Remote sensing mapping of cyanobacteria blooms in chaohu based on spatio-temporal-spectrum fusion: Improvement on spatial scales

Abstract

Occurrence and relevant risks of cyanobacteria blooms have been increasing continuously in the world wide. Monitoring on spatial and temporal distribution changes of cyanobacteria blooms has been the key of environmental monitoring. In particular, spatio-temporal distribution of cyanobacteria blooms in complicated small inland water areas changes so frequently and a more accurate inversion monitoring method and product was required with the fact that quality of inversion results of existing major optical remote sensing monitoring means is determined by the source images. Due to mutual restraints of temporal, spatial and spectral resolution of satellite images, common multispectral images cannot realize high-accuracy monitoring in complicated small inland water areas. A method to improve inversion accuracy and spatial resolution of inversion products was carried out in this study. Based on multisource image data, inversion models of chlorophyll a (Chl-a) and cyanobacterial biomarker pigment phycocyanin (PC) concentration in Lake Chaohu were constructed with the image fusion algorithm and machine learning algorithm. Effects of increasing spatial scale of source image on inversion accuracy were verified by comparing accuracy of the inversion models based on fusion image and original moderate-resolution imaging spectroradiometer (MODIS) images in the same period under the same conditions. Moreover, inversion mapping with high accuracy and high spatial scale was accomplished for several days successively. Results demonstrate that accuracy of the inversion model be increased with improving spatial resolution of source images, which further increased spatial scale of inversion products significantly. This study provides a feasible and effective method to realize high-accuracy monitoring of cyanobacteria blooms in small-scaled but complicated inland water environment.