Background/Question/Methods   Toxic algal blooms (Microcystis and Anabaena spp)can pose a health hazard and result in a loss of recreation dollars every year on Lake Champlain, the 11th largest natural lake in America. Since 2003 the University of Vermont's Rubenstein Ecosystems Science Laboratory has been monitoring phytoplankton from July through October to track toxic algal blooms. Using data from the years 2003-2006, we undertook a mechanistic time series modeling approach to answer the question: What drives toxic algal blooms in Lake Champlain? By analyzing population growth rates we tested specific hypotheses about how abiotic factors (nutrients) and biotic factors (competitor phytoplankton species) affect the dynamics of a toxic algal bloom once it has begun. We examined phase-space portraits of each year series and observed two different dynamics in each year, a bloom phase and a decline phase; bloom phases represented the first 5 weeks of each year. By using a time series "stitching" procedure we combined multiple years of data into a bloom phase series (N=16) and a decline phase series (N=40). We then constructed 31 different models each representing a different hypothesis about what controls the dynamics of a bloom and assessed them with AICc for the best fit.

Results/Conclusions   We found that the initial bloom phase exhibited strong density dependence and a strong positive influence of N:P ratio (R2 = 0.80, AIC weight = 0.76). In the decline phase we found no strong support for any one model. The best fitting model had a positive influence of total nitrogen on carrying capacity and negative population growth with a weak fit (R2 = 0.17, AIC weight = 0.14), and the next best fitting model as one of negative population growth (AIC weight =0.06, DAICc = 1.7). We demonstrated that toxic algal blooms undergo two distinct dynamical phases, a bloom initiation period where growth is highly influenced by N:P, and a long decline phase characterized by negative growth. In this second phase nutrients don't seem have any impact on bloom dynamics suggesting that some initial conditions allow for a bloom formation and nutrients help drive the initial phase, but don't control dynamics for the entire length of the bloom.