DATA VALIDATION - Wind Resource Assessment: A Practical Guide to Developing a Wind Project - page 121

9.2.2 Treatment of Suspect Data
After the raw data are subjected to the automated validation checks, a reviewer should
decide what to do about the suspect data records. Some suspect values may represent
real (albeit unusual) weather occurrences, which should not be excluded from the
resource assessment, while others may reflect sensor or logger problems and should
be eliminated.
Here are some guidelines for handling suspect data:

• Check to see whether data from different sensors on the same mast confirm the
suspect reading. If a transient feature such as a large jump in wind speed is noted
at one anemometer, is a similar jump seen at other anemometers? If only one
sensor shows the feature, it is more likely that the data for that sensor are invalid.
• Use data from a variety of sources to verify weather conditions. If icing is
suspected, is this supported by the observed temperature? If large changes in
wind or temperature are seen in the record, do local weather stations indicate a
passing weather front that might explain the pattern?
• Examine relationships between sensors over time. Very often, sensor degradation
happens so slowly that it goes unnoticed if the data are only examined in periods
of, say, 2 weeks or a month at a time. By examining the relationships over
several months or longer, the degradation becomes obvious. Other problems, such
as icing, take a limited time to develop and disappear and, moreover, may not
affect sensors at different heights to the same degree. Periods around flagged icing
episodes should be scrutinized carefully, however, to be sure the times of onset
and conclusion have been accurately identified. This is because anemoemeters
sometimes experience slowdown before the thresholds signaling an icing event
are crossed.
• Assign invalid data a code indicating the suspected reason. Table 9-4 gives some
examples of validation codes. An examination of operation and maintenance
logs, site temperature data, and data transmission logs may help determine the
appropriate code.
• Maintain a complete record of all data validation actions for each monitoring
site in a log file.

Table 9-4. Examples of validation codes

Code Rejection criteria

− 990 Unknown event

− 991 Icing or wet snow event

− 992 Static voltage discharge

− 993 Wind shading from tower

− 995 Wind vane deadband

− 996 Operator error

− 997 Equipment malfunction

− 998 Equipment service

− 999 Missing data (no value possible)