CERTH-ITI published the first FUELPHORIA-related scientific paper 

CERTH-ITI has recently published a scientific paper, supported by FUELPHORIA, that will serve as a foundation for developing an algal growth forecasting algorithm using camera images. The partner institution is responsible for the activities linked to the design of an AI-based system for monitoring and forecasting algal growth in the project’s demo 3 ponds, located in Thessaloniki, Greece.   

The publication, titled “Forecasting Lakes’ Chlorophyll Concentrations Using Satellite Images and Generative Adversarial Networks”, reflects previous work of CERTH-ITI in the domain of chlorophyll-α concentration forecasting in inland water bodies using satellite images. The algorithm described in the publication was initially planned to be used in the context of the FUELPHORIA project. Still, modifications to the construction plans of the demo 3 unit prevented the utilisation of satellite images. However, the work carried out for this paper, particularly the research regarding the use of RGB-NIR bands, will serve as a foundation for developing an algal growth forecasting algorithm using camera images

The scientific paper showcases a model that forecasts chlorophyll-α values in water bodies, a common water quality indicator. This method is proposed because it implies a reduced cost compared to in-situ data sampling. A data set of satellite (Sentinel 2) images of 15 lakes around Europe taken during 12 consecutive months was used to train a Generative Adversarial Network (GAN) to recognize chlorophyll-α spatiotemporal patterns, by using a pix2pix algorithm to match consecutive past and current chlorophyll-α maps to future ones. The result is a model, which can be used to make chlorophyll-α maps predictions with low computational cost and without using in-situ data. The model was tested by applying it to three water bodies in Europe that were not included in the training data set. Results showed it performed accurately.  

The accuracy of the model can be further improved by training on wider timeframes and a more diverse set of water bodies, as well as using more metadata (such as temperature) as inputs.  The objective is to enhance its ability to recognize more complex spatiotemporal patterns and better forecast chlorophyll-α seasonality.   

In the context of the FUELPHORIA project, the development of an algal growth forecasting algorithm plays a key role in the execution of the Demo 3 pilot. On this site, the project aims to demonstrate the efficiency of producing biodiesel from microalgae growing in high-rate open ponds.    

The complete scientific paper is also available here.