tikz pgfplots ternary diagram, density map

Question

I am plotting points in a ternary diagram using pgfplots and the ternary axis environment. Each point does posses a 4th value (besides the 3 that are the coordinates). At the moment I am stating this 4th value as a node next to each plotted point. This comes with the problem that the minima/maxima of this value is not easy to spot as well as it is hard to read general trends. Also nodes can't reach outside of the ternary axis which leads to inconsistency with labeling points close to the axis.

I would like to have the points coloured according to their value (i.e. a value of 5 would be blue, 7.5 would be yellow and 10 would be red) this should be done gradually of course. Ideally I would like to have a fully coloured map with the space between the points interpolated.

Does anyone know of a way or at least a workaround to have the points coloured according to their 4th value?

Image:

LaTeX source:

http://depositfiles.com/files/5gaazfu5q

Answer 1

a working example:

\documentclass [12pt]{scrartcl}
\usepackage [utf8x] {inputenc}
\usepackage{pgfplots}
\pgfplotsset{compat=1.7}
\usepgfplotslibrary{ternary, units}
\usepackage{tikz}
\usepackage{tikz-3dplot}
\usetikzlibrary{decorations.pathmorphing, pgfplots.ternary, pgfplots.units}

\begin{document}
\begin{tikzpicture}
 \begin{ternaryaxis}[colorbar, colormap/jet,
 xmin=0,
 xmax=100,
 ymin=0,
 ymax=100,
 zmin=0,
 zmax=100, 
 xlabel=component1,
ylabel=component2,
zlabel=component3,
label style={sloped},
minor tick num=3,
grid=both,
]

    \addplot3+[only marks, 
    point meta=\thisrow{myvalue}, %  uses ’point meta’ as color data.
     nodes near coords*={\tiny{\pgfmathprintnumber\myvalue}}, % does what it says
     visualization depends on={\thisrow{myvalue} \as \myvalue} %defines visualization dependency
     ] table {

x       y       z       myvalue
10      0       90      7.1
40      0       60      9.2
50      0       50      9.8
70      0       30      8.5
20      30      50      5.5
20      20      40      5
20      50      30      4.8
30      40      30      6.3
30      20      50      7.1
40      20      40      7.8
40      30      30      7.4
40      40      20      6.9
40      50      10      6.7
10      10      80      4.7
10      20      70      4.2
10      30      60      3.7
10      40      50      3.5
10      50      40      3.2
10      70      20      4.8
10      80      10      5.2
50      30      20      7.8
50      20      30      8.3
60      10      30      9
70      20      10      9.2
80      10      10      9.9
20      10      70      6.2
40      60      0       6.6
70      30      0       9.3
50      10      40      8.9
20      20      60      5.9
};

\end{ternaryaxis}
\end{tikzpicture}
\end{document}

Result:

Answer 2

The following can be achieved with R, knitr and ggtern (for producing ternary diagrams):

Produced with the following Rnw code, using data from .csv file (available HERE).

\documentclass[12pt,a4paper]{article}
\usepackage[margin=2cm]{geometry}
\usepackage{helvet}
\renewcommand\familydefault{\sfdefault}
\usepackage{fancyhdr}
\pagestyle{empty}
\usepackage{subcaption}

\begin{document}
  <<SETUP,echo=FALSE>>=
    suppressMessages(library(ggtern))
    suppressMessages(library(tikzDevice))
    opts_chunk$set(fig.width=10,fig.height=9)
  @

\begin{figure}[p]
    <<PLOT,echo=FALSE,fig.align='center',dev='tikz'>>=
      df <- read.table("data2.csv",header=T)
      base <- ggtern(data=df,aes(x=y,y=x,z=z)) + 
        geom_point(aes(fill=myvalue),color="black",shape=21,size=5) + 
        geom_text(aes(label=myvalue),vjust=-1,size=3) + 
        theme_rgbw() +
        guides(fill=guide_legend()) + 
        scale_fill_gradient2(low="blue",mid="yellow",high="red",midpoint=mean(df$myvalue),breaks=c(3:10)) +
        theme(legend.position=c(0,1),
              legend.justification=c(0,1)) +
        labs(title="Example Ternary Plot",T="THS",L="LHS",R="RHS",fill="Density") +
        custom_percent("Percent")
      print(base)
    @
  \caption{Example Ternary Plot with Aesthetic Mapping}
\end{figure}

\end{document}

Using the tikzDevice, the font of the latex document is reflected in the plot itself.