The spatial representativeness of turbulent fluxes on the basis of point measurements and the mechanism of the so-called energy imbalance problem have been investigated through numerical experiments using Large Eddy Simulation (LES). Daytime atmospheric boundary layers heated uniformly over flat surface with no synoptic vertical motion were simulated, under several geostrophic winds from 1 to 4 (m/s). The following results are obtained;

(1) The horizontally averaged turbulent fluxes based on point measurements systematically underestimate the real fluxes (negative imbalances), irrespective of the spatial density of the measurement points. This is attributed to the contribution of heat transport by time-averaged vertical velocity at local points.

(2) As the geostrophic wind decreases, both the horizontally averaged negative imbalance and the standard deviation of turbulent fluxes increase.

(3) These tendencies can be explained physically by the existence of Turbulent Organized Structure (TOS), whose time scale is much longer than that of each thermal plume. The points over the updraft regions of TOS are likely to have positive imbalance due to highly activated thermal plumes. The temporal and spatial change of TOS pattern causes low-frequency trends in time-series data, which can account for the large horizontal scatter of flux estimates. High-pass filter excluding the temperature trend is theoretically useful to improve the representativeness of point measurements, although the filter choice actually needs a great care in order not to damage the turbulent components.