How to present results from multiple algos in paper

[jbusecke]

@ocean-transport/collab_team Our most recent results indicate that differences between algos are very minor and thus I suggest not focussing on those front and center in the paper.

I suggest the following:

• For maps we show the averaged values across all algos we processed (this might change slightly) + Some in detail differences for the appendix
• For histograms we could show multiple distributions?
• For any aggregated metrics (e.g. global small scale heatflux) we could give the mean + std/var between algos as uncertainty?

Happy to make adjustments, just wanted to start a discussion

[jbusecke]

Here is a plot I made for the Appendix

Rows are latent and sensible heat flux. Columns are: (left) Small scale contribution averaged over algos; (middle) stdv between algos; (right) relative 'error' between algos with respect to the small scale contribution.

We can see that the difference between algos is not always negligible, but the high relative values are generally where the small scale contribution is small to begin with, so not that relevant to our results. Wonder how we could show that more conclusively.

[jbusecke]

We had a good discussion of this today, and there was an alternative suggestion to use the algo that makes most sense for the model (ecmwf?) and show results from that algo (instead of averaging across them).
Either way the uncertainties would have to be discussed (in the Appendix presumably), but I would like to get @rabernat s opinion on this.