Radius Aesthetics"
In ggfields: Add Vector Field Layers to Ggplots

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.ext = "svg",
  dev = "svg"
)

Radius scales and guides

Size does matter

A vector has a radius (i.e., length) and an angle (direction). When you depict a vector field, the radius often has a specific meaning: it reflects a certain magnitude. For instance, a vector can represent the magnitude and direction of a velocity. It could help your audience to read your plot if they knew which magnitudes the length of the arrows in your plot represent. This is why ggfields provides a legend key for the radius aesthetic specified with ggplot2::aes().

In order to provide a frame of reference, you need to specify max_radius in your call to geom_fields() (default is 0.5 cm). This max_radius should have a absolute or relative length unit (grid::unit()). Arrows in your vector field will be scaled between zero and this max_radius. This same max_radius is also used to scale arrows in your legend keys.

Note that the axes are not trained on the size of the vectors (only on the vector origins). You need to expand your x and y scales manually if vectors are cut off. Also if the max_radius is larger than the key width, you may run into visual problems. For large max_radius values you have to increase the key width manually, as shown in the example below.

library(ggplot2)
library(ggfields)
library(stars)
theme_set(theme_light())
data("seawatervelocity")

p <-
  ggplot() +
  geom_fields(
    aes(radius = as.numeric(v),
        angle  = as.numeric(angle)),
    seawatervelocity[,3:11,6:11],
    max_radius = grid::unit(1.5, "cm")) +
  labs(radius  = "v [m/s]") +
  ## We have to increase the 'keywidth' as otherwise `max_radius` won't fit:
  guides(radius = guide_legend(keywidth = grid::unit(1.5, "cm")))
p

Stay positive

It is important that the radius you wish to plot has only non-negative values. This is because it is impossible to distinguish between a vector with a negative magnitude and a vector with the same positive magnitude with an angle rotate 180 degrees (i.e, a vector with radius = 1 and and angle = 0 will look the same as a vector with radius = -1 and angle = pi). Therefore, from the plot, you cannot tell which vectors have a positive and which have a negative value. This is why geom_fields() will throw an error if you try to plot negative radii.

This is also why you need to be careful with transformations. Although, it is technically possible to log-transform your radii, you might run into problems due to potentially resulting negative values. You should also wonder whether transformed values accurately convey information about the magnitude of your vectors. If you are considering to apply transformations, try using binned or discrete scales instead.

One key to rule them all

The radius key in the guide legend can be combined into one key with many other aesthetics, such as colour, linewidth, linetype, etc. All you have to do is make sure that all aesthetics use are assigned to the same variable and use ggplot2::guide_legend(). The example below illustrates how colour and linewidth can be combined with radius in one legend.

p +
  aes(col = as.numeric(v), linewidth = as.numeric(v)) +
  ## Let's give the aesthetics all the same name
  labs(col = "v [m/s]", linewidth = "v [m/s]") +
  ## Make sure that the colour aesthetic uses the same guide as radius ("legend")
  scale_colour_viridis_c(guide = "legend") +
  ## Make sure all keys have the same width
  guides(colour    = guide_legend(keywidth = grid::unit(1.5, "cm")),
         linewidth = guide_legend(keywidth = grid::unit(1.5, "cm")))

Vector components

What if you only have the x and y component of a vector instead of its angle and radius? Can you provide those components as aesthetics to geom_fields()? Unfortunately, the answer is no. At least not directly, as radius is required early on in the data wrangling machinery of ggplot2::ggplot(). However, the x and y components of a vector can easily be converted into its angle and radius, as shown in the example below.

ggplot() +
  geom_fields(aes(radius = as.numeric(pythagoras(vo, uo)),
                  angle  = atan2(as.numeric(vo), as.numeric(uo))),
              seawatervelocity)