Quantification of Margalefidinium polykrikoides blooms in the South Sea of Korea using hyperspectral imagery

Abstract

Since 1995, the blooms caused by Margalefidium polykrikoides have been frequently occurred in the South Sea of Korea, where fish farms are concentrated. The fish in aquaculture cages are known to die from gill-clogging by M. polykrikoides at high abundance. Thus, it is important to understand the spatio-temporal distribution of these blooms through quantification of the red tide cell abundance in order to minimize damage caused by the red tide blooms. Hyperspectral imagery can provide useful data for precise detection of the red tide because it has substantial information on optical features of the red tide species. However, the existing red tide algorithms are not suitable for application to hyperspectral data because they are mainly developed with ocean colour sensor. The purpose of this study is to develop the red tide detection algorithm suitable for hyperspectral imagery. We compared the red tide spectra between the in situ survey and hyperspectral image with the various levels of cell abundance to verify the similarity of data. The R2 value between two data showed a good correlation with 0.68-0.98. In order to investigate the optimal band combination for the red tide detection, the correlations between cell abundance of in situ survey and each band of hyperspectral imagery were analyzed. The bands of 708.39 and 596.69 nm showed the highest correlation with cell abundance, and the two-band ratios were applied to