Identification and implications of biases in US surface wind observation, archival, and summarization methods

Observational biases in hourly surface wind speed and direction measurements from 41 northeastern U.S. stations are evaluated. Based on this analysis, a number of consistent wind observation biases are detected which should be considered when using these data in modeling and other climatological applications. These biases include a preference to report wind speed as a multiple of either 2 or 5 kt and a similar bias for reporting even, in terms of tens of degrees, wind directions. Temporal and station-dependent trends in estimating winds less than the starting speed of the anemometer are also evident. Given these biases, potential inaccuracies in analyses using wind data are illustrated. In addition, analyses are conducted to identify biases related to archival and summarization practices. In particular, the comparability of wind climatologies based on varying record lengths is determined. These assessments suggest that at least nine years of data are required to obtain representative wind frequency tabulations. The potential consequences of using too small of a record length are also illustrated. To achieve this minimum record length, climatologies can be constructed using either hourly or 3-hourly observations.