Missing Counties (Coastal Locations)

[yougsanghvi]

Hi!

I'm seeing a few dozen counties with missing data post aggregation (even though the raw data I am using has complete data) when I aggregate by population weights. Here is a map of counties in the US which are missing. Happy to send you a list of fips codes too, but as you can see they are mainly counties along the water. Does stagg give an NA even if there is a single grid cell with for eg. an NA population value (if grid cell is fully in the water for eg)? Maybe we can ignore those cells and simply aggregate over the rest? Not sure if that's already implemented and something else is going on here.

P.S there's also a few in Hawaii which are not pictured above. Some FIPS include: {'12015', '12045', '12075', '12087', '12089', '13039', '13127', '13179', '13191', '15001', '15003', '15007', '17031', '17097', '2013', '2016', '2050', '2070', '2122', '2130', '2150', '2158', '2164', '2180', '2185', ..., '69100', '69110', '69120', '72097', '78010'}

Thanks again!!
Youg

[traceybit]

hi @yougsanghvi! i think this might be directly related to a decision that we made to, as you suggest, return NAs for the polygon if any grid cells are NA. i think in your case the area-weighted only outputs will not contain NAs. this has come up for others, and i can't remember if we made changes to this assumption/rule more recently. the idea was that if there were NAs in the grid, stagg would return NAs for the polygons in order to "force" the user to make informed decisions about how to proceed.

@tliddell4 -- can you confirm that this is correct? thanks!

[yougsanghvi]

Thanks! For my code, i simply copy-pasted the function and added a line to change the NA weights to 0 (Maren had done the same previously, so I just used her code). This seems to happen since our era5_pop data doesn't have population for certain areas (areas where people probably don't live, see attached photo).

Here is the line I added:
w_merged <- w_merged |> dplyr::mutate(weight = ifelse(is.na(weight),0,weight))

Right after:
# Merge with secondary weights, NA for missing values
w_merged <- merge(area_weight, weights_dt,
by = c('x', 'y'),
all.x = T)

Maybe it is useful to implement this as an option for users? Not sure if this is an issue that only Maren and I had to fix or if it is something you see other people benefiting from too.

Thanks!!
Youg