Real-time, long-term prediction of total suspended solids in an aquaculture reactor using an Internet of Things (IoT) monitoring framework

Total suspended solids (TSS) is a core operational indicator of microbial biomass in biofloc-based recirculating aquaculture systems (B-RAS), yet gravimetric TSS analysis is slow and not available in real time. This study evaluated whether an open-source, low-cost Internet of Things (IoT) monitoring system can support real-time, on-device TSS estimation in a solids-rich side-stream biofloc reactor. The monitoring system measured pH, oxidation-reduction potential (ORP), dissolved oxygen (DO), turbidity, electrical conductivity (EC), and temperature over a six-month period. After confirming sensor stability, TSS sampling campaigns were conducted, and two regression models were developed: a linear regression (turbidity-only) and a multiple linear regression using turbidity, EC, and temperature. Cross-validation showed that the multiple regression model improved model-development performance relative to the linear regression model, increasing R² from 0.65 to 0.77 and reducing MAE and RMSE from 599 and 796 mg/L to 514 and 645 mg/L, respectively. However, a separate field validation showed that model performance depended on the validation window, with the linear regression model producing lower MAE under narrower local operating conditions. These results indicate that predicted TSS is useful for real-time trend monitoring, threshold guidance, and early warning of solids accumulation or biomass loss, but should not be used as the sole automatic control parameter for solids removal or carbon dosing. The developed workflow provides a reproducible embedded sensing approach for precision aquaculture, while emphasizing the need for reactor-specific calibration and periodic gravimetric confirmation.