Long-term studies of soil organic carbon dynamics in two- and three-crop rotations in irrigated cotton (Gossypium hirsutum L.) based cropping systems under varying stubble management practices in Australian Vertosols are relatively few. Our objective was to quantify soil organic carbon dynamics during a 9-year period in four irrigated, cotton-based cropping systems sown on permanent beds in a Vertosol with restricted subsoil drainage near Narrabri in north-western New South Wales, Australia. The experimental treatments were: cotton–cotton (CC); cotton–vetch (Vicia villosa Roth. in 2002–06, Vicia benghalensis L. in 2007–11) (CV); cotton–wheat (Triticum aestivum L.), where wheat stubble was incorporated (CW); and cotton–wheat–vetch, where wheat stubble was retained as in-situ mulch (CWV). Vetch was terminated during or just before flowering by a combination of mowing and contact herbicides, and the residues were retained as in situ mulch. Estimates of carbon sequestered by above- and below-ground biomass inputs were in the order CWV >> CW = CV > CC. Carbon concentrations in the 0–1.2 m depth and carbon storage in the 0–0.3 and 0–1.2 m depths were similar among all cropping systems. Net carbon sequestration rates did not differ among cropping systems and did not change significantly with time in the 0–0.3 m depth, but net losses occurred in the 0–1.2 m depth. The discrepancy between measured and estimated values of sequestered carbon suggests that either the value of 5% used to estimate carbon sequestration from biomass inputs was an overestimate for this site, or post-sequestration losses may have been high. The latter has not been investigated in Australian Vertosols. Future research efforts should identify the cause and quantify the magnitude of these losses of organic carbon from soil.