Air Quality Index ranking of China’s main cities during the 1st quarter of 2014

AQI bubble map is showing during Q1-2014 a concentration of very polluted or unhealthy urban pollution  in Hebei (Shijiazhuang, Baoding) , Shandong (Zibo), Hubei (Wuhan), Sichuan (Chengdu) and Shaanxi (Xi’an) cities

The ranking of the Chinese 190 main cities is issued each month by the China’s Ministry of Environmental Protection (see MEP link). It covers an overall population about 416 Mil people in main cities totaling around 63% of Chinese urban population.

Main results from AQI issued by MEP over Q1-2014

Based on the methodology described in my previous posts, the population weighted average AQI overall value over these 190 cities was AQI= 119 during Q1-2014 (90 days).

The  population-weighted average daily AQI repartition over these cities - during the period- was the following :

-    8 days /90 days or 9% where AQI < 50

-    37 days/90 days or 41% where 50< AQI <100

-    45 days/90 days or 50% where AQI >100

If we arrange these 90 population-weighted averages according to their decreasing daily  AQI values, the last day of the first and worst decile- assuming a lognormal distribution is valid - should be close to 150.

We know that EU is requesting a maximum value of 50 for the last day of the first decile. In average over a year an exposure to AQI >5O (or PM2.5>35µg/m3) during more than 36 days a year (10%) is considerated as dangerous, due to the cumulative effect  -even of a small pollution -on human health over a long period.

U.S. appellation to qualify the dangerousness same AQI levels in China and USA are used, these appearing to be more pertinent that the Chine's. 

AQI ranking is addressed in decreasing order unlike the publications from MEP. It is thus rather a "ranking" of Air Pollution urban areas.

We use the same color convention to represent the various AQI levels as in our previous posts. Nevertheless we should bear in mind that averaging tends to reduce daily or monthly average over time and cities. 

The best indicator of Air Pollution health hazards is undoubtedly the daily value in each city.  We know only AQI repartition inside three intervals: AQI <50; 50< AQI<100  and AQI >100 - although a splitting of the last interval in two would be helpful.

A bubble map of air pollution in China is given above. It gives a good feeling of the AQI situation during Q1-2014. It is nevertheless difficult to read as urban grid is very tight in the Beijing-Tianjin-Hebei area and in coastal regions.

Figure 1 : 30 of the most Air polluted cities are located in a semicircular area around Beijing, Tianjin,  Zibo, Wuhan, Chengdu and Xi’an

Zooming on the most polluted area is showing that the 30 most polluted cities were located within an semicircular area (see Figure 1 above) , located between the Yangze and the Huang He rivers, which surface is around 1,200,000 km2  or altogether: France, Germany, Netherlands, Belgium and Italy .

Figure 2: China Air Quality Index bubble zooming the Southern Chongqing-Shanghai-Guangzhou triangle moderate pollution

Figure 2 above is showing "unhealthy for sensitive groups" urban pollution just South to the Yangze river, becoming only moderatly polluted down to the coastal region. 

AQI main cities ranking over Q1-2014

Both following Figures 3 & 4  give the AQI ranking of the first two "quintiles" presented by MEP  concerning the 76 most polluted cities : 

- 1^st quintile: urban population 74 Mil with 38 cities all their AQI but one being either "very polluted" or "unhealthy”, the largest cities being: Xi'an N°12 (Shaanxi), Chengdu N°24 (Sichuan) and Wuhan N°26 (Hubei) ;

- 2^nd quintile: urban population 81Mil with 38 cities which AQI are all “unhealthy for sensitive groups”, the largest being Beijing N°42 ,Tianjin N°44, Harbin N°52,  Nanjing N°64 (Jiangsu) and Chongqing .

Figure 3: First quintile of AQI ranking of 190 main cities over Q1-2014

Figure 3 : Second quintile of AQI ranking of 190 main cities over Q1-2014

Discussion of AQI values on Q1-2014

In general, the main pollutant was PM25 except for some few exceptions where it was PM10.

The result for Beijing, which exhibits 14 days with AQI < 50 when in average AQI = 147 seems quite unlikely, while similar cities for their AQI level - such as Tianjin- are all showing 0-5 days for this indicator.

Furthermore, comparing the sub-AQI corresponding to the value of the monthly concentration of PM2.5, to the monthly value of the AQI (see my last post) it is possible to highlight some possible errors. 

This test establishes that 99 values out from 570 (190 cities x 3 months) or 17% might be inaccurate with an AQI value underestimated in the following cases:

Ma'anshan- in Mar 2014-(underestimation: 24.1 /AQI=108.4); Zhangqiu-Feb- (19.3/149.2); Langfang- Feb- (14.6/195.3); Ma'anshan-Jan- (12.7/187.78); Beijing- Feb- (12.0/183.7); Handan- Feb- (10.8/210.2);Tangshan-Feb-(9.71/190.3); Xianyang-Jan- (9.58/194.4); Baoji- Jan-(9.21/173.7)…Xi'an-Jan- (8.48/185.8)…Zibo-Jan- (7.09/185.8); Shijiazhuang-Feb (7/266.0)…etc

As the concerned cities are mosly  classified on a monthly basis either "Unhealthy" or "Very Polluted", such AQI underestimation could rock the ranking list and expend the number of "Unhealthy " or"Very Polluted" cities on a quarterly basis, such as Beijing for instance.  

Last Chinese Air Quality Index (AQI) showing persistent Air pollution during Jan & Feb 2014 with high level of Particulate Matter pollution in Northern regions

Shijiazhuang is struggling to contain a dense smog because of its heavy reliance on coal; both Xingtai and Shijiazhuang cities- located in Hebei 300 km South of Beijing- had the worst AQI throughout China during Jan & Feb 2014. From CCTV reporter Ning Hong - 16 Oct 2013

In the first two months of 2014, the 190 most important Chinese cities sample were recorded with an Air Quality Index (AQI) value around 130. Most affected areas were the industrial North region including Harbin, Beijing and Tianjin.

Particulate Matters (PM2.5 and PM10) were almost always the principal pollutants determining the value of Air Quality Index, PM2.5 with 68% of the city-month records being mostly concerned with PM10 only 29%. As a consequence, the PM2.5 inclusion in the new AQI definition has dramatically changed the Air quality perception.  


During Jan. & Feb. 2014, on a 59 days' period: the 76 most polluted cities (40% of 190 sample) had largely exceeded the yearly budget of "exceedance" days (with AQI>50 or PM2.5> 35μg/m3) normally permitted in the WHO, the US or the UE for instance.

This means that based on the first two months and for 40% of major urban areas China is more or less six times more polluted as compared with Western standards! 

The new Air Quality Index publications, concerning China’s major urban areas, are being speeding up gradually by the Ministry of Environmental Protection (MEP).

However every month already published AQIs ​​are modified retrospectively due to errors found after publication.

But we believe that many errors are still present: from a test performed by computing the highest values ​​of all pollutant sub-indices, we found that at least 181 cities indications already published - or 16% of all published city-month records- should still be corrected, including: Harbin (Oct & Dec), Xi’an (Dec & Jan) and Beijing (Feb).

What are the main information resulting from last AQI data released on the most important Chinese cities?

After 76 cities recorded from May 2013, then 144 cities from Sep 2013, we have now 190 cities- all among the largest urban areas- recorded from Jan 2014 with the new Air Quality Index (AQI) format that incorporates the finest Particulate Matter (PM2.5).

As explained in our previous blog dated 12 Nov 2013, we may establish average per capita pollution over specific region or on an overall national basis only after urban population weighting the results of each city inside the concerned area.

In the first two months Jan. & Feb. 2014, the population weighted average AQI value is 129 for an overall urban population of 408 Mil of the 190 concerned cities.

The most affected areas were the industrial North regions including Harbin, Beijing and Tianjin. The list of the top 76 most polluted cities is given in the following Figures 1 (a) and 1 (b):

Figure 1 (a): 38 most polluted cities or first quintile (1/5 of  the 190 cities) from mean Jan & Feb 2014

Figure 1 (b): 39th to 76th most polluted cities or second  quintile from mean Jan & Feb 2014

The first 40% - or 2 fifths (quintile) - of the ranked 190 Chinese cities are located mostly in the Northern (60%) or in the Middle (37%) regions.

These cities were classified with an Air Quality Index- following US denominations- either "Very polluted" (11 cities including Xi'an), or "Unhealthy" (40 cities including Beijing and Harbin), or "Unhealthy for sensitive groups" (25 cities including Chongqing).

Xi'an (ranking n°11), Harbin (n°40), Beijing (n°45), Chongqing (n°53), Tianjin (n°56) and Nanjing (n°64) are among the greatest Chinese cities but are also those which air is the most heavily polluted by Particulate Matter. 

The G06 Northern region (Beijing, Tianjin, Hebei Provinces & Dalian City) plus being the People's Republic of China capital location were certainly the most polluted during the period.

With 10 cities ranking respectively n°1 (Xingtai), n°2 (Shijiazhuang), n°3 (Baoding), n°5 (Handan), n°7 (Hengshui), n°16 (Langfang), n°25 (Tangshan), n°37 (Cangzhou), n°45 (Beijing) &  n°56 (Tianjin) and a population totaling 48 Mil: the population-weighted AQI value is around 162  with only  4.8 "good days" over the 59 days period. 

Most of the pollution is coming from the surrounding Hebei Province which burned over 300 million tons of coal in 2011, more than all of Germany!

The number of "good " days - where AQI <50 or PM2.5 <35μg/m3- during Jan. & Feb. 2014 where nil for Xi'an, Harbin and Tianjin, 3 for Chongking and Nanjing and -quite unexpectedly- 10 for Beijing with 5 in Jan. and again 5 in Feb.. 

As a result, after over the first two months of the year all these cities have largely exceeded the yearly budget of "exceedance" days such as defined for instance in the UE . This means that based on these first two months China is more of less six times more polluted following normal western standards.

Figure 2 : Most polluted 78 Chinese Cities from mean Jan & Feb 2014

 What are the most important pollutants concerned?

For a specific day the AQI value is the value reached by the highest underlying sub-index of each of the 7 main different pollutants recorded: PM2.5, PM10, CO, NO2, SO2 , O3-1h & O3-8h.

Since the publication of the new monthly AQI values, we have 1102 city-month records of monthly average AQI and associated pollutants over a period of 10 months from May 2013 to Feb 2014.

It is possible to compute monthly sub index values from each monthly recorded pollutant average concentration. In the vast majority of city-month records, PM2.5 or PM10 were almost always the most important pollutants as defined by the value of the various sub-index (see Figures 3 (a) & 3 (b)).

Under this approach we found that PM2.5 was the most important pollutant for 750 city-month records (68%), then PM10 for 325 (29%), NO2 for 17 (1.5%) and O3-8h then SO2 respectively for 6 (0.5%) & 3 (0.25%) city-month records only.

Figure 3 (a) : Beijing, Canton and Chongqing 10 month records from May 2013 to Feb 2014 

Figure 3 (b) : Beijing, Canton and Chongqing 10 month sub indexes as derived from May 2013 to Feb 2014 detailed AQI and associated pollutant records

As explained above, each new AQI monthly issue from China’s MEP might is bringing corrections to several of the preceding issued city-month records.

And nevertheless lots of mistakes are still uncorrected! We found that for 181 city-month records, the value of the month was less than the value of the sub highest index computed from the monthly average of the corresponding pollution (See Harbin in Oct. and Dec., in Beijing in Feb in following Figure 4).

These are certainly errors since the highest sub index including the daily volatility during the specific month is certainly higher than the sub index without volatility!  

181 errors on 1102 values is a level of mistakes around 16% with lots of errors still concerning Oct., Nov. and Dec..

Figure 4 : 40 among 181 errors found in the last AQI issues where the sub index is higher than the recorded AQI

These values ​​are average monthly which might hide severe daily pollution

As we have not for each city the detailed daily value of each of 7 main pollutants we can only assess the most severe daily pollution from the monthly average issued each month.    

Figure 5: Existing correlation between monthly  AQI or PM2.5 sub-index and PM2.5 monthly average PM2.5 for all 1102 city-month records; a large number of PM2.5 sub-index higher than that recorded AQI necessarily corresponds to an error

Figure 6: Correlation between overall AQI and calculated monthly Sub Index PM10 for all monthly-cities 1102 records

But in each of the 1102 city-month records we have the number of days during the month period where AQI was "good" (AQI<50) which permits an estimation of the size of population exposed over long periods to unhealthy air quality.

Figure 7: Number of days where AQIcity is "good" or <50

The Figure 7 above is showing that "good air quality" had only 22-29 "good" days during each last month; when "moderate air quality" had only 5-10 "good days"; and "unhealthy air quality" only 1-5 "good days"!   

These performances are more than extremely exceeding the health limitations put forward by WHO, EU or US regulations which are -in the EU-  90.4 percentiles with PM10 < 50μg/m3  or less than 35 exceedance days (9.6%) over one year and - in the US- 98 percentiles  for PM2.5 < 35 μg/m3 or less than 7 exceedance days (2%) over one year (see my last blog dated 22 Sep 2013). 

During Jan & Feb 2014 the population-weighted "good days" (AQI<50 or PM2.5<35μg/m3) were around 5.5 days for all the 190 cities. 

This mean as exposed above that on the overall 190 cities during Jan. to Feb. 2014 there were 53 (=59-5.5) unhealthy days (with AQI>50 or PM2.5>35μg/m3)  or 1.5 times the 35 days exceedance yearly period accepted by WHO, the US or the UE!

Finally is it possible to know whether air pollution is on an upward or downward trend during the last years?

As we did in our last post dated  10 Nov. 2013, we may use the benchmark of air pollution introduced by W. J. Qu et al. in their 2010 study "Spatial distribution and inter-annual variation of PM10 concentrations over 86 Chinese cities" as established  from Jun. 2000 to Feb. 2007 daily API data available.

As during that period the finest particulate matter were not recorded, we might only compare PM10 pollution using this 86 cities bench mark which are the biggest cities inside the 190 cities recorded by MEP.

On the overall 86 cities the total population weighted PM10 is +2.17% higher than the 2000-2007 bench-mark (see Figure 8 and 9 below):

Figure 8: Monotonous curve comparison with the bench-mark

Figure 9 : Comparison for 2000-2007 86 cities benchmark ; in blue the 2000-2007 population weighted 10 months PM10 value ; in red the last 10 month values from same 86 cities recorded completed progressively with small missing cities deducted from previous seasonal factors; for each cluster the number of cities and the overall population are given

On the 16 city’s clusters (see Figure 9 & 10):

There are reduction for Northern G01 (-21.76%); Northern G03 (-23.35%); Northern G05 (-17.87%); Southern G12 (-13%) & West GX1 (-10.14%). 

But also we note a huge increase for Northern G07 (37%); Northern G08 (34%); Southern G13 (30%); Northern G02 (11%) & Middle G11   (8.8%).      

Figure 10 : The 16 clusters of correlated cities used as a bench-mark 

2014 Environment Performance Index (EPI) analysis is showing a 4% worldwide average reduction over 2012 which means EPI Yale experts are more pessimistic especially concerning urgent actions on water issues

Figure 1: Water scarcity has become one of our greatest challenges. In less than twenty years, nearly two billion people could face water shortages.

© Binh Thuan, Thien Anh Huynh / Vietnam / 

Yale last 2014 EnvironmentPerformance Index (EPI) is showing deterioration from the assessment made in 2012: the worldwide ranking based on population weighted index over all the ranked countries is around 45.7 in 2014 after 47.5 in 2012 which is a reduction of 4% over the last two years.

There had been modifications in the methodology and some additional countries missing in 2012 are now included in 2014’s assessment. Altogether the general situation of Environment Performance is looking bleaker than it was 2 years ago mostly because of water issues. 

Water scarcity has become one of the greatest challenge: nowadays about 1.6 Mil people, or a quarter of the world's population, live in countries that have insufficient water supplis (see Figure 1 above).  

As illustrated by the concept of “Tragedy of the Commons”, the "Commons" which include the atmosphere, oceans, rivers, forests, fish stocks and national parks are shared resource, that should be exploited in connection with sustainable development, meshing economic growth and environmental protection, as well as in the debate over global warming.  The tragedy of the commons is an example of emergent behavior, the outcome of individual interactions in a complex system which are provoking huge negative "externalities" that nobody is prepared to pay.

If we want to restore and protect the environment against pollution and seamlessly share the burden of efforts among the various stakeholders, it is necessary to define new indicators and tools. EPI ranking is essential to detect in which countries further actions are needed to avoid disastrous environmental issues.

Starting from year 2002, Yale 2014 country EPI rankings along with others ranking from OECD or Siemens (Asian Green City index) are necessary steps to discuss and address these issues.

EPI methodology adjustments in 2014

My blog dated 6 Nov 2012 has presented the EPI 2012 methodology and main results in Asia Pacific. The last EPI 2012 issue’s overall architecture is respected, with 4 levels of indexes or indicators:

(1) Global EPI;

(2) Environmental Health index EH, Environmental Vitality EV;

(3) 9 sub indexes: 3 for EH and 6 for EV 

(4) Twenty indicators with 1, 2 or 4 depending of the sub indexes

More countries are now in the scope with a total of 178 EPI ranked in 2014 against 132 in 2012. The 2014 EPI evaluates 178 countries, with 46 new additions, in large part, from Small-Island Developing States and sub-Saharan Africa.

The overall population ranked is now 6900 Mil in 2014 against 6550 Mil in 2012, which is an increase of 5% over 2012.

Most of the new countries ranked are low Income (18), lower middle income (13) or upper middle income countries (10). There are only 5 high income countries: Bahamas, Barbados, Antigua & Barbuda, Bahrain and Equatorial Guinea.     

The overall method has been adjusted (see Figure 2) in order to address some shortcoming, notably with the following sub indexes or indicators:

-     Drinking, sanitation and water resources weighting are increased from 16% to 28% (+12%) with a new wastewater treatment indicator, which is a major driver of ecosystem water quality;

-     Forest and Fisheries are more or less the same;

-    Air pollution: SO2 pollution has been deleted to use only PM2.5 particulate pollution: Air quality weighting is reduced from 16.26% to 13.32% (-2.94%);

-     Health impact (child mortality) is reduced from 15% to 13.33% (-1.67%);

-     Agriculture is reduced from 5.83% to 3% (-2.83%);

-     Biodiversity is reduced from 17% to 15% (-2%);

-     Climate and energy is also reduced from 17.52% to 15% (-2.52%) with a new index calculation that is dependant on the country's economic development level.

The water increased weighting in EPI is mirroring the water situation following a recent UN report asking urgent actions: water demand is likely to increase by 55% by 2050, with 40% of population living in areas of "severe" water stress. 

Asia will be the biggest hotspot over water extraction, where water sources straddle national borders. "Areas of conflict include the Aral Sea and the Ganges-Brahmaputra River, Indus River and Mekong River basins".

Apart from this adjustment for water, the overall architecture is more or less the same and past comparisons are still very useful.

It is true that it should be necessary – as explained by EPI Yale- to “back-cast” the new definition and weighting of indicators to know exactly the overall variation from the past.

However as explained by EPI Yale not every indicator in the 2014 EPI lends itself to back-cast or trend calculations. But more importantly we can say that the expert judgment EPI Yale is issuing now in 2014 is more pessimistic as compared to the 2012 assessment meanly because of water urgent issues to be addressed.

Figure 2 : EPI 2014 coefficients values and EPI 2012 comparison

Main results over the countries

If we compare the new EPIs and their variations over 2012-2014 – with all precautions discussed above- the following Figure 3 is showing the last two year trend over the 132 countries already ranked in 2012:

-    73 countries have increased their current EPI scores over the last two years : mostly higher income and upper middle  income;

-    59 countries have decreased their current EPI scores over the last two years: mostly lower income and lower middle income.

Large high income countries are progressing: Spain: +32%; Germany: +20%; USA: +20%; Japan: +14%; UK: +12%.

But heavily populated and low income countries are showing a strong reduction: Bangladesh: -39%;; Philippines: -23%; Indonesia: -15%; India: -14%; Brazil: -13%.

Figure 3 : EPI 2014 and trend 2012-2014 over various income-groups; note that only EPI 2012 ranked countries are presented. The above figure is showing: 32 high income OCDE countries 31 with increasing & 1 with decreasing (Taiwan) trends; 15 high-income non OCDE countries: 11 with increasing & 4 with decreasing (Croatia, Uruguay, Lithuania & Latvia) trends; 39 upper middle income countries: 24 with increasing & 15 with decreasing trends; 30 lower middle income countries: 7 with increasing & 23 with decreasing trends; 16 lower income countries all with decreasing trends

If we want to know the overall result over the planet, Yale EPI ranking does not give any clue and we have to make further assumptions as explained hereafter.

Relationship between EPI and income per capita

If we plot each country EPI against its respective income per capita with a logarithm scale for both EPI and income, we have a much better view over the small EPI countries.

We see- which is an amazing outcome- that all countries are more or less aligned in their average  line that best fits the showed income level EPI points with the method of least squares (see table 4).

Nevertheless there is a slight tendency for high income countries to be above the average line.

We see also that EPI’s volatility around the average line is increasing when income level decrease.  

Figure 4 : Curve of 2014 EPI scores plotted against GDP per capita with logarithm scales  

Another aspect is to show among a group of homogenous countries an average EPI value and which countries are doing well or bad at their income level.

For instance (see following Figures 5, 6 & 7) :

Figure 5 : Curve of High Income countries 2014 EPI scores plotted against GDP per capita with logarithm scales; in high income OCDE countries (red signs): USA, Belgium, Taiwan, South Korea and Israel have bad EPI as compared with the average line; France, Canada & Japan are a better EPI but still down the line; Switzerland, Luxembourg, Australia & Germany are much above the line; in high income non-OCDE countries(yellow signs): Equatorial Guinea; Bahamas and Barbados are far down the average line; Qatar, Bahrain & Trinidad & Tobago are not doing well also when compared with the average line; Singapore and United Arab Emirates are much above the average line

Figure 6: Curve of upper middle income countries 2014 EPI scores plotted against GDP per capita with logarithm scales; Grenada, Angola & Irak are down from the average line; China is slightly down;  Hungaria, Serbia and Belarus are well above the line

Figure 7 : Curve of lower middle income and lower income countries 2014 EPI scores plotted against GDP per capita with logarithm scales; in lower middle income countries: Mali, Sierra Leone and Afghanistan are down the average line; Zimbabwe is well above the average line; in lower middle income countries (blue signs) Lesotho & Sudan are down from the average lines; India is less down the line; Armenia and Egypt are well above the average line

Can we define 2013 EPI average value for all the ranked countries?  What is the overall evolution over 2012?

The country's EPI is an averaged value of each part or unit EPI of the country.  

When comparing relative situation of countries such as Switzerland or Singapore (5-7 Mil) with say USA (313 Mil), Indonesia (246 Mil), India (1236 Mil) or China (1350 Mil), we need to factor EPI by the population.

So there is a relationship between EPI and population. This means that to know the overall impact on the planet we need to use the EPI cumulated-weighted average (see Figure 8).

This Figure is showing that during the last two years the average value - or the worldwide EPI- was 45.7 in 2014 after 47.5 in 2012. 

This means that during the 2012-2014 period:  EPI was in average less than the middle value between 0 and 100.

We see also that in the last two years there have not been any improvement but a reduction of 3.8%:  some countries have improved- mostly small populated and high-income countries such as: Australia, Singapore or even USA- but the reduction of the highly populated middle and low income countries is showing a huge deterioration of the overall value.

Figure 8 : Population weighted 2012 and 2014 EPI monotonous curves