Reducing Airborne Particulate Matter Concentrations: The Problem of Non-Linearity

Reducing exposure to airborne particulate matter is a high priority. The phenomenon of non-linearity, in which concentrations of some components of particles which are formed in the atmosphere do not decline as quickly as their precursor gases can create major difficulties for pollution control.

The WHO has recently recommended an air quality guideline for PM2.5 (fine particles suspended in the atmosphere) of 5µg/m3, a concentration well below that prevailing in most countries, and more than an order of magnitude lower than concentrations in many of the more polluted cities of the world.  This recommendation, designed to protect public health, places pressure on governments to implement measures which will reduce airborne concentrations by emissions reductions.  Such reductions need to be as cost-effective as possible, which requires that there is knowledge firstly of the sources of PM2.5, and secondly an understanding of the relationship between emissions and airborne concentrations.  For primary pollutants, which are those emitted directly into the atmosphere, linear rollback can be assumed.  This implies that an x% reduction in emissions will lead to an x% reduction in airborne concentration attributable to that source.  This generally appears to be the case in practice, with examples from the UK being reductions in sulphur dioxide (largely from power stations) and lead (used as a motor fuel additive).

Not all air pollutants arise from direct emissions.  An example is nitrogen dioxide (NO2).  While some is emitted directly, the major proportion is secondary, forming in the atmosphere from oxidation of nitric oxide (NO).  As a consequence of this formation mechanism, while concentrations of NOx (the sum of NO and NO2) are subject to linear rollback, nitrogen dioxide concentrations decline much less than emissions of NOx, a phenomenon termed non-linearity.

In the case of airborne particles (including the PM2.5 fraction), a major portion of the mass is contributed by secondary particles.  This secondary mass often accounts for more than 50% of PM2.5 mass, and is constituted by nitrates, sulphates and secondary organic aerosol (SOA).  Figure 1 shows the major chemical component composition of PM2.5 measured in sampling campaigns in London, Beijing and Delhi.  Figure 2 gives a source-related breakdown measured in London, including the split between primary and secondary organic aerosol.  The secondary nitrate, sulphate and organics are formed in the atmosphere by oxidation of nitrogen dioxide, sulphur dioxide and volatile organic compounds respectively.  The formation processes are complex and may be non-linear. If the oxidant is present at low concentrations, then the oxidation process can be oxidant-limited, in which case oxidation slows or ceases altogether as the concentration of oxidant approaches zero.  Some oxidation processes occur in the aqueous phase in cloud droplets. There may be mass transfer limitations to the entry of the precursor gas into the particles, or the oxidation reaction may slow as the pH declines with the formation of an acid such as sulphuric acid.  The consequence is non-linearity in the formation process of the secondary product, which is reflected in non-linearity in the rollback as primary pollutant concentrations decline.



Figure 1: Major chemical component composition of PM2.5 collected during winter campaigns in London (North Kensington), Beijing and Delhi.  From Airborne Particulate Matter by Roy M. Harrison is licenced under CC BY 4.01.




Figure 2: Source contributions to PM2.5 at North Kensington (%) derived from application of a Chemical Mass Balance model (data from Yin et al., 2015).  From Airborne Particulate Matter by Roy M. Harrison is licenced under CC BY 4.01.


Given this complexity, how can the influence of a reduction in primary pollutant emissions upon secondary pollutant concentrations be predicted?  Numerical models of the dispersion and oxidation of primary pollutants are helpful, but can be misleading.  Current knowledge of the formation processes of nitrate, sulphate and SOA is incomplete, and due their complexity cannot all be included fully in numerical models which also take account of atmospheric dynamics.  Consequently, the oxidation processes have to be highly simplified and may not work well in the models.  This means that measurements made in the atmosphere are very valuable, although long time series of consistent data are needed, and the trends are affected by weather as well as emissions changes.  Consequently, time series of ten or more years of data measured by a consistent technique are desirable, but rarely available.


Two such datasets are available from UK networks, and the study by Harrison and coworkers2 seeks trends in these datasets with a view to identifying non-linearities.  Despite the availability of long runs of data, it is surprisingly difficult to identify non-linearities with confidence unless they are very large.  The UK datasets provide evidence for non-linearities between NOx emissions and airborne nitrate, but in the case of sulphur dioxide emissions and sulphate, there appears to be linearity.  An example of the data is in Figure 3.  For organic aerosol, the data base is less and the analysis more complex because of a contribution of natural, biogenic precursor gases emitted by trees, but there is little sign of non-linearity for the anthropogenic organic compounds.




Figure 3: Trends in nitrate and sulphate from the AGANET sites.  (a) UK emissions and concentrations of SO2, and sulphate concentrations, and (b) emissions and concentrations of NOx, and nitrate concentrations, from 2000 to 2020 from the AGANET data.

As part of this research, the air mass back trajectories were calculated, and for air arriving at a site in south-east England, 42% of the mass flux of sulphate, 55% of nitrate and 35% of SOC are associated with air masses entering the UK from the European mainland, which adds an important international dimension to abatement policy.



  1. Airborne Particulate Matter, M. Harrison, Phil. Trans. R. Soc. A., 378, 20190319. (2020). 

  1. Harrison, R.M., Beddows, D.C.S., Tong, C., Damayanti, S. (2022). Non-linearity of secondary pollutant formation estimated from emissions data and measured precursor-secondary pollutant relationships, npj Clim. Atmos. Sci., 5, 71, (2022).

Please sign in or register for FREE

If you are a registered user on Nature Portfolio Earth and Environment Community, please sign in