Comparing predictive cyanobacterial models from temperate regions

Abstract

The global increase in cyanobacterial bloom reports heightens the need for a critical evaluation of models used for their prediction. In particular, it is unclear whether empirical cyanobacterial models vary regionally because of differences in environmental conditions and (or) community composition. To address this question, we applied linear and nonlinear models as well as mixed-effect models to a dataset of seasonally integrated environmental and cyanobacterial measurements collected from 149 lakes spread across three regions in Canada. Across all lakes, we found that linear models outperformed nonlinear approaches and that nutrients (phosphorus, nitrogen) were the best predictors of cyanobacterial biomass. Importantly, there was no significant regional difference in predicted cyanobacterial responses to nutrients, even though the means for these variables were different among regions. From canonical correspondence analyses of taxonomic biomass data, temperature, water column stability, and forms of inorganic nitrogen were also important in explaining cyanobacterial community structure at the regional scale. Based on these analyses, we conclude that North American models are suitable for estimating total cyanobacterial biomass from any particular temperate region in Canada.