I evaluate the performance of the model through the computation of statistical parameter such as bias, R, IOA and other. (i compare it with in situ measurements) However, wind direction (range 0 to 360 in degrees) has difficult characteristics. Thus, I would like to ask if you know statistics tailored to wind direction.


  • 1
    $\begingroup$ I've run into this issue with latitude and longitude. I solved that issue by treating them as 3D points on the sphere so there's no issue when longitude changes from -180 to +180 at the antimeridian, and so that the differing width of a degree of longitude isn't an issue. You can do something similar by using x and y coordinates $\endgroup$ Mar 22 at 14:56
  • 1
    $\begingroup$ Probably this might be helpful? math.stackexchange.com/questions/44621/… $\endgroup$
    – dimfalk
    Mar 24 at 17:32
  • $\begingroup$ @BarryCarter if I understand well, you propose to get a circle in cartesian coordinates like sin and cos. Interesting idea. To be fair, the link in mathematics by dimfalk states (and it is logical) that apart from the wind direction, the wind speed as absolute value is also important. $\endgroup$ Mar 26 at 7:52
  • $\begingroup$ @dimfalk yes, thanks, the idea of the vector is interesting, but mean value is proposed. I want to evaluate the performance of a model, thus I do need something more. But, I could use the idea of the vector, with other ways. $\endgroup$ Mar 26 at 7:54
  • 1
    $\begingroup$ @DimitrisTsiaousidis Yes, that's one way to solve the "discontinuity at 0/360" issue $\endgroup$ Mar 26 at 13:48


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.