Help choosing most suitable bivariate choropleth mapping coloring scheme

Question

I am trying to overlay a nominal variable (2 categories - rainfed vs irrigated) in a choropleth map over a numeric variable.

In order to highlight the difference in pattern among the two categories, I have outlined one category with a different color (irrigated areas in "red"), while the uncoloured ones are rainfed areas.

However, the map feels cluttered and the outline mixes with darker shades of the numeric variable.

I am trying to avoid using two different color shading schemes to represent the two categories as I feel that would hide the overall distribution of the underlying variable.

Is there a better way to represent this bivariate classification using the ggplots2 package in R (which I am currently using it for making the attached map)?

These maps are eventually going to be printed on paper.

Answer 1

Instead of outlines to indicate the irrigated areas you should use something like a transparent fill pattern (e.g. lines, hachures). An example would look similar to this:

or just google "map fill patterns" to get an overview of the options. Using outlines only for the irrigated areas would give the impression that irrigation is not a continous phenomenon.

Answer 2

I recently had to do a similar map. The solution I came up with uses the rasterVis package, rather than ggplot2 (which is an awesome package, by the way).

In my case, I had a map of trends over time (which is a numerical variable too) and also a map showing the significance of the trend (obtained from a statistical test).

In my case, I wanted to plot the significance map on top of the numerical map, something similar to a "stippling" pattern, to show where the confidence level was higher.

This is roughly the code that I used to create the map.

require(raster)
require(rasterVis)

# Scratch raster objects
data(volcano)
r1 <- raster(volcano)

over <- ifelse(volcano >=160 & volcano <=180, 1, NA) # This is the "mask" raster
r2 <- raster(over)

# And this is the key step:
# To convert the "mask" raster to spatial points
r.mask <- rasterToPoints(r2, spatial=TRUE)

# Plot
levelplot(r1, margin=F) +
layer(sp.points(r.mask, pch=20, cex=0.3, alpha=0.8))

In your case, you could change the code by:

1) creating a "binary" raster, where for example 0 is rainfed and 1 is irrigated, and plot it just like I've done or;

2) creating two different masks, one for rainfed and one for irrigated, and plot both of them as different objects (for example circles for rainfed and crosses for rainfed).

I think the second alternative would make the map look too visually "polluted", but you can always control the parameters of the points by consulting the sp package documentation - specifically ?sp.points.

Hope it helps.