Visualising marine heatwaves (MHWs)
Default MHW visuals
One may use event_line() and lolli_plot() directly on the output of detect_event() in order to visualise MHWs. Here are the functions being used to visualise the massive Western Australian heatwave of 2011:
event_line(mhw, spread = 180, metric = "intensity_max",
start_date = "1982-01-01", end_date = "2014-12-31")
lolli_plot(mhw, metric = "intensity_max")
Custom MHW visuals
The event_line() and lolli_plot() functions were designed to work directly on the list returned by detect_event(). If more control over the figures is required, it may be useful to create them in ggplot2 by stacking geoms. We specifically created two new ggplot2 geoms to reproduce the functionality of event_line() and lolli_plot(). These functions are more general in their functionality and can be used outside of the heatwaveR package, too. To apply them to MHWs and MCSs first requires that we access the climatology or event dataframes within the list that is produced by detect_event(). Here is how:
# Select the region of the time series of interest
mhw2 <- mhw$climatology %>%
slice(10580:10720)
ggplot(mhw2, aes(x = t, y = temp, y2 = thresh)) +
geom_flame() +
geom_text(aes(x = as.Date("2011-02-25"), y = 25.8, label = "the Destroyer\nof Kelps"))
ggplot(mhw$event, aes(x = date_start, y = intensity_max)) +
geom_lolli(colour = "salmon", colour_n = "red", n = 3) +
geom_text(colour = "black", aes(x = as.Date("2006-08-01"), y = 5,
label = "The marine heatwaves\nTend to be left skewed in a\nGiven time series")) +
labs(y = expression(paste("Max. intensity [", degree, "C]")), x = NULL)
Spicy MHW visuals
The default output of these function may not be to your liking. If so, not to worry. As ggplot2 geoms, they are highly malleable. For example, if we were to choose to reproduce the format of the MHWs as seen in Hobday et al. (2016), the code would look something like this:
# It is necessary to give geom_flame() at least one row on either side of
# the event in order to calculate the polygon corners smoothly
mhw_top <- mhw2 %>%
slice(5:111)
ggplot(data = mhw2, aes(x = t)) +
geom_flame(aes(y = temp, y2 = thresh, fill = "all"), show.legend = T) +
geom_flame(data = mhw_top, aes(y = temp, y2 = thresh, fill = "top"), show.legend = T) +
geom_line(aes(y = temp, colour = "temp")) +
geom_line(aes(y = thresh, colour = "thresh"), size = 1.0) +
geom_line(aes(y = seas, colour = "seas"), size = 1.2) +
scale_colour_manual(name = "Line Colour",
values = c("temp" = "black",
"thresh" = "forestgreen",
"seas" = "grey80")) +
scale_fill_manual(name = "Event Colour",
values = c("all" = "salmon",
"top" = "red")) +
scale_x_date(date_labels = "%b %Y") +
guides(colour = guide_legend(override.aes = list(fill = NA))) +
labs(y = expression(paste("Temperature [", degree, "C]")), x = NULL)
It is also worth pointing out that when we use geom_flame() directly like this, but we don’t want to highlight events greater less than our standard five day length, allowing for a two day gap, we want to use the arguments n and n_gap respectively.
mhw3 <- mhw$climatology %>%
slice(850:950)
ggplot(mhw3, aes(x = t, y = temp, y2 = thresh)) +
geom_flame(fill = "black", alpha = 0.5) +
# Note the use of n = 5 and n_gap = 2 below
geom_flame(n = 5, n_gap = 2, fill = "red", alpha = 0.5) +
ylim(c(22, 25)) +
geom_text(colour = "black", aes(x = as.Date("1984-05-16"), y = 24.5,
label = "heat\n\n\n\n\nspike"))
Should we not wish to highlight any events with geom_lolli(), plot them with a colour other than the default, and use a different theme, it would look like this:
ggplot(mhw$event, aes(x = date_peak, y = intensity_max)) +
geom_lolli(colour = "firebrick") +
labs(x = "Peak Date",
y = expression(paste("Max. intensity [", degree, "C]")), x = NULL) +
theme_linedraw()
Because these are simple ggplot2 geoms possibilities are nearly infinite.
