Air pollution exposure and impact

Calculating background PM2.5

scenario_pm_calculations uses total distance by mode to compute the total PM2.5 in the scenario. First we compute the emission inventory in terms of fractions. If the confidence is 1 then we use the inventory directly. If P<1, we sample the emission inventory fractions as is (https://www.overleaf.com/read/mrjtkhffzfzr).

Then we multiply vehicle distance by vehicle emission factor (where the emission factor is the emission inventory (fraction) divided by distance at baseline).

This table is augmented with rows corresponding to any vehicle missing from the trip set. For Accra, these modes include ''big trucks'' and ''other''. The baseline row is equal to the emission inventory fraction. For each scenario, it is scaled. We take the sum over modes to get a scalar for emissions in scenarios. The background PM2.5 in each scenario is the sum of the transport component and the non-transport component.

Calculating PM2.5 per person

Individual exposures to PM2.5 are calculated using the background PM2.5 and the trip sets. There are three major components to daily exposure: one, a person's total inhalation off road; two, a person's inhalation on road in a vehicle, and, three, a person's inhalation on road while undertaking active transport. Each category has an amount of background PM2.5 and a ventilation rate which together inform overall exposure.

The ratio of exposure off road to that on road is a function of total PM2.5, defined in https://www.overleaf.com/read/mrjtkhffzfzr (Goel, 2015). This defines the exposure of a person in an open vehicle (i.e. pedestrian, cyclist or motorist), and it is used to calculate in-vehicle exposure, assuming an exposure with the window closed, and the proportion of vehicles having closed windows. The exposure in a subway is constant and not dependent on the road ratio.

Ventilation rates are calculated for each mode, assuming a base-level inhalation rate, and the mMETs for the mode (Ainsworth compendium 2011 sites.google.com/site/ compendiumofphysicalactivities for walking and cycling). Then the air inhaled during travel per person is the sum over their travel, and the rate of PM2.5 inhalation during travel is a function of the exposure when travelling.

The air inhaled when not travelling is the base-level ventilation times the time spent not travelling. Together, the PM2.5 exposure is calculated as the total PM2.5 inhaled per hour (https://www.overleaf.com/read/mrjtkhffzfzr).

AP--disease dose--response relative risk

gen_ap_rr uses each person's exposure to PM2.5 to compute their relative risk of five diseases (IHD, lung cancer, COPD, stroke, LRI), using curves parametrised by four disease-specific variables (Burnett, 2014). Of the five diseases, two (IHD and stroke) have parameters specific to age groups starting at age 25. (For any person of age lower than 25, we set the relative risk to 1.) The other three (lung cancer, LRI and COPD) have one set of parameters for all ages.

The curves are in the form of samples of the set of four parameters. We model the densities of these samples (using a quantile for parameter 3, kernel density estimation for parameter 2, and GAMs for parameters 1 and 4) in order to draw either the median or random samples via their quantiles. (The four parameters refer, in numerical order, to alpha, beta, gamma and tmrel in Burnett (2014).)

From these parameters, the relative risk of mortality is defined as in Burnett (2014) as RR = 1 + alpha * (1 - exp( - beta * (AP - tmrel) ^ gamma ) ).