The coarse spatial resolution of large-scale models can be a significant source of error in calculating grid cell averaged rates of physical processes. We investigate here the influence of resolution in the case of chemical reactions. We use high resolution MOZAIC ozone and water vapour data to characterise the effects of unresolved structures on model-calculated production of the hydroxyl radical (OH) in the upper troposphere/lower stratosphere region. The statistical analysis of MOZAIC data indicates that, on average, large-scale models may overestimate OH production by 5% to 20% for resolutions ranging from ~200 to ~800 km respectively. This systematic bias is found to be most significant in the tropopause region.