Factors affecting the accuracy of thermal imaging cameras in volcanology

[1] Volcano observatories and researchers are recognizing the potential usefulness of thermal imaging cameras both before and during volcanic eruptions. Obvious applications include measurements of the surface temperatures of active lava domes and lava flows to determine the location of the most active parts of these potentially hazardous features. If appropriate precautions are taken, the new generation of thermal imaging cameras can be used to extract quantitative as well as qualitative information on volcanic activity. For example, they can be used to measure the temperature of lava on eruption and to reveal how the crust cools during flow emplacement. This is important for the validation of lava flow models. To ensure that meaningful temperatures are collected, thermal imaging data must be corrected for instrumental errors, emissivity of the surface being imaged, atmospheric attenuation, viewing angle and surface roughness. Controlled laboratory experiments have been undertaken to determine the emissivity of smooth and rough samples and the effects of viewing angle and to quantify the errors. Measured emissivities range from 0.973 ± 0.002 for smooth samples of basalt and 0.984 ± 0.004 for rough samples. Errors in emissivity-corrected temperatures are within ±15 C for lava at 1100 C. Variations from individual sensor receptors, which provide individual pixel temperature data, were found to be 0.6% and instrumental errors of the cameras used were 0.1%. Apparent temperatures were found to vary by less than the instrumental error for viewing angles up to 30 degrees from normal to lava, and thereafter increased by 1 C per degree. By increasing the apparent viewing distance of a small vent on Mount Etna from 1.5 to 30 m, the maximum temperature is shown to decrease by 53 C due to integrated averaging of radiance over increased pixel areas. At a viewing distance of 250 m the maximum temperature decreased by 200 C with a further 75 C decrease due to atmospheric attenuation for a relative humidity of 50%. However, errors in relative humidity measurements can lead to atmospheric attenuation correction inaccuracies up to 200 C at viewing distances of 1 km. We show how temperatures measured using thermal imaging cameras can be corrected to give improved estimates of temperature distributions on the surface of active lava flows.