Separate land and water using two gradients in ggplot

Question

I'm trying to get two separate gradients for land and water on a map. I am able to get the water gradient (first figure), but not the land.

How can I set a grey gradient for land in this code (similar to figure 2 below)?

Sample data

library(marmap)
library(ggplot2)

greys <- c(grey(0.6), grey(0.93), grey(0.99))

# Get map data
bat <- getNOAA.bathy(-68, -51, -48, -39, res = 1, keep = TRUE)

Generate Plot with water gradient

autoplot(bat, geom = c("raster", "contour"), colour = "white", size = 0.1) + 
  scale_fill_gradientn(limits = c(-6600, 0), colors = c("steelblue4", "#C7E0FF")) +
  NULL

I have tried setting different limits in scale_fill_gradientn, but without much luck:

autoplot(bat, geom = c("raster", "contour"), colour = "white", size = 0.1) + 
  scale_fill_gradientn(limits = c(min(bat), max(bat)), 
                       colors = c("steelblue4", "#C7E0FF", greys)) +
  NULL

Desired output (done using base R's plotting functions)

plot(bat, image = TRUE, land = TRUE, lwd = 0.1, bpal = list(c(0, max(bat), greys), c(min(bat), 0, blues)))
plot(bat, lwd = 0.8, deep = 0, shallow = 0, step = 0, add = TRUE) # highlight coastline

Your question reminds me of [this post](https://stackoverflow.com/questions/18347147/using-two-scale-colour-gradients-ggplot2). Although in that post, they exploit the `color` and `fill` aesthetics of `geom_point()`, and you want 2 different mappings of `fill` to 2 different `geom_raster()` layers. This is a good question. — Rich Pauloo, Apr 25 '19 at 01:51
One workaround would be to `cut()` the continuous elevation data into categorical bins, and manually specify all of the colors with `scale_fill_manual()`. — Rich Pauloo, Apr 25 '19 at 01:56
@RichPauloo Yes, I've thought about using `scale_fill_manual()` but was hoping for an automatic solution of the gradient since I have different maps to make. Do you have an example using it? — Vedda, Apr 25 '19 at 02:06
So far as I know there are two ways to have multiple scales. I'd invite you to have a look at https://github.com/clauswilke/relayer and the ggnewscale package. I think both options would require your water fill colour to be a seperate column from the land fill colour, but I'm unfamiliar with your data format or autoplot, so I wouldn't know exactly how to do it. — teunbrand, Apr 25 '19 at 07:41

Rich Pauloo · Answer 1 · 2019-04-30T07:22:11.483

Approach with `cut` and `scale_fill_manual`:

It's much more involved than @Z.Lin's answer (which I personally would go with), but this method potentially gives you more control, and the geom_raster plots pretty fast.

I've done it all manually, but you can imagine a function that takes a vector of elevations and a number of desired breaks, then cuts the vector into a set of categorical breaks and makes appropriate labels. This is what sp::spplot() does by default with continuous fields, and it uses nbreaks = 16. You'll need to force the zero elevation break to distinguish between land and ocean.

Here's the general idea that can be developed.

# convert to raster, then data frame
library(raster)
d <- as.raster(bat)
d <- as.data.frame(d, xy=TRUE)

# upper and lower elevation bounds
z <- max(d$layer)
a <- min(d$layer)

# breaks and labels for color scale
brks <- c(a, 1000, 500, 0, -500, -1000, -2000, -3000, -4000, -5000, z)
labs <- c("> 1000", "500:1000", "0:500", "-500:0", "-1000:-500", "-2000:-1000",
          "-3000:-2000", "-4000:-3000", "-5000:-4000", "< -6514")
d$bin <- cut(d$layer, breaks = brks, labels = labs)
d <- d[!is.na(d$bin), ] # filter sneaky NA values

library(colormap)
gr   <- colormap(colormaps$greys, nshades = 10)[4:6]
bl   <- colormap(colormaps$velocity_blue, nshades = 13)[3:9]
cols <- c(bl, gr)

# plot
ggplot(d, aes(x, y, fill = bin)) +
  geom_raster() +
  scale_fill_manual(values = cols, limits = labs, labels = rev(labs)) +
  theme_minimal() +
  labs(fill = "Elevation (ft)")

Nice one. Just want to add that `marmap::fortify.bathy` can be used for dataframe conversion. I found it while looking at the code behind `marmap::autoplot.bathy`. :) — Z.Lin, Apr 25 '19 at 02:48
Ah that's nice to know and removes the raster dependency. That just gave me another thought towards categorical binning of the elevations could be defining cuts, converting to polygons, and then using `geom_polygon()`. I still think your answer is the most simple and helpful though! — Rich Pauloo, Apr 25 '19 at 02:51

score 2 · Accepted Answer · answered Apr 25 '19 at 02:30

Maybe something like this? Range of fill values is obtained from checking summary(fortify.bathy(bat)) for z's range.

autoplot(bat, geom = c("raster", "contour"), colour = "white", size = 0.1) + 
  scale_fill_gradientn(values = scales::rescale(c(-6600, 0, 1, 1500)),
                       colors = c("steelblue4", "#C7E0FF", "grey50", "grey80"))

Note: I used c("grey50", "grey80") as the range for grey values here, but of course you can change that based on your requirements.

Separate land and water using two gradients in ggplot

2 Answers2

Approach with cut and scale_fill_manual:

Approach with `cut` and `scale_fill_manual`: