Development of a sub-seasonal cyanobacteria prediction model by leveraging local and global scale predictors

In recent decades, cultural eutrophication of coastal waters and inland lakes around the world has contributed to a rapid expansion of potentially toxic cyanobacteria, threatening aquatic and human systems. For many locations, a complex array of physical, chemical, and biological variables leads to significant inter-annual variability of cyanobacteria biomass, modulated by local and large-scale climate phenomena. Currently, however, minimal information regarding expected summertime cyanobacteria biomass conditions is available prior to the season, limiting proactive management and preparedness strategies for lake and beach safety. To address this, sub-seasonal (two-month) cyanobacteria biomass prediction models are developed, drawing on pre-season predictors including stream discharge, phosphorus loads, a floating algae index, and large-scale sea-surface temperature regions, with an application to Lake Mendota in Wisconsin. A two-phase statistical modeling approach is adopted to reflect identified asymmetric relationships between predictors (drivers of inter-annual variability) and cyanobacteria biomass levels. The model illustrates promising performance overall, with particular skill in predicting above normal cyanobacteria biomass conditions which are of primary importance to lake and beach managers.