THE CLIMATE ADJUSTMENT PROCESS - Wind Resource Assessment: A Practical Guide to Developing a Wind Project - page 167

satellites became increasingly important in the 1970s and 1980s, decades that were
marked also by the retirement of weather ships, growth in the use of commercial
aircraft to supplement weather observations, and a large increase in the frequency of
weather observations from both surface and rawinsonde stations (9).
To some degree, the atmospheric model should be able to attenuate the impact
of such changes, as observations from one new platform or sensor are reconciled
with those available from existing sensors. At different times and in different regions,
however, the availability of new data can significantly alter the model's analysis,
resulting in spurious trends and shifts in wind and other parameters (10).
In response to these concerns, the concept of a “controlled reanalysis” has been
introduced. This approach is similar to reanalysis except that additional care is taken
to employ data from a consistent set of observational systems and platforms (such as
a fixed number of levels from a fixed set of rawinsonde stations). Research suggests
that this method can reduce inconsistencies in traditional reanalysis (11).
In sum, modeled data sets can be useful compliments to surface and rawinsonde
observations, but the resource analyst should be wary of relying on them entirely for
MCP except when direct observations are unavailable or inadequate to the task. As
always, the consistency of the modeled data should be verified through comparisons
with independent data sources.

12.4 THE TARGET-REFERENCE RELATIONSHIP

Once the reference station (or stations) is selected, the next step is to establish a
relationship between the reference and target winds. This relationship is used to predict
the long-term wind resource at the target site based on the entire valid (homogeneous)
record of the reference station.
Many types of functional relationships can be used, too many to be described
comprehensively here. (Summaries of various methods can be found in References 12
and 13.) The most popular approaches are based on a linear transformation between
the reference and target wind speeds (and, occasionally, directions). The general form
of the linear equation is the familiar y b , where x is the reference wind

speed, y is the target wind speed, m is the slope, and b is the intercept. If the “true”
long-term mean wind speed at the reference station is known, then the predicted mean
is given by the equation

= mx +

y = mx + b (12.4)

where the bar over the variable indicates an average. Usually, this equation is deter-
mined through a least squares fitting procedure called linear regression (see below).
A variety of nonlinear methods (e.g., artificial neural networks or support vector
machines) have also been proposed and studied, but they tend to be more complicated
than linear methods and require more expertise. Only the linear methods are discussed
here.