National forecasting of cyanobacterial harmful algal bloom events: a three-year model evaluation.

Publicly available cyanobacterial harmful algal blooms (cyanoHAB) forecasts are now available at www.epa.gov/habs/hab-forecasts, offering water quality and public health managers a tool for anticipating cyanoHAB conditions in the largest United States lakes and reservoirs. These near-term forecasts support proactive management actions, such as sampling and additional monitoring, aligned with the U.S. Harmful Algal Bloom and Hypoxia Research and Control Act call for advancements in cyanoHAB forecasting. Accurate forecasting technology could save communities health related costs, monitoring and response resources, and drinking water treatment costs through early detection of cyanoHABs. A hierarchical Bayesian spatiotemporal model predicts cyanoHABs by determining probabilities of weekly lake exceedance of 12 μg/L chlorophyll-a for 2,192 satellite resolved lakes across the contiguous United States. The 2021-2023 predictions from this model were evaluated to determine occurrence of false negative and false positive predictions to update model interpretation. In total, 331,049 events were analyzed, of which false negatives represented 1.34% and false positives 8.06% of forecasts. False negatives were predominantly events lasting 1-3 weeks in the late fall. False positives varied in temporal length but were typically associated with actual blooms, often incorrectly forecasting blooms before, after, or between actual bloom events. The three-year model run demonstrated performance consistency and operational readiness. A stoplight approach for probability reporting could improve data visualization. Insights into forecast limitations support transparent communication and stakeholder engagement. This evaluation helps managers interpret model limitations when integrating results into decision frameworks to mitigate cyanoHAB impacts, protecting public and environmental health.