We describe two approaches for exposure assessment that we used in a large-scale retrospective cattle study conducted in Alberta, Canada. Sulfur dioxide (SO(2)) was the surrogate measure of exposure to a complex mixture of combusted sour-gas emissions. Monthly air pollution dispersion modeling (1985-1994) (based on individual industrial source processing-plant engineering specifications, emission volumes, and meteorologic information) provided exposure isopleths of sulfur dioxide concentration from each of 231 sour-gas processing-plants across the province. In contrast, a simpler measure of proximity to source(s) of varying emission rates was applied in a geographical information system based on simplified pollution decay at increasing distances from each point source. Province-wide (663,000 km(2)) surface analysis (by exposure-level classification) produced a contingency coefficient of 0.68 between the two exposure estimates. Annual exposure estimates at the 1382 dairy and 5726 beef cow-calf farms studied were highly correlated over the 10-years period (r(spearman)=0.82 and 0.83, respectively), while monthly exposure estimates were somewhat less correlated (r(spearman)=0.80 and 0.82, respectively) for the two exposure assessment methods. Crude exposure estimates from each method were similar in both direction and magnitude.