In this paper we have analysed the effects of the different atmospheric species (water vapour, fixed gases and aerosols) and the surface emissivity on the split-window method for determining the sea surface temperature. The widely used split-window method is based on the differential absorption of water vapour in the atmospheric window 10.5ndash;12.5 μm. Other atmospheric species with absorption coefficients different to that of water vapour can then have a large influence on the split-window. The effect of gases, such as C02, N20, CH4, CO and 03, and maritime aerosols is evaluated by comparing the effect of the water vapour alone. To do this we simulated AVHRR measurements in channels 4 and 5 for a set of mid-latitude atmospheres using LOWTRAN 7 code. Our results indicated that the fixed gases have a negative effect on the split-window specially for dry atmospheres; in this case the error in retrieved temperatures was shown to increase by about 70 per cent with respect to that obtained considering water vapour only. The effect of maritime aerosols was parameterised in terms of the surface meteorological range and the path optical thickness was measured at 0.55 μm, which can be obtained from both visible channels of AVHRR. The total effect on the split-window appeared to be a linear function of the path optical thickness. On the other hand, we analysed the impact of sea surface emissivity showing that it is strongly dependent on the observation angle, especially for angles larger than 40°. In addition to this it has been shown that the emissivity effect depends on the atmospheric moisture. However, for angles lower than 40° the atmospherically averaged emissivity effect is close to zero. Finally we have given a correction algorithm accounting for all the studied effects, yielding an error estimated at 0–34 degK. over the simulated mid-latitude data set.