Hourly remote sensing monitoring of harmful algal blooms (HABs) in Taihu Lake based on GOCI images

Abstract

The increasingly serious harmful algal blooms (HABs) in Taihu Lake has brought huge losses to the local economy and people’s life in Taihu Lake. Satellite remote sensing technology has become one of the most important monitoring methods for HAB disasters due to its large-scale and long-term advantages. GOCI image has become the new data source of HAB monitoring because of its large size and high time resolution. Due to the low spatial resolution (500 m) and the existence of mixed pixels, the error of HAB area obtained by the NDVI method is large. In this paper, the linear mixing model (LMM) and the normalized difference vegetation index (NDVI) threshold method are combined to extract the HAB area from GOCI images with 500-m spatial resolution. Compared with the results of the HAB area extracted by Landsat8 OLI and MODIS data, three small areas in the study area were selected to verify the accuracy of the HAB area extracted from the GOCI image on October 2, 2015. The results show that when the NDVI threshold is 0.1, the area error of HABs is the smallest when the extracted HAB pixels mask the decomposition results of mixed pixels; besides, the area error of HABs extracted from the GOCI image is smaller than that from MODIS image; finally, GOCI image can extract the spatial dynamic distribution of HABs in Taihu Lake within 8 h a day, which has higher temporal resolution than the MODIS image. Compared with the NDVI threshold method and LMM method, the inversion accuracy is greatly improved, and the accuracy is stable in different regions. It can provide technical support for the decision-making and assessment of HAB ecological disasters.