Satellite remote sensing of ocean and lake surface temperature is of considerable importance for climate research, but the main problem with thermal infrared measurements is the determination of the atmospheric correction. We present a novel extension to the split-window method to retrieve the total atmospheric transmittance, and demonstrate the use of this information in improved algorithms for sea surface temperature (SST) and total column content of water vapour. Simulation studies, using radiosonde profiles and an atmospheric transmission model, show that the new SST algorithm has an intrinsic precision of ∼01 degKr.m.s. (representing an improvement by a factor of 2-4 over conventional algorithms) and that atmospheric water vapour content can be retrieved with an accuracy of ∼ 5 per cent independent of its absolute value. We also discuss potential sources of error and future applications of the technique.