Modelling Data for Predicting Cyanobacteria Blooms - JPIWater Project BLOOWATER

The European Union Water JPI (http://www.waterjpi.eu/) has funded the project BLOOWATER (Supporting tools for the integrated management of drinking water reservoirs contaminated by Cyanobacteria and cyanotoxins (https://www.bloowater.eu/) The main objective of the BLOOWATER project is to produce information resources for Public water supply systems to prepare and respond to the risk of the cyanotoxins in drinking water. Practically the project proposes innovative technological solutions aim to develop a methodological approach based on the integration of monitoring techniques and treatment of water affected by toxic blooms. BLOOWATER aims to create forecasting models and systems of surveillance and early warning of toxic blooms to perform immediate actions such as opportune potabilization treatment. The project intends to develop and implement methods to treat cyanobacteria laden water with more efficient processes, to define diagnostic protocols through the use of innovative techniques for water monitoring, and create forecasting models and systems of surveillance and early warning of toxic blooms. Combined these actions will allow water treatment fallibilities to optimally adjust treatment plant operations in response to the onset of cyanobacteria blooms.

To develop cyanobacteria forecasts two different but complimentary methods are being tested

1) The use of Process based models, in this case the combination of the GOTM Hydrodynamic model and the SELMA biogeochemical model coupled using the Framework for Biogechemical Models (FABM) SELMA simulates the biomass of a generic cyanobacteria group and we will test if this can be of useful predictor of cyanobacteria blooms

2) Use of machine learning based models that will be forced and trained on the same data sets used to simulate and verify the process based models, but which may also take as imput data generated by the process based models.

Here we provide an archive of forcing data and measured lake chemistry and phytoplankton data that will be used by BLOOWATER to develop and test model forecasts using both process based modeling and machine learning approaches.

Data are provided for Lake Erken Sweden a primary case study site in the BLOOWATER project

All data files are formatted for use with the GOTM version 5.3 (https://gotm.net/) and SELMA models that are coupled by the frame work for biogeochemical models (https://github.com/fabm-model). The lake model was calibrated using the Parallel Sensitivity Analysis and Calibration tool ParSAC (https://bolding-bruggeman.com/portfolio/parsac/) The measured data used for calibration in the format used by ParSAC are also included in this archive

Additional data and machine learning workflows developed by the BLOOWATER project are available at https://github.com/Shuqi-Lin/Algal-bloom-prediction-machine-learning