Closed ledm closed 4 years ago
the fallback option for preprocessor function (mask landsea) when no fx files are available or they're broken (sometimes they are!) is to use Natural Earth shapefiles (for masking land at a higher resolution and seas at a lower resolution, but both resolutions much higher than the highest model resolution ie 10m and 50m << 1km). This option is straightforward and precise (even more precise than the fx files themselves, which may not necessarily be a good thing though) BUT it's not implemented for irregular grids and will spit a ValueError
if masking with NE files is attempted on irregular grids.
Can you not use the masking functions before doing any statistical operations on the data? Or you actually need the fx files in your diagnostic to manipulate different ratios and stuff
The area_statistics
and volume_statistics
operators are independent of the masking preprocessor.
In order to calculate an average over the data, we need to weight the mean by the size of the grid cells. We can't use the natural earth shape files to describe the grid area or volume as they are grid specific.
ok, makes sense, dunno really, not sure what can be done if the fx files are not there at all, do however see the comment https://github.com/ESMValGroup/ESMValCore/issues/24#issuecomment-503613912 since that will take care of about 90% of the problem :grin:
At the moment, some of the preprocessors use
fx_files
(like masking, the area/volume statistics and the upcoming zonal/meridional statitistics).It's not obvious what should happen when
fx_files
are absent. In the case of model data, I think that the preprocessor should fail withoutfx_files
.In the case of observational data, if
fx_files
are absent, then the tool should be able to estimate the grid's area or volume.This is complicated by the fact that the statistics tools can apply several operators:
It's a bit confusing!