Video classification of hypoxic habitats and benthic communities in two productive freshwater embayments

Abstract

Ongoing anthropogenic eutrophication and warming temperatures are expected to increase the extent and severity of hypoxia globally. Monitoring hypoxia has traditionally relied on costly surveys or sensor networks. While benthic macroinvertebrates are valuable indicators of hypoxia, community analysis is limited by small spatial scales of traditional grab sampling and labor-intensive processing. To address this, we combined benthic grab samples with underwater video to detect hypoxic habitats in two productive embayments of the Laurentian Great Lakes: Saginaw Bay, Lake Huron with periodic short-term hypoxia, and Hamilton Harbour, Lake Ontario with prolonged hypoxia. Using supervised classification, we identified significant grouping of in situ video data with cluster analysis, and then aligned video groups with environmental and biological datasets. These video groups were supported by cluster analysis of measured environmental variables, with clusters differing in duration of low near-bottom dissolved oxygen concentration and by depth. Independent cluster analysis confirmed significant separation of benthic communities among the selected video groups, with hypoxic habitats showing reduced species diversity and a higher proportion of tubificids. The gradient of conditions sampled in our study revealed assemblages of benthic invertebrates sensitive to and tolerant of hypoxia. The agreement among video, biological, and environmental data confirmed that video analysis can provide a novel, quick and reliable method to detect benthic habitats affected by hypoxia and determine their spatial extent once a baseline is established.

Keywords

• Hypoxia
• Underwater video
• Benthos
• Monitoring
• Indicator species