Infrared radiance data acquired by the Landsat thematic mapper (TM) provide information regarding the thermal state of active lavas, which can be used to estimate flow surface temperatures and thus lava cooling. Saturation of one or both of TM's short-wave infrared (SWIR) wavebands has meant, however, that previous studies of lava flow cooling have approximated what is almost certainly an n thermal component flow surface temperature distribution (where n>1), with a single pixel-integrated temperature (PIT). We show that the high sensitivity of TM's SWIR wavebands to the small amount of high-temperature (e.g., ~1000 ¿C) material exposed in hot fractures within the flows cooler (e.g., ~200 ¿C) surface crust results in the calculation of PITs that are higher than the temperature of this crust, which is usually the dominant radiative component. This causes heat loss from such pixels to be overestimated. We present adaptations that allow more realistic heat loss estimates to be produced using TM data by assuming that a second high-temperature component contributes to the radiance measured from saturated pixels and by defining an appropriate range for the temperature of the lava crust. By analyzing a suite of five TM images acquired during the 1991--1993 eruption of Mount Etna we present more realistic estimates of how this flow cooled, as well as results that show how the flow regime of the lava evolved over time. The methods we present will continue to be of use as, although the next generation of high spatial resolution sensors such as the Landsat 7 enhanced thematic mapper plus (ETM+) and the Terra advanced space-borne thermal emission and reflection radiometer (ASTER) will be less susceptible to saturation than TM, the problem will still persist. ¿ 2000 American Geophysical Union