Ambient air pollution concentrations

The resulting emissions of all precursor emissions of PM2.5 in ambient air, such as primary PM2.5, SO2, NOx, NH3 and volatile organic compounds (VOCs), are input into an atmospheric dispersion model to compute annual mean concentrations of PM2.5 across the globe. GAINS employs reduced-form source-receptor relationships derived from the EMEP atmospheric CTM (see Simpson et al., 2012 [11]) hosted by the Norwegian Meteorological Institute since the beginning of the EMEP Programme in 1979. Resulting concentration fields distinguishing about 6000 individual cities with more than 100, 000 inhabitants, are then compared with air quality standards, and the corresponding population exposure is computed for the population distribution assumed in the socio-economic projection. In national and local versions of GAINS, different atmospheric models have been used, such as the Flexible Air quality Regional Model (FARM) in GAINS-Italy, AERMOD model in GAINS-Vietnam, or the Comprehensive Air Quality Model with extensions (CAMx) in GAINS-JET for South-Africa.

The atmospheric transfer coefficients used in GAINS are usually based on full year simulations of the EMEP CTM. Those coefficients describe the relationship between one unit of emissions of a given pollutant emitted in one source region (and possibly one source sector) and the related change in ambient concentrations of pollution (e.g., \(PM_{2.5}\) or \(O_{3}\)) on a receptor grid, taking into account the dispersion of emissions and all relevant chemistry in ambient air:

\begin{align} q_j= \delta_j + \sum_r \sum_p E_{i,p} \cdot T_{i,p,j} \tag{1} \end{align}

where:

i, p, j	Region, pollutant, receptor grid cell respectively
\(E_{i, p}\)	Emissions of pollutant p in region i
\(T_{i,p,j}\)	Transfer coefficient for region i, pollutant p, and receptor grid cell j
\(\delta_j\)	Grid-specific constant arising from natural background concentrations, inflow from outside the domain, and linearization effects due to the non-linear atmospheric chemistry
\(q_j\)	Concentrations q in receptor grid cell j

Transfer coefficients are typically derived from perturbation simulations of atmospheric models, where a base case (base) is run, and then reduction runs (red) are performed, reducing emissions of one pollutant from one source region at a time. The response in terms of ambient concentration (or deposition) changes are recorded and used for a linear approximation as follows:

\begin{align} T_{i,p,j} = \frac {q_{base}(j) - q_{red}(j)} {E_{base}(r,p) - E_{red} (r,p)} \tag{2} \end{align}

The receptor grid j is typically a regular Eulerian grid defined by the respective atmospheric model. Its spatial resolution must be suitable for calculating the impact indicators such as population exposure for calculating human health impacts or deposition for calculating ecosystem impacts. However, the suitable resolution depends on the pollutant and the indicator being considered. In GAINS, we follow the approach that the spatial resolution of the underlying atmospheric calculations should be detailed enough so that the population exposure calculated by GAINS within one region is similar to that calculated with a high-resolution atmospheric model:

\begin{align} \sum_j c_{p, j} \cdot pop_j \approx \sum_{j fine} c_{p, j fine} \cdot pop_{j fine} \tag{3} \end{align}

where:

p, j, j fine	Pollutant, receptor grid cell, and receptor grid cell of a high-resolution atmospheric CTM respectively
\(c_{p, j}\)	Concentration of a pollutant p in the receptor grid cell j
\(pop_{j}\)	Population in receptor grid cell j

The details of the atmospheric model version and setups for deriving the atmospheric coefficients differ between individual model interfaces and domains. More information can be found on the GAINS model directly.