Some travel parameters, some AP exposure parameters, some 'what if' scenarios

[rahulatiitd]

• Uncertainty to implement for c. 5 parameters (will get new tab) building on Rob’s voi methods o Safety in numbers normal dist 2 sd= 15% relative change or Rahul to advise o PA RRs use the uncertainty from the meta-analysis – this will underestimate true uncertainty though o AP RR uses the uncertainty from the RRs o Travel patterns- assume motorbike baseline levels log normal 2 sd 1-4% mode share o Walking associated with public transport assume log normal dist roughly 2 sd 5-15 min

• What if /sensitivity (will get new tab) could include e.g. o Cleaner fleet- Rahul to advise o Improved safety- e.g. halve baseline number of deaths o More chronic disease e.g. double CVD and cancer o Lower background AP e.g halve non transport concentrations o Lower background PA e.g. halve non transport PA

[rahulatiitd]

@markotainio We need to decide if we want to use Goel et al's estimate of in-vehicle exposure based on background concentration. What I have done for Accra and what I suggest we should as we go ahead with other case studies is: 1) Divide the modes into two categories-- open and closed based on their ventilation status. AC cars and metro trains (well, Delhi for example is Air conditioned, Accra has no metro, London is not air conditioned) are closed and all other modes are open. 2) For cars in Accra, I observed from Google Street view that about half were closed window and half open window. 3) For open modes, we use Goel's curve to get the in-vehicle exposure ratio which effectively makes all open modes to have same exposure ratio, and for closed modes we use a value of 0.5 (thats what I found in Delhi, we could see Audrey's meta-analysis on this and revise) 4) For cars, then I averaged the value of 0.5 and the value I get from the curve (because they have equal share in Accra)

I would like to get your input on this.

[rahulatiitd]

@usr110 @JDWoodcock Walking associated with public transport assume log normal dist roughly 2 sd 5-15 min For this, we should use the empirical data we have to parametrise the distribution-- like UK survey where they look at short trips, as well as what we have from Delhi

[rahulatiitd]

@usr110 @leandromtg @robj411 and Rahul met today and it has been decided that we will use the current version of ITHIM-R that has been developed for Accra and edit the code to include uncertainty analysis. Also recommend that we share the relevant material on this page rather than on emails. @JDWoodcock @markotainio

[markotainio]

@rahulatiitd Your approach using Goel et al. method is ok for me. I assume the new meta-analysis from Audrey will have similar form so updating for that should be easier, when it will be available.

Dividing fleet to open and non-open is good improvement. However, I'm slightly worried that this will increase data requirements beyond what we can provide in many settings. Also, this could create issues in cities/areas where there is strong seasonal air pollution patterns, which also drives open/closed use of windows. We can implement it, but maybe as an optional parameter? Or at least to provide some average values to work with, if data is not available?