Different methods for temperature and spectral emissivity retrieval over active fires from IR hyperspectral remotely sensed data

摘要

We have developed a theoretical model which allows us to retrieve the temperature of burning areas employing the thermal infrared spectral-radiance data gathered by the used sensor. The algorithm performs the temperature-emissivity separation by coupling together radiance data from a pair of neighbouring similar image pixels, which are assumed to have the same emissivity spectrum and independent temperatures. The characteristic uncertainty in the basic thermal infrared equations is thus resolved using a double number of spectral measurements (radiance samples) for the temperature estimation. Results produced by this model have been compared with fire-front temperatures assessed utilising spectral data collected at short-wave infrared wavelengths between 1.99 μm and 2.48 μm. Moreover, we have also applied the GBE-method for retrieving the pixel temperature and spectral emissivity utilising thermal infrared spectral data. In order to test the three considered methods, we have utilised a rare example of airborne hyperspectral image acquired over a natural active fire. This image was collected by the MIVIS operated on board of a Casa 212 airplane over the Alps (Italy). The paper shows the image acquired by the MIVIS sensor, the related processing and is completed with a theoretical discussion of the involved topics.