An experimental study of entrainment of sand-sized particles in turbulent boundary layers has been performed in a high-speed wind tunnel at square-pulse flow speeds of 27 to 101 ms−1 and for soil bed lengths varying from 2.1 to 5.8 m. Because of high particle drag-to-weight ratios (D/W = 100–1000) and friction velocities (uf) well above soil threshold friction velocities (uft; 10 [les ] uf/uft [les ] 40), the present results correspond to the suspension regime of dust lofting, in contrast to low-speed saltation flows (1 [les ] uf/uft [les ] 5; D/W / 15). Results are obtained characterizing particle entrainment for both a natural soil (White Sands Missile Range (WSMR) sand; 50% finer-by-weight diameter, D50 = 180 μm) and a monosized sand sample (Ottawa sand, D50 = 250 μm). Measurements of local boundary layer velocities and dust densities were performed with traversing state-of-the-art diagnostics. Scouring rate data (0.015 [les ] ms [les ] 0.30 g cm−2 s−1) and streamwise soil flux (10 [les ] Q [les ] 150 g cm−1 s−1) as a function of bed length and velocity were determined.

Scouring rates were found to increase as the 3/2-power of velocity, but decay as the inverse square root of dust bed length. Corresponding streamwise soil fluxes (also known as soil loss rates) increased to the 3/2-power of velocity in contrast to the cube power dependence for low-speed results (ufree-stream [les ] 15 m s−1; Q [les ] 1.5 g cm−1 s−1). Comparison of scouring rate data (from pre/post-test soil loss measurements) with derived data based on the rate of change of streamwise flux with distance was favourable. WSMR rates were always lower than Ottawa sand rates, a result consistent with the lower repose angle for the Ottawa sand sample.

Both sets of soil data demonstrate that dust edges extend vertically to higher elevations than corresponding velocity edges. This result implies that the turbulent Schmidt number for the present flows is less than unity and of the order of 0.7. Favourable collapsing of the scouring rate data base was achieved when measured rates were normalized by the friction velocity mass flux, square root of edge Mach number and sand repose angle ratio. A universal rate of 0.3±0.1 correlated well with the bulk of the data.