ericnost
commented
2 weeks ago First, look at options for creating a dasymetric map to ensure a more accurate mapping of population distribution.
Both of the following employ a 6 cell rectangle majority smoothing algorithm (focal statistics) to clean up "noise" in the raster values:
Approach 1 based on Land Cover 2020 (https://open.canada.ca/data/en/dataset/ee1580ab-a23d-4f86-a09b-79763677eb47/resource/81252d30-5102-46db-a9c5-6ab1ccd5dcd7) for all of Canada, but with only one "urban" class
Approach 2 based on AAFC's 2020 land use (https://agriculture.canada.ca/atlas/data_donnees/landuse/data_donnees/tif/2020/). Only covers S of 60N but has more class precision.
Visual assessment: the land use approach (2) seems to pick up more settlement, especially in more rural areas.
Approach 1 in greenbelt:
Approach 2 in greenbelt:
However, I think Approach 2 picks up more golf courses and quarries, for instance (but both are bad in this respect)
Approach 1 near Caledon:
Approach 2 near Caledon:
Coverage (share of DA in the buffer) vs ratio (share of the buffer the DA accounts for)
Look at this in urban, rural, suburban contexts