For the period from January to September 1991 we describe spatial and temporal variations of sea surface carbon dioxide fugacity (fCO2) in the Antarctic, Subantarctic, subtropical, and tropical regions of the Indian Ocean (including the Red Sea). The measurements were made continuously with an infrared technique during seven cruises. We study the temporal variations of fCO2 at daily, monthly and seasonal scales in selected areas. High-frequency variabilities of 20 muatm/d have been observed near polar frontal zone. Both spatial and temporal fCO2 variations are large near the subtropical and Subantartic fronts. In the subtropical domain, fCO2 decreases regularly from austral summer to winter. In January this region is a small CO2 sink with values near equilibrium with the atmosphere. In July, low fCO2 (300 muatm) leads to a CO2 flux of -4.5 mmol/m2/d into the ocean for the zonal band 23-degrees-S-35-degrees-S. A quantitative study of monthly and seasonal fCO2 budgets is presented for the subtropical area. Considering first the observations at seasonal scale, it is shown that changes in fCO2 can be explained by temperature variations and air-sea exchanges; the sum of biological and mixing processes, considered as the balance of the seasonal fCO2 budget, is close to zero. The monthly fCO2 budgets are then calculated. In that case, other processes must be taken into account to close the budget: the observations indicate that the effect of productivity exceeds the one of mixing in austral summer and the opposite in winter. We then describe the seasonal air-sea fCO2 differences (DELTAfCO2) for the whole western Indian Ocean and corresponding Antarctic sector (18,000 observations). In the equatorial and tropical regions the ocean is a CO2 source as was previously observed in the 1960s. In the subtropical area the CO2 sink dominates but varies strongly on a monthly scale. In the circumpolar front zones there is a large potential CO2 sink in summer. In the Antarctic waters, fCO2 spatial variability is very high at mesoscale, especially in the area of the Kerguelen plateau. Finally, it is shown that in some oceanic areas, well-defined relations exist between fCO2 distribution and temperature and salinity. If we want to use them to constrain mappings of continuous fCO2 fields from sparse observations, such relations must be considered at regional and at least seasonal scales.