72
J.D. Albertson, G. Kiely and M.B. Parlange
a wide range of atmospheric stabilities were observed with low -z/ L values during the days
immediately following irrigation and increasingly large values as the drying out of the bare soil
continued and more of the available energy was forced to sensible heat.
The dry Owens Lake site, which had daytime highs of about 40°C and night-time lows of about
15°C during August 1994, enabled investigation over a wider range of convective atmospheric
conditions than in Davis. Owens Lake is an arid flat landscape with uniform fetch exceeding
10 km, and a lakebed area of about 200 km 2 • The surface roughness length of the lakebed has
been estimated at Zo = 0.13 mm (Katul et al., 1995b). The 3-D sonic anemometer was set at
z=2.65 m, and data were recorded at 56 Hz.
3.3 Dissipation Results
3.3.1 Data screening
Of the collected data, 180 files (each of 20 minute duration) were selected for analysis. The
selection was based on a requirement that the files support unambiguous decomposition into
means and fluctuations and that they possess turbulence intensity values ( T.I. = O'u/ < U » of
less than 50%, as necessary for the application of Taylor's hypothesis (Stull, 1988). The Davis
experiment contributed 105 of these files and Owens Lake the remainder. The atmospheric
stability range encountered was 0.004 < -z/L < 8.1 . For each file the power spectra of
the longitudinal velocity and temperature were computed using square windowing of 2048
points, bell tapering the first and last 10% of the window (Stull, 1988, p.309), using an FFT
to compute the spectra of the window, repeating the process on the remaining windows and
averaging all windows for each wavenumber. The power spectra of w was also calculated for
each file to verify local isotropy, which is necessary for inertial subrange scaling and is indicated
by t- = ~ (Tennekes and Lumley, 1972).
Typical power spectra of velocity and temperature are shown in Figures 3.4a, and 4b. These
spectra have been frequency smoothed for presentation; in their original form they are quite
jumpy. Note that they follow the expected -5/3 scaling over a wide range of wavenumbers.
Typical second order structure functions for velocity and temperature are shown in Figures
3.5a and 5b. As expected they both follow the 2/3 scaling. Figures 3.6a and 6b are typical
velocity and temperature third order structure functions. They both follow the rl scaling for
the short lag portion which is expected to scale inertially. The third order structure functions
are considered to be a more stringent test of the inertial subrange (Katul et al., 1995a). The
second and third order structure functions do not require smoothing as the time averaging
process provides stable measures that vary smoothly with r.
3.3.2 Normalized dissipation rates for TKE
From the computed power spectra, log transformation of (3.24) allows the straightforward
determination of e from the regressed intercept of 10g(E) vs log(k). A similar approach to the
second order structure function provides an estimate of e. Due to the uncertain nature of the
actual values of au and S2 and the susceptibility of power spectra and the second order structure
function to intermittency effects on the inertial subrange scaling, we place more confidence in
the third order structure function. A comparison of the three methods was made by Albertson
et al. (1996). Here we focus on the third order approach and compute e from (3.28). For the
third order structure function, the if>. results for the 180 data files are binned in equal log
increments of -z/L and presented in Figure 3.7. The scaling forms of (3.12) were fit to the
J.D. Albertson, G. Kiely and M.B. Parlange
a wide range of atmospheric stabilities were observed with low -z/ L values during the days
immediately following irrigation and increasingly large values as the drying out of the bare soil
continued and more of the available energy was forced to sensible heat.
The dry Owens Lake site, which had daytime highs of about 40°C and night-time lows of about
15°C during August 1994, enabled investigation over a wider range of convective atmospheric
conditions than in Davis. Owens Lake is an arid flat landscape with uniform fetch exceeding
10 km, and a lakebed area of about 200 km 2 • The surface roughness length of the lakebed has
been estimated at Zo = 0.13 mm (Katul et al., 1995b). The 3-D sonic anemometer was set at
z=2.65 m, and data were recorded at 56 Hz.
3.3 Dissipation Results
3.3.1 Data screening
Of the collected data, 180 files (each of 20 minute duration) were selected for analysis. The
selection was based on a requirement that the files support unambiguous decomposition into
means and fluctuations and that they possess turbulence intensity values ( T.I. = O'u/ < U » of
less than 50%, as necessary for the application of Taylor's hypothesis (Stull, 1988). The Davis
experiment contributed 105 of these files and Owens Lake the remainder. The atmospheric
stability range encountered was 0.004 < -z/L < 8.1 . For each file the power spectra of
the longitudinal velocity and temperature were computed using square windowing of 2048
points, bell tapering the first and last 10% of the window (Stull, 1988, p.309), using an FFT
to compute the spectra of the window, repeating the process on the remaining windows and
averaging all windows for each wavenumber. The power spectra of w was also calculated for
each file to verify local isotropy, which is necessary for inertial subrange scaling and is indicated
by t- = ~ (Tennekes and Lumley, 1972).
Typical power spectra of velocity and temperature are shown in Figures 3.4a, and 4b. These
spectra have been frequency smoothed for presentation; in their original form they are quite
jumpy. Note that they follow the expected -5/3 scaling over a wide range of wavenumbers.
Typical second order structure functions for velocity and temperature are shown in Figures
3.5a and 5b. As expected they both follow the 2/3 scaling. Figures 3.6a and 6b are typical
velocity and temperature third order structure functions. They both follow the rl scaling for
the short lag portion which is expected to scale inertially. The third order structure functions
are considered to be a more stringent test of the inertial subrange (Katul et al., 1995a). The
second and third order structure functions do not require smoothing as the time averaging
process provides stable measures that vary smoothly with r.
3.3.2 Normalized dissipation rates for TKE
From the computed power spectra, log transformation of (3.24) allows the straightforward
determination of e from the regressed intercept of 10g(E) vs log(k). A similar approach to the
second order structure function provides an estimate of e. Due to the uncertain nature of the
actual values of au and S2 and the susceptibility of power spectra and the second order structure
function to intermittency effects on the inertial subrange scaling, we place more confidence in
the third order structure function. A comparison of the three methods was made by Albertson
et al. (1996). Here we focus on the third order approach and compute e from (3.28). For the
third order structure function, the if>. results for the 180 data files are binned in equal log
increments of -z/L and presented in Figure 3.7. The scaling forms of (3.12) were fit to the