Two decades of cyanobacterial bloom dynamics in the Great Lakes: insights from multi-mission ocean color sensors

Abstract Cyanobacterial harmful algal blooms (cyanoHABs) pose an ongoing threat to water quality, freshwater ecosystems, economies, and public health around the United States Great Lakes region. This study presents a 25 year satellite-based time series (2000–2024) of cyanoHAB observations across the region, generated using data from multiple ocean color sensors, including Envisat-Medium Resolution Imaging Spectroradiometer (MERIS), Sentinel-3A/B Ocean Land Color Imager (OLCI), and MODIS-Terra. To fill the observation gap between MERIS and OLCI sensors, we used CyanNet—a science-informed deep learning framework—to derive the Cyanobacteria Index (CI cyano ) from MODIS-Terra, harmonized with CI cyano products from MERIS and OLCI. This approach enabled the creation of the longest continuous and cross-sensor-consistent record of cyanoHAB observations in the Great Lakes to date. We found that differences between sensors were not solely due to CyanNet model limitations but also to variations in temporal coverage and observational timing of the three satellite missions. Analysis of this multi-decadal dataset reveals spatially variable trends in cyanoHAB intensity, frequency, and bloom extent in the basins having the most prolific, historical cyanoHABs: Western Lake Erie, Sandusky Bay (SNB), Saginaw Bay (SGB), Lake Winnebago (LW), and Green Bay (GB). For example, Lake Erie experienced its most severe bloom in 2011, with a cumulative CI cyano of 95.9 and the largest bloom extent on record (6121 km 2 ). The western basin of Lake Erie shows an overall increasing trend in bloom intensity and extent since 2000; however, the blooms exhibited a high degree of interannual variability. In contrast, SNB and SGB exhibit slight decreasing trends in bloom severity, while GB and LW show negligible changes. These long-term observations underscore the value of harmonized satellite data time series and science-informed deep learning models for monitoring inland water quality, and provide a critical foundation for understanding bloom phenology and managing future cyanoHAB risks in the Great Lakes region.