1,118
Views
61
CrossRef citations to date
0
Altmetric
Technical Papers

Visual Range Trends in the Yangtze River Delta Region of China, 1981–2005

, , , , , , , & show all
Pages 843-849 | Published online: 10 Oct 2011

ABSTRACT

Visual range (VR) data from 1981 to 2005 were examined for 20 meteorological monitoring sites in the Yangtze River Delta Region of China. Cumulative percentile analysis was used to construct VR trend. The 25-yr average domain-average 50% VR was approximately 21.9 ±1.9 km. Domain-average 50% VR decreased from 1981 to 2005 with a trend of −2.41 km/decade. The worst 20% and 50% and best 20% VR and variation rates for the 20 sites were analyzed. The 50% VR of the town, county-level city, and prefecture-level city sites were 24.1, 21.5, and 19.4 km, respectively. The best 20% VR decreased fastest with a rate of −3.5 km/decade. Regional median VR decreased from the coastal sites to the inland sites. Ridit analysis and cumulative percentile were adopted to study the VR variation properties between economically developed areas (e.g., Nanjing and Hangzhou) and remote areas (e.g., Lvsi). The two analyses show that VR decreased in Nanjing and Hangzhou but remained constant in Lvsi from 1981 to 2005.

IMPLICATIONS

Decreasing VR corresponds to greater industrialization and particulate concentrations. Moving pollution sources from the coastal to inland locations only moves the decreasing visibility elsewhere. This work offered a good opportunity to comprehend the VR variation in the most developed region of China over the last 2 decades.

INTRODUCTION

Visual range (VR) refers to the maximum distance, usually horizontal, at which a given object or light is visible under particular conditions of atmospheric light transmission and background luminance.Citation1 VR impairment is caused by scattering and absorption of light by fine particles and gases in the atmosphere,Citation2 which are aggravated under some weather conditions.Citation1 In an uncontaminated atmosphere, VR normally ranges from 145 to 225 kmCitation3 and is limited by scattering from oxygen (O2) and nitrogen (N2) molecules. VR can be less than 1 km at polluted sites.Citation4

Long-term VR trendsCitation5 and the influence of major pollutants on local VRCitation6 have been examined. Husar et al.Citation4 summarized spatial and temporal patterns of haziness in the eastern United States using four historical databases. Mahowald et al.Citation7 studied global VR trends related to dust sources and manmade factors. Park et al.Citation8 compared the contribution of natural and transboundary pollution influences to regional VR in the United States. WatsonCitation9 discussed some of the limitations of VR as related to a more objective measure of light extinction at a point and along a select path. At the same time, VR analysis methods such as ridit analysesCitation10 and cumulative percentiles, which are based on statistical techniques, were developed. SloaneCitation11,Citation12 examined U.S. VR trends, and Doyle and DorlingCitation5showed VR trends for the United Kingdom.

In China, Che et al.Citation13 compared VR trends in China from 1981 to 2005, finding a significant decrease in VR (−2.1 km/decade from 1990 to 2005). Fan et al.Citation14 analyzed VR trends around Beijing, Tianjin, and Tangshan and showed that VR decreased from 1980 to 2003 with the steepest decline during summer. Song et al.Citation15 related VR degradation to particle pollution in Beijing. Hong et al.Citation16 found that VR decreased in Hangzhou, part of the Yangtze River Delta Region (YRDR), from average values of 10 km in the 1980s to 7 km in the first decade of the 21st century. Chang et al.Citation1 found the impact of industrialization on VR from 1973 to 2007 in Shanghai. This study adds to these studies of VR in China by composing tracks among major cities in the YRDR.

DATA AND METHODOLOGY

Data

The YRDR study area is located in eastern China () and includes Shanghai, the eastern part of the Anhui province (west of 118.5° east), the southern part of the Jiangsu province (north of 33° north), and the northern part of the Zhejiang province (south of 28° north). VR data and several meteorological variables, including relative humidity (RH) and precipitation, were obtained from China Meteorological Administration (CMA)/National Meteorological Center (cdc.cma.gov.cn) for each site () in this region for the 25 yr of 1981–2005. Well-trained observers measure the VR using reference objects (e.g., buildings and mountains) in different directions at a known distance from the observer. The data uncertainty is 0.1 km according to the visibility observation guideline of CMA, which was promulgated in 1980.Citation13 The industrial output data for Jiangsu, Shanghai, and Zhejiang and energy consumption in Jiangsu were obtained from http://www.stats.gov.cn and http://www.Jssb.gov.cn, respectively.

Figure 1. Locations of (a) the YRDR and (b) its 20 sites. Area in the small box of panel a is the YRDR.

Figure 1. Locations of (a) the YRDR and (b) its 20 sites. Area in the small box of panel a is the YRDR.

VR was recorded by the site observers four times (2:00 a.m., 8:00 a.m., 2:00 p.m., and 8:00 p.m., local standard time [LST]) per day at each station. The afternoon (2:00 p.m. LST) readings are used in these analyses to minimize effects of morning radiation fogs and air pollutants trapped in shallow radiation inversions.

If the purpose of trend analysis is to detect any alterations of air quality because of man's activities, then natural influence on air quality needs to be unfolded from the data.Citation11 In this study, VR measurements corresponding to precipitation were eliminated from the dataset.Citation10 Because particle growth is affected by high RH, which decreases VR,Citation17 the dataset was further limited to RH less than 90%.Citation5 The frequency of sand and dust storms was less than 1 time per year from 1954 to 1998 in the YRDR.Citation18 Asian dustCitation19 is not detected in large quantities in the YRDR because it is at a position further north, so sand or dust storms are retained in the dataset. There were approximately 50.3% of samples remaining out of the total possible in the analyzed dataset. The fewest and most samples occurred in Quzhou (∼42.9%) and Xuyi (∼60%), respectively.

Methodology

Ridit Analysis

Because the ridit theory is adequately documented,Citation10,Citation20,Citation21 ridit analysis was used. For example, VR observations for the Nanjing site during 1981–2005 play the role of reference distribution V, and VR observation in 1981 is a comparison group X. Five VR categories (i.e., K = 5) were selected according to Fan et al.Citation14: 0–1.9, 2–9, 10–19, 20–39, and more than 40 km. Referring formulas in the Appendix, the mean ridit of the VR observations in 1981 (r = 0.66) is obtained. This implies that VR in 1981 tends to be better than it was during the period 1981–2005 at the Nanjing site. With a sample size of 199 and 5226, respectively, for X and V, the variance of is estimated by the equation in the Appendix, and this gives the test statistic U of 9.24. If the confidence level α is set at 0.01, a U α of 2.32 is obtained; that is, at a level of α of 0.01, VR in 1981 was better than it was during 1981 and 2005.

In this study, a year was divided into the four seasons (i.e., March through April for spring, June through August for summer, September through November for autumn, and December through February for winter) and separate ridit analyses were conducted over time to examine the annual and seasonal VR trends of Nanjing, Hangzhou, and Lvsi.Citation5 Different sites have different locations, populations, and emission sources, so their VR distributions are different from each other. At each site, a reference distribution was obtained for 25 yr or each season by pooling all of the observations for the 25 yr or that season. Ridit values are not comparable between different sites and seasons because they have different reference distributions. U was calculated for each mean ridit. It is found that except for 2–3 mean ridits around the value of 0.5, all of the values pass the t test at the confidence level α = 0.01. Ridit analysis can be not only used to detect the general trend over a whole time period, but it can also be used to determine whether or not a pollution strategy instituted at a particular time was effective in improving visibility by examining the slope representing the periods before and after the policy changed.Citation10

Cumulative Percentiles

Cumulative percentiles are also examined to compose visibility among the different measurement sites. The ith cumulative percentile is the VR that is equal to or exceeds i percent of the time.Citation22 VR cumulative frequency distribution function is given by

(1)
where f(v) is the VR frequency distribution function. There are n i observations that are equal or greater than v i in n observations. Thus, the ith cumulative percentile is given by
(2)
Five categories similar to those used in ridit analysis were provided here. The U.S. Regional Haze RuleCitation23 uses the upper 20th percentile of chemical extinction to represent “poor” VR and the lower 20th percentile of chemical extinction to represent “good” VR; this study uses the same convention. The worst 20% and 50% and best 20% VR for the 20 sites from 1981 to 2005 were calculated.

Linear regression analysis has been adopted to characterize long-term trends of VR. In this study, VR rate denotes the slope of linear regression of VR against year. All VR rates pass the t test with a confidence level of 99% unless it is expressly stated. The worst 20% and 50% and best 20% 25-yr average VR and rate for 20 sites in the YRDR are shown in . Domain-average VR is the average value of the 20 sites' VR. The worst 20% and 50% and best 20% domain-average VR values and variation trends are shown in and .

Table 1. The worst 20% and 50% and best 20% 25-yr average VR and their rates in the YRDR

Figure 2. The worst 20% and 50% and best 20% domain-average VR trends of the YRDR. Dashed lines are the linear regression curves of the corresponding trend lines.

Figure 2. The worst 20% and 50% and best 20% domain-average VR trends of the YRDR. Dashed lines are the linear regression curves of the corresponding trend lines.

Median visibility is frequently used to summarize visibility observations. In this study, 50% VR (i.e., median VR) was used to represent the level of VR of each site because it is a commonly used statistic and is readily understood.Citation24 The results are shown in .

Figure 3. Spatial distribution of the 5-yr average median (i.e., 50%) VR (in km): (a) the first 5 yr (i.e., from 1981 to 1985) and (b) the last 5 yr (i.e., from 2001 to 2005). The interpolation error created by sparse sites in southwestern Liyang and the coastal and southeastern regions of Longquan should be considered in analysis.

Figure 3. Spatial distribution of the 5-yr average median (i.e., 50%) VR (in km): (a) the first 5 yr (i.e., from 1981 to 1985) and (b) the last 5 yr (i.e., from 2001 to 2005). The interpolation error created by sparse sites in southwestern Liyang and the coastal and southeastern regions of Longquan should be considered in analysis.

RESULTS AND DISCUSSION

VR Properties of the YRDR

Variation trends of the worst 20% and 50% and best 20% VR are summarized in and . These can then be examined to identify increasing, decreasing, or a combination of trends over time. The worst 20% and 50% and best 20% VR have similar fluctuations from 1981 to 2005. The worst 10% domain-average VR degraded at a rate of 1.6 km/decade with a decline of 4 km over 25 yr. The highest VR was 16.1 km in 1984 and the lowest value was 11.9 km in 2002. Fifty percent VR decreased 6 km over the 25 yr at a rate of 2.4 km/decade, attaining the highest value of 24.6 km in 1981 and the minimum of 18.1 km in 2002. Best 20% VR declined at a rate of 3.5 km/decade with a decline of 8.8 km from 1981 to 2005. Fifty percent VR rates of decline are on the order of the −2.1 km/decade found by Che et al.Citation13

The 20 sites are categorized by town, county-level city, and prefecture-level city according to the population in . From , it is shown that town sites have the highest worst 20% and 50% and best 20% VR and the prefecture-level city sites have the lowest. Smaller communities appear to have better visibility than larger cities. Quzhou is located in the Jinqu Basin with the Qu River passing through it. Poor ventilation may lead to its low 50% VR just like Chengdu in the western Sichuan Basin because they have similar terrain features.Citation1,Citation25 Only the Pinghu site (a town site) shows an upward trend for the worst 20% and 50% and best 20% VR. Many of the town sites have smaller worst decrease rates or even increase rates. Almost all of the prefecture-level city sites have a higher VR decrease rate than the other two category sites have. The best 20% decrease rate is greater than the worst 20% or 50% VR decrease rate for most sites.

is 5-yr mean median VR of the first 5 yr (i.e., 1981–1985) and the last 5 yr (i.e., 2001–2005). Regional VR decreased from the coast to the inland. Kim et al.Citation26 found that VR improved at two inland urban sites when a cleaner air mass originated over the Pacific Ocean. The coastal sites rather than the inland sites will be the first to be affected by the clean air from the ocean, so clean air from the ocean may be one of the reasons why VR decreased from the coast to the inland. contrasts the predevelopment and postdevelopment periods in the region. VR decreased at every site over the YRDR except for a few undeveloped sites along the coast. Median VR in Nanjing, Changzhou, Hangzhou, and Dongshan (at Suzhou city), which have been developing fast since 1980, decreased from the 18- to 21-km grade to the grade less than 15 km.

Intercomparison of VR of Nanjing, Hangzhou, and Lvsi

Urban sites Nanjing and Hanzhou and the nonurban site Lvsi are examined in more detail in and . Nanjing and Hangzhou are provincial capitals of Jiangsu and Zhejiang. Lvsi town is located in the northeastern part of the YRDR and is a small town revolving around the fishing industry.

Figure 4. Ridit plot for Nanjing, Hangzhou, and Lvsi. A mean ridit value of 0.5 means there is no difference with respect to the reference distribution, whereas a ridit value <0.5 means lower VR and >0.5 means better VR.

Figure 4. Ridit plot for Nanjing, Hangzhou, and Lvsi. A mean ridit value of 0.5 means there is no difference with respect to the reference distribution, whereas a ridit value <0.5 means lower VR and >0.5 means better VR.

Figure 5. Five-year averaged VR of the (a) worst 20%, (b) 50% (median), and (c) best 20% at Nanjing, Hangzhou, and Lvsi.

Figure 5. Five-year averaged VR of the (a) worst 20%, (b) 50% (median), and (c) best 20% at Nanjing, Hangzhou, and Lvsi.

Results from the ridit analysis are shown in . The annual ridit curves of Nanjing decrease from 1981 to 2005. The ridit has a value above 0.5 before 1991, suggesting that the VR was better before 1991 in comparison to the entire period. The ridit values were around 0.5 from 1992 to 1995, which could be regarded as the transition period. After 1996, the ridit values were lower than 0.5, indicating that the annual average VR was worse than the entire record. The transition period for winter was within the period 1987–1996, and for spring it was 1987–1996. The transition period for summer occurred from 1991 to 1999. Even without a transition period, autumn VR became worse when compared with the post-1993 record length. Except for different transition periods, the seasonal ridit value curves have trends that are similar to the annual one. The annual and seasonal ridit trends at Hang-zhou are similar to those at Nanjing, but Lvsi is different. Instead of degradation during the entire observation period, the annual and seasonal ridit values remained almost constant. There was a steep change where the annual and seasonal ridit values in 1988 went above 0.5, showing that the VR was better in 1988 compared with the entire period.

shows VR distances during the worst 20% days. VR values at Lvsi exceeded those in Nanjing and Hangzhou. VR during the worst 20% days was approximately 22 km at Lvsi, but it was only 4–12 km at Nanjing and Hangzhou. VR improved at Lvsi but decreased at Nanjing and Hangzhou. Fifty percent VR at Lvsi was 30 km in . At Nanjing, VR decreased from 28 to 12 km between 1981 and 2005. At Hangzhou, 50% VR decreased from 20 to 12 km. Lvsi had the highest best 20% VR in the value of 38 km from 1986 to 1995, and downward trends were observed from 1981 to 2005 in Nanjing and Hangzhou.

The Relationship between VR and Development

shows industrial output for Jiangsu, Shanghai, and Zhejiang increasing by a factor of 40 from 1981 to 2005. Energy consumption also increased since 1995 in Jiangsu along with economic development. Pollution controls were not a high priority during this growth, and primary particles and precursors were emitted into the atmosphere. The correlation coefficient between 50% VR over the YRDR and the total industrial output is −0.87, consistent with increasing uncontrolled pollutant emissions that follow economic growth.

Figure 6. Total industrial output values of Jiangsu, Shanghai, and Zhejiang and energy consumption values in industrial enterprises of Jiangsu.

Figure 6. Total industrial output values of Jiangsu, Shanghai, and Zhejiang and energy consumption values in industrial enterprises of Jiangsu.

SUMMARY

Cumulative percentile analysis was used to construct VR trend. The 25-yr average domain-average 50% VR was approximately 21.9 ± 1.9 km. Domain-average 50% VR decreased from 1981 to 2005 with a trend of −2.41 km/ decade. The domain-average worst 20% and best 20% VR have decrease rates of −1.6 and −3.5 km/decade, respectively. The worst 20% and 50% and best 20% VR and variation rates for the 20 sites were analyzed. The 50% VR of the town, county-level city, and prefecture-level city sites are 24.1, 21.5, and 19.4 km, respectively. The best 20% decrease rate is greater than the worst 20% or 50% VR decrease rate for most sites. Regional median VR decreased from the coastal sites to the inland sites.

Ridit analysis and cumulative percentile analysis were adopted to study the VR variation properties between economically developed areas (e.g., Nanjing and Hang-zhou) and remote areas (e.g., Lvsi). The two analyses show that VR decreased in Nanjing and Hangzhou but remained constant in Lvsi from 1981 to 2005. Economies grew rapidly in the YADR, and pollution controls were not a high priority during this growth, with primary particles and precursors emitted into the atmosphere. This may be the main reason why VR experienced a decrease trend in the YRDR.

ACKNOWLEDGMENTS

This research is supported by the National Basic Research Program of China (2010CB428503 and 2009CB723904), the Research and Development Special Fund for the Public Welfare Industry of CMA (GYHY201006047), and the Chinese Academy of Meteorological Sciences Basis Research Project (2009Z01).

REFERENCES

  • Chang , D. , Song , Y. and Liu , B. 2009 . Visibility Trends in Six Megalopolis in China 1973–2007 . Atmos. Res. , 94 : 161 – 167 .
  • Leung , Y.K. , Wu , M.C. and Yeung , K.K. 2009 . A Study on the Relationship among Visibility, Atmospheric Suspended Particulate Concentration, and Meteorological Conditions in Hong Kong . Acta Meteorol. Sinica , 23 : 250 – 260 .
  • 2001 . Visibility in Mandatory Federal Class I Areas (1994–1998) , Research Triangle Park , NC : Office of Air Quality Planning and Standards . EPA-452/R-01-008; U.S. Environmental Protection Agency
  • Husar , R.B. , Holloway , J.M. , Patterson , D.E. and Wilson , W.E. 1981 . Spatial and Temporal Pattern of Eastern U.S. Haziness: a Summary . Atmos. Environ. , 15 : 1919 – 1928 .
  • Doyle , M. and Dorling , S. 2002 . Visibility Trends in the UK 1950–1997 . Atmos. Environ. , 36 : 3136 – 3172 .
  • Schichtel , A.S. , Husar , R.B. , Falke , S.R. and Wilson , W.E. 2001 . Haze Trends over the United States, 1980–1995 . Atmos. Environ. , 35 : 5205 – 5210 .
  • Mahowald , N.M. , Ballantine , J.A. , Feddema , J. and Ramankutty , N. 2007 . Global Trends in Visibility: Implications for Dust Sources . Atmos. Chem. Phys. , 7 : 3309 – 3339 .
  • Park , R.J. , Jacob , D.J. , Kumar , N. and Yantosca , R.M. 2006 . Regional Visibility Statistics in the United States: Natural and Transboundary Pollution In?uences, and Implications for the Regional Haze Rule . Atmos. Environ. , 40 : 5405 – 5423 .
  • Watson , J.G. 2002 . Visibility: Science and Regulation—2002 Critical Review . Journal of the Air & Waste Management Association , 52 : 628 – 713 .
  • Craig , C.D. and Faulkenberry , G.D. 1979 . The Application of Ridit Analysis to Detect Trends in Visibility . Atmos. Environ. , 13 : 1617 – 1622 .
  • Sloane , C.S. 1982 . Visibility Trends—I. Methods of Analysis . Atmos. Environ. , 16 : 41 – 51 .
  • Sloane , C.S. 1982 . Visibility Trends—II. Mideastern United States 1948– 1978 . Atmos. Environ. , 16 : 2309 – 2321 .
  • Che , H.Z. and Zhang , X.Y. 2007 . Horizontal Visibility Trends in China 1981– 2005 . Geophys. Res. Lett. , 34 doi: 10.1029/2007GL031450
  • Fan , Y.Q. and Li , C.Q. 2008 . Visibility Trends in Beijing, Tianjin and Hebei Province during 1980–2003 . Plateau Meteorol. , 27 : 1392 – 1400 . in Chinese with English abstract
  • Song , Y. , Tang , X.Y. , Fang , C. , Zhang , Y.H. , Hu , M. , Zeng , L.M. , Li , C.C. , Mao , J.T. and Bergin , M. 2003 . Relationship between the Visibility Degradation and Particle Pollution in Beijing . Acta Sci. Circum. , 23 : 468 – 471 . in Chinese with English abstract
  • Hong , S.M. , Jiao , L. , He , X. , Bao , Z. , Sun , H.L. , Xu , H. , Ma , W.L. , Yang , L. and He , J.P. 2009 . Variation of Visibility in Hangzhou Urban and Their Relation with Major Factors . China Powder Sci. Technol. , 15 : 56 – 61 . in Chinese with English abstract
  • Tang , I.N. , Wong , W.T. and Munkelwitz , H.R. 1981 . The Relative Importance of Atmospheric Sulfates and Nitrates in Visibility Reduction . Atmos. Environ. , 15 : 2463 – 2471 .
  • Zhou , Z.J. 2001 . Blowing-Sand and Sandstorm in China in Recent 45 Years . Auatern. Sci. , 21 : 10 – 17 .
  • Cai , C.X. , Jiang , W.M. and Huang , S.H. 2000 . Chemical Characteristic of Two Dust Storms in the South-East Coastal Area of China and Its Possible Origin . Plateau Meteorol. , 19 : 179 – 186 .
  • Faulkenberry , D. and Craig , C.D. 1980 . The Application of Ridit Analysis to Detect Trends in Visibility—Reply . Atmos. Environ. , 14 : 1205 – 1206 .
  • Beder , J.H. and Heim , R.C. 1990 . On the Use of RIDIT Analysis . Psychometrika , 55 : 603 – 616 .
  • Holzworth , G.E. and Maga , J.A. 1960 . A Method for Analyzing the Trend in Visibility . J. Air. Pollut. Cont. Assoc. , 10 : 430 – 435 .
  • 1999 . Regional Haze Regulations: Final Rule. 40 CFR Part 51 . Fed. Regist. , 64 ( 126 ) : 35714 – 35774 .
  • Lee , D.O. 1988 . The Choice of Visibility Statistics in the Analysis of Long Term Visibility Trends in Southern England . Weather , 43 : 332 – 338 .
  • Li , Z. , Fu , G. and Hao , L. 2007 . Climatologic Characteristics of Fog and Visibility in the Sichuan Basin Region . Per. Ocean Univ. China , 37 : 191 – 196 . in Chinese with English abstract
  • Kim , Y.J. , Kim , K.W. , Kim , SD , Lee , B.K. and Hane , J.S. 2006 . Fine Particulate Matter Characteristics and Its Impact on Visibility Impairment at Two Urban Sites in Korea: Seoul and Incheon . Atmos. Environ. , 40 ( Suppl. 2 ) : 593 – 605 .

APPENDIX

Let V and X be the role of the reference and comparison groups, respectively. K VR categories were selected. Referring to Beder and Heim,Citation21 let

(1)
and denote the column vectors (q 1,…,q K)′ and (p 1,…,p K)′ by and , respectively, which are probability distributions on {1,…,K}. Such vectors are probability distributions on {1,…,K}.

The kth ridit for the reference sample is

(2)
and for the comparison sample is
(3)

The mean ridit of the comparison with respect to the reference group is given by

(4)
and the variance of r(X) is
(5)

If the roles of the two groups interchanged, the variance of t(V) is obtained

(6)

The test statistic is U given by

(7)

It is considered likely that VR in X would be higher than VR in V, and it was desired to test this hypothesis. In the terms of the introduction, the conjecture was that P(V < X) would exceed 1/2. Then R(p|q) is a proxy for P(V < X) and the problem becomes the testing of the hypotheses

(8)

The test statistic U is given by Equationeq 7 with R(p|q) = 0.5, and H 0 is rejected if U is large. The sample sizes for X and V were m and n, respectively. From eqs Equation4–6, the variance of is estimated by

(9)

This gives the value U. If the confidence level α is set at 0.01, then a U α of 2.32 is obtained; that is, if U > U α, then H 0 is rejected at a level of α = 0.01 and the VR in X is better than the average VR level in V (eqs Equation1–9 are all from Beder and HeimCitation21).

Reprints and Corporate Permissions

Please note: Selecting permissions does not provide access to the full text of the article, please see our help page How do I view content?

To request a reprint or corporate permissions for this article, please click on the relevant link below:

Academic Permissions

Please note: Selecting permissions does not provide access to the full text of the article, please see our help page How do I view content?

Obtain permissions instantly via Rightslink by clicking on the button below:

If you are unable to obtain permissions via Rightslink, please complete and submit this Permissions form. For more information, please visit our Permissions help page.