-
Notifications
You must be signed in to change notification settings - Fork 1
Description
We're currently computing around 5 times as many points for a world run as we need to!
Currently we reproject all the raw NASA data to the same equidistant 100m scale .bt files. This causes an interpolated filling-in of the sparse data further from the equator. In the extreme case near the poles, the actual scale of the raw data is around 400m. We don't need to do visibility calculations on filler data!
What we should be doing is also defining the optimal scale of each .bt tile, which will range from 92m at the equator to the 400m near the poles. Only then, once we've done the expensive compute, should we add the filler data to bring the whole world into the same shared pixel size, which in actual fact should be the base scale of 92m enjoyed at the equator.