Post Expo pollution in Shanghai: an emissions analysis

The Shanghai World Expo ended on October 31st, and I, like most of my fellow Shanghai denizens, noticed an immediate and frightening deterioration in the air quality. Over night, it seemed, things went from relatively clean to crazybad.

China recently launched a site that delivers hourly pollution data for 113 cities, but for prior data, there is but one resource for Shanghai air pollution data, and I’ve decided to crunch those numbers to see if there are any demonstrable trends, and if I could validate what we all see and feel.

The aforementioned resource delivers 10 years of data, from 2000-2010, for Particulate Matter 10 microns and larger (PM10), Sulphur Dioxide (SO2) and Nitrogen Dioxide (NO2). I will let the graphs speak for themselves, and only explain any math I did or if I have a funny joke.  Click on any graph to see a much larger and more detailed graph.

10 Year Analysis

First up is a series of graphs featuring 10 years of daily readings, with a 200pt (day) moving average trendline, which sort of smooths out the peaks.  First is Particulate Matter 10microns or larger (PM10)

Graph of daily Particulate Matter (PM10) levels in Shanghai, 2000-2010
Graph of daily Particulate Matter (PM10) levels in Shanghai, 2000-2010

The highest recorded PM10 levels were 500, on April 2nd, 2007, and March 21st, 2010. I think that, at least in Shanghai, ‘Crazybad’ is not a value on the scale.

Sulphur Dioxide (SO2) is next:

Graph of daily Sulphur Dioxide (SO2) levels in Shanghai, 2000-2010
Graph of daily Sulphur Dioxide (SO2) levels in Shanghai, 2000-2010

The highest recorded S02 levels were 113, on December 15th, 2007, and 112, the day before, December 14th, 2007.

Note that for the NO2 graph below, the data available from 1/1/2000 to 6/1/2000 was for NOx, which is the combination of NO and NO2, which was a much higher number and skewed the graphs.

Graph of daily Nitrogen Dioxide (NO2) levels in Shanghai, 2000-2010
Graph of daily Nitrogen Dioxide (NO2) levels in Shanghai, 2000-2010

The highest recorded NO2 levels were December 24th, 2003, at 160, and November 22nd, 2003, at 157.  This is certainly the most seasonal pollutant.

Year on Year Analysis

Those trendlines help a little bit, but how about we compare years against years? This series stacks all 10 years of readings together.  The bright yellow line represents 2010. As before, click any graph to biggify.

Year on Year comparison graph of PM10 emissions in Shanghai, China, from 2000-2010
Year on Year comparison graph of PM10 emissions in Shanghai, China, from 2000-2010
Year on Year comparison graph of S02 emissions in Shanghai, China, from 2000-2010
Year on Year comparison graph of SO2 emissions in Shanghai, China, from 2000-2010
Year on Year comparison graph of NO2 emissions in Shanghai, China, from 2000-2010
Year on Year comparison graph of N02 emissions in Shanghai, China, from 2000-2010

I suggest you click on a few of these graphs and notice that the Y value scales are much different (PM10 by far the largest). In these graphs it is easier to see the 2010 performance vs previous years: if you can see color above the yellow line, 2010 was better than at least some other year, if there’s color below, worse.  But these graphs still look like Jackson Pollack paintings with all those squiggles. Basic Statistics ho!

Year on Year 30 day trendlines

In this series, I’ve graphed the 30day moving average trendlines for all 1- years.  I’ve not wrapped the data, each year is calculated from 1/1 to 12/31 (note the X axis starts at 1/31).

Year on Year 30 day moving average comparison graph of PM10 emissions in Shanghai, China
Year on Year 30 day moving average comparison graph of PM10 emissions in Shanghai, China

I see a pretty hefty spike in emissions starting in november 2010, don’t you?

Year on Year 30 day moving average comparison graph of S02 emissions in Shanghai, China
Year on Year 30 day moving average comparison graph of S02 emissions in Shanghai, China

2010 was still a pretty good year for SO2 emissions.

Reminder for the next graph:  NO2 data is unavailable from 1/1/2000 to 6/1/2000, hence the flatline.

Year on Year 30 day moving average comparison graph of N02 emissions in Shanghai, China
Year on Year 30 day moving average comparison graph of N02 emissions in Shanghai, China

these numbers clearly show that 2010 was a pretty good year, in comparison to the previous 9, until the post expo trash-burnarama. But let’s smooth those graphs out even more.

Year on Year 90 day trendlines

90 day moving average  trendlines really smooth out the peaks, but still offer an accurate visual comparison.

Year on Year 90 day moving average comparison graph of PM10 emissions in Shanghai, China
Year on Year 90 day moving average comparison graph of PM10 emissions in Shanghai, China

Now this looks pretty demonstrative. By this data & the graph, the PM10 emissions during the last quarter of 2010 was worse than 7 or 8 out of the last 10 years, depending on when you look. Also, the Expo effect is pretty clear to see.  Don’t see it? Here, I’ll point it out for you.  Here’s the same graph again, with some cheeky notes:

Year on Year 90 day moving average comparison graph of PM10 emissions in Shanghai, China, with pithy arrows!
Year on Year 90 day moving average comparison graph of PM10 emissions in Shanghai, China, with pithy arrows!

Does the SO2 graph look similar?

Year on Year 90 day moving average comparison graph of SO2 emissions in Shanghai, China
Year on Year 90 day moving average comparison graph of SO2 emissions in Shanghai, China

I don’t see the same clear drop/increase from the expo influence, but it is the most parabolic line on the graph.  Also it shows what we expected based on the daily and 30 day graphs, that at least for this time period, 2010 is one of the best, if not the best year for SO2 emissions.

The last reminder, I promise: There is no NO2 data for the first 1/2 of 2000, hence the flatline in the next graph.

Year on Year 90 day moving average comparison graph of NO2 emissions in Shanghai, China
Year on Year 90 day moving average comparison graph of NO2 emissions in Shanghai, China

2010 started well for our NO2 numbers, but as we can see, starting around the end of the Expo, the air quality climbed well into the middle territory, comparatively.

Conclusions

First, the obvious caveats:

  1. I don’t know how/when/where this data was gathered, so we can’t make any claims to the veracity or accuracy of the data itself. we must simply run with what they publish (until we convince the US Consulate to mount some equipment on their roof as the embassy does in Beijing).
  2. This data source does not perform particularly stringent or useful tests: they don’t measure PM2.5, which measures much smaller (and more harmful) particulate matter. They also only measure SO2 and NO2, instead of coupling them with their equally damaging cousins, NO and SO (the combinations are typically called NOx and SOx).

Because of these clear issues, I don’t want to draw any huge sweeping conclusions about much of anything, but nonetheless, I think the data shows a few key things:

  1. Year over year, all emissions trend down, with a few bumps and bruises, including the current walloping we’re all taking in the Post- Expo air.
  2. Post-Expo air quality has been very bad, and the curve of the change suggests that the various moratoria on industrial activities during the Expo were successful in the short run, but we’re sure paying for it now.

All of these things we knew, or at least intuited, before I crunched these numbers, but at least now we have some pretty pictures.  If you’d like the full data set, email me.

About Mark Englehart Evans

Sustainability minded, China based technologist focused on emerging mobile technologies and digital trends that shape our rapidly evolving world. Currently with Blue Hive APA. Techyizu co-founder, cyclist & rabid Portland Timbers FC supporter. Your new best friend

  • Anonymous

    Great data visualizations, Mark! Have you thought about making some bar charts for yearly averages of the data? It’s a little hard to pick out from the year on year data to determine whether emissions are truly going down over time. It’d be great to take those moving averages and calculate a yearly average – then compare this to what’s being reported in the Official Statistical Yearbook for Shanghai as well as the Shanghai Official Annual Statistical data for each of those years. I might have some time and some of this data myself … :)

  • Pingback: Tweets that mention Post Expo pollution in Shanghai: an emissions analysis | Mark Englehart Evans -- Topsy.com

  • Toffler

    Fascinating, Mark. But I guess the question is, is China learning to manage its pollution output in the long run or is the constant coming online of new source outweighing any attempts to reduce emissions?
    The second link in the article about hourly pollution data doesn’t work. Can you fix it?

    • http://www.markenglehartevans.com/ Mark Englehart Evans

      link is fixed.

      As for the balance between new capacity and emissions control, I don’t believe that new grid capacity is more than 20% of total emissions. Cars and industrial processes are a more serious (and harder to control) issue.

  • Louie

    Hi Mark,

    Just stumbled onto your blog — seeing the trend data graphically like this is very helpful. Few comments/questions:

    1. What units are you using on the y-axis? API ? It doesn’t really impact the time-comparison analysis you’re doing here, but I’ve been looking for a data source that doesn’t have to be converted from API

    2. There’s been much said of Expo impact on air quality. Looking at 2010 in isolation, the numbers would support this. But, this effect is duplicated in almost every past year, suggesting a cyclical effect. I know in Beijing, this is related to the cold snap hitting and power generation plants kicking in to overdrive, seasonal hay burning, but it’s not something isolated to Expo. Since the pattern changes slightly, I would guess that it’s more seasonal (temperature based) rather than a specific regulatory date.

    3. Would it be much work for you to add in the 90-day moving averages for Q1? This would help show cyclicality and might confirm/deny the previous question.

    • http://www.markenglehartevans.com/ Mark Englehart Evans

      Hi Louie!

      1) it is API.

      2) There are seasonal effect on these pollutants, which his expected. The interesting thing is that the expo pollution numbers are lower than previous years.

      3) yeah, I got lazy. I’ll make the data set available if I don’t have time to finish wrapping the data around. I’m not sure of the best presentation for this — if you are using moving averages that start at 1/1, it’s not a data set that is within one calendar year. Anyway, that’s the minor issue that kept me from doing the data wrap in this iteration of the graphs.