Geoscience Reference
In-Depth Information
The concept of the eddy-covariance method is simple, but there are a number of
caveats:
1. Frequency response: The sensor should have an immediate response to a changing
signal. For example, if the air temperature changes by 1 K in 1 second, the thermometer
should be fast enough to follow that change (a mercury thermometer will not sufice in
this case, and neither will a cup anemometer to measure fast wind speed luctuations).
This is illustrated in
Figures 3.10a
and
b
: the sensor of part b) is not able to follow
the fast luctuations of the turbulence and hence it underestimates the magnitude in the
temperature variations. If this temperature sensor would be used to measure the sensible
heat lux, the lux would be underestimated.
2. Spatial response: The sensors should be able to sense variations at the smallest scales of
turbulence that carry signiicant parts of the lux (i.e., usually down to the scale of mm).
Thus a sensor that averages the wind speed over a distance of 1 m will not sufice for
measurements a few metres above the ground. As small scale luctuations also have small
time scales, the signal of a sensor that averages over a too long a path will look similar to
the signal of a slow sensor (as in
Figure 3.10b
).
3. Alignment: The sensors should be well aligned with the surface to ensure that the vertical
axis (the direction of the
w
-component of the wind speed) is indeed vertical.
4. Sensor separation: The locations where vertical wind speed luctuations and concentra-
tion of the transported quantity are measured should not be too far separated. If they sam-
ple different volumes, part of the correlation will be lost (the larger the distance between
the sensors, the smaller the correlation).
Because sensors and measurement setups are usually not as ideal as required, cor-
rections to the computed covariances are needed. Those corrections address issues
related to points (1) to (4). On top of that, instrument-related corrections may be
needed, because some sensors do not measure exactly the quantity one is interested
in (e.g., a gas analyser may be sensitive not only to humidity luctuations, but also to
oxygen luctuations). If the setup of an eddy-covariance system has been carefully
designed, the corrections do not add up to more than 10% of the measured lux (and
thus possible errors in the corrections do not have a major inluence on the resulting
luxes). The accuracy of observed luxes (30-minute averages) is 5-10% for the sen-
sible heat lux and 10-15% for the latent heat lux (Mauder et al.,
2006
).
Apart from the requirements related to the sensors and their installation, there are
also important requirements related to the sampling: both sampling frequency and
averaging period. As can be seen in
Figure 3.4
, turbulent signals vary wildly and the
correlation between vertical wind speed and the transported quantity is generally not
large (
Figure 3.7a
). The relative statistical error of a covariance (e.g., a lux like
w
′
θ
)
can be estimated as (after Lenschow et al.,
1994
):
1
[
]
RE(
wx
′′=
)
max(
RE
σ
),
RE
()
σ
(3.15)
w
x
R
wx
Search WWH ::
Custom Search