, ,

SummaryIn this article, we collected the data of the air quality in Beijing over the past 5 years.We assumed that the significant changes or improvements in air quality were related to China’s two top Annual Sessions – the Chinese People’s Political Consultative Congress (CPPCC) and the National People’s Congress (NPC).Besides, there were several news and current affairs affected the air quality in Beijing. We will also take a closer look at its air quality at that time to see if there are significant pollutants reductions during those periods.

Four years ago, the Chinese premier, Li Keqiang, said at the National People’s Congress and many more Chinese citizens watching live on state television, “We will resolutely declare war against pollution as we declare war against poverty”.

Since then, cities have cut concentrations of fine particles in the air by 32 percent on average. Just a few months before the premier’s speech, the country has released a national air quality action plan to require all urban areas to reduce concentrations of fine particulate matter pollution by at least 10 percent. As the capital city, Beijing was required to reduce pollution by 25 percent, of which the city set aside a massive $120 billion yuan to achieve this target.

All the statics in this project comes from Online China’s Air Quality , which is a public platform monitoring and analyzing China’s air quality covering statistics from 367 cities with AQI, PM2.5, PM10, S02, N02, O3, CO, temperature, humidity, wind scale, wind direction, satellite cloud picture and so on.

In this project, we mainly focus on Beijing’s air quality including AQI, air pollution category, and main pollutants such as PM2.5 during December 2013 to March 2018.

Less than half of the days in 5 years were pollution-free in Beijing


Air Pollution Category in Beijing from 2013 to 2018

This chart shows the percentage of Air Pollution Category in Beijing during last 5 years. We can see that nearly half of the days the weather was good or excellent. Beijing citizens spent half of the time in air pollution in the past five years. The days which the air was lightly polluted were approximately one-fifth of the hole 5 years. One-third of this period, the air is more than moderately polluted.

Air pollution in Beijing decreased after the  documentary “Under the dome” released


CCTV journalist Chai Jing made a documentary “Under the Dome” about the air pollution in Beijing and released it on February 28, 2015. The documentary made a stir in China, which lead Beijing to pay more attention to air quality. These two pie charts show the difference before and after the documentary released. We can see there is a moderately improve both of the “good” and “excellent”. Besides, air polluted days decreased.

PM2.5 is not the only leading factor in air pollution


Average AQI per month in 2013-2018

The highest AQI appears in December in 2015 in this bar chart. On December 8, 2015, Beijing used red alert on air pollution, which is the highest of a four-level alert system instituted two years ago, and has not been used in the city before.

At that time, Beijing plans to either shut down or curtail operations at dozens of steel plants from November 2017, over the next five months under an aggressive action plan to reduce winter pollution in Beijing and its surrounding areas.

It works in the following two years, as we can see the decrease of AQI from 2015 to 2017 in December.


Average PM2.5 per month in 2013-2018

Remove the extreme case in December, the air quality in summer period is the worst especially in Jun and July. When compared with the index of PM 2.5 during these time, the trends in total are similar, except the lowest concentration of PM2.5 in summer but the worst air quality at that time.

It shows that although more and more attention on PM2.5 recently, it is still not the leading factor in air pollution.

What’s more, in this five years, the AQI in January to April decreased faster. While in other months, especially in summer period, the improvement of air quality is limited, even become in some months like May, June and September.

AQI shows declining trend from 2014 to 2018


AQI & Major Air Pollutants Trends between 2013 and 2018

The above line graph illustrates the trends of the air quality index (AQI) as well as major air pollutants including PM2.5, PM10, Sulfur Dioxide (SO2), Carbon Monoxide (CO), Nitrogen Dioxide (NO2) and 8-hr Ozone (O3_8h) in Beijing between 2013 and 2018. The graph is created based on the average figures during the year for each of the air pollutants. Taking out the figure in 2013 due to insufficient statistics during that year (only December data is collected in 2013), despite not significant, there is a declining trend from 2014 to 2018, which is reflected in the AQI as well. It is worth highlighting the declining trends is relatively observable for PM2.5 and PM10 whereas Nitrogen Dioxide and Ozone remain relatively stable during the period. According to the graph, the yearly average AQI was down by about 15% from 2014 to 2018.


Weekly Average PM2.5 and PM10 between 2013 and 2018

2013 and 2018. Both graphs are similar in terms of patterns and trends except there is a significant surge in PM10 around May 2017 and March of 2018 which are inconsistent with the corresponding figures in PM2.5.

Overall, the PM2.5 and PM10 patterns are similar over the period with relatively high levels of PM2.5 and PM10 during winter closer to the end of the year and lower levels during summer in the middle of the year. Both PM2.5 and PM10 fluctuate significantly over the period but the fluctuations in 2017 were less intensive compared to previous years. Lastly, despite the moderate decrease over the period, levels of PM2.5, which poses the greatest risk to human health among the major air pollutants, were still far above the recommended standards by the World Health Organization (WHO) at 10 micrograms per cubic meter[1].




[1] WHO releases country estimates on air pollution exposure and health impact

Codes and data:

Interested readers can download codes and data here: Group 9 – Air quality of beijing

Notes from Lecturer

The headline, summary and main discussions do not reconcile with each other. For example, the summary points out

“We assumed that the significant changes or improvements in air quality were related to China’s two top Annual Sessions – the Chinese People’s Political Consultative Congress (CPPCC) and the National People’s Congress (NPC).Besides, there were several news and current affairs affected the air quality in Beijing. “.

But the article does not try to validate the assumption. The readers can expect a time series (polyline chart) annotated with major events/ conferences like “Liang Hui”, “Under the Dome”. In that way, we might see a relationship. If the (expected) drop of AQI is not apparent, one can compare horizontally, with same day of month, same week of year, same month of year, etc. See one past post on the same topic that correlated with some news events.

One glitch to note about the visualisation is the figure captioned “AQI & Major Air Pollutants Trends between 2013 and 2018”. The y-axis is not explained, especially different components may have different unit. If the story point is relative increase/ decrease for one indicator, it is better to set one year as “unit 1” so the other values will be comparable ratios. Or the larger valued indicators will hide the trend of smaller valued indicators. In this figure, it is hard to tell how CO (purple line) changes.

One major advantage of time series is the ability to aggregate on different granularity. The “.resample” function can be used to group data records at various levels, the “.aggregate” function can generate useful metrics for each group. The most common aggregation is “average” but it is not always the best choice. In general, average can be misled by skewed data distribution, so people look at median or other quantile values. Specifically for this case, not all the data points are interesting. The effect on people’s health does not change much, if the AQI fluctuates in small value ranges. This is similar to other health/ security related metric. For example, we require the density of harmful substance in sold drinking water to be smaller than certain number, but do not quite care how much smaller one bottle is to another bottle, as long as two bottles are both away from the threshold. In this study, a more meaningful metric may be “# of days with AQI > 100”, where “100” is an example threshold. This ways hides some distracting details and may have better chance to lead to insights.

— Pili Hu (April 15, 2018)


Author/ Shen Yue; Xu Manning; Wong Pui-lam

Editor/ Yucan Xu