Stable Boundary Layer Depth from High-Resolution Measurements of the Mean Wind Profile

Abstract

The depth h of the stable boundary layer (SBL) has long been an elusive measurement. In this diagnostic study the use of high-quality, high-resolution (?z = 10 m) vertical profile data of the mean wind U(z) and streamwise variance σu2(z) is investigated to see whether mean-profile features alone can be equated with h. Three mean-profile diagnostics are identified: hJ, the height of maximum low-level-jet (LLJ) wind speed U in the SBL; h1, the height of the first zero crossing or minimum absolute value of the magnitude of the shear ?U/?z profile above the surface; and h2, the minimum in the curvature ?2U/?z2 profile. Boundary layer BL here is defined as the surface-based layer of significant turbulence, so the top of the BL was determined as the first significant minimum in the σu2(z) profile, designated as hσ. The height hσ was taken as a reference against which the three mean-profile diagnostics were tested. Mean-wind profiles smooth enough to calculate second derivatives were obtained by averaging high-resolution Doppler lidar profile data, taken during two nighttime field programs in the Great Plains, over 10-min intervals. Nights are chosen for study when the maximum wind speed in the lowest 200 m exceeded 5 m s?1 (i.e., weak-wind, very stable BLs were excluded). To evaluate the three diagnostics, data from the 14-night sample were divided into three profile shapes: Type I, a traditional LLJ structure with a distinct maximum or ?nose,? Type II, a ?flat? structure with constant wind speed over a significant depth, and Type III, having a layered structure to the shear and turbulence in the lower levels. For Type I profiles, the height of the jet nose hJ, which coincided with h1 and h2 in this case, agreed with the reference SBL depth to within 5%. The study had two major results: 1) among the mean-profile diagnostics for h, the curvature depth h2 gave the best results; for the entire sample, h2 agreed with hσ to within 12%; 2) considering the profile shapes, the layered Type III profiles gave the most problems. When these profiles could be identified and eliminated from the sample, regression and error statistics improved significantly: mean relative errors of 8% for hJ and h1, and errors of <5% for h2, were found for the sample of only Type I and II profiles.