Questions regarding wind data and its processing by the climate analys

[dachiller]

I am writing my Bachelor-Thesis on a project with WAsP and I am kind of new to the software but really like to work accurately. I got a ton of questions and hope it is okay if I ain't to shy to ask them. I would really appreciate some advice from experienced users. Right now its concerning wind data and the climate analyst.

1. Does the averaging interval of wind speeds/directions have an impact on the produced observed wind climate? I guess so, because it is requested. To put it differently: I want to simulate a wind climate and compare it with measured data from the same time periods to test the accuracy of predictions. Can I do this as simple as that with different averaging intervals and also in general?

2. I guess a few missing data won't hurt with this intention and I plan to exclude bigger data gaps respectively from simulation and comparison data. But I also got different time steps. I guess thats no problem?

3. I got data from ncdc which sometimes doesn't include direction values but an indicator for 'variable wind directions'? How does the climate analyst cope with that. Excluding these data don't seem like an solution, because it occurs especially in calm conditions.

4. Also the minimum wind speed seems to be 1.5m/s which seems pretty high to me. Is this a problem?

5. Although I was wondering about wind shadows.
a) Do you know if ncdc data is corrected concerning wind shadows?
b) Do you think wind shadow correction is necessary for such an analysis/ are there any experience values of the magnitude of that mistake? I would assume it is pretty small in comparison to the interpolating error of wasp?

6. Do you know the averaging interval of ncdc data? It's kinda hard to find and what I found (2 mins) seems pretty unusual.

This got quite long :o. Thanks in advance for every response!

[Duncan]

Hello,

I don't know anything about the NCDC data, but I can tell you a bit about the way WACA handles the data.


The averaging interval is not actually used by our analysis code, but it's recorded and preserved along with the data so that you can check it later from the OWC reports in WAsP if you need. Typically, the averaging period (sampling period) is the same as the recording interval ("stride"), but this is not necessarily the case.


Missing data are OK if they are relatively few, and more or less randomly scattered over the time series. If you've got a solid bunch of data missing then you might decide to reject the whole year in which it falls. It depends if the missing data are biased with respect to the annual climate.


The minimum wind speeds and variable directions are both questions of calms. The variable wind direction probably means that the wind speed has fallen below some threshold for the vane to work properly. This may or may not be the same as the speed threshold for the anemometer to work properly, but they are often treated as the same. If the minimum wind speed in the data set is 1.5 and then you have lots of zeros, then I'd assume that the calm threshold has already been applied. So in the WACA, set the calm threshold for both data fields to be 1.5, and proceed. I would say that 1.5 is pretty high.


In the WAsP system, the mast and boom shadow would be corrected for when the GWC is analysed from the OWC data. There's no directional-dependent way to adjust the data in the WACA.


This doesn't answer all your questions, but I hope it helps with some.


Duncan.

[dachiller]

First of all: sorry I wasn't answering for such a long time, and second of all thanks for your detailed answers!

I found some literature regarding the averaging intervall and its influences on the Weibull-Parameters and I might perform a pre-study myself to resolve this problem.
My real concern right now is the calm threshold: If I understand WACA and Weibull correctly a threshold above 1.0 would have an influence on the Weibull-Parameters and also the following simulation (or does it take the calm threshold into account while calculating the weibull-parameters?). Is there a way to avoid this problem? Is it f.e. possible to combine the 1m/s and the 2m/s classes and which influences would this have?
Thanks in advance for every input you can give me!

[Duncan]

The calm threshold is an interesting issue.

Let's assume that both the direction and speed measurement instrument have the same calm threshold. If there are some measurements where the speed falls below this threshold, we lack information about these records:

1. The direction is completely unknown.
2. The speed is somewhere between zero and the calm threshold.

These records can't be ignored, but they can't be treated in the same way as the others.

The records are distributed uniformly with respect to direction (they are evenly distributed among the direction bins.)

I think that the speed for these records is assumed to be half the calm threshold, but I am not completely sure.

As far as I understand things, the important thing to register is the proportion of the time when the wind speed is very low. For AEP, the actual speed is unlikely to be significant.

Note that in WaCA, you can specify a separate width for the first wind speed bin of the histogram. I think that you should ensure that your first (lowest) wind speed bin upper limit is at least the calm threshold, so that all calms get bundled in to the same (first) bin.

[dachiller]

To be honest: I don't understand the Weibull function completly yet. But as far as I understand, the fitting is performed using the wind speed bin. So a lower first bin than my calm threshold would lead to an erroneous Weibull-function. Is this correct so far?
So I would, as you are also suggesting, widen my first bin. But should I put the upper limit to the treshold, to ensure that it won't get to big, or should I put it on 2m, so after that bin the other bins would be the same as in other simulations (for cross-analysis). The third option would be to use the same lower bin for every owc I want to cross-analyse. That would be much more work though, because I would need to examine so much data before processing it.
Which one would you suggest and why?
Of course these low speeds wouldn't have an impact on AEP but a wrong Weibull-distribution would.