We will adopt the newly-developed array technique proposed by Tong et al. (1999). The array consists of a primary line of nine equally-spaced sonic anemometers at one height, close to the ground, and a second line of five sonic anemometers slightly above (or below) the primary line (although Tong et al. have not yet used the second line in their study). Tong et al. (1998) evaluated the performance of this two-dimensional (x and y) filter with data from a high-resolution LES of a convective PBL and showed that "the two and three-dimensionally filtered data are essentially indistinguishable" if the cutoff lies on the peak of the w-spectrum.

The array is ideally aligned to be perpendicular to the mean wind direction, such that the filtered and SGS fields along the wind can be obtained by time filtering and Taylor's hypothesis and those across the wind can be retrieved from the nine sensor line as described in Tong et al. (1999). However, it is unlikely that the wind direction will be exactly normal to the array for a significant period of the field measurements. Nevertheless, moderate deviations from the ideal wind direction can be accommodated by again using Taylor's hypothesis and lagging the data from each of the individual anemometers to analyze the data in a reference frame normal to the wind. This will effectively reduce the crosswind spacing of the anemometers by the cosine of the wind direction deviation angle and can even be an advantage by providing data from a (limited) range of crosswind spacings.

In addition to the three components of velocity, sonic anemometers also measure the speed of sound and hence an acceptable approximation of virtual temperature. Thus with the primary line of sensors, one can easily compute the horizontal gradients of both filtered velocity and temperature fields. But the vertical gradients of these filtered fields require a second line of five sonic anemometers just above or below the primary line.

In this first experiment, we will try three different array setups. We plan to keep the experiment running both day and night so that we can collect data over a wide range of stability.