Distinguishing surface cyanobacterial blooms and aquatic macrophytes using Landsat/TM and ETM + shortwave infrared bands

Abstract

Satellite remote sensing can be considered a suitable approach to monitor the extent of cyanobacterial blooms compared with conventional ship surveys because of the patchiness and high spatial and temporal variability of the blooms. However, most of the existing algorithms are not capable of distinguishing cyanobacterial blooms and aquatic macrophytes due to their similar spectral characteristics in the red and near infrared (NIR) wavelengths. In this study, we conducted in situ spectral measurements and satellite data analyses for cyanobacterial blooms and aquatic macrophytes to find an effective method to distinguish them using medium-resolution Landsat satellite images. The reflectance spectra were measured for lake waters and cyanobacterial blooms with 13 different chlorophyll-a concentration levels (from 54 to 21,736μgL−1) and four types of aquatic macrophytes (two emerged and two floating-leaved macrophytes) in the wavelength range from 350 to 2500nm. In addition, seven Landsat images were collected for nine lakes in Japan or Indonesia. We calculated several selected indices, i.e., the normalized difference vegetation index (NDVI), five types of normalized difference water index (NDWI), and the floating algal index (FAI) to find an appropriate index for distinguishing cyanobacterial blooms and aquatic macrophytes.The results showed that the spectral characteristics of cyanobacterial blooms were significantly different from those of aquatic macrophytes in the short-wave infrared (SWIR) region, indicating that the SWIR bands are important for distinguishing cyanobacterial blooms and aquatic macrophytes. The results also showed that the combination of FAI and NDWI4,5 was an effective method for classifying lake areas. We first used the FAI for extracting lake waters, and we then used the NDWI4,5 to classify the remaining areas as cyanobacterial blooms or aquatic macrophytes. The results also showed that the threshold of NDWI4,5 was less sensitive to the effects of both the atmosphere and mixed pixels compared to the other indices. Our application of the FAI threshold of 0.05 and the NDWI4,5 threshold of 0.63 to six lakes in Japan and Indonesia showed that the proposed method could successfully distinguish lake water, cyanobacterial blooms, and aquatic macrophytes.