Monitoring cyanobacterial blooms: a strategy combining predictive modeling and remote sensing approaches

Development of effective monitoring programs and early alert systems of cyanobacterial harmful blooms (CyanoHABs) is a challenge due to their rapid temporal dynamics and high spatial heterogeneity. We provide a new approach for monitoring CyanoHABs in large ecosystems using a strategy combining modeling with in situ data and remote sensing methods. Between 2014 and 2021, we sampled phytoplankton and measured temperature and conductivity (continuously) at a coastal site at the Rio de la Plata estuary (South America). We used a Bayesian model to predict favorable conditions for bloom occurrences, using temperature and conductivity (a proxy for salinity). We defined a polygon area of 40 km(2) and obtained 121 cloud-free satellite images (Sentinel-2) in which 10 "small" (< 1% of polygon), 4 "medium" (> 1% and < 5%), and 2 "large" (> 5%) blooms were detected. A 7-day period of favorable environmental conditions was the best time frame to predict large blooms and medium size blooms. Integrating the bloom extent with modeling outputs generates valuable new information for management. The continuous model predictions allow for evaluation of the persistence/growth of blooms and they fill in an important gap for management when images are not available (i.e., cloud cover). We propose a monitoring strategy that combines information about the size of the remotely detected blooms with in situ conditions to evaluate the actions that stakeholders should take. Our approach is a rapid and cost-effective strategy with high potential for developing early warning systems for monitoring CyanoHABs in large ecosystems.