Compatibility issue while deploying new example to docs

[wilsonduan10]

The Climate Modeling Alliance

Purpose

The example file aerosol_activation.jl is only compatible with CloudMicrophysics v0.5, which constrains other CliMA packages to older versions. For example, CloudMicrophysics v0.5 is only compatible with the package SurfaceFluxes on versions up to 0.3. As a result, when adding an example file that depends on SurfaceFluxes v0.6, make.jl cannot build the docs for both aerosol_activation.jl and whatever file using SurfaceFluxes v0.6 due to compatibility mismatch. This poses a problem for branch wd/surface-flux-example, which intends to add the example kappa_calibration.jl, which relies on SurfaceFluxes v0.6.3.

Cost/Benefits/Risks

Although it takes time to update aerosol_activation.jl to match with more recent package releases, it makes the example file more relevant and easier to reproduce. In addition, by updating the file, it would allow developers such as myself to put out additional example files such as the one introduced above, kappa_calibration.jl, which depends on a newer version of SurfaceFluxes.

People and Personnel

• Lead: @odunbar
• Reviewers: @akshaysridhar, @LenkaNovak

Components

The solution would be updating aerosol_activation.jl to match current releases.

Results and Deliverables

We would expect the updated example to have the same functionality but simply operate under the newer package releases.

[odunbar]

Thanks for the suggestion.

I do not maintain this example, as i don't think it was the intention for this to track current versions. However I presume perhaps people involved in Microphysics API (e.g. @trontrytel) have an idea on how the Microphysics API has changed, and the expedience of updating it

In general, I am not super keen on having the documentation becoming heavy with evolving dependencies exactly for this reason. So if there are alternative solutions whereby examples are run in the example folders, figures and code snippets ported over to the docs this would be an easy fix. It's not as fancy as literating/example blocks but it is easier to maintain moving forwards (particularly if the maintainers of e.g. SurfaceFluxes are not planning on keeping the example here up-to-date).

[trontrytel]

I'll take a look. It shouldn't be too much work to update the dependecies (famous last words ;)

[akshaysridhar]

Thank you @trontrytel!

[odunbar]

Following on from those "famous last words" :P ... It looks like there is a non-identifiability issue of the aerosol activation inverse problem that was set up. I would like this example to be modified by the microphysics folks, either by improving the data used for one of the parameters, else by removing it and reducing to a 1D learning problem. In either case there may be an additional turnaround time here