Related Experiment Video
Updated: Aug 18, 2025
Physiological Characterization of the Coral Holobiont Using a New Micro-Respirometry Tool
Published on: April 28, 2023
Two step calibration method for ozone low-cost sensor: Field experiences with the UrbanSense DCUs
J P Sá1, H Chojer1, P T B S Branco1
1LEPABE - Laboratory for Process Engineering, Environment, Biotechnology and Energy, Faculty of Engineering, University of Porto, Rua Dr. Roberto Frias, 4200-465, Porto, Portugal; ALiCE - Associate Laboratory in Chemical Engineering, Faculty of Engineering, University of Porto, Rua Dr. Roberto Frias, 4200-465, Porto, Portugal.
This study validated and calibrated low-cost ozone sensors in a city-wide air pollution network, revealing challenges and proposing a new multi-step calibration strategy for improved data accuracy and reduced reliance on reference instruments.
Area of Science:
- Environmental Science
- Sensor Technology
- Data Calibration
Background:
- Urban air pollution poses significant health risks, necessitating effective monitoring solutions.
- Low-cost sensor networks offer potential for real-time, large-scale air quality data but face validation and calibration challenges.
- The UrbanSense network in Porto, Portugal, utilizes low-cost sensors for various pollutants and meteorological variables.
Purpose of the Study:
- To perform post-deployment validation and calibration of ozone (O3) low-cost sensors within a city-wide air pollution monitoring network.
- To address challenges in data reliability and accuracy encountered with low-cost sensors after deployment.
- To propose and evaluate a novel multi-step calibration strategy for improving sensor data quality.
Main Methods:
- Four data collection units (DCUs) from the UrbanSense network were evaluated for particulate matter (PM), carbon monoxide (CO), ozone (O3), and meteorological variables.
- A two-step calibration strategy was implemented: inter-DCU calibration followed by calibration with a reference-grade instrument.
- Multivariate linear regression (MLR) and machine learning algorithms (artificial neural networks, Stochastic Gradient Boosting Regressor) were employed for calibration model development.
Main Results:
- Preliminary validation indicated unreliable data from PM, CO, and precipitation sensors, and limited data from wind sensors.
- Inter-DCU calibration using MLR showed good performance (R² > 0.80) for O3, temperature, and relative humidity.
- Post-tuning Stochastic Gradient Boosting Regressor yielded the best results for O3 calibration with reference instruments, though overall O3 data predictability remained low (R² ≈ 0.32).
Conclusions:
- Post-deployment validation and calibration of low-cost sensors present significant challenges.
- A novel multi-step calibration strategy, including pre- and post-deployment calibration, is proposed to reduce reliance on reference instruments and minimize data drift.
- Further experimental campaigns are necessary to collect more data and refine calibration models for enhanced air quality monitoring.
