Default plot style

[daniel-caichac-DHI]

Hi, is it ok if I change the Default style of the Scatter plots from 2D histogram to Scatter Density, and default cmap from viridis to rainbow ? I think that it would make quite a change, as of now the default is a 2D Histogram with viridis, see example:

• Whereas if we put scatter density and rainbow as default same plot looks like this

• Maybe the viridis could stay as default (see next plot with viridis and scatter density) but the 2D histogram is not ideal as default I would say:

• Opinions? I can make the change quickly

[jsmariegaard]

I wanted to change the default colormap earlier (and still do) - I got stuck in making it consistent with the plotly backend. See #91 . There can be many opinions on the best cmap - personally, I like YlOrRd more, but we can cetainly agree that the current default is not good. @ecomodeller what do you say?

[daniel-caichac-DHI]

What do you think about plasma as default cmap? I have been using it already and it is one of the perceptually uniform colormaps from matplotlib

I think that the high-density area should be something like bright yellow and the low density should be darker (as the camp I propose) and not the other way around as the uber cmap, otherwise, the points that are far away (~outliers) won't be noticed when in reality those could actually dominate design conditions or whatnot, so we want to see them; I don't want to have these points as transparent just because they have low density: eg, the 100yr Hm0 maybe be located on top of the 1:1 line (perfect model fit), but will have very low density (maybe just once in my simulation), hence I need to see this point, making it transparent is risky.

I think the uber heatmap is good for a spatial-heatmap, but not ideal for a scatter plot heatmap for our modelling-purposes

[jsmariegaard]

I think we may need different colormaps depending on whether it's a 2d-histogram-type (where transparent-yellow-red works well) and density-type where outliers also need to be highlighted (as you point out). I guess it also depends on whether you're interested in extremes or normal conditions.

[jsmariegaard]

But I think that plasma looks good for the density plot!

[jsmariegaard]

I think we should do something about the quantiles - they are too dominating this way. Maybe a piecewise linear line would be better?

[daniel-caichac-DHI]

But I think that plasma looks good for the density plot!

Viridis actually also look quite decent so any of those is good with me

I think we should do something about the quantiles - they are too dominating this way. Maybe a piecewise linear line would be better?

I agree with the QQ being too dominant, but I am not sure I follow with the piecewise linear (as in, a line instead of points?)

I think we may need different colormaps depending on whether it's a 2d-histogram-type (where transparent-yellow-red works well) and density-type where outliers also need to be highlighted (as you point out). I guess it also depends on whether you're interested in extremes or normal conditions.

Agree