Author(s): Yaoyao Ma; Lu Chang; Joseph H. W. Lee

Linked Author(s): yaoyao Ma

Keywords: Harmful algal blooms; Imaging FlowCytobot; Machine learning; Algae species identification; Real-time water quality data; Dissolved Oxygen; Stratification; Bloom risk prediction

Abstract: Algal blooms in coastal waters can cause harmful effects to marine ecosystems and pose threats to fisheries management. The formation of harmful algal blooms (HAB) involves the interaction of complex biological and physical processes; on the other hand, field observations of algal blooms are typically sparse and cannot capture the characteristic spatial and temporal variations of blooms. Continuous real time data on algal cell counts and water quality are necessary for the development of HAB early warning systems. We report recent findings of an integrated ecological monitoring system that has been successfully deployed in a subtropical coastal bay over a 3-year period. The system incorporates: (i) a submerged in-situ Imaging FlowCytobot (IFCB) that is able to detect and classify up to 25 algal species with the use of a machine learning-based automatic classifier; (ii) a multi-parameter water quality monitor at two depths; and (iii) algal bloom risk prediction system based on antecedent environmental conditions. The algal cell count and relative species dominance (diatoms vs dinoflagellates) have been continuously monitored in a marine fish culture zone for the first time. The predicted algal bloom risks are found to correlate well with observed blooms; the changes in hydro-meteorological conditions, water temperature and salinity, water column stability, cell counts, chlorophyll, and dissolved oxygen (DO) throughout an algal bloom cycle are illustrated through field observations.

DOI: https://doi.org/10.3850/IAHR-39WC2521711920222037

Year: 2022