Thu, Aug 18, 2022: 5:00 PM-6:30 PM
ESA Exhibit Hall
Background/Question/Methods: Events that impact aquatic systems on short time scales have traditionally been difficult to capture, but recent developments in water quality remote sensing technologies have increased our ability to study these dynamic, complex phenomena at their relevant time scales. In this study we used high-frequency water quality buoy data and developed short-term (timescale of days to weeks) forecasting models to examine drivers of bloom onset and senescence in two shallow, eutrophic bays in Lake Champlain, USA. Both bays experience cyanobacteria blooms in the late summer and are 18km apart, thus experience similar weather. Their configuration and proximity to one another provide an ideal natural laboratory. We used random forest machine learning algorithms to rank feature importance of key environmental parameters in the systems (water temperature, water column stratification, riverine discharge, wind speed and direction, and solar radiation) and determined the environmental conditions that were most important in predicting cyanobacteria levels days and weeks into the future.
Results/Conclusions: Results from the short-term forecasting models suggest that despite experiencing the same weather, the environmental conditions important for predicting bloom onset in each bay were different both within the same bloom period and from year to year. Preliminary results from the weekly forecast models suggest that in one of the systems, during both the 2017 and 2018 bloom periods, water temperature and wind speed were the most important environmental conditions in predicting bloom formation one week before bloom onset. During the same bloom periods in the other system, solar radiation and wind direction were the main predictive variables one week prior to bloom onset. Preliminary results also suggest that water stratification (an environmental condition often associated with cyanobacteria bloom formation) was important for bloom onset in only one of the two bays, likely due to the system differences in bay hydrodynamics and hydrologic connectivity to the rest of the lake.
Results/Conclusions: Results from the short-term forecasting models suggest that despite experiencing the same weather, the environmental conditions important for predicting bloom onset in each bay were different both within the same bloom period and from year to year. Preliminary results from the weekly forecast models suggest that in one of the systems, during both the 2017 and 2018 bloom periods, water temperature and wind speed were the most important environmental conditions in predicting bloom formation one week before bloom onset. During the same bloom periods in the other system, solar radiation and wind direction were the main predictive variables one week prior to bloom onset. Preliminary results also suggest that water stratification (an environmental condition often associated with cyanobacteria bloom formation) was important for bloom onset in only one of the two bays, likely due to the system differences in bay hydrodynamics and hydrologic connectivity to the rest of the lake.