Cyanobacteria are ubiquitous in aquatic environments and typically present no health concerns. Yet when conditions are suitable, cyanobacteria can proliferate rapidly and form blooms. Since many cyanobacteria have the potential to produce cyanotoxins, these blooms can lead to human health concerns. We hypothesized that toxic cyanobacteria blooms may alter the composition of the entire phytoplankton community. If true, then we expected to find recognizable patterns in phytoplankton community structure among water samples from lakes with detectable levels of cyanotoxins.
Our main goal was to determine if there were continental-scale patterns of phytoplankton community composition across the United States that were associated with the occurrence of detectable levels of three cyanotoxins: microcystins, cylindrospermopsin, and saxitoxins. We analyzed landscape composition, along with phytoplankton counts, and physico-chemical monitoring data collected from 1000+ lakes during the 2007 National Lake Assessment. We used nonmetric multidimensional scaling (NMS) to reduce the dimensionality of the phytoplankton abundance data and elucidate any dominant patterns in community composition. Cluster analysis was then used to delineate boundaries between distinct community types within the NMS space. Finally, a statistical tests were used to explore relationship between community composition and environmental factors (including water quality measures, land use variables, and cyanotoxin detection).
Our final NMS configuration of four dimensions had a 0.178 stress value (NMS measure of model fit that ranges from 1 (poorly fit) to 0 (perfect fit)). There were three distinct clusters within the first two dimensions and these explained the strongest variance in community composition. Statistically significant gradients of microcystin and saxotoxin detections occurred predominately along the first NMS dimension. A cylindrospermopsin detection gradient was evident along the second dimension. Evaluation of indicator species among clusters revealed that potential toxin-producing cyanobacteria were present, but did not dominate community composition in clusters associated with toxin detections. Despite these relationships, random forest based classification revealed that geography, elevation, phytoplankton abundance, and landscape variables were stronger predictors of phytoplankton community cluster types. In conclusion, the phytoplankton community types across the conterminous US have a robust spatial pattern but only weak support for the hypothesis that community structure is dominated by cyanobacteria or associated with detectable cyanotoxins.