How can affected sensors be identified?

[johng]

Over time, a sensor which expresses this defect will spin more slowly than an unaffected sensor and discrepancies in the performance between the two may be noticed. 

When looking for discrepencies, it is a good idea to include filtering for icing, tower shadow effects and speed range. 

For example, one might want to look at 10 minute data records where the temperaure is above 3 degrees C, the vane Standard Deviation (SD) is not 0 in an unshadowed sector (typically 20 degrees wide), and where wind speeds are between 4 m/s and 26 m/s.

Once this filtering is accomplished you might compare the monthly speed averages for the month.  If the sensors do not agree within 0.2 m/s, the pair might be suspect - at that point you may want to look at the data closer or have another data analyst check it.