We used field measurements of methane (CH4) flux from upland and wetland soils in the Northern Study Area (NSA) of BOREAS (BOReal Ecosystem-Atmosphere Study), near Thompson, Manitoba, during the summers of 1994 and 1996 to estimate the overall CH4 emission from a 1350 km2 landscape. June--September 1994 and 1996 were both drier and warmer than normal, but summer 1996 received 68 mm more precipitation than 1994, a 40% increase, and had a mean daily air temperature 0.6¿C warmer than 1994. Upland soils consumed CH4 at rates from 0 to 1.0 mg m-2 d-1, with small spatial and temporal variations between years, and a weak dependence on soil temperature. In contrast, wetlands emitted CH4 at seasonal average rates ranging from 10 to 350 mg CH4 m-2 d-1, with high spatial and temporal variability, and increased an average of 60% during the wetter and warmer 1996. We used Landsat imagery, supervised classification, and ground truthing to scale point CH4 fluxes (1000 km2). We performed a sensitivity analysis for error terms in both areal coverage and CH4 flux, showing that the small areas of high CH4 emission (e.g., small ponds, graminoid fens, and permafrost collapse margins) contribute the largest uncertainty in both flux measurements and mapping. Although wetlands cover less than 30% of the landscape, areally extrapolated CH4 flux for the NSA increased by 61% from 10 to16 mg CH4 m-2 d-1 between years, entirely attributed to the increase in wetland CH4 emission. We conclude that CH4 fluxes will tend to be underestimated in areas where much of the landscape is covered by wetlands. This is due to the large spatial and temporal variability encountered in chamber-based measurements of wetland CH4 fluxes, strong sensitivity of wetland CH4 emission to small changes in climate, and because most remote sensing images do not adequately identify small areas of high CH4 flux.