Here comes the bloom: forecasting conditions for harmful algal blooms in a large oligotrophic lake

Harmful algal blooms (HABs) may be increasing in frequency and intensity in inland freshwaters. Our knowledge of the mechanisms driving the development of HABs remains limited due to their often-ephemeral nature. The conditions for HABs likely occur well before a HAB is observed at the surface. Yet without foreknowledge, attempting to collect water samples leading up to a HAB is difficult. Lake George, New York, USA is an oligotrophic lake but has experienced several HABs since at least fall, 2020. We deployed and analyzed high-frequency profiling buoy data and high-resolution weather model output to identify potential warning signals that occurred prior to observed HABs. We sought to determine the predictive value of these observations using a near-term iterative forecasting approach. We used an ensemble of one-dimensional lake models driven by NOAA weather forecasts to predict lake temperature and stratification attributes up to 30 days ahead. Prior fall HABs were typically preceded by an increase in Schmidt stability, and we could identify this increase in stability 8-12 days in advance. Using these insights, in 2023 we successfully predicted conditions that resulted in two HABs. Our results demonstrate the capability of parsimonious models to predict HABs days in advance based on proximate weather attributes in oligotrophic lakes. When combined with other models and sensor measurements these forecasts may be generalizable, enabling the ability to accurately forecast rapid changes in phytoplankton community dynamics up to weeks in advance across many types of lakes.

Primary Presenter: Jonathan Borrelli, Rensselaer Polytechnic Institute (borrejj@gmail.com)

Authors:

Jonathan Borrelli, Rensselaer Polytechnic Institute  (borrejj@gmail.com)

Guillaume Auger, IBM Research  (guillaume.auger@ibm.com)

Vincent Moriarty, IBM Research  (Vincent.W.Moriarty@ibm.com)

Kevin Rose, Rensselaer Polytechnic Institute  (rosek4@rpi.edu)