A method for monitoring the temporal variation of surface spectral emissivity: Application to thermal infrared multispectral scanner (TIMS) data in HAPEX-Sahel

During the Hydrologic Atmospheric Pilot Experiment (HAPEX)-Sahel experiment in 1992, thermal infrared multispectral scanner (TIMS) data were acquired from aircraft at an altitude of 600 m. The main steps to retrieve the surfa ce relative spectral emissivities E from TIMS data using the emissivity nor malization method were recalled. Several TIMS data on September 2 and 4 wer e processed, and the retrieved E for different types of surfaces showed tha t E in channels 1-3 (8.2-9.4 mu m) are less than E in channels 4-6 (9.6-12. 0 mu m) for bare soils, whereas for vegetation the E profile is not as flat as it should be. In order to correct for the effect of uncertainties in at mospheric profiles given by radiosonde data and to get the emissivity profi le flat for vegetation the temperature and humidity profiles were modified by considering the vegetation to be a gray body. Although E obtained with t he modified atmospheric profiles seemed to be improved significantly for ea ch day, E obtained for the same type of surface at two days are still very different, showing that it is difficult to compare E derived from different images with the method proposed. Through the comparison of E derived with radiosonde data and E derived with the modified atmospheric profiles the li near relationship found by Li et al. [1999] between E calculated with inacc urate atmospheric profiles and its actual value was confirmed. On the basis of this linear relationship property a method called temporal atmospheric normalization was proposed to correct for the errors caused by the imperfec t atmospheric corrections under horizontally invariant atmospheric conditio ns. This method relates linearly the measurement of relative emissivity mad e at a time t(2), E-c(t(2)), to the measurement E-c(t(2) --> t(1)) that wou ld have been made if the atmospheric condition of lime t(2) was that of a r eference time tl, In order to calculate the coefficients of the linear rela tionship between E-c(t(2) --> t(1)) and E-c(t(2)) it is necessary to select in the image some samples whose relative emissivity are known to be time i nvariant. This method was validated using TIMS data. It may now be used to monitor the temporal variation of surface relative emissivity correcting fo r possible temporal variation of the atmosphere particularly using Advanced Spaceborn Thermal Emission Reflection Radiometer data.