Significant variations are apparent between the various reported regional and global ship SO2 emission inventories. Important parameters for SO2 emission modelling are sulphur contents and marine fuel consumption. Since 1993, the global average sulphur content for heavy fuel has shown an overall downward trend, while the bunker sale has increased. We present an improved bottom up approach to estimate marine sulphur emissions from ship transportation, including the geographical distribution. More than 53,000 individual bunker samples are used to establish regionally and globally (volume) weighted average sulphur contents for heavy and distillate marine fuels. We find that the year 2002 sulphur content in heavy fuels varies regionally from 1.90% (South America) to 3.07% (Asia), with a globally weighted average of 2.68% sulphur. The calculated globally weighted average content for heavy fuels is found to be 5% higher than the average (arithmetic mean) sulphur content commonly used. The reason for this is likely that larger bunker stems are mainly of high-viscosity heavy fuel, which tends to have higher sulphur values compared to lower viscosity fuels. The uncertainties in SO2 inventories are significantly reduced using our updated SO2 emission factors (volume-weighted sulphur content). Regional marine bunker sales figures are combined with volume-weighted sulphur contents for each region to give a global SO2 emission estimate in the range of 5.9–7.2 Tg (SO2) for international marine transportation. Also taking into account the domestic sales, the total emissions from all ocean-going transportation is estimated to be 7.0–8.5 Tg (SO2). Our estimate is significantly lower than recent global estimate reported by Corbett and Koehler [2003. Journal of Geophysical Research: Atmospheres 108] (6.49 Tg S or about 13.0 Tg SO2). Endresen et al. [2004. Journal of Geophysical Research 109, D23302] claim that uncertainties in input data for the activity-based method will give too high emission estimates. We also indicate that this higher estimate will almost give doubling of regional emissions, compared to detailed movement-based estimates. The paper presents an alternative approach to estimate present overall SO2 ship emissions with improved accuracy.