Better hillshading for map plots in R

Question

I'm working on improved hillshading for some topographical map plots. The basic hillshade workflow documented in image() is:

require(raster)
alt = getData('alt', country='CHE')
slope = terrain(alt, opt='slope')
aspect = terrain(alt, opt='aspect')
hill = hillShade(slope, aspect, 40, 270)
plot(hill, col=grey(0:100/100), legend=FALSE, main='Switzerland')
plot(alt, col=rainbow(25, alpha=0.35), add=TRUE)

This image shows plot(hill..) before plot(alt..) is applied:

The method creates a solid grey under-layer of hillshades on which other data layers (e.g. elevation shading) are plotted semi-transparently. The problem with this approach is (a) that the neutral colour for flat terrain (RBG (202,202,202), '#CACACA') severely shades the whole model, which (b) prevents multiple shade layering, such as used by the 'Swiss hillshade' approach.

I can imagine an workaround that converts rasters to matrices and applies hillshading as a numerical multiplier to the brightness of other layers, but this doesn't seem very elegant (although I may be wrong). I wonder if anyone has any ideas or (preferably) experience in this area? Thanks in advance.

Answer 1

No experience with this, but why not give the gray undermap an alpha value that depends on slopes? Here's my try:

# before
require(raster)
alt = getData('alt', country='CHE')
slope = terrain(alt, opt='slope')
aspect = terrain(alt, opt='aspect')
hill = hillShade(slope, aspect, 40, 270)
plot(hill, col=grey(0:100/100), legend=FALSE, main='Switzerland')
plot(alt, col=rainbow(25, alpha=0.35), add=TRUE)

As you say, very dark.

# after
grayalphas <- seq(-1,1,by=.01)^2*100
grayalphas[grayalphas==100] <- 99
plot(hill, col=paste0(grey(0:100/100),sprintf("%.2d",grayalphas)), legend=FALSE, main='Switzerland')

plot(alt, col=rainbow(25, alpha=0.35), add=TRUE)

I set the gray alphas to have a parabolic shape, with minimum where the gray value is .5 and max of 99 at gray values of 0 or 1. If you choose something like this, you'll want to tinker with levels, etc, but it is easy to implement. Plus you'll want to put more effort than I did into the alphas, as mine are strictly numeric and not hex.

[Edit] I found a nifty function for adding alphas, addTrans() here in Sacha Epskamp's answer. This keeps the parabola, but it ranges from 0 in the middle to 255 on the extremes.

grayalphas <- seq(-1,1,length=101)^2*255
plot(hill, col=addTrans(grey(0:100/100),grayalphas), legend=FALSE, main='Switzerland')
plot(alt, col=rainbow(25, alpha=0.35), add=TRUE)