crstn
commented
7 years ago It turns out there are a bunch of cells with 1 or less persons in it that are allocated to an SDEI city or a GRUMP urban extent. So a selection of cells to exclude could be all cells with 1 or less persons, minus those in (buffered?) SDEI cities / GRUMP urban extents. This would drastically reduce the number of cells we are dealing with (by more than 85%), as the output of Check2010Zeros.py shows:
# Cells
730,944,000
# Cells without population (<= 1 persons / cell)
632,620,061
# Cells ON LAND without population (<= 1 persons / cell)
114,232,528
# Cells ON LAND that are WITHIN an SDEI city, but without population (<= 1 persons / cell)
359,657
# Cells in GRUMP urban extents, but without population
287,774
# Cells that are WITHIN an SDEI city, but not within an GRUMP urban area
525,999
pmarcotullio
It would make a big difference if we would just take certain areas out of the simulation because they are essentially uninhabitable. So I had the idea that if we take the areas with zero population in 2010, that might be a good starting point. It turns out, though, that there are not as many regions as a I thought – in the attached image, the white cells are cells on land with 0 population. The Sahara, Central Australia / Greenland, the Amazon, etc. are all not showing up here and quick check confirmed that the cells here have all been assigned very small population numbers in the 2010 GRUMP population grid that we start from.
If we take this into account and pick all cells with 1 person or less, the result is showing quite dramatically that the vast majority of cells are (almost) uninhabited: