Automatic Detection of Algal Blooms Using Sentinel-2 MSI and Landsat OLI Images

Abstract

Algal bloom is a serious global issue for inland waters, posing poses a serious threat to aquatic ecosystems. The timely and accurate detection of algal blooms is critical for their control, management and forecasting. Optical satellite imagery with short revisit times has been widely used to monitor algal blooms in marine and large inland waters. However, such images typically are of coarse resolution, limiting their utility to map algal blooms in small inland waters. We developed a new method to map the spatial extent of algal blooms using sentinel-2 multispectral instrument (MSI) and Landsat operational land imager (OLI) images with higher spatial resolution but lower temporal resolution based on the concept of local indicator of spatial association. The mapping results was applied to measure the duration and frequency of algal blooms in Lake Taihu from 2017 to 2020. Our results show that the developed methodology is able to extract the spatial distribution of moderate algal blooms using near-infrared and red-edge bands (bands 6, 7, 8, and 8a of sentinel-2 MSI images or band 5 of Landsat OLI images) by comparison with MODIS FAI data (R2 = 0.888 for sentinel-2 MSI and R2 = 0.85 for Landsat OLI, P < 0.05). However, the temporal resolution of combined Landsat OLI and sentinel-2 MSI images (i.e., up to 2–3 days) is insufficient to monitor algal blooms during the summer time in Lake Taihu due to cloud effects and rapid algal change. Our research has benefits for the management of small inland waters with complex water conditions.

Keywords

• Algal bloom
• cyanobacteria
• spatial autocorrelation
• sentinel-2 multispectral instrument (MSI)
• Landsat operational land imager (OLI)