Predicting taxon-specific benthic cyanobacterial mat cover and anatoxin concentrations in northern California rivers

Ecological forecasts often rely on established relationships between the ecological process of interest and predictor variables that are more easily measured. Proliferations of benthic (i.e., bottom-dwelling) cyanobacteria have been increasingly observed in rivers globally and are an emerging ecological forecasting issue as they pose a public health threat due to the production of potent neurotoxins known as anatoxins. Controls on these benthic cyanobacteria are poorly understood, thus predicting or forecasting their extent and anatoxin production is a significant challenge. Here, we measured benthic cyanobacterial cover and anatoxin concentrations for two common taxa associated with anatoxins ( Microcoleus and Anabaena ) at biweekly to weekly intervals during June to September in 2022 and 2023 in three northern Californian rivers. We then built predictive models to test how incorporation of a biotic predictor (river reach-scale gross primary productivity [GPP]) affected predictive accuracy in addition to widely measured abiotic predictors (i.e., nutrients, discharge, and temperature). Temporal patterns in taxon-specific benthic cyanobacterial cover and anatoxin concentrations were highly variable among rivers and between taxa. While Microcoleus cover peaked in rivers during periods of relatively low GPP, there were no clear relationships between GPP and Anabaena cover nor either taxon’s anatoxin concentrations among rivers. Among multiple reaches sampled weekly in the South Fork Eel River, magnitudes of taxon-specific benthic cyanobacterial cover and anatoxin concentrations differed, but the timing of peak taxon-specific cover and anatoxin concentrations were generally consistent. Furthermore, Anabaena displayed a “hysteresis” relationship where increases in cover were followed by increases in anatoxin concentrations while Microcoleus lacked such relationship. While incorporating GPP as a covariate improved our predictions of Anabaena cover, we had more success predicting Microcoleus cover than Anabaena cover due to its strong negative relationship with discharge. In contrast, our models predicting Anabaena anatoxin concentrations outperformed those predicting Microcoleus anatoxin concentrations due to the “hysteresis” relationship between Anabaena cover and anatoxins. Overall, our predictive modeling results highlight the application of ecological forecasting for benthic cyanobacterial cover and anatoxin concentrations in rivers and demonstrate the importance of incorporating taxon-specific predictors into future forecasts of benthic cyanobacteria.