Lib Interpolator

[sarah]

Dear WAsP Team and WAsP Forum Members,
I am trying to use Lib Interpolator to interpolate between two weather stations, and at some locations I get unrealistic k values (=40). There doesn't seem to be an obvious reason as to why this is happening, as the LIB files from the two weather stations look fine. Any suggestions would be appreciated!

[Niels Gylling Mortensen]

I always use three stations for the interpolation and have never experienced any 'funny looking' interpolated wind atlases. If you haven't already done so, I suggest you contact WAsP support directly at waspsupport@risoe.dtu.dk, including your data files. This seems to be a problem for the Lib Interpolator developer.

[sarah]

Thanks for the reply. I have contacted WAsP support and am waiting to hear from them. By the way I did do the interpolation with three stations, and just reduced to two to try and identify the problem (so I know which of the three stations is the one causing the problem, I just don't know why).

[Morten]

Sarah submitted her data to wengsupport@risoe.dtu.dk


The unrealistic Weibull distributions in the interpolated wind atlas occurred in a sector with very low wind speed and frequency of occurrence, so the error would probably not disturb WAsP predictions much. It was possible to work with the interpolated Wind Atlas in WAsP itself, but Sarah had a problem using it in Windfarmer (WF). I suspect WF simply rejects very strange Weibull distributions. I suggested a crude correction for Sarah, but I do not yet know whether it worked, so I will skip it from this reply.


The explanation for the strange interpolation results lies in the LibInt method, which is to

1) calculate first and third moments of wind speed in each sector by reference station A and k values
2) make a directional interpolation of frequency and wind speed moments
3) find A and k values matching the interpolated wind speed moments

The k values at the site of interpolation are bounded by kmin=0.5 and kmax=40 and a problem may occur if the interpolated third order moment is incompatible with the interpolated first-order moment. In this case the solution will be impossible or clearly out of range, so LibInt sets the k value to the nearest default boundary value and reports an A values matching this and the mean wind speed.


Finally, a little warning about LibInt taken from the help file:
“The method is just a mathematical interpolation unrelated to physics and it is difficult to describe its credibility in an objective manner. However, I suggest that the following questions are checked:
• Are the data from the reference sites measured with similar equipment?
• Are measurement periods and sample intervals comparable?
• Are the WAsP RIX numbers of the reference sites comparable?
• Are the wind atlases of the reference sites comparable?
• Are other climates parameters, e.g. temperature, humidity, and height above sea level, comparable?
• Finally, are the conditions at the new site expected to be comparable to conditions at the reference sites?
Obviously, it is a good thing if you truly can ask YES to all these questions. Unfortunately, we have no guidelines as to how much you could violate the ideal conditions.”


Sarah’s data seemed to be of good quality and the terrain was not complicated. But the observation periods for the reference wind atlases did not match and two of these did not even overlap. The frequency for the sector giving trouble was only 3%, so the corresponding Weibull distribution may not represent the long-term distribution well, even for a two-year observation period. This is probably what causes the problems for the interpolation.