Spatiotemporal Characterization of Ambient PM2.5 Concentrations in

Subscriber access provided by NEW YORK MED COLL

Article

Spatiotemporal Characterization of Ambient PM2.5 Concentrations in Shandong Province (China) Yong Yang, and George Christakos Environ. Sci. Technol., Just Accepted Manuscript • Publication Date (Web): 26 Oct 2015 Downloaded from http://pubs.acs.org on October 27, 2015

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Environmental Science & Technology is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 18

Environmental Science & Technology

Spatiotemporal Characterization of Ambient PM2.5 Concentrations in Shandong Province (China) Yong Yang1,2, and George Christakos3* 1. Department of Resources & Environmental Information, College of Resources & Environment, Huazhong Agricultural University, Wuhan (China). 2. Key Laboratory of Arable Land Conservation (Middle & Lower Reaches of Yangtse River), Ministry of Agriculture (China). 3. Ocean College, Zhejiang University, Hangzhou, Zhejiang (China). Corresponding author: George Christakos Address: 866 Yuhangtang Road, Hangzhou, Zhejiang Province, 310058, China Affiliations: Ocean College, Zhejiang University, Hangzhou, Zhejiang (China) Phone: (+86) 571-88981701/139-68048138 Fax: (+86) 0580-2092891 Email: [email protected]

Abstract China experiences severe particulate matter (PM) pollution problems closely linked to its rapid economic growth. Advancing the understanding and characterization of spatiotemporal air pollution distribution is an area where improved quantitative methods are of great benefit to risk assessment and environmental policy. This work uses the Bayesian Maximum Entropy (BME) method to assess the space-time variability of PM 2.5 concentrations and predict their distribution in the Shandong province, China. Daily PM 2.5 concentrations obtained at air quality monitoring sites during 2014 were used. Based on the space-time PM 2.5 distributions generated by BME, three kinds of querying analysis were performed to reveal the main distribution features. The results showed that the entire region of interest is seriously polluted (BME maps identified heavy pollution clusters during 2014). Quantitative characterization of pollution severity included both pollution level and duration. The number of days during which regional PM 2.5 exceeded 75, 115, 150, and 250 µg∗ m −3 varied: 43-253, 13-128, 4-66, and 0-15 days, respectively. The PM 2.5 pattern exhibited an increasing trend from east to west, with the western part of Shandong being a heavily polluted area ( PM 2.5 exceeded 150

µg∗ m −3 during long time periods). Pollution was much more serious during winter than during other seasons. Site indicators of PM 2.5 pollution intensity and space-time variation were used to assess regional uncertainties and risks with their interpretation depending on the pollutant threshold. The observed PM 2.5 concentrations exceeding a specified threshold increased almost linearly with increasing threshold value, whereas the relative probability of excess pollution decreased sharply with increasing threshold. Keywords: PM 2.5 , spatiotemporal, risk, BME, indicators, China.

ACS Paragon Plus Environment


1. INTRODUCTION Numerous studies have suggested that particulate matter (PM) in the atmosphere is linked to adverse impacts on human health [1, 2]. Along with rapid economic growth, regional industrialization, surging of cars, and urbanization, China experiences serious particulate matter pollution problems, leading to considerable human health impacts. In order to evaluate the PM situation, the Chinese government has made a serious effort to investigate the PM pollution situation in the country. On February 29, 2012, the 3rd revision of the “Ambient Air Quality Standard” (AAQS, GB 3095-2012) was released [3]. Since January 2013, 113 of the major cities have started releasing information on seven pollutant concentrations, including sulfur dioxide ( SO2 ), nitrogen dioxide ( NO2 ), PM10 , PM 2.5 , carbon monoxide (CO), and 1and 8-hour peak ozone concentrations ( O3 ) concentrations [4]. Based on the available monitoring data, the spatiotemporal variability of these pollutants has been studied rather extensively [5-8]. Some studies focused on pollution prediction based on spatiotemporal geostatistics methods, including Bayesian maximum entropy (BME) [9-11] and Kriging [12, 13]. However, the results of space-time prediction are not easily understood by the public because of a lack of further space-time analysis of the pollution distribution generated by interpolation techniques. In addition, measurements and predictions of atmospheric pollutants are subject to uncertainty, and, thus, any environmental decisions based on such measurements are also subject to uncertainty [14, 15]. Therefore, identifying regions at risk and reducing uncertainties across broad spatial and temporal scales is an important component of many pollution studies [16-19]. In view of the above considerations, this work has three main goals: the first one focuses on PM 2.5 prediction by means of the space-time BME method; the second goal is to reveal key space-time PM 2.5 features in the study area based on the results of BME prediction using different space-time querying patterns; the third goal is to calculate site indicators of PM 2.5 variation that can provide a quantitative assessment of the uncertainties and risks in the polluted regions. 2 MATERIALS AND METHODS 2.1 Study Area and Data Sources The region under study is the Chinese province of Shandong with an area of 157,900 Km 2 . Data used in this study were obtained at 96 national air quality monitoring sites during the period January 1-December 31, 2014 (Fig 1). Ambient PM 2.5 concentrations were measured according to the Chinese Environmental Protection Standard HJ655-2013 [20]. At each site, the daily PM 2.5 concentration was calculated by averaging hourly data. A summary of the descriptive statistics of PM 2.5 concentrations during the 365 days of the year 2014 is shown in Fig S1. The temporal trend of the averaged PM 2.5 values at all monitoring sites is presented in Fig S2. The mean PM 2.5 concentration of all data collected was 74.84 µg ∗ m −3 ,


Page 2 of 18

Page 3 of 18


and the coefficient of variation (CV) was equal to 0.69, indicating a medium variability of the monitored PM 2.5 concentrations (i.e., 0.1 < CV ζ ) over the probability that it does not, PEO shows the sharpest decrease among all site indicators starting at ζ = 0. On the other hand, the space-time average of the PM 2.5 > ζ over D (i.e., MEP) shows the slowest decrease up to ζ ≈ 50 µg∗ m −3 , and then displays a fast decrease, reaching a zero value at ζ ≈ 300 µg∗ m −3 . The RAEP (i.e. the ratio of the area in which PM 2.5 > ζ over the study area D ) decreases slowly up to ζ ≈ 30 µg ∗ m −3 , and then

experiences a quick, almost linear decrease, reaching a zero value at ζ ≈ 240 µg ∗ m −3 . The averaged pollution exceedance difference, PM 2.5 − ζ , over D (MEDP) decreases starting at ζ = 0, remains non-zero even for significant ζ -levels, and reaches a zero value only at very high thresholds, ζ ≈ 220 µg ∗ m −3 (i.e., the MEDP

variation confirms the existence of



Page 12 of 18

heavily polluted parts in the region of interest during 2014). And, the averaged PM 2.5 > ζ concentrations in the ζ -exceedance subarea Θ ⊂ D (CMEP) increase almost linearly with increasing ζ . I.e., the higher the ζ of a subarea Θ , the higher the average pollution in the subarea. As a numerical example, in the part of the region in which pollution exceeds, say,

ζ =180 µg ∗ m −3 the average PM 2.5 pollution is about 200 µg ∗ m −3 , which is much bigger than the average pollution of about 100 µg ∗ m −3 occurring in the part of the region in which pollution exceeds ζ = 60 µg ∗ m −3 (see Fig. 3). Thus, the variation of CMEP indicates the existence of significantly polluted areas throughout the Shandong province during 2014. From the one-point site indicators, we can obtain a numerical illustration of global pollution in the study area during the study period. For example, for a threshold value

ζ = 67.11 µg ∗ m −3 , the values of RAEP and PEO are equal to 0.5 and 1, respectively, indicating that, with 100% odds, in about half of the space-time domain (i.e., the study area during year 2014), the PM 2.5 concentrations were higher than 67.11. In this case, the PID value is 0.144, implying that the average excess PM 2.5 dispersion is a small percentage of the mean PM 2.5 concentration. This is due to the fact that in the specific case most MEDP values are close to 0 µg ∗ m −3 , whereas the threshold values are higher than 250 µg ∗ m −3 . The rates by which the site indicators change can be calculated in terms of the site elasticity indicator, SEI (i.e., the % of site indicator change corresponding to an 1% change of the pollution threshold). For illustration, for the studied PM 2.5 distribution the SEI of the RAEP, MEDP and PEO indicators are given by, respectively,

ERD (ζ )

=

dRX (ζ ) RX (ζ ) dζ

=

ζ

E LD (ζ )

=

dLDX (ζ ) LDX (ζ ) dζ

d lnRX (ζ ) R′X (ζ ) ζ= ζ, dζ RX (ζ )

=−

ζ

RX (ζ ) ζ, LDX (ζ )

(7)

(8)

and EOD (ζ )

=

dOXD (ζ ) OXD (ζ ) dζ

ζ

where RX′ (ζ ) =

=

R′X (ζ ) ζ, (1−RX (ζ )) 2 OXD (ζ )

(9)

dRX (ζ ) . Interestingly, the SEI of MEDP and PEO are not only functions of dζ

MEDP, PEO and ζ (as expected) but of the RAEP, as well. For numerical illustration, the E RD (ζ ) , E LD (ζ ) , and EOD (ζ ) vs. ζ are plotted in Fig. S8. As is shown in these plots, all

SEI are negative, decreasing slowly up to a certain value ζ ≈151 µg∗ m −3 (i.e., when ζ increases by a certain percentage, the PM 2.5 indicator values decrease by the corresponding


Page 13 of 18


percentage). In this ζ -range of values the RAEP and MEDP exhibit almost the same elasticity, whereas the elasticity of PEO is larger. At larger ζ values, the RAEP and PEO fluctuate with increasing amplitude, i.e., their elasticity varies widely for very high ζ values. Interesting, this means that at very high thresholds a threshold change can cause rather unpredictable (very small or very large) reductions in the areal extent of PM 2.5 pollution. On the other hand, the MEDP fluctuates slightly and then increases exponentially (i.e., its elasticity is very sensitivity to very high ζ values). Moreover, an SEI may be expressed in terms of another. For example EOD (ζ ) =

RX (ζ ) E D (ζ ) , (1−RX (ζ )) 2 OXD (ζ ) R

(10)

i.e., the elasticity of PEO is a linear function of the elasticity of RAEP. 3.3.2 Two-Point Site Indicators In Fig S9 the two-point site indicators for selected PM 2.5 threshold ζ -values are plotted. The first thing one notices is that, while the indicators NIC, CIC, NEC, EDC and STIR experienced a steadily decreasing trend with increasing ζ , the CEC showed the reverse behavior. Specifically, the NIC plots (probability that PM 2.5 > ζ at both points connected by the space-time lag δp in the study area D ) offer valuable information concerning site risk assessment. For smaller ζ –values, the NIC values are high, indicating a close connectivity between the events that PM 2.5 > ζ at space-time points p and p + δp (obviously, the NIC is higher at smaller lags δp and lower at larger δp ). As the ζ –values increase, the space-time connectivity of PM 2.5 concentrations decreases. Otherwise put, the connectivity between PM 2.5 > ζ concentrations at two different space-time points is higher for smaller ζ –values and lower for larger ζ –values. For further illustration, in Fig S10 the NIC indicator is plotted for selected ζ -values. >ζ While the NIC refers to PM 2.5 probabilities, the NEC plots (in (µg∗ m −3 ) 2 ) represent >ζ the space-time covariance (dependency) between PM 2.5 concentrations at both points

connected by δp within the sub-area Θ in which over-pollution occurs. For smaller ζ –values, the NEC values are higher, indicating a closer dependency (the NEC is higher at smaller δp and lower at smaller δp ). As the ζ –values increase, the space-time PM 2.5 concentration dependency decrease. The EDC plots (in (µg∗ m −3 ) 2 ) represent the space-time covariance (dependency) between PM 2.5 − ζ concentrations at both points connected by δp within the sub-area Θ . For smaller ζ –values, the EDC values are higher, indicating a closer dependency (the EDC is higher at smaller δp and lower at smaller δp ). As the ζ –values increase, the space-time PM 2.5 concentration dependency decreases. As regards the IR (Fig S9f), it establishes a linear relationship between the probability that PM 2.5 pollution threshold exceedance occurs at two points of the space-time domain simultaneously and the probability that ζ -exceedance occurs at either one of these points. The IR decreases with increasing ζ . On the other hand, the space-time variation of IR is smoother for small



Page 14 of 18

ζ –values, and exhibits a larger variation for higher ζ –values (the degree of space-time variation increases with increasing ζ –value). Lastly, the level-crossing contours are lines along which PM 2.5 = ζ . The continuum specific length (CSL) of the level-crossing PM 2.5 contours is an indicator given by [26] lS (ζ ) =

Π X (ζ ) |D|

,

(11)

where Π X (ζ ) denotes the ensemble expectation of the total length of the level-crossing contours, and the quantity | D | denotes the total length of the contours in the space-time domain. As we saw above, in the present work the site indicators were calculated based on the PM 2.5 prediction maps generated by the space-time BME method. Then, assuming that the site is simulated and expressed on a cube, the level-crossing contours on the cube can be approximated by the bonds that join neighbor sites with opposite values of the indicator field. Hence, the CSL can be approximated by the cube specific length, lb (ζ ) calculated as in Table S2 (in units m −1 ). As is shown in Fig S11, in the case of PM 2.5 pollution in the Shandong province, the CLS increases with the threshold, up to the value ζ = 67.11 µg∗ m −3 ; and then the CLS starts decreasing quickly, reaching a zero value at ζ = 305 µg∗ m −3 . Remarkably, the highest CLS value was found for ζ = 67.11 µg∗ m −3 , which, as was shown earlier, is equal to the threshold for which the RAEP value is RX (ζ = 67.11) = 0.5, i.e., half of the Shandong province is over-polluted ( PM 2.5 concentration exceeds the specified threshold).

3.4 The severity of PM2.5 pollution in Shandong Province The analysis of the present work concluded that the PM 2.5 pollution status in the Shandong province is a serious problem. Based on the space-time cube data generated by the BME method, three kinds of space-time querying were performed to reveal space-time PM 2.5 distribution characteristics of the study area during 2014. The results showed that the western part of Shandong province was more polluted by PM 2.5 than the eastern part (BME maps identified heavy pollution clusters during 2014). Quantitative characterization of pollution severity included both pollution level and duration. During half of 2014 the western part of the Shandong province was significantly polluted by PM 2.5 ( ζ > 75 µg ∗ m −3 ); even worse, during 50 days of that time period the western part was characterized as heavily polluted ( ζ > 150 µg ∗ m −3 ). The temporal CV of the PM 2.5 distribution in the western part was lower than that in the eastern part of the Shandong province. In term of seasonality, in the winter the PM 2.5 pollution was much more serious than during the other seasons. And,


Page 15 of 18


the livability index showed an increasing trend from west to east in the study area. The site indicators (one- and two-point) of PM 2.5 pollution intensity and space-time variation provided valuable risk assessment information, based on the fact that their interpretation depends decisively on the pollutant threshold value considered. In the study area D (Shandong province): the relative probability of excess pollution decreases sharply with increasing threshold ζ ; the averaged excess PM 2.5 > ζ concentration initially decreases slowly up to ζ ≈ 50 µg∗ m −3 , and then displays a fast decrease towards zero; the relative area of excess pollution decreases slowly up to ζ ≈ 30 µg∗ m −3 and then experiences a fast, almost linear decrease towards zero; the averaged PM 2.5 exceedance difference ( PM 2.5 − ζ ) shows the sharpest decrease (among the one-point indicator) towards zero; the probabilistic connectivity between PM 2.5 > ζ concentrations at different space-time points is higher for smaller ζ ( δp ) values and lower for larger ζ ( δp ) values; and the continuum specific length of the level-crossing PM 2.5 contours increases with ζ up to the value

ζ = 67.11 µg∗ m −3 (threshold for which the relative probability of excess pollution is 0.5) and then decreases quickly towards zero. In the over-polluted subarea: the averaged excess PM 2.5 concentration increases almost linearly with increasing ζ ; and the space-time dependency between concentrations PM 2.5 > ζ (or PM 2.5 − ζ ) at both points separated by lag δp is higher for small ζ ( δp ) values and lower for larger ζ ( δp ) values. The elasticities of the one-point site indicators were found to be all negative indicating that when ζ increases by a certain percentage, the indicator values decrease by the corresponding percentage. The elasticities increase slowly up to a certain ζ value, whereas at larger ζ values, the elasticities either fluctuate with increasing amplitude or increase exponentially. Remarkably, elasticity analysis showed that, if the Chinese pollution standard decreases slightly from 75 to about 67 ( µg ∗ m −3 ), then half of the Shandong province is PM 2.5 over-polluted.

Acknowledgments The research was supported by the Fundamental Research Funds for the Central Universities (Grant No. 2662014PY062 and 2662015PY156), National Natural Science Foundation of China (Grant No. 41101193), and China Scholarship Council. Opinions in the paper do not constitute an endorsement or approval by the funding agencies and only reflect the personal views of the authors.

References [1] Ko, F.W.S.; Tam, W.; Wong, T.W.; Lai, C.K.W.; Wong, G.W.K.; Leung, T.F.; Ng, S.S.S.; Hui, D.S.C. Effects of air pollution on asthma hospitalization rates in different age groups in Hong Kong.

Clinical and Experimental Allergy. 2007, 37: 1312-1319. [2] Lippmann, M.; Frampton, M.; Schwartz, J.; Dockery, D.; Schlesinger, R.; Koutrakis, P.; Froines, J.; Nel, A.; Finkelstein, J.; Godleski, J.; Kaufman, J.; Koenig, J.; Larson, T.; Luchtel, D.; Liu, L.J.S.; Oberdorster, G.; Peters, A.; Sarnat, J.; Sioutas, C.; Suh, H.; Sullivan, J.; Utell, M.; Wichmann, E.; Zelikoff, J. The US Environmental Protection Agency particulate matter health effects research



centers program: a midcourse report of status, progress, and plans. Environmental Health

Perspectives. 2003, 111: 1074-1092. [3] Ministry of Environmental Protection of the People’s Republic of China (MEP). China National

Ambient Air Quality Standards. MEP, Beijing, China., 2012. [4] Ministry of Environmental Protection of the People’s Republic of China (MEP). Technical

Regulation on Ambient Air Quality Index (On Trail). China Environmental Science Press, Beijing, China., 2012. [5] Snepvangers, J.J.J.C.; Heuvelink, G.B.M.; Huisman, J.A. Soil water content interpolation using spatio-temporal kriging with external drift. Geoderma. 2003,112: 253-271. [6] Juneng, L.; Latif, M.T.; Tangang, F.T.; Mansor, H. Spatio-temporal characteristics of PM10 concenttration across Malaysia. Atmospheric Environment. 2009, 43: 4584-4594. [7] Chu, H.J.; Yu, H.L.; Kuo, Y.M. Identifying spatial mixture distributions of PM2.5 and PM10 in Taiwan during and after a dust storm. Atmospheric Environment. 2012, 54: 728-737. [8] Hu, J.L.; Wang, Y.G.; Ying, Q.; Zhang, H.L. Spatial and temporal variability of PM2.5 and PM10 over the North China Plain and the Yangtze River Delta, China. Atmospheric Environment. 2014, 95: 598-609. [9] Akita, Y.; Chen, J.C.; Serre, M.L. The moving-window Bayesian maximum entropy framework: estimaton of PM2.5 yearly average concentration across the contiguous United States. Jour. of

Exposure Science and Environmental Epidemiology. 2012, 22: 496-501. [10] Yu, H.L.; Wang, C.H. Quantile-based Bayesian maximum entropy approach for spatiotemporal modeling of ambient air quality levels. Environmental Science & Technology. 2013, 47: 1416-1424. [11] Christakos, G.; Serre, M.L. BME analysis of spatiotemporal particulate matter distributions in North Carolina. Atmospheric Environment. 2000, 34: 3393-3406. [12] Vyas, V.; Christakos, G.;. Spatiotemporal analysis and mapping of sulfate deposition data over the conterminous USA. Atmospheric Environment. 1997, 31(21): 3623-3633. [13] Pearce, J.L.; Rathbun, S.L.; Aguilar-Villalobos, M.; Naeher, L.P. Characterizing the spatiotemporal variability of PM2.5 in Cusco, Peru using kriging with external drift. Atmospheric Environment. 2009, 43: 2060-2069.

[14] Smith, J.L.; Halvorson, J.J.; Papendick, R.L. Using multiple-variable indicator kriging for evaluating soil quality. Soil Science Society of America Journal. 1993, 57: 743-749.

[15] Christakos, G.; Hristopulos, D.T. Spatiotemporal Environmental Health Modelling; Kluwer Academic Publ., Boston, MA., 1998. [16] Wang, G.; Gertner, G.; Parysow, P.; Anderson, A.B. Spatial prediction and uncertainty analysis of topographic factors for the revised universal soil loss equation (RUSEL). Journal of Soil Water

Conservation. 2000, 55: 374-384. [17] Webster, R.; Oliver, M.A. Geostatistics for Environmental Scientists (2nd Edition). J. Wiley & Sons, Ltd., Chichester, U.K., 2007. [18] Zhao, Y.C.; Shi, X.Z.; Yu, D.S.; Wang, H.J.; Sun, W.X. Uncertainty assessment of spatial patterns of soil organic carbon density using sequential indicator simulation, a case study of Hebei province, China. Chemoshpere. 2005, 59: 1527-1535. [19] Zhao, C.X.; Wang, Y.Q.; Wang, Y.J.; Zhang, H.L.; Zhao, B.Q. Temporal and spatial distribution of PM2.5 and PM10 pollution status and the correlation of particulate matters and meteorological factors during winter and spring in Beijing. Environmental Science, 2014, 35(2): 418-427.


Page 16 of 18

Page 17 of 18


[20] National Health & Family Planning Commission of China (NHFPCC). Chinese Health and

Family Planning Statistical Yearbook for 2013. Beijing, China., 2013. [21] Kolovos, A.; Skupin, A.; Jerrett, M.; Christakos, G. Multi-perspective analysis and spatiotemporal mapping of air pollution monitoring data. Environmental Science & Technology. 2010, 44(17): 6738-6744. [22] Christakos, G. A Bayesian/maximum-entropy view to the spatial estimation problem.

Mathematical Geology, 1990, 22(7): 763-777. [23] Christakos, G. Modern Spatiotemporal Geostatistics. Oxford University Press, New York, NY., 2000. [24] Christakos, G. Random Field Models in Earth Sciences. Academic Press, San Diego, CA., 1992. [25] Yu, H.L.; Ku, S.J.; Kolovos, A. Advanced space-time predictive analysis with STAR-BME. SIGSPAIAL’12 Proceedings of the 20th International Conference on Advances in Geographic

Information Systems, 2012, 593-596. [26] Christakos, G.; Hristopulos, D.T. Characterization of atmospheric pollution by means of stochastic indicator parameters. Atmospheric Environment. 1996, 30(22): 3811-3823.

[27] Christakos, G.; Hristopulos, D.T. Stochastic indicator analysis of contaminated sites. Jour. of Applied Probability, 1997, 34(4): 988-1008.



83x39mm (300 x 300 DPI)


Page 18 of 18

Spatiotemporal Characterization of Ambient PM2.5 Concentrations in

Recommend Documents