The objective of this study was to evaluate the interannual variation during 1981--1997 in methane emission from an oligotrophic lawn plant community in a boreal mire. Daily mean air temperatures and daily accumulated precipitation figures were used as forcing variables for a mechanistic soil physical model, to generate estimates for the depths of snow, soil frost, and water table, together with soil temperature profiles. Methane emissions were then simulated with a mechanistic mixed mire methane model, described in this paper. The methane model simulated total methane emission rates separated into diffusion, ebullition, and plant-mediated transport. The climate record for 1981--1997 represented almost the total range in growing season temperature sums in northern Sweden during the twentieth century. The average temperature sum for 1980--1997 was 735 (range 553--981) degree days. The average accumulated annual precipitation during the same period was 283 (range 140--397) mm. The integrated simulated annual methane emission during 1981--1997 varied almost threefold among years, with a mean of 17.4¿1.1 (SE) and a range of 11 to 27 g m-2. The simulated annual emissions for the three calibration years 1995--1997 were 18, 14, and 22 g m-2, respectively, slightly higher than the integrated measured emissions over the growing season (May--September), which were 16, 13, and 18 g m-2 y-1, respectively. Given the model formulation and parameterization, the single most important climatic predictor of simulated annual methane emission was mean water table position (r2=0.58). Adding annual soil temperature sum at 26 cm in a multiple regression solution significantly increased the explained variance (R2=0.85). The study emphasizes that interannual variability in methane emission may be large, and therefore extrapolations of annual methane emissions must be based on time series that adequately span the interannual variability in the local climate. ¿ 2001 American Geophysical Union