Reputation: 3168
I have a raster in terra
that has values that ranges around 0. I want my negative values to be blue and positive values to be red. This question has already been ask here and here, however the answer is for the raster
package and do not apply for the terra
package has there is no breakpoints argument.
Here is what I tried so far:
# preparing the session
library(terra)
library(magrittr)
# preparing the raster, I'm making sure my data go through 0 but are not centered at 0
set.seed(1234)
rr <- rast(matrix(rnorm(400, 1.5, 1), nrow=20, ncol=20))
Even if my values are not centered at 0, I want my color intensity to be centered at 0, to do so, I find the range that I need :
the_range <- range(values(rr)) %>% abs %>% max %>% multiply_by(c(-1,1))
Then I create my color palette:
the_palette_fc <- leaflet::colorNumeric(palette = "RdBu", domain = the_range, reverse = T)
If I set, for example, 31 color classes to my map and plot it normally with terra I get:
plot(rr, col=the_palette_fc(seq(the_range[1], the_range[2], length.out=31)))
You can see that the white color was not fixed to the zero value. Thats is probably because terra::plot
calculate it's own range which is different from mine.
Reading around online, I found a possible option which is to play with the coltab
argument. I tried it 2 ways:
coltab(rr) <- data.frame(
val=seq(the_range[1], the_range[2], length.out=31),
col=the_palette_fc(seq(the_range[1], the_range[2], length.out=31)))
plot(rr)
and:
coltab(rr) <- data.frame(
val=values(rr),
col=the_palette_fc(values(rr)))
plot(rr)
But the result do not seem to give the wanted results. Also, there is no legend automatically added which is also necessary. Maybe the coltab argument should be only used for categorical values?
So is there a way in terra
to fined tune the colors palette so white is fix for 0 values?
Upvotes: 3
Views: 1636
Reputation: 47481
terra::plot
actually does have a "breaks" argument; but it is only documented in the examples so it was easy to overlook. You can do
library(terra)
set.seed(1234)
rr <- rast(matrix(rnorm(400, 1.5, 1), nrow=20, ncol=20))
ceil <- values(rr) |> abs() |> max() |> ceiling()
pal <- leaflet::colorNumeric(palette = "RdBu", domain=c(-ceil, ceil), reverse = T)
b <- seq(-ceil, ceil, 1)
plot(rr, type="interval", breaks=b, col=pal(b))
In response to your comment:
For your purposes, you can also use the range argument, I think.
plot(rr, range=c(-ceil, ceil), col=pal(seq(-ceil,ceil,.1)))
And if you use "breaks", you can overwrite the default legend type.
plot(rr, type="cont", breaks=b, col=pal(b))
Upvotes: 7
Reputation: 174468
To set 0 as the midpoint, you can set the domain of your palette to c(-max(rr[]), max(rr[]))
:
the_palette_fc <- leaflet::colorNumeric(palette = "RdBu",
domain = c(-max(rr[]), max(rr[])),
reverse = TRUE)
the_colors <- the_palette_fc(seq(min(rr[]), max(rr[]), length.out = 50))
plot(rr, col = the_colors)
If you want the scale to go from darkest blue to darkest red with a transition of white at 0, it is a bit more involved, but you could do:
the_palette_fc <- leaflet::colorNumeric(palette = "RdBu",
domain = c(-max(rr[]), max(rr[])),
reverse = TRUE)
breakpoints <- seq(min(rr[]), max(rr[]), length.out = 50)
the_colors <- the_palette_fc(c(seq(-max(rr[]), 0, length = sum(breakpoints < 0)),
seq(0, max(rr[]), length = sum(breakpoints > 0))))
plot(rr, col = the_colors)
Upvotes: 4