This thesis is an examination of the airflows in mushroom growing rooms. An experimental investigation of the nature of the flows in Irish tunnels showed them to be of low magnitude at the crop but controllable in principle for single layer growing. It was found that stratification of the airflow in growing tunnels could cause severe reductions in cropping surface airspeed and the operation of the heating system was identified as the main source of this. An alternative air distribution system was shown to have the potential to overcome the effects of heating. Airflow for three level growing systems in tunnels was found to be non-uniform and the use of wall-mounted deflecting plates was shown to have the potential to correct this. The provision of air flow solutions for the wide range of new growing systems would be difficult using empirical methods alone and therefore a modelling approach was sought to complement and aid the experimental work. The initial modelling work was carried out in two dimensions with TEACH-T code (SIMPLE flow solver) to calculate the turbulent flow. The code was extended to three dimensions because it was not possible to model usefully in a two-dimensional approximation. Convergence times for the SIMPLE solver were found to be excessively long. Trial applications of multi-level acceleration produced approximately 15% savings in computational effort so a new solver was investigated. The CELS (Coupled Equation Line Solver) method had been reported as superior to SIMPLE in two dimensions and already has a multi-level technique to accelerate convergence, i.e. Additive Correction Multigrid (ACM). CELS was first applied in two dimensions in order to test its usefulness with the turbulence model in the equation set. Improvements in the time to convergence, relative to SIMPLE, justified its extension to three dimensions. The Additive Correction Multi grid technique also produced significant improvements and this was extended to three dimensions. CELS3D is essentially a plane solver applied to a three-dimensional grid and a number of procedures for its application were investigated. All produced savings relative to the SIMPLE solver. The QUICK differencing scheme was incorporated in the TEACH-based code and CELS3D was tested with various geometries and values of the Reynolds number. The best results gave a 79% reduction in the time to convergence of the solver. The ACM technique in three dimensions was investigated but no useful savings in computational effort were made. In the application to mushroom growing structures, the principles of the application of CELS3D to flows around obstructions in the flow domain were examined and the difficulties identified. A solution was found but its implementation proved impractical for all but the simplest cases.