I've read several articles about how to perform Weibull distribution but they all did it with wind speed data in a time series manner (example: recorded data every 10mins and then averaged to every hour). My prob is that I've got only mean monthly wind speed for a year (from January to December). So how should I deal with this situation to perform Weibull distribution on these data?

I have data for 5 years namely 2006, 2007, 2008, 2009, 2010 and like i said, its monthly wind speed from January to December. I'd appreciate if you could help out. Thank you in advacnce...

  • $\begingroup$ Actually I need to use the collected data as stated above to do some forecasting. I need to show if a particular area where the data were recorded are apt to be used as a site for installation of wind turbines. So i needed a model which will allow me to predict this with the least error possible and I found Weibull distribution is used mostly for this purpose. But I can't find a way how to apply Weibull distribution to my recorded data. Regards: Arvind $\endgroup$
    – user8834
    Jan 30, 2012 at 11:13
  • $\begingroup$ Arvind, could you confirm that you are the same user as the original poster (in this case, we could merge your two accounts, but you will still need to register). $\endgroup$
    – chl
    Jan 30, 2012 at 14:05
  • $\begingroup$ Hi @Arvind, having seen your comment above I think my answer (below) is probably the right one; I'd worry about other things than the distribution of your data. I'd take a pretty empirical approach to your monthly data, and if it can be approximately Normal work form there with some reasonably robust method that can test the site against the minimum value of the key parameters necessary for it to be suitable. $\endgroup$ Feb 2, 2012 at 11:37

1 Answer 1


The first question is whether the Weibull distribution is still a good model of wind speed when it is a monthly average. I'm not familiar with this field, but from what you say it sounds as though average hourly wind speed is often modelled as having a Weibull distribution. If you take the average of 700 or so such random variables (24*30) the distribution will be very nearly normally distributed because of the central limit theorem, even with the autocorrelation of the underlying hourly observations.

I'd suggest looking at the actual distribution of your 60 data points and comparing it to a normal distribution, using something like the qqnorm() function in R to draw a plot being the obvious starting point.

But basically, I doubt the probability distribution of your variables is your main problem here. The challenge with modelling your data will be more related to the need for you to estimate and control for seasonal effects from quite a small data set (by time series standards). Exactly what sort of challenge that is though depends on what your research question is (for example, do you need to see if one year's wind speed was different from others? or are you looking for a trend? or what?).


Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.