The presented study investigates the evolution of artificial gravel placements for Atlantic salmon and sea trout in Aurlandselva in Western Norway. Various monitoring methods have been applied including (i) quantifying the spatial extent and dynamics of spawning sites over the monitoring period, (ii) grain size distributions as well as (iii) applying numerical hydraulic and sediment transport modelling with the aim to test the predictability of such numerical tools. The spawning sites were not clogged by fine sediments, but were reshaped due to scouring and sediment transport. The scouring resulted in a volume loss of the gravel banks between 32 and 95% in the monitoring period of 5 years. The application of hydrodynamic-numerical modelling, however, showed that the modelling methods were not sufficient to predict erosion of the gravel or the site. The study showed that the areas are sensitive especially to local scale micro-topographical roughness elements. The complex three-dimensional hydraulic processes and the coarse substrate in the non-fluvial river environment makes it impracticable for multi-dimensional modelling to predict dynamics of gravel. A novel sediment criterion was introduced to estimate the near-bottom turbulence by relating the d m of introduced gravel compared to the d 90 of the bed surface substrate composition.