Conditions of the atmosphere can be either weather or climate.
Another departure from the tmap vs ggplot2 face-off, given the specialised nature of weather data, and presentations.
If you live in Wellington as I do, then wind is the most salient aspect of the weather. Weather data in Aotearoa New Zealand of any kind (wind or otherwise) are regrettably not easy to get hold of.
After some fruitless searching I stumbled upon the rWind package (of course there’s a package about wind in R called rWind, because… it’s R). I imagine my erstwhile colleagues in climate science at VUW could have told me this right away, if I’d known to ask.
rWind allows relatively easy download of GFS (Global Forecast System) data. In combination with the rasterVis package I managed to cobble together something interesting to look at, if not up to the level of the amazing visualisation of earth.nullschool.net!
That it is indeed windy around Wellington is confirmed—a lot to do with the Cook Strait funneling the ‘Roaring Forties’ our way.
Some different packages this time around. There’s a lot of older formats involved in this particular toolchain hence some of the unfamiliar packages.
library(raster) # note: not terra!
library(sf)
library(sp) # older spatial features type
library(rWind) # wind data downloads
library(rasterVis) # visualisation of vector fields
library(RColorBrewer) # familiar colour palettes
library(latticeExtra) # a whole other graphics systemnz <- st_read("data/nz-small.gpkg") %>%
st_transform(4326) %>%
as("Spatial")
w <- wind.dl(2023, 10, 25, 18, # year, month, day, hour
169, 177, # longitude range
-44, -38) # latitude range
wind <- wind2raster(w) # a raster stack with direction and speedUsing rasterVis vector and stream flow plots.
vectorplot shows only directions but not speeds in its ‘default’ form. To get arrows that reflect wind speed, we have to supply ‘dXY’ format data.
I also densify the data a bit to get something a bit more ‘complete’ seeming.
# get vector components
north <- -cos(pi * wind$direction / 180) * wind$speed
east <- -sin(pi * wind$direction / 180) * wind$speed
wind_vec <- stack(north, east) %>%
disaggregate(2, method = "bilinear")
names(wind_vec) <- c("N", "E")
vectorplot(wind_vec, isField = "dXY", region = TRUE,
lwd.arrows = 1, col.arrows = "blue",
uLayers = 1, vLayers = 2, par.settings = YlOrRdTheme) +
layer(sp.polygons(nz, lwd = 2))
streamplot traces the paths that ‘droplets’ would follow on the surface, so it wants the data in speed, direction format. This is almost the format supplied by rWind::wind2Raster except the order needs reversing! However, since we’ve gone to the trouble of densifying the data into the vector component form, I recombine them here to give a denser interpolated speed-direction pair than we started with.
I don’t entirely know, nor do I have time to get to the bottom of how to drive this function! I’m sure that with a bit more experimentation it could be made to produce a much cooler looking output.
wind_field <- stack(sqrt(wind_vec$N ^ 2 + wind_vec$E ^ 2),
raster::atan2(wind_vec$N, wind_vec$E))
streamplot(wind_field, isField = TRUE,
streamlet = list(L = 8), droplet = list(pc = 100),
par.settings = streamTheme(symbol = brewer.pal(n = 9, 'YlOrRd'))) +
layer(sp.polygons(nz, col = "lightblue2", lwd = 2))