Numerical simulations are used to examine the roughness-induced transient growth in a laminar boundary-layer flow. Based on the spectral element method, these simulations model the stationary disturbance field associated with a nonsmooth roughness geometry, such as the spanwise periodic array of circular disks used by White and co-workers during a series of wind tunnel experiments at Case Western Reserve University. Besides capturing the major trends from the recent measurements by White and Ergin, the simulations provide additional information concerning the relative accuracy of the experimental findings derived from two separate wall-finding procedures. The paper also explores the dependence of transient growth on geometric characteristics of the roughness distribution, including the height and planform shape of the roughness element and the ratio of roughness s u e to spacing between a n adjacent pair of elements. Results are used for a P r e h i M r y assessment of the differences between recently reported theoretical results of Tumin and Reshotko and the measurements by White and Ergin.Nomenclature roughness height Measure of streamwise velocity perturbation associated with any Fourier mode involved in transient growth constants involved in scaling of transiently growing perturbation amplitudes with roughness height integrated energy for spanwise mode of harmonic index k at a given streamwise location (= JU, dq) height of roughness element ratio of spanwise wave number corresponding to Fourier mode of interest to fundamental spanwise wave number based on element spacing within roughness array order of discretization polynomial within each element power spectral density Reynolds number based on roughness height and mean-flow velocity at this height Reynolds number based on U , and x radial distance from center of roughness element, i.e. ((x-X , $ + Z~) '~ roughness radius root mean square = Modal profiles of streamwise, wall-normal, and spanwise velocity perturbations corresponding to spanwise mode of wavelength h = A& free-stream speed velocity components along streamwise, wall-normal, and spanwise directions streamwise, wall-normal, and spanwise coordinates (x = 0 corresponds to plate leading edge, i.e., the virtual origin of the boundary layer in an experiment; z = 0 is aligned with the center of the