Kemal Maulana Alhasa, Mohd Shahrul Mohd Nadzir, Sawal Hamid Md Ali, Wahyu Sasongko Putro
This study evaluates the robustness of a previously developed calibration model for low-cost ozone sensors, based on the Adaptive Neuro-Fuzzy Inference System (ANFIS). The model was deployed at a different site without retraining. It was tested in Putrajaya, Malaysia, over a 2-month period in 2018 and compared with conventional laboratory-calibrated models for CO and NO₂ sensors. The ANFIS model demonstrated consistently strong performance for the OX-A431 sensor, with R2 values approaching 0.9 and lower RMSE, confirming its transferability. However, deviations of up to 70 ppb were recorded during high ozone episodes. This may be due to the limited input variables used—namely O₃, NO₂, temperature, and relative humidity—while other influential factors such as co-pollutants or atmospheric pressure were not included. In contrast, laboratory-calibrated models for the CO-A4 and NO₂-A43F sensors exhibited poorer performance (R2 = 0.13–0.73), indicating low adaptability under field conditions. These findings underscore the importance of field calibration for low-cost sensors and suggest that incorporating additional environmental and chemical parameters may further improve calibration accuracy. This study contributes to the advancement of generalized machine learning calibration frameworks to enable scalable low-cost sensor networks for ambient air quality monitoring. © The Author(s), under exclusive licence to Springer Nature Switzerland AG 2025.
Research Center for Environmental and Clean Technology, National Research and Innovation Agency, Kawasan Puspiptek Gedung 820, Tangerang Selatan, 15314, Indonesia; Department of Earth Sciences and Environment, Faculty of Science and Technology, Universiti Kebangsaan Malaysia, Bangi, Selangor, 43600, Malaysia; Department of Electrical, Electronic and Systems Engineering, Faculty of Engineering and Built Environment, Universiti Kebangsaan Malaysia, Bangi, Selangor, 43600, Malaysia; Departement of Electrical Engineering, Universitas Negeri Surabaya, Surabaya, Indonesia
