The field modeling technique for predicting the temperature distribution and smoke movement in enclosures containing a fire source is validated against experiments carried out in a fully instrumented sports building covered by an air supported dome. The building is oval in plan and the dome has an ellipsoidal shape. A 2MW methanol pool fire located centrally on the floor of the building was used to obtain detailed measurements of temperature at a number of locations. The mathematical model simulates the transient problem in three dimensions using two different finite volume grids. The first grid is a polar cylindrical one with cells partially blocked to simulate features not coincident with grid lines. The second uses a non-orthogonal grid which follows closely the contours of the building. Results are obtained for pre-fire, fire and post-fire conditions and the two grid solutions are compared with experiments. Qualitative agreement is good throughout and trends are correctly simulated. Quantitative agreement is also good in all areas except in common with earlier studies in the immediate vicinity of the fire source. The body fitted grid solution predicts correctly the lack of stratification due to strong convection along the ceiling.