New Feature: Option to output separate layers as nc files for use as palm_csd input

[stefanfluck]

Request was voiced to enable the possibility to output the separate layers of the static driver as separate nc files. This would allow using them in the palm_csd routine that is currently shipped with palm. palm_csd takes nc files for each layer (e.g. vegetation_type, lad, etc).

To be clarified before implementation (help wanted)

• which attributes does palm_csd expect of a nc file
• which nc files can palm_csd take

Idea for implementation:

• write function that takes the xarray dataarrays and save them separately before they are assembled into the xarray dataset that represents the static driver in the end.

[sebschub]

(Misusing the this issue a bit) I've been working on palm_csd lately and just learned about palmpy today. I am happy to explain palm_csd, modify palm_csd or discuss other ways to work together. I would be interested in a meeting. I am not sure what your current email address is so could you contact me? Email should be visible here or via my TU Berlin address.

[stefanfluck]

@pimonte provided this link with data requirements in #70 : https://palm.muk.uni-hannover.de/trac/wiki/doc/app/iofiles/pids/palm_csd

[stefanfluck]

Hm: https://palm.muk.uni-hannover.de/trac/wiki/doc/app/iofiles/pids/palm_csd#input_01-input_XX In this list, almost all files are marked as mandatory, especially also e.g. file_street_type and file_street_crossings. Are they really mandatory? If yes, this seems unpractical and a lot of extra effort to create files that are not really required for a normal urban climate simulation (e.g. street crossing data is only required when doing a multi agent simulation).

Also in general, since there is no sample data itself or descriptions of it available to the public (e.g. Berlin_bridges_height_15m_DLR.nc), I cannot really work on this at this time without more information.

The approach needs to be different than proposed in the first post (take data arrays and export to nc). The data arrays are too processed for palm_csd. Better approach: output input data already after first rasterization operations before any landuse/pavement etc logic is applied.

[SebaStad]

Be aware that the documentation on the IMUK trac might be outdated, the page you linked is 15 months old. palm_csd has been heavily developed on the gitlab repository, and not all of those inputs might be necessary anymore. But documentation on that is sparse at this moment in time.

As for sample data, the netcdf data is very simply structured, if you want i cant provide you one or two files. Its usually just a 2D-Dataset stored in the variable "Band1", independent on what data is in the file.

[stefanfluck]

Thank you for your input! I am happy to report that I have access to the current palm_csd code and also samples of the required files after discussion with @sebschub about this topic, but thank you for your offer! I would be in a position to start looking into it, but workload is a bit high at the moment. Next rainy weekend I'm on it.