The displacement height d and roughness length z0 are parameters of the
logarithmic wind profile and as such these are characteristics of the
surface, that are required in a multitude of meteorological modeling
applications. Classically, both parameters are estimated from
multi-level measurements of wind speed over a terrain sufficiently
homogeneous to avoid footprint-induced differences between the levels.
As a rule-of thumb, d of a dense, uniform crop or forest canopy is 2/3
to 3/4 of the canopy height h, and z0 about 10% of canopy height in
absence of any d. However, the uncertainty of this rule-of-thumb becomes
larger if the surface of interest is not "dense and uniform", in which
case a site-specific determination is required again. By means of the
eddy covariance method, alternative possibilities to determine z0 and d
have become available. Various authors report robust results if either
several levels of sonic anemometer measurements, or one such level
combined with a classic wind profile is used to introduce direct
knowledge on the friction velocity into the estimation procedure. At the
same time, however, the eddy covariance method to measure various fluxes
has superseded the profile method, leaving many current stations without
a wind speed profile with enough levels sufficiently far above the
canopy to enable the classic estimation of z0 and d. From single-level
eddy covariance measurements at one point in time, only one parameter
can be estimated, usually z0 while d is assumed to be known. Even so,
results tend to scatter considerably. However, it has been pointed out,
that the use of multiple points in time providing different stability
conditions can enable the estimation of both parameters, if they are
assumed constant over the time period regarded. These methods either
rely on flux-variance similarity (Weaver 1990 and others following), or
on the integrated universal function for momentum (Martano 2000 and
others following). In both cases, iterations over the range of possible
d values are necessary. We extended this set of methods by a
non-iterative, regression based approach. Only a stability range of data
is used in which the universal function is known to be approximately
linear. Then, various types of multiple linear regression can be used to
relate the terms of the logarithmic wind profile equation to each other,
and derive z0 and d from the regression parameters. Two examples each of
the two existing iterative approaches, and the new non-iterative one are
compared to each other and to plausibility limits in three different
agricultural crops. The study contains periods of growth as well as of
constant crop height, also allowing for an examination of the relations
between z0, d, and canopy height. Results indicate that estimated z0
values, even in absence of prescribed d values, are fairly robust,
plausible and consistent across all methods. The largest deviations are
produced by the two flux-variance similarity based methods. Estimates of
d, in contrast, can be subject to implausible deviations with all
methods, even after quality-filtering of input data. Again, the largest
deviations occur with flux-variance similarity based methods. Ensemble
averaging between all methods can reduce this problem, offering a
potentially useful way of estimating d at more complex sites where the
rule-of-thumb cannot be applied easily. Martano P (2000): Estimation
of surface roughness length and displacement height from single-level
sonic anemometer data. Journal of Applied Meteorology 39:708-715.
Weaver HL (1990): Temperature and Humidity flux-variance relations
determined by one-dimensional eddy correlation. Boundary-Layer
Meteorology 53:77-91.