How to plot top species in Triplot RDA?

[Stabilo12]

Hi,

I'm quite new to R and hope somebody can help me. I would like to plot only the top e.g. five species in an RDA figure, as the other species are all overlapping and make the figure messy (see example).

I tried to use: select = abs(sum_rda$sp) > 0.5
However, it doesn't work but also does not show any error message.

data(varespec)
data(varechem)
ord <- rda(varespec)
summary(ord)
sum_rda <- summary(ord)

plot(ord, scaling=2, main="Triplot RDA", type="none", xlab=c("RDA1"), ylab=c("RDA2"), 
     xlim=c(-10,20), ylim=c(-10,10))
points(scores(ord, display="sites", choices=c(1,2), scaling=2), 
       pch=21, col="black", cex=1.2)
arrows(0,0, 
       scores(ord, select = abs(sum_rda$sp) > 0.5, display="species", choices=c(1),
              scaling=2)*2,  
       scores(ord, select = abs(sum_rda$sp) > 0.5, display="species", choices=c(2),
              scaling=2)*2,
       col="black",length=0, lty = 1)

text(scores(ord, select = abs(sum_rda$sp) > 0.5,  display="species", choices=c(1),
            scaling=2)*2.1,
     scores(ord, select = abs(sum_rda$sp) > 0.5,  display="species", choices=c(2), 
            scaling=2)*2.1,
     labels=rownames(scores(ord, display="species", scaling=2)),
     col="black", cex=0.8)   

Thanks for considering my request

[Imnotdogno]

Hi did you resolve this?
Can you please tell me how to plot only the most important species?

[jarioksa]

Seems to be a long and stale issue. The reason that the select did not work was that it was wrongly used. You are not using vegan functions text and points, but default R functions. You are just extracting the scores with vegan, but the scores function has no select argument: it will return all asked scores. If you want to take a subset of those in standard R functions, you need to use standard R tools, such as indexing, as they know no select.

I would like to suggest a different selections criterion: now you use axiswise length limit, meaning that the arrows should be within a square. I think it is better to base this test on arrow length, that is, to have arrows within a circle. Get the arrow lengths in 2-dim with

splen <- sqrt(rowSums(sum_rda$sp[,1:2]^2)) # your sum_rda object

Now to the solution: vegan plot functions return an invisible ordiplot object which you can catch, and this object has generic text and points functions that also have select and arrows argument. So this is a complete sequence:

library(vegan)
## copied from your message
data(varespec, varechem)
ord <- rda(varespec)
## modified from this on: catch the plot structure in pl
pl <- plot(ord, scaling=2, main="Triplot RDA", type="none", xlab=c("RDA1"), ylab=c("RDA2"), 
     xlim=c(-10,20), ylim=c(-10,10))
splen <- sqrt(rowSums(pl$species^2))  # species distance from the origin in pl[ot]
points(pl, "sites", pch=21, col="black", cex=1.2)
text(pl, "species", select = splen > 0.5, arrow=TRUE, length=0.05) # harmless warning on length argument, but works
## Alternatively you can use magrittr pipes (if you have splen  or select vector calculated by other means):
library(magrittr)
plot(ord, scaling=2, main="Triplot RDA", type="none", xlab=c("RDA1"), ylab=c("RDA2"), 
   xlim=c(-10,20), ylim=c(-10,10)) %>%
 points(what="sites", pch=21, col="black", cex=1.2) %>%
 text(what="species",  select = splen > 0.5, arrow=TRUE, length=0.05) # you need splen for this to work