Search

Phaeocystis globose (P. glo) are the most frequent harmful algae responsible for red tides in Qinzhou Bay, Guangxi. They pose a significant threat to the coastal marine ecosystem, making it essential to develop an efficient indicator method tailored to P. glo outbreaks. In remote sensing water quality monitoring, there is a strong correlation between P. glo and cyanobacteria, with phycocyanin (PC) serving as an indicator of cyanobacterial biomass. Consequently, existing research has predominantly focused on remote sensing monitoring of medium to high PC concentrations. However, it is still challenging to monitor low PC concentrations. This paper introduced the BP neural network (BPNN) and particle swarm optimization algorithm (PSO). It selects spectral bands and indices sensitive to PC concentrations and constructs a PC concentration retrieval model, in combination with meteorological factors, offering a comprehensive exploration of the indicative role of low PC concentrations in predicting P. glo red tide outbreaks in Qinzhou Bay. The results demonstrated that the PC concentration retrieval model, based on the backpropagation neural network optimized by the particle swarm optimization algorithm (PSO-BPNN), demonstrated better performance (MAE = 0.469, RMSE = 0.615). In Qinzhou Bay, PC concentrations were mainly concentrated around 2~5 μg/L. During the P. glo red tide event, the area with undetectable PC concentrations (PC < 0.04 μg/L) increased by 4.97 km2, with regions below 0.9 μg/L experiencing exponential growth. Considering the variations in PC concentrations along with meteorological factors, we proposed a straightforward early warning threshold for P. glo red tides: PC < 0.9 μg/L and T < 20 °C. This method, from a remote sensing perspective, analyzes the process of P. glo outbreaks, simplifies PC concentration monitoring, and provides a reasonably accurate prediction of the risk of P. glo red tide disasters. [ABSTRACT FROM AUTHOR]