Field measurements are undertaken with the specific purpose of addressing open issues in subgrid-scale (SGS) modeling of turbulence for large eddy simulation. Wind velocity and temperature signals are obtained using a horizontal linear array of six three-dimensional sonic anemometers placed at a height of 2.15 m in the surface layer over a grass field. From these data, the SGS heat flux and a two-dimensional surrogate of the SGS dissipation of temperature variance (x) are computed by means of two-dimensional horizontal filtering and by invoking Taylor’s hypothesis. Conditional averaging is used to isolate the effects of large-scale structures (sweeps and ejections) of the flow on the SGS dissipation under different stability conditions. During flow events associated with strong increments of vertical velocity (possibly associated with the onset of ejection events), negative values of x, indicative of transfer of temperature variance from the small scales to the resolved field (backscatter), have an important relative contribution regardless of atmospheric stability. Strong drops in the vertical velocity (possibly associated with the onset of sweeps) are accompanied by large positive values of the SGS dissipation. The two-dimensional SGS dissipation is compared with a one-dimensional surrogate based on a single sensor used in earlier work. The one- and two-dimensional results show qualitatively the same trends. Quantitative differences underscore the advantages of a two-dimensional approach based on the sensor array used in this work.