Accelerating airflow simulations for agricultural applications: Performance assessment of an open-source lattice Boltzmann framework

Efficient airflow simulations in and around livestock housing are essential for optimizing barn climate, improving animal welfare, and enabling large-scale emission monitoring. However, airflow in such environments is typically simulated by solving the incompressible turbulent Navier–Stokes (NS) equations using implicit finite-volume methods, which often require several hours to compute a single steady-state solution. This study evaluates the potential of a lattice Boltzmann method (LBM) solver as a computationally efficient alternative. The open-source framework OpenLB was implemented on local cluster infrastructure using parallel CPU (MPI) and GPU architectures. Model accuracy was validated against wind tunnel measurements conducted in an atmospheric boundary-layer wind tunnel and benchmarked against the widely used finite-volume solver OpenFOAM. The OpenLB simulations reproduced the measured flow field with comparable accuracy to the reference approach. For the validated benchmark case, OpenLB achieved a 23-fold acceleration in computational time compared to OpenFOAM on CPU hardware, while GPU accelaration further reduced computation times, yielding an overall speedup of approximately 46-fold relative to the reference solver. The simulations were run considering a uniform grid over the whole domain. Nevertheless, the achieved performance gains demonstrate strong potential of LBM-based approaches for accelerated transient simulations, such as pollutant transport modelling and future near-real-time coupling with spatially distributed gas sensor networks to improve flow field interpolation and data-gap filling.