Extract odd-shaped (irregular) regions about esmvalcore

[LisaBock]

Okay, I will try it like this. Thanks!

(Just for clarification @valeriupredoi: you mean extract_region instead of select_region, don't you? Or is this a function I do not know?)

[valeriupredoi]

@LisaBock it is very important from the implementation point of view to know which file format your regions are stored in - netCDF/Natural Earth/txt etc and what are the actual parameters of the region ie how it is defined in the file: like a mask of 0s and 1s, like a percentage that we apply a threshold on etc. Without this information I can not implement anything just yet. Cheers!

[ledm]

With regards to the Natural Earth Shapefiles, I can't see anything that names specific marine regions. The closest dataset I can find is the marine indicators, but I suspect that this is simply the national sea boundaries at 200 nautical miles.

I think that the best way to define the oceans right now would be to regrid CMIP models to a standard resolution in the preprocessor, then use an older regional definition mask file in the diagnostic. Do we have a definition file anywhere?

[valeriupredoi]

sftof is used for seas/oceans selection, that is available for most models, for the ones that dont have it we currently use Natural Earth low-res oceans shapefiles; the seas/oceans are not a problem - the problem is continents (and Russia? from the figure above) - we can use shapefiles for those as well but I thought @LisaBock already has a set of region-defining files at hand

[mattiarighi]

I think @LisaBock simply needs to extract regions which are not rectangular, but more generally quadrilateral.

No shapefiles should be required here, but only the possibility of specifying the coordinates of the four (or more) vertices which define the region.

[LisaBock]

Indeed, for the continental regions it should look like this (polygons with four or five corners):

But for the most diagnostics I would need a combination of this subregions.

And for the oceanic regions, I do not know if they are polygons. I asked already for the definitions but have not received them yet.

[bouweandela]

We already have the preprocessor function extract_region which is able to extract rectangular regions, so I think that should be able to extract most regions.

We also have a recipe that extracts regions based on shapefiles, recipe_shapeselect.yml. This could be used for the more complicated shapes, you will just need to define the shapefiles. You can then run this diagnostic before your own diagnostic, using the ancestors key in the recipe.

[valeriupredoi]

This would still be useftl in the preproc tho, to shrink the data and speed up the other preproc steps that follow. I agree we can use the shp extractor functionality we already have - Lisa, any chance you can define these regions for me in terms of lat-lon points? Then I can add a shp constructor function and use the shp masking we already have Dr Valeriu Predoi. Computational scientist NCAS-CMS University of Reading Department of Meteorology Reading RG6 6BB United Kingdom

…

On Fri, 8 Mar 2019, 10:14 Bouwe Andela, ***@***.***> wrote: We already have the preprocessor function extract_region <https://esmvaltool.readthedocs.io/en/version2_development/codedoc2/esmvaltool.preprocessor.html#esmvaltool.preprocessor.extract_region> which is able to extract rectangular regions, so I think that should be able to extract most regions. We also have a recipe that extracts regions based on shapefiles, recipe_shapeselect.yml <https://github.com/ESMValGroup/ESMValTool/blob/version2_development/esmvaltool/recipes/recipe_shapeselect.yml>. This could be used for the more complicated shapes, you will just need to define the shapefiles. You can then run this diagnostic before your own diagnostic, using the ancestors key in the recipe. — You are receiving this because you were assigned. Reply to this email directly, view it on GitHub <https://github.com/ESMValGroup/ESMValTool/issues/943#issuecomment-470877029>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AbpCo1cyZo8bSLRBcL-pRXzDoL4XS2Afks5vUjf3gaJpZM4bjNIj> .

[valeriupredoi]

actually here's how I see a least-effort/maximum-outcome path to solve this:

• provide the corners of the rough rectangular approximation of your region in the preprocessor, apply select_region to that to minimize the amount of data that is passed on to the next preproc steps;
• in diagnostic, use the masking with shp from _mask.py to select your refined region;
  All we'd need in this case is to have the region defined in a shp file and store it somewhere where the diagnostic can see it.

what say you @LisaBock

[valeriupredoi]

yes that one sorry, I am starting to pull function names out my arse 😁

[valeriupredoi]

hi @LisaBock - is this been done (by you) or you need any help implementing it? 🍺

[mattiarighi]

@LisaBock is this solved? If yes, please close.

[valeriupredoi]

closing since it looks to be solved, please re-open if need be 🍺