Document provenance of results by Allen Pope

From Geoscience Paper of the Future
Jump to: navigation, search


Details on how to do this task: Document the provenance of the results

I looked through the workflow tools. They look good in theory, but with code already done, it was too much of a hassle. Also, not used by my community so not as helpful. MIght also be important that MATLAB is proprietary and didn't look like it was supported? Not sure about gdal. So, I have chosen to created a fairly detailed workflow diagram, instead (I chose to do this in Illustrator). [1]

This was a good exercise to align datasets and code. It made sure that I knew all the bits of code I needed (was good to do before sharing code - now that is ready to do, too) as well as all the data. Gave me an appreciation for the complexity behind what is otherwise a fairly simple description.

I think it will potentially make easier for others to use. It also made me realize my code is probably not terribly efficient. Something to work on better in the future, not necessarily now. I also realized that putting together this structure helped me change the way I think about blocks of code, etc - which will be helpful for sharing code later. It made me think of better ways to structure my code (where to put parameters / how to comment) and how to make it more automated, less hard-coded to particular filenames/landsat scenes.

Consider who did what - and it's a matter of scale (number of authors, processes, etc.)

As a not-formally-trained coder, this made me realize more best practice in terms of creating code. (see above)

One thing which isn't 100% reproducible is the figures. For the map - the GIS is not something I can code up. For the other plots, I provide a way to get to the point of the figure where possible - but I then have taken it through plot.ly and Illustrator to get to what is in the paper. I would change how I did this - save the exact code that got me to a figure as much as possible. As well as share the plot.ly itself as an easier way to share things.

Will need to describe in text, too. Break down each sub-section of the workflow, and potentially sketch it out even more broadly at the time of writing. Potentially make more modular figures, too

Also - need to add code DOIs as appropriate.