Scatterplot version of heat maps
Introduction
The autoimage package makes it easy to plot a sequence of images with corresponding color scales, i.e., a sequence of heat maps, with straightforward, native options for projection of geographical coordinates. The package makes it simple to add lines, points, and other features to the images, even when the coordinates are projected.
The package also makes it easy to create a “heated scatterplot”, which is similar to a heat map in that color is used to distinguish smaller and larger values of a variable observed over a two-dimensional domain, but is different from a heat map in that the observed data locations are not gridded. Instead of interpolating the data locations onto a grid, the colors are displayed through a typical two-dimensional scatterplot.
The autopoints function is the main function for
creating a sequence of heated scatterplots; it relies on the helper
function heat_ppoints, which can be used in more complex
scenarios. We provide the examples below with brief explanation.
autopoints Examples
We provide examples using the non-gridded Aluminum and Cadmium measurements for the state of Colorado, which are available in the gear package.
Here is a basic heated scatterplot for Aluminum with custom x and y axis labels.
library(autoimage)
data(co, package = "gear")
autopoints(co$longitude, co$latitude, co$Al,
xlab = "lon", ylab = "lat")
In the sets of examples, we display heated scatterplots for Aluminum
and Cadmium while allowing them to have different z limits and z limit
breaks for the color scale. We illustrate a few different combinations
using defaults and by specifying zlim and
breaks as lists.
The following example uses the autopoints default
options, which automatically determines the color scale break labels
using the the base pretty function.
autopoints(co$longitude, co$latitude, co[, c("Al", "Ca")],
xlab = "lon", ylab = "lat", common.legend = FALSE,
main = c("Al", "Ca"))
In the next example, we manually specify the z limits, after using
the summary function to determine appropriate limis.
## Al Ca
## Min. :1.120 Min. : 0.110
## 1st Qu.:4.660 1st Qu.: 0.710
## Median :5.540 Median : 1.230
## Mean :5.531 Mean : 1.963
## 3rd Qu.:6.353 3rd Qu.: 2.375
## Max. :9.980 Max. :21.900autopoints(co$longitude, co$latitude, co[, c("Al", "Ca")],
xlab = "lon", ylab = "lat", common.legend = FALSE,
main = c("Al", "Ca"), zlim = list(c(0, 10), c(0, 22)))
Finally, we manually specify the breaks of each variable, which allows for different numbers of colors for each plot.
autopoints(co$longitude, co$latitude, co[, c("Al", "Ca")],
xlab = "lon", ylab = "lat", common.legend = FALSE,
main = c("Al", "Ca"),
breaks = list(seq(0, 10, by = 2.5),
c(0, 2, 4, 6, 8, 22)))
In the following example, we project the coordinates using the Bonne
projection. This automatically adds longitude/latitude lines
(graticules), which we lighten dramatically through the col
argument of the paxes.args list. The graticules do not
extend far enough by default, so we extend them further through the
axis.args via the xat and yat
arguments. We also add text above the Aluminum and Cadmium plots via the
main argument and a common title via the
outer.title argument. The mtext.args argument
list can be used to customize the outer title.
autopoints(co$longitude, co$latitude, co[, c("Al", "Ca")],
xlab = "lon", ylab = "lat",
proj = "bonne", parameters = 40,
paxes.args = list(col = "lightgrey"),
axis.args = list(yat = 35:43,
xat = -(110:101)),
main = c("(a) Aluminum", "(b) Cadmium"),
outer.title = "Geochemical measurements",
mtext.args = list(col = "blue", cex = 2))## Warning in autolayout(size, legend = legend, common.legend = common.legend, : There is no room in the outer margin for an outer title.
## Setting par(oma = c(0, 0, 3, 0)).
We can add selected borders via the map arguments (which
makes use of the the maps::map function. We can also add
borders/lines via the lines arguments. We demnstrate using USA county
maps (map = "county") and Colorado state borders
(lines = copoly) where copoly is a polygon
list available in the autoimage package. The lines can
be customized through the lines.args argument list.
autopoints(co$longitude, co$latitude, co$Al,
xlab = "lon", ylab = "lat", map = "county",
lines.args = list(col = "magenta"))
data(copoly)
autopoints(co$longitude, co$latitude, co$Al,
xlab = "lon", ylab = "lat", lines = copoly,
lines.args = list(col = "orange", lwd = 3))
Naturally, one may wish to change the color palette or the breaks
associated with the colors. This can be done via the col
and breaks arguments. breaks must be a
sequential vector with length one element longer than col.
We also move the legend scale to the right side.
autopoints(co$longitude, co$latitude, co$Al, legend = "v",
xlab = "lon", ylab = "lat",
col = colorspace::sequential_hcl(n = 4, palette = "Plasma"),
breaks = c(0, 1, 2, 3, 8))
We add the locations and names of two Colorado cities to the Colorado
geochemical data using the points and text
arguments. The appearance of these can be customized through the
points.args and text.args argument lists,
respectively.
citypoints = list(x = c(-104.98, -104.80), y = c(39.74, 38.85),
labels = c("Denver", "Colorado Springs"))
autopoints(co$lon, co$lat, co[,c("Al", "Ca")], common.legend = FALSE,
main = c("Aluminum", "Cadmium"),
points = citypoints,
points.args = list(pch = 6, col = "magenta"),
text = citypoints,
text.args = list(pos = 3, col = "orange"),
xlab = "lon", ylab = "lat")
Richer plots using autolayout and
autolegend
Suppose we want to add custom features to a sequences of images and
heated scatterplots, with each plots receiving different features. One
can create a richer sequence of images using the autolayout
and autolegend functions.
The autolayout function partitions the graphic device
into the sections needed to create a sequence of images. The most
important function arguments include size,
legend, common.legend, and
lratio, which correspond to the same arguments in the
autoimage function. The outer argument
specifies whether an outer.title is desired. The default is
FALSE. By default, numbers identify the plotting order of
the sections, though these can be hidden by setting
show = FALSE. As an initial example, we create a 2 × 3 grid of images with a common vertical
legend.
We now create a complicated (though unrealistic) example of this. We
first extract the borders of Hawaii and Alaska from the
"world" map in the maps package and store
it as the hiak list. We then select a small subset of
cities in Colorado from the us.cities dataset in the
maps package and store this in the codf
data frame. Lastly, we select the U.S. capitals from the
us.cities dataset and store this in the capdf
data frame.
# load world map
data(worldMapEnv, package = "maps")
# extract hawaii and alaskan borders
hiak <- maps::map("world", c("USA:Hawaii", "USA:Alaska"),
plot = FALSE)
# load us city information
data(us.cities, package = "maps")
# extract colorado cities from us.cities
codf <- us.cities[us.cities$country.etc == "CO", ]
# select smaller subset of colorado cities
# extract capitals from us.cities
capdf <- us.cities[us.cities$capital == 2,]Having obtained the relevant information, we setup a 1 × 2 matrix of images with individual
horizontal legends and an area for a common title. We create an image
plot of NARCCAP data using the mercator projection and including grey
state borders. The borders of Hawaii and Alaska are added using the
plines function. The state capitals are added to the image
using the ppoints function. The first image is then titled
using the title function. The legend is then added using
the autolegend function. Next, a heated scatterplots of the
Colorado Aluminum measurements is created using the
heat_ppoints function. The coordinates are projected using
the Bonne projection, the color scheme is customized, grey county
borders are added to the plot, but the grid lines are removed. The
ppoints function is then used to add locations for several
Colorado cities to the image. The ptext function is then
used to add the names of these cities to the image. The second heated
scatterplot is then titled using the title function. The
appropriate legend is then added using the autolegend
function. Lastly, a common title is added using the mtext
function.
## Warning in autolayout(c(1, 2), legend = "h", common.legend = FALSE, outer = TRUE): There is no room in the outer margin for an outer title.
## Setting par(oma = c(0, 0, 3, 0)).
# create image of NARCCAP data.
# xlim is chosen so to include alaska and hawaii
# add grey state borders
# extend graticules (longitude/latitude grid lines)
pimage(lon, lat, tasmax[,,1], legend = "none",
proj = "mercator", map = "state",
xlim = c(-180, 20),
axis.args = list(xat = seq(-175, -25, by = 25),
yat = seq(-10, 80, len = 10)),
lines.args = list(col = "grey"))
# add hawaii and alaskan borders
plines(hiak, proj = "mercator", col = "grey")
# add state captials to image
ppoints(capdf$lon, capdf$lat, proj = "mercator", pch = 16)
# title image
title("tasmax for North America")
# add legend for plot
autolegend()
# load colorado geochemical data
data(co, package = "gear")
# create image for colorado aluminum measurements
# use bonne projection
# customize legend colors
# add grey county borders
# exclude longitude/latitude
heat_ppoints(co$lon, co$lat, co$Al, map = "county", legend = "none",
proj = "bonne", parameters = 39,
paxes.args = list(grid = FALSE),
col = cm.colors(5),
lines.args = list(col = "grey"),
xlab = "lon", ylab = "lat")
# add colorado city points to image
ppoints(codf$lon, codf$lat, pch = 16, proj = "bonne")
# add names of colorado cities to image
ptext(codf$lon, codf$lat, labels = codf$name, proj = "bonne", pos = 4)
# title plot
title("Colorado Aluminum levels (%)")
# add legend to current image
autolegend()
# add common title for plots
mtext("Two complicated maps", col = "purple", outer = TRUE, cex = 2)