Infill areas up to a regional target

[brynpickering]

What can be improved?

Area infilling is a secret sauce in some regions where individual facility data is sparse but land use data is available.

Since we tend to use OSMOX in tandem with census / other geospatial statistical data, we can use area infilling to reach the number of facilities expected from that data (e.g. region A has X households in the census, so we need X "home" facilities in that region).

From #47 @val-ismaili raised this:

we want to be able to let the user input a reference dataset i.e. count of households in msoa. Ideally you have more detailed data but i think thats a good base point that should be common across most places. Presuming we input no additional landuse and just use OSM land-use, we can get OSMOX to check an intersection between the reference dataset and landuse. The area of residential landuse that is within an msoa polygon should have X houses, but osmox only finds <X, so we infill the difference in count within that portion of landuse within the msoa. There may be more than one resi landuse polygon in the msoa so we should look at totals.

The plan would be to use relative area of target areas to handle distribution of required facilities.

Open questions:

1. how do we handle there being too many facilities compared to the reference dataset?
2. how do we handle the need to pack facilities too tightly into areas to reach the reference target? Do we expand beyond target areas to reach our target?

Version

v0.2.0