JOSS Paper Review 2

[ziyiyin97]

My apologies for the late reviews openjournals/joss-reviews#4869

Overall the package shows how to run derivative-free ensemble kalman process in Julia and entails quite clean documentation and examples. There are also a lot of useful references in the manuscript. A few comments:

1. In general how does this package compare to other Kalman-related packages? like Kalman.jl or other Julia/Python based packages?

Comments for the manuscript:

1. Line 30 "under certain conditions" -- probably need a reference.

In terms of the code/examples/docs:

1. The package claims that the prior distribution can be either from Julia Distributions or empirical/sample-based distributions. I would appreciate it if authors could provide an example related to sample-based distributions. (or point to reference)
2. Also appreciate an example w/ TOML I/O for a few iterations. (or point to reference) I assume these are both related to this module https://github.com/CliMA/EnsembleKalmanProcesses.jl/blob/main/src/UQParameters.jl

I hope these concerns can be easily addressed with more explanations or so.

[odunbar]

Hi @ziyiyin97 Thank you very much for your review.

Proposed response:
In paper.md

• We are currently writing a state-of-field section for the other review, and where our code fits in, we shall include Kalman.jl, GaussianFilter.jl, EnKF.jl in this.
• "under certain conditions" refers to the standard Kalman assumptions, we shall tighten up this statement.

In the docs

• We will add a docs for defining a sample-based distribution.
• We will add a docs tutorial for the TOML interface

In the code

• We will add a simple example that uses the TOML interface

We shall keep you updated! Thanks again

[ziyiyin97]

Sounds good!

[odunbar]

An update of where we are:

• PR Add State-of-Field, and rearranges paper.md sections #231 adds a state-of-field, and tightened Kalman assumptions (along with changes for the other review)
• PR toml-based example #238 adds a replica of our simple sinusoid example, but uses the TOML and file-system interface between model and calibration tools.
• PR JOSS - new Samples examples #240 adds some simple intuitive examples for creating prior distributions using "samples-based" distributions into the docs.

[odunbar]

@ziyiyin97 I believe we have now addressed all the issues raised.

• The changes to the paper are in paper.pdf
• The new TOML example is found in examples/SinusoidInterface/
• The docs for the TOML examples here
• The docs for the samples-example here

[ziyiyin97]

Thanks for your quick update. The new TOML example in https://github.com/CliMA/EnsembleKalmanProcesses.jl/tree/main/examples/SinusoidInterface needs an updated version of EnsembleKalmanProcess otherwise julia --project will use the previous version of EnsembleKalmanProcess w/o TOMLInterface --i.e. ERROR: UndefVarError: TOMLInterface not defined. But everything else looks good to me and this issue can be fixed by burning a new release. Thanks again for your great work which enriches the Kalman community in Julia!