Operational short-term hydro-ecological forecasting for algal-related threats in seawater desalination

Coastal resources are productive drivers of the so-called blue economy, impacting rapidly growing industries, as the Seawater desalination. Yet, the efficiency of desalination operations is at stake as a result of an imminent operational threat at a global scale, i.e., the proliferation of microscopic algae in seawater. Algal blooms are associated with operational difficulties in the desalination industry, i.e., clogging and bio-fouling, which increase the costs of chemicals, energy and maintenance.

To alleviate the impact of algal blooms, desalination could be supported by innovative tools that foretell the onset and evolution of bloom events. However, the desalination sector lacks near-real time decision-support tools. This work aims to address this gap. To this end, an operational forecasting service was developed, deployed and tested in a seawater desalination plant,  located at the Saronic Gulf (Greece).

The operational forecasting service comprises three components: (a) a hydrodynamic component, (b) a water quality component, and (c) an early warning system for algal bloom events.

The hydrodynamic model predicts the hydrodynamic regime in the Gulf, including vertical mixing, circulation patterns, temperature and salinity profiles. The hydrodynamic model accounts for the heat exchange between the water body and the atmosphere, the salinity, wind and wave action. Both the hydrodynamic and the wave component have been calibrated and validated using satellite-derived and reanalysis data for the first and in situ data for the latter. Specifically, on the validation of the hydrodynamic component, comparisons with satellite-derived water temperatures proved the model’s ability to accurately predict water temperature profiles in the domain, with MAE=1.11oC and MAPE=4% at the validation period from 01/07/2018 to 30-11-2018. To further improve the predictive capacity of the forecast model, the service assimilates satellite-derived sea surface temperature (obtained by Landsat-8 imagery) using the Ensemble Kalman Filtering method.

The prediction of algal-related water quality attributes (i.e., chlorophyll-a) is based on a data-driven approach. An ensemble learning method (i.e., a random forest) was trained to map hydrodynamic data (temperature, mixed layer thickness), biogeochemical data (inorganic nutrients) and meteorological data (air temperature, wind speed, solar radiation) to chlorophyll-a concentrations at the area of interest. The random-forest-based model produced accurate predictions in hindcast (the mean absolute percentage error was 14% for the held-out data), allowing for its further deployment in an operational setting.

Ultimately, forecasted hydrodynamic and water quality attributes of the coastal zone are integrated into an early warning system that generates and disseminates readily interpretable warning information to enable operators threatened by a probable shift in the regime of the coastal environment to act promptly and appropriately to reduce the vulnerability of those due to be impacted.

In conclusion, this work delivers an operational platform that predicts accurately algal-related parameters in coastal waters. Following its deployment and testing in hindcast, the service line will be tested and validated in operational conditions, aiming to assess the limitations in its forecasting abilities.

Acknowledgements: This work is supported by a grant from Iceland, Liechtenstein and Norway through the EEA Financial Mechanism 2014-2021, within the framework of the Programme “Business Innovation Greece”.