How is this going to work?

[mpadge]

Ping @Robinlovelace @layik @eyesofbambi So it turns out that all the EmissV package does is take a static estimate of total emissions for a given area and distribute that into raster cells according to total lengths of roads in each cell. This approach is presumed good enough for them to implement as the spatial extension of the really impressively detailed base work here. And yet ... it is ridiculously simplified, and we can certainly do much, much better with osmdata and dodgr. The question is: is it worth it?

A much more realistic spatial extension would be just:

1. Use dodgr to route a layer or two of trucks/lorries/heavy-goods-vehicles, cars, maybe motorbikes;
2. Aggregate the dodgr flows into raster cells; and
3. Proportionally distribute observed pollutant values according to the total flows in each cell.

Dead easy, and would or should be some unquantifiable yet enormous degree more accurate that simply distributing according to total lengths of roads.

Note, however, that what we and by extension the WHO would ultimately desire would be a way of identifying local populations at risk or something like that - or equivalently, local populations likely to benefit the most from whatever interventions or modifications might be devised or intended. This would be possible by multiplying for each grid cell the expected densities of active transport (per capita and per unit length, time, whatever) by pollutant densities. High values then simply reflect proportionally high exposure to pollutants. Now the catch in all this: The entire procedure of EmissV is nothing other than simple linear scaling and aggregation. This is thus no real need whatsoever to do this actual aggregation into raster cells; rather all we need is a means of smoothing or interpolating the vehicle flows across space (through kriging or whatever). Those smoothed values can then be mapped directly back on to the pedestrian or cyclist flows, and the whole thing can remain in far more spatially detailed vector form.

Thoughts?

[layik]

Hi @mpadge, glad you are revisiting this as I did not get a chance to say that was my understanding of that repo. Once you have something for me to either work on/with you or just consume, happy to add it into ATT.

[Robinlovelace]

Will be great if we build on what they've started. Good to have a baseline model to compare against. Will be amazing if we can get a better spatial extension. There's high demand for that.

[Robinlovelace]

the whole thing can remain in far more spatially detailed vector form.

I do think raster heatmap output is highly desirable though. As discussed with Layik my main issue with most maps of air pollution is that they are too vectory (just points).

[mpadge]

Agreed!