Eddy covariance flux towers were used to measure net ecosystem production over three adjacent agricultural fields in Manitoba, Canada, from 2009 to 2011. Two fields were converted from long-term perennial hay/pasture to annual cropping, while the third field served as a control field that was maintained as hay/pasture. One converted field had a rotation of oat–canola–oat crops, while the second was hay–oat–fallow. Weather was an important driver of inter-annual variability, with poor yields on all fields in 2011 because of dry conditions in summer, with the summer-fallow condition on one field caused by excess spring moisture not allowing planting. The cumulative net ecosystem production of the oat–canola–oat field showed a net CO2 emission of 100 g C m−2, the hay–oat–fallow field emitted 500 g C m−2, and the hay field gained 550 g C m−2 by the end of the 30-month study period. The hay field had the highest cumulative gross primary production of 2500 g C m−2, whereas the oat–canola–oat and hay–oat–fallow fields had only about 1400 g C m−2. The perennial field had the advantage of both early- and late-season growth when crops were absent on the other fields. The hay and hay–oat–fallow fields had comparable cumulative ecosystem respiration (1400 g C m−2). Manure additions contributed 300 g C m−2 on the two converted fields. With harvest exports and manure additions included, the oat–canola–oat field was a carbon source of 240 g C m−2, the hay–oat–fallow field was a source of 415 g C m−2, and the hay/pasture field was a sink of 120 g C m−2 over the 30-month period.