An uncertainty analysis of the Hull acid rain model (HARM) is presented using the Generalised Likelihood Uncertainty Estimation methodology, constrained by observations at 25 sites across Wales. It was found that no simulations could fit the 1995 observed deposition data simultaneously at all sites. To retain predictive capability, the observational constraint was relaxed, resulting in 2101 simulations that described adequately the data. HARM�s predictive capability was further tested in a �hindcast� to data collected at 44 sites in 1984, using the successful parameterisations for 1995. At the majority of sites, the uncertainty prediction bounds spanned the observed data satisfactorily for both 1995 and 1984, but there was a tendency for wet deposition to be overestimated at high rainfall sites and underestimated or well estimated at lower rainfall sites close to major source areas. For wet-deposited oxidised nitrogen there was systematic overestimation at most sites. It is concluded that the model could be improved to remove the systematic nature of the overestimation of wet oxidised nitrogen for Wales and to better represent the spatial pattern of all wet-deposited species considered.