Adaptive forecasting of phytoplankton communities.

Area of Science:

• Environmental Science
• Limnology
• Computational Ecology

Background:

• Harmful algal blooms (HABs) are a growing global threat to water quality, ecosystems, and recreational activities.
• Effective management of HABs necessitates accurate prediction of their occurrence.
• Existing forecasting methods require enhancement to address model uncertainties and nonlinear dynamics.

Purpose of the Study:

• To evaluate the forecasting potential of the PROTECH phytoplankton community model for harmful algal blooms.
• To implement a pseudo-real-time data assimilation scheme using the Ensemble Kalman Filter for improved forecast accuracy.
• To assess forecast skill for chlorophyll a and phytoplankton community structure in two English lakes.

Main Methods:

• Utilized the PROTECH model within a data assimilation framework incorporating the Ensemble Kalman Filter.
• Conducted pseudo-real-time forecasting experiments on two mesotrophic lakes with varying characteristics.
• Propagated model uncertainties and nonlinearities to forecast outputs for chlorophyll a and phytoplankton composition.

Main Results:

• Achieved some success in forecasting chlorophyll a concentrations, outperforming persistence forecasts in certain instances.
• Forecast skill generally decreased with longer forecasting horizons, with notable promise for predictions up to four to five days.
• Phytoplankton community structure forecasts showed general consistency with observations, though cyanobacteria prediction proved challenging.

Conclusions:

• The PROTECH model, integrated with Ensemble Kalman Filter data assimilation, shows potential for short-term algal bloom forecasting.
• Forecasting accuracy is sensitive to the prediction period, highlighting the value of short-range outlooks.
• Further refinement is needed to improve the prediction of specific phytoplankton groups, such as cyanobacteria, due to functional species interchangeability.