Tropospheric carbon monoxide (CO) was measured throughout 2001 using ground-based Fourier transform IR (FTIR) spectrometers at Moshiri (44.4¿N) and Rikubetsu (43.5¿N) observatories in northern Japan, which are separated by 150 km. Seasonal and day-to-day variations of CO are studied using these data, and contributions from various CO sources are evaluated using three-dimensional global chemistry transport model (GEOS-CHEM) calculations. Seasonal maximum and minimum FTIR-derived tropospheric CO amounts occurred in April and September, respectively. The ratio of partial column amounts between the 0--4 and 0--12 km altitude ranges is found to be slightly greater in early spring. The GEOS-CHEM model calculations generally reproduce these observed features. Source-labeled CO model calculations suggest that the observed seasonal variation is caused by seasonal contributions from various sources, in addition to a seasonal change in chemical CO loss by OH. Changes in meteorological fields largely control the relative importance of various source contributions. The contributions from fossil fuel (FF) combustion in Asia and photochemical CO production have the greatest yearly averaged contribution at 1 km among the CO sources (31% each). The Asian FF contribution increases from winter to summer, because weak southwesterly wind in summer brings more Asian pollutants to the observation sites. The seasonal variation from photochemical CO production is small (¿17% at 1 km), likely because of concurrent increases (decreases) of photochemical production and loss rates in summer (winter), with the largest contribution between August and December. The contribution from intercontinental transport of European FF combustion CO is found to be comparable to that of Asian FF sources in winter. Northwesterly wind around the Siberian high in this season brings pollutants from Europe directly to Japan, in addition to southward transport of accumulated pollution from higher latitudes. The influences are generally greater at lower altitudes, resulting in a vertical gradient in the CO profile during winter. The model underestimates total CO by 12--14% between March and June. Satellite-derived fire-count data and the relationship between FTIR-derived HCN and CO amounts are generally consistent with biomass burning influences, which could have been underestimated by the model calculations.