This dataset provides a high-fidelity view of the viscous sublayer in a canonical turbulent boundary layer at Reynolds numbers up to Reθ = 2400. The data were generated by direct numerical simulation using the open-source Incompact3d solver on the UK's ARCHER2 supercomputer.
The core of the dataset is a time series of 16,384 three-dimensional snapshots. Each snapshot contains all three velocity components and pressure (double precision) on a 3482 × 26 × 256 grid, forming a five-dimensional array of shape (16384, 3482, 26, 256, 4). Data are provided in Zarr format to support efficient partial reads, sensible chunking for common access patterns, and on-the-fly compression to minimise data transfer.
Alongside the instantaneous fields, we include pre-computed, time-averaged three-dimensional statistics (mean velocities and Reynolds stresses) and complete metadata covering domain dimensions, time step, and viscosity.
The dataset is intended to support a wide range of uses, including turbulence theory and modelling, experimental design, training or benchmarking data-driven methods, and validation of faster, lower-fidelity simulations. Access instructions and simple Python/Xarray examples are provided in the repository.