Regression methods in the calibration of low-cost sensors for ambient particulate matter measurements

Abstract

The article presents comparison of regression methods used to obtain calibration formulas for low-cost optical particulate matter sensors. Data for analysis were taken from 1-year collocation study of PMS7003 sensors (Plantower) with researchgrade instrument TEOM 1400a. The PM 2.5 fraction was considered in this study. The results of measurements showed that PMS7003 was characterized by high reproducibility between units (coefficient of variation was lower than 10%), but the raw sensor outputs significantly overestimated PM 2.5 concentrations. Data analysis revealed that simple univariate models were sufficient to obtain a good fitting quality to TEOM data; however, the best results were achieved for raw PM 1 outputs (R 2 ≈ 0.81). The fitting quality was improved when multi-variable equations were examined (R 2 ≈ 0.84). The addition of temperature and relative humidity in the models was also beneficial (R 2 ≈ 0.87). Stepwise selection algorithm was used to choose the best subset of variables in the model. The results of that method were compared with "all possible regression" approach, demonstrating the convenience of stepwise regression. Data from Plantower sensor were also used for training of artificial neural network. That algorithm proved to be very effective for fitting data from one sensor (R 2 ≈ 0.9), but it was susceptible to deviations in the data from the other units. In general, regression analysis proved to be useful for sensor systems for ambient particulate matter measurements.

Spatial calibration and PM2.5 mapping of low-cost air quality sensors

“…Therefore, low-cost sensors with high relative humidity must be established to eliminate the possibility of problematic observations. Multi-variable models may be fitted to obtain reasonable formulas, which can be easily interpreted and implemented from each sensor 22 . However, the current spatial calibration model performs the best (4.8 μg/m 3 ), because the model is already considered a local weighting for the spatial heterogeneity of calibration.…”

Evaluation of optical particulate matter sensors under realistic conditions of strong and mild urban pollution

“…Regarding the weighting factors, sensor manufacturers calculate values to correct for certain effects, such as the fact that OPS cannot detect particles which are too small. Laboratory tests and calibrations of OPS are performed under controlled conditions with known particles, such as polystyrene latex spheres (Walser et al, 2017;Bezantakos et al, 2018), continuously changing monodisperse particles (Kuula et al, 2017;Kuula et al, 2020) or multi-modal particles (Cai et al, 2019). A burning chamber is used in some investigations as well (Wang et al, 2015).…”

The Deployment of a LoRaWAN-Based IoT Air Quality Sensor Network for Public Good

“…We did not perform collocation tests for our sensors, but based our decisions on laboratory tests that have been conducted by the US Environmental Protection Agency as well as independent researchers who have shown that the PMS7003 factory calibration demonstrates low intramodel variability and strong correlation with reference instruments (a coefficient of variation less than 10% between units) [14], [15], [25]. Both PM and CO2 sensors are reported to be sensitive to environmental conditions, such as humidity and temperature, so we use these metrics to validate data points, but also report these metrics with our data set for future studies [26], [27].…”

“…The "community science" approach has become more feasible than ever in recent years due to the release of low-cost and high-precision sensors which can consistently measure particulate matter among other pollutants. While these sensors should still be properly calibrated with reference instruments before having their accuracy assumed, their performance is a marked improvement over past generations, especially considering their price, size, and power consumption as attested by laboratory tests by independent researchers and the EPA [14], [15]. Along with these new sensors have come new proven transmission technologies such as LoRaWAN, enabling scientists and citizens to not just collect data but also reliably transport it to high performance computing infrastructures for processing [16].…”