Multi-season mobile monitoring campaign of on-road air pollution in Bengaluru, India: High-resolution mapping and estimation of quasi-emission factors.

Mobile monitoring can supplement regulatory measurements, particularly in low-income countries where stationary monitoring is sparse. Here, we report results from a ~ year-long mobile monitoring campaign of on-road concentrations of black carbon (BC), ultrafine particles (UFP), and carbon dioxide (CO2) in Bengaluru, India. The study route included 150 unique kms (average: ~22 repeat measurements per monitored road segment). After cleaning the data for known instrument artifacts and sensitivities, we generated 30 m high-resolution stable 'data only' spatial maps of BC, UFP, and CO2 for the study route. For the urban residential areas, the mean BC levels for residential roads, arterials, and highways were ~ 10, 22, and 56 µg m-3, respectively. A similar pattern (highways being characterized by highest pollution levels) was also observed for UFP and CO2. Using the data from repeat measurements, we carried out a Monte Carlo subsampling analysis to understand the minimum number of repeat measures to generate stable maps of pollution in the city. Leveraging the simultaneous nature of the measurements, we also mapped the quasi-emission factors (QEF) of the pollutants under investigation. The current study is the first multi-season mobile monitoring exercise conducted in a low or middle -income country (LMIC) urban setting that oversampled the study route and investigated the optimum number of repeat rides required to achieve representative pollution spatial patterns characterized with high precision and low bias. Finally, the results are discussed in the context of technical aspects of the campaign, limitations, and their policy relevance for our study location and for other locations. Given the day-to-day variability in the pollution levels, the presence of dynamic and unorganized sources, and active government pollution mitigation policies, multi-year mobile measurement campaigns would help test the long-term representativeness of the current results.

Urban Air-Quality Estimation Using Visual Cues and a Deep Convolutional Neural Network in Bengaluru (Bangalore), India.

Mobile monitoring provides robust measurements of air pollution. However, resource constraints often limit the number of measurements so that assessments cannot be obtained in all locations of interest. In response, surrogate measurement methodologies, such as videos and images, have been suggested. Previous studies of air pollution and images have used static images (e.g., satellite images or Google Street View images). The current study was designed to develop deep learning methodologies to infer on-road pollutant concentrations from videos acquired with dashboard cameras. Fifty hours of on-road measurements of four pollutants (black carbon, particle number concentration, PM2.5 mass concentration, carbon dioxide) in Bengaluru, India, were analyzed. The analysis of each video frame involved identifying objects and determining motion (by segmentation and optical flow). Based on these visual cues, a regression convolutional neural network (CNN) was used to deduce pollution concentrations. The findings showed that the CNN approach outperformed several other machine learning (ML) techniques and more conventional analyses (e.g., linear regression). The CO2 prediction model achieved a normalized root-mean-square error of 10-13.7% for the different train-validation division methods. The results here thus contribute to the literature by using video and the relative motion of on-screen objects rather than static images and by implementing a rapid-analysis approach enabling analysis of the video in real time. These methods can be applied to other mobile-monitoring campaigns since the only additional equipment they require is an inexpensive dashboard camera.

Application of an Ultra-Low-Cost Passive Sampler for Light-Absorbing Carbon in Mongolia.

Low-cost, long-term measures of air pollution concentrations are often needed for epidemiological studies and policy analyses of household air pollution. The Washington passive sampler (WPS), an ultra-low-cost method for measuring the long-term average levels of light-absorbing carbon (LAC) air pollution, uses digital images to measure the changes in the reflectance of a passively exposed paper filter. A prior publication on WPS reported high precision and reproducibility. Here, we deployed three methods to each of 10 households in Ulaanbaatar, Mongolia: one PurpleAir for PM2.5; two ultrasonic personal aerosol samplers (UPAS) with quartz filters for the thermal-optical analysis of elemental carbon (EC); and two WPS for LAC. We compared multiple rounds of 4-week-average measurements. The analyses calibrating the LAC to the elemental carbon measurement suggest that 1 µg of EC/m3 corresponds to 62 PI/month (R2 = 0.83). The EC-LAC calibration curve indicates an accuracy (root-mean-square error) of 3.1 µg of EC/m3, or ~21% of the average elemental carbon concentration. The RMSE values observed here for the WPS are comparable to the reported accuracy levels for other methods, including reference methods. Based on the precision and accuracy results shown here, as well as the increased simplicity of deployment, the WPS may merit further consideration for studying air quality in homes that use solid fuels.

Bias in PM2.5 measurements using collocated reference-grade and optical instruments.

Optical PM2.5 measurements are sensitive to aerosol properties that can vary with space and time. Here, we compared PM2.5 measurements from collocated reference-grade (beta attenuation monitors, BAMs) and optical instruments (two DustTrak II and two DustTrak DRX) over 6 months. We performed inter-model (two different models), intra-model (two units of the same model), and inter-type (two different device types: optical vs. reference-grade) comparisons under ambient conditions. Averaged over our study period, PM2.5 measured concentrations were 46.0 and 45.5 µg m-3 for the two DustTrak II units, 29.8 and 38.4 µg m-3 for DRX units, and 18.3 and 19.0 µg m-3 for BAMs. The normalized root square difference (NRMSD; compares PM2.5 measurements from paired instruments of the same type) was ~ 5% (DustTrak II), ~ 27% (DRX), and ~ 15% (BAM). The normalized root mean square error (NRMSE; compares PM2.5 measurements from optical instruments against a reference instrument) was ~ 165% for DustTrak II, ~ 74% after applying literature-based humidity correction and ~ 27% after applying both the humidity and BAM corrections. Although optical instruments are highly precise in their PM2.5 measurements, they tend to be strongly biased relative to reference-grade devices. We also explored two different methods to compensate for relative humidity bias and found that the results differed by ~ 50% between the two methods. This study highlights the limitations of adopting a literature-derived calibration equation and the need for conducting local model-specific calibration. Moreover, this is one of the few studies to perform an intra-model comparison of collocated reference-grade devices.