Correlation and Regression Evaluation of Small PM Sensors in Metro Manila

The popularity of compact, off-the-shelf air quality sensors have gained traction in recent years, providing wider access while also being a cheaper alternative to reference monitors which are often inaccessible in some regions. This study investigates the performance of two SPS30 PM2.5 sensors over a 6-month period as they were collocated with Teledyne T640 FEM Monitors located in Metro Manila, Philippines. These sensors are part of an existing IoT network that manages over sensor enclosures that aim to measure air quality. Results show that the sensors can attain strong correlations with its reference monitors, but can also perform poorly due to factors such as seasonal changes or sensor placements. Using simple linear regression to correct these errors managed to achieve
${\mathrm{R}}^{2}$
values up to 0.91 and errors as low as 3
$\mu \mathrm{g}/\mathrm{m}^{3}$
. The promising performance and simple correction of these sensors highlight the potential of equipping the IoT network with cloud-based auto-calibration to provide accurate measurements.