Adding functionality to use ECMWF forecast or era5 grid instead of pre-processing grid + additional dustflag from Gareth #102
Conversation
…re present in the image
…re present in the image
|
@simonrp84 Yeah, I'm looking at it now. It compiles, just checking nothing breaks and aiming to add the flags for EMOS to be truly optional. |
adamcpovey
left a comment
adamcpovey
left a comment
There was a problem hiding this comment.
This compiles and runs for me, with no regressions using the previous processing style. Using the ERA5 grid introduces bigger changes in stemp then I expected, which warrants some investigation, but I'm happy to merge this once Dora confirms I haven't broken anything with my tidying.
| 'overlap ' // & | ||
| 'prob_opaque_ice' | ||
| 'prob_opaque_ice ' // & | ||
| 'N/A ' // & |
There was a problem hiding this comment.
Why isn't this prob_clear?
There was a problem hiding this comment.
I think I took this from Gareth's original implementation of the dustflag, and didn't realise it corresponds to prob_clear - I can change it to "prob_clear"
There was a problem hiding this comment.
In the main ORAC repo it is also written as prob_opaque_ice. Should I change or resolve this conversation and merge?
Updating main branch of fork to include new commits from main ORAC repo
3 participants
The main aim of this PR is to eliminate the need to use the emos library by interpolating onto a pre-defined pre-processing grid. The new edits automatically use the ECMWF forecast/era5(t) grids for the pre-processing. In order to do this, I changed the indexing that is used for the preproc_structures variables, as the x- and y-dimensions were using the start and end pixel numbers to define the arrays, while the k-dimension used 1 to kdim. I've changed it so all dimensions have array indices between 1 and the length of the dimension (xdim, ydim, kdim) for simplicity. I haven't enabled it by default, but if you set USE_ECMWF_PREPROC_GRID=True as one of the arguments for the pre-processor, then it will not use emos to interpolate.
Another minor change in this PR is Gareth's dust flag, which is to help reduce the misclassification of thick dust as cloud in the pre-processor.