need for additional treatments or a failed eradication attempt.…”

Classification of Lakebed Geologic Substrate in Autonomously Collected Benthic Imagery Using Machine Learning

“…The complete CMECS scheme includes a biotic component, which previous research has integrated into another modified CMECS scheme [5]. Incorporating the biotic component should be explored further, as it can directly aid in identifying and mapping important habitat modifiers like zebra and quagga mussels and/or submerged aquatic vegetation in the Laurentian Great Lakes [24,54]. Explicitly including both abiotic and biotic components of the benthic habitat mapping process could aid researchers and managers in better understanding and managing aquatic ecosystems.…”