The Atmospheric Pollution and Human Health in a Chinese Megacity (APHH-Beijing) programme is an international collaborative project focusing on understanding the sources, processes and health effects of air pollution in the Beijing megacity. APHH-Beijing brings together leading China and UK research groups, state-of-the-art infrastructure and air quality models to work on four research themes: (1) sources and emissions of air pollutants; (2) atmospheric processes affecting urban air pollution; (3) air pollution exposure and health impacts; and (4) interventions and solutions. Themes 1 and 2 are closely integrated and support Theme 3, while Themes 1–3 provide scientific data for Theme 4 to develop cost-effective air pollution mitigation solutions. This paper provides an introduction to (i) the rationale of the APHH-Beijing programme and (ii) the measurement and modelling activities performed as part of it. In addition, this paper introduces the meteorology and air quality conditions during two joint intensive field campaigns – a core integration activity in APHH-Beijing. The coordinated campaigns provided observations of the atmospheric chemistry and physics at two sites: (i) the Institute of Atmospheric Physics in central Beijing and (ii) Pinggu in rural Beijing during 10 November–10 December 2016 (winter) and 21 May–22 June 2017 (summer). The campaigns were complemented by numerical modelling and automatic air quality and low-cost sensor observations in the Beijing megacity. In summary, the paper provides background information on the APHH-Beijing programme and sets the scene for more focused papers addressing specific aspects, processes and effects of air pollution in Beijing.

There is increasing concern about the health impacts of ambient Particulate Matter (PM) exposure. Traditional monitoring networks, because of their sparseness, cannot provide sufficient spatial-temporal measurements characteristic of ambient PM. Recent studies have shown portable low-cost devices (e.g., optical particle counters, OPCs) can help address this issue; however, their application under ambient conditions can be affected by high relative humidity (RH) conditions. Here, we show how, by exploiting the measured particle size distribution information rather than PM as has been suggested elsewhere, a correction can be derived which not only significantly improves sensor performance but which also retains fundamental information on particle composition. A particle size distribution–based correction algorithm, founded on $\kappa$ -Köhler theory, was developed to account for the influence of RH on sensor measurements. The application of the correction algorithm, which assumed physically reasonable $\kappa$ values, resulted in a significant improvement, with the overestimation of PM measurements reduced from a factor of ~5 before correction to 1.05 after correction. We conclude that a correction based on particle size distribution, rather than PM mass, is required to properly account for RH effects and enable low cost optical PM sensors to provide reliable ambient PM measurements.

Recent studies have shown that (three-electrode) electrochemical sensors can be utilised for air quality monitoring and exposure assessment. The long-term performance of these sensors is however, often limited by the effects of ambient meteorological parameters on the sensor baseline, in particular temperature. If electrochemical (EC) sensors are to be adopted for air quality measurement over extended periods (months), this effect must be accounted for. Recent long-term, ambient measurements of CO, NO and NO2 using EC sensors have revealed that temperature (and relative humidity (RH)) had an effect on the baseline which was more pronounced in the case of NO sensors with coefficient of determination, R2 of 0.9 when compared to CO and NO2 with R2 < 0.2. In this paper we present a correction methodology that quantifies this effect (referred to here as fitted baseline), implementing these correction on the EC measurements. We found that EC sensors corrected for baseline-temperature effect using the method describe in this paper show good agreement when compared with traditional reference instrument. The coefficient of determination R2 of 0.7–0.8 and gradient of 0.9 was observed for baseline-temperature corrected NO compared to R2 = 0.02 prior to baseline-temperature correction. Furthermore, the correction methodology was validated by comparing the temperature-baseline with proxy temperature compensating measurements obtained from the fourth electrode of a set of novel four-electrode electrochemical sensors. A good agreement (R2 = 0.9, with gradients = 0.7–1.08 for NO and 0.5 < R2 < 0.73 for CO) was observed between temperature fitted baselines and outputs from the fourth electrodes (also known non-sensing/auxiliary electrode). Meanwhile, the long-term stability (calibrated signal output) of temperature-corrected data was evaluated by comparing the change in sensor gain to meteorological parameters including temperature, relative humidity, wind speed and wind direction. The results showed that there was no statistically significant change in sensitivity (two-sided t-test, p = 0.34) of the temperature-corrected electrochemical sensor with respect to these parameters (over several months). This work demonstrates that using the baseline-temperature correction methodology described in this paper, electrochemical sensors can be used for long-term (months), quantitative measurements of air quality gases at the parts per billion volume (ppb) mixing ratio level typical of ambient conditions in the urban environment

Low cost, personal air pollution sensors may reduce exposure measurement errors in epidemiological investigations and contribute to citizen science initiatives. Here we assess the validity of a low cost personal air pollution sensor. Study participants were drawn from two ongoing epidemiological projects in Barcelona, Spain. Participants repeatedly wore the pollution sensor - which measured carbon monoxide (CO), nitric oxide (NO), and nitrogen dioxide (NO2). We also compared personal sensor measurements to those from more expensive instruments. Our personal sensors had moderate to high correlations with government monitors with averaging times of 1-h and 30-min epochs (r ~ 0.38-0.8) for NO and CO, but had low to moderate correlations with NO2 (~0.04-0.67). Correlations between the personal sensors and more expensive research instruments were higher than with the government monitors. The sensors were able to detect high and low air pollution levels in agreement with expectations (e.g., high levels on or near busy roadways and lower levels in background residential areas and parks). Our findings suggest that the low cost, personal sensors have potential to reduce exposure measurement error in epidemiological studies and provide valid data for citizen science studies.

Satellite retrieval of vertical column densities (VCDs) of tropospheric nitrogen dioxide (NO₂) is critical for NO_x pollution and impact evaluation. For regions with high aerosol loadings, the retrieval accuracy is greatly affected by whether aerosol optical effects are treated implicitly (as additional “effective” clouds) or explicitly, among other factors. Our previous POMINO algorithm explicitly accounts for aerosol effects to improve the retrieval, especially in polluted situations over China, by using aerosol information from GEOS-Chem simulations with further monthly constraints by MODIS/Aqua aerosol optical depth (AOD) data. Here we present a major algorithm update, POMINO v1.1, by constructing a monthly climatological dataset of aerosol extinction profiles, based on level 2 CALIOP/CALIPSO data over 2007–2015, to better constrain the modeled aerosol vertical profiles. We find that GEOS-Chem captures the month-to-month variation in CALIOP aerosol layer height (ALH) but with a systematic underestimate by about 300–600 m (season and location dependent), due to a too strong negative vertical gradient of extinction above 1 km. Correcting the model aerosol extinction profiles results in small changes in retrieved cloud fraction, increases in cloud-top pressure (within 2 %–6 % in most cases), and increases in tropospheric NO₂ VCD by 4 %–16 % over China on a monthly basis in 2012. The improved NO₂ VCDs (in POMINO v1.1) are more consistent with independent ground-based MAX-DOAS observations (R²=0.80, NMB = −3.4 %, for 162 pixels in 49 days) than POMINO (R²=0.80, NMB = −9.6 %), DOMINO v2 (R²=0.68, NMB = −2.1 %), and QA4ECV (R²=0.75, NMB = −22.0 %) are. Especially on haze days, R² reaches 0.76 for POMINO v1.1, much higher than that for POMINO (0.68), DOMINO v2 (0.38), and QA4ECV (0.34). Furthermore, the increase in cloud pressure likely reveals a more realistic vertical relationship between cloud and aerosol layers, with aerosols situated above the clouds in certain months instead of always below the clouds. The POMINO v1.1 algorithm is a core step towards our next public release of the data product (POMINO v2), and it will also be applied to the recently launched S5P-TROPOMI sensor.