Li and Avissar [1994] investigated the impact of microscale (or ``patch''
scale) variability of the most important land-surface characteristics
[as identified by Henderson-Sellers, 1993, and
Collins and Avissar, 1994] on the atmospheric turbulent heat fluxes near the ground surface.
They characterized this variability by different probability density
functions (pdf's) for each one of the land-surface characteristics.
They found that, in general, the variability of a particular
characteristic affects the radiative and the heat fluxes simultaneously,
yet not with the same intensity. On the average, the latent heat flux
is the most sensitive and the radiative flux the least sensitive to
spatial variability. Furthermore, they showed that the more positively
skewed the pdf, the larger the difference between the heat fluxes
calculated with the pdf or the mean. For instance, differences as large
as 150 Wm
, 135 Wm
, 105 Wm
, and 100 Wm
were
found between the heat fluxes calculated with the log-normal
distributions of the mean leaf area index, stomatal conductance,
soil-surface wetness, and surface roughness, respectively. The spatial
variability of albedo created differences of not more than 20 Wm
,
indicating a relatively linear relation between this characteristic and
land-surface energy fluxes.
A detailed analysis of the land-surface energy fluxes integrated at the patch scale with either the pdf's or the corresponding mean characteristics indicated that, under unstable atmospheric conditions, stomatal conductance and leaf area index variability have the most significant effect on spatially integrated energy fluxes from vegetated land. In bare land, soil-surface wetness, instead of stomatal conductance, strongly affected the surface fluxes. Under neutral and stable atmospheric conditions, surface roughness appeared to have the predominant effect on the surface fluxes.
In their analysis, however, Li and Avissar [1994] assumed that the impact of the variability of land-surface characteristics did not extend beyond the atmospheric surface layer. This, of course, allowed them to use a relatively simple land-atmosphere model, but limited their study to the patch scale. The impact of the spatial variability of land-surface characteristics at a scale larger than the patch scale, though still considered microscale, can be studied with large-eddy simulations (LES) models. This modeling technique, which was pioneered by Deardorff [1972], is accurate and, therefore, very attractive to study the turbulence activity in the PBL. For instance, Deardorff [1974] showed that turbulence statistics obtained with an LES model compared nicely to those observed in the 1967 Wangara Experiment [ Clarke et al., 1971]. Note that with this modeling technique, only small (less energetic) eddies need to be parameterized and largest-size eddies are resolved. Numerical experiments with LES by Moeng [1984, 1986, and 1987] and, more recently, by Hechtel et al. [1990], Hadfield et al. [1991, 1992], and Walko et al. [1992] confirmed the high potential of this technique, especially with the powerful computers readily available nowaday.
Unfortunately, because of the extremely large computer resources needed to produce LES, so far only a few investigations have been conducted to better understand the impact of spatial variability of land-surface characteristics on the PBL. Hechtel et al. [1990] studied the development of the PBL during a light-wind, cloud-free skies day of the 1983 Oklahoma boundary Layer Experiment (BLX83). In this particular case, they did not find any significant impact of the surface heterogeneity on the PBL. Hadfield et al. [1992] simulated the effect of a sinusoidal wave (with two different wavelengths) of surface heat flux with zero or weak wind background conditions. They found that the heat flux variation drives a circulation, which is considerably strengthened with the increase of the wavelength of the perturbation. However, even a light wind weakens this circulation drastically and moves it downwind. Walko et al. [1992] studied the impact of small topographical features (assuming horizontal homogeneity of heat fluxes) on the structure of turbulence in the PBL. They found that the horizontal spectra of vertical motion was strongly biased toward the horizontal scales of the terrain. However, Hechtel et al. [1990], Hadfield et al. [1992], and Walko et al. [1992] indicated that vertical profiles of atmospheric variables obtained by horizontal averaging were not affected by the spatial variability of heat flux or the presence of small hills.
More research is needed in this area and with the availability of new, more powerful computers, additional insights are expected to come in the near future. Nevertheless, these preliminary results indicate that as long as the length scale of the surface sensible heat flux perturbation is smaller than about 2 km and the topographical features are not higher than about 200 m, the assumption used by Li and Avissar [1994] seems reasonable.