meteohub.de

Hi,
I have a MB Pro connected to a Vantage Vue, and there seems to be something strange with the wind direction stored by the MB Pro.
Wind values only ever seem to range between 0 - 337.5 degrees. and it seems to have a very low resolution, after a week of blustery weather with the winds shifting all over the place I have the following values stored on the MB Pro

0.0
22.5
45.0
67.5
90.0
112.5
135.0
157.5
180.0
202.5
225.0
247.5
270.0
292.5
315.0
337.5

These are the only values in a weeks worth of data at a minutes resolution.

In addition the 'History' page on the MB Pro does not show any values for Wind0dir in the table, however the graph is populated.

any Ideas?
Tom

EDIT:

The data exported to Wunderground is correct, as is live data [wind0dir-act]. it seems to be only the data stored on the MB Pro internal database which is incorrect, and data exported using this data eg. [wind0dir-avg5]

After thinking about it for a while it seems the the values stored to the database equate to the 16 compass points, N, NNE, NE, NEE, E, ESE etc.
Is this correct? if so why is it not possible to store the true bearing as there is a function to extract the compass point if it is needed. eg. [wind0dir-ave5=endir]

Your assumption of 16 compass points being used to describe the main wind direction for a certain amount of time is correct.
There is nothing like an exact bearing for the last hour. What should this be?
Exact bearings are only possible for a point of time, which might be now or bound to another event, like maximum gust of
today, where an exact bearing can be recorded to in theory.

its not the data for the last hour, I'm talking about the data stored for each minute in the database. I would expect to see an average bearing for that minute, not an average bearing rounded to the nearest 22 degrees

There is no average bearing for a minute. When you get there for example 4 wind direction readings like 235, 103, 90, 107. What should the average be?
Only thing that makes sense is to define buckets and to count how often a bucket is hit. When you have a better formula, please let us know.

What you're currently calculating is the dominant wind direction in an interval rather than the average. I think the usual solution to calculating the mean direction is to separate each direction into u and v components, average each and then recombine with an arctan function to give a mean value. Occasionally when the wind is veering or backing quickly (or is close to calm and odd small gusts are almost in random directions) then this can give an odd result, but most of the time it is a good alternative to just finding the dominant direction on rather a coarse scale.

The advantage of your proposal is that it generates a number which look like being more precise.

At the core your approach is to transform the polar coordinates into an x/y coordinate system, do averaging on the
x/y axis and re-transforming to polar coordinates again. Unfortunately results will even depend on how the base
orientation of the x/y axis is and this transformation also has no meaning in real life, it is just math not connected
to physics. You could as well just sum up all degrees and divide by number of samples. I am not convinced at all.

You may not be convinced but it is the method that is widely used in analyses of wind data. See for example:

http://www.ndbc.noaa.gov/wndav.shtml

Averaging wind direction becomes more difficult as time increases and many systems have difficulties with very constant northerly wind that flicker a few degrees either side of north. I think a better option over time is the range of the wind direction over time which can be output in degree units.

But the point is that there is something approximating to a standard way of calculating the mean direction (as opposed to the dominant direction) and that is simply to average the unit vectors.