In interpreting Fig. 6, the following should be noted. First, the good agreement between the sonic anemometer and lidars in the low-frequency spectrum shows that large turbulent eddies are captured by the lidars well up to an energetic large-eddy frequency cut off ($fe=0.008\u2009\u2009Hz$); a plot of $S(f)$ versus $f$ shows that most of the along-beam component energy ($\u223c79%$ along the ARL lidar beam, $\u223c63%$ along the UU lidar beam) is contained below the frequency of the “energy containing” eddies ($f<fe$) which is defined as the frequency corresponding to the maximum of the plot. Taking the typical mean velocity as $U\xaf=3\u2009\u2009m\u2009s\u22121$ (see Fig. 5), for this time series and using Taylor’s hypothesis, this translates into a relationship between the wave number ($k$) and frequency of $kU\xaf=2\pi fe$ or an eddy size of $\u2113e=U\xaf/2\pi fe=60\u2009\u2009m$, which is a reasonable value based on previous observations.^{39} Second, note that the sampling Nyquist frequency ($fN$, defined as half the sampling frequency) for the UU lidar is 0.125 Hz and for the ARL lidar is 0.5 Hz, and frequencies higher than $fN$ are to be discarded for the lidar results. Third, the lidar radial velocity is spatially averaged by a weighting function which has a peak value at the center of a range gate; this spatial average is equivalent to spatial filtering with a filter function and the resulting radial velocity has fewer fluctuations than exhibited by the sonic anemometer. The spatial resolution of the lidar (range-gate size) imposes a restriction on the size of eddies that can be resolved. For UU, this is $k=2\pi /18=0.35\u2009\u2009m\u22121$, and hence above the frequency of $fr=kU\xaf/2\pi =0.17\u2009\u2009Hz$, which is slightly greater than its sampling Nyquist frequency. For the ARL lidar, the wave number $k$ becomes $2\pi /50=0.125\u2009\u2009m\u22121$ and $fr=0.06\u2009\u2009Hz$. The limiting factor for the ARL lidar is its spatial resolution rather than the sampling frequency for this relatively low wind speed of $3\u2009\u2009ms\u22121$. We expect the classical Kolmogorov spectra [$\u22125/3$ slope, i.e., $\u22122/3$ slope for $fS(f)$] is applicable for $fr>f>fe$ sonic anemometer data, but the slope is slightly steeper for Doppler lidar data, reflecting the spectral attenuation due to the spatial average of the range gate (Fig. 6). A rough estimation can be made of the amount of energy that is becoming “opaque” because of the low resolution of the lidar. Integrating the spectrum up to $f=fr$, we find that TKE is underestimated in lidar measurements by about 7% for the ARL lidar, and 11% for the UU lidar compared with the sonic anemometer. Analogously, the work of Kit et al.^{40} showed that about 10% of the energy is unaccounted for due to the relatively low resolution of sonic anemometers compared to hot-film probes.