R - Taylor Diagram Plotting

Question

I am trying to plot multiple models in a Taylor Diagram and am struggling slightly with the code. I have managed to produce the diagram (see image) but cannot figure out how to reduce the axes as they are too large, standardise the axes to be labelled 1,2,3,4 and add tick marks on the correlations - with the tick marks I wanted to have major ticks every 0.1 and minor ticks every 0.05 up to 0.9, after which I was trying to have major ticks at 0.95 and minor ticks every 0.01 at this point (if this makes sense). Any help/advice with the above would help. I have used 'taylor.diagram' within the 'plotrix' package (and read the guides to it - but I'm relatively inexperienced with R) and attached my (somewhat basic) code so far but my plot looks rather messy. Thanks

all.models <- as.data.frame(cbind(Sy.One, Sy.Two, Sy.Three, Sy.Four, Sy.Five, Sy.Six, Sy.Seven, Sy.Eight, Sy.Nine, Sy.Ten))

taylor.diagram(CSR, Sy.One, sd.arcs=T, ref.sd=T, pcex=1.5, main=NULL, pos.cor=F,
              xlab="Standard Deviation (cm)", ylab="Standard Deviation (cm)")

for (i in 1:dim(all.models)[2]) {
  model.wanted <- all.models[,i]
  taylor.diagram(CSR, model.wanted, sd.arcs=T, ref.sd=T, pcex=1.5, col=i, add=T, pos.cor=F)}

# Add legend
model.names <- c("Sy=1%","Sy=2%","Sy=3%","Sy=4%","Sy=5%","Sy=6%","Sy=7%","Sy=8%","Sy=9%","Sy=10%")
legend("top", model.names, pch=19, col=i, cex=1.0, bty="n", ncol=5)

Answer 1

One option is to use a different package, such as openair, which might be more flexible. Because of your specific requirements it might be easier to use code designed for your requirements. I wrote some code to generate the following plot, which is close to your desired plot. You can hack the code to adjust the plot to your desired format.

# code to make a Taylor diagram
# formulas found in http://www-pcmdi.llnl.gov/about/staff/Taylor/CV/Taylor_diagram_primer.pdf
# and http://rainbow.llnl.gov/publications/pdf/55.pdf

# correlations and tick marks (only major will have a line to the center)
# minor will have a tick mark
correlation_major <- c(seq(-1,1,0.1),-0.95,0.95)
correlation_minor <- c(seq(-1,-0.95,0.01),seq(-0.9,9,0.05),seq(0.95,1,0.01))

# test standard deviation tick marks (only major will have a line)
sigma_test_major <- seq(1,4,1)
sigma_test_minor <- seq(0.5,4,0.5)

# rms lines locations
rms_major <- seq(1,6,1)

# reference standard deviation (observed)
sigma_reference <- 2.9

# color schemes for the liens
correlation_color <- 'black'
sigma_test_color <- 'blue'
rms_color <- 'green'

# line types
correlation_type <- 1
sigma_test_type <- 1
rms_type <- 1

# plot parameters
par(pty='s')
par(mar=c(3,3,3,3)+0.1)

# creating plot with correct space based on the sigma_test limits
plot(NA
     ,NA
     ,xlim=c(-1*max(sigma_test_major),max(sigma_test_major))
     ,ylim=c(-1*max(sigma_test_major),max(sigma_test_major))
     ,xaxt='n'
     ,yaxt='n'
     ,xlab=''
     ,ylab=''
     ,bty='n')

#### adding sigma_test (standard deviation)
# adding semicircles
for(i in 1:length(sigma_test_major)){
  lines(sigma_test_major[i]*cos(seq(0,pi,pi/1000))
       ,sigma_test_major[i]*sin(seq(0,pi,pi/1000))
       ,col=sigma_test_color
       ,lty=sigma_test_type
       ,lwd=1
    )
}

# adding horizontal axis
lines(c(-1*max(sigma_test_major),max(sigma_test_major))
     ,c(0,0)
     ,col=sigma_test_color
     ,lty=sigma_test_type
     ,lwd=1)

# adding labels
text(c(-1*sigma_test_major,0,sigma_test_major)
     ,-0.2
     ,as.character(c(-1*sigma_test_major,0,sigma_test_major))
     ,col=sigma_test_color
     ,cex=0.7)

# adding title
text(0
     ,-0.6
     ,"Standard Deviation"
     ,col=sigma_test_color
     ,cex=1)

#### adding correlation lines, tick marks, and lables
# adding lines
for(i in 1:length(correlation_major)){

  lines(c(0,1.02*max(sigma_test_major)*cos(acos(correlation_major[i])))
        ,c(0,1.02*max(sigma_test_major)*sin(acos(correlation_major[i])))
        ,lwd=2
        ,lty=correlation_type
        ,col=correlation_color
  )
}

# adding minor tick marks for correlation
for(i in 1:length(correlation_minor)){

  lines(max(sigma_test_major)*cos(acos(correlation_minor[i]))*c(1,1.01)
        ,max(sigma_test_major)*sin(acos(correlation_minor[i]))*c(1,1.01)
        ,lwd=2
        ,lty=correlation_type
        ,col=correlation_color
  )
}

# adding labels for correlation
text(1.05*max(sigma_test_major)*cos(acos(correlation_major))
     ,1.05*max(sigma_test_major)*sin(acos(correlation_major))
     ,as.character(correlation_major)
     ,col=correlation_color
     ,cex=0.5)

# adding correlation title
text(0
     ,max(sigma_test_major)+0.5
     ,"Correlation"
     ,col=correlation_color
     ,cex=1)


#### adding rms difference lines
# adding rms semicircles
for(i in 1:length(rms_major)){
  inds <- which((rms_major[i]*cos(seq(0,pi,pi/1000))+sigma_reference)^2 + (rms_major[i]*sin(seq(0,pi,pi/1000)))^2 < max(sigma_test_major)^2)
  lines(rms_major[i]*cos(seq(0,pi,pi/1000))[inds]+sigma_reference
        ,rms_major[i]*sin(seq(0,pi,pi/1000))[inds]
        ,col=rms_color
        ,lty=rms_type
        ,lwd=1
  )
}

# adding observed point
points(sigma_reference
       ,0
       ,pch=19
       ,col=rms_color
       ,cex=1)

# adding labels for the rms lines
text(-1*rms_major*cos(pi*rms_major/40)+sigma_reference
     , rms_major*sin(pi*rms_major/40)
     ,as.character(rms_major)
     ,col=rms_color
     ,cex=0.7
     ,adj=1)

# adding title
text(0
     ,-1.5
     ,'Centered RMS Difference'
     ,col=rms_color
     ,cex=1
     ,adj=0.5)


###################### adding points #####################
names <- paste("model",seq(1,8),sep='')
correl_names <- seq(-0.6,0.8,by=0.2)
std_names <- seq(2,4,by=0.26)
color_names <- topo.colors(length(names))
points(std_names*cos(acos(correl_names))
       ,std_names*sin(acos(correl_names))
       ,col=color_names
       ,pch=19
       ,cex=1.5)

# making legend
par(xpd=TRUE)
legend(0,-2
       ,names
       ,pc=19
       ,col=color_names
       ,ncol=3
       ,bty='n'
       ,xjust=0.5)