Use centroids for GCM query

[lindsayplatt]

Change GCM query to work with centroids + use nice geoknife helper, result() instead of download() to avoid a bunch of parsing work. Currently only working with one variable, but have in issue here to help resolve that.

[lindsayplatt]

@hcorson-dosch here is the start of using centroids to query. This should help you do the next part, which is smartly splitting/grouping and using a constructed grid to query grid centroids, not lakes.

[jordansread]

Quick follow-up on the geoknife parsing issue. It seems the issue resides with just one of the variables you are using, which is soil moisture (mrso). That variable in funky because it has multiple "levels" which are causing some of the assumptions in the parser to fail as captured by Dave here.

gcm_job <- geoknife(
    stencil = simplegeom(query_pts),
    fabric = webdata(
        url = "https://cida.usgs.gov/thredds/dodsC/notaro_GFDL_1980_1999",
        variables = c("evspsbl", "hfss"),
        times = c('1999-01-01', '1999-01-15')
    ),
    wait = TRUE
)
result(gcm_job)
               DateTime         1         2             3         4         5 variable statistic
1   1999-01-01 00:00:00  5.40e-06  6.20e-06  8.000000e-06  1.20e-05  7.00e-06  evspsbl      MEAN
2   1999-01-01 01:00:00  3.90e-06  5.60e-06  7.300000e-06  1.25e-05  6.40e-06  evspsbl      MEAN
3   1999-01-01 02:00:00  3.20e-06  7.40e-06  5.700000e-06  6.80e-06  5.70e-06  evspsbl      MEAN
4   1999-01-01 03:00:00  8.30e-06  6.00e-06  4.200000e-06  1.80e-06  4.10e-06  evspsbl      MEAN
5   1999-01-01 04:00:00  5.10e-06 -1.00e-07  3.100000e-06  3.70e-06  3.00e-06  evspsbl      MEAN
6   1999-01-01 05:00:00  0.00e+00  1.90e-06  3.000000e-06  3.00e-06 -5.00e-07  evspsbl      MEAN
...

works, even though you are using more than one variable in this case. Since mrso isn't a variable used by our models, you can skip that. (We should also touch base on the variables necessary for the driver data, since not all variables in the Notaro dataset are used, they are here but we renamed those for readability back then...).

[hcorson-dosch-usgs]

Just looks like you missed updating this comment here to refer to the centroids, rather than the lake polygons

[hcorson-dosch-usgs]

But then again we'll be tweaking this to instead use the grid cell centroids anyhow, so will change shortly

[hcorson-dosch-usgs]

Looks like this target is still used to build our test query_map_png below, FYI

[hcorson-dosch-usgs]

This all looks good to me, Lindsay. Wasn't sure if you wanted to test w/ mrso dropped to see if the result() parser works w/ multiple variables if that one isn't included, before we merge this in?

[lindsayplatt]

@hcorson-dosch made those changes and added hfss back in - it does work!

library(targets)
tar_make(gcm_data_raw_feather)
tar_load(gcm_data_raw_feather)
x <- arrow::read_feather(gcm_data_raw_feather)
unique(x$variable)

[1] "evspsbl" "hfss"