Comments Figure 5 -- Beeswarm plot

[chrisweij]

Dear Alfredo, thanks for providing the Beeswarm plots for the paper! It is a very good start, there were some comments/requests made last week during the meeting. If you need any help or have any questions let me/us know.

• Add a vertical line between rest and sink to make clearer that sink does not count to the 100 %
• Try to enlarge the labels, also just one legend
• Add IPCC regions to the plot? E.g. in upper right corner —> An other suggestions is that if the IPCC region names be coloured to indicate ocean regions (for example in blue) and land regions (for example in green)?
• Use twin vertical axis, so that you can see the different values more easily, especially the rainfall recycling/sink. (basically a way to deal with the fact that one IPCC regions has a very large contribution which makes it harder to see the differences among models for the IPCC regions with low contributions)
• Use horizontal grid lines?
• There were some comments on the colours used in the plot: E.g. Colours Lagranto-watersip and Flexpart-latin are very similar, Think of colour blindness and the use of colours in this plot, e.g. different symbols could be used as an alternative. Personally, I am already quite happy with the colours.
• Keep boxplots in the back.

Optional:

• Scotland case: NA - ocean is large, there was a question if we should split this region? E.g. based on latitude —> Disadvantage would be that we can not include the WRF-WVT experiments then directly.
• Similar for S-Pacific ocean for the Australia case
• Make x-ticklabels right-Bound
• The use of a logarithmic y-axis was suggested, however some people indicated that this might also distort the plot

[crespoalfredo]

Hi Chris, I implemented some of the changes proposed in the last meeting, they can be reproduced with the new notebook BeeswarmAndBoxPlots_AllCases.ipynb. Some comments regarding these changes:

• The sink graph is now separated from the other CMIP regions.
• The labels are now larger (they could be even larger if we used the abbreviations of the regions on the X axis, instead of the whole name). In addition, there is only one legend.
• There is now a map with the IPCC regions in each panel. I have not changed the names of the regions because I think it is now clear from the map whether the regions are land or ocean, and adding more colours to the graph could be confusing.
• I included a twin vertical axis for the recycling, but I believe that using this new axis for other regions can be confusing, and that the same can be achieved by experimenting with the scale of the main Y axis.
• Horizontal grid lines are now included.

I include the resulting figure.

To see the different values more easily, I experimented with nonlinear scales. A logarithmic one distorts the plot and accumulate the points for high contributions, so I finally made a piecewise transformation of the Y axis. Between 0 and 20 %, this transformation is a square root, until 100 %, a linear one. I believe that now it is easier to see the differences.

Finally, it would be good to know if large regions (such as the North Atlantic in the case of Scotland) will be subdivided, as performing these simulations and getting the results will probably take me longer.