Eleven long-term sulfur deposition models are evaluated for each season of 1980 using screened precipitation chemistry data from five North American networks. Two objectives of the model evaluation are (1) to identify which models best replicate the observed temporal and spatial patterns of sulfur wet deposition within data uncertainty limits (i.e., from measurement errors and spatial variability) and (2) to delineate for each model areas and periods of good and poor agreement with observations. Analyses indicated that the MOE, SERTAD and TVA models significantly under/overpredict in neither more than 20% of the evaluation region for any season nor in 25% of any of the four subregions. Many of the models replicated rather well the temporal pattern. However, the temporal patterns of the SIMPMOD and STATMOD models were significantly different.