What Influences Low-cost Sensor Data Calibration? - A Systematic Assessment of Algorithms, Duration, and Predictor Selection

Abstract

The low-cost sensor has changed the air quality monitoring paradigm with the capacity for efficient network expansion and community engagement. The surge in its use has sparked a new research wave in understanding its data quality. Many studies have employed field calibration to improve sensor agreement with co-located reference monitors. Yet, studies that systematically examine the performance of different calibration techniques are limited in scope and depth. This study comprehensively assessed ten widely used data techniques, namely AdaBoost, Bayesian ridge, gradient tree boosting, K-nearest neighbors, Lasso, multivariable linear regression, neural network, random forest, ridge regression, and support vector machine. We compared their performance using a standardized baseline dataset and their responses to various parameter combinations. We further assessed the training sample size effect to understand the optimal duration for field calibration for achieving good accuracy. Finally, we tested different predictor combinations to address whether the inclusion of more predictors will lead to better performance. Using baseline data, the neural network achieved the best performance, followed by the four regression-based methods, showing very consistent and stable performance. While confirming that the latest research tendency is deep learning, regression is still a viable option for studies with limited effort in parameter tuning and method selection, especially considering its computational efficiency and simplicity. The sample size effect is most evident when the sample size drops below 30%, which is equivalent to six weeks of continuously collected hourly data. Although algorithms react differently to the number of predictors, their performance was typically boosted by adding more predictors, especially the particle count and humidity. Our study not only describes an approach of sophisticated data-driven calibration for practical applications, but also provides insights into the compounding impacts of parameters, samples, and predictors in algorithm performance.

Urban PM2.5 concentrations in a small Colombian city and the impact associated with particle emissions generated by small-scale lime production

“…The overall conclusions from this study would not change as a consequence of which regression was used, but the comparison underline the uncertainties linked to low-cost sensor output. A recommendation for future assessments with low-cost optical sensors is to perform the intercalibration both before and after the campaign and also, as recommended by Liang and Daniels (2022), during various weeks to yield better statistics.…”

“…A general conclusion is the need to evaluate the performance of the particular sensor used, leaving recommendations on individual calibration against reference monitors. The procedure of calibrating a low-cost sensor through collocating it at the side of a reference monitor has been discussed in depth by Diez et al (2022) and Liang and Daniels (2022). Low-cost and small sensors allow a huge number of devices to operate simultaneously, describing spatial patterns with great detail that traditional reference monitoring cannot, but they are also beneficial in low-and middle-income countries where few reference measurements are made (Giordano et al, 2021).…”

Field testing of low-cost particulate matter sensors for Digital Twin applications in nanomanufacturing processes

“…However, calibration aerosols used in the laboratory differ considerably from aerosols found in real scenarios; so, LCPMS require additional field calibrations before implementation to ensure measurement accuracy. Postprocessing calibration strategies to improve LCPMS performance include models with corrections for hygroscopicity, traditional simple or multiple linear regression models, and algorithms based on machine learning, where the simplicity and applicability of the model is always prioritized as basic selection criterion (parsimonious approach) [5,15,16,18,20,23,24].…”

“…The absence of a standardized evaluation/calibration methodology to compare LCPMS and describe its performance, leads to the use of a variety of metrics and test procedures by the scientific community [7,15,22,23,24], which limits the inter-comparison of results and constitutes a barrier for the expansion of the technology. In this context, CEN/TC 137 is currently developing a future standard to provide guidelines and testing procedures for LCPMS to be implemented in workplaces for measuring engineered NOAA [7].…”

Low-processing data enrichment and calibration for PM2.5 low-cost sensors

“…Field calibration is an approach to correct the signal of low-cost sensors for the influences of their interfering environment [10][11][12]. This technique is based on models such as multi-linear or machine learning regression, using the measurements of the low-cost sensor and variables from external environment as features and monitoring stations as reference [13][14][15][16][17].…”