Cyanobacterial blooms are increasing in their intensity, frequency, and duration. Recent research suggests that analysis of high-frequency time series data can be used to identify statistical early warning indicators of cyanobacterial blooms. This approach uses temporal variation in algal data and rolling window statistics to identify when lakes begin to shift from a clear-water state to a bloom state. The effectiveness of this approach has been shown in models, laboratory studies, and experimental lakes where early warning indicators foreshadowed the onset of blooms. Less is known about how well these early warning indicators work in turbid, eutrophic reservoirs that experience cyanobacterial blooms on a regular basis and where differences between bloom and non-bloom states are less pronounced and can shift quickly between states. Therefore, we used high frequency in-situ probe data that were collected from a eutrophic reservoir almost continuously for four (phycocyanin fluorescence) or eleven years (chlorophyll fluorescence) to characterize variation in bloom dynamics, and determine if early warning indicators using rolling window statistics predicted blooms in the reservoir.
Results/Conclusions
Bloom events, which we defined as a doubling of algal pigment over a period of five days or less, occurred frequently in the reservoir. A total of 59 bloom event (3.7 blooms per year) were detected using chlorophyll and 17 bloom events (2.4 blooms per year) were detected using phycocyanin. However, only 2 common blooms events were detected with both variables. The majority of chlorophyll blooms occurred between May-August, with less than 3.5 percent of the blooms occurring between November-January. Chlorophyll blooms varied in length from 1-20+ days and magnitude from a 1 to > 8-fold increase. Rolling window statistics provided early warning indicators for some blooms, but generally only 1-2 days before the blooms started to develop. The rolling window standard deviation indicator also identified a large number of false positive bloom events. We are conducting additional analyses to better understand how different bloom definitions and rolling window widths influence the efficacy of early warning indicators in eutrophic reservoirs.