A new observational-modelling framework for algae bloom monitoring and forecast in the Baltic Sea

Algae blooms, specifically cyanobacterial blooms, are frequent in the Baltic Sea and pose major environmental problems for the marine ecosystem and coastal societies. Surface accumulations of algae exacebate eutrophication, limit access to oxygen and can be toxic to humans and marine life. They affect marine services including drinking water resources, marine operations, tourism and fishing. Monitoring of algae blooms based on satellite-borne and in-situ data have been ongoing for years. However, a proper assessment of monitoring needs to cover complex spatio-temporal variability of the blooms as well as reliable early warning and forecasts systems are still lacking, owing to the complexity of physical and biological processes involved in algae growth and too sparse data to constrain complex marine ecosystem models. As algal blooms are expected to intensify under the observed long-term warming of surface waters, developing relevant monitoring-early warning systems is a priority.

Our interdisciplinary collaboration aims at a tantalizing task of building a framework tailored for monitoring and forecasting of algae blooms in the Baltic Sea. The framework combines surface drift observations, in-situ observations, remotely-sensed chlorophyll products as well as numerical simulations of Lagrangian (drifting) trajectories driven by the ocean state forecast available at the Copernicus Marine Environment Monitoring Service (CMEMS). The first step toward this goal consisted of collecting observations of the surface drift in the Baltic Sea relevant to the dispersion of algae accumulations. To this end, we deployed 6 CARTHE Smart surface drifter platforms in Western Gotland Basin in August 2021. The CARTHE drifter platforms are designed to sample sea currents close to the surface compared to other standard drift measurements, provide a very accurate positioning data at 15 minute intervals, and their floating parts are biodegradable. We will present data from this experiment as well the results from a comparison between the observed surface drift and CMEMS-driven Lagrangian simulations. The results using relative dispersion statistics point to a good skill of the model-driven drift forecast (a few km error in mean dispersion over a two day scale). We extend the analysis including Lagrangian ecosystem modelling, spectral analysis and clustering approaches, taking into the consideration sparseness of in-situ data.