Appendix statements - Githubissues

In this issue lets discuss all the statements + Figures that are relegated to the Appendix.

These plots are taken from the paper/appendix_plots.ipynb in the final_refactor branch or will be added there.

List of statements from the manuscript

Our results don't differ much between models

(Will redo and insert the figures from https://github.com/ocean-transport/scale-aware-air-sea/issues/59 here)

The small scale signal differs less between algos compared to the full flux signal

I looked at this with histograms

I think the histograms are quite nice visually, and the algo based RMSE (grid point diff to our main algo ecmwf squared and averaged over all time, space, algo dimensions) is relatively small to the full global signal. I also plotted the global average timeseries for different algos:

Difference between online and offline fluxes

🚨 As seen in the above the difference of our values compared to the online fluxes (black) is still large and I think we will have to find at least a likely explanation? Candidates for mismatch

Skin correction

Rectifying effect of daily cycle (e.g. are there nonlinear effects that we miss when computing the flux on daily averages?)

Aerobulk is not getting the right results?

[ ] Double Check that there is no output whatsoever on subdaily frequency?
[ ] Can we exclude the skin correction?

Filtering vs Coarsening

[ ] Show that the global mean timeseries of the coarsened results is very similar to the filtered?
- Might have to tune the coarsening factor to get the best fit?

Sensitivity to the iteration parameter in aerobulk

Can we test this on the coarsened data? That would make it a lot cheaper

I will keep updating this as I revise the data processing and the plots just wanted to summarize this before the meeting.

ocean-transport / scale-aware-air-sea