Polar Analysis

Introduction

An openair polar plot looks like this:

openair::polarPlot(ggopenair::marylebone)

To achieve the same result in ggopenair one would write:

library(ggopenair)
library(ggplot2)

polar_plot(marylebone, "nox") +
  theme_polar() +
  scale_opencolours_c()

This is more long winded, but the flexibility allows users to customise their outputs very closely. For example:

Scales

Use any ggplot2 scale function to change how the plot behaves. For example, use scale_color_binned() to bin the colour bar.

polar_plot(marylebone, "nox") +
  theme_polar() +
  scale_color_steps(
    low = "black",
    high = "goldenrod",
    breaks = seq(0, 1000, 40)
  )

Alternatively, one could use the “trans” argument to shift the colour scale. This could be particularly useful for polar_freq(), which had its own “trans” argument in openair.

shift_axis <- function(trans) {
  polar_freq(marylebone, "nox") +
    theme_polar() +
    scale_fill_gradientn(
      colours = c("darkgreen", "hotpink"),
      trans = trans
    ) +
    labs(title = trans)
}

patchwork::wrap_plots(
  shift_axis("identity"),
  shift_axis("sqrt"),
  shift_axis("log10")
)
#> Warning: ✖ `statistic == 'frequency'` incompatible with a defined pollutant.
#> ℹ Setting statistic to `'mean'`.
#> ✖ `statistic == 'frequency'` incompatible with a defined pollutant.
#> ℹ Setting statistic to `'mean'`.
#> ✖ `statistic == 'frequency'` incompatible with a defined pollutant.
#> ℹ Setting statistic to `'mean'`.

Annotations

Use annotate() to easily draw on your polar plots and to draw attention to certain aspects. In-built annotation functions make it easy to, for example, draw a highlighting wedge or direct axis labels.

polar_plot(marylebone, "nox") +
  theme_polar() +
  scale_opencolours_c("inferno") +
  annotate_polar_wedge("S", "W") +
  annotate_polar_axis(seq(5, 15, 5), color = "grey50")

Extensions

Use any of the ggplot2 extension packages out there, such as patchwork. For example, a polar plot could be inserted into a time variation plot.

library(patchwork)

polar <-
  polar_plot(marylebone, "nox") +
  theme_polar() +
  theme(panel.border = element_rect(fill = NA, color = "black")) +
  scale_opencolours_c()

tv <- trend_variation(marylebone, "nox", return = "list")
tv <-
  purrr::map(
    tv,
    ~ .x + theme_classic() + theme(
      legend.position = "none",
      panel.border = element_rect(fill = NA)
    )
  )

tv$day_hour / (tv$month | tv$day | polar) +
  plot_layout(heights = c(.8, 1), guides = "collect")

Other Polar Plots

Currently the main eight openair polar plots have been implemented in ggopenair.

polarplot <-
  polar_plot(marylebone, "nox") + theme_polar() + scale_opencolours_c()
polarannulus <-
  polar_annulus(marylebone, "nox") + theme_polar() + scale_opencolours_c()
polarfreq <-
  polar_freq(marylebone, "nox") + theme_polar() + scale_opencolours_c(trans = "sqrt")
#> Warning: ✖ `statistic == 'frequency'` incompatible with a defined pollutant.
#> ℹ Setting statistic to `'mean'`.

polarperc <-
  rose_percentile(marylebone, "nox") + theme_polar() + scale_opencolors_d()
pollrose <-
  rose_pollution(marylebone, "nox") + theme_polar(panel_ontop = FALSE) + scale_opencolors_d()
rose_wind <-
  rose_wind(marylebone) + theme_polar(panel_ontop = FALSE) + scale_opencolors_d()

patchwork::wrap_plots(polarplot, polarannulus, polarfreq, polarperc, pollrose, rose_wind, nrow = 3)

Note that polar_diff() and polar_cluster() have been developed, but aren’t shown here.

The Three Roses

Note that two of the above “roses” - wind and pollution - have their own annotation functions to put on their. The rose_angle arguments are required if you change the angle argument of wind/rose_pollution due to the way geom_col() interacts with coord_polar() — simply copy the chosen argument for angle and it’ll fix any strangeness when wd is near 0/North.

rose_wind(marylebone, facet = "daylight") +
  theme_polar(panel_ontop = FALSE) +
  annotate_rose_text(y = 15, wd = "SE", size = 2.5)

The functionality to compare two sets of ws/wd data felt a bit misplaced in pollutionRose(), so ggopenair exports a separate rose_metbias() function. This also lets the plot better reflect what it is actually representing (no more N/S/E/W axis labels).