Cluster wind direction

anyone can help me about how to cluster wind direction? i want to cluster wind in degrees 1-360 so i convert to x and y in cartesian just like this ... is it okay ??

i measure (d), it is a distance between point (0,0) to poitn at circle any suggestion??

• Do you have a dataset you're trying to cluster? It's not clear what you're actually asking here – shadowtalker Sep 5 '16 at 5:01
• Clustering wind direction strikes me as unphysical for most stations. The exceptions would be where local conditions promote strong bimodality, e.g anabatic and katabatic winds, onshore and offshore winds. – Nick Cox Sep 5 '16 at 8:35

Use hierarchical clustering, and the difference in angle modulo 360. i.e.

dist(a,b) = min(abs(a-b),abs(a-(360-b)),abs((360-a)-b))


If you want to use Cartesian coordinates, center at 0, i.e. use

x, y = sin(a), cos(a)


Why would you use a circle that is not centered at 0?

• Please check my edits preserve your meaning. – Nick Cox Sep 5 '16 at 8:33
• i have no idea about how to clouster wind direction, i have 30 years dataset in 425 grids, so make just like the picture, i measure a distance (d) which measure the distance between point (0,0) to the circle, the circle represent represent wind direction in 1-360 degrees, what do you think? im sorry i just newbie in clustering @NickCox – Novi Fitrianti Sep 7 '16 at 4:49