This dataset contains the sediment models, DNS flow field data, subsurface path data, and calculated transit time distributions for both the regular- and random-interface cases used in the paper: "Quantifying the effects of bed roughness on transit time distributions via direct numerical simulations of turbulent hyporheic exchange" by Guangchen Shen, Junlin Yuan, and Mantha S. Phanikumar (Submitted to Water Resources Research). Detailed introduction of each data file is as follows. 1. DNS flow field data Flowfield_Reg.h5 and Flowfield_Ran.h5 contains the following fields for the regular and random cases, respectively. 'ni', 'nj', 'nk' are the numbers of grid points along x, y, and z directions. 'xc', 'yc', 'zc' are the cell center locations. 'u3d','v3d','w3d' are the three-dimensional time-averaged velocities at each grid point. 'vof' is the volume-of-fluid field used by the immersed-boundary method to prescribe the fluid-solid interface (vof=1 in fluid and 0 in solid), at each grid point. 'vof' contains the information of sediment grain distribution and bed roughness geometry. Only the subsurface data (those below the sediment crest) are shared due to dataset size limit. 2. Particle-tracked subsurface flow paths and corresponding transit time distributions The mat files “xxx_pathline” store the (x,y,z) location of each point (saved as 'StrX', 'StrY, 'StrZ') along the subsurface paths, discretized by uniform steps of travel time (with time intervals of 0.1 for 'A' and 0.01 for 'MD' cases, normalized by channel height and friction velocity). The files “xxx_TT” store the array of transit times corresponding to the tracked paths, where the 1d array T is the transit time. 'A' denotes calculations based on time-mean advection only, while 'MD' denotes calculations accounting for additional molecular diffusion. 'Interface' and '3DiameterBelow' indicate that the particles were released at the interface and -3 D below the interface, respectively.