Abstract Cyanobacterial blooms are causing increasing issues across the globe. Bloom forecasting can facilitate adaptation to blooms. Most bloom forecasting models depend on weekly or fortnightly sampling, but these sparse measurements can miss important dynamics. Here we develop forecasting models from five years of high frequency summer monitoring in a shallow lake (which serves as an important regional water supply). A suite of models were calibrated to predict cyanobacterial fluorescence (a biomass proxy) using measurements of: cyanobacterial fluorescence, water temperature, light, and wind speed. High temporal autocorrelation contributed to relatively strong predictive power over 1, 4 and 7 day intervals. Higher order derivatives of water temperature helped improve forecasting accuracy. While traditional monitoring and modelling have supported forecasting on longer timescales, we show high frequency monitoring combined with telemetry allows forecasting over timescales of 1 day to 1 week, supporting early warning, enhanced monitoring, and adaptation of water treatment processes.

Forecasting changes to ecological communities is one of the central challenges in ecology. However, nonlinear dependencies, biotic interactions and data limitations have limited our ability to assess how predictable communities are. Here, we used a machine learning approach and environmental monitoring data (biological, physical and chemical) to assess the predictability of phytoplankton cell density in one lake across an unprecedented range of time-scales. Communities were highly predictable over hours to months: model R2 decreased from 0.89 at 4 hours to 0.74 at 1 month, and in a long-term dataset lacking fine spatial resolution, from 0.46 at 1 month to 0.32 at 10 years. When cyanobacterial and eukaryotic algal cell densities were examined separately, model-inferred environmental growth dependencies matched laboratory studies, and suggested novel trade-offs governing their competition. High-frequency monitoring and machine learning can set prediction targets for process-based models and help elucidate the mechanisms underlying ecological dynamics.

Abstract Forecasting anthropogenic changes to ecological communities is one of the central challenges in ecology. However, nonlinear dependencies, biotic interactions and data limitations have limited our ability to assess how predictable communities are. Here we used a machine learning approach and environmental monitoring data (biological, physical and chemical) to assess the predictability of phytoplankton cell density in one lake across an unprecedented range of time scales. Communities were highly predictable over hours to months: model R 2 decreased from 0. 89 at 4 hours to 0.75 at 1 month, and in a long-term dataset lacking fine spatial resolution, from 0.46 at 1 month to 0.32 at 10 years. When cyanobacterial and eukaryotic algal cell density were examined separately, model-inferred environmental growth dependencies matched laboratory studies, and suggested novel trade-offs governing their competition. High-frequency monitoring and machine learning can help elucidate the mechanisms underlying ecological dynamics and set prediction targets for process-based models.