Influence of non-representative atmospheric profiles on estimates of land surface temperature by MIVIS data using radiative transfer models

When retrieving ground surface temperatures from remote sensing images by a radiative transfer model, error's can arise from the use of atmospheric vertical profiles that are not well representative of the imaged area; due to non-local and/or non-simultaneous measurements. The magnitude of these errors is dependent on many factors such as the platform of the sensor (satellite or aircraft), the kind:of sensor (channel wavelength and band width) and the actual meteorological condition on the area. An evaluation of these errors is performed, both by means of a simulation and by studying a real case using the data extracted from images acquired over Mt Etna (Sicily, southern Italy) by the MIVIS (Multispectral Infrared and Visible Imaging Spectrometer) airborne spectroradiometer during the 'Sicilia-94' remote sensing campaign. We mainly connect the error in;surface temperature and the error in the columnar water vapor content caused by a non-representative profile. We show that, in the case studied, these errors cannot be neglected; e.g., differences up to 5 degrees C in the most transparent channel (the other channels show greater values) are found. This work furnishes an important evaluation to the group working on the 'Sicilia-94' campaign and,it may be useful for all the MIVIS users. Moreover, it makes possible comparison with other airborne TIR sensors like TIMS or ATLAS. Finally, we think that this contribution may help to enhance the relation between land surface temperature/emissivity and atmospheric profile products as recommended at the International Land-Surface Temperature Workshop hold at the University of California. (C) 1998 Elsevier Science Ltd. All rights reserved.