Diurnal and seasonal variability of PM2.5 and AOD in North China plain: Comparison of MERRA-2 products and ground measurements

Atmospheric Environment 191 (2018) 70–78

Contents lists available at ScienceDirect

Atmospheric Environment journal homepage: www.elsevier.com/locate/atmosenv

Diurnal and seasonal variability of PM2.5 and AOD in North China plain: Comparison of MERRA-2 products and ground measurements

T

Zijue Songa,b, Disong Fua, Xiaoling Zhangb, Yunfei Wud, Xiangao Xiaa,c,f,∗, Jianxin Heb, Xinlei Hana,f, Renjian Zhangd, Huizheng Chee,∗∗ a

LAGEO, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing, 100029, China Chengdu University of Information Technology, Chengdu, 610225, China c Collaborative Innovation Center on Forecast and Evaluation of Meteorological Disasters, Nanjing University of Information Science & Technology, Nanjing, 210044, China d RCE-TEA, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing, 100029, China e State Key Laboratory of Severe Weather (LASW) and Institute of Atmospheric Composition, Chinese Academy of Meteorological Sciences, CMA, Beijing, 100081, China f College of Earth and Planetary Sciences, University of Chinese Academy of Sciences, Beijing, 100049, China b

A R T I C LE I N FO

A B S T R A C T

Keywords: MERRA-2 Aerosol optical depth PM2.5 North China plain

North China Plain (NCP) is one of heavily polluted regions that is characterized by a mixture of a myriad of anthropogenic and natural aerosols. A substantial spatial and temporal variations of aerosols and their compositions there poses a good testbed for the validation of model simulations. Aerosol optical depth (AOD) and PM2.5 (particulate matter with aerodynamic diameter < 2.5 μm) concentration products from the Modern Era Retrospective-Analysis for Research and Applications, version 2 (MERRA-2) are evaluated using available independent ground-based in situ and remote sensing products in the NCP. The comparison of MERRA-2 aerosol species to the observations is also performed. Although several satellite and ground-based AOD products are assimilated into the MERRA-2, MERRA-2 AOD is systematically smaller than independent sunphotometer measurements. The biases range from 0.09 (13%) in the summer to 0.17 (33%) in the spring and show little spatial dependence. Daytime AOD variations are captured by the MERRA-2, although MERRA-2 has relatively lower AOD. MERRA-2 produces lower PM2.5 concentration relative to surface measurements in all seasons except in summer. The largest bias is found in the winter (44 μgm−3). On the contrary, summer MERRA-2 PM2.5 is close to surface-measured PM2.5 (with bias of 0.4 μgm−3). MERRA-2 was unable to reproduce diurnal PM2.5 variation. Evaluation of MERRA-2 aerosol species in the winter of 2014 suggests that MERRA-2 could not keep track of dramatic day-to-day variation of aerosols and their species. Potential causes for this deficiency may include a lack of nitrate aerosols (accounting for 20% of PM2.5 concentrations during heavily polluted days). This fault cannot be remedied by assimilation of satellite AODs because they are often missing.

1. Introduction

2008). PM2.5 shows a substantial spatiotemporal variability partly due to its short lifetime in the atmosphere. Fully understanding of spatiotemporal variability of PM2.5 requires multiple approaches. Surface measurement is the fundamental requirement for monitoring PM2.5, which is often taken to be the ground truth to validate satellite retrievals and model simulations. However, surface measurements are generally taken in populated regions and thereby their spatial coverage is limited. Therefore, detection of large-scale spatial distribution of PM2.5 requires extra methods. Satellite retrievals and model simulations are two most promising methods in the production of regional to global

A large spread of haze, as a result of huge emission of particulate matter (PM) from industrial and agricultural activities and adverse weather condition in East Asia, results in significant impacts on environment, climate and human health (Li et al., 2016). More specifically, particles with an aerodynamic equivalent diameter smaller than 2.5 μm (PM2.5) are of greatest health concern because they can pass through the nose and throat and be absorbed deep inside the lung. The smaller the particles, the deeper they can penetrate the respiratory system and the more hazardous they are to breathe (Valavanidis et al.,

∗

Corresponding author. LAGEO, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing, 100029, China. Corresponding author. E-mail addresses: [email protected] (X. Xia), [email protected] (H. Che).

∗∗

https://doi.org/10.1016/j.atmosenv.2018.08.012 Received 9 May 2018; Received in revised form 2 August 2018; Accepted 6 August 2018 Available online 07 August 2018 1352-2310/ © 2018 Elsevier Ltd. All rights reserved.

Atmospheric Environment 191 (2018) 70–78

Z. Song et al.

Fig. 1. Topographic maps of the NCP overlaid by 81 MEC PM2.5 stations (blue circle) and 10 sun-photometer stations (red triangle). (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

than surface observations. One of outstanding features of MERRA-2 is that it assimilates satellite AOD, which could likely constrain the total column aerosol mass in MERRA-2. Comparison of MERRA-2 PM2.5 products against surface measurements over Africa, South America, Central and Eastern Asia showed that assimilation of satellite AOD did contribute to improvement of MERRA-2 PM2.5 reanalysis (Buchard et al., 2016 and references therein). Discrepancies between the MERRA2 and surface measurements were also revealed, for example, the lack of nitrate emissions in MERRA-2 and an underestimation of carbonaceous emissions in the Western US led to the reanalysis bias in PM2.5 (Buchard et al., 2018). The North China Plain (NCP) is one of heavily polluted regions in the world. NCP is recognized for its large-scale biomass burning of field residues in crop harvest season, occasional large spread of dust events in spring, frequent regional haze all year round because of large anthropogenic emissions (Xia et al., 2013). Therefore, the NCP poses a good testbed for the evaluation of MERRA-2 aerosol products, which becomes possible by much progress in surface observations of aerosols during recent years. A regional network of surface measurement of hourly PM2.5 concentration has been established in 2013. Sunphotometer network has be expanded and accumulated more than ten years data in some stations (Che et al., 2014; Xia et al., 2016). These observations provide independent data to compare MERRA-2 aerosol products. The objective of this study is to evaluate MERRA-2 aerosol products in the NCP. Since this region is characterized by a very complicated mixture of natural and anthropogenic emissions that poses a big challenge for the reanalysis, a thorough validation of MERRA-2 aerosol products in this heavily polluted region would indicate the way to improve aerosol reanalysis. The focus of this study is whether MERRA-2 can reproduce diurnal and seasonal variabilities of AOD and PM2.5

PM2.5 distribution. Much progress has been made on these two aspects during the last decade, which are benefited from rapid development of satellite missions. For example, the MODerate resolution Imaging Spectroradiometer (MODIS) onboard Terra (1999) and Aqua (2002) can produce global aerosol optical depth (AOD) data with a spatial resolution of 3–10 km (Levy et al., 2013). Surface and satellite remote sensing of AODs were widely used to estimate PM2.5 concentration (Wang and Christopher, 2003; Fu et al., 2018). Satellite AODs have also been assimilated into models that greatly enhances models' capability in the reproduction of aerosol distribution. One of examples is that the National Aeronautics and Space Administration (NASA) has extended the Modern-Era Retrospective Analysis for Research and Application, version 2 (MERRA-2) with an atmospheric aerosol reanalysis. MERRA-2 is based on a version of the Goddard Earth Observing System Data Assimilation System version 5 (GEOS-5) model driven by the MERRA meteorological reanalysis (Rienecker et al., 2011; Randles et al., 2018; Buchard et al., 2018). GEOS-5 is radiatively coupled to the Goddard Chemistry, Aerosol, Radiation and Transport (GOCART) aerosol module that produces five particulate species, i.e., sulfate, organic carbon (OC), black carbon (BC), mineral dust and sea salt. An important feature of the MERRA-2 is that it assimilates bias-corrected AOD from MODIS and the Advanced Very High Resolution Radiometer instruments. Additionally, Non-bias-corrected AOD from the Multiangle Imaging SpectroRadiometer over bright surfaces and AOD from Aerosol Robotic Network (AERONET) sunphotometer stations are newly assimilated in the MERRA-2. The data can be used to study the impact of aerosols on the atmospheric circulation, climate, and air quality around the world for its global and constant coverage and its distinction of aerosol species (Buchard et al., 2016, 2018). MERRA-2 has potential to provide improved estimates of AOD and PM2.5 compared to the model alone and with much greater coverage

71

Atmospheric Environment 191 (2018) 70–78

Z. Song et al.

Table 1 Description of 10 sun-photometer stations in NCP. Stations

description

Lon(°E)/Lat(°N)/Elev(m)

Days and duration

AOD

ZB SDZ GC TJ DL HM TS BJ XH XL

suburban hill rural urban urban rural mountain urban suburban hill

114.7/41.2/1394.2 117.1/40.7/286.5 115.7/39.1/30.0 117.2/39.1/5.2 121.6/38.9/97.3 117.5/37.5/12.2 117.1/36.3/1536.5 116.4/40.0/92.0 117.0/39.8/36.0 117.6/40.4/970.0

453 (2004/1/1–2005/3/28) 2954 (2004/3/17–2012/4/17) 1639 (2007/10/2–2012/3/27) 3652 (2002/4/19–2012/4/17 1824 (2007/4/24–2012/4/20) 1806 (2007/4/28–2012/4/6) 633 (2010/6/15–2012/3/8) 4962 (2001/3/7–2014/10/6) 4078 (2001/3/20–2012/5/18) 2272 (2006/2/19–2012/5/9)

0.25 0.30 0.50 0.52 0.36 0.49 0.20 0.42 0.42 0.20

± ± ± ± ± ± ± ± ± ±

0.15 0.27 0.35 0.33 0.24 0.27 0.16 0.37 0.36 0.18

the analysis. The simulated PM2.5 chemical compositions from the MERRA-2 were further compared to the ground observations, including dust, sea salt, sulfate, black and organic carbon, during a winter field campaign (24 February to 12 March 2014) at Beijing (39.97°N, 116.37°E). The PM2.5 samples were collected twice per day (07:00 to 19:00 and 19:00 to 07:00 of the next day) and then sent to the laboratory for chemical analysis. Details of the chemical analysis were referred in Ma et al. (2017) and Zhang et al. (2013).

recorded by surface measurements in NCP. Potential causes for the discrepancies and discussions on potential improvements concerning assimilation of aerosol products are finally presented. 2. Region, data and methodology The NCP (32 °N -41.5 °N, 114 °E -123 E°) is the second largest plain in China, which is bordered to the north by the Yanshan Mountains, to the west by the Taihang Mountains, to the south by the Tai and Ta-pieh Mountains, and to the east by the Yellow Sea (Fig. 1).

2.2. MERRA-2 data 2.1. Ground data Following the success of the original MERRA reanalysis, the NASA Global Modeling and Assimilation Office (GMAO) released the new atmospheric reanalysis product, i.e., MERRA-2, in 2017 (Randles et al., 2018). The MERRA-2 is mainly based on the NASA GMAO Earth system model version 5 (GEOS 5), a weather and climate capable model. The GEOS-5 system is composed of atmospheric circulation and composition, oceanic and land components. The atmospheric data assimilation subsystem of GEOS-5 builds upon the Grid-point Statistical Interpolation (GSI) algorithm that is able to assimilate all the in-situ and remotely-sensed atmospheric data (Buchard et al., 2016). MERRA-2 provides not only reanalysis of meteorological field but also several parallel re-analyses of other components of the Earth system, such as ocean, land and atmospheric components (Randles et al., 2018). Of particular interest for this study is the MERRA-2 analyzed aerosol fields. This is the first multi-decadal reanalysis within which meteorological and aerosol observations are jointly assimilated into a global assimilation system (Randles et al., 2018). GEOS-5 is radiatively coupled to the Goddard Chemistry, Aerosol, Radiation and Transport (GOCART) aerosol module to provide a companion gridded aerosol data set. Aerosol species are assumed to be external mixtures that do not interact with each other. Both dust and sea salt emissions depend on surface wind speed, while sulfate and carbonaceous species have emissions principally from fossil fuel combustion, biomass burning, and biofuel consumption, with additional biogenic sources of organic carbon. BC, organic carbon (OC) and sulfate are prescribed from standard inventories (Randles et al., 2016). The Emission Database for Global Atmospheric Research 4.1 (EDGAR) inventory is used to prescribe emissions of SO2 from anthropogenic sources. Differences in the injection profiles of emissions sources from energy and non-energy sectors are accounted for (Buchard et al., 2014). Biomass burning emissions are from the NASA Quick Fire Emission Dataset (QFED) version 2.1 (Randles et al., 2018). The MERRA-2 experiment is performed at 0.5°×0.625° latitude by longitude with 72 vertical layers between the surface and about 80 km. GEOS-5 is run in replay-mode using six-hourly atmospheric analyses from MERRA to update the meteorological fields, with an aerosol assimilation performed every 3 h. Optical properties such as the mass extinction efficiency, single scattering albedo, and asymmetry parameter, associated with five aerosol species are primarily from the

Ground remote sensing AOD data at 3 Aerosol Robotic Network (AERONET) and 7 Chinese Aerosol Research Network (CARSNET) stations with at least 2-year measurements (Table 1) in the NCP are used to compare MERRA-2 AOD product. The AERONET program provides a long-term, continuous and readily available AOD database for aerosol study (Holben et al., 2001). The CARSNET is a regional aerosol research network that consists of > 50 sunphotometer stations across China (Che et al., 2015). The CE-318 sunphotometer used in AERONET and CARSNET measures direct solar spectral radiation at wavelengths from 340 nm to 1020 nm, which are used to derive AOD with accuracy of 0.01–0.02 (Eck et al., 1999). AOD at 550 nm was interpolated from AOD at 440, 675 and 870 nm to compare with MERRA-2 AOD product. We used the cloud-screened and quality assured sunphotometer Level 2.0 AOD data in the comparison. Surface measurements of hourly PM2.5 concentration at 81 stations in the NCP are available from the data archive of Ministry of Environmental Protection of China (MEPC) (http://datacenter.mep. gov.cn/) (Fig. 1). The PM2.5 (MEPC PM2.5 hereafter) are measured using the tapered element oscillating microbalance method (TEOM) and/or the beta absorption method. The instrumental operation, maintenance, data assurance and quality control are properly conducted according to the most revisions of China Environmental Protection Standards. Hourly PM2.5 data from May 2014 and June 2017 are used in the comparison. Surface PM2.5 values are averaged if there are more than one stations within one MERRA-2 grid. The comparison is performed at 44 MERRA-2 grids (0.5°×0.625°) where at least one station is available. Two further steps are adopted for the quality control of MEPC PM2.5 data. First, there are some cases in which PM2.5 is constant within a few continuous hours or even from one day to another. These measurements are taken to be unacceptable if they are constant within at least 5 continuous hours since they likely do not reflect the normal fluctuations. Second, a spatial consistency algorithm suggested by Provencal et al. (2017a) is applied to assure reliability of the observed and MERRA-2 data. If the difference between log-simulated and log-observed concentrations at all locations on a given hour does not fall within the 2 standard deviation limits, these observed and simulated PM2.5 values are excluded. Totally, 4.1% of data points are excluded in 72

Atmospheric Environment 191 (2018) 70–78

Z. Song et al.

Optical Properties of Aerosols and Clouds (OPAC) data set (Hess et al., 1998). AOD is calculated by integration of the aerosol extinction profile provided by MERRA-2 from the surface to 80 km. MERRA-2 total PM2.5 concentrations are calculated from five species as follows. PM2.5 = [DUST2.5] + [SS2.5] + [BC] + 1.6×[OC] + 1.375×[SO4](1) Where [DUST2.5], [SS2.5], [BC], [OC], and [SO4] are respectively the GOCART concentrations of dust, sea-salt, BC, OC and sulfate ion, all with diameter not larger than 2.5 μm. The sulfate concentration is assumed to be primarily present in the form of neutralized ammonium sulfate, which is calculated from the mass of sulfate ion (provided by the MERRA-2) multiplied by a factor of 1.375. MERRA-2 simulates OC aerosol that is used to estimate organic matter (OM), in which OC aerosol is multiplied by a factor (molecular weight per carbon weight ratio) that accounts for contributions from other elements associated with OM. This factor varies spatially and temporally with values between 1.2 and 2.6 (Malm et al., 2011). A constant value of 1.6 is applied here because surface measurements has indicated an averaged ratio of 1.59 ± 0.18 in PM2.5 over China (Zhang et al., 2013). Dust and sea salt with a particle diameter not larger than 2.5 μm are also accounted for in the calculation of PM2.5 concentrations. Note that the nitrate particles predominantly emitted by anthropogenic activities are not included in the equation.

Fig. 2. Density scatter-plot between MERRA-2 and sun-photometer AOD products at 10 stations. The color bar represents the density of points ranging from a few percent (blue) to 100 percent (yellow) of points. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

2.3. Methodology Mean bias and its standard deviation of PM2.5 are calculated from collocated MEPC and MERRA-2 PM2.5 concentrations (MEPC – MERRA2). Correlation coefficient (R) between them is calculated. Furthermore, the proportion of simulated data which falls within a factor of 2 of observed data (FAC2) is calculated. The model performance is thought to be reasonably good if FAC2 ≥ 0.50. FAC2 is calculated as follows. FAC2 = (MERRA-2 PM2.5)/(MEPC PM2.5)

(2)

3. Results 3.1. AOD comparison

Fig. 3. Comparison of monthly MERRA-2 AOD (grey) against sun-photometer AOD (red). Shading indicates monthly standard deviation for these two datasets. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Although MERRA-2 has assimilated bias-corrected AOD from MODIS onboard Terra and Aqua satellites to constrain the total column aerosol mass (Buchard et al., 2016), it is still necessary to validate MERRA-2 AOD against independent AOD products. Since CARSNET AOD data here are not assimilated into MERRA-2, so we compare MERRA-2 AOD products against these ground measurements (Fig. 2). Though MERRA-2 AOD generally compares well to CARSNET AOD (with R of 0.77), the deficiency of MERRA-2 AOD product is remarkable. MERRA-2 produces lower AOD and this bias is more outstanding for high CARSNET AODs, for example, MERRA-2 AODs are nearly always lower than sunphotometer AODs when AOD > 1.0. This pattern was not in existence in the comparison of hourly sunphotometer AOD against MODIS collection 5.1 AOD product within 50 km of the station (similar resolution as MERRA-2) (Xia et al., 2013). Note that mean AOD at XH (suburban) is close to that at BJ (urban), although these two sites are separated by 70 km (Table 1), therefore, MERRA-2 AOD bias should not be attributable to inconsistence in AOD spatial representativeness between MERRA-2 and station data. Fig. 3 presents the seasonal cycle of AOD from reanalysis and measurements. Both products exhibit a similar seasonal cycle, i.e., lower values during the winter and higher values during the summer. However, MERRA-2 AOD is nearly consistently smaller than CARSNET AOD and the biases are 0.17, 0.09, 0.11 and 0.14 from spring to winter. Site-specific comparison of seasonal mean AOD (and standard deviation) is presented in Fig. 4. The spatial variation of site AOD is

reproduced well by MERRA-2, i.e., larger AODs in eastern (TJ) and southern NCP (HM, GC) and relatively smaller AODs in northern NCP. Seasonal MERRA-2 AODs are nearly always smaller than sunphotometer AODs except at Mountain Tai (TS). Sunphotometer was installed at the top of the mountain (1536 m altitude, Fig. 1), therefore, sunphotometer AOD is definitely smaller relative to surrounding plain areas that are covered by MERRA-2. Similar negative bias of MERRA-2 AOD in urban (BJ, TJ), suburban (XH), rural (HM, GC) and background stations (SDZ, XL) indicates potential problem in the assimilation of AOD. MERRA-2 assimilated non-bias corrected MISR AODs, while MISR AOD is substantially smaller than AERONET in NCP, especially under situations with higher AOD (Liu et al., 2010). Besides, remote sensing of AOD is often not available due to cloud contamination that makes assimilation of AOD not possible. More importantly, heavy air pollution is often misclassified into cloud by the satellite cloud detection algorithm that results in very limited satellite retrievals of high AODs. Winter MODIS collection 6.0 AOD sampling percentage in Beijing is smaller than 20% and most of missing retrievals corresponds to heavy pollution episodes (Fu et al., 2018). One would expect that AERONET can fill the gap of satellite retrievals owing to its high temporal resolution, however, AERONET AOD product in NCP is very limited in spatial coverage 73

Atmospheric Environment 191 (2018) 70–78

Z. Song et al.

Fig. 4. Scatter-plot of seasonal mean AOD between collocated CARSNET and MERRA-2 at 10 stations. The vertical and horizontal bars represent one standard deviation of seasonal mean.

anthropogenic emissions have varied dramatically. This may be another alternative explanation for this deficiency of MERRA-2. Diurnal variation of hourly PM2.5 concentration from MERRA-2 and MEPC is shown in Fig. 7. Both products show pronounced variations. PM2.5 decreases along with a replacement of stable boundary layer during nighttime by a convective boundary during daytime. Nighttime PM2.5 is larger than that during daytime. Several notable differences in subtle diurnal variations are evident. First, MERRA-2 PM2.5 decreases soon after sunrise and is unable to reproduce a maximum PM2.5 at rush hours as recorded by MEPC PM2.5, which is probably resulted from lack of detailed temporal changes in emissions in MERRA-2, for example, heavy transportation emissions during rush hours. Second, the minimum MEPC PM2.5 occurs in the late afternoon, while MERRA-2 produces a minimum at noon. Diurnal PM2.5 is to a great extent controlled by the boundary layer height. Our results show that MERRA-2 cannot follow the diurnal variation of PM2.5 but reproduce a good daytime variation of AOD, which likely implies that evolution of the boundary layer height may need improvement. Third, the diurnal variability magnitude of MERRA-2 is smaller than that of MEPC PM2.5. In order to have a closer looking potential causes for the difference in PM2.5 from MERRA-2 and surface measurement, temporal variation of dust, OM, sulfate concentrations from in situ measurements and MERRA-2 is presented in Fig. 8. MERRA-2 did not simulate two peaks of dust concentration (60 μgm−3 on day 63 and 40 μgm−3 on day 68) recorded by MEPC data. With regard to the temporal evolution of OM and sulfate, MERRA-2 can follow the tendency of MEPC in most cases. More specifically, MERRA-2 reproduced the evolution of the first pollution episode. OM and sulfate aerosols maintain a high level on day 57 and 58 and drop dramatically to background level on day 59. However, MERRA-2 OM and sulfate concentrations are only one half of MEPC concentrations. MERRA-2 nearly always underestimates OM concentrations relative to MEPC. While it is not unusual that MEPC sulfate concentrations are lower than MERRA-2 data. MERRA-2 underestimate

that would not be expected to contribute a lot to the MERRA-2 AOD assimilation. Not to say that AERONET version 2 often screened heavy pollution cases due to its strict cloud-screening algorithm (Eck et al., 2018). The lack of nitrate aerosols and some necessary chemical progresses in GOCART also have effects on negative biases (Randles et al., 2016). Daytime variations of MERRA-2 and sun-photometer AOD datasets in four seasons are presented in Fig. 5. MERRA-2 reproduces well daytime variation, with average R between both AOD products being 0.77 in four seasons. Hourly differences between two AOD products are stable and both products show nearly the same hourly variabilities of AOD. 3.2. PM2.5 comparison Comparisons of PM2.5 from MERRA-2 and MEPC are presented in Table 2. Agreement in PM2.5 between both products is reasonably good. Annual and seasonal FAC2 values are all larger than 0.5, which is comparable to similar results in other regions. On average, MERRA-2 produces lower annual PM2.5 concentration (17.7 μgm−3) (Table 2) and the bias shows seasonal dependence (Fig. 6). MERRA-2 PM2.5 is in better agreement with MEPC data in the summer when the bias is only 0.4 μgm−3. The performance of MERRA-2 PM2.5 is comparable in the spring and autumn when the bias is 14.9–16.0 μgm−3 (26%). The largest bias is found in the winter, reaching 43.8 μgm−3 (47%). This substantial large bias is probably owing to unresolved sources in the GOCART. For example, it was suggested that lack of nitrate and underestimation of OC emissions in urban/suburban location would certainly explain a significant portion of this bias (Buchard et al., 2016, 2018; Provencal et al., 2017b). This speculation is verified by validation of MERRA-2 aerosol species in next section. Furthermore, this is debatable to extend inventories of anthropogenic emissions in 2006 and 2008 to represent emission status during 2013–2014, given that 74

Atmospheric Environment 191 (2018) 70–78

Z. Song et al.

Fig. 5. Daytime variations of MERRA-2 (grey dashed-line) and sun-photometer (red dashed-line) AOD products in four seasons. Shading indicates month standard deviation of stations. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

sulfate and OM concentrations by ∼100 μgm−3 on day 57 and 58, which cannot fully account for underestimation of PM2.5 concentration by ∼200 μgm−3. The remaining 100 μgm−3 is mainly due to missing nitrate in MERRA-2 as shown in Fig. 9. Nitrate accounts for a few to 28% of PM2.5 concentration. The percentage of nitrate increases with the increase of PM2.5, which clearly suggests that secondary inorganic aerosol species play an important role in the enhancement of PM2.5 pollution. Moreover, new aqueous sulfate formation mechanism was revealed recently to compensate the considerable underestimation of sulfate and PM2.5 in the high polluted NCP region in popular models (Wang et al., 2016; Cheng et al., 2016), which have not been included in the GOCART. These processes should be carefully considered in MERRA-2 in order to improve its performance in PM2.5 simulation.

Table 2 Comparison of hourly PM2.5 concentration (μg m−3) between MERRA-2 and MEPC products. N represents data points. Mean PM2.5 is calculated from ground measurements. The standard deviation is calculated from PM2.5 differences between MERRA-2 and MEPC.

N Mean PM2.5 (μg m−3) Mean Bias(μg m−3) standard deviation (μg m−3) correlation coefficient FAC2

Overall

Spring

Summer

Autumn

Winter

2284686 62.2 −17.7 40.4 0.59 0.73

552177 57.8 −14.9 30.1 0.55 0.77

660636 44.6 −0.4 23.9 0.58 0.81

567567 59.8 −16.0 35.4 0.63 0.75

504306 91.3 −43.8 54.9 0.64 0.57

Fig. 6. Monthly PM2.5 variation from MERRA-2 (grey dashed-line) and MEPC products (red dashed-line). Shading indicates standard deviation of stations. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.) 75

Atmospheric Environment 191 (2018) 70–78

Z. Song et al.

Fig. 7. Diurnal variations of PM2.5 (μg m−3) from MEPC (red dashed-line) and MERRA-2 products (grey dashed-line) during 2014–2017. Shading indicates standard deviation of stations. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

pollution episodes, which may results in assimilation of AOD not possible. Second, even AODs are available, for example, during last two pollution episodes (Fig. 10), MERRA-2 still underestimates AOD. This fact implies that chemistry in the GOCART needs further improvement, especially in heavily polluted regions.

Fig. 10 shows AOD time series from sunphotometer, Terra and Aqua in Beijing. The contribution of each aerosol species to the total AOD from MERRA-2 is also shown for the validation. We can see that AOD from sunphotometer, Terra and Aqua show higher values than MERRA2 AOD, especially during the last two pollution episodes. Two outstanding features that merit mention are as follows. First, sunphotometer and satellite often cannot provide AOD retrievals during

Fig. 8. Comparison of temporal evolution of PM2.5 and its species against in situ measurements during a two-weeks campaign from February 24 and March 12 of 2014 in Beijing. 76

Atmospheric Environment 191 (2018) 70–78

Z. Song et al.

Fig. 9. Temporal variation of 12-h average nitrate aerosol concentration and its percentage to PM2.5 between 24 February and 12 March 2014 in Beijing. Correlation coefficient between nitrate and PM2.5 mass is 0.97. Fig. 10. Time series of total 550-nm AOD from sunphotometer, Terra and Aqua (red dot line, green Square line, blue triangle line) and the contributions of each AOD components (color shading) from MERRA-2 for February 24 to March 12 of 2014 pollution event observed in Beijing. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

4. Discussion and summary

AOD not possible under these conditions. Therefore, MERRA-2 improves little in high level of AOD and PM2.5 relative to MERRA-Aero or a control experiment (the experiment was performed by using the exact same modeling system as for MERRA-2 that is driven by MERRA-2 meteorology and identical aerosol emissions, but without the assimilation of AOD observations). In the latest version of AERONET AOD products, cloud-screening is relaxed to a reasonable extent that leads to that much more elevated AOD values are preserved relative to past versions (Eck et al., 2018). Similarly, MODIS cloud-screening algorithm has also been modified to cover these pollution events in the retrieval of AOD as many as possible. The retrieval percentage of the latest MODIS collection 6.1 AOD has been improved in polluted regions. Assimilation

Assimilation of satellite AOD products is a key component in the modern reanalysis model. The fundamental requirement is to have high-quality AOD with good spatial and temporal coverage. Satellite remote sensing of AOD is always hindered by the presence of cloud and thereby cloud-screening is one of important steps in retrieval of AOD. Traditionally, cloud-screening is very strict to minimize cloud contamination on AOD retrieval. Every coin has two sides. This strict requirement excluded many heavy aerosol events in the retrievals, for example, strong dust storms, and wide-spread biomass burning as well as regional anthropogenic pollution events. This makes assimilation of

77

Atmospheric Environment 191 (2018) 70–78

Z. Song et al.

of these latest AOD products would definitely improve the performance of MERRA-2 AOD and PM2.5 products. In situ measurement of aerosol species suggested that nitrate aerosol accounted for 20% of PM2.5 concentration during winter pollution events, which suggested that inclusion of nitrate simulation should be carefully considered in future aerosol reanalysis. Using AOD products at 10 sunphotometer stations and PM2.5 products at 88 stations in NCP, the quality of MERRA-2 AOD and PM2.5 is evaluated in detail. MERRA-2 species are also assessed by comparison against surface in situ analysis during a 2-week campaign. Major conclusions are as follows. Overall, MERRA-2 simulates AOD well. MERRA-2 produces similar temporal AOD variations as sunphotometer, i.e., daytime and seasonal variations of AOD are captured by MERRA-2. Spatial AOD variation is also captured by MERRA-2. However, MERRA-2 has lower AOD relative to sunphotometer despite it assimilates several advanced satellite and ground AOD products. The underestimation stands out when AOD is high. Lack of remote sensing AOD products under this situation should be one of primary causes for this phenomenon. Underestimation of PM2.5 is also remarkable and it shows seasonal dependence. This phenomenon is most striking in the winter when MERRA-2 PM2.5 is lower than MEPC by ∼50 μgm−3. On the contrary, MERRA-2 and MEPC PM2.5 data are close to each other in the summer. Diurnal evolution of MERRA-2 PM2.5 does not follow the step of MEPC, indicating that diurnal evolution of the boundary layer depth should be improved. Nitrate aerosol accounts for > 20% of PM2.5 concentration during winter pollution cases, which is likely the primary cause for the underestimation of PM2.5 by MERRA-2 since it excludes simulation of nitrate aerosol.

Eck, T., Holben, B., Reid, J., Xian, P., Giles, D., Sinyuk, A., Smirnov, A., Schafer, J., Slutsker, I., Kim, J., Koo, J., Choi, M., Kim, K., Sano, I., Arola, A., Syer, A., Levy, R., Munchak, L., O'Neill, N., Lyapustin, A., Hsu, N., Randles, C., Silva, A., Buchard, V., Govindaraju, R., Hyer, E., Crawford, J., Wang, P., Xia, X., 2018. Observations of the interaction and transport of fine mode aerosols with cloud and/or fog in Northeast Asia from Aerosol Robotic Network and satellite remote sensing. J. Geophys. Res. 123. https://doi.org/10.1029/2018JD028313. Fu, D., Xia, X., Wang, J., Zhang, X., Li, X., Liu, J., Song, Z., 2018. Synergy of AERONET and MODIS AOD products for estimating PM2.5 concentration in Beijing. Sci. Rep. https://doi.org/10.1038/s41598-01828535-2. Hess, M., Koepke, P., Schult, I., 1998. Optical properties of aerosols and clouds: the software package OPAC. Bull. Am. Meteorol. Soc. 79, 831–844. Holben, B.N., Tanré, D., Smirnov, a., Eck, T.F., Slutsker, I., Abuhassan, N., Newcomb, W.W., Schafer, J.S., Chatenet, B., Lavenu, F., Kaufman, Y.J., Castle, J. Vande, Setzer, a., Markham, B., Clark, D., Frouin, R., Halthore, R., Karneli, a., O'Neill, N.T., Pietras, C., Pinker, R.T., Voss, K., Zibordi, G., 2001. An emerging ground-based aerosol climatology: aerosol optical depth from AERONET. J. Geophys. Res. 106, 12067. https://doi.org/10.1029/2001JD900014. Levy, R.C., Mattoo, S., Munchak, L.A., Remer, L.A., Sayer, A.M., Patadia, F., Hsu, N.C., 2013. The Collection 6 MODIS aerosol products over land and ocean. Atmos. Meas. Tech. 6, 2989–3034. Li, Z., Lau, W.K., Ramanathan, V., Wu, G., Ding, Y., Manoj, M.G., Liu, J., Qian, Y., Li, J., Zhou, T., Fan, J., Rosenfeld, D., Ming, Y., Wang, Y., Huang, J., Wang, B., Xu, X., Lee, S., Cribb, M., Zhang, F., Yang, X., Takemura, T., Wang, K., Xia, X., Yin, Y., Zhang, H., Guo, J., Zhai, P.M., Sugimoto, N., Babu, S.S., Brasseur, G.P., 2016. Aerosol and monsoon climate interactions over Asia. Rev. Geophys. 119–161. Liu, J., Xia, X., Li, Z., Wang, P., Min, M., Hao, W., Wang, Y., Xin, J., Li, X., Zhang, Y., chen, Z., 2010. Validation of multi-angle imaging spectroradiometer aerosol products in China. Tellus B 62, 117–124. Ma, Q., Wu, Y., Tao, J., Xia, Y., Liu, X., Zhang, D., Han, Z., Zhang, X., Zhang, R., 2017. Variations of chemical composition and source apportionment of PM2.5 during winter haze episodes in Beijing. Aerosol Air Qual. Res. 17, 2791–2803. Malm, W.C., Schichtel, B.A., Pitchford, M.L., 2011. Uncertainties in PM2.5 gravimetric and speciation measurements and what we can learn from them. J. Air Waste Manag. Assoc. 61, 1131–1149. Provencal, S., Buchard, V., Silva, A.M.D., Leduc, R., Barrette, N., 2017a. Evaluation of PM2.5 surface concentrations simulated by version 1 of NASA's MERRA aerosol reanalysis over Europe. Atmos. Poll. Res. 8, 374–382. Provencal, S., Buchard, V., da Silva, A.M., Leduc, R., Barrette, N., Elhacham, E., Wang, S.H., 2017b. Evaluation of PM2.5 surface concentrations simulated by version 1 of NASA's MERRA aerosol reanalysis over Israel and Taiwan. Aerosol Air Qual. Res. 17, 253–261. Randles, C.A., da Silva, A.M., Buchard, V., Darmenov, A., Colarco, P.R., Aquila, V., Bian, H., Nowottnick, E.P., Pan, X., Smirnov, A., Yu, H., Govindaraju, R., 2016. The MERRA - 2 aerosol assimilation. Tech. Rep. Ser. Glob. Model. Data Assim. 45. Randles, C.A., da Silva, A.M., Buchard, V., Colarco, P.R., Darmenov, A., Govindaraju, R., Smirnov, A., Holben, B., Ferrare, R., Hair, J., Shinozuka, Y., Flynn, C.J., 2018. The merra-2 aerosol reanalysis, 1980 onward. part I: system description and data assimilation evaluation. J. Clim. 30, 6823–6850. Rienecker, M.M., Suarez, M.J., Gelaro, R., Todling, R., Bacmeister, J., Liu, E., Bosilovich, M.G., Schubert, S.D., Takacs, L., Kim, G.K., Bloom, S., Chen, J., Collins, D., Conaty, A., da Silva, A., Gu, W., Joiner, J., Koster, R.D., Lucchesi, R., Molod, A., Owens, T., Pawson, S., Pegion, P., Redder, C.R., Reichle, R., Robertson, F.R., Ruddick, A.G., Sienkiewick, M., Woollen, J., 2011. MERRA: NASA's Modern-Era retrospective analysis for research and application. J. Clim. 24, 3624–3648. Valavanidis, A., Fiotakis, K., Vlachogianni, T., 2008. Airborne particulate matter and human health: toxicological assessment and importance of size and composition of particles for oxidative damage and carcinogenic mechanisms. J. Environ. Sci. Health Part C Environ. Carcinog. Ecotoxicol. Rev. 26, 339–362. Wang, J., Christopher, S., 2003. Intercomparison between satellite-derived aerosol optical thickness and PM2.5 mass: implications for air quality studies. Geophys. Res. Lett. 30. https://doi.org/10.1029/2003GL018174. Wang, G., Zhang, R., Gomez, M.E., Yang, L., Levy Zamora, M., Hu, M., Lin, Y., Peng, J., Guo, S., Meng, J., Li, J., Cheng, C., Hu, T., Ren, Y., Wang, Y., Gao, J., Cao, J., An, Z., Zhou, W., Li, G., Wang, J., Tian, P., Marrero-Ortiz, W., Secrest, J., Du, Z., Zheng, J., Shang, D., Zeng, L., Shao, M., Wang, W., Huang, Y., Wang, Y., Zhu, Y., Li, Y., Hu, J., Pan, B., Cai, L., Cheng, Y., Ji, Y., Zhang, F., Rosenfeld, D., Liss, P.S., Duce, R.A., Kolb, C.E., Molina, M.J., 2016. Persistent sulfate formation from London Fog to Chinese haze. Proc. Natl. Acad. Sci. Unit. States Am. 113, 13630–13635. Xia, X., Chen, H., Goloub, P., Zong, X., Zhang, W., Wang, P., 2013. Climatological aspects of aerosol optical properties in North China Plain based on ground and satellite remote-sensing data. J. Quant. Spectrosc. Radiat. Transf. 127, 12–23. Xia, X., Che, H., Zhu, J., Chen, H., Cong, Z., Deng, X., Fan, X., Fu, Y., Goloub, P., Jiang, H., Liu, Q., Mai, B., Wang, P., Wu, Y., Zhang, J., Zhang, R., Zhang, X., 2016. Groundbased remote sensing of aerosol climatology in China: aerosol optical properties, direct radiative effect and its parameterization. Atmos. Environ. 124, 243–251. Zhang, R.J., Jing, J., Tao, J., Hsu, S.C., Wang, G., Cao, J., Lee, C.S.L., Zhu, L., Chen, Z., Zhao, Y., Shen, Z., 2013. Chemical characterization and source apportionment of PM2.5 in Beijing: seasonal perspective. Atmos. Chem. Phys. 13, 7053–7074.

Acknowledgements This work is supported by the national key research and development program of China (2016YFC0200403) and the National Science Foundation of China (41475138, 41675033, 41175028, and 91644217). We greatly appreciate NASA GMAO providing MERRA-2 aerosol reanalysis data. References Buchard, V., Silva, A., Colarco, P., Krotkov, N., Dickerson, R., Stehr, J., Mount, G., Spinei, E., Arkinson, H., He, H., 2014. Evaluation of GEOS-5 sulfur dioxide simulations during the Frostburg, MD 2010 field campaign. Atmos. Chem. Phys. 14, 1929–1941. Buchard, V., da Silva, A.M., Randles, C.A., Colarco, P., Ferrare, R., Hair, J., Hostetler, C., Tackett, J., Winker, D., 2016. Evaluation of the surface PM2.5 in version 1 of the NASA MERRA aerosol reanalysis over the United States. Atmos. Environ. 125, 100–111. Buchard, V., Randles, C.A., da Silva, A.M., Darmenov, A., Colarco, P.R., Govindaraju, R., Ferrare, R., Hair, J., Beyersdorf, A.J., Ziemba, L.D., Yu, H., 2018. The MERRA-2 aerosol reanalysis, 1980 onward. Part II: evaluation and case studies. J. Clim. 30, 6851–6872. Che, H., Xia, X., Zhu, J., Li, Z., Dubovik, O., Holben, B., Goloub, P., Chen, H., Estelles, V., Cuevas-Agullo, E., Blarel, L., Wang, H., Zhao, H., Zhang, X., Wang, Y., Sun, J., Tao, R., Zhang, X., Shi, G., 2014. Column aerosol optical properties and aerosol radiative forcing during a serious haze-fog month over North China Plain in 2013 based on ground-based sunphotometer measurements. Atmos. Chem. Phys. 14, 2125–2138. Che, H., Zhang, X.Y., Xia, X., Goloub, P., Holben, B., Zhao, H., Wang, Y., Zhang, X.C., Wang, H., Blarel, L., Damiri, B., Zhang, R., Deng, X., Ma, Y., Wang, T., Geng, F., Qi, B., Zhu, J., Yu, J., Chen, Q., Shi, G., 2015. Ground-based aerosol climatology of China: aerosol optical depths from the China aerosol remote sensing network (CARSNET) 2002-2013. Atmos. Chem. Phys. 15, 7619–7652. Cheng, Y., Zheng, G., Wei, C., Mu, Q., Zheng, B., Wang, Z., Gao, M., Zhang, Q., He, K., Carmichael, G., Pöschl, U., Su, H., 2016. Reactive nitrogen chemistry in aerosol water as a source of sulfate during haze events in China. Sci. Adv. 2, 1601530. Eck, T.F., Holben, B.N., Reid, J.S., Dubovik, O., Smirnov, A., Neill, N.T.O., Slutsker, I., Kinne, S., 1999. Wavelength dependence of the optical depth of biomass burning, urban , and desert dust aerosols. J. Geophys. Res. 104, 31333–31349.

78