Long-term behavior and stability of calibration models for NO and NO₂ low cost sensors

Abstract. Assessing the intracity spatial distribution and temporal variability in air quality can be facilitated by a dense network of monitoring stations. However, the cost of implementing such a network can be prohibitive if traditional high-quality, expensive monitoring systems are used. To this end, the Real-time Affordable Multi-Pollutant (RAMP) monitor has been developed, which can measure up to five gases including the criteria pollutant gases carbon monoxide (CO), nitrogen dioxide (NO2), and ozone (O3), along with temperature and relative humidity. This study compares various algorithms to calibrate the RAMP measurements including linear and quadratic regression, clustering, neural networks, Gaussian processes, and hybrid random forest–linear regression models. Using data collected by almost 70 RAMP monitors over periods ranging up to 18 months, we recommend the use of limited quadratic regression calibration models for CO, neural network models for NO, and hybrid models for NO2 and O3 for any low-cost monitor using electrochemical sensors similar to those of the RAMP. Furthermore, generalized calibration models may be used instead of individual models with only a small reduction in overall performance. Generalized models also transfer better when the RAMP is deployed to other locations. For long-term deployments, it is recommended that model performance be re-evaluated and new models developed periodically, due to the noticeable change in performance over periods of a year or more. This makes generalized calibration models even more useful since only a subset of deployed monitors are needed to build these new models. These results will help guide future efforts in the calibration and use of low-cost sensor systems worldwide.

Abstract. Assessing the intra-city spatial distribution and temporal variability of air quality can be facilitated by a dense network of monitoring stations. However, the cost of implementing such a network can be prohibitive if traditional high-quality, expensive monitoring systems are used. To this end, the Real-time Affordable Multi-Pollutant (RAMP) monitor has been developed, which can measure up to five gases including the criteria pollutant gases carbon monoxide (CO), nitrogen dioxide (NO2), and ozone (O3), along with temperature and relative humidity. This study compares various algorithms to calibrate the RAMP measurements including linear and quadratic regression, clustering, neural networks, Gaussian processes, and random forests. Using data collected by more than sixty RAMP monitors over periods ranging up to eighteen months, it was found that quadratic regression models or a hybrid of random forest and linear models tend to be the most effective calibration models overall. In specific cases, other types of models can have comparable or even superior performance. Furthermore, generalized calibration models may be used instead of individual models with only a small reduction in overall performance. For long-term deployments, it is recommended that new models be developed each year, due to the noticeable change in performance when models for one year were used for processing data collected in the subsequent year. This makes annually-developed generalized calibration models even more useful since only a subset of deployed monitors are needed to build these models. These results will help guide future efforts in the calibration and use of low-cost sensor systems worldwide.

Low-cost sensor technology has been available for several years and has the potential to complement official monitoring networks. The current generation of nitrogen dioxide (NO2) sensors suffers from various technical problems. This study explores the added value of calibration models based on (multiple) linear regression including cross terms on the performance of an electrochemical NO2 sensor, the B43F manufactured by Alphasense. Sensor data were collected in duplicate at four reference sites in the Netherlands over a period of one year. It is shown that a calibration, using O3 and temperature in addition to a reference NO2 measurement, improves the prediction in terms of R2 from less than 0.5 to 0.69–0.84. The uncertainty of the calibrated sensors meets the Data Quality Objective for indicative methods specified by the EU directive in some cases and it was verified that the sensor signal itself remains an important predictor in the multilinear regressions. In practice, these sensors are likely to be calibrated over a period (much) shorter than one year. This study shows the dependence of the quality of the calibrated signal on the choice of these short (monthly) calibration and validation periods. This information will be valuable for determining short-period calibration strategies.

With attention increasing regarding the level of air pollution in different metropolitan and industrial areas worldwide, interest in expanding the monitoring networks by low-cost air quality sensors is also increasing. Although the role of these small and affordable sensors is rather supplementary, determination of the measurement uncertainty is one of the main questions of their applicability because there is no certificate for quality assurance of these non-reference technologies. This paper presents the results of almost one-year field testing measurements, when the data from different low-cost sensors (for SO2, NO2, O3, and CO: Cairclip, Envea, FR; for PM1, PM2.5, and PM10: PMS7003, Plantower, CHN, and OPC-N2, Alphasense, UK) were compared with co-located reference monitors used within the Czech national ambient air quality monitoring network. The results showed that in addition to the given reduced measurement accuracy of the sensors, the data quality depends on the early detection of defective units and changes caused by the effect of meteorological conditions (effect of air temperature and humidity on gas sensors and effect of air humidity with condensation conditions on particle counters), or by the interference of different pollutants (especially in gas sensors). Comparative measurement is necessary prior to each sensor’s field applications.

The assessment of changes in air pollution quality for 4 selected sites in Southern and Central Poland was presented in this paper. The evaluation was based on the sets of long-term data, recorded by the state air monitoring network. Concentrations of O3, PM10, SO2, NOx, and CO, were considered. The basis for the calculations were 12-year time series of hourly concentrations. Using the hourly data, the monthly averages were calculated to illustrate seasonal changes of pollutant concentrations. Linear trends were adjusted to the concentration courses with the least squares method. Long-time trends were calculated for each pollutant separately. Based on the analysis of the trend lines slopes, risks those may arise in the future were identified.

Based on the 19 year observation from 1998 to 2016 at the Tsuan Wan and Central/Western District monitoring stations in Hong Kong, the aim of this paper was to assess the wet deposition pathway of Benzo(a)pyrene (BaP) on a large time-scale. In order to achieve this goal, multi-fractal detrended cross-correlation analysis (MF-DCCA) was used to characterize the long-term cross-correlations behaviors and multi-fractal temporal scaling properties between BaP (or PM2.5) and precipitation. The results showed that the relationships between BaP and precipitation (or PM2.5) displayed long-term cross-correlation at the time-scale ranging from one month to one year; no cross-correlation between each other was observed in longer temporal scaling regimes (greater than one year). These results correspond to the atmospheric circulation of the Asian monsoon system and are explained in detail. Similar dynamic processes of the wet deposition of BaP and PM2.5 suggested that the main removal process of atmospheric BaP was rainfall deposits of PM2.5-bound BaP. Furthermore, cross-correlations between BaP (or PM2.5) and precipitation at the long time-scale have a multi-fractal nature and long-term persistent power-law decaying behavior. The temporal evolutions of the multi-fractality were investigated by the approach of a sliding window. Based on the evolution curves of multi-fractal parameters, the wet deposition pathway of PM2.5-bound BaP is discussed. Finally, the contribution degree of wet deposition to PM2.5-bound BaP was derived from the coefficient of determination. It was demonstrated that about 45% and 60% of atmospheric BaP removal can be attributed to the wet deposition pathway of PM2.5-bound BaP for the Tsuan Wan and Central/Western District areas, respectively. The findings in this paper are of great significance for further study on the removal mechanism of atmospheric BaP in the future. The MF-DCCA method provides a novel approach to assessing the geochemical cycle dynamics of BaP.

OBJECTIVES/SPECIFIC AIMS: The objective of this project is the application of complex fusion models, which combine observed and modeled data, to areas with sparse monitoring networks with multiple chemical components is under-developed. Such models could provide improved accuracy and coverage for air quality measurement predictions, an area greatly limited by the amount of missing data. METHODS/STUDY POPULATION: This project focuses on the development of methods for improved estimation of pollutant concentrations when only sparse monitor networks are found. Sparse monitoring networks are defined as areas where fewer than three criteria air pollutants (based on EPA standards) are monitored. Particularly, a multivariate air pollutant statistical model to predict spatio-temporally resolved concentration fields for multiple pollutants simultaneously is developed and evaluated. The multivariate predictions allow monitored pollutants to inform the prediction of nonmonitored pollutants in sparse networks. RESULTS/ANTICIPATED RESULTS: Daily, ZIP code level pollutant concentration estimates will be provided for 8 pollutants across South Carolina, and goodness of fit metrics for model variants and previously established methods will be compared. DISCUSSION/SIGNIFICANCE OF IMPACT: These methods utilize only widely available data resources, meaning that the improved predictive accuracy of sparsely monitored pollutant concentrations can benefit future studies in any US area by improving estimation of health effects and saving resources needed for supplemental air pollutant monitoring campaigns. Our method for estimation attempts to improve predictive accuracy and data availability for sparsely monitored pollutants and areas.

<p>Land use regression (LUR) models estimate air pollutant concentrations for areas without air quality measurements, which provides valuable information for exposure assessment and epidemiological studies. In the present study, we developed LUR models for ambient air pollutants in Hong Kong, China, a typical high-density and high-rise city. Air quality measurements at sixteen air quality monitoring stations, operated by the Hong Kong Environmental Protection Department, were collected. Moreover, five categories of predictor variables, including population distribution, traffic emissions, land use variables, urban/building morphology, and meteorological parameters, were employed to establish the LUR models of various air pollutants. Then the spatial distribution of air pollutant concentrations at 1 km × 1 km grid cells were plotted. Taking fine particle (PM2.5) as an example, the developed LUR model explained 89% of variability of PM2.5 concentrations, with a leave-one-out-cross-validation R2 of 0.64. LUR modelling results for other air pollutants will be presented. In addition, further improvements on the development of LUR models will be discussed. This study can help to assess long-term exposures to air pollutants for high-density and high-rise urban areas like Hong Kong.</p>

Low cost pollution sensors have been widely publicized, in principle offering increased information on the distribution of air pollution and a democratization of air quality measurements to amateur users. We report a laboratory study of commonly-used electrochemical sensors and quantify a number of cross-interferences with other atmospheric chemicals, some of which become significant at typical suburban air pollution concentrations. We highlight that artefact signals from co-sampled pollutants such as CO2 can be greater than the electrochemical sensor signal generated by the measurand. We subsequently tested in ambient air, over a period of three weeks, twenty identical commercial sensor packages alongside standard measurements and report on the degree of agreement between references and sensors. We then explore potential experimental approaches to improve sensor performance, enhancing outputs from qualitative to quantitative, focusing on low cost VOC photoionization sensors. Careful signal handling, for example, was seen to improve limits of detection by one order of magnitude. The quantity, magnitude and complexity of analytical interferences that must be characterised to convert a signal into a quantitative observation, with known uncertainties, make standard individual parameter regression inappropriate. We show that one potential solution to this problem is the application of supervised machine learning approaches such as boosted regression trees and Gaussian processes emulation.

Abstract In the beginning of March 2020, cases of CoVID-19 infections began rising worldwide, reacting to which the Government of India called for nationwide lockdown initially for March 25th to April 14th, 2020 and later extended it in phases till May 31st, 2020. Due to the forced restrictions pan-India on every level, the move led to drastic drop in pollution levels. The results demonstrated that during lockdown air quality is significantly improved and the pollution levels of PM2.5, PM10, SO2 and NO2 reduced drastically during the lockdown period than the preceding year for the same time frame. A direct relationship has been established between the high level of air pollutants (PM2.5, PM10, NO2 and SO2) and CoVID-19 infections being reported in these Indian cities. The correlation indicates that the air pollutants like PM2.5, PM10, NO2 and SO2 are aggravating the number of casualties due to the CoVID-19 infections. The high-level exposure of PM2.5 over a long period is found to be significantly correlated with the mortality per unit confirmed CoVID-19 cases as compared to other air pollutant parameters like PM10, NO2 and SO2.