Reliable estimates of carbon exchange between terrestrial ecosystems and the atmosphere due to land-use change have become increasingly important. One source of land-use change estimates comes from comparing multi-date remote sensing imagery, though the effect of land-cover classification errors on carbon flux estimates has not been considered to our knowledge. We evaluated the integration of a land-cover change detection methodology using Landsat MultiSpectral Scanner (MSS) imagery with a regional carbon budget analysis. This work includes the incorporation of image classification accuracy information into the carbon budget of our sample landscape. Our analysis indicates that the Los Tuxtlas, Mexico study region experienced an approximate net loss of 2.146 106 tC from 1986 to 1990. A carbon loss value which does not include classification error estimates is 34 per cent less than the adjusted value.