Oddity with aw_intersects and adding up areas? about areal HOT 4 CLOSED

Also, FYI, I've got a second version of that function that uses areal here: https://github.com/dblodgett-usgs/intersectr/blob/areal/R/calculate_area_intersection_weights_2.R

from areal.

Thanks for following up @dblodgett-usgs!

This behavior (and the total and sum approaches to calculating weights) is explained (I hope!) here. The weight = "total" approach for aw_total() would use the total area of the source polygons.

The calculate_area_intersection_weights_2 function makes sense - and I see where the value is used here.

Can you tell me a bit about the use case here and what your plans were for intersectr?

from areal.

I see. I guess I misunderstood the distinction between intensive and extensive.

Quantities like mean-areal precip per time step are actually extensive? A little googling and a basic definition clarifies it for me.

Extensive properties, such as mass and volume, depend on the amount of matter being measured. Intensive properties, such as density and color, do not depend on the amount of the substance present.

I switched over to intensive and "total" and things are looking better.

Re: intersectr, we work with a lot of very large spatio-temporal array datasets. The worst case right now is hourly X 1km from 1979-present but more typical is 4km X daily data is common. This is primarily historical weather and downscaled climate projections.

I've implemented various workflows that take such grid X/Y/T grids and convert them to other grids or to spatial geometries over the years. The big heavy-lift system we use is here: https://cida.usgs.gov/gdp/

intersectr has been a little side project that's outside my day to day during the government shutdown to move some code I had sitting around into something formal and hopefully find a home for some of the kernels in there or replace it with other people's stuff!

I'll close this issue for now. Will submit another when I get a little further testing areal.

from areal.

OK - glad the weights are working better. FYI the weight = "total" argument is being ignored - the intensive approach uses weight = "sum" (as sf::st_intersect_as() does). If you have a use case for weight = "total" with the intensive option we can experiment with that as well.

Thanks for describing the more general use - makes sense - happy to think about how parts fit into areal if you want to go that route. Like I said on Twitter, creating grids is of particular interest to me, as are expanding the number of interpolation tools.

from areal.