"nan" is not supported in NCL but more and more dataset make nan as fillvalue

[Yonghan33]

Describe the bug Recently, I found that the era5 dataset is different from the old CDS.

The new website to download. (https://cds-beta.climate.copernicus.eu/)

1. The dataset's fillvalue is 'nan' , and NCL cannot recognize this fill value. Everytime I need to use the 'where' function to take it place to 9.96921e+36 or other fillvalue NCL supported. This problem also happends in the output data from python.
2. The type of 'time' dimension is int64, which is not supported in cd_calendar function.

ERA5 dataset is a dataset that we commonly use. I hope the official team can resolve the issue to facilitate the use of the datasets. Thanks!

Provide the following:

• a concise NCL script demonstrating the issue (remove unnecessary code)
• necessary data files
• relevant graphical output The code I used to deal with, ; ----take place the fillvalue u@_FillValue = 9.96921e+36 u = where(u.eq."nan", u@_FillValue, false_value) ; ----change the time type to integer not int64 time0 = f->valid_time time = toint(time0) copy_VarMeta(time0, time) ; printVarSummary(time) timeARR = cd_calendar(time, 3)

Computing environment

• OS: [Linux, Mac, or Windows]
• OS version: [e.g. Ubuntu 16.04, MacOS 10.13.6, Windows 10]
• NCL Version: [e.g. 6.5.0]
• Installation method: [binary from NCAR website, conda, other package manager (apt, brew, macports, yum), built from source]

Additional context Add any other context about the problem here.

[Yonghan33]

It's a good way to repalce "nan" value to fillvalue quickly. ncatted -O -a _FillValue,value_in,o,f,9.96921e+36 infile.nc Here, value_in is the float value that you need to change its fillvalue and infile is the file you have. Others as original.

[KMFleischer]

What about the function replace_ieeenan?

Have a look at https://www.ncl.ucar.edu/Document/Functions/Built-in/replace_ieeenan.shtml and https://www.ncl.ucar.edu/Document/Functions/Built-in/isnan_ieee.shtml

[Yonghan33]

Thanks for your reply!

But NCL is slow to precess the data. Using 'ncatted' by NCO is more convenient.

 

------------------ Original ------------------ From: "NCAR/ncl" @.>; Date: Wed, Nov 13, 2024 07:07 PM @.>; @.**@.>; Subject: Re: [NCAR/ncl] "nan" is not supported in NCL but more and more dataset make nan as fillvalue (Issue #217)

What about the function replace_ieeenan?

Have a look at https://www.ncl.ucar.edu/Document/Functions/Built-in/replace_ieeenan.shtml and https://www.ncl.ucar.edu/Document/Functions/Built-in/isnan_ieee.shtml

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>