Both biotic and abiotic predictors explain significant variation in cyanobacteria biomass across lakes from temperate to subarctic zones

• Published in: Limnology and Oceanography
• Main Authors: MacKeigan, Paul W., Taranu, Zofia E., Pick, Frances R., Beisner, Beatrix E., Gregory‐Eaves, Irene
• Format: Article in Journal/Newspaper
• Language: English
• Published: Wiley 2023
• Subjects: Canada; Subarctic
• Online Access: http://dx.doi.org/10.1002/lno.12352; https://aslopubs.onlinelibrary.wiley.com/doi/pdf/10.1002/lno.12352

Abstract The development of cyanobacteria blooms is of increasing concern in many lakes worldwide, and as a result, modeling their predictors is vital for understanding where and why they occur. In this study, we developed and analyzed a 640‐lake data set that spans Canada and 12 ecozones to identify the drivers of cyanobacteria biomass and of several key toxin‐ and bloom‐forming genera ( Microcystis , Aphanizomenon , and Dolichospermum ). The database consisted of an exhaustive list of potential predictors ( n = 55), including water chemistry, land‐use, and zooplankton variables. We applied a series of empirical modeling approaches to identify significant predictors and thresholds (generalized linear and additive models, mixed effect regression trees), all while accounting for ecozone variability. Across all modeling approaches, and ecozones total phosphorus was identified as the most important predictor of total cyanobacterial and focal genera biomass. In addition, cyanobacteria across Canada showed significant associations with increasing dissolved organic and inorganic carbon, and several ions. Despite the widely held notion that cyanobacteria are often toxic and/or a poor food source for zooplankton, we found a positive relationship between cyanobacteria and zooplankton, particularly with daphnid and copepod biomass. Localized top‐down forces and evolutionary adaptations resulting from long‐term exposure in eutrophic lakes are among the possible explanations for this observed positive association. By considering a suite of complementary modeling approaches, we found that nonlinear models provided greater predictive power and the random ecozone effect was minor due to the overarching importance of local abiotic and biotic factors.