The plot a month later
Over the past month I have worked solid on rewriting the plot facilities
Figure 1, low resolution web image, comparable as shown at the beginning of May except a different geographic projection, done using wildly different software.
This is a quick look at what is new, the full thing is about to be used as part of a major article at the Talkshop.
Why? I was thoroughly fed up with fighting gnuplot, drives many people half crazy trying to get it to do what is wanted, all too often it simply can’t. In this case the final straw was no way of disabling automatic smearing of colour, nor can it draw literal filled polygons. (very amusingly given the authors say it can’t it then when writing out SVG format proceeds to write out filled polygons… can’t make it up stuff)
I was already using GLE, a tool with ancient origins. This will do what I say so all those polygons are drawn one at a time. The whole thing is very complicated, more so since my code writes GLE code and data before passing the whole lot over for plotting (GLE in turn calls ghostscript with postscript code, wheels within wheels)
There is more…
The colouring is beyond the capability of GLE so there was another tricky development task, a new sub-library.
Choosing a colour given data scaled between 0 and 1 is trivial? Oh no, it is a very hard problem. If you look at the key it is smooth and illumination smoothly varies. True enough I can program breakpoints and colours yet nothing much is being used at the moment, the trick is in the mathematics.
This is a fairly full implementation of CIE xyY colour internally [1,2] dealing as is necessary in RGB for input/output. This particular definition of colour is still not quite how we see but close. Using xyY allows tricks, eg. Y has the nice property of being visual brightness, I can literally scale all channels of Y and the whole image lightness changes across all colours.
Whether the present colour scheme is good is a matter of opinion, I’ve not had time to fiddle, I do subtle anyway. At least it is reasonably tasteful, avoiding vile yellow/greens is an objective.
Anyone who has persisted and read this far is now going to be offered an “adequate” new plot provided download size is not OTT.
Hadcrut4 because the filesize is smaller, make it a weighted plot, bump up coastal resolution, turn on major rivers too, let it autoscale, here is the small filesize web image, grab the PDF and zoom.
PDF of annotated plot 2.6MB
I don’t use Adobe software so I am assuming it works with Reader.
And that is the best I can do at the moment.
Criticise so I can hit you back 🙂
Maybe I’m mostly dry, I know foul, humour, former is rarely good in public writing, the latter rarely works in public.
1 Excellent short text by Douglas A. Kerr on CIE colour space
2. A very useful person http://www.brucelindbloom.com