Air

How do we know air quality is changing?

Expanding monitoring for more data alone won t help. Need science, statistical methods for legal reporting on air quality trends, compliance with clean air targets

Anumita Roychowdhury

The questions are simple but fundamental: Where is the debate on science of using air quality data to officially report air quality trend and compliance with the national ambient air quality standards in India? How will cities know if their pollution levels are rising or declining?

Air quality monitoring — realtime as well as manual — are expanding across India under the National Clean Air Programme (NCAP). But no one is asking for legally acceptable and scientifically robust method for trend analysis and legal requirement for reporting compliance against the National Ambient Air Quality Standard or the NCAP target (20-30 per cent reduction in particulate matter (PM) by 2024 from 2017 levels). 

If all 122 cities implementing clean air action plans under the NCAP aim for 30 per cent cut in five years, they need on an average 6 per cent annual reduction. But we do not know how this target will be achieved nationally and get reported. This has not been defined yet.

But cities need to know what monitors and locations they should consider; how data should be used and averaged; and what geography or airshed they should represent. How will they leverage manual and realtime monitors and deal with data gaps that plague our datasets? 

After making huge investments in regulatory monitors across India, air quality regulatory agencies cannot claim that data is uncertain and therefore unusable for reporting compliance with the standards.

While quality assurance and quality control of monitoring will require immediate attention and annual maintenance contracts for monitors need accountability, the legal requirement for reporting air quality trends based on standardised methods need to be defined quickly.  

Rise and fall of air pollution whip up emotion and outrage in our cities. But no one knows what’s going on — how action is helping or if at all. How to verify change?

Globally, regulators need to report trends and compliance based on data from monitoring stations. Meteorological events may influence annual and seasonal trends to some extent but control for these must be built in the methodology of reporting. 

How is reporting done currently?

Currently, the Central Pollution Control Board (CPCB) puts out annual average data for criteria pollutants for all monitoring locations and cities based on manual data. Its latest report (December 2019) has put out the data for 2018.

Also city-wise averages for those with populations of more than a million are reported periodically to Parliament, based on manual monitors; the latest was in June 2019 for 2018. 

Reporting trend based on manual monitors is well accepted global practice. But there is huge concern around the quality of manual data as the basic legal requirement for generating manual data are not met in a large number of monitoring locations.

CPCB protocol for manual monitors requires minimum monitoring of twice a week or 104 days in a year. Thus, legally compliance with national ambient air quality standards can be assessed based on data for a minimum of 28.5 per cent of the days in a year.

Even this current benchmark is not met in 73 per cent of PM2.5 manual monitors in India (as reported in CPCB’s December 2019 report). This limited data is then averaged by the CPCB. A city like Kozhikode monitored for only 10 days in 2018. Even Delhi’s manual monitors do not meet the benchmark. 

The objective of monitoring is to establish an indicative trend in pollution levels that can be compared temporally and with other areas. If done well, even fewer monitoring days than 104 can deliver on the objective.

In the United States, the minimum requirement of manual monitoring is every sixth day that helps to cover all patterns of days. But more data samples is always desirable for accuracy. 

On the other hand, realtime data which is more continuous and voluminous, is available on the CPCB portal and annual trend can be generated online if required. But there is no structured official method for using it for establishing long-term trends.

Realtime data is used officially for daily reporting of spatial averages against the national air quality index; that too with a rider that such spatial averaging is not the most scientifically sound method.

Going forward, how will cities — with both manual and realtime monitors, or cities with only realtime monitors for PM2.5 — will select primary monitors and collocated monitors to analyse and report trends?

What is the legal minimum data requirement from realtime monitors? What happens if minimum data requirements are not met? We need these answers quickly. 

What do other governments do? 

That’s what we set out to find out in our new report. We found out that the US, European Union, Canada and Australia have adopted detailed rules and methods for addressing these compelling issues. 

The USEPA has the most elaborate method for trend and compliance reporting as this is a legal requirement. For long-term trend reporting they use spatial averaging of data from different monitors; for compliance with the ambient air quality standards they take the trend in three-year averages for the worst-performing monitoring location to protect vulnerable groups. Realtime and manual monitors are treated on a par.

They identify long-term primary monitors to represent not just the city but also the larger air basin. The compliance with standards has to be established for the entire air basin. They also ensure minimal overlap among monitors in an area. 

More important, the USEPA has elaborate methods for addressing data gaps and data completeness to make data usable. It requires a minimum data availability of 75 per cent in each quarter of the year. If this benchmark is not met they don’t reject that time period but apply data substitution tests based on data of preceding quarters.

If the primary monitor does not produce valid value for a particular day, but value is available from a collocated monitor, it can be considered valid for combined site data record. 

It works for even large data gaps. For instance, data for Corcoran-Patterson in San Joaquin Valley for one whole year was not available as it was gutted in fire. But three-year average calculation methods could still be used for constructing a trend.

Station pairs with corelation coefficient higher than 0.75 are treated as collocated and USEPA treats data from the station with lower average as redundant. Stations in close proximity give identical results.

The European Union on the other hand has the minimum data availability requirement of 90 per cent. They have strong quality control and assurance regime. It is clear that data from all available stations are not averaged to report trends or compliance.

Even Beijing, which has 40 monitoring stations to generate more granular local data, has identified 13 stations for legal reporting on trends.  

What if global methods are applied to Indian data? 

There are concerns that India is not generating enough well distributed air quality data. But it has also not made enough efforts to make use of the hugely accumulated pile of data of 15-minute granularity.

Even after massive investments in 38 realtime monitoring stations and also with a few realtime monitors in existence since 2011-12, Delhi is still guessing if its pollution is rising or falling. 

What will Delhi’s long-term trend look like if we apply USEPA method for data completeness and trend analysis? Conversations with USEPA scientists brought out that while the application of their method to Delhi data is reliable and accurate, a few caveats are required as the infrastructure and data availability in the US is different.

Particularly, when India develops its own method of trend analysis and defines thresholds for data completeness, the trend results may vary. Also if there is large variability in data a statistical test on the location with more complete data is needed to establish the threshold; and a few more. 

Therefore, after deep mining of realtime raw data of 15-minute average for PM2.5 from the CPCB’s online portal ‘Central Control Room for Air Quality Management - All India’ for 2010-2018,  final cleaned dataset includes 3.16 million data points (with outliers and erroneous entries removed).

The USEPA method of data completeness (station-wise and quarter-wise for each year) was applied to the five oldest stations — IHBAS, ITO, Mandir Marg, Punjabi Bagh and RK Puram — for which realtime data is continuously available since 2012. During the earlier years, quarters with more noise and missing data could be addressed with data substitution method. And this gives us a clear indicative trend. 

If trend in three years spatial averages of five stations is considered for long-term trend since 2012-14, there is an indicative drop of more than a quarter; but the city needs a cut of another 65-67 per cent to meet the annual PM2.5 standard.

If the trend in three-year averages of worst polluted monitoring location is considered then the reduction target is more than 75 per cent.

The fact that PM2.5 levels have peaked and the curve has bent is also evident from the year-on-year trend analysis based on realtime data by SAFAR. It shows an about 20 per cent reduction since 2014.

The CPCB has reported to Parliament in 2019, based on realtime data, that PM2.5 has dropped by 15 per cent between 2016 and 2018. After the initial reduction in high gross pollution, the subsequent reduction will get tougher and complex. 

These indicative trends throw up a major lesson for all. Multi-sector action in Delhi that has contributed to this trend is not small; but do not add up to meet the benchmark.

The actions include: Closure of power plants and big industry, natural gas transition in transport and industry, phase-out of old vehicles, drastic reduction in the number of trucks, BS-VI fuels and BS-IV standards, pollution charge on diesel cars and SUVs, trucks and diesel fuel, some action on construction waste and more. These bent the curve.

What will be a significantly more disruptive action for clean energy and technology transition, mobility transition and waste management at a regional scale that can lead to this next big cut? This is also a cue for other cities and regions facing the challenge of big pollution cut. 

Code to decode air quality

Don’t miss the crucial point. Though patchy in some cases there is still a lot of data out there. But without the regulatory adoption of scientific and statistical tools to make them usable a lot of it gets rejected and makes the investment in monitoring wasteful.

Decide what tool to use and make it consistent throughout the country. The CPCB has to define this quickly. Without this it will be difficult to build public confidence in change and action. This is one crucial step towards strengthening compliance regime under NCAP.