DATA VALIDATION - Wind Resource Assessment: A Practical Guide to Developing a Wind Project - page 124

occur in flatter terrain for brief periods under low wind and strong surface heating,
but this effect is usually small and thus requires no correction.
9.4 DATA SUBSTITUTION AND AVERAGING
Up to this point, the data validation process has sought to keep valid data from each
sensor intact and separate from the data from other sensors on the same tower. In
this section, two methods of combining the data from different sensors are discussed:
substitution and averaging. Data substitution aims to create the longest possible data
record by filling gaps in one sensor's record with data from one or more other sensors;
data averaging seeks to reduce the uncertainty in the observed speeds by combining
measurements from two different anemometers at the same height.
9.4.1 Data Substitution
Since a key objective of the wind resource monitoring program is to develop a time
series of wind data covering as long a period as possible, it is desirable to fill any
gaps in the record with valid data from other sensors when available. Data substitution
is virtually a requirement for anemometers at the top mast height, as well as for the
top direction vanes, as they are the most important for assessing the site's wind
resource. Whether data substitution is performed for lower level anemometers or
temperature and pressure sensors is largely a matter of preference (note that for reasons
discussed in Chapter 10, substituted data should not be used for estimating the wind
shear).
For anemometers, the substituted data ideally should come from an instrument at
the same height, although in rare instances—such as when both anemometers at the
top height have malfunctioned for an extended period—data from an anemometer
at a different height may be used. In any case, before the substitution is carried
out, a relationship (such as a linear regression forced through the origin or a simple
ratio) between the two anemometers should be established from concurrent, valid data.
The analyst should verify that the relationship between them is tight and linear, as
otherwise the results will be unreliable. This “field calibration” is especially important
when there is a significant, persistent bias between the anemometer readings, which
can happen with anemometers of different types (such as heated and unheated) and
with anemometers at different heights.
It is generally straightforward to fill gaps in the directional data record using valid
data from another vane. The analyst should merely check to make sure that there is no
significant, persistent bias between the two vanes' directional readings during periods
when both produce valid data. Such a bias could indicate a discrepancy in the boom
orientations or vane deadbands and should be investigated and corrected, if possible.
Note that large transient deviations in direction can occasionally arise under light,
variable winds; when the wind is strong, however, the directions recorded at heights
within 20-30 m of each other should be nearly equal (within 5 ◦ ).