Stickplots

library(oce)
d <- read.csv('wind.csv', stringsAsFactors = FALSE)
d$time <- as.POSIXct(d$date, tz='UTC')

Note that the date is provided only as days, but they appear to be hourly observations, so I’m going to fudge the times:

d$time <- d$time + rep(seq(0, 86399, 3600), length(unique(d$date)))

Now calculate the \(u\) and \(v\) components:

d$u <- d$speed*cos(d$direction*pi/180)
d$v <- d$speed*sin(d$direction*pi/180)

Before we try a stick plot, let’s just look at the \(u\) and \(v\) time series:

par(mfrow=c(2, 1))
oce.plot.ts(d$time, d$u, ylab='u')
oce.plot.ts(d$time, d$v, ylab='v')

If we want to filter the data to smooth it for the stick plot, we can do so using a number of methods. One option is to use the signal package, and specify a filter, e.g. Butterworth.

library(signal)
bw <- butter(1, 0.1) # lowpass with normalized cutoff of 0.1
t <- d$time
uf <- filtfilt(bw, d$u)
vf <- filtfilt(bw, d$v)
par(mfrow=c(2, 1))
oce.plot.ts(t, d$u, ylab='u')
lines(t, uf, col=2)

oce.plot.ts(t, d$v, ylab='v')
lines(t, vf, col=2)

And make a plot:

plotSticks(t, 0, d$u, d$v, yscale=10, ylab='', yaxt='n', length=0)

If we use the filtered data instead, it looks a little cleaner:

plotSticks(t, 0, uf, vf, yscale=10, xlab='', ylab='', yaxt='n', length=0)

Another way of plotting wind data is to use a windrose object, and the associated plotting methods:

plot(as.windrose(d$u, d$v))