Researchers at the Complutense University of Madrid (UCM) have developed a low-cost device capable of integrating shallow learning algorithms to anticipate cyanobacterial blooms in surface waters. The work, published in the journal Water Research, demonstrates that artificial intelligence can be applied in a practical way in aquatic environments, opening new possibilities for early warning systems.
ESP32 microcontroller development board, used for the work. Source: Sandubete-López, J. et al.
Low-cost artificial intelligence in the field
The proposed model is based on lightweight machine learning, optimized to run on microcontrollers with low energy consumption and minimal cost. This architecture enables the device to process data directly in the field, without the need for large servers or complex infrastructure.
The research shows that, with this technology, it is possible to anticipate bloom events far enough in advance for managers to take preventive measures. In addition, the compact and affordable nature of the device facilitates deployment across different water bodies, making it an accessible tool even for operators with limited resources.
Conceptual diagram of the proposed framework. Source: Sandubete-López, J. et al.
Implications for water management
The integration of artificial intelligence algorithms into low-cost devices marks a turning point in the digitalization of the water cycle. Faced with the growing threat posed by cyanobacterial blooms with consequences for the quality, supply, and recreational use of water bodies, this solution offers utilities and administrations the possibility of moving from a reactive to a predictive approach, incorporating automatic prediction directly into existing infrastructure, even in regions with limited technical resources or reduced connectivity.
Although the system still needs to be validated under various hydrological and climatic conditions to confirm its large-scale applicability, the authors emphasize that the model provides reliable medium-term predictions of cyanobacterial blooms. Its integration into a low-cost device represents a realistic step forward for improving real-time monitoring systems.
