Dealing with nans

[AndrewILWilliams]

Hi all!

Just looking around the code and am wondering how the package/the Bryan(2008) code deals with nans? Lots of CMIP data on pressure levels is nan because of orography, so just wondering if this is taken care of in the code or if it's required to remove the nans by hand before running it through?

Looking forward to using xcape!

Cheers!
Andrew

[AndrewILWilliams]

Just looking now, I don't think missing values are accounted for. I'm not sure what the easiest way to do this would be... one could think of adding a line to the fortran call here, where it's told only to use levels with data? Perhaps there's a smarter way of doing this though.

[chiaral]

We focused on the model level part. So the pressure part might have some lingering issues. But the pressure lev code has a part of the code which should only use the pressure levels that are lower than (surface pressure) ps. It checks that in the code. So if there are missing values within the column where there shouldn’t be I think i used to interpolate or bfill before feeding it to xcape (sometime hus is negative in models and you have to correct that, xcape is not that smart. You have to clean the model data first). But if it is nan because of orography, you need ps and the code will adapt for each grid point. Unless the model has some discontinuity between ps and the fixed pressure level. I think that was the idea. Not sure it is fool proof.

…

Sent from my iPhone

On Dec 11, 2021, at 11:48 AM, Andrew Williams ***@***.***> wrote:  Just looking now, I don't think missing values are accounted for. I'm not sure what the easiest way to do this would be... one could think of adding a line to the fortran call here, where it's told only to use levels with data? Perhaps there's a smarter way of doing this though. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or unsubscribe. Triage notifications on the go with GitHub Mobile for iOS or Android.

[AndrewILWilliams]

Cool! I'll take another look at the data and just check if anything funny is happening... some interpolation might be necessary

Also sorry for messaging on the weekend, AGU brings out the worst in all of us eh!

[chiaral]

I misspoke. It should check for nan if the zero level at surface is nan but it won’t if there are missing values. I don’t recall well I would need to check.

…

Sent from my iPhone

On Dec 11, 2021, at 12:01 PM, Chiara Lepore ***@***.***> wrote: We focused on the model level part. So the pressure part might have some lingering issues. But the pressure lev code has a part of the code which should only use the pressure levels that are lower than (surface pressure) ps. It checks that in the code. So if there are missing values within the column where there shouldn’t be I think i used to interpolate or bfill before feeding it to xcape (sometime hus is negative in models and you have to correct that, xcape is not that smart. You have to clean the model data first). But if it is nan because of orography, you need ps and the code will adapt for each grid point. Unless the model has some discontinuity between ps and the fixed pressure level. I think that was the idea. Not sure it is fool proof. Sent from my iPhone >> On Dec 11, 2021, at 11:48 AM, Andrew Williams ***@***.***> wrote: >> >  > Just looking now, I don't think missing values are accounted for. I'm not sure what the easiest way to do this would be... one could think of adding a line to the fortran call here, where it's told only to use levels with data? Perhaps there's a smarter way of doing this though. > > — > You are receiving this because you are subscribed to this thread. > Reply to this email directly, view it on GitHub, or unsubscribe. > Triage notifications on the go with GitHub Mobile for iOS or Android.

[AndrewILWilliams]

One more (potentially dumb) question, but how do you find the model level data for CMIP6? I'm looking on the google cloud catalogue (https://storage.googleapis.com/cmip6/cmip6-zarr-consolidated-stores.csv) and I just see the output on pressure levels...maybe I'm missing something though.

[chiaral]

6hrlev

…

Sent from my iPhone

On Dec 11, 2021, at 1:36 PM, Andrew Williams ***@***.***> wrote:  One more (potentially dumb) question, but how do you find the model level data for CMIP6? I'm looking on the google cloud catalogue (https://storage.googleapis.com/cmip6/cmip6-zarr-consolidated-stores.csv) and I just see the output on pressure levels...maybe I'm missing something though. — You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe. Triage notifications on the go with GitHub Mobile for iOS or Android.