Partitioning the Optical Signal of Lake Erie Using Satellite-Based Remote Sensing Instruments

The complex optical signal of Lake Erie is due to multiple color producing agents (CPAs). Cyanobacteria and Harmful Algal Blooms (CyanoHABs) recurring in Lake Erie prompt the need for rapid and effective analysis of CPAs. Images from the MODIS Aqua NASA and the ESA Sentinel 3A instruments are analyzed using the Kent State University spectral decomposition method, which employs a varimax-rotated, principal component analysis (VPCA) of visible derivative reflectance spectra. The results of the analysis are spectral signatures (component loadings) and spatial maps (component scores).

A MODIS Aqua image acquired on 27 July 2015 and an S3A image acquired on 26 September 2017 were analyzed using the KSU method. Four components were extracted from each image. Component 1 from 27 July 2015 MODIS image is the CyanoHAB signal, a mixture of dinoflagellates and cyanobacteria. MODIS component 2 is clay sediment and phaeophorbide a. MODIS component 3 represents hematite and chlorophyllide a. MODIS component 4 represents chlorophyta. The 26 September 2017 S3A component 1 represents phycocyanin. S3A component 2 represents a mixture of illite and diatoms. S3A component 3 represents chlorophyll a& cyanophyta. S3A component 4 represents geothite & phycocyanin.

Field & lab measurements collected by the Cooperative Institute for Great Lakes Research (CIGLR) on 27 July 2015 (coincident to the MODIS image) and on 25 September 2017 (one day prior to the S3A image) were correlated against the score values at sampling locations for both remote sensing images. Component 3 of the MODIS image correlates to the CIGLR in-water chlorophyll a(RFU) measurements, component 3 of the S3A image correlates to both in situand lab chlorophyll aand microcystin measurements. The optical signal of CPAs can be partitioned using moderate spatial & spectral resolution imagery.

Our results match the NOAA HAB Bulletin well, on visual inspection. The spatial distribution of the MODIS component 1 is very similar to the 29 July 2015 CI Index. Together, the three S3A image CyanoHAB components (1, 3, & 4) visually match the 28 September 2017 CI index. This indicates the KSU method effectively partitions CPAs, and provides information beyond bloom intensity. This method has implications for early CyanoHAB season assessment and validation of future hyperspectral missions.