Metrological Characterization of Low-Cost CO2 Sensors for Environmental Monitoring Applications

Abstract

Low-cost sensors are increasingly used in atmospheric monitoring to provide spatially distributed measurements of gas concentrations, often through sensor networks. However, their application is still limited by the lack of metrologically robust characterization procedures. This work addresses a metrological characterization of SCD30 (Sensirion) non-dispersive infrared (NDIR) low-cost sensors for atmospheric carbon dioxide measurements, tested against an NDIR reference analyzer. A dedicated experimental facility and a systematic characterization procedure were developed using a dynamic dilution method in an isolator, covering a concentration range of approximately (350–950) µmol/mol, representative of typical ambient conditions. The analysis focused on sensor performance, calibration functions, uncertainty evaluation, and statistical indicators. Results show that all sensors exhibit good linearity but significant systematic deviations. The uncertainty evaluation highlights reproducibility as the dominant contribution (>85% of the uncertainty budget). The results demonstrate that, after applying calibration, root mean square error (RMSE) and mean absolute error (MAE) are reduced below 20 µmol/mol, demonstrating a substantial improvement in accuracy. The Bland–Altman analysis shows a good agreement between the reference instrument and the low-cost sensors. The proposed methodology provides a robust framework for the metrological evaluation and calibration of low-cost sensors, which can be extended to other atmospheric gases.