One of the “conceptual branches” of metR
is the visualization tools. These are a set of functions that interface with ggplot2 for easier and better plotting of meteorological (an other) fields.
# Packages and data use throught
library(metR)
library(ggplot2)
library(data.table)
temperature <- copy(temperature)
temperature[, air.z := Anomaly(air), by = .(lat, lev)]
This geom has a bit of a long story. Before ggplot2 introduced geom_contour_filled()
, the ‘canonical’ way to get filled contours was by using stat_contour()
with a polygon
geom and mapping fill
to level
(see this issue). This was ugly and bad and didn’t work in most cases, so I threw together some hacks to make it work, and thus metR::geom_contour_fill()
was born.
Then, Claus Wilke created {isoband}, which lead to ggplot2 finally getting a proper implementation of filled contours. Since the implementation worked much better than my Frankenstein’s monster hacks and workarounds, I decided to scrub it and adopt {isoband} as well.
So after all this history the R landscape is left with metR::geom_contour_fill()
and ggplot2::geom_contour_filled()
with almost identical implementations. Why should you use one or the other? Why didn’t I just deprecate this seemingly redundant function? Because, in the intervening time metR::geom_contour_fill()
has evolved and it now provides many functionality that is missing from ggplot2’s counterpart.
breaks
can be a function.The main control parameter of the contour family of functions it their breaks
argument, which sets the levels at which to draw contour lines. In geom_contour_fill()
this argument can be a function that takes a binwidth and the range of the data. This allows for great flexibility.
For example, is is very common in meteorological science to not draw contours at the zero level. Passing MakeBreaks()
as the breaks
argument allows you to control not only the binwidth, but also which levels not to draw.
ggplot(temperature[lon %~% 180], aes(lat, lev, z = air.z)) +
geom_contour_fill(breaks = MakeBreaks(binwidth = 2, exclude = 0)) +
scale_fill_divergent(breaks = MakeBreaks(binwidth = 2, exclude = 0))
Both geom_contour_fill()
and geom_contour_filled()
will work fine with missing values, but they will leave a somewhat ugly empty patch when they are found.
data(volcano)
volcano <- setDT(reshape2::melt(volcano))
volcano[, value.gap := value]
volcano[(Var1 - 40)^2 + (Var2 - 35)^2 < 50, value.gap := NA]
ggplot(volcano, aes(Var1, Var2, z = value.gap)) +
geom_contour_fill()
This is an OK behaviour, but sometimes it’s more visually appealing to fill in these missing areas. The na.fill
argument allows for some control over how this values are imputed. It can take a number, a function function that takes the vector of values and returns one number, or TRUE
, in which case it will use metR::Impute2D()
to interpolate the missing values.
ggplot(volcano, aes(Var1, Var2, z = value.gap)) +
geom_contour_fill(na.fill = TRUE)
#> Warning: Imputing missing values.
Contouring only works on regular grids (although see {contoureR} for an implementation that accepts non-regular girds). Sometimes the data is regular in some coordinate system and you need to plot in another. For example, this data is not regular on a a longitude by latitude coordinate system, but it is regular on the x y coordinates which is a lambertian grid.
data(surface)
ggplot(surface, aes(lon, lat)) +
geom_point(aes(color = height))
ggplot(surface, aes(x, y)) +
geom_point(aes(color = height))
If you know the projection string, you can pass it to the proj
argument and get filled contours in longitude by latitude space (this needs {proj4} to be installed).
proj_string <- "+proj=lcc +lat_1=-30.9659996032715 +lat_2=-30.9659996032715 +lat_0=-30.9660034179688 +lon_0=-63.5670013427734 +a=6370000 +b=6370000"
ggplot(surface, aes(x, y)) +
geom_contour_fill(aes(z = height))
ggplot(surface, aes(x, y)) +
geom_contour_fill(aes(z = height), proj = proj_string)
You can actually pass any arbitrary function to proj
. It should take an iso object as returned by the isoband package with the contour information and return the modified contours.
The clip
argument takes a simple feature polygon and clips the contours to the interior of the polygon.
Clipping is done after proj
is applied.
argentina <- rnaturalearth::ne_countries(country = "argentina", returnclass = "sf")
ggplot(surface, aes(x, y)) +
geom_contour_fill(aes(z = height), proj = proj_string, clip = argentina)
If everything fails and your data is not on a regular grid on any coordinate system, then you can use kriging to interpolate it to a regular grid and then plot it using contours. Setting the kriging
argument toTRUE
uses {kriging} to krige (is that a verb?) the data before computing the contours. Here, as an example, the volcano dataset is sampled and perturbed to create a dataset of irregularly spaced points. Using kriging, geom_contour_fill()
can recover the general shape.
set.seed(42)
some_volcano <- volcano[sample(.N, .N/7)] # select 70% of the points
some_volcano[, Var1 := Var1 + rnorm(.N)] # add some random noise
some_volcano[, Var2 := Var2 + rnorm(.N)]
ggplot(some_volcano, aes(Var1, Var2)) +
geom_point(aes(color = value))
ggplot(some_volcano, aes(Var1, Var2)) +
geom_contour_fill(aes(z = value), kriging = TRUE) +
geom_point(size = 0.2)
breaks
between facetsggplot2’s geom_contour_filled()
computes breaks using all the data. This is great because it means that breaks are consistent between panels when using facet_grid()
. But if you want to compute breaks separately for each panel, you can use geom_contour_fill()
’s global.breaks
argument.
With the default, separation between levels is always the same. Good for comparison, but in this case the panel on the left lacks detail.
ggplot(temperature[lev %in% c(1000, 300)], aes(lon, lat, z = air.z)) +
geom_contour_fill() +
scale_fill_divergent() +
facet_grid(~lev)
With global.breaks = FALSE
, binwidth is no longer the same in each panel and now you can see more detail in the leftmost panel.
ggplot(temperature[lev %in% c(1000, 300)], aes(lon, lat, z = air.z)) +
geom_contour_fill(global.breaks = FALSE) +
scale_fill_divergent() +
facet_grid(~lev)
All of these features are also available to regular contours in the geom_contour2()
function.
Like ggplot::geom_contour_filled()
, geom_contour_fill()
computes several relevant variables. The two most important ones are level_mid
(also called int.level
for backwards compatibility reasons) and level
. The former (the default) is a numeric value that corresponds to the midpoint of the levels while the latter is an ordered factor that represents the range of the contour.
ggplot(temperature[lev == 300], aes(lon, lat, z = air.z)) +
geom_contour_fill()
ggplot(temperature[lev == 300], aes(lon, lat, z = air.z)) +
geom_contour_fill(aes(fill = after_stat(level)))
A big difference between the two variables is that one is compatible with continuous scales while the other only works with discrete scales. The advantage of using level_mid
is that it preserves the actual continuous nature of the data, which allows the use of meaningful continuous scales such as scale_fill_divergent()
and results in proper colours when breaks are not equally spaced. The disadvantage is that the scale function has no information about the discretisation, and thus is very hard to use synchronise the breaks that define the levels of the contours and the breaks shown in the color guide.
A third important computed variable is level_d
, which is identical to level
but forces the scale to match the breaks in the data automatically.
ggplot(temperature[lev == 300], aes(lon, lat, z = air.z)) +
geom_contour_fill(aes(fill = after_stat(level_d)))
See the discretised_scale section for a way of using continuous sales for discretised data.
Labelling contours is also a problematic aspect of ggplot2
. geom_text_contour()
and geom_label_contour()
can be use to automatically add text or labels to the flattest part of a contour.
ggplot(temperature[lev == 300], aes(lon, lat, z = air.z)) +
geom_contour_fill() +
geom_contour(color = "black") +
geom_text_contour() +
scale_fill_divergent()
By default it labels every 2nd contour (this can be changed by the skip
parameter) and it rotates to follow the angle of the contour (this is not available on geom_label_contour()
). Since in some datasets there can be very small contours that should not be labelled for clarity, the min.size
argument specifies the minimum points a contour has to have in order to be labelled.
Notice how labels are drawn on top of contours? The problem is that geom_contour()
doesn’t know it’s being labelled; geom_text_contour()
addresses this issue by allowing you to draw a stroke around the text.
ggplot(temperature[lev == 300], aes(lon, lat, z = air.z)) +
geom_contour_fill() +
geom_contour2(color = "black") +
geom_text_contour(stroke = 0.2) +
scale_fill_divergent()
You can have full control over where labels are placed with the label.placement
argument. It takes a function that takes a vector of x and y values and returns a logical vector of the same length that indicate the points where labels should be. There are 5 build in placement functions:
label_placement_flattest()
: The default. It computes the curvature and the angle at each points and then selects the point which minimizes the sum of both numbers. Therefore, it tries to label the flattest and most horizontal part of the contour.
label_placement_fraction()
: It takes a vector of numbers ranging from 0 to 1 representing where tu put the label in relation to the length of the contour. That is, using 0.5 means to put the label at the midpoint.
label_placement_n()
: Places n
labels approximately equally spaced along the contour.
label_placement_random()
: It puts the label in random places. ¯\_(ツ)_/¯ .
label_placement_all()
: It puts the label at every point. Probably is best not to use it.
ggplot(temperature[lev == 300], aes(lon, lat, z = air.z)) +
geom_contour_fill() +
geom_contour2(color = "black") +
geom_text_contour(stroke = 0.2, label.placement = label_placement_random()) +
scale_fill_divergent()
#> Warning: 'label_placement_random' is deprecated.
#> Use 'label_placer_random' instead.
#> See help("Deprecated")
#> Warning in geom_text_contour(stroke = 0.2, label.placement =
#> label_placement_random()): The 'label.placement' argument is now
#> 'label.placer'.
#> Warning in geom_text_contour(stroke = 0.2, label.placement =
#> label_placement_random()): Ignoring unknown parameters: `label.placement`
Illuminated contours (aka Tanaka contours) use varying brightness and width to create an illusion of relief. This can help distinguishing between concave and convex areas (local minimums and maximums), specially in black and white plots. It also allows for photocopy safe plots with divergent colour palettes, and it just looks cool.
ggplot(temperature[lev == 300], aes(lon, lat, z = air.z)) +
geom_contour_fill() +
geom_contour_tanaka() +
scale_fill_divergent()
metR
also has stat_subset()
which makes a subset of the data according to the subset
aesthetic. This makes it possible to show only part of the data in one geom without needing to specify a data
argument (specially useful if the data being plotted is the result of a long pipe and not actually assigned to a data.frame
). It has a somewhat limited use in that it cannot perform further statistical transformations of the data.
For example, it can be used if you have a correlation field and want to mark only the points with significant correlations:
data(geopotential) # geopotential height at 700hPa for the Southern Hemisphere.
ggplot(geopotential[, gh.base := gh[lon == 120 & lat == -50], by = date][
, .(correlation = cor(gh.base, gh)),
by = .(lon, lat)],
aes(lon, lat, z = correlation)) +
geom_contour_fill(breaks = MakeBreaks(0.1)) +
stat_subset(aes(subset = correlation > 0.5),
geom = "point", size = 0.1) +
scale_fill_divergent()
Another possible use it to quickly mask missing values.
ggplot(volcano, aes(Var1, Var2, z = value.gap)) +
geom_contour_fill(na.fill = TRUE) +
stat_subset(aes(subset = is.na(value.gap)), geom = "raster",
fill = "#a56de2")
Plotting arrows can be a pain. Again, the ‘canonical’ way of plotting vectors is to use geom_segment()
and specify x
, y
, xend
and yend
aesthetics which can be a lot of typing when one has the data on location and displacement (or velocity). Instead, metR
’s geom_vector()
and geom_arrow()
draw vectors defined by their lateral displacements (dx, dy) or their magnitude and angle. It also has some useful parameters like min.mag
, which controls the minimum magnitude for an arrow to be drawn (useful for highlighting only areas of strong ‘flow’) and skip
, which draws only the nth arrow in the x and y directions.
Both geoms are essentially the same, except that geom_arrow()
defaults to preserving direction in coordinate transformations and regardless of the plot aspect ratio. This is recommended if the arrows are meant to convey the direction of the flow in each point instead of being a description of the shape of the flow related to the plot coordinates.
So, as an example, we can plot the temperature gradient like this:
temperature[, c("t.dx", "t.dy") := Derivate(air.z ~ lon + lat,
cyclical = c(TRUE, FALSE)),
by = lev]
(g <- ggplot(temperature[lev == 500], aes(lon, lat)) +
geom_contour_fill(aes(z = air.z)) +
geom_vector(aes(dx = t.dx, dy = t.dy), skip.x = 2,
skip.y = 1) +
scale_mag())
In the above example, t.dx and t.dy represent displacements in the same units as the x and y dimension, so using geom_vector()
each arrow points perpendicular to the contours even if we change the coordinate system.
g + coord_polar()