Ground Monitoring Network Research Articles

Estimating hourly PM2.5 concentrations using Himawari-8 AOD and a DBSCAN-modified deep learning model over the YRDUA, China

In view of the spatial and temporal gaps left by the ground monitoring network for fine particulate matter (PM2.5), remote sensing has been regarded as an effective monitoring alternative to provide data of the PM2.5 distribution at a high resolution. Satellite-derived aerosol optical depth (AOD) is a parameter that has a good correlation with the PM2.5 concentration. In this study, a modified deep learning model was established to estimate hourly PM2.5 concentrations in the Yangtze River Delta Urban Agglomeration (YRDUA) region based on 5-km AOD data from Himawari-8. The model incorporated a density-based spatial clustering of applications-with-noise (DBSCAN) cluster analysis and a deep neural network (DNN) (denoted as DBSCAN–DNN) to construct individualized DNNs for the retrieval of PM2.5. The DBSCAN algorithm was used to identify the outlying datasets by involving the relationship between AOD–NO2 and PM2.5. The cluster analysis divided the inputs into separated clusters with distinct pollution levels, which helped to build a reference for the construction of individualized DNNs. The results showed that the DBSCAN–DNN model could greatly improve the estimation accuracy of the PM2.5 concentration based on identical inputs when compared with the pure DNN model. The 10-fold cross-validation R-value was enhanced by over 30%, with the highest R-value reaching 0.94 when applied to Shanghai dataset of 2018. The root-mean-square prediction error (RMSE) was also reduced by over 30%. In addition, the model performed well in generating hourly spatial estimations, thereby showing more detailed information for the dynamic changes of the PM2.5 concentration during the day. Moreover, according to the comparisons of five regional estimation results, the model was proven to have a good applicability to deal with a differentiated data volume and data complexity. This study not only proposes a new method to achieve PM2.5 estimations with a higher accuracy and spatiotemporal resolution, but also provides a new perspective for exploiting the sophisticated correlations among large environmental datasets by introducing pre-clustering into the deep learning approach.

Multi-stage satellite-based machine learning approach to reconstruct PM2.5 concentrations across Great Britain

Background/Aim: Exposure to fine particulate matter (PM2.5) is linked to adverse health outcomes. Usually, epidemiological studies rely on PM2.5 measurements collected from fixed monitors. However, in many countries such as Great Britain the existing monitoring network provides limited spatio-temporal coverage of PM2.5, with scarce data in small towns and rural areas. Data from satellites, climate and atmospheric reanalysis models, chemical transport models, and geospatial features offer additional information that can be used to reconstruct PM2.5 concentrations, filling the gaps in the ground monitoring network. The aim of this study is to develop and apply a multi-stage satellite-based machine learning (ML) model to estimate daily PM2.5 over a 1km2 grid across Great Britain for 2003-2018. Methods: We managed, processed, and synchronised data from several resources with different formats, projections, and spatio-temporal resolutions, collecting a dataset with more than 100 billion rows. We then applied a multi-stage random forest (RF) model to obtain daily modelled PM2.5 at 1km2. Stage-1 predicts PM2.5 concentrations in monitors with PM10 only records. Stage-2 imputes satellite aerosol optical depth missing due to cloudiness and bad retrievals. Stage-3 applies the RF algorithm to estimate PM2.5 concentrations using a combined dataset from Stage-1, Stage-2, and a list of spatiotemporally synchronised predictors. Stage-4 predicts daily PM2.5 using Stage-3 model across the whole Great Britain. Results/Discussion: The RF model performed well in all stages. Stage-1 obtained an R2=0.91 (Ntree=500/mtry=4). Stage-2 and Stage-3 obtained a mean overall R2 of 0.93 (Ntree=50/mtry=20) and 0.79 (Ntree=500/mtry=30), respectively. Stage-4 reconstructed approximately 1.5 billion PM2.5 values across Great Britain. Conclusion: The modelling tools and data developed in this project provide continuous estimations of PM2.5 at surface-level across Great Britain, which can then be linked with existing health databases. This will enable an accurate estimation of health risks and impacts linked to both short- and long-term exposures.

Estimating the spatial and temporal variability of the ground-level NO2 concentration in China during 2005–2019 based on satellite remote sensing

Based on the ground-level observed NO2 concentration, satellite-observed NO2 column concentration from the Ozone Monitoring Instrument (OMI) and meteorological parameters, we comprehensively consider the seasonal and regional differences in the relationship between NO2 column concentration and measured NO2 concentration and establish a two-stage combined ground NO2 concentration estimation (TSCE-NO2) model using a support vector machine for regression (SVR) and a genetic algorithm optimized back propagation neural network (GABP). On this basis, the spatial-temporal variation in the modelled ground-level NO2 concentration over China during the period of 2005–2019 was analysed. The results show that the TSCE-NO2 model proposed in this study provides a reliable estimation of the modelled ground-level NO2 concentration over China, effectively filling the spatial and temporal gaps in China's air quality ground monitoring network (the model's correlation coefficient, R, is 0.92, the mean absolute error, MAE, is 3.62 μg/m3, the mean square percentage error, MSPE, is 0.72%, and the root-mean-square error, RMSE, is 5.93 μg/m3). The analysis results of the spatial and temporal variation indicate that (1) the perennial ground-level NO2 concentration over China is high in the eastern area and low in the western area, and the high values are mainly distributed along the northern coast, the eastern coast, the middle reaches of the Yangtze River, the middle reaches of the Yellow River, the Pearl River Delta and the Sichuan Basin. (2) The modelled ground-level NO2 concentrations over China are highest in winter, followed by those in autumn and spring, and they are lowest in summer. Before 2011, the ground-level NO2 concentration over China increased at a rate of 0.348 ± 0.132 μg/(m3∙a) but decreased at a rate of 0.312 ± 0.188 μg/(m3∙a) after 2011. (3) From 2011 to 2019, measures such as energy savings and emission reductions alleviated NO2 pollution on the premise of ensuring sustained China's GDP growth.

Retrievals of key biophysical parameters at mesoscale from the Ts/VI scatterplot domain

Earth Observation (EO) is often combined with simulated process models to gain insight on key parameters for the physical mechanisms and interactions of the Earth's system. The objective of this study is two-fold: First, it explores the combined use of EO data from the Advanced Along-Track Scanning Radiometer (AATSR) with the SimSphere land biosphere model via the “triangle” to derive latent (LE) and sensible heat (H) fluxes and soil moisture content (SMC) for diverse European ecosystems. Secondly, it investigates the influence of atmospheric correction on the “triangle”-derived retrievals. Both non-atmospherically (1P) and atmospherically corrected (2P) AATSR data products are used. The data cover selected days from 2007 to 2011 at 12 sites of a European ground monitoring network. For all the predicted parameters, the 2P product leads to improved statistical validation measures. The present study for the first time: (1) assesses the triangle technique at 1km mesoscale resolution using AATSR data; (2) investigates the effects of atmospheric correction and surface heterogeneity on prediction accuracy. The findings provide important information regarding potential operational use of the technique.

Spatio-temporal modeling of PM2.5 concentrations with missing data problem: a case study in Beijing, China

ABSTRACTOne of the major challenges in conducting epidemiological studies of air pollution and health is the difficulty of estimating the degree of exposure accurately. Fine particulate matter (PM2.5) concentrations vary in space and time, which are difficult to estimate in rural, suburban and smaller urban areas due to the sparsity of the ground monitoring network. Satellite retrieved aerosol optical depth (AOD) has been increasingly used as a proxy of ground PM2.5 observations, although it suffers from non-trivial missing data problems. To address these issues, we developed a multi-stage statistical model in which daily PM2.5 concentrations can be obtained with complete spatial coverage. The model consists of three stages – an inverse probability weighting scheme to correct non-random missing patterns of AOD values, a spatio-temporal linear mixed effect model to account for the spatially and temporally varying PM2.5-AOD relationships, and a gap-filling model based on the integrated nested Laplace approximation-stochastic partial differential equations (INLA-SPDE). Good model performance was achieved from out-of-sample validation as shown in R2 of 0.93 and root mean square error of 9.64 μg/m3. The results indicated that the multi-stage PM2.5 prediction model proposed in the present study yielded highly accurate predictions, while gaining computational efficiency from the INLA-SPDE.

Two-stage model for estimating the spatiotemporal distribution of hourly PM1.0 concentrations over central and east China

Widespread and severe PM1.0 (particulate matter ≤1.0 μm) pollution in China has a significant negative influence on human health. However, knowledge of the regional spatiotemporal distribution of PM1.0 has been hindered by sparsely distributed PM1.0 concentration data. In this work, a two-stage model (linear mixed effect–bagged tree model) was proposed for estimating hourly PM1.0 pollution levels from July 2015 to June 2017 over central and east China by using Himawari-8 aerosol products and coincident geographic data, meteorology, and site-based PM1.0 concentrations from ground monitoring network. The cross-validation for the developed model displayed R2 and mean absolute error value of 0.80 and 9.3 μg/m3, respectively. Validation demonstrated that the model accurately estimated hourly PM1.0 concentrations with high R2 of 0.63–0.85 and low bias of 8.7–10.1 μg/m3. The estimated PM1.0 concentrations on daily scale showed peaks with PM1.0 of 36.9 ± 8.4 μg/m3 at rush hours during daytime. Seasonal distribution displayed that summer was cleanest with an average PM1.0 of 20.9 ± 6.8 μg/m3 and winter was the most polluted season with an average PM1.0 of 45.6 ± 16.8 μg/m3. These results indicated that the proposed satellite-based model can estimate reliable spatial distribution of PM1.0 concentrations over a large-scale region.

Initial analysis of the BDS satellite autonomous integrity monitoring capability

The integrity of Global Navigation Satellite System (GNSS) has the capability of sending a timely alarm to users when the GNSS is out of use, and it is one of the significant parameters of GNSS. Compared with ground-based integrity monitoring, integrity monitoring within the satellite constellation itself could lessen the effects of possible failures in the signal propagation path and ground systems and shorten the alarm time, which would quickly alert users of the system failure. Developing satellite autonomous integrity monitoring (SAIM) has been one of the main objectives of the upgrade plan for future GNSS such as GPS III, Galileo and the third generation of Beidou satellite navigation system (BDS3). To test the new technology, BDS3 experimental satellites are equipped with SAIM payload which is independently developed by China and it is the first satellite navigation system that applies SAIM method to monitor the navigation signals integrity. SAIM monitors signal quality and clock and frequency stability. We briefly introduce the payload design of SAIM on BDS3 experimental satellites, and evaluate in detail the monitoring measurements stability for two new civil signals of BDS3 open service, B1C_pilot signal, and B2a_pilot signal. By artificially setting signal faults, we analyze the time-to-alert performance of SAIM under its two kinds of alarming modes. The results show that the onboard observation stabilities of signal power, pseudorange, code?carrier bias, and code?carrier divergence are better than 0.12 dB, 0.1?ns, 0.14?ns, and 3.5?mm/s, respectively, which were reliable at 1-year interval. The onboard satellite clock step monitoring precision is approximately 0.5?ns. Once the abnormal signal is detected, the BDS SAIM may provide a warning within 6?s. With the characteristics of overcoming the ground monitoring network limitation and fast alarming ability, SAIM is expected to be an effective method to enhance the BDS3 system integrity monitoring capability.

Exploring the effects of crop residue burning on local haze pollution in Northeast China using ground and satellite data

In this study, we used ground-measured air pollutants and various remote sensing and meteorological datasets to explore the possible causes of the severe particulate matter (PM) pollution episodes of October and November 2015 in Northeast China. The three pollution episodes in different regions were elaborately characterized by analyzing the time variation of PM2.5 and other pollutants, e.g., CO, NO2, and SO2. The results from MODerate-resolution Imaging Spectrometer (MODIS) revealed that during the study period, dramatic crop residue burning occurred in Northeast China. Combining its time variation and spatial distribution with air pollutants, the crop residue burning was inferred to have a direct influence on the first and second pollution episodes, especially in Heilongjiang Province. Since a cold wave swept Northeast China beginning on 4th November, the PM pollutants were assumed to be transported southward. This assumption was further verified by the backward and forward trajectory models. Additionally, the suddenly increased relative humidity beginning on 7th November and the week of no wind after the cold wave also greatly contributed to the most severe pollution of Liaoning and Jilin on approximately 8th November. Lastly, we compared air quality indexes and pollutants from remote sensing with ground-measured datasets; the results indicated that there were certain correlations and spatial consistency between the two types of datasets, except for UVAI, which is meaningful to the area without an effective ground monitoring network.

Assessment of GPM and SM2RAIN-ASCAT rainfall products over complex terrain in southern Italy

The assessment of precipitation over land is extremely important for a number of scientific purposes related to the mitigation of natural hazards, climate modelling and prediction, famine and disease monitoring, to cite a few. Due to the difficulties and the cost to maintain ground monitoring networks, i.e., raingauges and meteorological radars, remote sensing is receiving more and more attention in the recent decade(s). However, the accuracy of satellite observations of rainfall should be assessed with ground information as it is affected by a number of factors (topography, vegetation density, land-sea interface).Calabria is a peninsular region in southern Italy characterized by complex topography, dense vegetation and a narrow North-South elongated shape, thus being a very challenging place for rainfall retrieval from remote sensing. In this study, we built a high-quality rainfall datasets from raingauges and meteorological radars for testing three remotely sensed rainfall products: 1) the Integrated Multi-satellitE Retrievals for Global Precipitation Measurement product (IMERG), 2) the SM2RASC product obtained from the application of SM2RAIN (Soil Moisture TO RAIN) algorithm to the Advanced SCATterometer (ASCAT) derived satellite soil moisture data, and 3) a product derived from a simple combination of IMERG and SM2RASC. The assessment of the products is carried out at different rainfall time accumulation (e.g., from 0.5 to 24 h) for a ~2-year period from 10th March 2015, to 31st December 2016.Results show that IMERG has good performance at time resolutions higher than 6 h. At daily time scale, IMERG and SM2RASC show similar results with median correlations, R, ~0.60, and root mean square error, RMSE, ~7.6 mm/day (BIAS is −0.85 and +0.51 mm/day, respectively). The combined product outperforms the parent products (median R > 0.70, RMSE<6.5 mm/day, BIAS -0.07 mm/day). Among the different factors affecting products quality, topographic complexity seems to play the more relevant role, particularly for SM2RASC but also for IMERG. Overall, this study shows that the investigated satellite-based products agree reasonably well with observations notwithstanding the challenging features of the region, and the combination of IMERG and SM2RASC provides a way to overcome their limitations and to produce a higher quality satellite rainfall product.

Time synchronization of new-generation BDS satellites using inter-satellite link measurements

Autonomous satellite navigation is based on the ability of a Global Navigation Satellite System (GNSS), such as Beidou, to estimate orbits and clock parameters onboard satellites using Inter-Satellite Link (ISL) measurements instead of tracking data from a ground monitoring network. This paper focuses on the time synchronization of new-generation Beidou Navigation Satellite System (BDS) satellites equipped with an ISL payload. Two modes of Ka-band ISL measurements, Time Division Multiple Access (TDMA) mode and the continuous link mode, were used onboard these BDS satellites. Using a mathematical formulation for each measurement mode along with a derivation of the satellite clock offsets, geometric ranges from the dual one-way measurements were introduced. Then, pseudoranges and clock offsets were evaluated for the new-generation BDS satellites. The evaluation shows that the ranging accuracies of TDMA ISL and the continuous link are approximately 4 cm and 1 cm (root mean square, RMS), respectively. Both lead to ISL clock offset residuals of less than 0.3 ns (RMS). For further validation, time synchronization between these satellites to a ground control station keeping the systematic time in BDT was conducted using L-band Two-way Satellite Time Frequency Transfer (TWSTFT). System errors in the ISL measurements were calibrated by comparing the derived clock offsets with the TWSTFT. The standard deviations of the estimated ISL system errors are less than 0.3 ns, and the calibrated ISL clock parameters are consistent with that of the L-band TWSTFT. For the regional BDS network, the addition of ISL measurements for medium orbit (MEO) BDS satellites increased the clock tracking coverage by more than 40% for each orbital revolution. As a result, the clock predicting error for the satellite M1S was improved from 3.59 to 0.86 ns (RMS), and the predicting error of the satellite M2S was improved from 1.94 to 0.57 ns (RMS), which is a significant improvement by a factor of 3–4.

Data Integration Model for Air Quality: A Hierarchical Approach to the Global Estimation of Exposures to Ambient Air Pollution

SummaryAir pollution is a major risk factor for global health, with 3 million deaths annually being attributed to fine particulate matter ambient pollution (PM2.5). The primary source of information for estimating population exposures to air pollution has been measurements from ground monitoring networks but, although coverage is increasing, regions remain in which monitoring is limited. The data integration model for air quality supplements ground monitoring data with information from other sources, such as satellite retrievals of aerosol optical depth and chemical transport models. Set within a Bayesian hierarchical modelling framework, the model allows spatially varying relationships between ground measurements and other factors that estimate air quality. The model is used to estimate exposures, together with associated measures of uncertainty, on a high resolution grid covering the entire world from which it is estimated that 92% of the world's population reside in areas exceeding the World Health Organization's air quality guidelines.

Estimating adult mortality attributable to PM2.5 exposure in China with assimilated PM2.5 concentrations based on a ground monitoring network

Estimates of mortality attributable to air pollution in China showed large differences among various studies, mainly arising from differences in exposure assessments and choice of the concentration-response function. The Chinese national monitoring network recently has included direct measurements of PM2.5 (particulates with aerodynamic diameter≤2.5μm), providing a potentially more reliable exposure assessment. We estimated adult premature mortalities due to PM2.5 across China in 2013 and mortality benefits for scenarios in which China meets the World Health Organization (WHO) Air Quality Guidelines (AQG) and three interim targets (ITs) for PM2.5. Attributable adult mortalities were estimated with assimilated spatial PM2.5 concentrations across China based on direct PM2.5 measurements from 506 PM2.5 monitoring sites and a regional air quality model, and using the integrated exposure-response model. Our results show that in China, 83% of the population lived in areas where PM2.5 concentrations exceeded the Chinese Ambient Air Quality Standard of 35μgm−3. Premature mortalities attributed to PM2.5 nationwide were 1.37 million in total, and 0.69, 0.38, 0.13, and 0.17 million for stroke, ischemic heart disease, lung cancer, and chronic obstructive pulmonary disease, respectively. High population density areas exhibited the highest health risks attributed to air pollution. The mortality benefits will be 23%, 39%, 66%, and 83% of the total present premature mortalities (1.37 million mortalities) when PM2.5 concentrations in China meet the WHO IT-1, IT-2, IT-3, and AQG, respectively. Our study shows that integrating PM2.5 concentrations based on the national monitoring network with the regional air quality model provides an advanced exposure estimate method with potentials to further improve the accuracy for mortality estimate; much higher health benefits could be achieved if China adopted more stringent WHO guidelines for PM2.5.

Satellite and Ground Observations of Severe Air Pollution Episodes in the Winter of 2013 in Beijing, China

ABSTRACTBeginning in early January 2013, Beijing experienced multiple prolonged and severe smog events that were characterized by very high levels of PM2.5, with peak daily PM2.5 over 400 µg m–3. With PM2.5 concentration contours created from ground observations and satellite remote sensing data, we describe the spatial and temporal characteristics of these episodes and further investigated the factors that contributed to these episodes. Our results indicated that these smog episodes affected a much larger geographic region, far beyond Beijing metropolitan area, corresponding to a total area of ~550,000 km2 and ~180 million people. The extremely cold weather in December 2012 and regional pollution transport were likely the main causes of these severe PM pollutions. In addition to aggressive emission control measures for Beijing, coordinated regional policy must be put in place to achieve more blue-sky days. Although the configuration of the current ground monitoring network may be sufficient to record PM2.5 levels in urban centers, these monitors alone cannot fully characterize the spatial pattern and track the transport of air pollution on a regional scale. Satellite remote sensing data can provide valuable information to fill the gaps left by ground monitors to create a more comprehensive picture of PM2.5.

The impact of observing characteristics on the ability to predict ozone under varying polluted photochemical regimes

Abstract. We conduct analyses to assess how characteristics of observations of ozone and its precursors affect air quality forecasting and research. To carry out this investigation, we use a photochemical box model and its adjoint integrated with a Lagrangian 4D-variational data assimilation system. Using this framework in conjunction with pseudo-observations, we perform an ozone precursor source inversion and estimate surface emissions. We then assess the resulting improvement in ozone air quality prediction. We use an analytical model to conduct uncertainty analyses. Using this analytical tool, we address some key questions regarding how the characteristics of observations affect ozone precursor emission inversion and in turn ozone prediction. These questions include what the effect is of choosing which species to observe, of varying amounts of observation noise, of changing the observing frequency and the observation time during the diurnal cycle, and of how these different scenarios interact with different photochemical regimes. In our investigation we use three observed species scenarios: CO and NO2; ozone, CO, and NO2; and HCHO, CO and NO2. The photochemical model was set up to simulate a range of summertime polluted environments spanning NOx-(NO and NO2)-limited to volatile organic compound (VOC)-limited conditions. We find that as the photochemical regime changes, here is a variation in the relative importance of trace gas observations to be able to constrain emission estimates and to improve the subsequent ozone forecasts. For example, adding ozone observations to an NO2 and CO observing system is found to decrease ozone prediction error under NOx- and VOC-limited regimes, and complementing the NO2 and CO system with HCHO observations would improve ozone prediction in the transitional regime and under VOC-limited conditions. We found that scenarios observing ozone and HCHO with a relative observing noise of lower than 33 % were able to achieve ozone prediction errors of lower than 5 ppbv (parts per billion by volume). Further, only observing intervals of 3 h or shorter were able to consistently achieve ozone prediction errors of 5 ppbv or lower across all photochemical regimes. Making observations closer to the prediction period and either in the morning or afternoon rush hour periods made greater improvements for ozone prediction: 0.2–0.3 ppbv for the morning rush hour and from 0.3 to 0.8 ppbv for the afternoon compared to only 0–0.1 ppbv for other times of the day. Finally, we made two complementary analyses that show that our conclusions are insensitive to the assumed diurnal emission cycle and to the choice of which VOC species emission to estimate using our framework. These questions will address how different types of observing platform, e.g. geostationary satellites or ground monitoring networks, could support future air quality research and forecasting.

Estimate Mortality Attributable To PM2.5 Exposure In China With Assimilated PM2.5 Concentrations Based On Ground Monitoring Network And A Regional Air Quality Model

Background. Significant progresses have been made to estimate the health burden attributable to exposure to ambient air pollutants, such as retrieval surface PM2.5 concentration from satellite observed aerosol optical depth. While in China, surface monitoring network of air pollutants has been expanding drastically. Assimilating the concentrations of ground monitoring network with regional air quality model could provide surface concentration of PM2.5 with much higher spatial and temporal resolution. Methods. PM2.5 concentrations in 2013 were estimated with the Nested Air Quality Prediction Modeling System (NAQPMS) using measured concentrations of air pollutants from more than 300 monitoring stations, data assimilation method based on ensemble Kalman filter was used to improve surface PM2.5 simulation. The provincial permanent population and age-specific mortalities were obtained from statistical year book in China at both central and provincial levels. Attributable mortalities including ischemic heart disease (IHD), cerebrovascular disease (stroke), chronic obstructive pulmonary disease (COPD), and lung cancer (LC) in adults were estimated using the integrated exposure-response (IER) model. Results. In 2013 in China, exposure to PM2.5 was associated with 0.69, 0.35, 0.14 and 0.14 million of premature deaths from stroke, IHD, LC and COPD disease. All together, about 1.33 million premature deaths were attributed to PM2.5 exposure. Northern China and Yangtze River Delta have the highest mortality attributable to PM2.5. Conclusions. Assimilation results of PM2.5 concentration from monitoring network with regional air quality model has been successfully used to estimate mortality attributable to PM2.5, the results are comparable with that of the Global Burden of Disease Study 2010.

Satellite-Based Spatiotemporal Trends in PM2.5 Concentrations: China, 2004-2013.

BackgroundThree decades of rapid economic development is causing severe and widespread PM2.5 (particulate matter ≤ 2.5 μm) pollution in China. However, research on the health impacts of PM2.5 exposure has been hindered by limited historical PM2.5 concentration data.ObjectivesWe estimated ambient PM2.5 concentrations from 2004 to 2013 in China at 0.1° resolution using the most recent satellite data and evaluated model performance with available ground observations.MethodsWe developed a two-stage spatial statistical model using the Moderate Resolution Imaging Spectroradiometer (MODIS) Collection 6 aerosol optical depth (AOD) and assimilated meteorology, land use data, and PM2.5 concentrations from China’s recently established ground monitoring network. An inverse variance weighting (IVW) approach was developed to combine MODIS Dark Target and Deep Blue AOD to optimize data coverage. We evaluated model-predicted PM2.5 concentrations from 2004 to early 2014 using ground observations.ResultsThe overall model cross-validation R2 and relative prediction error were 0.79 and 35.6%, respectively. Validation beyond the model year (2013) indicated that it accurately predicted PM2.5 concentrations with little bias at the monthly (R2 = 0.73, regression slope = 0.91) and seasonal (R2 = 0.79, regression slope = 0.92) levels. Seasonal variations revealed that winter was the most polluted season and that summer was the cleanest season. Analysis of predicted PM2.5 levels showed a mean annual increase of 1.97 μg/m3 between 2004 and 2007 and a decrease of 0.46 μg/m3 between 2008 and 2013.ConclusionsOur satellite-driven model can provide reliable historical PM2.5 estimates in China at a resolution comparable to those used in epidemiologic studies on the health effects of long-term PM2.5 exposure in North America. This data source can potentially advance research on PM2.5 health effects in China.CitationMa Z, Hu X, Sayer AM, Levy R, Zhang Q, Xue Y, Tong S, Bi J, Huang L, Liu Y. 2016. Satellite-based spatiotemporal trends in PM2.5 concentrations: China, 2004–2013. Environ Health Perspect 124:184–192; http://dx.doi.org/10.1289/ehp.1409481

The view from afar: satellite-derived estimates of global PM2.5.

More than 3 million people died prematurely in 2010 due to ambient exposure to fine particulate matter (PM2.5), according to estimates from the Global Burden of Disease Study.1 Although air pollution measurements taken from ground-level monitors can help inform such estimates, a paucity of monitoring stations outside of North America and Western Europe make it difficult to compare levels and trends in PM2.5 and their health effects around the world.2 Fortunately, satellite data provide a way of filling in data gaps for areas with no ground-based monitoring. In this issue of EHP, a team of researchers report their satellite-derived estimates of global exposure trends to PM2.5 over 15 years.3 “We found notable trends of increasing PM2.5 in South and East Asia, where billions of people live. Meanwhile, parts of North America are getting cleaner,” says study author Randall Martin, an atmospheric scientist at Dalhousie University in Halifax, Nova Scotia. Satellite-derived estimates show PM2.5 pollution improving in some regions (top) and worsening in others (bottom) over a 15-year period. Satellite sensors don’t measure PM2.5 directly. Instead, they assess how particles in the air, including PM2.5, scatter sunlight as it passes through the atmosphere. “In a sense, the satellites we use are little more than extremely well calibrated cameras that take pictures of the earth below. When aerosol particles are present, these pictures begin to look a little hazy,” explains first author Aaron van Donkelaar, also an atmospheric scientist at Dalhousie University. The extent to which aerosols scatter the light is called the aerosol optical depth (AOD). The researchers used AOD data from the National Aeronautics and Space Administration to estimate ground-level PM2.5 at a spatial resolution of approximately 10 km × 10 km. Although some regions experienced a decrease in PM2.5 over the period 1998–2012, the global population-weighted average increased by an estimated 2.1% per year. Rising levels of air pollution in developing regions in South and East Asia largely drove the upward trend.3 After adjusting for population changes, the researchers estimated that the proportion of people in South and East Asia exposed to PM2.5 at levels exceeding the World Health Organization (WHO) interim target of 35 µg/m3 rose from 51% in 1998–2000 to 70% in 2010–2012. In contrast, the proportion of North Americans exposed to PM2.5 at levels above the WHO air quality guideline of 10 µg/m3 fell from 62% in 1998–2000 to 19% in 2010–2012.4 Where ground-level PM2.5 data were available, the researchers found a significant association with satellite-based estimates, though satellite-derived PM2.5 estimates tended to be slightly lower than ground-level readings.3 “Satellite-based estimates reported here will enable researchers to design and conduct large epidemiological studies in low- and middle-income countries that lack the extensive ground monitoring networks found in higher income countries,” says Aaron Cohen, an epidemiologist at the Health Effects Institute in Boston. He was not involved in the current study. As satellite-derived PM2.5 estimates have become available, groups such as the Global Burden of Disease Study1 and the WHO5 have begun to use them as the basis for estimates of the global burden of disease. The current study builds upon a previous analysis by these authors, which estimated global PM2.5 levels from 2001 through 2006.6 Collecting data over an even longer period will allow researchers to talk with more confidence about trends in air pollution levels and link those trends to changes in health burden, according to Cohen. Researchers and policy makers can also track future progress as developing countries such as China begin to implement more stringent environmental regulations, he says. Although the data set of 10 km × 10 km resolution may be useful for evaluating large global trends, it is not ideal for estimating associations between air pollution and health outcomes in individuals because there is too much heterogeneity in PM2.5 concentrations within such an area at any given point in time. Some air pollution estimates using satellite data at a resolution of 2.5 km × 2.5 km have been made in U.S. cities including Cleveland, Ohio, to assess more localized health risks from air pollution exposure.7,8 “In the future, we hope to infer global PM2.5 data at a higher spatial resolution and to investigate other air pollution compounds, such as nitrogen dioxide, using satellite data,” says van Donkelaar.

Meteorological influence on predicting surface SO2 concentration from satellite remote sensing in Shanghai, China

In this study, we explored the potential applications of the Ozone Monitoring Instrument (OMI) satellite sensor in air pollution research. The OMI planetary boundary layer sulfur dioxide (SO2_PBL) column density and daily average surface SO2 concentration of Shanghai from 2004 to 2012 were analyzed. After several consecutive years of increase, the surface SO2 concentration finally declined in 2007. It was higher in winter than in other seasons. The coefficient between daily average surface SO2 concentration and SO2_PBL was only 0.316. But SO2_PBL was found to be a highly significant predictor of the surface SO2 concentration using the simple regression model. Five meteorological factors were considered in this study, among them, temperature, dew point, relative humidity, and wind speed were negatively correlated with surface SO2 concentration, while pressure was positively correlated. Furthermore, it was found that dew point was a more effective predictor than temperature. When these meteorological factors were used in multiple regression, the determination coefficient reached 0.379. The relationship of the surface SO2 concentration and meteorological factors was seasonally dependent. In summer and autumn, the regression model performed better than in spring and winter. The surface SO2 concentration predicting method proposed in this study can be easily adapted for other regions, especially most useful for those having no operational air pollution forecasting services or having sparse ground monitoring networks.

Estimating ground-level PM2.5 concentrations in the Southeastern United States using MAIAC AOD retrievals and a two-stage model

Previous studies showed that fine particulate matter (PM2.5, particles smaller than 2.5μm in aerodynamic diameter) is associated with various health outcomes. Ground in situ measurements of PM2.5 concentrations are considered to be the gold standard, but are time-consuming and costly. Satellite-retrieved aerosol optical depth (AOD) products have the potential to supplement the ground monitoring networks to provide spatiotemporally-resolved PM2.5 exposure estimates. However, the coarse resolutions (e.g., 10km) of the satellite AOD products used in previous studies make it very difficult to estimate urban-scale PM2.5 characteristics that are crucial to population-based PM2.5 health effects research. In this paper, a new aerosol product with 1km spatial resolution derived by the Multi-Angle Implementation of Atmospheric Correction (MAIAC) algorithm was examined using a two-stage spatial statistical model with meteorological fields (e.g., wind speed) and land use parameters (e.g., forest cover, road length, elevation, and point emissions) as ancillary variables to estimate daily mean PM2.5 concentrations. The study area is the southeastern U.S., and data for 2003 were collected from various sources. A cross validation approach was implemented for model validation. We obtained R2 of 0.83, mean prediction error (MPE) of 1.89μg/m3, and square root of the mean squared prediction errors (RMSPE) of 2.73μg/m3 in model fitting, and R2 of 0.67, MPE of 2.54μg/m3, and RMSPE of 3.88μg/m3 in cross validation. Both model fitting and cross validation indicate a good fit between the dependent variable and predictor variables. The results showed that 1km spatial resolution MAIAC AOD can be used to estimate PM2.5 concentrations.

A Time Series Analysis of PM2.5 Concentrations in the Southeastern U.S. Using MAIAC AOD in a Two-stage Spatial Statistical Model

Background. Long term PM2.5 exposure has been associated with various health outcomes. However, ground monitoring networks leave large rural and suburban areas uncovered even in developed countries.In recent years, satellite-retrieved aerosol optical depth (AOD) has been used for PM2.5 concentration estimation due to its large spatial coverage. A limitation of the current AOD products is their coarse spatial resolution (10 – 20 km). Aims. We examined the PM2.5 predicting power of a new AOD product with 1 km spatial resolution retrieved by the Multi-Angle Implementation of Atmospheric Correction (MAIAC) algorithm. Methods. We developed a two stage spatial model with MAIAC AOD, meteorological and land use variables as covariates. The first stage linear mixed effects model accounts for the day-to-day variability of the PM2.5-AOD relationship, and the second stage geographically weighted regression model reduces the spatially varying residuals from the first stage. A 10-year model simulation was performed in the southeastern U.S. centered at the Atlanta Metro area. Results. Annual model R2 ranged from 0.52 to 0.83, and annual mean prediction error ranged from 1.97 to 2.58 ?g/m3. Model performance is significantly better than a similar model developed without MAIAC AOD. Our time series analysis results showed that PM2.5 concentration level in the study area was in a general declining trend from 2001 to 2010 with the exception of 2005, which could be attributed to higher sulfate concentrations related to increased power production during the warm season. Conclusions. MAIAC AOD is a strong predictor of ground level PM2.5 when used together with effect modifiers such as temperature and wind speed. The prediction error of our MAIAC AOD model at 1 km resolution is comparable with that of a similar MODIS AOD model at 10 km resolution.