ISSS608 Assign Pu Yiran-Task 2

   Task 1: Spatio-temporal Analysis of Official Air Quality

Background & Introduction

Data Preparation

Task 1

Task 2

Task 3

Get To Know About Sensors

Where did sensors cover

After decoding geohash into corresponding longitude and latitude, we are able to locate all the geohash onto map, each of which represents a sensor’s location. As shown in the first graph, there is one remote point which could be an error and will be excluded in further analysis. In total, there are 538 sensors located across the entire Bulgaria, giving 1,048,574 sensor records of 2017 and 2018.

National-widely, sensors were detecting from 382 locations in 2017 and 528 locations in 2018. In Sofia city, sensors were detecting from 240 locations in 2017 and 308 locations in 2018.

Although the sensors are covering national wide of Bulgaria, most of them are centralized in capital city Sofia, and especially, the central region of Sofia city. A small number of sensors also gathers at Plovdiv, a province near Sofia. At the rest provinces/cities, sensors are evenly distributed.

To perform further analysis on Sofia capital city, a set of all the sensors located in Sofia city is created.

How did sensors work

Although the given time interval of measurement in data is one hour, not all the sensors have been either working well nor giving correct data all the time.

Not all the sensors started working at the beginning — the number of sensors that were working per day was generally increasing during the given time period in data.

Most of the sensors were not working well all the time. As the example given below, even if a sensor was working on a particular day, it might not be working for 24 hours continuously, which makes all the blanks and gaps in the graph.

Even worse, sometimes sensors performed abnormally and gave ridiculous data. Plus, as shown in below, some abnormal status has been lasted for hours and even a few days continuously, which can badly affect some analysis on daily average.

By looking at all the abnormal records, we can find out that most of the error values are the same or in the same range, which can be considered as a systematic breakdown of machines. To make sure the accurate of further analysis, all the error records are excluded.

Analysis of Air Pollutants--PM10 and PM2.5

Insight 1- How severe Sofia has been polluted

From 6th September 2017 to 13th February 2018, the overall average of PM10 and PM2.5 was 50.5 µg/m3 and 26.7 µg/m3 respectively, both of which has exceeded the EU yearly limits (40 µg/m3 for PM10 and 25 µg/m3 for PM2.5 http://ec.europa.eu/environment/air/quality/standards.htm)

In addition, daily average concentration of pollutant PM10 and PM2.5 in Sophia city has a significant increase from September 2017 to January 2018. Daily averaging concentration of PM10 was always much higher than that of PM2.5.

Based on this result, we can conclude that Sofia city as well as Bulgaria has been suffered from heavy air pollution and its pollution level has exceeded EU limit.

Interestingly, concentration of PM2.5 has been always moving with that of PM10, especially when PM10 increased/decreased tremendously, PM2.5 also increased/decreased almost in the same rate. In some sense, PM10 and PM2.5 are highly positively correlated, which can be seen from scatterplot. Consider the characteristics of these two particle, the cause of this pattern could be that PM2.5 is an appendage of PM10.

Knowing the EU limit daily average of PM10 is 50 µg/m3, in this visualization, colour of daily average of PM10 is differentiated from blue to red, representing daily average exceeds 50 µg/m3. Therefore, it is can be seen clearly that before November, daily average concentration of PM10 was all below EU limit, but from November onwards, nearly half number of the days had concentration of PM10 exceeding EU limit.

Even worse, most of the days with excessive concentration had critical heavy concentrations, which exceeded EU limit by three times even five times.

A set of days when daily average concentration of PM10 exceeded 50 µg/m3 is created.

Insight 2- Where were pollutants distributed in Sofia city

Let’s look at the most heavily polluted two months in given data—December 2017 and January 2018, after excluding records from error machines. Daily average concentration of PM10 and PM2.5 measured at each location are mapped onto colour and size respectively in the trellis map.

From certain days that were heavily polluted, we could clearly observe some patterns of the distribution of PM10 and PM2.5, as shown in trellis map:
1) The relatively heavily polluted area is north central region
2) On certain days severely polluted, pollutants spread to central Sofia from north central region.
3) Relatively, southern regions are less polluted even on certain severely polluted days.

For the rest months, north central region also witnessed relatively higher concentration of PM2.5 and PM10. According to this observation, the distribution of pollutants could be time independent.

Insight 3- Reveal areas with excessive pollution

To find out the source area of pollutants in Sofia city, which pulled up the average pollution level of Sofia city, locations coloured in red had PM10 higher than 50 µg/m3 on that particular day, while others coloured in light green had PM10 lower than 50 µg/m3.

Similar patterns with above can be observed:
1) North central region and central region are heavily polluted.
2) Certain places at southern region witnessed high PM10 on certain days, but they are in minority.